2008-05-12 21:20:42 +02:00
|
|
|
/*
|
|
|
|
* Infrastructure for profiling code inserted by 'gcc -pg'.
|
|
|
|
*
|
|
|
|
* Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
|
|
|
|
* Copyright (C) 2004-2008 Ingo Molnar <mingo@redhat.com>
|
|
|
|
*
|
|
|
|
* Originally ported from the -rt patch by:
|
|
|
|
* Copyright (C) 2007 Arnaldo Carvalho de Melo <acme@redhat.com>
|
|
|
|
*
|
|
|
|
* Based on code in the latency_tracer, that is:
|
|
|
|
*
|
|
|
|
* Copyright (C) 2004-2006 Ingo Molnar
|
2012-12-06 10:39:54 +01:00
|
|
|
* Copyright (C) 2004 Nadia Yvette Chambers
|
2008-05-12 21:20:42 +02:00
|
|
|
*/
|
|
|
|
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
#include <linux/stop_machine.h>
|
|
|
|
#include <linux/clocksource.h>
|
|
|
|
#include <linux/kallsyms.h>
|
2008-05-12 21:20:43 +02:00
|
|
|
#include <linux/seq_file.h>
|
2009-01-14 22:33:27 +01:00
|
|
|
#include <linux/suspend.h>
|
2015-01-20 18:13:40 +01:00
|
|
|
#include <linux/tracefs.h>
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
#include <linux/hardirq.h>
|
2008-02-23 16:55:50 +01:00
|
|
|
#include <linux/kthread.h>
|
2008-05-12 21:20:43 +02:00
|
|
|
#include <linux/uaccess.h>
|
2011-12-17 01:27:42 +01:00
|
|
|
#include <linux/bsearch.h>
|
2011-05-26 23:53:52 +02:00
|
|
|
#include <linux/module.h>
|
2008-02-23 16:55:50 +01:00
|
|
|
#include <linux/ftrace.h>
|
2008-05-12 21:20:43 +02:00
|
|
|
#include <linux/sysctl.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
|
|
|
#include <linux/slab.h>
|
2008-05-12 21:20:43 +02:00
|
|
|
#include <linux/ctype.h>
|
2011-12-16 23:06:45 +01:00
|
|
|
#include <linux/sort.h>
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
#include <linux/list.h>
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
#include <linux/hash.h>
|
2010-03-06 00:03:25 +01:00
|
|
|
#include <linux/rcupdate.h>
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2009-04-15 01:39:12 +02:00
|
|
|
#include <trace/events/sched.h>
|
2009-03-24 06:10:15 +01:00
|
|
|
|
2009-05-28 19:37:24 +02:00
|
|
|
#include <asm/setup.h>
|
2008-06-21 20:17:27 +02:00
|
|
|
|
2009-03-24 04:12:58 +01:00
|
|
|
#include "trace_output.h"
|
2009-03-20 17:50:56 +01:00
|
|
|
#include "trace_stat.h"
|
2008-05-12 21:20:42 +02:00
|
|
|
|
2008-10-23 15:33:03 +02:00
|
|
|
#define FTRACE_WARN_ON(cond) \
|
2011-04-29 16:36:31 +02:00
|
|
|
({ \
|
|
|
|
int ___r = cond; \
|
|
|
|
if (WARN_ON(___r)) \
|
2008-10-23 15:33:03 +02:00
|
|
|
ftrace_kill(); \
|
2011-04-29 16:36:31 +02:00
|
|
|
___r; \
|
|
|
|
})
|
2008-10-23 15:33:03 +02:00
|
|
|
|
|
|
|
#define FTRACE_WARN_ON_ONCE(cond) \
|
2011-04-29 16:36:31 +02:00
|
|
|
({ \
|
|
|
|
int ___r = cond; \
|
|
|
|
if (WARN_ON_ONCE(___r)) \
|
2008-10-23 15:33:03 +02:00
|
|
|
ftrace_kill(); \
|
2011-04-29 16:36:31 +02:00
|
|
|
___r; \
|
|
|
|
})
|
2008-10-23 15:33:03 +02:00
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
/* hash bits for specific function selection */
|
|
|
|
#define FTRACE_HASH_BITS 7
|
|
|
|
#define FTRACE_FUNC_HASHSIZE (1 << FTRACE_HASH_BITS)
|
2011-05-02 23:34:47 +02:00
|
|
|
#define FTRACE_HASH_DEFAULT_BITS 10
|
|
|
|
#define FTRACE_HASH_MAX_BITS 12
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2014-01-10 23:01:58 +01:00
|
|
|
#define FL_GLOBAL_CONTROL_MASK (FTRACE_OPS_FL_CONTROL)
|
2012-02-15 15:51:48 +01:00
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
2014-08-15 23:23:02 +02:00
|
|
|
#define INIT_OPS_HASH(opsname) \
|
|
|
|
.func_hash = &opsname.local_hash, \
|
|
|
|
.local_hash.regex_lock = __MUTEX_INITIALIZER(opsname.local_hash.regex_lock),
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
#define ASSIGN_OPS_HASH(opsname, val) \
|
|
|
|
.func_hash = val, \
|
|
|
|
.local_hash.regex_lock = __MUTEX_INITIALIZER(opsname.local_hash.regex_lock),
|
2013-05-09 07:44:17 +02:00
|
|
|
#else
|
2014-08-15 23:23:02 +02:00
|
|
|
#define INIT_OPS_HASH(opsname)
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
#define ASSIGN_OPS_HASH(opsname, val)
|
2013-05-09 07:44:17 +02:00
|
|
|
#endif
|
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
static struct ftrace_ops ftrace_list_end __read_mostly = {
|
|
|
|
.func = ftrace_stub,
|
2013-03-27 14:31:28 +01:00
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE | FTRACE_OPS_FL_STUB,
|
2014-08-15 23:23:02 +02:00
|
|
|
INIT_OPS_HASH(ftrace_list_end)
|
2011-08-08 22:57:47 +02:00
|
|
|
};
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
/* ftrace_enabled is a method to turn ftrace on or off */
|
|
|
|
int ftrace_enabled __read_mostly;
|
2008-05-12 21:20:43 +02:00
|
|
|
static int last_ftrace_enabled;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
/* Current function tracing op */
|
|
|
|
struct ftrace_ops *function_trace_op __read_mostly = &ftrace_list_end;
|
2013-11-08 20:17:30 +01:00
|
|
|
/* What to set function_trace_op to */
|
|
|
|
static struct ftrace_ops *set_function_trace_op;
|
2008-11-05 22:05:44 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
/* List for set_ftrace_pid's pids. */
|
|
|
|
LIST_HEAD(ftrace_pids);
|
|
|
|
struct ftrace_pid {
|
|
|
|
struct list_head list;
|
|
|
|
struct pid *pid;
|
|
|
|
};
|
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
static bool ftrace_pids_enabled(void)
|
|
|
|
{
|
|
|
|
return !list_empty(&ftrace_pids);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_update_trampoline(struct ftrace_ops *ops);
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
/*
|
|
|
|
* ftrace_disabled is set when an anomaly is discovered.
|
|
|
|
* ftrace_disabled is much stronger than ftrace_enabled.
|
|
|
|
*/
|
|
|
|
static int ftrace_disabled __read_mostly;
|
|
|
|
|
2009-02-14 07:15:39 +01:00
|
|
|
static DEFINE_MUTEX(ftrace_lock);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2012-02-15 15:51:48 +01:00
|
|
|
static struct ftrace_ops *ftrace_control_list __read_mostly = &ftrace_list_end;
|
2011-05-04 15:27:52 +02:00
|
|
|
static struct ftrace_ops *ftrace_ops_list __read_mostly = &ftrace_list_end;
|
2008-05-12 21:20:42 +02:00
|
|
|
ftrace_func_t ftrace_trace_function __read_mostly = ftrace_stub;
|
2011-05-04 04:49:52 +02:00
|
|
|
static struct ftrace_ops global_ops;
|
2012-02-15 15:51:48 +01:00
|
|
|
static struct ftrace_ops control_ops;
|
2008-05-12 21:20:42 +02:00
|
|
|
|
2014-07-23 02:16:57 +02:00
|
|
|
static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip,
|
|
|
|
struct ftrace_ops *op, struct pt_regs *regs);
|
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
#if ARCH_SUPPORTS_FTRACE_OPS
|
|
|
|
static void ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *op, struct pt_regs *regs);
|
2011-08-08 22:57:47 +02:00
|
|
|
#else
|
|
|
|
/* See comment below, where ftrace_ops_list_func is defined */
|
|
|
|
static void ftrace_ops_no_ops(unsigned long ip, unsigned long parent_ip);
|
|
|
|
#define ftrace_ops_list_func ((ftrace_func_t)ftrace_ops_no_ops)
|
|
|
|
#endif
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2012-11-02 22:03:03 +01:00
|
|
|
/*
|
|
|
|
* Traverse the ftrace_global_list, invoking all entries. The reason that we
|
2013-05-28 20:38:43 +02:00
|
|
|
* can use rcu_dereference_raw_notrace() is that elements removed from this list
|
2012-11-02 22:03:03 +01:00
|
|
|
* are simply leaked, so there is no need to interact with a grace-period
|
2013-05-28 20:38:43 +02:00
|
|
|
* mechanism. The rcu_dereference_raw_notrace() calls are needed to handle
|
2012-11-02 22:03:03 +01:00
|
|
|
* concurrent insertions into the ftrace_global_list.
|
|
|
|
*
|
|
|
|
* Silly Alpha and silly pointer-speculation compiler optimizations!
|
|
|
|
*/
|
|
|
|
#define do_for_each_ftrace_op(op, list) \
|
2013-05-28 20:38:43 +02:00
|
|
|
op = rcu_dereference_raw_notrace(list); \
|
2012-11-02 22:03:03 +01:00
|
|
|
do
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Optimized for just a single item in the list (as that is the normal case).
|
|
|
|
*/
|
|
|
|
#define while_for_each_ftrace_op(op) \
|
2013-05-28 20:38:43 +02:00
|
|
|
while (likely(op = rcu_dereference_raw_notrace((op)->next)) && \
|
2012-11-02 22:03:03 +01:00
|
|
|
unlikely((op) != &ftrace_list_end))
|
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
static inline void ftrace_ops_init(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_INITIALIZED)) {
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_init(&ops->local_hash.regex_lock);
|
|
|
|
ops->func_hash = &ops->local_hash;
|
2013-05-09 07:44:17 +02:00
|
|
|
ops->flags |= FTRACE_OPS_FL_INITIALIZED;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2012-07-20 19:08:05 +02:00
|
|
|
/**
|
|
|
|
* ftrace_nr_registered_ops - return number of ops registered
|
|
|
|
*
|
|
|
|
* Returns the number of ftrace_ops registered and tracing functions
|
|
|
|
*/
|
|
|
|
int ftrace_nr_registered_ops(void)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *ops;
|
|
|
|
int cnt = 0;
|
|
|
|
|
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
|
|
|
|
for (ops = ftrace_ops_list;
|
|
|
|
ops != &ftrace_list_end; ops = ops->next)
|
|
|
|
cnt++;
|
|
|
|
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
|
|
|
|
return cnt;
|
|
|
|
}
|
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
static void ftrace_pid_func(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *op, struct pt_regs *regs)
|
2008-11-26 06:16:23 +01:00
|
|
|
{
|
2008-12-03 21:36:58 +01:00
|
|
|
if (!test_tsk_trace_trace(current))
|
2008-11-26 06:16:23 +01:00
|
|
|
return;
|
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
op->saved_func(ip, parent_ip, op, regs);
|
2008-11-26 06:16:23 +01:00
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:42 +02:00
|
|
|
/**
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* clear_ftrace_function - reset the ftrace function
|
2008-05-12 21:20:42 +02:00
|
|
|
*
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* This NULLs the ftrace function and in essence stops
|
|
|
|
* tracing. There may be lag
|
2008-05-12 21:20:42 +02:00
|
|
|
*/
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
void clear_ftrace_function(void)
|
2008-05-12 21:20:42 +02:00
|
|
|
{
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
ftrace_trace_function = ftrace_stub;
|
|
|
|
}
|
|
|
|
|
2012-02-15 15:51:48 +01:00
|
|
|
static void control_ops_disable_all(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
*per_cpu_ptr(ops->disabled, cpu) = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int control_ops_alloc(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int __percpu *disabled;
|
|
|
|
|
|
|
|
disabled = alloc_percpu(int);
|
|
|
|
if (!disabled)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
ops->disabled = disabled;
|
|
|
|
control_ops_disable_all(ops);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-11-08 20:17:30 +01:00
|
|
|
static void ftrace_sync(struct work_struct *work)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* This function is just a stub to implement a hard force
|
|
|
|
* of synchronize_sched(). This requires synchronizing
|
|
|
|
* tasks even in userspace and idle.
|
|
|
|
*
|
|
|
|
* Yes, function tracing is rude.
|
|
|
|
*/
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_sync_ipi(void *data)
|
|
|
|
{
|
|
|
|
/* Probably not needed, but do it anyway */
|
|
|
|
smp_rmb();
|
|
|
|
}
|
|
|
|
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
static void update_function_graph_func(void);
|
2015-09-30 01:06:50 +02:00
|
|
|
|
|
|
|
/* Both enabled by default (can be cleared by function_graph tracer flags */
|
|
|
|
static bool fgraph_sleep_time = true;
|
|
|
|
static bool fgraph_graph_time = true;
|
|
|
|
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
#else
|
|
|
|
static inline void update_function_graph_func(void) { }
|
|
|
|
#endif
|
|
|
|
|
2015-02-19 15:56:14 +01:00
|
|
|
|
|
|
|
static ftrace_func_t ftrace_ops_get_list_func(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* If this is a dynamic ops or we force list func,
|
|
|
|
* then it needs to call the list anyway.
|
|
|
|
*/
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_DYNAMIC || FTRACE_FORCE_LIST_FUNC)
|
|
|
|
return ftrace_ops_list_func;
|
|
|
|
|
|
|
|
return ftrace_ops_get_func(ops);
|
|
|
|
}
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static void update_ftrace_function(void)
|
|
|
|
{
|
|
|
|
ftrace_func_t func;
|
|
|
|
|
2014-09-10 16:42:46 +02:00
|
|
|
/*
|
|
|
|
* Prepare the ftrace_ops that the arch callback will use.
|
|
|
|
* If there's only one ftrace_ops registered, the ftrace_ops_list
|
|
|
|
* will point to the ops we want.
|
|
|
|
*/
|
|
|
|
set_function_trace_op = ftrace_ops_list;
|
|
|
|
|
|
|
|
/* If there's no ftrace_ops registered, just call the stub function */
|
|
|
|
if (ftrace_ops_list == &ftrace_list_end) {
|
|
|
|
func = ftrace_stub;
|
|
|
|
|
2011-05-06 03:14:55 +02:00
|
|
|
/*
|
|
|
|
* If we are at the end of the list and this ops is
|
2012-07-20 17:04:44 +02:00
|
|
|
* recursion safe and not dynamic and the arch supports passing ops,
|
|
|
|
* then have the mcount trampoline call the function directly.
|
2011-05-06 03:14:55 +02:00
|
|
|
*/
|
2014-09-10 16:42:46 +02:00
|
|
|
} else if (ftrace_ops_list->next == &ftrace_list_end) {
|
2015-02-19 15:56:14 +01:00
|
|
|
func = ftrace_ops_get_list_func(ftrace_ops_list);
|
2014-09-10 16:42:46 +02:00
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
} else {
|
|
|
|
/* Just use the default ftrace_ops */
|
2013-11-08 20:17:30 +01:00
|
|
|
set_function_trace_op = &ftrace_list_end;
|
2011-05-04 15:27:52 +02:00
|
|
|
func = ftrace_ops_list_func;
|
2011-08-08 22:57:47 +02:00
|
|
|
}
|
2011-05-04 04:49:52 +02:00
|
|
|
|
2014-07-15 17:05:12 +02:00
|
|
|
update_function_graph_func();
|
|
|
|
|
2013-11-08 20:17:30 +01:00
|
|
|
/* If there's no change, then do nothing more here */
|
|
|
|
if (ftrace_trace_function == func)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we are using the list function, it doesn't care
|
|
|
|
* about the function_trace_ops.
|
|
|
|
*/
|
|
|
|
if (func == ftrace_ops_list_func) {
|
|
|
|
ftrace_trace_function = func;
|
|
|
|
/*
|
|
|
|
* Don't even bother setting function_trace_ops,
|
|
|
|
* it would be racy to do so anyway.
|
|
|
|
*/
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef CONFIG_DYNAMIC_FTRACE
|
|
|
|
/*
|
|
|
|
* For static tracing, we need to be a bit more careful.
|
|
|
|
* The function change takes affect immediately. Thus,
|
|
|
|
* we need to coorditate the setting of the function_trace_ops
|
|
|
|
* with the setting of the ftrace_trace_function.
|
|
|
|
*
|
|
|
|
* Set the function to the list ops, which will call the
|
|
|
|
* function we want, albeit indirectly, but it handles the
|
|
|
|
* ftrace_ops and doesn't depend on function_trace_op.
|
|
|
|
*/
|
|
|
|
ftrace_trace_function = ftrace_ops_list_func;
|
|
|
|
/*
|
|
|
|
* Make sure all CPUs see this. Yes this is slow, but static
|
|
|
|
* tracing is slow and nasty to have enabled.
|
|
|
|
*/
|
|
|
|
schedule_on_each_cpu(ftrace_sync);
|
|
|
|
/* Now all cpus are using the list ops. */
|
|
|
|
function_trace_op = set_function_trace_op;
|
|
|
|
/* Make sure the function_trace_op is visible on all CPUs */
|
|
|
|
smp_wmb();
|
|
|
|
/* Nasty way to force a rmb on all cpus */
|
|
|
|
smp_call_function(ftrace_sync_ipi, NULL, 1);
|
|
|
|
/* OK, we are all set to update the ftrace_trace_function now! */
|
|
|
|
#endif /* !CONFIG_DYNAMIC_FTRACE */
|
|
|
|
|
2011-04-28 03:43:36 +02:00
|
|
|
ftrace_trace_function = func;
|
|
|
|
}
|
|
|
|
|
2014-04-20 17:10:43 +02:00
|
|
|
int using_ftrace_ops_list_func(void)
|
|
|
|
{
|
|
|
|
return ftrace_trace_function == ftrace_ops_list_func;
|
|
|
|
}
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static void add_ftrace_ops(struct ftrace_ops **list, struct ftrace_ops *ops)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2011-05-04 04:49:52 +02:00
|
|
|
ops->next = *list;
|
2008-05-12 21:20:42 +02:00
|
|
|
/*
|
2011-05-04 15:27:52 +02:00
|
|
|
* We are entering ops into the list but another
|
2008-05-12 21:20:42 +02:00
|
|
|
* CPU might be walking that list. We need to make sure
|
|
|
|
* the ops->next pointer is valid before another CPU sees
|
2011-05-04 15:27:52 +02:00
|
|
|
* the ops pointer included into the list.
|
2008-05-12 21:20:42 +02:00
|
|
|
*/
|
2011-05-04 04:49:52 +02:00
|
|
|
rcu_assign_pointer(*list, ops);
|
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static int remove_ftrace_ops(struct ftrace_ops **list, struct ftrace_ops *ops)
|
2008-05-12 21:20:42 +02:00
|
|
|
{
|
|
|
|
struct ftrace_ops **p;
|
|
|
|
|
|
|
|
/*
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* If we are removing the last function, then simply point
|
|
|
|
* to the ftrace_stub.
|
2008-05-12 21:20:42 +02:00
|
|
|
*/
|
2011-05-04 04:49:52 +02:00
|
|
|
if (*list == ops && ops->next == &ftrace_list_end) {
|
|
|
|
*list = &ftrace_list_end;
|
2009-02-14 07:42:44 +01:00
|
|
|
return 0;
|
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
for (p = list; *p != &ftrace_list_end; p = &(*p)->next)
|
2008-05-12 21:20:42 +02:00
|
|
|
if (*p == ops)
|
|
|
|
break;
|
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
if (*p != ops)
|
|
|
|
return -1;
|
2008-05-12 21:20:42 +02:00
|
|
|
|
|
|
|
*p = (*p)->next;
|
2011-05-04 04:49:52 +02:00
|
|
|
return 0;
|
|
|
|
}
|
2008-05-12 21:20:42 +02:00
|
|
|
|
2012-02-15 15:51:48 +01:00
|
|
|
static void add_ftrace_list_ops(struct ftrace_ops **list,
|
|
|
|
struct ftrace_ops *main_ops,
|
|
|
|
struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int first = *list == &ftrace_list_end;
|
|
|
|
add_ftrace_ops(list, ops);
|
|
|
|
if (first)
|
|
|
|
add_ftrace_ops(&ftrace_ops_list, main_ops);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_ftrace_list_ops(struct ftrace_ops **list,
|
|
|
|
struct ftrace_ops *main_ops,
|
|
|
|
struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int ret = remove_ftrace_ops(list, ops);
|
|
|
|
if (!ret && *list == &ftrace_list_end)
|
|
|
|
ret = remove_ftrace_ops(&ftrace_ops_list, main_ops);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
static void ftrace_update_trampoline(struct ftrace_ops *ops);
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static int __register_ftrace_function(struct ftrace_ops *ops)
|
|
|
|
{
|
2014-01-10 22:17:45 +01:00
|
|
|
if (ops->flags & FTRACE_OPS_FL_DELETED)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2011-05-04 15:27:52 +02:00
|
|
|
if (WARN_ON(ops->flags & FTRACE_OPS_FL_ENABLED))
|
|
|
|
return -EBUSY;
|
|
|
|
|
2012-09-28 10:15:17 +02:00
|
|
|
#ifndef CONFIG_DYNAMIC_FTRACE_WITH_REGS
|
2012-04-30 22:20:23 +02:00
|
|
|
/*
|
|
|
|
* If the ftrace_ops specifies SAVE_REGS, then it only can be used
|
|
|
|
* if the arch supports it, or SAVE_REGS_IF_SUPPORTED is also set.
|
|
|
|
* Setting SAVE_REGS_IF_SUPPORTED makes SAVE_REGS irrelevant.
|
|
|
|
*/
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_SAVE_REGS &&
|
|
|
|
!(ops->flags & FTRACE_OPS_FL_SAVE_REGS_IF_SUPPORTED))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_SAVE_REGS_IF_SUPPORTED)
|
|
|
|
ops->flags |= FTRACE_OPS_FL_SAVE_REGS;
|
|
|
|
#endif
|
|
|
|
|
2011-05-06 03:14:55 +02:00
|
|
|
if (!core_kernel_data((unsigned long)ops))
|
|
|
|
ops->flags |= FTRACE_OPS_FL_DYNAMIC;
|
|
|
|
|
2014-01-10 23:01:58 +01:00
|
|
|
if (ops->flags & FTRACE_OPS_FL_CONTROL) {
|
2012-02-15 15:51:48 +01:00
|
|
|
if (control_ops_alloc(ops))
|
|
|
|
return -ENOMEM;
|
|
|
|
add_ftrace_list_ops(&ftrace_control_list, &control_ops, ops);
|
2014-11-14 01:40:56 +01:00
|
|
|
/* The control_ops needs the trampoline update */
|
|
|
|
ops = &control_ops;
|
2011-05-04 15:27:52 +02:00
|
|
|
} else
|
|
|
|
add_ftrace_ops(&ftrace_ops_list, ops);
|
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
/* Always save the function, and reset at unregistering */
|
|
|
|
ops->saved_func = ops->func;
|
|
|
|
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_PID && ftrace_pids_enabled())
|
|
|
|
ops->func = ftrace_pid_func;
|
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
ftrace_update_trampoline(ops);
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
if (ftrace_enabled)
|
|
|
|
update_ftrace_function();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __unregister_ftrace_function(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
2011-05-04 15:27:52 +02:00
|
|
|
if (WARN_ON(!(ops->flags & FTRACE_OPS_FL_ENABLED)))
|
|
|
|
return -EBUSY;
|
|
|
|
|
2014-01-10 23:01:58 +01:00
|
|
|
if (ops->flags & FTRACE_OPS_FL_CONTROL) {
|
2012-02-15 15:51:48 +01:00
|
|
|
ret = remove_ftrace_list_ops(&ftrace_control_list,
|
|
|
|
&control_ops, ops);
|
2011-05-04 15:27:52 +02:00
|
|
|
} else
|
|
|
|
ret = remove_ftrace_ops(&ftrace_ops_list, ops);
|
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2011-04-28 03:43:36 +02:00
|
|
|
if (ftrace_enabled)
|
|
|
|
update_ftrace_function();
|
2008-05-12 21:20:42 +02:00
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
ops->func = ops->saved_func;
|
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
return 0;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2008-11-26 06:16:23 +01:00
|
|
|
static void ftrace_update_pid_func(void)
|
|
|
|
{
|
2015-07-24 16:38:12 +02:00
|
|
|
bool enabled = ftrace_pids_enabled();
|
|
|
|
struct ftrace_ops *op;
|
|
|
|
|
2011-04-28 03:43:36 +02:00
|
|
|
/* Only do something if we are tracing something */
|
2008-11-26 06:16:23 +01:00
|
|
|
if (ftrace_trace_function == ftrace_stub)
|
2009-03-06 07:29:04 +01:00
|
|
|
return;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
if (op->flags & FTRACE_OPS_FL_PID) {
|
|
|
|
op->func = enabled ? ftrace_pid_func :
|
|
|
|
op->saved_func;
|
|
|
|
ftrace_update_trampoline(op);
|
|
|
|
}
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
|
2011-04-28 03:43:36 +02:00
|
|
|
update_ftrace_function();
|
2008-11-26 06:16:23 +01:00
|
|
|
}
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_PROFILER
|
|
|
|
struct ftrace_profile {
|
|
|
|
struct hlist_node node;
|
|
|
|
unsigned long ip;
|
|
|
|
unsigned long counter;
|
2009-03-24 04:12:58 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
unsigned long long time;
|
2010-04-26 20:02:05 +02:00
|
|
|
unsigned long long time_squared;
|
2009-03-24 04:12:58 +01:00
|
|
|
#endif
|
2009-02-16 21:28:00 +01:00
|
|
|
};
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile_page {
|
|
|
|
struct ftrace_profile_page *next;
|
|
|
|
unsigned long index;
|
|
|
|
struct ftrace_profile records[];
|
2008-05-12 21:20:43 +02:00
|
|
|
};
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat {
|
|
|
|
atomic_t disabled;
|
|
|
|
struct hlist_head *hash;
|
|
|
|
struct ftrace_profile_page *pages;
|
|
|
|
struct ftrace_profile_page *start;
|
|
|
|
struct tracer_stat stat;
|
|
|
|
};
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
#define PROFILE_RECORDS_SIZE \
|
|
|
|
(PAGE_SIZE - offsetof(struct ftrace_profile_page, records))
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
#define PROFILES_PER_PAGE \
|
|
|
|
(PROFILE_RECORDS_SIZE / sizeof(struct ftrace_profile))
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2009-03-25 18:26:41 +01:00
|
|
|
static int ftrace_profile_enabled __read_mostly;
|
|
|
|
|
|
|
|
/* ftrace_profile_lock - synchronize the enable and disable of the profiler */
|
2009-03-20 17:50:56 +01:00
|
|
|
static DEFINE_MUTEX(ftrace_profile_lock);
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
static DEFINE_PER_CPU(struct ftrace_profile_stat, ftrace_profile_stats);
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2013-04-10 01:55:50 +02:00
|
|
|
#define FTRACE_PROFILE_HASH_BITS 10
|
|
|
|
#define FTRACE_PROFILE_HASH_SIZE (1 << FTRACE_PROFILE_HASH_BITS)
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
static void *
|
|
|
|
function_stat_next(void *v, int idx)
|
|
|
|
{
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile *rec = v;
|
|
|
|
struct ftrace_profile_page *pg;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
pg = (struct ftrace_profile_page *)((unsigned long)rec & PAGE_MASK);
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
again:
|
2009-06-26 05:15:37 +02:00
|
|
|
if (idx != 0)
|
|
|
|
rec++;
|
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
if ((void *)rec >= (void *)&pg->records[pg->index]) {
|
|
|
|
pg = pg->next;
|
|
|
|
if (!pg)
|
|
|
|
return NULL;
|
|
|
|
rec = &pg->records[0];
|
2009-03-23 22:12:36 +01:00
|
|
|
if (!rec->counter)
|
|
|
|
goto again;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return rec;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *function_stat_start(struct tracer_stat *trace)
|
|
|
|
{
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat *stat =
|
|
|
|
container_of(trace, struct ftrace_profile_stat, stat);
|
|
|
|
|
|
|
|
if (!stat || !stat->start)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return function_stat_next(&stat->start->records[0], 0);
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2009-03-24 04:12:58 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
/* function graph compares on total time */
|
|
|
|
static int function_stat_cmp(void *p1, void *p2)
|
|
|
|
{
|
|
|
|
struct ftrace_profile *a = p1;
|
|
|
|
struct ftrace_profile *b = p2;
|
|
|
|
|
|
|
|
if (a->time < b->time)
|
|
|
|
return -1;
|
|
|
|
if (a->time > b->time)
|
|
|
|
return 1;
|
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
/* not function graph compares against hits */
|
2009-03-20 17:50:56 +01:00
|
|
|
static int function_stat_cmp(void *p1, void *p2)
|
|
|
|
{
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile *a = p1;
|
|
|
|
struct ftrace_profile *b = p2;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
if (a->counter < b->counter)
|
|
|
|
return -1;
|
|
|
|
if (a->counter > b->counter)
|
|
|
|
return 1;
|
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
}
|
2009-03-24 04:12:58 +01:00
|
|
|
#endif
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
static int function_stat_headers(struct seq_file *m)
|
|
|
|
{
|
2009-03-24 04:12:58 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, " Function "
|
|
|
|
"Hit Time Avg s^2\n"
|
|
|
|
" -------- "
|
|
|
|
"--- ---- --- ---\n");
|
2009-03-24 04:12:58 +01:00
|
|
|
#else
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, " Function Hit\n"
|
|
|
|
" -------- ---\n");
|
2009-03-24 04:12:58 +01:00
|
|
|
#endif
|
2009-03-20 17:50:56 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int function_stat_show(struct seq_file *m, void *v)
|
|
|
|
{
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile *rec = v;
|
2009-03-20 17:50:56 +01:00
|
|
|
char str[KSYM_SYMBOL_LEN];
|
2010-08-23 10:50:12 +02:00
|
|
|
int ret = 0;
|
2009-03-24 04:12:58 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2009-03-26 02:00:47 +01:00
|
|
|
static struct trace_seq s;
|
|
|
|
unsigned long long avg;
|
2010-04-26 20:02:05 +02:00
|
|
|
unsigned long long stddev;
|
2009-03-24 04:12:58 +01:00
|
|
|
#endif
|
2010-08-23 10:50:12 +02:00
|
|
|
mutex_lock(&ftrace_profile_lock);
|
|
|
|
|
|
|
|
/* we raced with function_profile_reset() */
|
|
|
|
if (unlikely(rec->counter == 0)) {
|
|
|
|
ret = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2015-06-22 13:28:08 +02:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
avg = rec->time;
|
|
|
|
do_div(avg, rec->counter);
|
|
|
|
if (tracing_thresh && (avg < tracing_thresh))
|
|
|
|
goto out;
|
|
|
|
#endif
|
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
kallsyms_lookup(rec->ip, NULL, NULL, NULL, str);
|
2009-03-24 04:12:58 +01:00
|
|
|
seq_printf(m, " %-30.30s %10lu", str, rec->counter);
|
|
|
|
|
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, " ");
|
2009-03-26 02:00:47 +01:00
|
|
|
|
2010-04-26 20:02:05 +02:00
|
|
|
/* Sample standard deviation (s^2) */
|
|
|
|
if (rec->counter <= 1)
|
|
|
|
stddev = 0;
|
|
|
|
else {
|
2013-06-12 12:03:18 +02:00
|
|
|
/*
|
|
|
|
* Apply Welford's method:
|
|
|
|
* s^2 = 1 / (n * (n-1)) * (n * \Sum (x_i)^2 - (\Sum x_i)^2)
|
|
|
|
*/
|
|
|
|
stddev = rec->counter * rec->time_squared -
|
|
|
|
rec->time * rec->time;
|
|
|
|
|
2010-04-26 20:02:05 +02:00
|
|
|
/*
|
|
|
|
* Divide only 1000 for ns^2 -> us^2 conversion.
|
|
|
|
* trace_print_graph_duration will divide 1000 again.
|
|
|
|
*/
|
2013-06-12 12:03:18 +02:00
|
|
|
do_div(stddev, rec->counter * (rec->counter - 1) * 1000);
|
2010-04-26 20:02:05 +02:00
|
|
|
}
|
|
|
|
|
2009-03-26 02:00:47 +01:00
|
|
|
trace_seq_init(&s);
|
|
|
|
trace_print_graph_duration(rec->time, &s);
|
|
|
|
trace_seq_puts(&s, " ");
|
|
|
|
trace_print_graph_duration(avg, &s);
|
2010-04-26 20:02:05 +02:00
|
|
|
trace_seq_puts(&s, " ");
|
|
|
|
trace_print_graph_duration(stddev, &s);
|
2009-03-24 04:12:58 +01:00
|
|
|
trace_print_seq(m, &s);
|
|
|
|
#endif
|
|
|
|
seq_putc(m, '\n');
|
2010-08-23 10:50:12 +02:00
|
|
|
out:
|
|
|
|
mutex_unlock(&ftrace_profile_lock);
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2010-08-23 10:50:12 +02:00
|
|
|
return ret;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
static void ftrace_profile_reset(struct ftrace_profile_stat *stat)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile_page *pg;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
pg = stat->pages = stat->start;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
while (pg) {
|
|
|
|
memset(pg->records, 0, PROFILE_RECORDS_SIZE);
|
|
|
|
pg->index = 0;
|
|
|
|
pg = pg->next;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
memset(stat->hash, 0,
|
2009-03-23 22:12:36 +01:00
|
|
|
FTRACE_PROFILE_HASH_SIZE * sizeof(struct hlist_head));
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
int ftrace_profile_pages_init(struct ftrace_profile_stat *stat)
|
2009-03-23 22:12:36 +01:00
|
|
|
{
|
|
|
|
struct ftrace_profile_page *pg;
|
2009-03-26 01:06:34 +01:00
|
|
|
int functions;
|
|
|
|
int pages;
|
2009-03-23 22:12:36 +01:00
|
|
|
int i;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
/* If we already allocated, do nothing */
|
2009-03-25 01:50:39 +01:00
|
|
|
if (stat->pages)
|
2009-03-23 22:12:36 +01:00
|
|
|
return 0;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
stat->pages = (void *)get_zeroed_page(GFP_KERNEL);
|
|
|
|
if (!stat->pages)
|
2009-03-23 22:12:36 +01:00
|
|
|
return -ENOMEM;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-26 01:06:34 +01:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
|
|
|
functions = ftrace_update_tot_cnt;
|
|
|
|
#else
|
|
|
|
/*
|
|
|
|
* We do not know the number of functions that exist because
|
|
|
|
* dynamic tracing is what counts them. With past experience
|
|
|
|
* we have around 20K functions. That should be more than enough.
|
|
|
|
* It is highly unlikely we will execute every function in
|
|
|
|
* the kernel.
|
|
|
|
*/
|
|
|
|
functions = 20000;
|
|
|
|
#endif
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
pg = stat->start = stat->pages;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-26 01:06:34 +01:00
|
|
|
pages = DIV_ROUND_UP(functions, PROFILES_PER_PAGE);
|
|
|
|
|
2013-04-01 14:46:24 +02:00
|
|
|
for (i = 1; i < pages; i++) {
|
2009-03-23 22:12:36 +01:00
|
|
|
pg->next = (void *)get_zeroed_page(GFP_KERNEL);
|
|
|
|
if (!pg->next)
|
2009-03-26 01:06:34 +01:00
|
|
|
goto out_free;
|
2009-03-23 22:12:36 +01:00
|
|
|
pg = pg->next;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
2009-03-26 01:06:34 +01:00
|
|
|
|
|
|
|
out_free:
|
|
|
|
pg = stat->start;
|
|
|
|
while (pg) {
|
|
|
|
unsigned long tmp = (unsigned long)pg;
|
|
|
|
|
|
|
|
pg = pg->next;
|
|
|
|
free_page(tmp);
|
|
|
|
}
|
|
|
|
|
|
|
|
stat->pages = NULL;
|
|
|
|
stat->start = NULL;
|
|
|
|
|
|
|
|
return -ENOMEM;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
static int ftrace_profile_init_cpu(int cpu)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat *stat;
|
2009-03-23 22:12:36 +01:00
|
|
|
int size;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
stat = &per_cpu(ftrace_profile_stats, cpu);
|
|
|
|
|
|
|
|
if (stat->hash) {
|
2009-03-23 22:12:36 +01:00
|
|
|
/* If the profile is already created, simply reset it */
|
2009-03-25 01:50:39 +01:00
|
|
|
ftrace_profile_reset(stat);
|
2009-03-23 22:12:36 +01:00
|
|
|
return 0;
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
/*
|
|
|
|
* We are profiling all functions, but usually only a few thousand
|
|
|
|
* functions are hit. We'll make a hash of 1024 items.
|
|
|
|
*/
|
|
|
|
size = FTRACE_PROFILE_HASH_SIZE;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
stat->hash = kzalloc(sizeof(struct hlist_head) * size, GFP_KERNEL);
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
if (!stat->hash)
|
2009-03-23 22:12:36 +01:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2009-03-26 01:06:34 +01:00
|
|
|
/* Preallocate the function profiling pages */
|
2009-03-25 01:50:39 +01:00
|
|
|
if (ftrace_profile_pages_init(stat) < 0) {
|
|
|
|
kfree(stat->hash);
|
|
|
|
stat->hash = NULL;
|
2009-03-23 22:12:36 +01:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
static int ftrace_profile_init(void)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
int ret = 0;
|
|
|
|
|
2013-12-16 08:20:01 +01:00
|
|
|
for_each_possible_cpu(cpu) {
|
2009-03-25 01:50:39 +01:00
|
|
|
ret = ftrace_profile_init_cpu(cpu);
|
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
/* interrupts must be disabled */
|
2009-03-25 01:50:39 +01:00
|
|
|
static struct ftrace_profile *
|
|
|
|
ftrace_find_profiled_func(struct ftrace_profile_stat *stat, unsigned long ip)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile *rec;
|
2009-03-20 17:50:56 +01:00
|
|
|
struct hlist_head *hhd;
|
|
|
|
unsigned long key;
|
|
|
|
|
2013-04-10 01:55:50 +02:00
|
|
|
key = hash_long(ip, FTRACE_PROFILE_HASH_BITS);
|
2009-03-25 01:50:39 +01:00
|
|
|
hhd = &stat->hash[key];
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
if (hlist_empty(hhd))
|
|
|
|
return NULL;
|
|
|
|
|
2013-05-28 20:38:43 +02:00
|
|
|
hlist_for_each_entry_rcu_notrace(rec, hhd, node) {
|
2009-03-20 17:50:56 +01:00
|
|
|
if (rec->ip == ip)
|
2009-03-23 22:12:36 +01:00
|
|
|
return rec;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
static void ftrace_add_profile(struct ftrace_profile_stat *stat,
|
|
|
|
struct ftrace_profile *rec)
|
2009-03-23 22:12:36 +01:00
|
|
|
{
|
|
|
|
unsigned long key;
|
|
|
|
|
2013-04-10 01:55:50 +02:00
|
|
|
key = hash_long(rec->ip, FTRACE_PROFILE_HASH_BITS);
|
2009-03-25 01:50:39 +01:00
|
|
|
hlist_add_head_rcu(&rec->node, &stat->hash[key]);
|
2009-03-23 22:12:36 +01:00
|
|
|
}
|
|
|
|
|
2009-03-26 01:06:34 +01:00
|
|
|
/*
|
|
|
|
* The memory is already allocated, this simply finds a new record to use.
|
|
|
|
*/
|
2009-03-23 22:12:36 +01:00
|
|
|
static struct ftrace_profile *
|
2009-03-26 01:06:34 +01:00
|
|
|
ftrace_profile_alloc(struct ftrace_profile_stat *stat, unsigned long ip)
|
2009-03-23 22:12:36 +01:00
|
|
|
{
|
|
|
|
struct ftrace_profile *rec = NULL;
|
|
|
|
|
2009-03-26 01:06:34 +01:00
|
|
|
/* prevent recursion (from NMIs) */
|
2009-03-25 01:50:39 +01:00
|
|
|
if (atomic_inc_return(&stat->disabled) != 1)
|
2009-03-23 22:12:36 +01:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
2009-03-26 01:06:34 +01:00
|
|
|
* Try to find the function again since an NMI
|
|
|
|
* could have added it
|
2009-03-23 22:12:36 +01:00
|
|
|
*/
|
2009-03-25 01:50:39 +01:00
|
|
|
rec = ftrace_find_profiled_func(stat, ip);
|
2009-03-23 22:12:36 +01:00
|
|
|
if (rec)
|
2009-03-25 01:50:39 +01:00
|
|
|
goto out;
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
if (stat->pages->index == PROFILES_PER_PAGE) {
|
|
|
|
if (!stat->pages->next)
|
|
|
|
goto out;
|
|
|
|
stat->pages = stat->pages->next;
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
rec = &stat->pages->records[stat->pages->index++];
|
2009-03-23 22:12:36 +01:00
|
|
|
rec->ip = ip;
|
2009-03-25 01:50:39 +01:00
|
|
|
ftrace_add_profile(stat, rec);
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
out:
|
2009-03-25 01:50:39 +01:00
|
|
|
atomic_dec(&stat->disabled);
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
return rec;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2011-08-08 22:57:47 +02:00
|
|
|
function_profile_call(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *ops, struct pt_regs *regs)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat *stat;
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_profile *rec;
|
2009-03-20 17:50:56 +01:00
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
if (!ftrace_profile_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
local_irq_save(flags);
|
2009-03-25 01:50:39 +01:00
|
|
|
|
2014-04-29 21:17:40 +02:00
|
|
|
stat = this_cpu_ptr(&ftrace_profile_stats);
|
2009-06-02 03:51:28 +02:00
|
|
|
if (!stat->hash || !ftrace_profile_enabled)
|
2009-03-25 01:50:39 +01:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
rec = ftrace_find_profiled_func(stat, ip);
|
2009-03-23 22:12:36 +01:00
|
|
|
if (!rec) {
|
2009-03-26 01:06:34 +01:00
|
|
|
rec = ftrace_profile_alloc(stat, ip);
|
2009-03-23 22:12:36 +01:00
|
|
|
if (!rec)
|
|
|
|
goto out;
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
rec->counter++;
|
|
|
|
out:
|
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
|
|
|
|
2009-03-24 04:12:58 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
static int profile_graph_entry(struct ftrace_graph_ent *trace)
|
|
|
|
{
|
2011-08-09 18:50:46 +02:00
|
|
|
function_profile_call(trace->func, 0, NULL, NULL);
|
2009-03-24 04:12:58 +01:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void profile_graph_return(struct ftrace_graph_ret *trace)
|
|
|
|
{
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat *stat;
|
2009-03-25 04:17:58 +01:00
|
|
|
unsigned long long calltime;
|
2009-03-24 04:12:58 +01:00
|
|
|
struct ftrace_profile *rec;
|
2009-03-25 01:50:39 +01:00
|
|
|
unsigned long flags;
|
2009-03-24 04:12:58 +01:00
|
|
|
|
|
|
|
local_irq_save(flags);
|
2014-04-29 21:17:40 +02:00
|
|
|
stat = this_cpu_ptr(&ftrace_profile_stats);
|
2009-06-02 03:51:28 +02:00
|
|
|
if (!stat->hash || !ftrace_profile_enabled)
|
2009-03-25 01:50:39 +01:00
|
|
|
goto out;
|
|
|
|
|
2010-04-28 03:04:24 +02:00
|
|
|
/* If the calltime was zero'd ignore it */
|
|
|
|
if (!trace->calltime)
|
|
|
|
goto out;
|
|
|
|
|
2009-03-25 04:17:58 +01:00
|
|
|
calltime = trace->rettime - trace->calltime;
|
|
|
|
|
2015-09-30 01:06:50 +02:00
|
|
|
if (!fgraph_graph_time) {
|
2009-03-25 04:17:58 +01:00
|
|
|
int index;
|
|
|
|
|
|
|
|
index = trace->depth;
|
|
|
|
|
|
|
|
/* Append this call time to the parent time to subtract */
|
|
|
|
if (index)
|
|
|
|
current->ret_stack[index - 1].subtime += calltime;
|
|
|
|
|
|
|
|
if (current->ret_stack[index].subtime < calltime)
|
|
|
|
calltime -= current->ret_stack[index].subtime;
|
|
|
|
else
|
|
|
|
calltime = 0;
|
|
|
|
}
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
rec = ftrace_find_profiled_func(stat, trace->func);
|
2010-04-26 20:02:05 +02:00
|
|
|
if (rec) {
|
2009-03-25 04:17:58 +01:00
|
|
|
rec->time += calltime;
|
2010-04-26 20:02:05 +02:00
|
|
|
rec->time_squared += calltime * calltime;
|
|
|
|
}
|
2009-03-25 04:17:58 +01:00
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
out:
|
2009-03-24 04:12:58 +01:00
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int register_ftrace_profiler(void)
|
|
|
|
{
|
|
|
|
return register_ftrace_graph(&profile_graph_return,
|
|
|
|
&profile_graph_entry);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unregister_ftrace_profiler(void)
|
|
|
|
{
|
|
|
|
unregister_ftrace_graph();
|
|
|
|
}
|
|
|
|
#else
|
2011-05-31 21:51:55 +02:00
|
|
|
static struct ftrace_ops ftrace_profile_ops __read_mostly = {
|
2009-03-25 18:26:41 +01:00
|
|
|
.func = function_profile_call,
|
2013-05-09 07:44:17 +02:00
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE | FTRACE_OPS_FL_INITIALIZED,
|
2014-08-15 23:23:02 +02:00
|
|
|
INIT_OPS_HASH(ftrace_profile_ops)
|
2009-03-20 17:50:56 +01:00
|
|
|
};
|
|
|
|
|
2009-03-24 04:12:58 +01:00
|
|
|
static int register_ftrace_profiler(void)
|
|
|
|
{
|
|
|
|
return register_ftrace_function(&ftrace_profile_ops);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unregister_ftrace_profiler(void)
|
|
|
|
{
|
|
|
|
unregister_ftrace_function(&ftrace_profile_ops);
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
|
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
static ssize_t
|
|
|
|
ftrace_profile_write(struct file *filp, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
|
|
|
unsigned long val;
|
|
|
|
int ret;
|
|
|
|
|
2011-06-07 21:58:27 +02:00
|
|
|
ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
|
|
|
|
if (ret)
|
2009-03-20 17:50:56 +01:00
|
|
|
return ret;
|
|
|
|
|
|
|
|
val = !!val;
|
|
|
|
|
|
|
|
mutex_lock(&ftrace_profile_lock);
|
|
|
|
if (ftrace_profile_enabled ^ val) {
|
|
|
|
if (val) {
|
2009-03-23 22:12:36 +01:00
|
|
|
ret = ftrace_profile_init();
|
|
|
|
if (ret < 0) {
|
|
|
|
cnt = ret;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-03-24 04:12:58 +01:00
|
|
|
ret = register_ftrace_profiler();
|
|
|
|
if (ret < 0) {
|
|
|
|
cnt = ret;
|
|
|
|
goto out;
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
ftrace_profile_enabled = 1;
|
|
|
|
} else {
|
|
|
|
ftrace_profile_enabled = 0;
|
2009-06-02 03:51:28 +02:00
|
|
|
/*
|
|
|
|
* unregister_ftrace_profiler calls stop_machine
|
|
|
|
* so this acts like an synchronize_sched.
|
|
|
|
*/
|
2009-03-24 04:12:58 +01:00
|
|
|
unregister_ftrace_profiler();
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
}
|
2009-03-23 22:12:36 +01:00
|
|
|
out:
|
2009-03-20 17:50:56 +01:00
|
|
|
mutex_unlock(&ftrace_profile_lock);
|
|
|
|
|
2009-10-24 01:36:16 +02:00
|
|
|
*ppos += cnt;
|
2009-03-20 17:50:56 +01:00
|
|
|
|
|
|
|
return cnt;
|
|
|
|
}
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
static ssize_t
|
|
|
|
ftrace_profile_read(struct file *filp, char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
2009-03-25 18:26:41 +01:00
|
|
|
char buf[64]; /* big enough to hold a number */
|
2009-03-23 22:12:36 +01:00
|
|
|
int r;
|
|
|
|
|
|
|
|
r = sprintf(buf, "%u\n", ftrace_profile_enabled);
|
|
|
|
return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
|
|
|
|
}
|
|
|
|
|
2009-03-20 17:50:56 +01:00
|
|
|
static const struct file_operations ftrace_profile_fops = {
|
|
|
|
.open = tracing_open_generic,
|
|
|
|
.read = ftrace_profile_read,
|
|
|
|
.write = ftrace_profile_write,
|
llseek: automatically add .llseek fop
All file_operations should get a .llseek operation so we can make
nonseekable_open the default for future file operations without a
.llseek pointer.
The three cases that we can automatically detect are no_llseek, seq_lseek
and default_llseek. For cases where we can we can automatically prove that
the file offset is always ignored, we use noop_llseek, which maintains
the current behavior of not returning an error from a seek.
New drivers should normally not use noop_llseek but instead use no_llseek
and call nonseekable_open at open time. Existing drivers can be converted
to do the same when the maintainer knows for certain that no user code
relies on calling seek on the device file.
The generated code is often incorrectly indented and right now contains
comments that clarify for each added line why a specific variant was
chosen. In the version that gets submitted upstream, the comments will
be gone and I will manually fix the indentation, because there does not
seem to be a way to do that using coccinelle.
Some amount of new code is currently sitting in linux-next that should get
the same modifications, which I will do at the end of the merge window.
Many thanks to Julia Lawall for helping me learn to write a semantic
patch that does all this.
===== begin semantic patch =====
// This adds an llseek= method to all file operations,
// as a preparation for making no_llseek the default.
//
// The rules are
// - use no_llseek explicitly if we do nonseekable_open
// - use seq_lseek for sequential files
// - use default_llseek if we know we access f_pos
// - use noop_llseek if we know we don't access f_pos,
// but we still want to allow users to call lseek
//
@ open1 exists @
identifier nested_open;
@@
nested_open(...)
{
<+...
nonseekable_open(...)
...+>
}
@ open exists@
identifier open_f;
identifier i, f;
identifier open1.nested_open;
@@
int open_f(struct inode *i, struct file *f)
{
<+...
(
nonseekable_open(...)
|
nested_open(...)
)
...+>
}
@ read disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ read_no_fpos disable optional_qualifier exists @
identifier read_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t read_f(struct file *f, char *p, size_t s, loff_t *off)
{
... when != off
}
@ write @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
expression E;
identifier func;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
<+...
(
*off = E
|
*off += E
|
func(..., off, ...)
|
E = *off
)
...+>
}
@ write_no_fpos @
identifier write_f;
identifier f, p, s, off;
type ssize_t, size_t, loff_t;
@@
ssize_t write_f(struct file *f, const char *p, size_t s, loff_t *off)
{
... when != off
}
@ fops0 @
identifier fops;
@@
struct file_operations fops = {
...
};
@ has_llseek depends on fops0 @
identifier fops0.fops;
identifier llseek_f;
@@
struct file_operations fops = {
...
.llseek = llseek_f,
...
};
@ has_read depends on fops0 @
identifier fops0.fops;
identifier read_f;
@@
struct file_operations fops = {
...
.read = read_f,
...
};
@ has_write depends on fops0 @
identifier fops0.fops;
identifier write_f;
@@
struct file_operations fops = {
...
.write = write_f,
...
};
@ has_open depends on fops0 @
identifier fops0.fops;
identifier open_f;
@@
struct file_operations fops = {
...
.open = open_f,
...
};
// use no_llseek if we call nonseekable_open
////////////////////////////////////////////
@ nonseekable1 depends on !has_llseek && has_open @
identifier fops0.fops;
identifier nso ~= "nonseekable_open";
@@
struct file_operations fops = {
... .open = nso, ...
+.llseek = no_llseek, /* nonseekable */
};
@ nonseekable2 depends on !has_llseek @
identifier fops0.fops;
identifier open.open_f;
@@
struct file_operations fops = {
... .open = open_f, ...
+.llseek = no_llseek, /* open uses nonseekable */
};
// use seq_lseek for sequential files
/////////////////////////////////////
@ seq depends on !has_llseek @
identifier fops0.fops;
identifier sr ~= "seq_read";
@@
struct file_operations fops = {
... .read = sr, ...
+.llseek = seq_lseek, /* we have seq_read */
};
// use default_llseek if there is a readdir
///////////////////////////////////////////
@ fops1 depends on !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier readdir_e;
@@
// any other fop is used that changes pos
struct file_operations fops = {
... .readdir = readdir_e, ...
+.llseek = default_llseek, /* readdir is present */
};
// use default_llseek if at least one of read/write touches f_pos
/////////////////////////////////////////////////////////////////
@ fops2 depends on !fops1 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read.read_f;
@@
// read fops use offset
struct file_operations fops = {
... .read = read_f, ...
+.llseek = default_llseek, /* read accesses f_pos */
};
@ fops3 depends on !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write.write_f;
@@
// write fops use offset
struct file_operations fops = {
... .write = write_f, ...
+ .llseek = default_llseek, /* write accesses f_pos */
};
// Use noop_llseek if neither read nor write accesses f_pos
///////////////////////////////////////////////////////////
@ fops4 depends on !fops1 && !fops2 && !fops3 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
identifier write_no_fpos.write_f;
@@
// write fops use offset
struct file_operations fops = {
...
.write = write_f,
.read = read_f,
...
+.llseek = noop_llseek, /* read and write both use no f_pos */
};
@ depends on has_write && !has_read && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier write_no_fpos.write_f;
@@
struct file_operations fops = {
... .write = write_f, ...
+.llseek = noop_llseek, /* write uses no f_pos */
};
@ depends on has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
identifier read_no_fpos.read_f;
@@
struct file_operations fops = {
... .read = read_f, ...
+.llseek = noop_llseek, /* read uses no f_pos */
};
@ depends on !has_read && !has_write && !fops1 && !fops2 && !has_llseek && !nonseekable1 && !nonseekable2 && !seq @
identifier fops0.fops;
@@
struct file_operations fops = {
...
+.llseek = noop_llseek, /* no read or write fn */
};
===== End semantic patch =====
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Julia Lawall <julia@diku.dk>
Cc: Christoph Hellwig <hch@infradead.org>
2010-08-15 18:52:59 +02:00
|
|
|
.llseek = default_llseek,
|
2009-03-20 17:50:56 +01:00
|
|
|
};
|
|
|
|
|
2009-03-25 01:50:39 +01:00
|
|
|
/* used to initialize the real stat files */
|
|
|
|
static struct tracer_stat function_stats __initdata = {
|
2009-03-25 18:26:41 +01:00
|
|
|
.name = "functions",
|
|
|
|
.stat_start = function_stat_start,
|
|
|
|
.stat_next = function_stat_next,
|
|
|
|
.stat_cmp = function_stat_cmp,
|
|
|
|
.stat_headers = function_stat_headers,
|
|
|
|
.stat_show = function_stat_show
|
2009-03-25 01:50:39 +01:00
|
|
|
};
|
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
static __init void ftrace_profile_tracefs(struct dentry *d_tracer)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
2009-03-25 01:50:39 +01:00
|
|
|
struct ftrace_profile_stat *stat;
|
2009-03-20 17:50:56 +01:00
|
|
|
struct dentry *entry;
|
2009-03-25 01:50:39 +01:00
|
|
|
char *name;
|
2009-03-20 17:50:56 +01:00
|
|
|
int ret;
|
2009-03-25 01:50:39 +01:00
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
stat = &per_cpu(ftrace_profile_stats, cpu);
|
|
|
|
|
|
|
|
/* allocate enough for function name + cpu number */
|
|
|
|
name = kmalloc(32, GFP_KERNEL);
|
|
|
|
if (!name) {
|
|
|
|
/*
|
|
|
|
* The files created are permanent, if something happens
|
|
|
|
* we still do not free memory.
|
|
|
|
*/
|
|
|
|
WARN(1,
|
|
|
|
"Could not allocate stat file for cpu %d\n",
|
|
|
|
cpu);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
stat->stat = function_stats;
|
|
|
|
snprintf(name, 32, "function%d", cpu);
|
|
|
|
stat->stat.name = name;
|
|
|
|
ret = register_stat_tracer(&stat->stat);
|
|
|
|
if (ret) {
|
|
|
|
WARN(1,
|
|
|
|
"Could not register function stat for cpu %d\n",
|
|
|
|
cpu);
|
|
|
|
kfree(name);
|
|
|
|
return;
|
|
|
|
}
|
2009-03-20 17:50:56 +01:00
|
|
|
}
|
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
entry = tracefs_create_file("function_profile_enabled", 0644,
|
2009-03-20 17:50:56 +01:00
|
|
|
d_tracer, NULL, &ftrace_profile_fops);
|
|
|
|
if (!entry)
|
2015-01-20 18:13:40 +01:00
|
|
|
pr_warning("Could not create tracefs "
|
2009-03-20 17:50:56 +01:00
|
|
|
"'function_profile_enabled' entry\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
#else /* CONFIG_FUNCTION_PROFILER */
|
2015-01-20 18:13:40 +01:00
|
|
|
static __init void ftrace_profile_tracefs(struct dentry *d_tracer)
|
2009-03-20 17:50:56 +01:00
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_FUNCTION_PROFILER */
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
static struct pid * const ftrace_swapper_pid = &init_struct_pid;
|
|
|
|
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
static int ftrace_graph_active;
|
|
|
|
#else
|
|
|
|
# define ftrace_graph_active 0
|
|
|
|
#endif
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
|
|
|
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
static struct ftrace_ops *removed_ops;
|
|
|
|
|
2014-08-05 23:19:38 +02:00
|
|
|
/*
|
|
|
|
* Set when doing a global update, like enabling all recs or disabling them.
|
|
|
|
* It is not set when just updating a single ftrace_ops.
|
|
|
|
*/
|
|
|
|
static bool update_all_ops;
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
#ifndef CONFIG_FTRACE_MCOUNT_RECORD
|
|
|
|
# error Dynamic ftrace depends on MCOUNT_RECORD
|
|
|
|
#endif
|
|
|
|
|
|
|
|
static struct hlist_head ftrace_func_hash[FTRACE_FUNC_HASHSIZE] __read_mostly;
|
|
|
|
|
|
|
|
struct ftrace_func_probe {
|
|
|
|
struct hlist_node node;
|
|
|
|
struct ftrace_probe_ops *ops;
|
|
|
|
unsigned long flags;
|
|
|
|
unsigned long ip;
|
|
|
|
void *data;
|
2013-03-13 17:42:58 +01:00
|
|
|
struct list_head free_list;
|
2009-03-23 22:12:36 +01:00
|
|
|
};
|
|
|
|
|
2011-04-29 21:12:32 +02:00
|
|
|
struct ftrace_func_entry {
|
|
|
|
struct hlist_node hlist;
|
|
|
|
unsigned long ip;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct ftrace_hash {
|
|
|
|
unsigned long size_bits;
|
|
|
|
struct hlist_head *buckets;
|
|
|
|
unsigned long count;
|
2011-05-06 00:03:47 +02:00
|
|
|
struct rcu_head rcu;
|
2011-04-29 21:12:32 +02:00
|
|
|
};
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
/*
|
|
|
|
* We make these constant because no one should touch them,
|
|
|
|
* but they are used as the default "empty hash", to avoid allocating
|
|
|
|
* it all the time. These are in a read only section such that if
|
|
|
|
* anyone does try to modify it, it will cause an exception.
|
|
|
|
*/
|
|
|
|
static const struct hlist_head empty_buckets[1];
|
|
|
|
static const struct ftrace_hash empty_hash = {
|
|
|
|
.buckets = (struct hlist_head *)empty_buckets,
|
2011-04-30 02:59:51 +02:00
|
|
|
};
|
2011-05-02 23:34:47 +02:00
|
|
|
#define EMPTY_HASH ((struct ftrace_hash *)&empty_hash)
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static struct ftrace_ops global_ops = {
|
2014-08-15 23:23:02 +02:00
|
|
|
.func = ftrace_stub,
|
|
|
|
.local_hash.notrace_hash = EMPTY_HASH,
|
|
|
|
.local_hash.filter_hash = EMPTY_HASH,
|
|
|
|
INIT_OPS_HASH(global_ops)
|
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE |
|
2015-07-24 16:38:12 +02:00
|
|
|
FTRACE_OPS_FL_INITIALIZED |
|
|
|
|
FTRACE_OPS_FL_PID,
|
2011-05-02 18:29:25 +02:00
|
|
|
};
|
|
|
|
|
2014-11-19 03:14:11 +01:00
|
|
|
/*
|
|
|
|
* This is used by __kernel_text_address() to return true if the
|
2014-11-20 16:05:36 +01:00
|
|
|
* address is on a dynamically allocated trampoline that would
|
2014-11-19 03:14:11 +01:00
|
|
|
* not return true for either core_kernel_text() or
|
|
|
|
* is_module_text_address().
|
|
|
|
*/
|
|
|
|
bool is_ftrace_trampoline(unsigned long addr)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *op;
|
|
|
|
bool ret = false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Some of the ops may be dynamically allocated,
|
|
|
|
* they are freed after a synchronize_sched().
|
|
|
|
*/
|
|
|
|
preempt_disable_notrace();
|
|
|
|
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
/*
|
|
|
|
* This is to check for dynamically allocated trampolines.
|
|
|
|
* Trampolines that are in kernel text will have
|
|
|
|
* core_kernel_text() return true.
|
|
|
|
*/
|
|
|
|
if (op->trampoline && op->trampoline_size)
|
|
|
|
if (addr >= op->trampoline &&
|
|
|
|
addr < op->trampoline + op->trampoline_size) {
|
|
|
|
ret = true;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
|
|
|
|
out:
|
|
|
|
preempt_enable_notrace();
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
struct ftrace_page {
|
|
|
|
struct ftrace_page *next;
|
2011-12-16 22:23:44 +01:00
|
|
|
struct dyn_ftrace *records;
|
2009-03-23 22:12:36 +01:00
|
|
|
int index;
|
2011-12-16 22:23:44 +01:00
|
|
|
int size;
|
2009-03-23 22:12:36 +01:00
|
|
|
};
|
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
#define ENTRY_SIZE sizeof(struct dyn_ftrace)
|
|
|
|
#define ENTRIES_PER_PAGE (PAGE_SIZE / ENTRY_SIZE)
|
2009-03-23 22:12:36 +01:00
|
|
|
|
|
|
|
/* estimate from running different kernels */
|
|
|
|
#define NR_TO_INIT 10000
|
|
|
|
|
|
|
|
static struct ftrace_page *ftrace_pages_start;
|
|
|
|
static struct ftrace_page *ftrace_pages;
|
|
|
|
|
2014-05-01 18:44:50 +02:00
|
|
|
static bool __always_inline ftrace_hash_empty(struct ftrace_hash *hash)
|
2011-12-20 01:07:36 +01:00
|
|
|
{
|
|
|
|
return !hash || !hash->count;
|
|
|
|
}
|
|
|
|
|
2011-04-29 21:12:32 +02:00
|
|
|
static struct ftrace_func_entry *
|
|
|
|
ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip)
|
|
|
|
{
|
|
|
|
unsigned long key;
|
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
struct hlist_head *hhd;
|
|
|
|
|
2011-12-20 01:07:36 +01:00
|
|
|
if (ftrace_hash_empty(hash))
|
2011-04-29 21:12:32 +02:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (hash->size_bits > 0)
|
|
|
|
key = hash_long(ip, hash->size_bits);
|
|
|
|
else
|
|
|
|
key = 0;
|
|
|
|
|
|
|
|
hhd = &hash->buckets[key];
|
|
|
|
|
2013-05-28 20:38:43 +02:00
|
|
|
hlist_for_each_entry_rcu_notrace(entry, hhd, hlist) {
|
2011-04-29 21:12:32 +02:00
|
|
|
if (entry->ip == ip)
|
|
|
|
return entry;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static void __add_hash_entry(struct ftrace_hash *hash,
|
|
|
|
struct ftrace_func_entry *entry)
|
2011-04-29 21:12:32 +02:00
|
|
|
{
|
|
|
|
struct hlist_head *hhd;
|
|
|
|
unsigned long key;
|
|
|
|
|
|
|
|
if (hash->size_bits)
|
2011-05-02 23:34:47 +02:00
|
|
|
key = hash_long(entry->ip, hash->size_bits);
|
2011-04-29 21:12:32 +02:00
|
|
|
else
|
|
|
|
key = 0;
|
|
|
|
|
|
|
|
hhd = &hash->buckets[key];
|
|
|
|
hlist_add_head(&entry->hlist, hhd);
|
|
|
|
hash->count++;
|
2011-05-02 23:34:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int add_hash_entry(struct ftrace_hash *hash, unsigned long ip)
|
|
|
|
{
|
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
|
|
|
|
entry = kmalloc(sizeof(*entry), GFP_KERNEL);
|
|
|
|
if (!entry)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
entry->ip = ip;
|
|
|
|
__add_hash_entry(hash, entry);
|
2011-04-29 21:12:32 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2011-05-02 23:34:47 +02:00
|
|
|
free_hash_entry(struct ftrace_hash *hash,
|
2011-04-29 21:12:32 +02:00
|
|
|
struct ftrace_func_entry *entry)
|
|
|
|
{
|
|
|
|
hlist_del(&entry->hlist);
|
|
|
|
kfree(entry);
|
|
|
|
hash->count--;
|
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static void
|
|
|
|
remove_hash_entry(struct ftrace_hash *hash,
|
|
|
|
struct ftrace_func_entry *entry)
|
|
|
|
{
|
|
|
|
hlist_del(&entry->hlist);
|
|
|
|
hash->count--;
|
|
|
|
}
|
|
|
|
|
2011-04-29 21:12:32 +02:00
|
|
|
static void ftrace_hash_clear(struct ftrace_hash *hash)
|
|
|
|
{
|
|
|
|
struct hlist_head *hhd;
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
struct hlist_node *tn;
|
2011-04-29 21:12:32 +02:00
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
int size = 1 << hash->size_bits;
|
|
|
|
int i;
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
if (!hash->count)
|
|
|
|
return;
|
|
|
|
|
2011-04-29 21:12:32 +02:00
|
|
|
for (i = 0; i < size; i++) {
|
|
|
|
hhd = &hash->buckets[i];
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
hlist_for_each_entry_safe(entry, tn, hhd, hlist)
|
2011-05-02 23:34:47 +02:00
|
|
|
free_hash_entry(hash, entry);
|
2011-04-29 21:12:32 +02:00
|
|
|
}
|
|
|
|
FTRACE_WARN_ON(hash->count);
|
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static void free_ftrace_hash(struct ftrace_hash *hash)
|
|
|
|
{
|
|
|
|
if (!hash || hash == EMPTY_HASH)
|
|
|
|
return;
|
|
|
|
ftrace_hash_clear(hash);
|
|
|
|
kfree(hash->buckets);
|
|
|
|
kfree(hash);
|
|
|
|
}
|
|
|
|
|
2011-05-06 00:03:47 +02:00
|
|
|
static void __free_ftrace_hash_rcu(struct rcu_head *rcu)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *hash;
|
|
|
|
|
|
|
|
hash = container_of(rcu, struct ftrace_hash, rcu);
|
|
|
|
free_ftrace_hash(hash);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void free_ftrace_hash_rcu(struct ftrace_hash *hash)
|
|
|
|
{
|
|
|
|
if (!hash || hash == EMPTY_HASH)
|
|
|
|
return;
|
|
|
|
call_rcu_sched(&hash->rcu, __free_ftrace_hash_rcu);
|
|
|
|
}
|
|
|
|
|
ftrace, perf: Add filter support for function trace event
Adding support to filter function trace event via perf
interface. It is now possible to use filter interface
in the perf tool like:
perf record -e ftrace:function --filter="(ip == mm_*)" ls
The filter syntax is restricted to the the 'ip' field only,
and following operators are accepted '==' '!=' '||', ending
up with the filter strings like:
ip == f1[, ]f2 ... || ip != f3[, ]f4 ...
with comma ',' or space ' ' as a function separator. If the
space ' ' is used as a separator, the right side of the
assignment needs to be enclosed in double quotes '"', e.g.:
perf record -e ftrace:function --filter '(ip == do_execve,sys_*,ext*)' ls
perf record -e ftrace:function --filter '(ip == "do_execve,sys_*,ext*")' ls
perf record -e ftrace:function --filter '(ip == "do_execve sys_* ext*")' ls
The '==' operator adds trace filter with same effect as would
be added via set_ftrace_filter file.
The '!=' operator adds trace filter with same effect as would
be added via set_ftrace_notrace file.
The right side of the '!=', '==' operators is list of functions
or regexp. to be added to filter separated by space.
The '||' operator is used for connecting multiple filter definitions
together. It is possible to have more than one '==' and '!='
operators within one filter string.
Link: http://lkml.kernel.org/r/1329317514-8131-8-git-send-email-jolsa@redhat.com
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2012-02-15 15:51:54 +01:00
|
|
|
void ftrace_free_filter(struct ftrace_ops *ops)
|
|
|
|
{
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
2014-08-15 23:23:02 +02:00
|
|
|
free_ftrace_hash(ops->func_hash->filter_hash);
|
|
|
|
free_ftrace_hash(ops->func_hash->notrace_hash);
|
ftrace, perf: Add filter support for function trace event
Adding support to filter function trace event via perf
interface. It is now possible to use filter interface
in the perf tool like:
perf record -e ftrace:function --filter="(ip == mm_*)" ls
The filter syntax is restricted to the the 'ip' field only,
and following operators are accepted '==' '!=' '||', ending
up with the filter strings like:
ip == f1[, ]f2 ... || ip != f3[, ]f4 ...
with comma ',' or space ' ' as a function separator. If the
space ' ' is used as a separator, the right side of the
assignment needs to be enclosed in double quotes '"', e.g.:
perf record -e ftrace:function --filter '(ip == do_execve,sys_*,ext*)' ls
perf record -e ftrace:function --filter '(ip == "do_execve,sys_*,ext*")' ls
perf record -e ftrace:function --filter '(ip == "do_execve sys_* ext*")' ls
The '==' operator adds trace filter with same effect as would
be added via set_ftrace_filter file.
The '!=' operator adds trace filter with same effect as would
be added via set_ftrace_notrace file.
The right side of the '!=', '==' operators is list of functions
or regexp. to be added to filter separated by space.
The '||' operator is used for connecting multiple filter definitions
together. It is possible to have more than one '==' and '!='
operators within one filter string.
Link: http://lkml.kernel.org/r/1329317514-8131-8-git-send-email-jolsa@redhat.com
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2012-02-15 15:51:54 +01:00
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static struct ftrace_hash *alloc_ftrace_hash(int size_bits)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *hash;
|
|
|
|
int size;
|
|
|
|
|
|
|
|
hash = kzalloc(sizeof(*hash), GFP_KERNEL);
|
|
|
|
if (!hash)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
size = 1 << size_bits;
|
2011-11-29 22:08:00 +01:00
|
|
|
hash->buckets = kcalloc(size, sizeof(*hash->buckets), GFP_KERNEL);
|
2011-05-02 23:34:47 +02:00
|
|
|
|
|
|
|
if (!hash->buckets) {
|
|
|
|
kfree(hash);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
hash->size_bits = size_bits;
|
|
|
|
|
|
|
|
return hash;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct ftrace_hash *
|
|
|
|
alloc_and_copy_ftrace_hash(int size_bits, struct ftrace_hash *hash)
|
|
|
|
{
|
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
struct ftrace_hash *new_hash;
|
|
|
|
int size;
|
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
new_hash = alloc_ftrace_hash(size_bits);
|
|
|
|
if (!new_hash)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/* Empty hash? */
|
2011-12-20 01:07:36 +01:00
|
|
|
if (ftrace_hash_empty(hash))
|
2011-05-02 23:34:47 +02:00
|
|
|
return new_hash;
|
|
|
|
|
|
|
|
size = 1 << hash->size_bits;
|
|
|
|
for (i = 0; i < size; i++) {
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
hlist_for_each_entry(entry, &hash->buckets[i], hlist) {
|
2011-05-02 23:34:47 +02:00
|
|
|
ret = add_hash_entry(new_hash, entry->ip);
|
|
|
|
if (ret < 0)
|
|
|
|
goto free_hash;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
FTRACE_WARN_ON(new_hash->count != hash->count);
|
|
|
|
|
|
|
|
return new_hash;
|
|
|
|
|
|
|
|
free_hash:
|
|
|
|
free_ftrace_hash(new_hash);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2011-07-13 21:03:44 +02:00
|
|
|
static void
|
2014-08-18 19:21:08 +02:00
|
|
|
ftrace_hash_rec_disable_modify(struct ftrace_ops *ops, int filter_hash);
|
2011-07-13 21:03:44 +02:00
|
|
|
static void
|
2014-08-18 19:21:08 +02:00
|
|
|
ftrace_hash_rec_enable_modify(struct ftrace_ops *ops, int filter_hash);
|
2011-07-13 21:03:44 +02:00
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
static int ftrace_hash_ipmodify_update(struct ftrace_ops *ops,
|
|
|
|
struct ftrace_hash *new_hash);
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static int
|
2011-07-13 21:03:44 +02:00
|
|
|
ftrace_hash_move(struct ftrace_ops *ops, int enable,
|
|
|
|
struct ftrace_hash **dst, struct ftrace_hash *src)
|
2011-05-02 23:34:47 +02:00
|
|
|
{
|
|
|
|
struct ftrace_func_entry *entry;
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
struct hlist_node *tn;
|
2011-05-02 23:34:47 +02:00
|
|
|
struct hlist_head *hhd;
|
2011-05-06 00:03:47 +02:00
|
|
|
struct ftrace_hash *new_hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
int size = src->count;
|
|
|
|
int bits = 0;
|
2014-11-21 11:25:16 +01:00
|
|
|
int ret;
|
2011-05-02 23:34:47 +02:00
|
|
|
int i;
|
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
/* Reject setting notrace hash on IPMODIFY ftrace_ops */
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_IPMODIFY && !enable)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
/*
|
|
|
|
* If the new source is empty, just free dst and assign it
|
|
|
|
* the empty_hash.
|
|
|
|
*/
|
|
|
|
if (!src->count) {
|
2014-06-17 13:04:42 +02:00
|
|
|
new_hash = EMPTY_HASH;
|
|
|
|
goto update;
|
2011-05-02 23:34:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make the hash size about 1/2 the # found
|
|
|
|
*/
|
|
|
|
for (size /= 2; size; size >>= 1)
|
|
|
|
bits++;
|
|
|
|
|
|
|
|
/* Don't allocate too much */
|
|
|
|
if (bits > FTRACE_HASH_MAX_BITS)
|
|
|
|
bits = FTRACE_HASH_MAX_BITS;
|
|
|
|
|
2011-05-06 00:03:47 +02:00
|
|
|
new_hash = alloc_ftrace_hash(bits);
|
|
|
|
if (!new_hash)
|
2014-06-17 13:04:42 +02:00
|
|
|
return -ENOMEM;
|
2011-05-02 23:34:47 +02:00
|
|
|
|
|
|
|
size = 1 << src->size_bits;
|
|
|
|
for (i = 0; i < size; i++) {
|
|
|
|
hhd = &src->buckets[i];
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
hlist_for_each_entry_safe(entry, tn, hhd, hlist) {
|
2011-05-02 23:34:47 +02:00
|
|
|
remove_hash_entry(src, entry);
|
2011-05-06 00:03:47 +02:00
|
|
|
__add_hash_entry(new_hash, entry);
|
2011-05-02 23:34:47 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-06-17 13:04:42 +02:00
|
|
|
update:
|
2014-11-21 11:25:16 +01:00
|
|
|
/* Make sure this can be applied if it is IPMODIFY ftrace_ops */
|
|
|
|
if (enable) {
|
|
|
|
/* IPMODIFY should be updated only when filter_hash updating */
|
|
|
|
ret = ftrace_hash_ipmodify_update(ops, new_hash);
|
|
|
|
if (ret < 0) {
|
|
|
|
free_ftrace_hash(new_hash);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-06-17 13:04:42 +02:00
|
|
|
/*
|
|
|
|
* Remove the current set, update the hash and add
|
|
|
|
* them back.
|
|
|
|
*/
|
2014-08-18 19:21:08 +02:00
|
|
|
ftrace_hash_rec_disable_modify(ops, enable);
|
2014-06-17 13:04:42 +02:00
|
|
|
|
2011-05-06 00:03:47 +02:00
|
|
|
rcu_assign_pointer(*dst, new_hash);
|
|
|
|
|
2014-08-18 19:21:08 +02:00
|
|
|
ftrace_hash_rec_enable_modify(ops, enable);
|
2011-07-13 21:03:44 +02:00
|
|
|
|
2014-06-17 13:04:42 +02:00
|
|
|
return 0;
|
2011-05-02 23:34:47 +02:00
|
|
|
}
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
static bool hash_contains_ip(unsigned long ip,
|
|
|
|
struct ftrace_ops_hash *hash)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* The function record is a match if it exists in the filter
|
|
|
|
* hash and not in the notrace hash. Note, an emty hash is
|
|
|
|
* considered a match for the filter hash, but an empty
|
|
|
|
* notrace hash is considered not in the notrace hash.
|
|
|
|
*/
|
|
|
|
return (ftrace_hash_empty(hash->filter_hash) ||
|
|
|
|
ftrace_lookup_ip(hash->filter_hash, ip)) &&
|
|
|
|
(ftrace_hash_empty(hash->notrace_hash) ||
|
|
|
|
!ftrace_lookup_ip(hash->notrace_hash, ip));
|
|
|
|
}
|
|
|
|
|
2011-05-04 15:27:52 +02:00
|
|
|
/*
|
|
|
|
* Test the hashes for this ops to see if we want to call
|
|
|
|
* the ops->func or not.
|
|
|
|
*
|
|
|
|
* It's a match if the ip is in the ops->filter_hash or
|
|
|
|
* the filter_hash does not exist or is empty,
|
|
|
|
* AND
|
|
|
|
* the ip is not in the ops->notrace_hash.
|
2011-05-06 03:14:55 +02:00
|
|
|
*
|
|
|
|
* This needs to be called with preemption disabled as
|
|
|
|
* the hashes are freed with call_rcu_sched().
|
2011-05-04 15:27:52 +02:00
|
|
|
*/
|
|
|
|
static int
|
2013-07-24 04:06:15 +02:00
|
|
|
ftrace_ops_test(struct ftrace_ops *ops, unsigned long ip, void *regs)
|
2011-05-04 15:27:52 +02:00
|
|
|
{
|
2014-07-24 18:25:47 +02:00
|
|
|
struct ftrace_ops_hash hash;
|
2011-05-04 15:27:52 +02:00
|
|
|
int ret;
|
|
|
|
|
2013-07-24 04:06:15 +02:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
|
|
|
|
/*
|
|
|
|
* There's a small race when adding ops that the ftrace handler
|
|
|
|
* that wants regs, may be called without them. We can not
|
|
|
|
* allow that handler to be called if regs is NULL.
|
|
|
|
*/
|
|
|
|
if (regs == NULL && (ops->flags & FTRACE_OPS_FL_SAVE_REGS))
|
|
|
|
return 0;
|
|
|
|
#endif
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
hash.filter_hash = rcu_dereference_raw_notrace(ops->func_hash->filter_hash);
|
|
|
|
hash.notrace_hash = rcu_dereference_raw_notrace(ops->func_hash->notrace_hash);
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
if (hash_contains_ip(ip, &hash))
|
2011-05-04 15:27:52 +02:00
|
|
|
ret = 1;
|
|
|
|
else
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-03-23 22:12:36 +01:00
|
|
|
/*
|
|
|
|
* This is a double for. Do not use 'break' to break out of the loop,
|
|
|
|
* you must use a goto.
|
|
|
|
*/
|
|
|
|
#define do_for_each_ftrace_rec(pg, rec) \
|
|
|
|
for (pg = ftrace_pages_start; pg; pg = pg->next) { \
|
|
|
|
int _____i; \
|
|
|
|
for (_____i = 0; _____i < pg->index; _____i++) { \
|
|
|
|
rec = &pg->records[_____i];
|
|
|
|
|
|
|
|
#define while_for_each_ftrace_rec() \
|
|
|
|
} \
|
|
|
|
}
|
|
|
|
|
2011-12-17 01:27:42 +01:00
|
|
|
|
|
|
|
static int ftrace_cmp_recs(const void *a, const void *b)
|
|
|
|
{
|
2012-04-25 19:48:13 +02:00
|
|
|
const struct dyn_ftrace *key = a;
|
|
|
|
const struct dyn_ftrace *rec = b;
|
2011-12-17 01:27:42 +01:00
|
|
|
|
2012-04-25 19:48:13 +02:00
|
|
|
if (key->flags < rec->ip)
|
2011-12-17 01:27:42 +01:00
|
|
|
return -1;
|
2012-04-25 19:48:13 +02:00
|
|
|
if (key->ip >= rec->ip + MCOUNT_INSN_SIZE)
|
|
|
|
return 1;
|
2011-12-17 01:27:42 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-25 20:39:54 +02:00
|
|
|
static unsigned long ftrace_location_range(unsigned long start, unsigned long end)
|
2011-08-16 15:53:39 +02:00
|
|
|
{
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec;
|
2011-12-17 01:27:42 +01:00
|
|
|
struct dyn_ftrace key;
|
2011-08-16 15:53:39 +02:00
|
|
|
|
2012-04-25 19:48:13 +02:00
|
|
|
key.ip = start;
|
|
|
|
key.flags = end; /* overload flags, as it is unsigned long */
|
2011-12-17 01:27:42 +01:00
|
|
|
|
|
|
|
for (pg = ftrace_pages_start; pg; pg = pg->next) {
|
2012-04-25 19:48:13 +02:00
|
|
|
if (end < pg->records[0].ip ||
|
|
|
|
start >= (pg->records[pg->index - 1].ip + MCOUNT_INSN_SIZE))
|
2012-04-25 16:14:43 +02:00
|
|
|
continue;
|
2011-12-17 01:27:42 +01:00
|
|
|
rec = bsearch(&key, pg->records, pg->index,
|
|
|
|
sizeof(struct dyn_ftrace),
|
|
|
|
ftrace_cmp_recs);
|
|
|
|
if (rec)
|
2012-04-25 20:39:54 +02:00
|
|
|
return rec->ip;
|
2011-12-17 01:27:42 +01:00
|
|
|
}
|
2011-08-16 15:53:39 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-25 19:48:13 +02:00
|
|
|
/**
|
|
|
|
* ftrace_location - return true if the ip giving is a traced location
|
|
|
|
* @ip: the instruction pointer to check
|
|
|
|
*
|
2012-04-25 20:39:54 +02:00
|
|
|
* Returns rec->ip if @ip given is a pointer to a ftrace location.
|
2012-04-25 19:48:13 +02:00
|
|
|
* That is, the instruction that is either a NOP or call to
|
|
|
|
* the function tracer. It checks the ftrace internal tables to
|
|
|
|
* determine if the address belongs or not.
|
|
|
|
*/
|
2012-04-25 20:39:54 +02:00
|
|
|
unsigned long ftrace_location(unsigned long ip)
|
2012-04-25 19:48:13 +02:00
|
|
|
{
|
|
|
|
return ftrace_location_range(ip, ip);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_text_reserved - return true if range contains an ftrace location
|
|
|
|
* @start: start of range to search
|
|
|
|
* @end: end of range to search (inclusive). @end points to the last byte to check.
|
|
|
|
*
|
|
|
|
* Returns 1 if @start and @end contains a ftrace location.
|
|
|
|
* That is, the instruction that is either a NOP or call to
|
|
|
|
* the function tracer. It checks the ftrace internal tables to
|
|
|
|
* determine if the address belongs or not.
|
|
|
|
*/
|
2013-01-10 00:09:20 +01:00
|
|
|
int ftrace_text_reserved(const void *start, const void *end)
|
2012-04-25 19:48:13 +02:00
|
|
|
{
|
2012-04-25 20:39:54 +02:00
|
|
|
unsigned long ret;
|
|
|
|
|
|
|
|
ret = ftrace_location_range((unsigned long)start,
|
|
|
|
(unsigned long)end);
|
|
|
|
|
|
|
|
return (int)!!ret;
|
2012-04-25 19:48:13 +02:00
|
|
|
}
|
|
|
|
|
ftrace: Allow no regs if no more callbacks require it
When registering a function callback for the function tracer, the ops
can specify if it wants to save full regs (like an interrupt would)
for each function that it traces, or if it does not care about regs
and just wants to have the fastest return possible.
Once a ops has registered a function, if other ops register that
function they all will receive the regs too. That's because it does
the work once, it does it for everyone.
Now if the ops wanting regs unregisters the function so that there's
only ops left that do not care about regs, those ops will still
continue getting regs and going through the work for it on that
function. This is because the disabling of the rec counter only
sees the ops registered, and does not see the ops that are still
attached, and does not know if the current ops that are still attached
want regs or not. To play it safe, it just keeps regs being processed
until no function is registered anymore.
Instead of doing that, check the ops that are still registered for that
function and if none want regs for it anymore, then disable the
processing of regs.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-01 04:35:48 +02:00
|
|
|
/* Test if ops registered to this rec needs regs */
|
|
|
|
static bool test_rec_ops_needs_regs(struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *ops;
|
|
|
|
bool keep_regs = false;
|
|
|
|
|
|
|
|
for (ops = ftrace_ops_list;
|
|
|
|
ops != &ftrace_list_end; ops = ops->next) {
|
|
|
|
/* pass rec in as regs to have non-NULL val */
|
|
|
|
if (ftrace_ops_test(ops, rec->ip, rec)) {
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_SAVE_REGS) {
|
|
|
|
keep_regs = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return keep_regs;
|
|
|
|
}
|
|
|
|
|
2011-05-03 19:25:24 +02:00
|
|
|
static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
|
|
|
|
int filter_hash,
|
|
|
|
bool inc)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *hash;
|
|
|
|
struct ftrace_hash *other_hash;
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec;
|
|
|
|
int count = 0;
|
|
|
|
int all = 0;
|
|
|
|
|
|
|
|
/* Only update if the ops has been registered */
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_ENABLED))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* In the filter_hash case:
|
|
|
|
* If the count is zero, we update all records.
|
|
|
|
* Otherwise we just update the items in the hash.
|
|
|
|
*
|
|
|
|
* In the notrace_hash case:
|
|
|
|
* We enable the update in the hash.
|
|
|
|
* As disabling notrace means enabling the tracing,
|
|
|
|
* and enabling notrace means disabling, the inc variable
|
|
|
|
* gets inversed.
|
|
|
|
*/
|
|
|
|
if (filter_hash) {
|
2014-08-15 23:23:02 +02:00
|
|
|
hash = ops->func_hash->filter_hash;
|
|
|
|
other_hash = ops->func_hash->notrace_hash;
|
2011-12-20 01:07:36 +01:00
|
|
|
if (ftrace_hash_empty(hash))
|
2011-05-03 19:25:24 +02:00
|
|
|
all = 1;
|
|
|
|
} else {
|
|
|
|
inc = !inc;
|
2014-08-15 23:23:02 +02:00
|
|
|
hash = ops->func_hash->notrace_hash;
|
|
|
|
other_hash = ops->func_hash->filter_hash;
|
2011-05-03 19:25:24 +02:00
|
|
|
/*
|
|
|
|
* If the notrace hash has no items,
|
|
|
|
* then there's nothing to do.
|
|
|
|
*/
|
2011-12-20 01:07:36 +01:00
|
|
|
if (ftrace_hash_empty(hash))
|
2011-05-03 19:25:24 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
|
|
|
int in_other_hash = 0;
|
|
|
|
int in_hash = 0;
|
|
|
|
int match = 0;
|
|
|
|
|
|
|
|
if (all) {
|
|
|
|
/*
|
|
|
|
* Only the filter_hash affects all records.
|
|
|
|
* Update if the record is not in the notrace hash.
|
|
|
|
*/
|
2011-05-04 15:27:52 +02:00
|
|
|
if (!other_hash || !ftrace_lookup_ip(other_hash, rec->ip))
|
2011-05-03 19:25:24 +02:00
|
|
|
match = 1;
|
|
|
|
} else {
|
2011-12-20 01:07:36 +01:00
|
|
|
in_hash = !!ftrace_lookup_ip(hash, rec->ip);
|
|
|
|
in_other_hash = !!ftrace_lookup_ip(other_hash, rec->ip);
|
2011-05-03 19:25:24 +02:00
|
|
|
|
|
|
|
/*
|
2014-05-07 21:06:14 +02:00
|
|
|
* If filter_hash is set, we want to match all functions
|
|
|
|
* that are in the hash but not in the other hash.
|
2011-05-03 19:25:24 +02:00
|
|
|
*
|
2014-05-07 21:06:14 +02:00
|
|
|
* If filter_hash is not set, then we are decrementing.
|
|
|
|
* That means we match anything that is in the hash
|
|
|
|
* and also in the other_hash. That is, we need to turn
|
|
|
|
* off functions in the other hash because they are disabled
|
|
|
|
* by this hash.
|
2011-05-03 19:25:24 +02:00
|
|
|
*/
|
|
|
|
if (filter_hash && in_hash && !in_other_hash)
|
|
|
|
match = 1;
|
|
|
|
else if (!filter_hash && in_hash &&
|
2011-12-20 01:07:36 +01:00
|
|
|
(in_other_hash || ftrace_hash_empty(other_hash)))
|
2011-05-03 19:25:24 +02:00
|
|
|
match = 1;
|
|
|
|
}
|
|
|
|
if (!match)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (inc) {
|
|
|
|
rec->flags++;
|
2014-05-07 19:46:45 +02:00
|
|
|
if (FTRACE_WARN_ON(ftrace_rec_count(rec) == FTRACE_REF_MAX))
|
2011-05-03 19:25:24 +02:00
|
|
|
return;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If there's only a single callback registered to a
|
|
|
|
* function, and the ops has a trampoline registered
|
|
|
|
* for it, then we can call it directly.
|
|
|
|
*/
|
2014-07-24 18:25:47 +02:00
|
|
|
if (ftrace_rec_count(rec) == 1 && ops->trampoline)
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
rec->flags |= FTRACE_FL_TRAMP;
|
2014-07-24 18:25:47 +02:00
|
|
|
else
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
/*
|
|
|
|
* If we are adding another function callback
|
|
|
|
* to this function, and the previous had a
|
2014-08-21 05:57:04 +02:00
|
|
|
* custom trampoline in use, then we need to go
|
|
|
|
* back to the default trampoline.
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
*/
|
2014-07-24 18:25:47 +02:00
|
|
|
rec->flags &= ~FTRACE_FL_TRAMP;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
2012-04-30 22:20:23 +02:00
|
|
|
/*
|
|
|
|
* If any ops wants regs saved for this function
|
|
|
|
* then all ops will get saved regs.
|
|
|
|
*/
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_SAVE_REGS)
|
|
|
|
rec->flags |= FTRACE_FL_REGS;
|
2011-05-03 19:25:24 +02:00
|
|
|
} else {
|
2014-05-07 19:46:45 +02:00
|
|
|
if (FTRACE_WARN_ON(ftrace_rec_count(rec) == 0))
|
2011-05-03 19:25:24 +02:00
|
|
|
return;
|
|
|
|
rec->flags--;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
ftrace: Allow no regs if no more callbacks require it
When registering a function callback for the function tracer, the ops
can specify if it wants to save full regs (like an interrupt would)
for each function that it traces, or if it does not care about regs
and just wants to have the fastest return possible.
Once a ops has registered a function, if other ops register that
function they all will receive the regs too. That's because it does
the work once, it does it for everyone.
Now if the ops wanting regs unregisters the function so that there's
only ops left that do not care about regs, those ops will still
continue getting regs and going through the work for it on that
function. This is because the disabling of the rec counter only
sees the ops registered, and does not see the ops that are still
attached, and does not know if the current ops that are still attached
want regs or not. To play it safe, it just keeps regs being processed
until no function is registered anymore.
Instead of doing that, check the ops that are still registered for that
function and if none want regs for it anymore, then disable the
processing of regs.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-01 04:35:48 +02:00
|
|
|
/*
|
|
|
|
* If the rec had REGS enabled and the ops that is
|
|
|
|
* being removed had REGS set, then see if there is
|
|
|
|
* still any ops for this record that wants regs.
|
|
|
|
* If not, we can stop recording them.
|
|
|
|
*/
|
2014-05-07 19:46:45 +02:00
|
|
|
if (ftrace_rec_count(rec) > 0 &&
|
ftrace: Allow no regs if no more callbacks require it
When registering a function callback for the function tracer, the ops
can specify if it wants to save full regs (like an interrupt would)
for each function that it traces, or if it does not care about regs
and just wants to have the fastest return possible.
Once a ops has registered a function, if other ops register that
function they all will receive the regs too. That's because it does
the work once, it does it for everyone.
Now if the ops wanting regs unregisters the function so that there's
only ops left that do not care about regs, those ops will still
continue getting regs and going through the work for it on that
function. This is because the disabling of the rec counter only
sees the ops registered, and does not see the ops that are still
attached, and does not know if the current ops that are still attached
want regs or not. To play it safe, it just keeps regs being processed
until no function is registered anymore.
Instead of doing that, check the ops that are still registered for that
function and if none want regs for it anymore, then disable the
processing of regs.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-01 04:35:48 +02:00
|
|
|
rec->flags & FTRACE_FL_REGS &&
|
|
|
|
ops->flags & FTRACE_OPS_FL_SAVE_REGS) {
|
|
|
|
if (!test_rec_ops_needs_regs(rec))
|
|
|
|
rec->flags &= ~FTRACE_FL_REGS;
|
|
|
|
}
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
/*
|
|
|
|
* If the rec had TRAMP enabled, then it needs to
|
|
|
|
* be cleared. As TRAMP can only be enabled iff
|
|
|
|
* there is only a single ops attached to it.
|
|
|
|
* In otherwords, always disable it on decrementing.
|
|
|
|
* In the future, we may set it if rec count is
|
|
|
|
* decremented to one, and the ops that is left
|
|
|
|
* has a trampoline.
|
|
|
|
*/
|
|
|
|
rec->flags &= ~FTRACE_FL_TRAMP;
|
|
|
|
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
/*
|
|
|
|
* flags will be cleared in ftrace_check_record()
|
|
|
|
* if rec count is zero.
|
|
|
|
*/
|
2011-05-03 19:25:24 +02:00
|
|
|
}
|
|
|
|
count++;
|
|
|
|
/* Shortcut, if we handled all records, we are done. */
|
|
|
|
if (!all && count == hash->count)
|
|
|
|
return;
|
|
|
|
} while_for_each_ftrace_rec();
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_hash_rec_disable(struct ftrace_ops *ops,
|
|
|
|
int filter_hash)
|
|
|
|
{
|
|
|
|
__ftrace_hash_rec_update(ops, filter_hash, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_hash_rec_enable(struct ftrace_ops *ops,
|
|
|
|
int filter_hash)
|
|
|
|
{
|
|
|
|
__ftrace_hash_rec_update(ops, filter_hash, 1);
|
|
|
|
}
|
|
|
|
|
2014-08-18 19:21:08 +02:00
|
|
|
static void ftrace_hash_rec_update_modify(struct ftrace_ops *ops,
|
|
|
|
int filter_hash, int inc)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *op;
|
|
|
|
|
|
|
|
__ftrace_hash_rec_update(ops, filter_hash, inc);
|
|
|
|
|
|
|
|
if (ops->func_hash != &global_ops.local_hash)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the ops shares the global_ops hash, then we need to update
|
|
|
|
* all ops that are enabled and use this hash.
|
|
|
|
*/
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
/* Already done */
|
|
|
|
if (op == ops)
|
|
|
|
continue;
|
|
|
|
if (op->func_hash == &global_ops.local_hash)
|
|
|
|
__ftrace_hash_rec_update(op, filter_hash, inc);
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_hash_rec_disable_modify(struct ftrace_ops *ops,
|
|
|
|
int filter_hash)
|
|
|
|
{
|
|
|
|
ftrace_hash_rec_update_modify(ops, filter_hash, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_hash_rec_enable_modify(struct ftrace_ops *ops,
|
|
|
|
int filter_hash)
|
|
|
|
{
|
|
|
|
ftrace_hash_rec_update_modify(ops, filter_hash, 1);
|
|
|
|
}
|
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
/*
|
|
|
|
* Try to update IPMODIFY flag on each ftrace_rec. Return 0 if it is OK
|
|
|
|
* or no-needed to update, -EBUSY if it detects a conflict of the flag
|
|
|
|
* on a ftrace_rec, and -EINVAL if the new_hash tries to trace all recs.
|
|
|
|
* Note that old_hash and new_hash has below meanings
|
|
|
|
* - If the hash is NULL, it hits all recs (if IPMODIFY is set, this is rejected)
|
|
|
|
* - If the hash is EMPTY_HASH, it hits nothing
|
|
|
|
* - Anything else hits the recs which match the hash entries.
|
|
|
|
*/
|
|
|
|
static int __ftrace_hash_update_ipmodify(struct ftrace_ops *ops,
|
|
|
|
struct ftrace_hash *old_hash,
|
|
|
|
struct ftrace_hash *new_hash)
|
|
|
|
{
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec, *end = NULL;
|
|
|
|
int in_old, in_new;
|
|
|
|
|
|
|
|
/* Only update if the ops has been registered */
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_ENABLED))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_IPMODIFY))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Since the IPMODIFY is a very address sensitive action, we do not
|
|
|
|
* allow ftrace_ops to set all functions to new hash.
|
|
|
|
*/
|
|
|
|
if (!new_hash || !old_hash)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* Update rec->flags */
|
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
|
|
|
/* We need to update only differences of filter_hash */
|
|
|
|
in_old = !!ftrace_lookup_ip(old_hash, rec->ip);
|
|
|
|
in_new = !!ftrace_lookup_ip(new_hash, rec->ip);
|
|
|
|
if (in_old == in_new)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (in_new) {
|
|
|
|
/* New entries must ensure no others are using it */
|
|
|
|
if (rec->flags & FTRACE_FL_IPMODIFY)
|
|
|
|
goto rollback;
|
|
|
|
rec->flags |= FTRACE_FL_IPMODIFY;
|
|
|
|
} else /* Removed entry */
|
|
|
|
rec->flags &= ~FTRACE_FL_IPMODIFY;
|
|
|
|
} while_for_each_ftrace_rec();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
rollback:
|
|
|
|
end = rec;
|
|
|
|
|
|
|
|
/* Roll back what we did above */
|
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
|
|
|
if (rec == end)
|
|
|
|
goto err_out;
|
|
|
|
|
|
|
|
in_old = !!ftrace_lookup_ip(old_hash, rec->ip);
|
|
|
|
in_new = !!ftrace_lookup_ip(new_hash, rec->ip);
|
|
|
|
if (in_old == in_new)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (in_new)
|
|
|
|
rec->flags &= ~FTRACE_FL_IPMODIFY;
|
|
|
|
else
|
|
|
|
rec->flags |= FTRACE_FL_IPMODIFY;
|
|
|
|
} while_for_each_ftrace_rec();
|
|
|
|
|
|
|
|
err_out:
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ftrace_hash_ipmodify_enable(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *hash = ops->func_hash->filter_hash;
|
|
|
|
|
|
|
|
if (ftrace_hash_empty(hash))
|
|
|
|
hash = NULL;
|
|
|
|
|
|
|
|
return __ftrace_hash_update_ipmodify(ops, EMPTY_HASH, hash);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Disabling always succeeds */
|
|
|
|
static void ftrace_hash_ipmodify_disable(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *hash = ops->func_hash->filter_hash;
|
|
|
|
|
|
|
|
if (ftrace_hash_empty(hash))
|
|
|
|
hash = NULL;
|
|
|
|
|
|
|
|
__ftrace_hash_update_ipmodify(ops, hash, EMPTY_HASH);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ftrace_hash_ipmodify_update(struct ftrace_ops *ops,
|
|
|
|
struct ftrace_hash *new_hash)
|
|
|
|
{
|
|
|
|
struct ftrace_hash *old_hash = ops->func_hash->filter_hash;
|
|
|
|
|
|
|
|
if (ftrace_hash_empty(old_hash))
|
|
|
|
old_hash = NULL;
|
|
|
|
|
|
|
|
if (ftrace_hash_empty(new_hash))
|
|
|
|
new_hash = NULL;
|
|
|
|
|
|
|
|
return __ftrace_hash_update_ipmodify(ops, old_hash, new_hash);
|
|
|
|
}
|
|
|
|
|
2008-11-15 01:21:19 +01:00
|
|
|
static void print_ip_ins(const char *fmt, unsigned char *p)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
printk(KERN_CONT "%s", fmt);
|
|
|
|
|
|
|
|
for (i = 0; i < MCOUNT_INSN_SIZE; i++)
|
|
|
|
printk(KERN_CONT "%s%02x", i ? ":" : "", p[i]);
|
|
|
|
}
|
|
|
|
|
2014-10-24 23:56:04 +02:00
|
|
|
static struct ftrace_ops *
|
|
|
|
ftrace_find_tramp_ops_any(struct dyn_ftrace *rec);
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
/**
|
|
|
|
* ftrace_bug - report and shutdown function tracer
|
|
|
|
* @failed: The failed type (EFAULT, EINVAL, EPERM)
|
2014-10-24 23:56:04 +02:00
|
|
|
* @rec: The record that failed
|
2011-08-16 15:53:39 +02:00
|
|
|
*
|
|
|
|
* The arch code that enables or disables the function tracing
|
|
|
|
* can call ftrace_bug() when it has detected a problem in
|
|
|
|
* modifying the code. @failed should be one of either:
|
|
|
|
* EFAULT - if the problem happens on reading the @ip address
|
|
|
|
* EINVAL - if what is read at @ip is not what was expected
|
|
|
|
* EPERM - if the problem happens on writting to the @ip address
|
|
|
|
*/
|
2014-10-24 23:56:04 +02:00
|
|
|
void ftrace_bug(int failed, struct dyn_ftrace *rec)
|
2008-11-15 01:21:19 +01:00
|
|
|
{
|
2014-10-24 23:56:04 +02:00
|
|
|
unsigned long ip = rec ? rec->ip : 0;
|
|
|
|
|
2008-11-15 01:21:19 +01:00
|
|
|
switch (failed) {
|
|
|
|
case -EFAULT:
|
|
|
|
FTRACE_WARN_ON_ONCE(1);
|
|
|
|
pr_info("ftrace faulted on modifying ");
|
|
|
|
print_ip_sym(ip);
|
|
|
|
break;
|
|
|
|
case -EINVAL:
|
|
|
|
FTRACE_WARN_ON_ONCE(1);
|
|
|
|
pr_info("ftrace failed to modify ");
|
|
|
|
print_ip_sym(ip);
|
|
|
|
print_ip_ins(" actual: ", (unsigned char *)ip);
|
2014-10-24 23:56:04 +02:00
|
|
|
pr_cont("\n");
|
2008-11-15 01:21:19 +01:00
|
|
|
break;
|
|
|
|
case -EPERM:
|
|
|
|
FTRACE_WARN_ON_ONCE(1);
|
|
|
|
pr_info("ftrace faulted on writing ");
|
|
|
|
print_ip_sym(ip);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
FTRACE_WARN_ON_ONCE(1);
|
|
|
|
pr_info("ftrace faulted on unknown error ");
|
|
|
|
print_ip_sym(ip);
|
|
|
|
}
|
2014-10-24 23:56:04 +02:00
|
|
|
if (rec) {
|
|
|
|
struct ftrace_ops *ops = NULL;
|
|
|
|
|
|
|
|
pr_info("ftrace record flags: %lx\n", rec->flags);
|
|
|
|
pr_cont(" (%ld)%s", ftrace_rec_count(rec),
|
|
|
|
rec->flags & FTRACE_FL_REGS ? " R" : " ");
|
|
|
|
if (rec->flags & FTRACE_FL_TRAMP_EN) {
|
|
|
|
ops = ftrace_find_tramp_ops_any(rec);
|
|
|
|
if (ops)
|
|
|
|
pr_cont("\ttramp: %pS",
|
|
|
|
(void *)ops->trampoline);
|
|
|
|
else
|
|
|
|
pr_cont("\ttramp: ERROR!");
|
|
|
|
|
|
|
|
}
|
|
|
|
ip = ftrace_get_addr_curr(rec);
|
|
|
|
pr_cont(" expected tramp: %lx\n", ip);
|
|
|
|
}
|
2008-11-15 01:21:19 +01:00
|
|
|
}
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
static int ftrace_check_record(struct dyn_ftrace *rec, int enable, int update)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
2009-07-15 06:32:15 +02:00
|
|
|
unsigned long flag = 0UL;
|
2008-11-16 06:02:06 +01:00
|
|
|
|
2008-11-15 22:31:41 +01:00
|
|
|
/*
|
2011-12-05 18:22:48 +01:00
|
|
|
* If we are updating calls:
|
2008-11-15 22:31:41 +01:00
|
|
|
*
|
2011-05-03 19:25:24 +02:00
|
|
|
* If the record has a ref count, then we need to enable it
|
|
|
|
* because someone is using it.
|
2008-11-15 22:31:41 +01:00
|
|
|
*
|
2011-05-03 19:25:24 +02:00
|
|
|
* Otherwise we make sure its disabled.
|
|
|
|
*
|
2011-12-05 18:22:48 +01:00
|
|
|
* If we are disabling calls, then disable all records that
|
2011-05-03 19:25:24 +02:00
|
|
|
* are enabled.
|
2008-11-15 22:31:41 +01:00
|
|
|
*/
|
2014-05-07 19:46:45 +02:00
|
|
|
if (enable && ftrace_rec_count(rec))
|
2011-05-03 19:25:24 +02:00
|
|
|
flag = FTRACE_FL_ENABLED;
|
2008-11-15 22:31:41 +01:00
|
|
|
|
2012-04-30 22:20:23 +02:00
|
|
|
/*
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
* If enabling and the REGS flag does not match the REGS_EN, or
|
|
|
|
* the TRAMP flag doesn't match the TRAMP_EN, then do not ignore
|
|
|
|
* this record. Set flags to fail the compare against ENABLED.
|
2012-04-30 22:20:23 +02:00
|
|
|
*/
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
if (flag) {
|
|
|
|
if (!(rec->flags & FTRACE_FL_REGS) !=
|
|
|
|
!(rec->flags & FTRACE_FL_REGS_EN))
|
|
|
|
flag |= FTRACE_FL_REGS;
|
|
|
|
|
|
|
|
if (!(rec->flags & FTRACE_FL_TRAMP) !=
|
|
|
|
!(rec->flags & FTRACE_FL_TRAMP_EN))
|
|
|
|
flag |= FTRACE_FL_TRAMP;
|
|
|
|
}
|
2012-04-30 22:20:23 +02:00
|
|
|
|
2009-07-15 06:32:15 +02:00
|
|
|
/* If the state of this record hasn't changed, then do nothing */
|
|
|
|
if ((rec->flags & FTRACE_FL_ENABLED) == flag)
|
2011-08-16 15:53:39 +02:00
|
|
|
return FTRACE_UPDATE_IGNORE;
|
2008-11-15 22:31:41 +01:00
|
|
|
|
2009-07-15 06:32:15 +02:00
|
|
|
if (flag) {
|
2012-04-30 22:20:23 +02:00
|
|
|
/* Save off if rec is being enabled (for return value) */
|
|
|
|
flag ^= rec->flags & FTRACE_FL_ENABLED;
|
|
|
|
|
|
|
|
if (update) {
|
2011-08-16 15:53:39 +02:00
|
|
|
rec->flags |= FTRACE_FL_ENABLED;
|
2012-04-30 22:20:23 +02:00
|
|
|
if (flag & FTRACE_FL_REGS) {
|
|
|
|
if (rec->flags & FTRACE_FL_REGS)
|
|
|
|
rec->flags |= FTRACE_FL_REGS_EN;
|
|
|
|
else
|
|
|
|
rec->flags &= ~FTRACE_FL_REGS_EN;
|
|
|
|
}
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
if (flag & FTRACE_FL_TRAMP) {
|
|
|
|
if (rec->flags & FTRACE_FL_TRAMP)
|
|
|
|
rec->flags |= FTRACE_FL_TRAMP_EN;
|
|
|
|
else
|
|
|
|
rec->flags &= ~FTRACE_FL_TRAMP_EN;
|
|
|
|
}
|
2012-04-30 22:20:23 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this record is being updated from a nop, then
|
|
|
|
* return UPDATE_MAKE_CALL.
|
|
|
|
* Otherwise,
|
|
|
|
* return UPDATE_MODIFY_CALL to tell the caller to convert
|
2014-05-07 22:09:49 +02:00
|
|
|
* from the save regs, to a non-save regs function or
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
* vice versa, or from a trampoline call.
|
2012-04-30 22:20:23 +02:00
|
|
|
*/
|
|
|
|
if (flag & FTRACE_FL_ENABLED)
|
|
|
|
return FTRACE_UPDATE_MAKE_CALL;
|
2014-05-07 22:09:49 +02:00
|
|
|
|
|
|
|
return FTRACE_UPDATE_MODIFY_CALL;
|
2011-08-16 15:53:39 +02:00
|
|
|
}
|
|
|
|
|
2012-04-30 22:20:23 +02:00
|
|
|
if (update) {
|
|
|
|
/* If there's no more users, clear all flags */
|
2014-05-07 19:46:45 +02:00
|
|
|
if (!ftrace_rec_count(rec))
|
2012-04-30 22:20:23 +02:00
|
|
|
rec->flags = 0;
|
|
|
|
else
|
ftrace: Clear REGS_EN and TRAMP_EN flags on disabling record via sysctl
When /proc/sys/kernel/ftrace_enabled is set to zero, all function
tracing is disabled. But the records that represent the functions
still hold information about the ftrace_ops that are hooked to them.
ftrace_ops may request "REGS" (have a full set of pt_regs passed to
the callback), or "TRAMP" (the ops has its own trampoline to use).
When the record is updated to represent the state of the ops hooked
to it, it sets "REGS_EN" and/or "TRAMP_EN" to state that the callback
points to the correct trampoline (REGS has its own trampoline).
When ftrace_enabled is set to zero, all ftrace locations are a nop,
so they do not point to any trampoline. But the _EN flags are still
set. This can cause the accounting to go wrong when ftrace_enabled
is cleared and an ops that has a trampoline is registered or unregistered.
For example, the following will cause ftrace to crash:
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
# echo 0 > /proc/sys/kernel/ftrace_enabled
# echo nop > /sys/kernel/debug/tracing/current_tracer
# echo 1 > /proc/sys/kernel/ftrace_enabled
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
As function_graph uses a trampoline, when ftrace_enabled is set to zero
the updates to the record are not done. When enabling function_graph
again, the record will still have the TRAMP_EN flag set, and it will
look for an op that has a trampoline other than the function_graph
ops, and fail to find one.
Cc: stable@vger.kernel.org # 3.17+
Reported-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-05 05:10:28 +01:00
|
|
|
/*
|
|
|
|
* Just disable the record, but keep the ops TRAMP
|
|
|
|
* and REGS states. The _EN flags must be disabled though.
|
|
|
|
*/
|
|
|
|
rec->flags &= ~(FTRACE_FL_ENABLED | FTRACE_FL_TRAMP_EN |
|
|
|
|
FTRACE_FL_REGS_EN);
|
2012-04-30 22:20:23 +02:00
|
|
|
}
|
2011-08-16 15:53:39 +02:00
|
|
|
|
|
|
|
return FTRACE_UPDATE_MAKE_NOP;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_update_record, set a record that now is tracing or not
|
|
|
|
* @rec: the record to update
|
|
|
|
* @enable: set to 1 if the record is tracing, zero to force disable
|
|
|
|
*
|
|
|
|
* The records that represent all functions that can be traced need
|
|
|
|
* to be updated when tracing has been enabled.
|
|
|
|
*/
|
|
|
|
int ftrace_update_record(struct dyn_ftrace *rec, int enable)
|
|
|
|
{
|
|
|
|
return ftrace_check_record(rec, enable, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_test_record, check if the record has been enabled or not
|
|
|
|
* @rec: the record to test
|
|
|
|
* @enable: set to 1 to check if enabled, 0 if it is disabled
|
|
|
|
*
|
|
|
|
* The arch code may need to test if a record is already set to
|
|
|
|
* tracing to determine how to modify the function code that it
|
|
|
|
* represents.
|
|
|
|
*/
|
|
|
|
int ftrace_test_record(struct dyn_ftrace *rec, int enable)
|
|
|
|
{
|
|
|
|
return ftrace_check_record(rec, enable, 0);
|
|
|
|
}
|
|
|
|
|
2014-07-24 22:00:31 +02:00
|
|
|
static struct ftrace_ops *
|
|
|
|
ftrace_find_tramp_ops_any(struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *op;
|
2014-07-24 18:25:47 +02:00
|
|
|
unsigned long ip = rec->ip;
|
2014-07-24 22:00:31 +02:00
|
|
|
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
|
|
|
|
if (!op->trampoline)
|
|
|
|
continue;
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
if (hash_contains_ip(ip, op->func_hash))
|
2014-07-24 22:00:31 +02:00
|
|
|
return op;
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
static struct ftrace_ops *
|
|
|
|
ftrace_find_tramp_ops_curr(struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *op;
|
2014-07-24 18:25:47 +02:00
|
|
|
unsigned long ip = rec->ip;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
/*
|
|
|
|
* Need to check removed ops first.
|
|
|
|
* If they are being removed, and this rec has a tramp,
|
|
|
|
* and this rec is in the ops list, then it would be the
|
|
|
|
* one with the tramp.
|
|
|
|
*/
|
|
|
|
if (removed_ops) {
|
|
|
|
if (hash_contains_ip(ip, &removed_ops->old_hash))
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
return removed_ops;
|
|
|
|
}
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
/*
|
|
|
|
* Need to find the current trampoline for a rec.
|
|
|
|
* Now, a trampoline is only attached to a rec if there
|
|
|
|
* was a single 'ops' attached to it. But this can be called
|
|
|
|
* when we are adding another op to the rec or removing the
|
|
|
|
* current one. Thus, if the op is being added, we can
|
|
|
|
* ignore it because it hasn't attached itself to the rec
|
2014-10-24 20:48:35 +02:00
|
|
|
* yet.
|
|
|
|
*
|
|
|
|
* If an ops is being modified (hooking to different functions)
|
|
|
|
* then we don't care about the new functions that are being
|
|
|
|
* added, just the old ones (that are probably being removed).
|
|
|
|
*
|
|
|
|
* If we are adding an ops to a function that already is using
|
|
|
|
* a trampoline, it needs to be removed (trampolines are only
|
|
|
|
* for single ops connected), then an ops that is not being
|
|
|
|
* modified also needs to be checked.
|
2014-07-24 18:25:47 +02:00
|
|
|
*/
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
2014-07-24 18:25:47 +02:00
|
|
|
|
|
|
|
if (!op->trampoline)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the ops is being added, it hasn't gotten to
|
|
|
|
* the point to be removed from this tree yet.
|
|
|
|
*/
|
|
|
|
if (op->flags & FTRACE_OPS_FL_ADDING)
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
continue;
|
|
|
|
|
2014-10-24 20:48:35 +02:00
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
/*
|
2014-10-24 20:48:35 +02:00
|
|
|
* If the ops is being modified and is in the old
|
|
|
|
* hash, then it is probably being removed from this
|
|
|
|
* function.
|
2014-07-24 18:25:47 +02:00
|
|
|
*/
|
|
|
|
if ((op->flags & FTRACE_OPS_FL_MODIFYING) &&
|
|
|
|
hash_contains_ip(ip, &op->old_hash))
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
return op;
|
2014-10-24 20:48:35 +02:00
|
|
|
/*
|
|
|
|
* If the ops is not being added or modified, and it's
|
|
|
|
* in its normal filter hash, then this must be the one
|
|
|
|
* we want!
|
|
|
|
*/
|
|
|
|
if (!(op->flags & FTRACE_OPS_FL_MODIFYING) &&
|
|
|
|
hash_contains_ip(ip, op->func_hash))
|
|
|
|
return op;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct ftrace_ops *
|
|
|
|
ftrace_find_tramp_ops_new(struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *op;
|
2014-07-24 18:25:47 +02:00
|
|
|
unsigned long ip = rec->ip;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
/* pass rec in as regs to have non-NULL val */
|
2014-07-24 18:25:47 +02:00
|
|
|
if (hash_contains_ip(ip, op->func_hash))
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
return op;
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-05-07 03:34:14 +02:00
|
|
|
/**
|
|
|
|
* ftrace_get_addr_new - Get the call address to set to
|
|
|
|
* @rec: The ftrace record descriptor
|
|
|
|
*
|
|
|
|
* If the record has the FTRACE_FL_REGS set, that means that it
|
|
|
|
* wants to convert to a callback that saves all regs. If FTRACE_FL_REGS
|
|
|
|
* is not not set, then it wants to convert to the normal callback.
|
|
|
|
*
|
|
|
|
* Returns the address of the trampoline to set to
|
|
|
|
*/
|
|
|
|
unsigned long ftrace_get_addr_new(struct dyn_ftrace *rec)
|
|
|
|
{
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
struct ftrace_ops *ops;
|
|
|
|
|
|
|
|
/* Trampolines take precedence over regs */
|
|
|
|
if (rec->flags & FTRACE_FL_TRAMP) {
|
|
|
|
ops = ftrace_find_tramp_ops_new(rec);
|
|
|
|
if (FTRACE_WARN_ON(!ops || !ops->trampoline)) {
|
2014-08-21 05:57:04 +02:00
|
|
|
pr_warn("Bad trampoline accounting at: %p (%pS) (%lx)\n",
|
|
|
|
(void *)rec->ip, (void *)rec->ip, rec->flags);
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
/* Ftrace is shutting down, return anything */
|
|
|
|
return (unsigned long)FTRACE_ADDR;
|
|
|
|
}
|
|
|
|
return ops->trampoline;
|
|
|
|
}
|
|
|
|
|
2014-05-07 03:34:14 +02:00
|
|
|
if (rec->flags & FTRACE_FL_REGS)
|
|
|
|
return (unsigned long)FTRACE_REGS_ADDR;
|
|
|
|
else
|
|
|
|
return (unsigned long)FTRACE_ADDR;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_get_addr_curr - Get the call address that is already there
|
|
|
|
* @rec: The ftrace record descriptor
|
|
|
|
*
|
|
|
|
* The FTRACE_FL_REGS_EN is set when the record already points to
|
|
|
|
* a function that saves all the regs. Basically the '_EN' version
|
|
|
|
* represents the current state of the function.
|
|
|
|
*
|
|
|
|
* Returns the address of the trampoline that is currently being called
|
|
|
|
*/
|
|
|
|
unsigned long ftrace_get_addr_curr(struct dyn_ftrace *rec)
|
|
|
|
{
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
struct ftrace_ops *ops;
|
|
|
|
|
|
|
|
/* Trampolines take precedence over regs */
|
|
|
|
if (rec->flags & FTRACE_FL_TRAMP_EN) {
|
|
|
|
ops = ftrace_find_tramp_ops_curr(rec);
|
|
|
|
if (FTRACE_WARN_ON(!ops)) {
|
|
|
|
pr_warning("Bad trampoline accounting at: %p (%pS)\n",
|
|
|
|
(void *)rec->ip, (void *)rec->ip);
|
|
|
|
/* Ftrace is shutting down, return anything */
|
|
|
|
return (unsigned long)FTRACE_ADDR;
|
|
|
|
}
|
|
|
|
return ops->trampoline;
|
|
|
|
}
|
|
|
|
|
2014-05-07 03:34:14 +02:00
|
|
|
if (rec->flags & FTRACE_FL_REGS_EN)
|
|
|
|
return (unsigned long)FTRACE_REGS_ADDR;
|
|
|
|
else
|
|
|
|
return (unsigned long)FTRACE_ADDR;
|
|
|
|
}
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
static int
|
|
|
|
__ftrace_replace_code(struct dyn_ftrace *rec, int enable)
|
|
|
|
{
|
2012-04-30 22:20:23 +02:00
|
|
|
unsigned long ftrace_old_addr;
|
2011-08-16 15:53:39 +02:00
|
|
|
unsigned long ftrace_addr;
|
|
|
|
int ret;
|
|
|
|
|
2014-05-08 13:01:21 +02:00
|
|
|
ftrace_addr = ftrace_get_addr_new(rec);
|
2011-08-16 15:53:39 +02:00
|
|
|
|
2014-05-08 13:01:21 +02:00
|
|
|
/* This needs to be done before we call ftrace_update_record */
|
|
|
|
ftrace_old_addr = ftrace_get_addr_curr(rec);
|
|
|
|
|
|
|
|
ret = ftrace_update_record(rec, enable);
|
2012-04-30 22:20:23 +02:00
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
switch (ret) {
|
|
|
|
case FTRACE_UPDATE_IGNORE:
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
case FTRACE_UPDATE_MAKE_CALL:
|
2009-07-15 06:32:15 +02:00
|
|
|
return ftrace_make_call(rec, ftrace_addr);
|
2011-08-16 15:53:39 +02:00
|
|
|
|
|
|
|
case FTRACE_UPDATE_MAKE_NOP:
|
2014-08-18 02:59:10 +02:00
|
|
|
return ftrace_make_nop(NULL, rec, ftrace_old_addr);
|
2012-04-30 22:20:23 +02:00
|
|
|
|
|
|
|
case FTRACE_UPDATE_MODIFY_CALL:
|
|
|
|
return ftrace_modify_call(rec, ftrace_old_addr, ftrace_addr);
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
return -1; /* unknow ftrace bug */
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2012-04-27 15:13:18 +02:00
|
|
|
void __weak ftrace_replace_code(int enable)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
|
|
|
struct dyn_ftrace *rec;
|
|
|
|
struct ftrace_page *pg;
|
2009-02-17 17:20:26 +01:00
|
|
|
int failed;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return;
|
|
|
|
|
2009-02-13 18:43:56 +01:00
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
2012-04-27 15:13:18 +02:00
|
|
|
failed = __ftrace_replace_code(rec, enable);
|
2009-03-13 10:16:34 +01:00
|
|
|
if (failed) {
|
2014-10-24 23:56:04 +02:00
|
|
|
ftrace_bug(failed, rec);
|
2009-10-07 22:57:56 +02:00
|
|
|
/* Stop processing */
|
|
|
|
return;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
2009-02-13 18:43:56 +01:00
|
|
|
} while_for_each_ftrace_rec();
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
struct ftrace_rec_iter {
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
int index;
|
|
|
|
};
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_rec_iter_start, start up iterating over traced functions
|
|
|
|
*
|
|
|
|
* Returns an iterator handle that is used to iterate over all
|
|
|
|
* the records that represent address locations where functions
|
|
|
|
* are traced.
|
|
|
|
*
|
|
|
|
* May return NULL if no records are available.
|
|
|
|
*/
|
|
|
|
struct ftrace_rec_iter *ftrace_rec_iter_start(void)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* We only use a single iterator.
|
|
|
|
* Protected by the ftrace_lock mutex.
|
|
|
|
*/
|
|
|
|
static struct ftrace_rec_iter ftrace_rec_iter;
|
|
|
|
struct ftrace_rec_iter *iter = &ftrace_rec_iter;
|
|
|
|
|
|
|
|
iter->pg = ftrace_pages_start;
|
|
|
|
iter->index = 0;
|
|
|
|
|
|
|
|
/* Could have empty pages */
|
|
|
|
while (iter->pg && !iter->pg->index)
|
|
|
|
iter->pg = iter->pg->next;
|
|
|
|
|
|
|
|
if (!iter->pg)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_rec_iter_next, get the next record to process.
|
|
|
|
* @iter: The handle to the iterator.
|
|
|
|
*
|
|
|
|
* Returns the next iterator after the given iterator @iter.
|
|
|
|
*/
|
|
|
|
struct ftrace_rec_iter *ftrace_rec_iter_next(struct ftrace_rec_iter *iter)
|
|
|
|
{
|
|
|
|
iter->index++;
|
|
|
|
|
|
|
|
if (iter->index >= iter->pg->index) {
|
|
|
|
iter->pg = iter->pg->next;
|
|
|
|
iter->index = 0;
|
|
|
|
|
|
|
|
/* Could have empty pages */
|
|
|
|
while (iter->pg && !iter->pg->index)
|
|
|
|
iter->pg = iter->pg->next;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!iter->pg)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_rec_iter_record, get the record at the iterator location
|
|
|
|
* @iter: The current iterator location
|
|
|
|
*
|
|
|
|
* Returns the record that the current @iter is at.
|
|
|
|
*/
|
|
|
|
struct dyn_ftrace *ftrace_rec_iter_record(struct ftrace_rec_iter *iter)
|
|
|
|
{
|
|
|
|
return &iter->pg->records[iter->index];
|
|
|
|
}
|
|
|
|
|
2008-05-24 20:40:04 +02:00
|
|
|
static int
|
2008-11-15 01:21:19 +01:00
|
|
|
ftrace_code_disable(struct module *mod, struct dyn_ftrace *rec)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
2008-10-23 15:32:59 +02:00
|
|
|
int ret;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return 0;
|
|
|
|
|
2009-01-09 04:29:40 +01:00
|
|
|
ret = ftrace_make_nop(mod, rec, MCOUNT_ADDR);
|
2008-10-23 15:32:59 +02:00
|
|
|
if (ret) {
|
2014-10-24 23:56:04 +02:00
|
|
|
ftrace_bug(ret, rec);
|
2008-05-24 20:40:04 +02:00
|
|
|
return 0;
|
2008-05-12 21:20:48 +02:00
|
|
|
}
|
2008-05-24 20:40:04 +02:00
|
|
|
return 1;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2009-02-17 19:35:06 +01:00
|
|
|
/*
|
|
|
|
* archs can override this function if they must do something
|
|
|
|
* before the modifying code is performed.
|
|
|
|
*/
|
|
|
|
int __weak ftrace_arch_code_modify_prepare(void)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* archs can override this function if they must do something
|
|
|
|
* after the modifying code is performed.
|
|
|
|
*/
|
|
|
|
int __weak ftrace_arch_code_modify_post_process(void)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-26 20:59:43 +02:00
|
|
|
void ftrace_modify_all_code(int command)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2013-08-31 07:04:07 +02:00
|
|
|
int update = command & FTRACE_UPDATE_TRACE_FUNC;
|
2014-02-24 17:12:21 +01:00
|
|
|
int err = 0;
|
2013-08-31 07:04:07 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the ftrace_caller calls a ftrace_ops func directly,
|
|
|
|
* we need to make sure that it only traces functions it
|
|
|
|
* expects to trace. When doing the switch of functions,
|
|
|
|
* we need to update to the ftrace_ops_list_func first
|
|
|
|
* before the transition between old and new calls are set,
|
|
|
|
* as the ftrace_ops_list_func will check the ops hashes
|
|
|
|
* to make sure the ops are having the right functions
|
|
|
|
* traced.
|
|
|
|
*/
|
2014-02-24 17:12:21 +01:00
|
|
|
if (update) {
|
|
|
|
err = ftrace_update_ftrace_func(ftrace_ops_list_func);
|
|
|
|
if (FTRACE_WARN_ON(err))
|
|
|
|
return;
|
|
|
|
}
|
2013-08-31 07:04:07 +02:00
|
|
|
|
2012-04-26 20:59:43 +02:00
|
|
|
if (command & FTRACE_UPDATE_CALLS)
|
2008-05-12 21:20:43 +02:00
|
|
|
ftrace_replace_code(1);
|
2012-04-26 20:59:43 +02:00
|
|
|
else if (command & FTRACE_DISABLE_CALLS)
|
2008-05-12 21:20:43 +02:00
|
|
|
ftrace_replace_code(0);
|
|
|
|
|
2013-11-08 20:17:30 +01:00
|
|
|
if (update && ftrace_trace_function != ftrace_ops_list_func) {
|
|
|
|
function_trace_op = set_function_trace_op;
|
|
|
|
smp_wmb();
|
|
|
|
/* If irqs are disabled, we are in stop machine */
|
|
|
|
if (!irqs_disabled())
|
|
|
|
smp_call_function(ftrace_sync_ipi, NULL, 1);
|
2014-02-24 17:12:21 +01:00
|
|
|
err = ftrace_update_ftrace_func(ftrace_trace_function);
|
|
|
|
if (FTRACE_WARN_ON(err))
|
|
|
|
return;
|
2013-11-08 20:17:30 +01:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2012-04-26 20:59:43 +02:00
|
|
|
if (command & FTRACE_START_FUNC_RET)
|
2014-02-24 17:12:21 +01:00
|
|
|
err = ftrace_enable_ftrace_graph_caller();
|
2012-04-26 20:59:43 +02:00
|
|
|
else if (command & FTRACE_STOP_FUNC_RET)
|
2014-02-24 17:12:21 +01:00
|
|
|
err = ftrace_disable_ftrace_graph_caller();
|
|
|
|
FTRACE_WARN_ON(err);
|
2012-04-26 20:59:43 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int __ftrace_modify_code(void *data)
|
|
|
|
{
|
|
|
|
int *command = data;
|
|
|
|
|
|
|
|
ftrace_modify_all_code(*command);
|
2008-11-26 06:16:24 +01:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
return 0;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
/**
|
|
|
|
* ftrace_run_stop_machine, go back to the stop machine method
|
|
|
|
* @command: The command to tell ftrace what to do
|
|
|
|
*
|
|
|
|
* If an arch needs to fall back to the stop machine method, the
|
|
|
|
* it can call this function.
|
|
|
|
*/
|
|
|
|
void ftrace_run_stop_machine(int command)
|
|
|
|
{
|
|
|
|
stop_machine(__ftrace_modify_code, &command, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* arch_ftrace_update_code, modify the code to trace or not trace
|
|
|
|
* @command: The command that needs to be done
|
|
|
|
*
|
|
|
|
* Archs can override this function if it does not need to
|
|
|
|
* run stop_machine() to modify code.
|
|
|
|
*/
|
|
|
|
void __weak arch_ftrace_update_code(int command)
|
|
|
|
{
|
|
|
|
ftrace_run_stop_machine(command);
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static void ftrace_run_update_code(int command)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2009-02-17 19:35:06 +01:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = ftrace_arch_code_modify_prepare();
|
|
|
|
FTRACE_WARN_ON(ret);
|
|
|
|
if (ret)
|
|
|
|
return;
|
|
|
|
|
2011-08-16 15:53:39 +02:00
|
|
|
/*
|
|
|
|
* By default we use stop_machine() to modify the code.
|
|
|
|
* But archs can do what ever they want as long as it
|
|
|
|
* is safe. The stop_machine() is the safest, but also
|
|
|
|
* produces the most overhead.
|
|
|
|
*/
|
|
|
|
arch_ftrace_update_code(command);
|
|
|
|
|
2009-02-17 19:35:06 +01:00
|
|
|
ret = ftrace_arch_code_modify_post_process();
|
|
|
|
FTRACE_WARN_ON(ret);
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2014-10-24 20:56:01 +02:00
|
|
|
static void ftrace_run_modify_code(struct ftrace_ops *ops, int command,
|
2015-01-13 20:03:38 +01:00
|
|
|
struct ftrace_ops_hash *old_hash)
|
2014-08-05 23:19:38 +02:00
|
|
|
{
|
|
|
|
ops->flags |= FTRACE_OPS_FL_MODIFYING;
|
2015-01-13 20:03:38 +01:00
|
|
|
ops->old_hash.filter_hash = old_hash->filter_hash;
|
|
|
|
ops->old_hash.notrace_hash = old_hash->notrace_hash;
|
2014-08-05 23:19:38 +02:00
|
|
|
ftrace_run_update_code(command);
|
2014-10-24 20:56:01 +02:00
|
|
|
ops->old_hash.filter_hash = NULL;
|
2015-01-13 20:03:38 +01:00
|
|
|
ops->old_hash.notrace_hash = NULL;
|
2014-08-05 23:19:38 +02:00
|
|
|
ops->flags &= ~FTRACE_OPS_FL_MODIFYING;
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
static ftrace_func_t saved_ftrace_func;
|
2008-11-05 22:05:44 +01:00
|
|
|
static int ftrace_start_up;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2014-07-03 21:48:16 +02:00
|
|
|
void __weak arch_ftrace_trampoline_free(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
2014-03-10 21:42:11 +01:00
|
|
|
static void control_ops_free(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
free_percpu(ops->disabled);
|
|
|
|
}
|
|
|
|
|
2008-11-26 06:16:23 +01:00
|
|
|
static void ftrace_startup_enable(int command)
|
|
|
|
{
|
|
|
|
if (saved_ftrace_func != ftrace_trace_function) {
|
|
|
|
saved_ftrace_func = ftrace_trace_function;
|
|
|
|
command |= FTRACE_UPDATE_TRACE_FUNC;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!command || !ftrace_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
ftrace_run_update_code(command);
|
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2014-08-05 23:19:38 +02:00
|
|
|
static void ftrace_startup_all(int command)
|
|
|
|
{
|
|
|
|
update_all_ops = true;
|
|
|
|
ftrace_startup_enable(command);
|
|
|
|
update_all_ops = false;
|
|
|
|
}
|
|
|
|
|
2011-05-23 21:24:25 +02:00
|
|
|
static int ftrace_startup(struct ftrace_ops *ops, int command)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2013-11-26 02:59:46 +01:00
|
|
|
int ret;
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
2011-05-23 21:24:25 +02:00
|
|
|
return -ENODEV;
|
2008-05-12 21:20:48 +02:00
|
|
|
|
2013-11-26 02:59:46 +01:00
|
|
|
ret = __register_ftrace_function(ops);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2008-11-05 22:05:44 +01:00
|
|
|
ftrace_start_up++;
|
2011-12-05 18:22:48 +01:00
|
|
|
command |= FTRACE_UPDATE_CALLS;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2014-08-05 23:19:38 +02:00
|
|
|
/*
|
|
|
|
* Note that ftrace probes uses this to start up
|
|
|
|
* and modify functions it will probe. But we still
|
|
|
|
* set the ADDING flag for modification, as probes
|
|
|
|
* do not have trampolines. If they add them in the
|
|
|
|
* future, then the probes will need to distinguish
|
|
|
|
* between adding and updating probes.
|
|
|
|
*/
|
|
|
|
ops->flags |= FTRACE_OPS_FL_ENABLED | FTRACE_OPS_FL_ADDING;
|
2014-05-07 03:57:49 +02:00
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
ret = ftrace_hash_ipmodify_enable(ops);
|
|
|
|
if (ret < 0) {
|
|
|
|
/* Rollback registration process */
|
|
|
|
__unregister_ftrace_function(ops);
|
|
|
|
ftrace_start_up--;
|
|
|
|
ops->flags &= ~FTRACE_OPS_FL_ENABLED;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2014-05-07 03:57:49 +02:00
|
|
|
ftrace_hash_rec_enable(ops, 1);
|
2011-05-03 19:25:24 +02:00
|
|
|
|
2008-11-26 06:16:23 +01:00
|
|
|
ftrace_startup_enable(command);
|
2011-05-23 21:24:25 +02:00
|
|
|
|
2014-08-05 23:19:38 +02:00
|
|
|
ops->flags &= ~FTRACE_OPS_FL_ADDING;
|
|
|
|
|
2011-05-23 21:24:25 +02:00
|
|
|
return 0;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2013-11-26 02:59:46 +01:00
|
|
|
static int ftrace_shutdown(struct ftrace_ops *ops, int command)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2013-11-26 02:59:46 +01:00
|
|
|
int ret;
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
2013-11-26 02:59:46 +01:00
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
ret = __unregister_ftrace_function(ops);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
2008-05-12 21:20:48 +02:00
|
|
|
|
2008-11-05 22:05:44 +01:00
|
|
|
ftrace_start_up--;
|
2009-06-20 06:52:21 +02:00
|
|
|
/*
|
|
|
|
* Just warn in case of unbalance, no need to kill ftrace, it's not
|
|
|
|
* critical but the ftrace_call callers may be never nopped again after
|
|
|
|
* further ftrace uses.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(ftrace_start_up < 0);
|
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
/* Disabling ipmodify never fails */
|
|
|
|
ftrace_hash_ipmodify_disable(ops);
|
2014-05-07 03:57:49 +02:00
|
|
|
ftrace_hash_rec_disable(ops, 1);
|
2011-05-03 19:25:24 +02:00
|
|
|
|
2014-06-12 16:56:12 +02:00
|
|
|
ops->flags &= ~FTRACE_OPS_FL_ENABLED;
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2011-12-05 18:22:48 +01:00
|
|
|
command |= FTRACE_UPDATE_CALLS;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
if (saved_ftrace_func != ftrace_trace_function) {
|
|
|
|
saved_ftrace_func = ftrace_trace_function;
|
|
|
|
command |= FTRACE_UPDATE_TRACE_FUNC;
|
|
|
|
}
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2014-01-13 18:56:21 +01:00
|
|
|
if (!command || !ftrace_enabled) {
|
|
|
|
/*
|
|
|
|
* If these are control ops, they still need their
|
|
|
|
* per_cpu field freed. Since, function tracing is
|
|
|
|
* not currently active, we can just free them
|
|
|
|
* without synchronizing all CPUs.
|
|
|
|
*/
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_CONTROL)
|
|
|
|
control_ops_free(ops);
|
2013-11-26 02:59:46 +01:00
|
|
|
return 0;
|
2014-01-13 18:56:21 +01:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
/*
|
|
|
|
* If the ops uses a trampoline, then it needs to be
|
|
|
|
* tested first on update.
|
|
|
|
*/
|
2014-08-05 23:19:38 +02:00
|
|
|
ops->flags |= FTRACE_OPS_FL_REMOVING;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
removed_ops = ops;
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
/* The trampoline logic checks the old hashes */
|
|
|
|
ops->old_hash.filter_hash = ops->func_hash->filter_hash;
|
|
|
|
ops->old_hash.notrace_hash = ops->func_hash->notrace_hash;
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
ftrace_run_update_code(command);
|
2014-01-13 18:56:21 +01:00
|
|
|
|
2014-09-12 20:21:13 +02:00
|
|
|
/*
|
|
|
|
* If there's no more ops registered with ftrace, run a
|
|
|
|
* sanity check to make sure all rec flags are cleared.
|
|
|
|
*/
|
|
|
|
if (ftrace_ops_list == &ftrace_list_end) {
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec;
|
|
|
|
|
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
|
|
|
if (FTRACE_WARN_ON_ONCE(rec->flags))
|
|
|
|
pr_warn(" %pS flags:%lx\n",
|
|
|
|
(void *)rec->ip, rec->flags);
|
|
|
|
} while_for_each_ftrace_rec();
|
|
|
|
}
|
|
|
|
|
2014-07-24 18:25:47 +02:00
|
|
|
ops->old_hash.filter_hash = NULL;
|
|
|
|
ops->old_hash.notrace_hash = NULL;
|
|
|
|
|
|
|
|
removed_ops = NULL;
|
2014-08-05 23:19:38 +02:00
|
|
|
ops->flags &= ~FTRACE_OPS_FL_REMOVING;
|
ftrace: Optimize function graph to be called directly
Function graph tracing is a bit different than the function tracers, as
it is processed after either the ftrace_caller or ftrace_regs_caller
and we only have one place to modify the jump to ftrace_graph_caller,
the jump needs to happen after the restore of registeres.
The function graph tracer is dependent on the function tracer, where
even if the function graph tracing is going on by itself, the save and
restore of registers is still done for function tracing regardless of
if function tracing is happening, before it calls the function graph
code.
If there's no function tracing happening, it is possible to just call
the function graph tracer directly, and avoid the wasted effort to save
and restore regs for function tracing.
This requires adding new flags to the dyn_ftrace records:
FTRACE_FL_TRAMP
FTRACE_FL_TRAMP_EN
The first is set if the count for the record is one, and the ftrace_ops
associated to that record has its own trampoline. That way the mcount code
can call that trampoline directly.
In the future, trampolines can be added to arbitrary ftrace_ops, where you
can have two or more ftrace_ops registered to ftrace (like kprobes and perf)
and if they are not tracing the same functions, then instead of doing a
loop to check all registered ftrace_ops against their hashes, just call the
ftrace_ops trampoline directly, which would call the registered ftrace_ops
function directly.
Without this patch perf showed:
0.05% hackbench [kernel.kallsyms] [k] ftrace_caller
0.05% hackbench [kernel.kallsyms] [k] arch_local_irq_save
0.05% hackbench [kernel.kallsyms] [k] native_sched_clock
0.04% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] preempt_trace
0.04% hackbench [kernel.kallsyms] [k] prepare_ftrace_return
0.04% hackbench [kernel.kallsyms] [k] __this_cpu_preempt_check
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
See that the ftrace_caller took up more time than the ftrace_graph_caller
did.
With this patch:
0.05% hackbench [kernel.kallsyms] [k] __buffer_unlock_commit
0.04% hackbench [kernel.kallsyms] [k] call_filter_check_discard
0.04% hackbench [kernel.kallsyms] [k] ftrace_graph_caller
0.04% hackbench [kernel.kallsyms] [k] sched_clock
The ftrace_caller is no where to be found and ftrace_graph_caller still
takes up the same percentage.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-05-07 03:56:17 +02:00
|
|
|
|
2014-01-13 18:56:21 +01:00
|
|
|
/*
|
|
|
|
* Dynamic ops may be freed, we must make sure that all
|
|
|
|
* callers are done before leaving this function.
|
|
|
|
* The same goes for freeing the per_cpu data of the control
|
|
|
|
* ops.
|
|
|
|
*
|
|
|
|
* Again, normal synchronize_sched() is not good enough.
|
|
|
|
* We need to do a hard force of sched synchronization.
|
|
|
|
* This is because we use preempt_disable() to do RCU, but
|
|
|
|
* the function tracers can be called where RCU is not watching
|
|
|
|
* (like before user_exit()). We can not rely on the RCU
|
|
|
|
* infrastructure to do the synchronization, thus we must do it
|
|
|
|
* ourselves.
|
|
|
|
*/
|
|
|
|
if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_CONTROL)) {
|
|
|
|
schedule_on_each_cpu(ftrace_sync);
|
|
|
|
|
2014-07-03 21:48:16 +02:00
|
|
|
arch_ftrace_trampoline_free(ops);
|
|
|
|
|
2014-01-13 18:56:21 +01:00
|
|
|
if (ops->flags & FTRACE_OPS_FL_CONTROL)
|
|
|
|
control_ops_free(ops);
|
|
|
|
}
|
|
|
|
|
2013-11-26 02:59:46 +01:00
|
|
|
return 0;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static void ftrace_startup_sysctl(void)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
int command;
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return;
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
/* Force update next time */
|
|
|
|
saved_ftrace_func = NULL;
|
2008-11-05 22:05:44 +01:00
|
|
|
/* ftrace_start_up is true if we want ftrace running */
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
if (ftrace_start_up) {
|
|
|
|
command = FTRACE_UPDATE_CALLS;
|
|
|
|
if (ftrace_graph_active)
|
|
|
|
command |= FTRACE_START_FUNC_RET;
|
2015-03-07 01:55:13 +01:00
|
|
|
ftrace_startup_enable(command);
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static void ftrace_shutdown_sysctl(void)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
int command;
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return;
|
|
|
|
|
2008-11-05 22:05:44 +01:00
|
|
|
/* ftrace_start_up is true if ftrace is running */
|
ftrace: Fix en(dis)able graph caller when en(dis)abling record via sysctl
When ftrace is enabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_START_FUNC_RET command to ftrace_run_update_code(). Similarly, when
ftrace is disabled globally through the proc interface, we must check if
ftrace_graph_active is set. If it is set, then we should also pass the
FTRACE_STOP_FUNC_RET command to ftrace_run_update_code().
Consider the following situation.
# echo 0 > /proc/sys/kernel/ftrace_enabled
After this ftrace_enabled = 0.
# echo function_graph > /sys/kernel/debug/tracing/current_tracer
Since ftrace_enabled = 0, ftrace_enable_ftrace_graph_caller() is never
called.
# echo 1 > /proc/sys/kernel/ftrace_enabled
Now ftrace_enabled will be set to true, but still
ftrace_enable_ftrace_graph_caller() will not be called, which is not
desired.
Further if we execute the following after this:
# echo nop > /sys/kernel/debug/tracing/current_tracer
Now since ftrace_enabled is set it will call
ftrace_disable_ftrace_graph_caller(), which causes a kernel warning on
the ARM platform.
On the ARM platform, when ftrace_enable_ftrace_graph_caller() is called,
it checks whether the old instruction is a nop or not. If it's not a nop,
then it returns an error. If it is a nop then it replaces instruction at
that address with a branch to ftrace_graph_caller.
ftrace_disable_ftrace_graph_caller() behaves just the opposite. Therefore,
if generic ftrace code ever calls either ftrace_enable_ftrace_graph_caller()
or ftrace_disable_ftrace_graph_caller() consecutively two times in a row,
then it will return an error, which will cause the generic ftrace code to
raise a warning.
Note, x86 does not have an issue with this because the architecture
specific code for ftrace_enable_ftrace_graph_caller() and
ftrace_disable_ftrace_graph_caller() does not check the previous state,
and calling either of these functions twice in a row has no ill effect.
Link: http://lkml.kernel.org/r/e4fbe64cdac0dd0e86a3bf914b0f83c0b419f146.1425666454.git.panand@redhat.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Pratyush Anand <panand@redhat.com>
[
removed extra if (ftrace_start_up) and defined ftrace_graph_active as 0
if CONFIG_FUNCTION_GRAPH_TRACER is not set.
]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2015-03-06 19:28:06 +01:00
|
|
|
if (ftrace_start_up) {
|
|
|
|
command = FTRACE_DISABLE_CALLS;
|
|
|
|
if (ftrace_graph_active)
|
|
|
|
command |= FTRACE_STOP_FUNC_RET;
|
|
|
|
ftrace_run_update_code(command);
|
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
static cycle_t ftrace_update_time;
|
|
|
|
unsigned long ftrace_update_tot_cnt;
|
|
|
|
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
static inline int ops_traces_mod(struct ftrace_ops *ops)
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
{
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
/*
|
|
|
|
* Filter_hash being empty will default to trace module.
|
|
|
|
* But notrace hash requires a test of individual module functions.
|
|
|
|
*/
|
2014-08-15 23:23:02 +02:00
|
|
|
return ftrace_hash_empty(ops->func_hash->filter_hash) &&
|
|
|
|
ftrace_hash_empty(ops->func_hash->notrace_hash);
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the current ops references the record.
|
|
|
|
*
|
|
|
|
* If the ops traces all functions, then it was already accounted for.
|
|
|
|
* If the ops does not trace the current record function, skip it.
|
|
|
|
* If the ops ignores the function via notrace filter, skip it.
|
|
|
|
*/
|
|
|
|
static inline bool
|
|
|
|
ops_references_rec(struct ftrace_ops *ops, struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
/* If ops isn't enabled, ignore it */
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_ENABLED))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* If ops traces all mods, we already accounted for it */
|
|
|
|
if (ops_traces_mod(ops))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* The function must be in the filter */
|
2014-08-15 23:23:02 +02:00
|
|
|
if (!ftrace_hash_empty(ops->func_hash->filter_hash) &&
|
|
|
|
!ftrace_lookup_ip(ops->func_hash->filter_hash, rec->ip))
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
return 0;
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
/* If in notrace hash, we ignore it too */
|
2014-08-15 23:23:02 +02:00
|
|
|
if (ftrace_lookup_ip(ops->func_hash->notrace_hash, rec->ip))
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int referenced_filters(struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *ops;
|
|
|
|
int cnt = 0;
|
|
|
|
|
|
|
|
for (ops = ftrace_ops_list; ops != &ftrace_list_end; ops = ops->next) {
|
|
|
|
if (ops_references_rec(ops, rec))
|
|
|
|
cnt++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return cnt;
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
}
|
|
|
|
|
2014-02-24 19:59:56 +01:00
|
|
|
static int ftrace_update_code(struct module *mod, struct ftrace_page *new_pgs)
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2011-12-16 22:30:31 +01:00
|
|
|
struct ftrace_page *pg;
|
2009-03-13 10:51:27 +01:00
|
|
|
struct dyn_ftrace *p;
|
2008-06-21 20:20:29 +02:00
|
|
|
cycle_t start, stop;
|
2014-02-24 19:59:56 +01:00
|
|
|
unsigned long update_cnt = 0;
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
unsigned long ref = 0;
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
bool test = false;
|
2011-12-16 22:30:31 +01:00
|
|
|
int i;
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* When adding a module, we need to check if tracers are
|
|
|
|
* currently enabled and if they are set to trace all functions.
|
|
|
|
* If they are, we need to enable the module functions as well
|
|
|
|
* as update the reference counts for those function records.
|
|
|
|
*/
|
|
|
|
if (mod) {
|
|
|
|
struct ftrace_ops *ops;
|
|
|
|
|
|
|
|
for (ops = ftrace_ops_list;
|
|
|
|
ops != &ftrace_list_end; ops = ops->next) {
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
if (ops->flags & FTRACE_OPS_FL_ENABLED) {
|
|
|
|
if (ops_traces_mod(ops))
|
|
|
|
ref++;
|
|
|
|
else
|
|
|
|
test = true;
|
|
|
|
}
|
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-07-15 05:02:27 +02:00
|
|
|
}
|
|
|
|
}
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2008-05-12 21:20:46 +02:00
|
|
|
start = ftrace_now(raw_smp_processor_id());
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2014-02-24 19:59:56 +01:00
|
|
|
for (pg = new_pgs; pg; pg = pg->next) {
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2011-12-16 22:30:31 +01:00
|
|
|
for (i = 0; i < pg->index; i++) {
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
int cnt = ref;
|
|
|
|
|
2011-12-16 22:30:31 +01:00
|
|
|
/* If something went wrong, bail without enabling anything */
|
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return -1;
|
2008-06-21 20:20:29 +02:00
|
|
|
|
2011-12-16 22:30:31 +01:00
|
|
|
p = &pg->records[i];
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
if (test)
|
|
|
|
cnt += referenced_filters(p);
|
|
|
|
p->flags = cnt;
|
2008-06-21 20:20:29 +02:00
|
|
|
|
2011-12-16 22:30:31 +01:00
|
|
|
/*
|
|
|
|
* Do the initial record conversion from mcount jump
|
|
|
|
* to the NOP instructions.
|
|
|
|
*/
|
|
|
|
if (!ftrace_code_disable(mod, p))
|
|
|
|
break;
|
2009-10-13 22:33:53 +02:00
|
|
|
|
2014-02-24 19:59:56 +01:00
|
|
|
update_cnt++;
|
2009-10-13 22:33:53 +02:00
|
|
|
|
2011-12-16 22:30:31 +01:00
|
|
|
/*
|
|
|
|
* If the tracing is enabled, go ahead and enable the record.
|
|
|
|
*
|
|
|
|
* The reason not to enable the record immediatelly is the
|
|
|
|
* inherent check of ftrace_make_nop/ftrace_make_call for
|
|
|
|
* correct previous instructions. Making first the NOP
|
|
|
|
* conversion puts the module to the correct state, thus
|
|
|
|
* passing the ftrace_make_call check.
|
|
|
|
*/
|
ftrace: Check module functions being traced on reload
There's been a nasty bug that would show up and not give much info.
The bug displayed the following warning:
WARNING: at kernel/trace/ftrace.c:1529 __ftrace_hash_rec_update+0x1e3/0x230()
Pid: 20903, comm: bash Tainted: G O 3.6.11+ #38405.trunk
Call Trace:
[<ffffffff8103e5ff>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8103e65a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810c2ee3>] __ftrace_hash_rec_update+0x1e3/0x230
[<ffffffff810c4f28>] ftrace_hash_move+0x28/0x1d0
[<ffffffff811401cc>] ? kfree+0x2c/0x110
[<ffffffff810c68ee>] ftrace_regex_release+0x8e/0x150
[<ffffffff81149f1e>] __fput+0xae/0x220
[<ffffffff8114a09e>] ____fput+0xe/0x10
[<ffffffff8105fa22>] task_work_run+0x72/0x90
[<ffffffff810028ec>] do_notify_resume+0x6c/0xc0
[<ffffffff8126596e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
[<ffffffff815c0f88>] int_signal+0x12/0x17
---[ end trace 793179526ee09b2c ]---
It was finally narrowed down to unloading a module that was being traced.
It was actually more than that. When functions are being traced, there's
a table of all functions that have a ref count of the number of active
tracers attached to that function. When a function trace callback is
registered to a function, the function's record ref count is incremented.
When it is unregistered, the function's record ref count is decremented.
If an inconsistency is detected (ref count goes below zero) the above
warning is shown and the function tracing is permanently disabled until
reboot.
The ftrace callback ops holds a hash of functions that it filters on
(and/or filters off). If the hash is empty, the default means to filter
all functions (for the filter_hash) or to disable no functions (for the
notrace_hash).
When a module is unloaded, it frees the function records that represent
the module functions. These records exist on their own pages, that is
function records for one module will not exist on the same page as
function records for other modules or even the core kernel.
Now when a module unloads, the records that represents its functions are
freed. When the module is loaded again, the records are recreated with
a default ref count of zero (unless there's a callback that traces all
functions, then they will also be traced, and the ref count will be
incremented).
The problem is that if an ftrace callback hash includes functions of the
module being unloaded, those hash entries will not be removed. If the
module is reloaded in the same location, the hash entries still point
to the functions of the module but the module's ref counts do not reflect
that.
With the help of Steve and Joern, we found a reproducer:
Using uinput module and uinput_release function.
cd /sys/kernel/debug/tracing
modprobe uinput
echo uinput_release > set_ftrace_filter
echo function > current_tracer
rmmod uinput
modprobe uinput
# check /proc/modules to see if loaded in same addr, otherwise try again
echo nop > current_tracer
[BOOM]
The above loads the uinput module, which creates a table of functions that
can be traced within the module.
We add uinput_release to the filter_hash to trace just that function.
Enable function tracincg, which increments the ref count of the record
associated to uinput_release.
Remove uinput, which frees the records including the one that represents
uinput_release.
Load the uinput module again (and make sure it's at the same address).
This recreates the function records all with a ref count of zero,
including uinput_release.
Disable function tracing, which will decrement the ref count for uinput_release
which is now zero because of the module removal and reload, and we have
a mismatch (below zero ref count).
The solution is to check all currently tracing ftrace callbacks to see if any
are tracing any of the module's functions when a module is loaded (it already does
that with callbacks that trace all functions). If a callback happens to have
a module function being traced, it increments that records ref count and starts
tracing that function.
There may be a strange side effect with this, where tracing module functions
on unload and then reloading a new module may have that new module's functions
being traced. This may be something that confuses the user, but it's not
a big deal. Another approach is to disable all callback hashes on module unload,
but this leaves some ftrace callbacks that may not be registered, but can
still have hashes tracing the module's function where ftrace doesn't know about
it. That situation can cause the same bug. This solution solves that case too.
Another benefit of this solution, is it is possible to trace a module's
function on unload and load.
Link: http://lkml.kernel.org/r/20130705142629.GA325@redhat.com
Reported-by: Jörn Engel <joern@logfs.org>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Steve Hodgson <steve@purestorage.com>
Tested-by: Steve Hodgson <steve@purestorage.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-07-30 06:04:32 +02:00
|
|
|
if (ftrace_start_up && cnt) {
|
2011-12-16 22:30:31 +01:00
|
|
|
int failed = __ftrace_replace_code(p, 1);
|
|
|
|
if (failed)
|
2014-10-24 23:56:04 +02:00
|
|
|
ftrace_bug(failed, p);
|
2011-12-16 22:30:31 +01:00
|
|
|
}
|
2009-10-13 22:33:53 +02:00
|
|
|
}
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:46 +02:00
|
|
|
stop = ftrace_now(raw_smp_processor_id());
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
ftrace_update_time = stop - start;
|
2014-02-24 19:59:56 +01:00
|
|
|
ftrace_update_tot_cnt += update_cnt;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
2008-05-12 21:20:42 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
static int ftrace_allocate_records(struct ftrace_page *pg, int count)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
2011-12-16 22:23:44 +01:00
|
|
|
int order;
|
2008-05-12 21:20:43 +02:00
|
|
|
int cnt;
|
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
if (WARN_ON(!count))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
order = get_count_order(DIV_ROUND_UP(count, ENTRIES_PER_PAGE));
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
/*
|
2011-12-16 22:23:44 +01:00
|
|
|
* We want to fill as much as possible. No more than a page
|
|
|
|
* may be empty.
|
2008-05-12 21:20:43 +02:00
|
|
|
*/
|
2011-12-16 22:23:44 +01:00
|
|
|
while ((PAGE_SIZE << order) / ENTRY_SIZE >= count + ENTRIES_PER_PAGE)
|
|
|
|
order--;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
again:
|
|
|
|
pg->records = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, order);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
if (!pg->records) {
|
|
|
|
/* if we can't allocate this size, try something smaller */
|
|
|
|
if (!order)
|
|
|
|
return -ENOMEM;
|
|
|
|
order >>= 1;
|
|
|
|
goto again;
|
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
cnt = (PAGE_SIZE << order) / ENTRY_SIZE;
|
|
|
|
pg->size = cnt;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
if (cnt > count)
|
|
|
|
cnt = count;
|
|
|
|
|
|
|
|
return cnt;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct ftrace_page *
|
|
|
|
ftrace_allocate_pages(unsigned long num_to_init)
|
|
|
|
{
|
|
|
|
struct ftrace_page *start_pg;
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
int order;
|
|
|
|
int cnt;
|
|
|
|
|
|
|
|
if (!num_to_init)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
start_pg = pg = kzalloc(sizeof(*pg), GFP_KERNEL);
|
|
|
|
if (!pg)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to allocate as much as possible in one continues
|
|
|
|
* location that fills in all of the space. We want to
|
|
|
|
* waste as little space as possible.
|
|
|
|
*/
|
|
|
|
for (;;) {
|
|
|
|
cnt = ftrace_allocate_records(pg, num_to_init);
|
|
|
|
if (cnt < 0)
|
|
|
|
goto free_pages;
|
|
|
|
|
|
|
|
num_to_init -= cnt;
|
|
|
|
if (!num_to_init)
|
2008-05-12 21:20:43 +02:00
|
|
|
break;
|
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
pg->next = kzalloc(sizeof(*pg), GFP_KERNEL);
|
|
|
|
if (!pg->next)
|
|
|
|
goto free_pages;
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
pg = pg->next;
|
|
|
|
}
|
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
return start_pg;
|
|
|
|
|
|
|
|
free_pages:
|
2014-06-11 10:06:53 +02:00
|
|
|
pg = start_pg;
|
|
|
|
while (pg) {
|
2011-12-16 22:23:44 +01:00
|
|
|
order = get_count_order(pg->size / ENTRIES_PER_PAGE);
|
|
|
|
free_pages((unsigned long)pg->records, order);
|
|
|
|
start_pg = pg->next;
|
|
|
|
kfree(pg);
|
|
|
|
pg = start_pg;
|
|
|
|
}
|
|
|
|
pr_info("ftrace: FAILED to allocate memory for functions\n");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
#define FTRACE_BUFF_MAX (KSYM_SYMBOL_LEN+4) /* room for wildcards */
|
|
|
|
|
|
|
|
struct ftrace_iterator {
|
2010-09-10 17:47:43 +02:00
|
|
|
loff_t pos;
|
2010-09-09 16:00:28 +02:00
|
|
|
loff_t func_pos;
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *func;
|
|
|
|
struct ftrace_func_probe *probe;
|
|
|
|
struct trace_parser parser;
|
2011-04-30 02:59:51 +02:00
|
|
|
struct ftrace_hash *hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
struct ftrace_ops *ops;
|
2010-09-09 16:00:28 +02:00
|
|
|
int hidx;
|
|
|
|
int idx;
|
|
|
|
unsigned flags;
|
2008-05-12 21:20:43 +02:00
|
|
|
};
|
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
static void *
|
2010-09-09 16:00:28 +02:00
|
|
|
t_hash_next(struct seq_file *m, loff_t *pos)
|
2009-02-16 21:28:00 +01:00
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter = m->private;
|
2010-09-09 16:00:28 +02:00
|
|
|
struct hlist_node *hnd = NULL;
|
2009-02-16 21:28:00 +01:00
|
|
|
struct hlist_head *hhd;
|
|
|
|
|
|
|
|
(*pos)++;
|
2010-09-10 17:47:43 +02:00
|
|
|
iter->pos = *pos;
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
if (iter->probe)
|
|
|
|
hnd = &iter->probe->node;
|
2009-02-16 21:28:00 +01:00
|
|
|
retry:
|
|
|
|
if (iter->hidx >= FTRACE_FUNC_HASHSIZE)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
hhd = &ftrace_func_hash[iter->hidx];
|
|
|
|
|
|
|
|
if (hlist_empty(hhd)) {
|
|
|
|
iter->hidx++;
|
|
|
|
hnd = NULL;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!hnd)
|
|
|
|
hnd = hhd->first;
|
|
|
|
else {
|
|
|
|
hnd = hnd->next;
|
|
|
|
if (!hnd) {
|
|
|
|
iter->hidx++;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
if (WARN_ON_ONCE(!hnd))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
iter->probe = hlist_entry(hnd, struct ftrace_func_probe, node);
|
|
|
|
|
|
|
|
return iter;
|
2009-02-16 21:28:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void *t_hash_start(struct seq_file *m, loff_t *pos)
|
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter = m->private;
|
|
|
|
void *p = NULL;
|
2009-06-24 03:54:54 +02:00
|
|
|
loff_t l;
|
|
|
|
|
2011-12-19 21:21:16 +01:00
|
|
|
if (!(iter->flags & FTRACE_ITER_DO_HASH))
|
|
|
|
return NULL;
|
|
|
|
|
2010-09-09 14:43:22 +02:00
|
|
|
if (iter->func_pos > *pos)
|
|
|
|
return NULL;
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2009-06-24 03:54:54 +02:00
|
|
|
iter->hidx = 0;
|
2010-09-09 14:43:22 +02:00
|
|
|
for (l = 0; l <= (*pos - iter->func_pos); ) {
|
2010-09-09 16:00:28 +02:00
|
|
|
p = t_hash_next(m, &l);
|
2009-06-24 03:54:54 +02:00
|
|
|
if (!p)
|
|
|
|
break;
|
|
|
|
}
|
2010-09-09 16:00:28 +02:00
|
|
|
if (!p)
|
|
|
|
return NULL;
|
|
|
|
|
2010-09-10 17:47:43 +02:00
|
|
|
/* Only set this if we have an item */
|
|
|
|
iter->flags |= FTRACE_ITER_HASH;
|
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
return iter;
|
2009-02-16 21:28:00 +01:00
|
|
|
}
|
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
static int
|
|
|
|
t_hash_show(struct seq_file *m, struct ftrace_iterator *iter)
|
2009-02-16 21:28:00 +01:00
|
|
|
{
|
2009-02-17 18:32:04 +01:00
|
|
|
struct ftrace_func_probe *rec;
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
rec = iter->probe;
|
|
|
|
if (WARN_ON_ONCE(!rec))
|
|
|
|
return -EIO;
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2009-02-17 05:06:01 +01:00
|
|
|
if (rec->ops->print)
|
|
|
|
return rec->ops->print(m, rec->ip, rec->ops, rec->data);
|
|
|
|
|
2009-09-17 06:05:58 +02:00
|
|
|
seq_printf(m, "%ps:%ps", (void *)rec->ip, (void *)rec->ops->func);
|
2009-02-16 21:28:00 +01:00
|
|
|
|
|
|
|
if (rec->data)
|
|
|
|
seq_printf(m, ":%p", rec->data);
|
|
|
|
seq_putc(m, '\n');
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static void *
|
2008-05-12 21:20:43 +02:00
|
|
|
t_next(struct seq_file *m, void *v, loff_t *pos)
|
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter = m->private;
|
2011-12-19 20:41:25 +01:00
|
|
|
struct ftrace_ops *ops = iter->ops;
|
2008-05-12 21:20:43 +02:00
|
|
|
struct dyn_ftrace *rec = NULL;
|
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return NULL;
|
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
if (iter->flags & FTRACE_ITER_HASH)
|
2010-09-09 16:00:28 +02:00
|
|
|
return t_hash_next(m, pos);
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
(*pos)++;
|
2011-02-16 17:35:34 +01:00
|
|
|
iter->pos = iter->func_pos = *pos;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-02-16 17:21:52 +01:00
|
|
|
if (iter->flags & FTRACE_ITER_PRINTALL)
|
2010-09-14 17:21:11 +02:00
|
|
|
return t_hash_start(m, pos);
|
2009-02-16 17:21:52 +01:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
retry:
|
|
|
|
if (iter->idx >= iter->pg->index) {
|
|
|
|
if (iter->pg->next) {
|
|
|
|
iter->pg = iter->pg->next;
|
|
|
|
iter->idx = 0;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
rec = &iter->pg->records[iter->idx++];
|
2011-12-16 20:42:37 +01:00
|
|
|
if (((iter->flags & FTRACE_ITER_FILTER) &&
|
2014-08-15 23:23:02 +02:00
|
|
|
!(ftrace_lookup_ip(ops->func_hash->filter_hash, rec->ip))) ||
|
2008-11-08 04:36:02 +01:00
|
|
|
|
2008-05-22 17:46:33 +02:00
|
|
|
((iter->flags & FTRACE_ITER_NOTRACE) &&
|
2014-08-15 23:23:02 +02:00
|
|
|
!ftrace_lookup_ip(ops->func_hash->notrace_hash, rec->ip)) ||
|
2011-05-03 20:39:21 +02:00
|
|
|
|
|
|
|
((iter->flags & FTRACE_ITER_ENABLED) &&
|
2013-05-10 01:31:48 +02:00
|
|
|
!(rec->flags & FTRACE_FL_ENABLED))) {
|
2011-05-03 20:39:21 +02:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
rec = NULL;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
if (!rec)
|
2010-09-14 17:21:11 +02:00
|
|
|
return t_hash_start(m, pos);
|
2010-09-09 16:00:28 +02:00
|
|
|
|
|
|
|
iter->func = rec;
|
|
|
|
|
|
|
|
return iter;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2010-09-10 17:47:43 +02:00
|
|
|
static void reset_iter_read(struct ftrace_iterator *iter)
|
|
|
|
{
|
|
|
|
iter->pos = 0;
|
|
|
|
iter->func_pos = 0;
|
2012-06-09 18:10:27 +02:00
|
|
|
iter->flags &= ~(FTRACE_ITER_PRINTALL | FTRACE_ITER_HASH);
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void *t_start(struct seq_file *m, loff_t *pos)
|
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter = m->private;
|
2011-12-19 20:41:25 +01:00
|
|
|
struct ftrace_ops *ops = iter->ops;
|
2008-05-12 21:20:43 +02:00
|
|
|
void *p = NULL;
|
2009-06-24 03:54:19 +02:00
|
|
|
loff_t l;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2011-04-22 05:16:46 +02:00
|
|
|
|
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return NULL;
|
|
|
|
|
2010-09-10 17:47:43 +02:00
|
|
|
/*
|
|
|
|
* If an lseek was done, then reset and start from beginning.
|
|
|
|
*/
|
|
|
|
if (*pos < iter->pos)
|
|
|
|
reset_iter_read(iter);
|
|
|
|
|
2009-02-16 17:21:52 +01:00
|
|
|
/*
|
|
|
|
* For set_ftrace_filter reading, if we have the filter
|
|
|
|
* off, we can short cut and just print out that all
|
|
|
|
* functions are enabled.
|
|
|
|
*/
|
2014-06-13 09:24:06 +02:00
|
|
|
if ((iter->flags & FTRACE_ITER_FILTER &&
|
2014-08-15 23:23:02 +02:00
|
|
|
ftrace_hash_empty(ops->func_hash->filter_hash)) ||
|
2014-06-13 09:24:06 +02:00
|
|
|
(iter->flags & FTRACE_ITER_NOTRACE &&
|
2014-08-15 23:23:02 +02:00
|
|
|
ftrace_hash_empty(ops->func_hash->notrace_hash))) {
|
2009-02-16 17:21:52 +01:00
|
|
|
if (*pos > 0)
|
2009-02-16 21:28:00 +01:00
|
|
|
return t_hash_start(m, pos);
|
2009-02-16 17:21:52 +01:00
|
|
|
iter->flags |= FTRACE_ITER_PRINTALL;
|
2010-09-10 01:34:59 +02:00
|
|
|
/* reset in case of seek/pread */
|
|
|
|
iter->flags &= ~FTRACE_ITER_HASH;
|
2009-02-16 17:21:52 +01:00
|
|
|
return iter;
|
|
|
|
}
|
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
if (iter->flags & FTRACE_ITER_HASH)
|
|
|
|
return t_hash_start(m, pos);
|
|
|
|
|
2010-09-10 17:47:43 +02:00
|
|
|
/*
|
|
|
|
* Unfortunately, we need to restart at ftrace_pages_start
|
|
|
|
* every time we let go of the ftrace_mutex. This is because
|
|
|
|
* those pointers can change without the lock.
|
|
|
|
*/
|
2009-06-24 03:54:19 +02:00
|
|
|
iter->pg = ftrace_pages_start;
|
|
|
|
iter->idx = 0;
|
|
|
|
for (l = 0; l <= *pos; ) {
|
|
|
|
p = t_next(m, p, &l);
|
|
|
|
if (!p)
|
|
|
|
break;
|
2008-11-28 05:13:21 +01:00
|
|
|
}
|
function tracing: fix wrong pos computing when read buffer has been fulfilled
Impact: make output of available_filter_functions complete
phenomenon:
The first value of dyn_ftrace_total_info is not equal with
`cat available_filter_functions | wc -l`, but they should be equal.
root cause:
When printing functions with seq_printf in t_show, if the read buffer
is just overflowed by current function record, then this function
won't be printed to user space through read buffer, it will
just be dropped. So we can't see this function printing.
So, every time the last function to fill the read buffer, if overflowed,
will be dropped.
This also applies to set_ftrace_filter if set_ftrace_filter has
more bytes than read buffer.
fix:
Through checking return value of seq_printf, if less than 0, we know
this function doesn't be printed. Then we decrease position to force
this function to be printed next time, in next read buffer.
Another little fix is to show correct allocating pages count.
Signed-off-by: walimis <walimisdev@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-15 08:19:06 +01:00
|
|
|
|
2011-12-19 21:21:16 +01:00
|
|
|
if (!p)
|
|
|
|
return t_hash_start(m, pos);
|
2010-09-09 16:00:28 +02:00
|
|
|
|
|
|
|
return iter;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void t_stop(struct seq_file *m, void *p)
|
|
|
|
{
|
2009-02-16 21:28:00 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2014-07-03 20:51:36 +02:00
|
|
|
void * __weak
|
|
|
|
arch_ftrace_trampoline_func(struct ftrace_ops *ops, struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void add_trampoline_func(struct seq_file *m, struct ftrace_ops *ops,
|
|
|
|
struct dyn_ftrace *rec)
|
|
|
|
{
|
|
|
|
void *ptr;
|
|
|
|
|
|
|
|
ptr = arch_ftrace_trampoline_func(ops, rec);
|
|
|
|
if (ptr)
|
|
|
|
seq_printf(m, " ->%pS", ptr);
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
static int t_show(struct seq_file *m, void *v)
|
|
|
|
{
|
2009-02-16 17:21:52 +01:00
|
|
|
struct ftrace_iterator *iter = m->private;
|
2010-09-09 16:00:28 +02:00
|
|
|
struct dyn_ftrace *rec;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-02-16 21:28:00 +01:00
|
|
|
if (iter->flags & FTRACE_ITER_HASH)
|
2010-09-09 16:00:28 +02:00
|
|
|
return t_hash_show(m, iter);
|
2009-02-16 21:28:00 +01:00
|
|
|
|
2009-02-16 17:21:52 +01:00
|
|
|
if (iter->flags & FTRACE_ITER_PRINTALL) {
|
2014-06-13 09:24:06 +02:00
|
|
|
if (iter->flags & FTRACE_ITER_NOTRACE)
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "#### no functions disabled ####\n");
|
2014-06-13 09:24:06 +02:00
|
|
|
else
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "#### all functions enabled ####\n");
|
2009-02-16 17:21:52 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-09-09 16:00:28 +02:00
|
|
|
rec = iter->func;
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
if (!rec)
|
|
|
|
return 0;
|
|
|
|
|
2011-05-03 20:39:21 +02:00
|
|
|
seq_printf(m, "%ps", (void *)rec->ip);
|
2014-05-09 22:54:59 +02:00
|
|
|
if (iter->flags & FTRACE_ITER_ENABLED) {
|
2014-07-03 20:51:36 +02:00
|
|
|
struct ftrace_ops *ops = NULL;
|
|
|
|
|
2014-11-21 11:25:16 +01:00
|
|
|
seq_printf(m, " (%ld)%s%s",
|
2014-05-07 19:46:45 +02:00
|
|
|
ftrace_rec_count(rec),
|
2014-11-21 11:25:16 +01:00
|
|
|
rec->flags & FTRACE_FL_REGS ? " R" : " ",
|
|
|
|
rec->flags & FTRACE_FL_IPMODIFY ? " I" : " ");
|
2014-05-09 22:54:59 +02:00
|
|
|
if (rec->flags & FTRACE_FL_TRAMP_EN) {
|
2014-07-24 22:00:31 +02:00
|
|
|
ops = ftrace_find_tramp_ops_any(rec);
|
2014-07-24 18:25:47 +02:00
|
|
|
if (ops)
|
2014-05-09 22:54:59 +02:00
|
|
|
seq_printf(m, "\ttramp: %pS",
|
|
|
|
(void *)ops->trampoline);
|
|
|
|
else
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "\ttramp: ERROR!");
|
2014-07-03 20:51:36 +02:00
|
|
|
|
2014-05-09 22:54:59 +02:00
|
|
|
}
|
2014-07-03 20:51:36 +02:00
|
|
|
add_trampoline_func(m, ops, rec);
|
2014-05-09 22:54:59 +02:00
|
|
|
}
|
|
|
|
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_putc(m, '\n');
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-23 01:43:43 +02:00
|
|
|
static const struct seq_operations show_ftrace_seq_ops = {
|
2008-05-12 21:20:43 +02:00
|
|
|
.start = t_start,
|
|
|
|
.next = t_next,
|
|
|
|
.stop = t_stop,
|
|
|
|
.show = t_show,
|
|
|
|
};
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static int
|
2008-05-12 21:20:43 +02:00
|
|
|
ftrace_avail_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter;
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2012-04-25 10:23:39 +02:00
|
|
|
iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter));
|
|
|
|
if (iter) {
|
|
|
|
iter->pg = ftrace_pages_start;
|
|
|
|
iter->ops = &global_ops;
|
2008-05-12 21:20:46 +02:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2012-04-25 10:23:39 +02:00
|
|
|
return iter ? 0 : -ENOMEM;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2011-05-03 20:39:21 +02:00
|
|
|
static int
|
|
|
|
ftrace_enabled_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter;
|
|
|
|
|
2012-04-25 10:23:39 +02:00
|
|
|
iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter));
|
|
|
|
if (iter) {
|
|
|
|
iter->pg = ftrace_pages_start;
|
|
|
|
iter->flags = FTRACE_ITER_ENABLED;
|
|
|
|
iter->ops = &global_ops;
|
2011-05-03 20:39:21 +02:00
|
|
|
}
|
|
|
|
|
2012-04-25 10:23:39 +02:00
|
|
|
return iter ? 0 : -ENOMEM;
|
2011-05-03 20:39:21 +02:00
|
|
|
}
|
|
|
|
|
2011-12-19 20:41:25 +01:00
|
|
|
/**
|
|
|
|
* ftrace_regex_open - initialize function tracer filter files
|
|
|
|
* @ops: The ftrace_ops that hold the hash filters
|
|
|
|
* @flag: The type of filter to process
|
|
|
|
* @inode: The inode, usually passed in to your open routine
|
|
|
|
* @file: The file, usually passed in to your open routine
|
|
|
|
*
|
|
|
|
* ftrace_regex_open() initializes the filter files for the
|
|
|
|
* @ops. Depending on @flag it may process the filter hash or
|
|
|
|
* the notrace hash of @ops. With this called from the open
|
|
|
|
* routine, you can use ftrace_filter_write() for the write
|
|
|
|
* routine if @flag has FTRACE_ITER_FILTER set, or
|
|
|
|
* ftrace_notrace_write() if @flag has FTRACE_ITER_NOTRACE set.
|
2013-12-21 23:39:40 +01:00
|
|
|
* tracing_lseek() should be used as the lseek routine, and
|
2011-12-19 20:41:25 +01:00
|
|
|
* release must call ftrace_regex_release().
|
|
|
|
*/
|
|
|
|
int
|
2011-05-02 18:29:25 +02:00
|
|
|
ftrace_regex_open(struct ftrace_ops *ops, int flag,
|
2011-04-30 02:59:51 +02:00
|
|
|
struct inode *inode, struct file *file)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter;
|
2011-05-02 18:29:25 +02:00
|
|
|
struct ftrace_hash *hash;
|
2008-05-12 21:20:43 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
|
|
|
|
2008-05-12 21:20:48 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
iter = kzalloc(sizeof(*iter), GFP_KERNEL);
|
|
|
|
if (!iter)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
if (trace_parser_get_init(&iter->parser, FTRACE_BUFF_MAX)) {
|
|
|
|
kfree(iter);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2013-05-09 07:44:21 +02:00
|
|
|
iter->ops = ops;
|
|
|
|
iter->flags = flag;
|
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_lock(&ops->func_hash->regex_lock);
|
2013-05-09 07:44:21 +02:00
|
|
|
|
2011-05-02 18:29:25 +02:00
|
|
|
if (flag & FTRACE_ITER_NOTRACE)
|
2014-08-15 23:23:02 +02:00
|
|
|
hash = ops->func_hash->notrace_hash;
|
2011-05-02 18:29:25 +02:00
|
|
|
else
|
2014-08-15 23:23:02 +02:00
|
|
|
hash = ops->func_hash->filter_hash;
|
2011-05-02 18:29:25 +02:00
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
if (file->f_mode & FMODE_WRITE) {
|
2014-06-11 10:06:54 +02:00
|
|
|
const int size_bits = FTRACE_HASH_DEFAULT_BITS;
|
|
|
|
|
|
|
|
if (file->f_flags & O_TRUNC)
|
|
|
|
iter->hash = alloc_ftrace_hash(size_bits);
|
|
|
|
else
|
|
|
|
iter->hash = alloc_and_copy_ftrace_hash(size_bits, hash);
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
if (!iter->hash) {
|
|
|
|
trace_parser_put(&iter->parser);
|
|
|
|
kfree(iter);
|
2013-05-09 07:44:21 +02:00
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out_unlock;
|
2011-05-02 23:34:47 +02:00
|
|
|
}
|
|
|
|
}
|
2011-04-30 02:59:51 +02:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
iter->pg = ftrace_pages_start;
|
|
|
|
|
|
|
|
ret = seq_open(file, &show_ftrace_seq_ops);
|
|
|
|
if (!ret) {
|
|
|
|
struct seq_file *m = file->private_data;
|
|
|
|
m->private = iter;
|
2009-09-22 07:54:28 +02:00
|
|
|
} else {
|
2011-05-02 23:34:47 +02:00
|
|
|
/* Failed */
|
|
|
|
free_ftrace_hash(iter->hash);
|
2009-09-22 07:54:28 +02:00
|
|
|
trace_parser_put(&iter->parser);
|
2008-05-12 21:20:43 +02:00
|
|
|
kfree(iter);
|
2009-09-22 07:54:28 +02:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
} else
|
|
|
|
file->private_data = iter;
|
2013-05-09 07:44:21 +02:00
|
|
|
|
|
|
|
out_unlock:
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_unlock(&ops->func_hash->regex_lock);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-05-22 17:46:33 +02:00
|
|
|
static int
|
|
|
|
ftrace_filter_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
2013-11-12 05:07:14 +01:00
|
|
|
struct ftrace_ops *ops = inode->i_private;
|
|
|
|
|
|
|
|
return ftrace_regex_open(ops,
|
2011-12-19 21:21:16 +01:00
|
|
|
FTRACE_ITER_FILTER | FTRACE_ITER_DO_HASH,
|
|
|
|
inode, file);
|
2008-05-22 17:46:33 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
ftrace_notrace_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
2013-11-12 05:07:14 +01:00
|
|
|
struct ftrace_ops *ops = inode->i_private;
|
|
|
|
|
|
|
|
return ftrace_regex_open(ops, FTRACE_ITER_NOTRACE,
|
2011-04-30 02:59:51 +02:00
|
|
|
inode, file);
|
2008-05-22 17:46:33 +02:00
|
|
|
}
|
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
/* Type for quick search ftrace basic regexes (globs) from filter_parse_regex */
|
|
|
|
struct ftrace_glob {
|
|
|
|
char *search;
|
|
|
|
unsigned len;
|
|
|
|
int type;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int ftrace_match(char *str, struct ftrace_glob *g)
|
2009-02-13 21:56:43 +01:00
|
|
|
{
|
|
|
|
int matched = 0;
|
2010-01-14 03:53:02 +01:00
|
|
|
int slen;
|
2009-02-13 21:56:43 +01:00
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
switch (g->type) {
|
2009-02-13 21:56:43 +01:00
|
|
|
case MATCH_FULL:
|
2015-09-29 18:46:14 +02:00
|
|
|
if (strcmp(str, g->search) == 0)
|
2009-02-13 21:56:43 +01:00
|
|
|
matched = 1;
|
|
|
|
break;
|
|
|
|
case MATCH_FRONT_ONLY:
|
2015-09-29 18:46:14 +02:00
|
|
|
if (strncmp(str, g->search, g->len) == 0)
|
2009-02-13 21:56:43 +01:00
|
|
|
matched = 1;
|
|
|
|
break;
|
|
|
|
case MATCH_MIDDLE_ONLY:
|
2015-09-29 18:46:14 +02:00
|
|
|
if (strstr(str, g->search))
|
2009-02-13 21:56:43 +01:00
|
|
|
matched = 1;
|
|
|
|
break;
|
|
|
|
case MATCH_END_ONLY:
|
2010-01-14 03:53:02 +01:00
|
|
|
slen = strlen(str);
|
2015-09-29 18:46:14 +02:00
|
|
|
if (slen >= g->len &&
|
|
|
|
memcmp(str + slen - g->len, g->search, g->len) == 0)
|
2009-02-13 21:56:43 +01:00
|
|
|
matched = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return matched;
|
|
|
|
}
|
|
|
|
|
2011-04-29 21:12:32 +02:00
|
|
|
static int
|
2015-09-29 18:46:13 +02:00
|
|
|
enter_record(struct ftrace_hash *hash, struct dyn_ftrace *rec, int clear_filter)
|
2011-04-26 22:11:03 +02:00
|
|
|
{
|
2011-04-29 21:12:32 +02:00
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
int ret = 0;
|
|
|
|
|
2011-04-30 02:59:51 +02:00
|
|
|
entry = ftrace_lookup_ip(hash, rec->ip);
|
2015-09-29 18:46:13 +02:00
|
|
|
if (clear_filter) {
|
2011-04-30 02:59:51 +02:00
|
|
|
/* Do nothing if it doesn't exist */
|
|
|
|
if (!entry)
|
|
|
|
return 0;
|
2011-04-29 21:12:32 +02:00
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
free_hash_entry(hash, entry);
|
2011-04-30 02:59:51 +02:00
|
|
|
} else {
|
|
|
|
/* Do nothing if it exists */
|
|
|
|
if (entry)
|
|
|
|
return 0;
|
2011-04-29 21:12:32 +02:00
|
|
|
|
2011-04-30 02:59:51 +02:00
|
|
|
ret = add_hash_entry(hash, rec->ip);
|
2011-04-29 21:12:32 +02:00
|
|
|
}
|
|
|
|
return ret;
|
2011-04-26 22:11:03 +02:00
|
|
|
}
|
|
|
|
|
2009-02-13 23:08:48 +01:00
|
|
|
static int
|
2015-09-29 18:46:15 +02:00
|
|
|
ftrace_match_record(struct dyn_ftrace *rec, struct ftrace_glob *func_g,
|
|
|
|
struct ftrace_glob *mod_g, int exclude_mod)
|
2009-02-13 23:08:48 +01:00
|
|
|
{
|
|
|
|
char str[KSYM_SYMBOL_LEN];
|
2011-04-29 02:32:08 +02:00
|
|
|
char *modname;
|
|
|
|
|
|
|
|
kallsyms_lookup(rec->ip, NULL, NULL, &modname, str);
|
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
if (mod_g) {
|
|
|
|
int mod_matches = (modname) ? ftrace_match(modname, mod_g) : 0;
|
|
|
|
|
|
|
|
/* blank module name to match all modules */
|
|
|
|
if (!mod_g->len) {
|
|
|
|
/* blank module globbing: modname xor exclude_mod */
|
|
|
|
if ((!exclude_mod) != (!modname))
|
|
|
|
goto func_match;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* not matching the module */
|
|
|
|
if (!modname || !mod_matches) {
|
|
|
|
if (exclude_mod)
|
|
|
|
goto func_match;
|
|
|
|
else
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mod_matches && exclude_mod)
|
2011-04-29 02:32:08 +02:00
|
|
|
return 0;
|
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
func_match:
|
2011-04-29 02:32:08 +02:00
|
|
|
/* blank search means to match all funcs in the mod */
|
2015-09-29 18:46:14 +02:00
|
|
|
if (!func_g->len)
|
2011-04-29 02:32:08 +02:00
|
|
|
return 1;
|
|
|
|
}
|
2009-02-13 23:08:48 +01:00
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
return ftrace_match(str, func_g);
|
2009-02-13 23:08:48 +01:00
|
|
|
}
|
|
|
|
|
2011-04-30 02:59:51 +02:00
|
|
|
static int
|
2015-09-29 18:46:14 +02:00
|
|
|
match_records(struct ftrace_hash *hash, char *func, int len, char *mod)
|
2009-02-13 21:56:43 +01:00
|
|
|
{
|
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec;
|
2015-09-29 18:46:14 +02:00
|
|
|
struct ftrace_glob func_g = { .type = MATCH_FULL };
|
2015-09-29 18:46:15 +02:00
|
|
|
struct ftrace_glob mod_g = { .type = MATCH_FULL };
|
|
|
|
struct ftrace_glob *mod_match = (mod) ? &mod_g : NULL;
|
|
|
|
int exclude_mod = 0;
|
2009-12-08 04:15:11 +01:00
|
|
|
int found = 0;
|
2011-04-29 21:12:32 +02:00
|
|
|
int ret;
|
2015-09-29 18:46:13 +02:00
|
|
|
int clear_filter;
|
2009-02-13 21:56:43 +01:00
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
if (func) {
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.type = filter_parse_regex(func, len, &func_g.search,
|
|
|
|
&clear_filter);
|
|
|
|
func_g.len = strlen(func_g.search);
|
2011-04-29 02:32:08 +02:00
|
|
|
}
|
2009-02-13 21:56:43 +01:00
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
if (mod) {
|
|
|
|
mod_g.type = filter_parse_regex(mod, strlen(mod),
|
|
|
|
&mod_g.search, &exclude_mod);
|
|
|
|
mod_g.len = strlen(mod_g.search);
|
2011-04-29 02:32:08 +02:00
|
|
|
}
|
2009-02-13 21:56:43 +01:00
|
|
|
|
2009-02-14 07:15:39 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2009-02-13 18:43:56 +01:00
|
|
|
|
2011-04-29 02:32:08 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
goto out_unlock;
|
2009-02-13 21:56:43 +01:00
|
|
|
|
2009-02-13 18:43:56 +01:00
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
2015-09-29 18:46:15 +02:00
|
|
|
if (ftrace_match_record(rec, &func_g, mod_match, exclude_mod)) {
|
2015-09-29 18:46:13 +02:00
|
|
|
ret = enter_record(hash, rec, clear_filter);
|
2011-04-29 21:12:32 +02:00
|
|
|
if (ret < 0) {
|
|
|
|
found = ret;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2009-12-08 04:15:11 +01:00
|
|
|
found = 1;
|
2009-02-13 18:43:56 +01:00
|
|
|
}
|
|
|
|
} while_for_each_ftrace_rec();
|
2011-04-29 02:32:08 +02:00
|
|
|
out_unlock:
|
2009-02-14 07:15:39 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2009-12-08 04:15:11 +01:00
|
|
|
|
|
|
|
return found;
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2009-02-13 23:08:48 +01:00
|
|
|
static int
|
2011-04-30 02:59:51 +02:00
|
|
|
ftrace_match_records(struct ftrace_hash *hash, char *buff, int len)
|
2009-02-13 23:08:48 +01:00
|
|
|
{
|
2015-09-29 18:46:13 +02:00
|
|
|
return match_records(hash, buff, len, NULL);
|
2009-02-13 23:08:48 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2009-02-14 06:40:25 +01:00
|
|
|
/*
|
|
|
|
* We register the module command as a template to show others how
|
|
|
|
* to register the a command as well.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int
|
2011-07-07 17:09:22 +02:00
|
|
|
ftrace_mod_callback(struct ftrace_hash *hash,
|
2015-09-29 18:46:13 +02:00
|
|
|
char *func, char *cmd, char *module, int enable)
|
2009-02-14 06:40:25 +01:00
|
|
|
{
|
2015-09-29 18:46:12 +02:00
|
|
|
int ret;
|
2009-02-14 06:40:25 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* cmd == 'mod' because we only registered this func
|
|
|
|
* for the 'mod' ftrace_func_command.
|
|
|
|
* But if you register one func with multiple commands,
|
|
|
|
* you can tell which command was used by the cmd
|
|
|
|
* parameter.
|
|
|
|
*/
|
2015-09-29 18:46:13 +02:00
|
|
|
ret = match_records(hash, func, strlen(func), module);
|
2011-04-29 21:12:32 +02:00
|
|
|
if (!ret)
|
2015-09-29 18:46:12 +02:00
|
|
|
return -EINVAL;
|
2011-04-29 21:12:32 +02:00
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
return 0;
|
2009-02-14 06:40:25 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct ftrace_func_command ftrace_mod_cmd = {
|
|
|
|
.name = "mod",
|
|
|
|
.func = ftrace_mod_callback,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init ftrace_mod_cmd_init(void)
|
|
|
|
{
|
|
|
|
return register_ftrace_command(&ftrace_mod_cmd);
|
|
|
|
}
|
2012-10-05 18:13:07 +02:00
|
|
|
core_initcall(ftrace_mod_cmd_init);
|
2009-02-14 06:40:25 +01:00
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
static void function_trace_probe_call(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *op, struct pt_regs *pt_regs)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
2009-02-17 18:32:04 +01:00
|
|
|
struct ftrace_func_probe *entry;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
struct hlist_head *hhd;
|
|
|
|
unsigned long key;
|
|
|
|
|
|
|
|
key = hash_long(ip, FTRACE_HASH_BITS);
|
|
|
|
|
|
|
|
hhd = &ftrace_func_hash[key];
|
|
|
|
|
|
|
|
if (hlist_empty(hhd))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Disable preemption for these calls to prevent a RCU grace
|
|
|
|
* period. This syncs the hash iteration and freeing of items
|
|
|
|
* on the hash. rcu_read_lock is too dangerous here.
|
|
|
|
*/
|
2010-06-03 15:36:50 +02:00
|
|
|
preempt_disable_notrace();
|
2013-05-28 20:38:43 +02:00
|
|
|
hlist_for_each_entry_rcu_notrace(entry, hhd, node) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
if (entry->ip == ip)
|
|
|
|
entry->ops->func(ip, parent_ip, &entry->data);
|
|
|
|
}
|
2010-06-03 15:36:50 +02:00
|
|
|
preempt_enable_notrace();
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
static struct ftrace_ops trace_probe_ops __read_mostly =
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
2009-03-25 18:26:41 +01:00
|
|
|
.func = function_trace_probe_call,
|
2013-05-09 07:44:17 +02:00
|
|
|
.flags = FTRACE_OPS_FL_INITIALIZED,
|
2014-08-15 23:23:02 +02:00
|
|
|
INIT_OPS_HASH(trace_probe_ops)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
};
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
static int ftrace_probe_registered;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2015-01-13 20:03:38 +01:00
|
|
|
static void __enable_ftrace_function_probe(struct ftrace_ops_hash *old_hash)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
2011-05-04 15:27:52 +02:00
|
|
|
int ret;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
int i;
|
|
|
|
|
2013-05-10 01:37:36 +02:00
|
|
|
if (ftrace_probe_registered) {
|
|
|
|
/* still need to update the function call sites */
|
|
|
|
if (ftrace_enabled)
|
2014-10-24 20:56:01 +02:00
|
|
|
ftrace_run_modify_code(&trace_probe_ops, FTRACE_UPDATE_CALLS,
|
|
|
|
old_hash);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
return;
|
2013-05-10 01:37:36 +02:00
|
|
|
}
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
|
|
|
for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) {
|
|
|
|
struct hlist_head *hhd = &ftrace_func_hash[i];
|
|
|
|
if (hhd->first)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
/* Nothing registered? */
|
|
|
|
if (i == FTRACE_FUNC_HASHSIZE)
|
|
|
|
return;
|
|
|
|
|
2013-11-26 02:59:46 +01:00
|
|
|
ret = ftrace_startup(&trace_probe_ops, 0);
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
ftrace_probe_registered = 1;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
static void __disable_ftrace_function_probe(void)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
if (!ftrace_probe_registered)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
return;
|
|
|
|
|
|
|
|
for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) {
|
|
|
|
struct hlist_head *hhd = &ftrace_func_hash[i];
|
|
|
|
if (hhd->first)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* no more funcs left */
|
2013-11-26 02:59:46 +01:00
|
|
|
ftrace_shutdown(&trace_probe_ops, 0);
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
ftrace_probe_registered = 0;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2013-03-13 17:42:58 +01:00
|
|
|
static void ftrace_free_entry(struct ftrace_func_probe *entry)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
|
|
|
if (entry->ops->free)
|
2013-03-12 20:07:59 +01:00
|
|
|
entry->ops->free(entry->ops, entry->ip, &entry->data);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
kfree(entry);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2009-02-17 18:32:04 +01:00
|
|
|
register_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
void *data)
|
|
|
|
{
|
2015-01-13 20:03:38 +01:00
|
|
|
struct ftrace_ops_hash old_hash_ops;
|
2009-02-17 18:32:04 +01:00
|
|
|
struct ftrace_func_probe *entry;
|
2015-09-29 18:46:14 +02:00
|
|
|
struct ftrace_glob func_g;
|
2014-08-15 23:23:02 +02:00
|
|
|
struct ftrace_hash **orig_hash = &trace_probe_ops.func_hash->filter_hash;
|
2014-07-24 21:33:41 +02:00
|
|
|
struct ftrace_hash *old_hash = *orig_hash;
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
struct ftrace_hash *hash;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
struct ftrace_page *pg;
|
|
|
|
struct dyn_ftrace *rec;
|
2015-09-29 18:46:14 +02:00
|
|
|
int not;
|
2009-02-17 17:20:26 +01:00
|
|
|
unsigned long key;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
int count = 0;
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
int ret;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.type = filter_parse_regex(glob, strlen(glob),
|
|
|
|
&func_g.search, ¬);
|
|
|
|
func_g.len = strlen(func_g.search);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
/* we do not support '!' for function probes */
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
if (WARN_ON(not))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_lock(&trace_probe_ops.func_hash->regex_lock);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2015-01-13 20:03:38 +01:00
|
|
|
old_hash_ops.filter_hash = old_hash;
|
|
|
|
/* Probes only have filters */
|
|
|
|
old_hash_ops.notrace_hash = NULL;
|
|
|
|
|
2014-07-24 21:33:41 +02:00
|
|
|
hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, old_hash);
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
if (!hash) {
|
|
|
|
count = -ENOMEM;
|
2013-05-10 00:20:37 +02:00
|
|
|
goto out;
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(ftrace_disabled)) {
|
|
|
|
count = -ENODEV;
|
2013-05-10 00:20:37 +02:00
|
|
|
goto out;
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
}
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2013-05-10 00:20:37 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
if (!ftrace_match_record(rec, &func_g, NULL, 0))
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
continue;
|
|
|
|
|
|
|
|
entry = kmalloc(sizeof(*entry), GFP_KERNEL);
|
|
|
|
if (!entry) {
|
2009-02-17 18:32:04 +01:00
|
|
|
/* If we did not process any, then return error */
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
if (!count)
|
|
|
|
count = -ENOMEM;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
count++;
|
|
|
|
|
|
|
|
entry->data = data;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The caller might want to do something special
|
|
|
|
* for each function we find. We call the callback
|
|
|
|
* to give the caller an opportunity to do so.
|
|
|
|
*/
|
2013-03-12 20:07:59 +01:00
|
|
|
if (ops->init) {
|
|
|
|
if (ops->init(ops, rec->ip, &entry->data) < 0) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
/* caller does not like this func */
|
|
|
|
kfree(entry);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
ret = enter_record(hash, rec, 0);
|
|
|
|
if (ret < 0) {
|
|
|
|
kfree(entry);
|
|
|
|
count = ret;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
entry->ops = ops;
|
|
|
|
entry->ip = rec->ip;
|
|
|
|
|
|
|
|
key = hash_long(entry->ip, FTRACE_HASH_BITS);
|
|
|
|
hlist_add_head_rcu(&entry->node, &ftrace_func_hash[key]);
|
|
|
|
|
|
|
|
} while_for_each_ftrace_rec();
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
|
|
|
|
ret = ftrace_hash_move(&trace_probe_ops, 1, orig_hash, hash);
|
2014-10-24 20:56:01 +02:00
|
|
|
|
2015-01-13 20:03:38 +01:00
|
|
|
__enable_ftrace_function_probe(&old_hash_ops);
|
2014-10-24 20:56:01 +02:00
|
|
|
|
2014-07-24 21:33:41 +02:00
|
|
|
if (!ret)
|
|
|
|
free_ftrace_hash_rcu(old_hash);
|
|
|
|
else
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
count = ret;
|
|
|
|
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
out_unlock:
|
2013-05-10 00:20:37 +02:00
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
out:
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_unlock(&trace_probe_ops.func_hash->regex_lock);
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
free_ftrace_hash(hash);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
|
|
|
return count;
|
|
|
|
}
|
|
|
|
|
|
|
|
enum {
|
2009-02-17 18:32:04 +01:00
|
|
|
PROBE_TEST_FUNC = 1,
|
|
|
|
PROBE_TEST_DATA = 2
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
static void
|
2009-02-17 18:32:04 +01:00
|
|
|
__unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
void *data, int flags)
|
|
|
|
{
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
struct ftrace_func_entry *rec_entry;
|
2009-02-17 18:32:04 +01:00
|
|
|
struct ftrace_func_probe *entry;
|
2013-03-13 17:42:58 +01:00
|
|
|
struct ftrace_func_probe *p;
|
2015-09-29 18:46:14 +02:00
|
|
|
struct ftrace_glob func_g;
|
2014-08-15 23:23:02 +02:00
|
|
|
struct ftrace_hash **orig_hash = &trace_probe_ops.func_hash->filter_hash;
|
2014-07-24 21:33:41 +02:00
|
|
|
struct ftrace_hash *old_hash = *orig_hash;
|
2013-03-13 17:42:58 +01:00
|
|
|
struct list_head free_list;
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
struct ftrace_hash *hash;
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
struct hlist_node *tmp;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
char str[KSYM_SYMBOL_LEN];
|
2015-09-29 18:46:14 +02:00
|
|
|
int i, ret;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2009-09-15 12:06:30 +02:00
|
|
|
if (glob && (strcmp(glob, "*") == 0 || !strlen(glob)))
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.search = NULL;
|
2009-09-15 12:06:30 +02:00
|
|
|
else if (glob) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
int not;
|
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.type = filter_parse_regex(glob, strlen(glob),
|
|
|
|
&func_g.search, ¬);
|
|
|
|
func_g.len = strlen(func_g.search);
|
|
|
|
func_g.search = glob;
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
/* we do not support '!' for function probes */
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
if (WARN_ON(not))
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_lock(&trace_probe_ops.func_hash->regex_lock);
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
|
|
|
|
hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, *orig_hash);
|
|
|
|
if (!hash)
|
|
|
|
/* Hmm, should report this somehow */
|
|
|
|
goto out_unlock;
|
|
|
|
|
2013-03-13 17:42:58 +01:00
|
|
|
INIT_LIST_HEAD(&free_list);
|
|
|
|
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) {
|
|
|
|
struct hlist_head *hhd = &ftrace_func_hash[i];
|
|
|
|
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 02:06:00 +01:00
|
|
|
hlist_for_each_entry_safe(entry, tmp, hhd, node) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
|
|
|
|
/* break up if statements for readability */
|
2009-02-17 18:32:04 +01:00
|
|
|
if ((flags & PROBE_TEST_FUNC) && entry->ops != ops)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
continue;
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
if ((flags & PROBE_TEST_DATA) && entry->data != data)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
continue;
|
|
|
|
|
|
|
|
/* do this last, since it is the most expensive */
|
2015-09-29 18:46:14 +02:00
|
|
|
if (func_g.search) {
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
kallsyms_lookup(entry->ip, NULL, NULL,
|
|
|
|
NULL, str);
|
2015-09-29 18:46:14 +02:00
|
|
|
if (!ftrace_match(str, &func_g))
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
rec_entry = ftrace_lookup_ip(hash, entry->ip);
|
|
|
|
/* It is possible more than one entry had this ip */
|
|
|
|
if (rec_entry)
|
|
|
|
free_hash_entry(hash, rec_entry);
|
|
|
|
|
2013-03-13 16:15:19 +01:00
|
|
|
hlist_del_rcu(&entry->node);
|
2013-03-13 17:42:58 +01:00
|
|
|
list_add(&entry->free_list, &free_list);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
}
|
2013-05-09 07:44:21 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
2009-02-17 18:32:04 +01:00
|
|
|
__disable_ftrace_function_probe();
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
/*
|
|
|
|
* Remove after the disable is called. Otherwise, if the last
|
|
|
|
* probe is removed, a null hash means *all enabled*.
|
|
|
|
*/
|
2014-07-24 21:33:41 +02:00
|
|
|
ret = ftrace_hash_move(&trace_probe_ops, 1, orig_hash, hash);
|
2013-03-13 17:42:58 +01:00
|
|
|
synchronize_sched();
|
2014-07-24 21:33:41 +02:00
|
|
|
if (!ret)
|
|
|
|
free_ftrace_hash_rcu(old_hash);
|
|
|
|
|
2013-03-13 17:42:58 +01:00
|
|
|
list_for_each_entry_safe(entry, p, &free_list, free_list) {
|
|
|
|
list_del(&entry->free_list);
|
|
|
|
ftrace_free_entry(entry);
|
|
|
|
}
|
2013-05-09 07:44:21 +02:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2015-09-29 18:46:14 +02:00
|
|
|
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
out_unlock:
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_unlock(&trace_probe_ops.func_hash->regex_lock);
|
ftrace: Fix function probe to only enable needed functions
Currently the function probe enables all functions and runs a "hash"
against every function call to see if it should call a probe. This
is extremely wasteful.
Note, a probe is something like:
echo schedule:traceoff > /debug/tracing/set_ftrace_filter
When schedule is called, the probe will disable tracing. But currently,
it has a call back for *all* functions, and checks to see if the
called function is the probe that is needed.
The probe function has been created before ftrace was rewritten to
allow for more than one "op" to be registered by the function tracer.
When probes were created, it couldn't limit the functions without also
limiting normal function calls. But now we can, it's about time
to update the probe code.
Todo, have separate ops for different entries. That is, assign
a ftrace_ops per probe, instead of one op for all probes. But
as there's not many probes assigned, this may not be that urgent.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-03-12 15:09:42 +01:00
|
|
|
free_ftrace_hash(hash);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void
|
2009-02-17 18:32:04 +01:00
|
|
|
unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
void *data)
|
|
|
|
{
|
2009-02-17 18:32:04 +01:00
|
|
|
__unregister_ftrace_function_probe(glob, ops, data,
|
|
|
|
PROBE_TEST_FUNC | PROBE_TEST_DATA);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void
|
2009-02-17 18:32:04 +01:00
|
|
|
unregister_ftrace_function_probe_func(char *glob, struct ftrace_probe_ops *ops)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
2009-02-17 18:32:04 +01:00
|
|
|
__unregister_ftrace_function_probe(glob, ops, NULL, PROBE_TEST_FUNC);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
2009-02-17 18:32:04 +01:00
|
|
|
void unregister_ftrace_function_probe_all(char *glob)
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
{
|
2009-02-17 18:32:04 +01:00
|
|
|
__unregister_ftrace_function_probe(glob, NULL, NULL, 0);
|
ftrace: trace different functions with a different tracer
Impact: new feature
Currently, the function tracer only gives you an ability to hook
a tracer to all functions being traced. The dynamic function trace
allows you to pick and choose which of those functions will be
traced, but all functions being traced will call all tracers that
registered with the function tracer.
This patch adds a new feature that allows a tracer to hook to specific
functions, even when all functions are being traced. It allows for
different functions to call different tracer hooks.
The way this is accomplished is by a special function that will hook
to the function tracer and will set up a hash table knowing which
tracer hook to call with which function. This is the most general
and easiest method to accomplish this. Later, an arch may choose
to supply their own method in changing the mcount call of a function
to call a different tracer. But that will be an exercise for the
future.
To register a function:
struct ftrace_hook_ops {
void (*func)(unsigned long ip,
unsigned long parent_ip,
void **data);
int (*callback)(unsigned long ip, void **data);
void (*free)(void **data);
};
int register_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data);
glob is a simple glob to search for the functions to hook.
ops is a pointer to the operations (listed below)
data is the default data to be passed to the hook functions when traced
ops:
func is the hook function to call when the functions are traced
callback is a callback function that is called when setting up the hash.
That is, if the tracer needs to do something special for each
function, that is being traced, and wants to give each function
its own data. The address of the entry data is passed to this
callback, so that the callback may wish to update the entry to
whatever it would like.
free is a callback for when the entry is freed. In case the tracer
allocated any data, it is give the chance to free it.
To unregister we have three functions:
void
unregister_ftrace_function_hook(char *glob, struct ftrace_hook_ops *ops,
void *data)
This will unregister all hooks that match glob, point to ops, and
have its data matching data. (note, if glob is NULL, blank or '*',
all functions will be tested).
void
unregister_ftrace_function_hook_func(char *glob,
struct ftrace_hook_ops *ops)
This will unregister all functions matching glob that has an entry
pointing to ops.
void unregister_ftrace_function_hook_all(char *glob)
This simply unregisters all funcs.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
2009-02-14 21:29:06 +01:00
|
|
|
}
|
|
|
|
|
2009-02-14 06:40:25 +01:00
|
|
|
static LIST_HEAD(ftrace_commands);
|
|
|
|
static DEFINE_MUTEX(ftrace_cmd_mutex);
|
|
|
|
|
2013-10-24 15:34:18 +02:00
|
|
|
/*
|
|
|
|
* Currently we only register ftrace commands from __init, so mark this
|
|
|
|
* __init too.
|
|
|
|
*/
|
|
|
|
__init int register_ftrace_command(struct ftrace_func_command *cmd)
|
2009-02-14 06:40:25 +01:00
|
|
|
{
|
|
|
|
struct ftrace_func_command *p;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
mutex_lock(&ftrace_cmd_mutex);
|
|
|
|
list_for_each_entry(p, &ftrace_commands, list) {
|
|
|
|
if (strcmp(cmd->name, p->name) == 0) {
|
|
|
|
ret = -EBUSY;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
list_add(&cmd->list, &ftrace_commands);
|
|
|
|
out_unlock:
|
|
|
|
mutex_unlock(&ftrace_cmd_mutex);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-10-24 15:34:18 +02:00
|
|
|
/*
|
|
|
|
* Currently we only unregister ftrace commands from __init, so mark
|
|
|
|
* this __init too.
|
|
|
|
*/
|
|
|
|
__init int unregister_ftrace_command(struct ftrace_func_command *cmd)
|
2009-02-14 06:40:25 +01:00
|
|
|
{
|
|
|
|
struct ftrace_func_command *p, *n;
|
|
|
|
int ret = -ENODEV;
|
|
|
|
|
|
|
|
mutex_lock(&ftrace_cmd_mutex);
|
|
|
|
list_for_each_entry_safe(p, n, &ftrace_commands, list) {
|
|
|
|
if (strcmp(cmd->name, p->name) == 0) {
|
|
|
|
ret = 0;
|
|
|
|
list_del_init(&p->list);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
out_unlock:
|
|
|
|
mutex_unlock(&ftrace_cmd_mutex);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static int ftrace_process_regex(struct ftrace_hash *hash,
|
|
|
|
char *buff, int len, int enable)
|
2009-02-13 23:08:48 +01:00
|
|
|
{
|
2009-02-14 06:40:25 +01:00
|
|
|
char *func, *command, *next = buff;
|
2009-02-17 17:20:26 +01:00
|
|
|
struct ftrace_func_command *p;
|
2011-06-01 13:18:47 +02:00
|
|
|
int ret = -EINVAL;
|
2009-02-13 23:08:48 +01:00
|
|
|
|
|
|
|
func = strsep(&next, ":");
|
|
|
|
|
|
|
|
if (!next) {
|
2011-04-30 02:59:51 +02:00
|
|
|
ret = ftrace_match_records(hash, func, len);
|
2011-04-29 21:12:32 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = -EINVAL;
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
return 0;
|
2009-02-13 23:08:48 +01:00
|
|
|
}
|
|
|
|
|
2009-02-14 06:40:25 +01:00
|
|
|
/* command found */
|
2009-02-13 23:08:48 +01:00
|
|
|
|
|
|
|
command = strsep(&next, ":");
|
|
|
|
|
2009-02-14 06:40:25 +01:00
|
|
|
mutex_lock(&ftrace_cmd_mutex);
|
|
|
|
list_for_each_entry(p, &ftrace_commands, list) {
|
|
|
|
if (strcmp(p->name, command) == 0) {
|
2011-07-07 17:09:22 +02:00
|
|
|
ret = p->func(hash, func, command, next, enable);
|
2009-02-14 06:40:25 +01:00
|
|
|
goto out_unlock;
|
|
|
|
}
|
2009-02-13 23:08:48 +01:00
|
|
|
}
|
2009-02-14 06:40:25 +01:00
|
|
|
out_unlock:
|
|
|
|
mutex_unlock(&ftrace_cmd_mutex);
|
2009-02-13 23:08:48 +01:00
|
|
|
|
2009-02-14 06:40:25 +01:00
|
|
|
return ret;
|
2009-02-13 23:08:48 +01:00
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
static ssize_t
|
2008-05-22 17:46:33 +02:00
|
|
|
ftrace_regex_write(struct file *file, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos, int enable)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
|
|
|
struct ftrace_iterator *iter;
|
2009-09-11 17:29:29 +02:00
|
|
|
struct trace_parser *parser;
|
|
|
|
ssize_t ret, read;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-09-22 07:52:20 +02:00
|
|
|
if (!cnt)
|
2008-05-12 21:20:43 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
struct seq_file *m = file->private_data;
|
|
|
|
iter = m->private;
|
|
|
|
} else
|
|
|
|
iter = file->private_data;
|
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
2013-05-09 07:44:21 +02:00
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
/* iter->hash is a local copy, so we don't need regex_lock */
|
2013-05-09 07:44:17 +02:00
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
parser = &iter->parser;
|
|
|
|
read = trace_get_user(parser, ubuf, cnt, ppos);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-09-22 07:52:20 +02:00
|
|
|
if (read >= 0 && trace_parser_loaded(parser) &&
|
2009-09-11 17:29:29 +02:00
|
|
|
!trace_parser_cont(parser)) {
|
2011-05-02 23:34:47 +02:00
|
|
|
ret = ftrace_process_regex(iter->hash, parser->buffer,
|
2009-09-11 17:29:29 +02:00
|
|
|
parser->idx, enable);
|
2009-12-08 04:15:30 +01:00
|
|
|
trace_parser_clear(parser);
|
2013-05-09 17:35:12 +02:00
|
|
|
if (ret < 0)
|
2013-05-09 07:44:21 +02:00
|
|
|
goto out;
|
2009-08-11 17:29:04 +02:00
|
|
|
}
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
ret = read;
|
2013-05-09 07:44:21 +02:00
|
|
|
out:
|
2008-05-12 21:20:43 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2011-12-19 20:41:25 +01:00
|
|
|
ssize_t
|
2008-05-22 17:46:33 +02:00
|
|
|
ftrace_filter_write(struct file *file, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
|
|
|
return ftrace_regex_write(file, ubuf, cnt, ppos, 1);
|
|
|
|
}
|
|
|
|
|
2011-12-19 20:41:25 +01:00
|
|
|
ssize_t
|
2008-05-22 17:46:33 +02:00
|
|
|
ftrace_notrace_write(struct file *file, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
|
|
|
return ftrace_regex_write(file, ubuf, cnt, ppos, 0);
|
|
|
|
}
|
|
|
|
|
2011-05-02 23:34:47 +02:00
|
|
|
static int
|
2012-06-05 12:28:08 +02:00
|
|
|
ftrace_match_addr(struct ftrace_hash *hash, unsigned long ip, int remove)
|
|
|
|
{
|
|
|
|
struct ftrace_func_entry *entry;
|
|
|
|
|
|
|
|
if (!ftrace_location(ip))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (remove) {
|
|
|
|
entry = ftrace_lookup_ip(hash, ip);
|
|
|
|
if (!entry)
|
|
|
|
return -ENOENT;
|
|
|
|
free_hash_entry(hash, entry);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return add_hash_entry(hash, ip);
|
|
|
|
}
|
|
|
|
|
2014-10-24 20:56:01 +02:00
|
|
|
static void ftrace_ops_update_code(struct ftrace_ops *ops,
|
2015-01-13 20:03:38 +01:00
|
|
|
struct ftrace_ops_hash *old_hash)
|
2013-07-26 02:22:00 +02:00
|
|
|
{
|
2015-01-13 17:20:43 +01:00
|
|
|
struct ftrace_ops *op;
|
|
|
|
|
|
|
|
if (!ftrace_enabled)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_ENABLED) {
|
2014-10-24 20:56:01 +02:00
|
|
|
ftrace_run_modify_code(ops, FTRACE_UPDATE_CALLS, old_hash);
|
2015-01-13 17:20:43 +01:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is the shared global_ops filter, then we need to
|
|
|
|
* check if there is another ops that shares it, is enabled.
|
|
|
|
* If so, we still need to run the modify code.
|
|
|
|
*/
|
|
|
|
if (ops->func_hash != &global_ops.local_hash)
|
|
|
|
return;
|
|
|
|
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
if (op->func_hash == &global_ops.local_hash &&
|
|
|
|
op->flags & FTRACE_OPS_FL_ENABLED) {
|
|
|
|
ftrace_run_modify_code(op, FTRACE_UPDATE_CALLS, old_hash);
|
|
|
|
/* Only need to do this once */
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
} while_for_each_ftrace_op(op);
|
2013-07-26 02:22:00 +02:00
|
|
|
}
|
|
|
|
|
2012-06-05 12:28:08 +02:00
|
|
|
static int
|
|
|
|
ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len,
|
|
|
|
unsigned long ip, int remove, int reset, int enable)
|
2008-05-22 17:46:33 +02:00
|
|
|
{
|
2011-05-02 23:34:47 +02:00
|
|
|
struct ftrace_hash **orig_hash;
|
2015-01-13 20:03:38 +01:00
|
|
|
struct ftrace_ops_hash old_hash_ops;
|
2014-07-24 21:33:41 +02:00
|
|
|
struct ftrace_hash *old_hash;
|
2011-05-02 18:29:25 +02:00
|
|
|
struct ftrace_hash *hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
int ret;
|
2011-05-02 18:29:25 +02:00
|
|
|
|
2008-05-22 17:46:33 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
2011-05-02 23:34:47 +02:00
|
|
|
return -ENODEV;
|
2008-05-22 17:46:33 +02:00
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_lock(&ops->func_hash->regex_lock);
|
2013-05-09 07:44:21 +02:00
|
|
|
|
2011-05-02 18:29:25 +02:00
|
|
|
if (enable)
|
2014-08-15 23:23:02 +02:00
|
|
|
orig_hash = &ops->func_hash->filter_hash;
|
2011-05-02 18:29:25 +02:00
|
|
|
else
|
2014-08-15 23:23:02 +02:00
|
|
|
orig_hash = &ops->func_hash->notrace_hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
|
2014-07-15 02:40:20 +02:00
|
|
|
if (reset)
|
|
|
|
hash = alloc_ftrace_hash(FTRACE_HASH_DEFAULT_BITS);
|
|
|
|
else
|
|
|
|
hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, *orig_hash);
|
|
|
|
|
2013-05-09 07:44:21 +02:00
|
|
|
if (!hash) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out_regex_unlock;
|
|
|
|
}
|
2011-05-02 18:29:25 +02:00
|
|
|
|
2012-01-02 10:04:14 +01:00
|
|
|
if (buf && !ftrace_match_records(hash, buf, len)) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out_regex_unlock;
|
|
|
|
}
|
2012-06-05 12:28:08 +02:00
|
|
|
if (ip) {
|
|
|
|
ret = ftrace_match_addr(hash, ip, remove);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out_regex_unlock;
|
|
|
|
}
|
2011-05-02 23:34:47 +02:00
|
|
|
|
|
|
|
mutex_lock(&ftrace_lock);
|
2014-07-24 21:33:41 +02:00
|
|
|
old_hash = *orig_hash;
|
2015-01-13 20:03:38 +01:00
|
|
|
old_hash_ops.filter_hash = ops->func_hash->filter_hash;
|
|
|
|
old_hash_ops.notrace_hash = ops->func_hash->notrace_hash;
|
2011-07-13 21:03:44 +02:00
|
|
|
ret = ftrace_hash_move(ops, enable, orig_hash, hash);
|
2014-07-24 21:33:41 +02:00
|
|
|
if (!ret) {
|
2015-01-13 20:03:38 +01:00
|
|
|
ftrace_ops_update_code(ops, &old_hash_ops);
|
2014-07-24 21:33:41 +02:00
|
|
|
free_ftrace_hash_rcu(old_hash);
|
|
|
|
}
|
2011-05-02 23:34:47 +02:00
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
|
2012-01-02 10:04:14 +01:00
|
|
|
out_regex_unlock:
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_unlock(&ops->func_hash->regex_lock);
|
2011-05-02 23:34:47 +02:00
|
|
|
|
|
|
|
free_ftrace_hash(hash);
|
|
|
|
return ret;
|
2008-05-22 17:46:33 +02:00
|
|
|
}
|
|
|
|
|
2012-06-05 12:28:08 +02:00
|
|
|
static int
|
|
|
|
ftrace_set_addr(struct ftrace_ops *ops, unsigned long ip, int remove,
|
|
|
|
int reset, int enable)
|
|
|
|
{
|
|
|
|
return ftrace_set_hash(ops, 0, 0, ip, remove, reset, enable);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_set_filter_ip - set a function to filter on in ftrace by address
|
|
|
|
* @ops - the ops to set the filter with
|
|
|
|
* @ip - the address to add to or remove from the filter.
|
|
|
|
* @remove - non zero to remove the ip from the filter
|
|
|
|
* @reset - non zero to reset all filters before applying this filter.
|
|
|
|
*
|
|
|
|
* Filters denote which functions should be enabled when tracing is enabled
|
|
|
|
* If @ip is NULL, it failes to update filter.
|
|
|
|
*/
|
|
|
|
int ftrace_set_filter_ip(struct ftrace_ops *ops, unsigned long ip,
|
|
|
|
int remove, int reset)
|
|
|
|
{
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
2012-06-05 12:28:08 +02:00
|
|
|
return ftrace_set_addr(ops, ip, remove, reset, 1);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(ftrace_set_filter_ip);
|
|
|
|
|
|
|
|
static int
|
|
|
|
ftrace_set_regex(struct ftrace_ops *ops, unsigned char *buf, int len,
|
|
|
|
int reset, int enable)
|
|
|
|
{
|
|
|
|
return ftrace_set_hash(ops, buf, len, 0, 0, reset, enable);
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:45 +02:00
|
|
|
/**
|
|
|
|
* ftrace_set_filter - set a function to filter on in ftrace
|
2011-05-06 04:54:01 +02:00
|
|
|
* @ops - the ops to set the filter with
|
|
|
|
* @buf - the string that holds the function filter text.
|
|
|
|
* @len - the length of the string.
|
|
|
|
* @reset - non zero to reset all filters before applying this filter.
|
|
|
|
*
|
|
|
|
* Filters denote which functions should be enabled when tracing is enabled.
|
|
|
|
* If @buf is NULL and reset is set, all functions will be enabled for tracing.
|
|
|
|
*/
|
2012-01-02 10:04:14 +01:00
|
|
|
int ftrace_set_filter(struct ftrace_ops *ops, unsigned char *buf,
|
2011-05-06 04:54:01 +02:00
|
|
|
int len, int reset)
|
|
|
|
{
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
2012-01-02 10:04:14 +01:00
|
|
|
return ftrace_set_regex(ops, buf, len, reset, 1);
|
2011-05-06 04:54:01 +02:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(ftrace_set_filter);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ftrace_set_notrace - set a function to not trace in ftrace
|
|
|
|
* @ops - the ops to set the notrace filter with
|
|
|
|
* @buf - the string that holds the function notrace text.
|
|
|
|
* @len - the length of the string.
|
|
|
|
* @reset - non zero to reset all filters before applying this filter.
|
|
|
|
*
|
|
|
|
* Notrace Filters denote which functions should not be enabled when tracing
|
|
|
|
* is enabled. If @buf is NULL and reset is set, all functions will be enabled
|
|
|
|
* for tracing.
|
|
|
|
*/
|
2012-01-02 10:04:14 +01:00
|
|
|
int ftrace_set_notrace(struct ftrace_ops *ops, unsigned char *buf,
|
2011-05-06 04:54:01 +02:00
|
|
|
int len, int reset)
|
|
|
|
{
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
2012-01-02 10:04:14 +01:00
|
|
|
return ftrace_set_regex(ops, buf, len, reset, 0);
|
2011-05-06 04:54:01 +02:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(ftrace_set_notrace);
|
|
|
|
/**
|
2014-04-20 17:10:44 +02:00
|
|
|
* ftrace_set_global_filter - set a function to filter on with global tracers
|
2008-05-12 21:20:45 +02:00
|
|
|
* @buf - the string that holds the function filter text.
|
|
|
|
* @len - the length of the string.
|
|
|
|
* @reset - non zero to reset all filters before applying this filter.
|
|
|
|
*
|
|
|
|
* Filters denote which functions should be enabled when tracing is enabled.
|
|
|
|
* If @buf is NULL and reset is set, all functions will be enabled for tracing.
|
|
|
|
*/
|
2011-05-06 04:54:01 +02:00
|
|
|
void ftrace_set_global_filter(unsigned char *buf, int len, int reset)
|
2008-05-12 21:20:45 +02:00
|
|
|
{
|
2011-05-02 18:29:25 +02:00
|
|
|
ftrace_set_regex(&global_ops, buf, len, reset, 1);
|
2008-05-22 17:46:33 +02:00
|
|
|
}
|
2011-05-06 04:54:01 +02:00
|
|
|
EXPORT_SYMBOL_GPL(ftrace_set_global_filter);
|
2008-05-12 21:20:48 +02:00
|
|
|
|
2008-05-22 17:46:33 +02:00
|
|
|
/**
|
2014-04-20 17:10:44 +02:00
|
|
|
* ftrace_set_global_notrace - set a function to not trace with global tracers
|
2008-05-22 17:46:33 +02:00
|
|
|
* @buf - the string that holds the function notrace text.
|
|
|
|
* @len - the length of the string.
|
|
|
|
* @reset - non zero to reset all filters before applying this filter.
|
|
|
|
*
|
|
|
|
* Notrace Filters denote which functions should not be enabled when tracing
|
|
|
|
* is enabled. If @buf is NULL and reset is set, all functions will be enabled
|
|
|
|
* for tracing.
|
|
|
|
*/
|
2011-05-06 04:54:01 +02:00
|
|
|
void ftrace_set_global_notrace(unsigned char *buf, int len, int reset)
|
2008-05-22 17:46:33 +02:00
|
|
|
{
|
2011-05-02 18:29:25 +02:00
|
|
|
ftrace_set_regex(&global_ops, buf, len, reset, 0);
|
2008-05-12 21:20:45 +02:00
|
|
|
}
|
2011-05-06 04:54:01 +02:00
|
|
|
EXPORT_SYMBOL_GPL(ftrace_set_global_notrace);
|
2008-05-12 21:20:45 +02:00
|
|
|
|
2009-05-28 19:37:24 +02:00
|
|
|
/*
|
|
|
|
* command line interface to allow users to set filters on boot up.
|
|
|
|
*/
|
|
|
|
#define FTRACE_FILTER_SIZE COMMAND_LINE_SIZE
|
|
|
|
static char ftrace_notrace_buf[FTRACE_FILTER_SIZE] __initdata;
|
|
|
|
static char ftrace_filter_buf[FTRACE_FILTER_SIZE] __initdata;
|
|
|
|
|
2013-06-28 04:18:06 +02:00
|
|
|
/* Used by function selftest to not test if filter is set */
|
|
|
|
bool ftrace_filter_param __initdata;
|
|
|
|
|
2009-05-28 19:37:24 +02:00
|
|
|
static int __init set_ftrace_notrace(char *str)
|
|
|
|
{
|
2013-06-28 04:18:06 +02:00
|
|
|
ftrace_filter_param = true;
|
2013-04-08 06:12:39 +02:00
|
|
|
strlcpy(ftrace_notrace_buf, str, FTRACE_FILTER_SIZE);
|
2009-05-28 19:37:24 +02:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("ftrace_notrace=", set_ftrace_notrace);
|
|
|
|
|
|
|
|
static int __init set_ftrace_filter(char *str)
|
|
|
|
{
|
2013-06-28 04:18:06 +02:00
|
|
|
ftrace_filter_param = true;
|
2013-04-08 06:12:39 +02:00
|
|
|
strlcpy(ftrace_filter_buf, str, FTRACE_FILTER_SIZE);
|
2009-05-28 19:37:24 +02:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("ftrace_filter=", set_ftrace_filter);
|
|
|
|
|
2009-10-12 22:17:21 +02:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2009-11-05 04:16:17 +01:00
|
|
|
static char ftrace_graph_buf[FTRACE_FILTER_SIZE] __initdata;
|
2014-06-12 18:23:50 +02:00
|
|
|
static char ftrace_graph_notrace_buf[FTRACE_FILTER_SIZE] __initdata;
|
2013-10-14 10:24:24 +02:00
|
|
|
static int ftrace_set_func(unsigned long *array, int *idx, int size, char *buffer);
|
2010-03-06 02:02:19 +01:00
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
static unsigned long save_global_trampoline;
|
|
|
|
static unsigned long save_global_flags;
|
|
|
|
|
2009-10-12 22:17:21 +02:00
|
|
|
static int __init set_graph_function(char *str)
|
|
|
|
{
|
2009-10-14 20:43:39 +02:00
|
|
|
strlcpy(ftrace_graph_buf, str, FTRACE_FILTER_SIZE);
|
2009-10-12 22:17:21 +02:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("ftrace_graph_filter=", set_graph_function);
|
|
|
|
|
2014-06-12 18:23:50 +02:00
|
|
|
static int __init set_graph_notrace_function(char *str)
|
|
|
|
{
|
|
|
|
strlcpy(ftrace_graph_notrace_buf, str, FTRACE_FILTER_SIZE);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("ftrace_graph_notrace=", set_graph_notrace_function);
|
|
|
|
|
|
|
|
static void __init set_ftrace_early_graph(char *buf, int enable)
|
2009-10-12 22:17:21 +02:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
char *func;
|
2014-06-12 18:23:50 +02:00
|
|
|
unsigned long *table = ftrace_graph_funcs;
|
|
|
|
int *count = &ftrace_graph_count;
|
|
|
|
|
|
|
|
if (!enable) {
|
|
|
|
table = ftrace_graph_notrace_funcs;
|
|
|
|
count = &ftrace_graph_notrace_count;
|
|
|
|
}
|
2009-10-12 22:17:21 +02:00
|
|
|
|
|
|
|
while (buf) {
|
|
|
|
func = strsep(&buf, ",");
|
|
|
|
/* we allow only one expression at a time */
|
2014-06-12 18:23:50 +02:00
|
|
|
ret = ftrace_set_func(table, count, FTRACE_GRAPH_MAX_FUNCS, func);
|
2009-10-12 22:17:21 +02:00
|
|
|
if (ret)
|
|
|
|
printk(KERN_DEBUG "ftrace: function %s not "
|
|
|
|
"traceable\n", func);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
|
|
|
|
|
2011-12-20 03:57:44 +01:00
|
|
|
void __init
|
|
|
|
ftrace_set_early_filter(struct ftrace_ops *ops, char *buf, int enable)
|
2009-05-28 19:37:24 +02:00
|
|
|
{
|
|
|
|
char *func;
|
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
|
|
|
|
2009-05-28 19:37:24 +02:00
|
|
|
while (buf) {
|
|
|
|
func = strsep(&buf, ",");
|
2011-05-02 18:29:25 +02:00
|
|
|
ftrace_set_regex(ops, func, strlen(func), 0, enable);
|
2009-05-28 19:37:24 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __init set_ftrace_early_filters(void)
|
|
|
|
{
|
|
|
|
if (ftrace_filter_buf[0])
|
2011-12-20 03:57:44 +01:00
|
|
|
ftrace_set_early_filter(&global_ops, ftrace_filter_buf, 1);
|
2009-05-28 19:37:24 +02:00
|
|
|
if (ftrace_notrace_buf[0])
|
2011-12-20 03:57:44 +01:00
|
|
|
ftrace_set_early_filter(&global_ops, ftrace_notrace_buf, 0);
|
2009-10-12 22:17:21 +02:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
if (ftrace_graph_buf[0])
|
2014-06-12 18:23:50 +02:00
|
|
|
set_ftrace_early_graph(ftrace_graph_buf, 1);
|
|
|
|
if (ftrace_graph_notrace_buf[0])
|
|
|
|
set_ftrace_early_graph(ftrace_graph_notrace_buf, 0);
|
2009-10-12 22:17:21 +02:00
|
|
|
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
|
2009-05-28 19:37:24 +02:00
|
|
|
}
|
|
|
|
|
2011-12-19 20:41:25 +01:00
|
|
|
int ftrace_regex_release(struct inode *inode, struct file *file)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
|
|
|
struct seq_file *m = (struct seq_file *)file->private_data;
|
2015-01-13 20:03:38 +01:00
|
|
|
struct ftrace_ops_hash old_hash_ops;
|
2008-05-12 21:20:43 +02:00
|
|
|
struct ftrace_iterator *iter;
|
2011-05-02 23:34:47 +02:00
|
|
|
struct ftrace_hash **orig_hash;
|
2014-07-24 21:33:41 +02:00
|
|
|
struct ftrace_hash *old_hash;
|
2009-09-11 17:29:29 +02:00
|
|
|
struct trace_parser *parser;
|
2011-05-03 19:25:24 +02:00
|
|
|
int filter_hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
int ret;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
iter = m->private;
|
|
|
|
seq_release(inode, file);
|
|
|
|
} else
|
|
|
|
iter = file->private_data;
|
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
parser = &iter->parser;
|
|
|
|
if (trace_parser_loaded(parser)) {
|
|
|
|
parser->buffer[parser->idx] = 0;
|
2011-04-30 02:59:51 +02:00
|
|
|
ftrace_match_records(iter->hash, parser->buffer, parser->idx);
|
2008-05-12 21:20:43 +02:00
|
|
|
}
|
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
trace_parser_put(parser);
|
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_lock(&iter->ops->func_hash->regex_lock);
|
2013-05-09 07:44:21 +02:00
|
|
|
|
2011-04-30 04:35:33 +02:00
|
|
|
if (file->f_mode & FMODE_WRITE) {
|
2011-05-03 19:25:24 +02:00
|
|
|
filter_hash = !!(iter->flags & FTRACE_ITER_FILTER);
|
|
|
|
|
|
|
|
if (filter_hash)
|
2014-08-15 23:23:02 +02:00
|
|
|
orig_hash = &iter->ops->func_hash->filter_hash;
|
2011-05-03 19:25:24 +02:00
|
|
|
else
|
2014-08-15 23:23:02 +02:00
|
|
|
orig_hash = &iter->ops->func_hash->notrace_hash;
|
2011-05-02 23:34:47 +02:00
|
|
|
|
2011-04-30 04:35:33 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
2014-07-24 21:33:41 +02:00
|
|
|
old_hash = *orig_hash;
|
2015-01-13 20:03:38 +01:00
|
|
|
old_hash_ops.filter_hash = iter->ops->func_hash->filter_hash;
|
|
|
|
old_hash_ops.notrace_hash = iter->ops->func_hash->notrace_hash;
|
2011-07-13 21:03:44 +02:00
|
|
|
ret = ftrace_hash_move(iter->ops, filter_hash,
|
|
|
|
orig_hash, iter->hash);
|
2014-07-24 21:33:41 +02:00
|
|
|
if (!ret) {
|
2015-01-13 20:03:38 +01:00
|
|
|
ftrace_ops_update_code(iter->ops, &old_hash_ops);
|
2014-07-24 21:33:41 +02:00
|
|
|
free_ftrace_hash_rcu(old_hash);
|
|
|
|
}
|
2011-04-30 04:35:33 +02:00
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
}
|
2013-05-09 07:44:21 +02:00
|
|
|
|
2014-08-15 23:23:02 +02:00
|
|
|
mutex_unlock(&iter->ops->func_hash->regex_lock);
|
2011-05-02 23:34:47 +02:00
|
|
|
free_ftrace_hash(iter->hash);
|
|
|
|
kfree(iter);
|
2011-04-30 04:35:33 +02:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-03-06 03:44:55 +01:00
|
|
|
static const struct file_operations ftrace_avail_fops = {
|
2008-05-12 21:20:43 +02:00
|
|
|
.open = ftrace_avail_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
2009-08-17 10:54:03 +02:00
|
|
|
.release = seq_release_private,
|
2008-05-12 21:20:43 +02:00
|
|
|
};
|
|
|
|
|
2011-05-03 20:39:21 +02:00
|
|
|
static const struct file_operations ftrace_enabled_fops = {
|
|
|
|
.open = ftrace_enabled_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
|
|
|
.release = seq_release_private,
|
|
|
|
};
|
|
|
|
|
2009-03-06 03:44:55 +01:00
|
|
|
static const struct file_operations ftrace_filter_fops = {
|
2008-05-12 21:20:43 +02:00
|
|
|
.open = ftrace_filter_open,
|
2009-03-13 10:47:23 +01:00
|
|
|
.read = seq_read,
|
2008-05-12 21:20:43 +02:00
|
|
|
.write = ftrace_filter_write,
|
2013-12-21 23:39:40 +01:00
|
|
|
.llseek = tracing_lseek,
|
2011-04-30 02:59:51 +02:00
|
|
|
.release = ftrace_regex_release,
|
2008-05-12 21:20:43 +02:00
|
|
|
};
|
|
|
|
|
2009-03-06 03:44:55 +01:00
|
|
|
static const struct file_operations ftrace_notrace_fops = {
|
2008-05-22 17:46:33 +02:00
|
|
|
.open = ftrace_notrace_open,
|
2009-03-13 10:47:23 +01:00
|
|
|
.read = seq_read,
|
2008-05-22 17:46:33 +02:00
|
|
|
.write = ftrace_notrace_write,
|
2013-12-21 23:39:40 +01:00
|
|
|
.llseek = tracing_lseek,
|
2011-04-30 02:59:51 +02:00
|
|
|
.release = ftrace_regex_release,
|
2008-05-22 17:46:33 +02:00
|
|
|
};
|
|
|
|
|
2008-12-03 21:36:57 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
|
|
|
|
|
|
|
static DEFINE_MUTEX(graph_lock);
|
|
|
|
|
|
|
|
int ftrace_graph_count;
|
2013-10-14 10:24:26 +02:00
|
|
|
int ftrace_graph_notrace_count;
|
2008-12-03 21:36:57 +01:00
|
|
|
unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly;
|
2013-10-14 10:24:26 +02:00
|
|
|
unsigned long ftrace_graph_notrace_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly;
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2013-10-14 10:24:24 +02:00
|
|
|
struct ftrace_graph_data {
|
|
|
|
unsigned long *table;
|
|
|
|
size_t size;
|
|
|
|
int *count;
|
|
|
|
const struct seq_operations *seq_ops;
|
|
|
|
};
|
|
|
|
|
2008-12-03 21:36:57 +01:00
|
|
|
static void *
|
2009-06-24 03:54:00 +02:00
|
|
|
__g_next(struct seq_file *m, loff_t *pos)
|
2008-12-03 21:36:57 +01:00
|
|
|
{
|
2013-10-14 10:24:24 +02:00
|
|
|
struct ftrace_graph_data *fgd = m->private;
|
|
|
|
|
|
|
|
if (*pos >= *fgd->count)
|
2008-12-03 21:36:57 +01:00
|
|
|
return NULL;
|
2013-10-14 10:24:24 +02:00
|
|
|
return &fgd->table[*pos];
|
2009-06-24 03:54:00 +02:00
|
|
|
}
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2009-06-24 03:54:00 +02:00
|
|
|
static void *
|
|
|
|
g_next(struct seq_file *m, void *v, loff_t *pos)
|
|
|
|
{
|
|
|
|
(*pos)++;
|
|
|
|
return __g_next(m, pos);
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void *g_start(struct seq_file *m, loff_t *pos)
|
|
|
|
{
|
2013-10-14 10:24:24 +02:00
|
|
|
struct ftrace_graph_data *fgd = m->private;
|
|
|
|
|
2008-12-03 21:36:57 +01:00
|
|
|
mutex_lock(&graph_lock);
|
|
|
|
|
2009-02-19 21:13:12 +01:00
|
|
|
/* Nothing, tell g_show to print all functions are enabled */
|
2013-10-14 10:24:24 +02:00
|
|
|
if (!*fgd->count && !*pos)
|
2009-02-19 21:13:12 +01:00
|
|
|
return (void *)1;
|
|
|
|
|
2009-06-24 03:54:00 +02:00
|
|
|
return __g_next(m, pos);
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void g_stop(struct seq_file *m, void *p)
|
|
|
|
{
|
|
|
|
mutex_unlock(&graph_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int g_show(struct seq_file *m, void *v)
|
|
|
|
{
|
|
|
|
unsigned long *ptr = v;
|
|
|
|
|
|
|
|
if (!ptr)
|
|
|
|
return 0;
|
|
|
|
|
2009-02-19 21:13:12 +01:00
|
|
|
if (ptr == (unsigned long *)1) {
|
2014-06-12 18:23:51 +02:00
|
|
|
struct ftrace_graph_data *fgd = m->private;
|
|
|
|
|
|
|
|
if (fgd->table == ftrace_graph_funcs)
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "#### all functions enabled ####\n");
|
2014-06-12 18:23:51 +02:00
|
|
|
else
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "#### no functions disabled ####\n");
|
2009-02-19 21:13:12 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-17 06:05:58 +02:00
|
|
|
seq_printf(m, "%ps\n", (void *)*ptr);
|
2008-12-03 21:36:57 +01:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-23 01:43:43 +02:00
|
|
|
static const struct seq_operations ftrace_graph_seq_ops = {
|
2008-12-03 21:36:57 +01:00
|
|
|
.start = g_start,
|
|
|
|
.next = g_next,
|
|
|
|
.stop = g_stop,
|
|
|
|
.show = g_show,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int
|
2013-10-14 10:24:24 +02:00
|
|
|
__ftrace_graph_open(struct inode *inode, struct file *file,
|
|
|
|
struct ftrace_graph_data *fgd)
|
2008-12-03 21:36:57 +01:00
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
mutex_lock(&graph_lock);
|
|
|
|
if ((file->f_mode & FMODE_WRITE) &&
|
2009-07-23 05:29:30 +02:00
|
|
|
(file->f_flags & O_TRUNC)) {
|
2013-10-14 10:24:24 +02:00
|
|
|
*fgd->count = 0;
|
|
|
|
memset(fgd->table, 0, fgd->size * sizeof(*fgd->table));
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
2009-09-18 08:06:28 +02:00
|
|
|
mutex_unlock(&graph_lock);
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2013-10-14 10:24:24 +02:00
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
ret = seq_open(file, fgd->seq_ops);
|
|
|
|
if (!ret) {
|
|
|
|
struct seq_file *m = file->private_data;
|
|
|
|
m->private = fgd;
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
file->private_data = fgd;
|
2008-12-03 21:36:57 +01:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-10-14 10:24:24 +02:00
|
|
|
static int
|
|
|
|
ftrace_graph_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct ftrace_graph_data *fgd;
|
|
|
|
|
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
fgd = kmalloc(sizeof(*fgd), GFP_KERNEL);
|
|
|
|
if (fgd == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
fgd->table = ftrace_graph_funcs;
|
|
|
|
fgd->size = FTRACE_GRAPH_MAX_FUNCS;
|
|
|
|
fgd->count = &ftrace_graph_count;
|
|
|
|
fgd->seq_ops = &ftrace_graph_seq_ops;
|
|
|
|
|
|
|
|
return __ftrace_graph_open(inode, file, fgd);
|
|
|
|
}
|
|
|
|
|
2013-10-14 10:24:26 +02:00
|
|
|
static int
|
|
|
|
ftrace_graph_notrace_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
struct ftrace_graph_data *fgd;
|
|
|
|
|
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
fgd = kmalloc(sizeof(*fgd), GFP_KERNEL);
|
|
|
|
if (fgd == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
fgd->table = ftrace_graph_notrace_funcs;
|
|
|
|
fgd->size = FTRACE_GRAPH_MAX_FUNCS;
|
|
|
|
fgd->count = &ftrace_graph_notrace_count;
|
|
|
|
fgd->seq_ops = &ftrace_graph_seq_ops;
|
|
|
|
|
|
|
|
return __ftrace_graph_open(inode, file, fgd);
|
|
|
|
}
|
|
|
|
|
2009-07-23 05:29:11 +02:00
|
|
|
static int
|
|
|
|
ftrace_graph_release(struct inode *inode, struct file *file)
|
|
|
|
{
|
2013-10-14 10:24:24 +02:00
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
struct seq_file *m = file->private_data;
|
|
|
|
|
|
|
|
kfree(m->private);
|
2009-07-23 05:29:11 +02:00
|
|
|
seq_release(inode, file);
|
2013-10-14 10:24:24 +02:00
|
|
|
} else {
|
|
|
|
kfree(file->private_data);
|
|
|
|
}
|
|
|
|
|
2009-07-23 05:29:11 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-12-03 21:36:57 +01:00
|
|
|
static int
|
2013-10-14 10:24:24 +02:00
|
|
|
ftrace_set_func(unsigned long *array, int *idx, int size, char *buffer)
|
2008-12-03 21:36:57 +01:00
|
|
|
{
|
2015-09-29 18:46:14 +02:00
|
|
|
struct ftrace_glob func_g;
|
2008-12-03 21:36:57 +01:00
|
|
|
struct dyn_ftrace *rec;
|
|
|
|
struct ftrace_page *pg;
|
2010-02-10 08:43:04 +01:00
|
|
|
int fail = 1;
|
2015-09-29 18:46:14 +02:00
|
|
|
int not;
|
2009-02-19 21:13:12 +01:00
|
|
|
bool exists;
|
|
|
|
int i;
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2009-02-19 21:13:12 +01:00
|
|
|
/* decode regex */
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.type = filter_parse_regex(buffer, strlen(buffer),
|
|
|
|
&func_g.search, ¬);
|
2013-10-14 10:24:24 +02:00
|
|
|
if (!not && *idx >= size)
|
2010-02-10 08:43:04 +01:00
|
|
|
return -EBUSY;
|
2009-02-19 21:13:12 +01:00
|
|
|
|
2015-09-29 18:46:14 +02:00
|
|
|
func_g.len = strlen(func_g.search);
|
2009-02-19 21:13:12 +01:00
|
|
|
|
2009-02-14 07:15:39 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2011-04-22 05:16:46 +02:00
|
|
|
|
|
|
|
if (unlikely(ftrace_disabled)) {
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
return -ENODEV;
|
|
|
|
}
|
|
|
|
|
2009-02-13 18:43:56 +01:00
|
|
|
do_for_each_ftrace_rec(pg, rec) {
|
|
|
|
|
2015-09-29 18:46:15 +02:00
|
|
|
if (ftrace_match_record(rec, &func_g, NULL, 0)) {
|
2010-02-10 08:43:04 +01:00
|
|
|
/* if it is in the array */
|
2009-02-19 21:13:12 +01:00
|
|
|
exists = false;
|
2010-02-10 08:43:04 +01:00
|
|
|
for (i = 0; i < *idx; i++) {
|
2009-02-19 21:13:12 +01:00
|
|
|
if (array[i] == rec->ip) {
|
|
|
|
exists = true;
|
2009-02-13 18:43:56 +01:00
|
|
|
break;
|
|
|
|
}
|
2010-02-10 08:43:04 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!not) {
|
|
|
|
fail = 0;
|
|
|
|
if (!exists) {
|
|
|
|
array[(*idx)++] = rec->ip;
|
2013-10-14 10:24:24 +02:00
|
|
|
if (*idx >= size)
|
2010-02-10 08:43:04 +01:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (exists) {
|
|
|
|
array[i] = array[--(*idx)];
|
|
|
|
array[*idx] = 0;
|
|
|
|
fail = 0;
|
|
|
|
}
|
|
|
|
}
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
2009-02-13 18:43:56 +01:00
|
|
|
} while_for_each_ftrace_rec();
|
2010-02-10 08:43:04 +01:00
|
|
|
out:
|
2009-02-14 07:15:39 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2010-02-10 08:43:04 +01:00
|
|
|
if (fail)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
return 0;
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
ftrace_graph_write(struct file *file, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
2009-09-11 17:29:29 +02:00
|
|
|
struct trace_parser parser;
|
2013-10-14 10:24:25 +02:00
|
|
|
ssize_t read, ret = 0;
|
2013-10-14 10:24:24 +02:00
|
|
|
struct ftrace_graph_data *fgd = file->private_data;
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2010-02-10 08:43:04 +01:00
|
|
|
if (!cnt)
|
2008-12-03 21:36:57 +01:00
|
|
|
return 0;
|
|
|
|
|
2013-10-14 10:24:25 +02:00
|
|
|
if (trace_parser_get_init(&parser, FTRACE_BUFF_MAX))
|
|
|
|
return -ENOMEM;
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
read = trace_get_user(&parser, ubuf, cnt, ppos);
|
2008-12-03 21:36:57 +01:00
|
|
|
|
2009-09-22 07:52:20 +02:00
|
|
|
if (read >= 0 && trace_parser_loaded((&parser))) {
|
2009-09-11 17:29:29 +02:00
|
|
|
parser.buffer[parser.idx] = 0;
|
|
|
|
|
2013-10-14 10:24:25 +02:00
|
|
|
mutex_lock(&graph_lock);
|
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
/* we allow only one expression at a time */
|
2013-10-14 10:24:24 +02:00
|
|
|
ret = ftrace_set_func(fgd->table, fgd->count, fgd->size,
|
|
|
|
parser.buffer);
|
2013-10-14 10:24:25 +02:00
|
|
|
|
|
|
|
mutex_unlock(&graph_lock);
|
2008-12-03 21:36:57 +01:00
|
|
|
}
|
|
|
|
|
2013-10-14 10:24:25 +02:00
|
|
|
if (!ret)
|
|
|
|
ret = read;
|
2009-09-22 07:52:57 +02:00
|
|
|
|
2009-09-11 17:29:29 +02:00
|
|
|
trace_parser_put(&parser);
|
2008-12-03 21:36:57 +01:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct file_operations ftrace_graph_fops = {
|
2009-07-23 05:29:11 +02:00
|
|
|
.open = ftrace_graph_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.write = ftrace_graph_write,
|
2013-12-21 23:39:40 +01:00
|
|
|
.llseek = tracing_lseek,
|
2009-07-23 05:29:11 +02:00
|
|
|
.release = ftrace_graph_release,
|
2008-12-03 21:36:57 +01:00
|
|
|
};
|
2013-10-14 10:24:26 +02:00
|
|
|
|
|
|
|
static const struct file_operations ftrace_graph_notrace_fops = {
|
|
|
|
.open = ftrace_graph_notrace_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.write = ftrace_graph_write,
|
2013-12-21 23:39:40 +01:00
|
|
|
.llseek = tracing_lseek,
|
2013-10-14 10:24:26 +02:00
|
|
|
.release = ftrace_graph_release,
|
|
|
|
};
|
2008-12-03 21:36:57 +01:00
|
|
|
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
|
|
|
|
|
2014-01-10 22:17:45 +01:00
|
|
|
void ftrace_create_filter_files(struct ftrace_ops *ops,
|
|
|
|
struct dentry *parent)
|
|
|
|
{
|
|
|
|
|
|
|
|
trace_create_file("set_ftrace_filter", 0644, parent,
|
|
|
|
ops, &ftrace_filter_fops);
|
|
|
|
|
|
|
|
trace_create_file("set_ftrace_notrace", 0644, parent,
|
|
|
|
ops, &ftrace_notrace_fops);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The name "destroy_filter_files" is really a misnomer. Although
|
|
|
|
* in the future, it may actualy delete the files, but this is
|
|
|
|
* really intended to make sure the ops passed in are disabled
|
|
|
|
* and that when this function returns, the caller is free to
|
|
|
|
* free the ops.
|
|
|
|
*
|
|
|
|
* The "destroy" name is only to match the "create" name that this
|
|
|
|
* should be paired with.
|
|
|
|
*/
|
|
|
|
void ftrace_destroy_filter_files(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_ENABLED)
|
|
|
|
ftrace_shutdown(ops, 0);
|
|
|
|
ops->flags |= FTRACE_OPS_FL_DELETED;
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
}
|
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
static __init int ftrace_init_dyn_tracefs(struct dentry *d_tracer)
|
2008-05-12 21:20:43 +02:00
|
|
|
{
|
|
|
|
|
2009-03-27 00:25:38 +01:00
|
|
|
trace_create_file("available_filter_functions", 0444,
|
|
|
|
d_tracer, NULL, &ftrace_avail_fops);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-05-03 20:39:21 +02:00
|
|
|
trace_create_file("enabled_functions", 0444,
|
|
|
|
d_tracer, NULL, &ftrace_enabled_fops);
|
|
|
|
|
2014-01-10 22:17:45 +01:00
|
|
|
ftrace_create_filter_files(&global_ops, d_tracer);
|
ftrace: user update and disable dynamic ftrace daemon
In dynamic ftrace, the mcount function starts off pointing to a stub
function that just returns.
On start up, the call to the stub is modified to point to a "record_ip"
function. The job of the record_ip function is to add the function to
a pre-allocated hash list. If the function is already there, it simply is
ignored, otherwise it is added to the list.
Later, a ftraced daemon wakes up and calls kstop_machine if any functions
have been recorded, and changes the calls to the recorded functions to
a simple nop. If no functions were recorded, the daemon goes back to sleep.
The daemon wakes up once a second to see if it needs to update any newly
recorded functions into nops. Usually it does not, but if a lot of code
has been executed for the first time in the kernel, the ftraced daemon
will call kstop_machine to update those into nops.
The problem currently is that there's no way to stop the daemon from doing
this, and it can cause unneeded latencies (800us which for some is bothersome).
This patch adds a new file /debugfs/tracing/ftraced_enabled. If the daemon
is active, reading this will return "enabled\n" and "disabled\n" when the
daemon is not running. To disable the daemon, the user can echo "0" or
"disable" into this file, and "1" or "enable" to re-enable the daemon.
Since the daemon is used to convert the functions into nops to increase
the performance of the system, I also added that anytime something is
written into the ftraced_enabled file, kstop_machine will run if there
are new functions that have been detected that need to be converted.
This way the user can disable the daemon but still be able to control the
conversion of the mcount calls to nops by simply,
"echo 0 > /debugfs/tracing/ftraced_enabled"
when they need to do more conversions.
To see the number of converted functions:
"cat /debugfs/tracing/dyn_ftrace_total_info"
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-05-28 02:48:37 +02:00
|
|
|
|
2008-12-03 21:36:57 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2009-03-27 00:25:38 +01:00
|
|
|
trace_create_file("set_graph_function", 0444, d_tracer,
|
2008-12-03 21:36:57 +01:00
|
|
|
NULL,
|
|
|
|
&ftrace_graph_fops);
|
2013-10-14 10:24:26 +02:00
|
|
|
trace_create_file("set_graph_notrace", 0444, d_tracer,
|
|
|
|
NULL,
|
|
|
|
&ftrace_graph_notrace_fops);
|
2008-12-03 21:36:57 +01:00
|
|
|
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-04-25 04:32:06 +02:00
|
|
|
static int ftrace_cmp_ips(const void *a, const void *b)
|
2011-12-16 23:06:45 +01:00
|
|
|
{
|
2012-04-25 04:32:06 +02:00
|
|
|
const unsigned long *ipa = a;
|
|
|
|
const unsigned long *ipb = b;
|
2011-12-16 23:06:45 +01:00
|
|
|
|
2012-04-25 04:32:06 +02:00
|
|
|
if (*ipa > *ipb)
|
|
|
|
return 1;
|
|
|
|
if (*ipa < *ipb)
|
|
|
|
return -1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-10-13 22:33:53 +02:00
|
|
|
static int ftrace_process_locs(struct module *mod,
|
2008-11-15 01:21:19 +01:00
|
|
|
unsigned long *start,
|
2008-08-14 21:45:08 +02:00
|
|
|
unsigned long *end)
|
|
|
|
{
|
2012-04-25 05:45:26 +02:00
|
|
|
struct ftrace_page *start_pg;
|
2011-12-16 22:23:44 +01:00
|
|
|
struct ftrace_page *pg;
|
2012-04-25 05:45:26 +02:00
|
|
|
struct dyn_ftrace *rec;
|
2011-12-16 22:23:44 +01:00
|
|
|
unsigned long count;
|
2008-08-14 21:45:08 +02:00
|
|
|
unsigned long *p;
|
|
|
|
unsigned long addr;
|
2011-06-25 05:28:13 +02:00
|
|
|
unsigned long flags = 0; /* Shut up gcc */
|
2011-12-16 22:23:44 +01:00
|
|
|
int ret = -ENOMEM;
|
|
|
|
|
|
|
|
count = end - start;
|
|
|
|
|
|
|
|
if (!count)
|
|
|
|
return 0;
|
|
|
|
|
2012-04-25 04:32:06 +02:00
|
|
|
sort(start, count, sizeof(*start),
|
2015-09-09 23:27:02 +02:00
|
|
|
ftrace_cmp_ips, NULL);
|
2012-04-25 04:32:06 +02:00
|
|
|
|
2012-04-25 05:45:26 +02:00
|
|
|
start_pg = ftrace_allocate_pages(count);
|
|
|
|
if (!start_pg)
|
2011-12-16 22:23:44 +01:00
|
|
|
return -ENOMEM;
|
2008-08-14 21:45:08 +02:00
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2011-12-16 22:23:44 +01:00
|
|
|
|
2011-12-16 20:42:37 +01:00
|
|
|
/*
|
|
|
|
* Core and each module needs their own pages, as
|
|
|
|
* modules will free them when they are removed.
|
|
|
|
* Force a new page to be allocated for modules.
|
|
|
|
*/
|
2011-12-16 22:23:44 +01:00
|
|
|
if (!mod) {
|
|
|
|
WARN_ON(ftrace_pages || ftrace_pages_start);
|
|
|
|
/* First initialization */
|
2012-04-25 05:45:26 +02:00
|
|
|
ftrace_pages = ftrace_pages_start = start_pg;
|
2011-12-16 22:23:44 +01:00
|
|
|
} else {
|
2011-12-16 20:42:37 +01:00
|
|
|
if (!ftrace_pages)
|
2011-12-16 22:23:44 +01:00
|
|
|
goto out;
|
2011-12-16 20:42:37 +01:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
if (WARN_ON(ftrace_pages->next)) {
|
|
|
|
/* Hmm, we have free pages? */
|
|
|
|
while (ftrace_pages->next)
|
|
|
|
ftrace_pages = ftrace_pages->next;
|
2011-12-16 20:42:37 +01:00
|
|
|
}
|
2011-12-16 22:23:44 +01:00
|
|
|
|
2012-04-25 05:45:26 +02:00
|
|
|
ftrace_pages->next = start_pg;
|
2011-12-16 20:42:37 +01:00
|
|
|
}
|
|
|
|
|
2008-08-14 21:45:08 +02:00
|
|
|
p = start;
|
2012-04-25 05:45:26 +02:00
|
|
|
pg = start_pg;
|
2008-08-14 21:45:08 +02:00
|
|
|
while (p < end) {
|
|
|
|
addr = ftrace_call_adjust(*p++);
|
2008-11-15 01:21:19 +01:00
|
|
|
/*
|
|
|
|
* Some architecture linkers will pad between
|
|
|
|
* the different mcount_loc sections of different
|
|
|
|
* object files to satisfy alignments.
|
|
|
|
* Skip any NULL pointers.
|
|
|
|
*/
|
|
|
|
if (!addr)
|
|
|
|
continue;
|
2012-04-25 05:45:26 +02:00
|
|
|
|
|
|
|
if (pg->index == pg->size) {
|
|
|
|
/* We should have allocated enough */
|
|
|
|
if (WARN_ON(!pg->next))
|
|
|
|
break;
|
|
|
|
pg = pg->next;
|
|
|
|
}
|
|
|
|
|
|
|
|
rec = &pg->records[pg->index++];
|
|
|
|
rec->ip = addr;
|
2008-08-14 21:45:08 +02:00
|
|
|
}
|
|
|
|
|
2012-04-25 05:45:26 +02:00
|
|
|
/* We should have used all pages */
|
|
|
|
WARN_ON(pg->next);
|
|
|
|
|
|
|
|
/* Assign the last page to ftrace_pages */
|
|
|
|
ftrace_pages = pg;
|
|
|
|
|
2011-06-07 15:26:46 +02:00
|
|
|
/*
|
2011-06-25 05:28:13 +02:00
|
|
|
* We only need to disable interrupts on start up
|
|
|
|
* because we are modifying code that an interrupt
|
|
|
|
* may execute, and the modification is not atomic.
|
|
|
|
* But for modules, nothing runs the code we modify
|
|
|
|
* until we are finished with it, and there's no
|
|
|
|
* reason to cause large interrupt latencies while we do it.
|
2011-06-07 15:26:46 +02:00
|
|
|
*/
|
2011-06-25 05:28:13 +02:00
|
|
|
if (!mod)
|
|
|
|
local_irq_save(flags);
|
2014-02-24 19:59:56 +01:00
|
|
|
ftrace_update_code(mod, start_pg);
|
2011-06-25 05:28:13 +02:00
|
|
|
if (!mod)
|
|
|
|
local_irq_restore(flags);
|
2011-12-16 22:23:44 +01:00
|
|
|
ret = 0;
|
|
|
|
out:
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-08-14 21:45:08 +02:00
|
|
|
|
2011-12-16 22:23:44 +01:00
|
|
|
return ret;
|
2008-08-14 21:45:08 +02:00
|
|
|
}
|
|
|
|
|
2009-04-15 19:24:06 +02:00
|
|
|
#ifdef CONFIG_MODULES
|
2011-12-16 20:42:37 +01:00
|
|
|
|
|
|
|
#define next_to_ftrace_page(p) container_of(p, struct ftrace_page, next)
|
|
|
|
|
2009-10-07 19:00:35 +02:00
|
|
|
void ftrace_release_mod(struct module *mod)
|
2009-04-15 19:24:06 +02:00
|
|
|
{
|
|
|
|
struct dyn_ftrace *rec;
|
2011-12-16 20:42:37 +01:00
|
|
|
struct ftrace_page **last_pg;
|
2009-04-15 19:24:06 +02:00
|
|
|
struct ftrace_page *pg;
|
2011-12-16 22:23:44 +01:00
|
|
|
int order;
|
2009-04-15 19:24:06 +02:00
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
|
2009-10-07 19:00:35 +02:00
|
|
|
if (ftrace_disabled)
|
2011-04-22 05:16:46 +02:00
|
|
|
goto out_unlock;
|
2009-04-15 19:24:06 +02:00
|
|
|
|
2011-12-16 20:42:37 +01:00
|
|
|
/*
|
|
|
|
* Each module has its own ftrace_pages, remove
|
|
|
|
* them from the list.
|
|
|
|
*/
|
|
|
|
last_pg = &ftrace_pages_start;
|
|
|
|
for (pg = ftrace_pages_start; pg; pg = *last_pg) {
|
|
|
|
rec = &pg->records[0];
|
2009-10-07 19:00:35 +02:00
|
|
|
if (within_module_core(rec->ip, mod)) {
|
2009-04-15 19:24:06 +02:00
|
|
|
/*
|
2011-12-16 20:42:37 +01:00
|
|
|
* As core pages are first, the first
|
|
|
|
* page should never be a module page.
|
2009-04-15 19:24:06 +02:00
|
|
|
*/
|
2011-12-16 20:42:37 +01:00
|
|
|
if (WARN_ON(pg == ftrace_pages_start))
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
/* Check if we are deleting the last page */
|
|
|
|
if (pg == ftrace_pages)
|
|
|
|
ftrace_pages = next_to_ftrace_page(last_pg);
|
|
|
|
|
|
|
|
*last_pg = pg->next;
|
2011-12-16 22:23:44 +01:00
|
|
|
order = get_count_order(pg->size / ENTRIES_PER_PAGE);
|
|
|
|
free_pages((unsigned long)pg->records, order);
|
|
|
|
kfree(pg);
|
2011-12-16 20:42:37 +01:00
|
|
|
} else
|
|
|
|
last_pg = &pg->next;
|
|
|
|
}
|
2011-04-22 05:16:46 +02:00
|
|
|
out_unlock:
|
2009-04-15 19:24:06 +02:00
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_init_module(struct module *mod,
|
|
|
|
unsigned long *start, unsigned long *end)
|
2008-08-14 21:45:09 +02:00
|
|
|
{
|
2008-08-16 03:40:04 +02:00
|
|
|
if (ftrace_disabled || start == end)
|
2008-08-15 04:47:19 +02:00
|
|
|
return;
|
2009-10-13 22:33:53 +02:00
|
|
|
ftrace_process_locs(mod, start, end);
|
2008-08-14 21:45:09 +02:00
|
|
|
}
|
|
|
|
|
2014-04-24 16:40:12 +02:00
|
|
|
void ftrace_module_init(struct module *mod)
|
2009-04-15 19:24:06 +02:00
|
|
|
{
|
2014-04-24 16:40:12 +02:00
|
|
|
ftrace_init_module(mod, mod->ftrace_callsites,
|
|
|
|
mod->ftrace_callsites +
|
|
|
|
mod->num_ftrace_callsites);
|
ftrace: Call ftrace cleanup module notifier after all other notifiers
Commit: c1bf08ac "ftrace: Be first to run code modification on modules"
changed ftrace module notifier's priority to INT_MAX in order to
process the ftrace nops before anything else could touch them
(namely kprobes). This was the correct thing to do.
Unfortunately, the ftrace module notifier also contains the ftrace
clean up code. As opposed to the set up code, this code should be
run *after* all the module notifiers have run in case a module is doing
correct clean-up and unregisters its ftrace hooks. Basically, ftrace
needs to do clean up on module removal, as it needs to know about code
being removed so that it doesn't try to modify that code. But after it
removes the module from its records, if a ftrace user tries to remove
a probe, that removal will fail due as the record of that code segment
no longer exists.
Nothing really bad happens if the probe removal is called after ftrace
did the clean up, but the ftrace removal function will return an error.
Correct code (such as kprobes) will produce a WARN_ON() if it fails
to remove the probe. As people get annoyed by frivolous warnings, it's
best to do the ftrace clean up after everything else.
By splitting the ftrace_module_notifier into two notifiers, one that
does the module load setup that is run at high priority, and the other
that is called for module clean up that is run at low priority, the
problem is solved.
Cc: stable@vger.kernel.org
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-02-13 21:18:38 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static int ftrace_module_notify_exit(struct notifier_block *self,
|
|
|
|
unsigned long val, void *data)
|
|
|
|
{
|
|
|
|
struct module *mod = data;
|
|
|
|
|
|
|
|
if (val == MODULE_STATE_GOING)
|
2009-10-07 19:00:35 +02:00
|
|
|
ftrace_release_mod(mod);
|
2009-04-15 19:24:06 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#else
|
ftrace: Call ftrace cleanup module notifier after all other notifiers
Commit: c1bf08ac "ftrace: Be first to run code modification on modules"
changed ftrace module notifier's priority to INT_MAX in order to
process the ftrace nops before anything else could touch them
(namely kprobes). This was the correct thing to do.
Unfortunately, the ftrace module notifier also contains the ftrace
clean up code. As opposed to the set up code, this code should be
run *after* all the module notifiers have run in case a module is doing
correct clean-up and unregisters its ftrace hooks. Basically, ftrace
needs to do clean up on module removal, as it needs to know about code
being removed so that it doesn't try to modify that code. But after it
removes the module from its records, if a ftrace user tries to remove
a probe, that removal will fail due as the record of that code segment
no longer exists.
Nothing really bad happens if the probe removal is called after ftrace
did the clean up, but the ftrace removal function will return an error.
Correct code (such as kprobes) will produce a WARN_ON() if it fails
to remove the probe. As people get annoyed by frivolous warnings, it's
best to do the ftrace clean up after everything else.
By splitting the ftrace_module_notifier into two notifiers, one that
does the module load setup that is run at high priority, and the other
that is called for module clean up that is run at low priority, the
problem is solved.
Cc: stable@vger.kernel.org
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-02-13 21:18:38 +01:00
|
|
|
static int ftrace_module_notify_exit(struct notifier_block *self,
|
|
|
|
unsigned long val, void *data)
|
2009-04-15 19:24:06 +02:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_MODULES */
|
|
|
|
|
ftrace: Call ftrace cleanup module notifier after all other notifiers
Commit: c1bf08ac "ftrace: Be first to run code modification on modules"
changed ftrace module notifier's priority to INT_MAX in order to
process the ftrace nops before anything else could touch them
(namely kprobes). This was the correct thing to do.
Unfortunately, the ftrace module notifier also contains the ftrace
clean up code. As opposed to the set up code, this code should be
run *after* all the module notifiers have run in case a module is doing
correct clean-up and unregisters its ftrace hooks. Basically, ftrace
needs to do clean up on module removal, as it needs to know about code
being removed so that it doesn't try to modify that code. But after it
removes the module from its records, if a ftrace user tries to remove
a probe, that removal will fail due as the record of that code segment
no longer exists.
Nothing really bad happens if the probe removal is called after ftrace
did the clean up, but the ftrace removal function will return an error.
Correct code (such as kprobes) will produce a WARN_ON() if it fails
to remove the probe. As people get annoyed by frivolous warnings, it's
best to do the ftrace clean up after everything else.
By splitting the ftrace_module_notifier into two notifiers, one that
does the module load setup that is run at high priority, and the other
that is called for module clean up that is run at low priority, the
problem is solved.
Cc: stable@vger.kernel.org
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-02-13 21:18:38 +01:00
|
|
|
struct notifier_block ftrace_module_exit_nb = {
|
|
|
|
.notifier_call = ftrace_module_notify_exit,
|
|
|
|
.priority = INT_MIN, /* Run after anything that can remove kprobes */
|
|
|
|
};
|
|
|
|
|
2008-08-14 21:45:08 +02:00
|
|
|
void __init ftrace_init(void)
|
|
|
|
{
|
2014-02-24 19:59:56 +01:00
|
|
|
extern unsigned long __start_mcount_loc[];
|
|
|
|
extern unsigned long __stop_mcount_loc[];
|
2014-02-24 19:59:59 +01:00
|
|
|
unsigned long count, flags;
|
2008-08-14 21:45:08 +02:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
local_irq_save(flags);
|
2014-02-24 19:59:59 +01:00
|
|
|
ret = ftrace_dyn_arch_init();
|
2008-08-14 21:45:08 +02:00
|
|
|
local_irq_restore(flags);
|
2014-02-24 19:59:58 +01:00
|
|
|
if (ret)
|
2008-08-14 21:45:08 +02:00
|
|
|
goto failed;
|
|
|
|
|
|
|
|
count = __stop_mcount_loc - __start_mcount_loc;
|
2014-02-24 19:59:57 +01:00
|
|
|
if (!count) {
|
|
|
|
pr_info("ftrace: No functions to be traced?\n");
|
2008-08-14 21:45:08 +02:00
|
|
|
goto failed;
|
2014-02-24 19:59:57 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
pr_info("ftrace: allocating %ld entries in %ld pages\n",
|
|
|
|
count, count / ENTRIES_PER_PAGE + 1);
|
2008-08-14 21:45:08 +02:00
|
|
|
|
|
|
|
last_ftrace_enabled = ftrace_enabled = 1;
|
|
|
|
|
2009-10-13 22:33:53 +02:00
|
|
|
ret = ftrace_process_locs(NULL,
|
2008-11-15 01:21:19 +01:00
|
|
|
__start_mcount_loc,
|
2008-08-14 21:45:08 +02:00
|
|
|
__stop_mcount_loc);
|
|
|
|
|
ftrace: Call ftrace cleanup module notifier after all other notifiers
Commit: c1bf08ac "ftrace: Be first to run code modification on modules"
changed ftrace module notifier's priority to INT_MAX in order to
process the ftrace nops before anything else could touch them
(namely kprobes). This was the correct thing to do.
Unfortunately, the ftrace module notifier also contains the ftrace
clean up code. As opposed to the set up code, this code should be
run *after* all the module notifiers have run in case a module is doing
correct clean-up and unregisters its ftrace hooks. Basically, ftrace
needs to do clean up on module removal, as it needs to know about code
being removed so that it doesn't try to modify that code. But after it
removes the module from its records, if a ftrace user tries to remove
a probe, that removal will fail due as the record of that code segment
no longer exists.
Nothing really bad happens if the probe removal is called after ftrace
did the clean up, but the ftrace removal function will return an error.
Correct code (such as kprobes) will produce a WARN_ON() if it fails
to remove the probe. As people get annoyed by frivolous warnings, it's
best to do the ftrace clean up after everything else.
By splitting the ftrace_module_notifier into two notifiers, one that
does the module load setup that is run at high priority, and the other
that is called for module clean up that is run at low priority, the
problem is solved.
Cc: stable@vger.kernel.org
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-02-13 21:18:38 +01:00
|
|
|
ret = register_module_notifier(&ftrace_module_exit_nb);
|
2009-05-17 09:31:38 +02:00
|
|
|
if (ret)
|
ftrace: Call ftrace cleanup module notifier after all other notifiers
Commit: c1bf08ac "ftrace: Be first to run code modification on modules"
changed ftrace module notifier's priority to INT_MAX in order to
process the ftrace nops before anything else could touch them
(namely kprobes). This was the correct thing to do.
Unfortunately, the ftrace module notifier also contains the ftrace
clean up code. As opposed to the set up code, this code should be
run *after* all the module notifiers have run in case a module is doing
correct clean-up and unregisters its ftrace hooks. Basically, ftrace
needs to do clean up on module removal, as it needs to know about code
being removed so that it doesn't try to modify that code. But after it
removes the module from its records, if a ftrace user tries to remove
a probe, that removal will fail due as the record of that code segment
no longer exists.
Nothing really bad happens if the probe removal is called after ftrace
did the clean up, but the ftrace removal function will return an error.
Correct code (such as kprobes) will produce a WARN_ON() if it fails
to remove the probe. As people get annoyed by frivolous warnings, it's
best to do the ftrace clean up after everything else.
By splitting the ftrace_module_notifier into two notifiers, one that
does the module load setup that is run at high priority, and the other
that is called for module clean up that is run at low priority, the
problem is solved.
Cc: stable@vger.kernel.org
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2013-02-13 21:18:38 +01:00
|
|
|
pr_warning("Failed to register trace ftrace module exit notifier\n");
|
2009-04-15 19:24:06 +02:00
|
|
|
|
2009-05-28 19:37:24 +02:00
|
|
|
set_ftrace_early_filters();
|
|
|
|
|
2008-08-14 21:45:08 +02:00
|
|
|
return;
|
|
|
|
failed:
|
|
|
|
ftrace_disabled = 1;
|
|
|
|
}
|
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
/* Do nothing if arch does not support this */
|
|
|
|
void __weak arch_ftrace_update_trampoline(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_update_trampoline(struct ftrace_ops *ops)
|
|
|
|
{
|
2014-07-03 21:48:16 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Currently there's no safe way to free a trampoline when the kernel
|
|
|
|
* is configured with PREEMPT. That is because a task could be preempted
|
|
|
|
* when it jumped to the trampoline, it may be preempted for a long time
|
|
|
|
* depending on the system load, and currently there's no way to know
|
|
|
|
* when it will be off the trampoline. If the trampoline is freed
|
|
|
|
* too early, when the task runs again, it will be executing on freed
|
|
|
|
* memory and crash.
|
|
|
|
*/
|
|
|
|
#ifdef CONFIG_PREEMPT
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
/* Currently, only non dynamic ops can have a trampoline */
|
|
|
|
if (ops->flags & FTRACE_OPS_FL_DYNAMIC)
|
|
|
|
return;
|
2014-07-03 21:48:16 +02:00
|
|
|
#endif
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
|
|
|
|
arch_ftrace_update_trampoline(ops);
|
|
|
|
}
|
|
|
|
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
#else
|
2008-10-28 20:17:38 +01:00
|
|
|
|
2011-05-04 04:49:52 +02:00
|
|
|
static struct ftrace_ops global_ops = {
|
2011-05-04 03:55:54 +02:00
|
|
|
.func = ftrace_stub,
|
2015-07-24 16:38:12 +02:00
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE |
|
|
|
|
FTRACE_OPS_FL_INITIALIZED |
|
|
|
|
FTRACE_OPS_FL_PID,
|
2011-05-04 03:55:54 +02:00
|
|
|
};
|
|
|
|
|
2008-10-28 20:17:38 +01:00
|
|
|
static int __init ftrace_nodyn_init(void)
|
|
|
|
{
|
|
|
|
ftrace_enabled = 1;
|
|
|
|
return 0;
|
|
|
|
}
|
2012-10-05 18:13:07 +02:00
|
|
|
core_initcall(ftrace_nodyn_init);
|
2008-10-28 20:17:38 +01:00
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
static inline int ftrace_init_dyn_tracefs(struct dentry *d_tracer) { return 0; }
|
2008-11-26 06:16:23 +01:00
|
|
|
static inline void ftrace_startup_enable(int command) { }
|
2014-08-05 23:19:38 +02:00
|
|
|
static inline void ftrace_startup_all(int command) { }
|
2008-11-26 06:16:24 +01:00
|
|
|
/* Keep as macros so we do not need to define the commands */
|
2013-11-26 02:59:46 +01:00
|
|
|
# define ftrace_startup(ops, command) \
|
|
|
|
({ \
|
|
|
|
int ___ret = __register_ftrace_function(ops); \
|
|
|
|
if (!___ret) \
|
|
|
|
(ops)->flags |= FTRACE_OPS_FL_ENABLED; \
|
|
|
|
___ret; \
|
2011-05-23 21:33:49 +02:00
|
|
|
})
|
2014-02-19 21:12:18 +01:00
|
|
|
# define ftrace_shutdown(ops, command) \
|
|
|
|
({ \
|
|
|
|
int ___ret = __unregister_ftrace_function(ops); \
|
|
|
|
if (!___ret) \
|
|
|
|
(ops)->flags &= ~FTRACE_OPS_FL_ENABLED; \
|
|
|
|
___ret; \
|
|
|
|
})
|
2013-11-26 02:59:46 +01:00
|
|
|
|
2008-05-12 21:20:45 +02:00
|
|
|
# define ftrace_startup_sysctl() do { } while (0)
|
|
|
|
# define ftrace_shutdown_sysctl() do { } while (0)
|
2011-05-04 15:27:52 +02:00
|
|
|
|
|
|
|
static inline int
|
2013-07-24 04:06:15 +02:00
|
|
|
ftrace_ops_test(struct ftrace_ops *ops, unsigned long ip, void *regs)
|
2011-05-04 15:27:52 +02:00
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
static void ftrace_update_trampoline(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
#endif /* CONFIG_DYNAMIC_FTRACE */
|
|
|
|
|
2014-01-10 23:01:58 +01:00
|
|
|
__init void ftrace_init_global_array_ops(struct trace_array *tr)
|
|
|
|
{
|
|
|
|
tr->ops = &global_ops;
|
|
|
|
tr->ops->private = tr;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func)
|
|
|
|
{
|
|
|
|
/* If we filter on pids, update to use the pid function */
|
|
|
|
if (tr->flags & TRACE_ARRAY_FL_GLOBAL) {
|
|
|
|
if (WARN_ON(tr->ops->func != ftrace_stub))
|
|
|
|
printk("ftrace ops had %pS for function\n",
|
|
|
|
tr->ops->func);
|
|
|
|
}
|
|
|
|
tr->ops->func = func;
|
|
|
|
tr->ops->private = tr;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ftrace_reset_array_ops(struct trace_array *tr)
|
|
|
|
{
|
|
|
|
tr->ops->func = ftrace_stub;
|
|
|
|
}
|
|
|
|
|
2012-02-15 15:51:48 +01:00
|
|
|
static void
|
2011-08-08 22:57:47 +02:00
|
|
|
ftrace_ops_control_func(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *op, struct pt_regs *regs)
|
2012-02-15 15:51:48 +01:00
|
|
|
{
|
|
|
|
if (unlikely(trace_recursion_test(TRACE_CONTROL_BIT)))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Some of the ops may be dynamically allocated,
|
|
|
|
* they must be freed after a synchronize_sched().
|
|
|
|
*/
|
|
|
|
preempt_disable_notrace();
|
|
|
|
trace_recursion_set(TRACE_CONTROL_BIT);
|
2013-11-05 00:34:44 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Control funcs (perf) uses RCU. Only trace if
|
|
|
|
* RCU is currently active.
|
|
|
|
*/
|
|
|
|
if (!rcu_is_watching())
|
|
|
|
goto out;
|
|
|
|
|
2012-11-02 22:03:03 +01:00
|
|
|
do_for_each_ftrace_op(op, ftrace_control_list) {
|
2013-03-27 14:31:28 +01:00
|
|
|
if (!(op->flags & FTRACE_OPS_FL_STUB) &&
|
|
|
|
!ftrace_function_local_disabled(op) &&
|
2013-07-24 04:06:15 +02:00
|
|
|
ftrace_ops_test(op, ip, regs))
|
2011-08-09 18:50:46 +02:00
|
|
|
op->func(ip, parent_ip, op, regs);
|
2012-11-02 22:03:03 +01:00
|
|
|
} while_for_each_ftrace_op(op);
|
2013-11-05 00:34:44 +01:00
|
|
|
out:
|
2012-02-15 15:51:48 +01:00
|
|
|
trace_recursion_clear(TRACE_CONTROL_BIT);
|
|
|
|
preempt_enable_notrace();
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct ftrace_ops control_ops = {
|
2013-05-09 07:44:17 +02:00
|
|
|
.func = ftrace_ops_control_func,
|
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE | FTRACE_OPS_FL_INITIALIZED,
|
2014-08-15 23:23:02 +02:00
|
|
|
INIT_OPS_HASH(control_ops)
|
2012-02-15 15:51:48 +01:00
|
|
|
};
|
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
static inline void
|
|
|
|
__ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *ignored, struct pt_regs *regs)
|
2011-05-04 15:27:52 +02:00
|
|
|
{
|
2011-05-06 03:14:55 +02:00
|
|
|
struct ftrace_ops *op;
|
2012-11-02 22:47:21 +01:00
|
|
|
int bit;
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2012-11-02 22:47:21 +01:00
|
|
|
bit = trace_test_and_set_recursion(TRACE_LIST_START, TRACE_LIST_MAX);
|
|
|
|
if (bit < 0)
|
|
|
|
return;
|
ftrace: Add internal recursive checks
Witold reported a reboot caused by the selftests of the dynamic function
tracer. He sent me a config and I used ktest to do a config_bisect on it
(as my config did not cause the crash). It pointed out that the problem
config was CONFIG_PROVE_RCU.
What happened was that if multiple callbacks are attached to the
function tracer, we iterate a list of callbacks. Because the list is
managed by synchronize_sched() and preempt_disable, the access to the
pointers uses rcu_dereference_raw().
When PROVE_RCU is enabled, the rcu_dereference_raw() calls some
debugging functions, which happen to be traced. The tracing of the debug
function would then call rcu_dereference_raw() which would then call the
debug function and then... well you get the idea.
I first wrote two different patches to solve this bug.
1) add a __rcu_dereference_raw() that would not do any checks.
2) add notrace to the offending debug functions.
Both of these patches worked.
Talking with Paul McKenney on IRC, he suggested to add recursion
detection instead. This seemed to be a better solution, so I decided to
implement it. As the task_struct already has a trace_recursion to detect
recursion in the ring buffer, and that has a very small number it
allows, I decided to use that same variable to add flags that can detect
the recursion inside the infrastructure of the function tracer.
I plan to change it so that the task struct bit can be checked in
mcount, but as that requires changes to all archs, I will hold that off
to the next merge window.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1306348063.1465.116.camel@gandalf.stny.rr.com
Reported-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-05-25 20:27:43 +02:00
|
|
|
|
2011-05-06 03:14:55 +02:00
|
|
|
/*
|
|
|
|
* Some of the ops may be dynamically allocated,
|
|
|
|
* they must be freed after a synchronize_sched().
|
|
|
|
*/
|
|
|
|
preempt_disable_notrace();
|
2012-11-02 22:03:03 +01:00
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
2014-01-10 23:01:58 +01:00
|
|
|
if (ftrace_ops_test(op, ip, regs)) {
|
2014-06-25 17:54:03 +02:00
|
|
|
if (FTRACE_WARN_ON(!op->func)) {
|
|
|
|
pr_warn("op=%p %pS\n", op, op);
|
2014-01-10 23:01:58 +01:00
|
|
|
goto out;
|
|
|
|
}
|
2011-08-09 18:50:46 +02:00
|
|
|
op->func(ip, parent_ip, op, regs);
|
2014-01-10 23:01:58 +01:00
|
|
|
}
|
2012-11-02 22:03:03 +01:00
|
|
|
} while_for_each_ftrace_op(op);
|
2014-01-10 23:01:58 +01:00
|
|
|
out:
|
2011-05-06 03:14:55 +02:00
|
|
|
preempt_enable_notrace();
|
2012-11-02 22:47:21 +01:00
|
|
|
trace_clear_recursion(bit);
|
2011-05-04 15:27:52 +02:00
|
|
|
}
|
|
|
|
|
2011-08-08 22:57:47 +02:00
|
|
|
/*
|
|
|
|
* Some archs only support passing ip and parent_ip. Even though
|
|
|
|
* the list function ignores the op parameter, we do not want any
|
|
|
|
* C side effects, where a function is called without the caller
|
|
|
|
* sending a third parameter.
|
2011-08-09 18:50:46 +02:00
|
|
|
* Archs are to support both the regs and ftrace_ops at the same time.
|
|
|
|
* If they support ftrace_ops, it is assumed they support regs.
|
|
|
|
* If call backs want to use regs, they must either check for regs
|
2012-09-28 10:15:17 +02:00
|
|
|
* being NULL, or CONFIG_DYNAMIC_FTRACE_WITH_REGS.
|
|
|
|
* Note, CONFIG_DYNAMIC_FTRACE_WITH_REGS expects a full regs to be saved.
|
2011-08-09 18:50:46 +02:00
|
|
|
* An architecture can pass partial regs with ftrace_ops and still
|
|
|
|
* set the ARCH_SUPPORT_FTARCE_OPS.
|
2011-08-08 22:57:47 +02:00
|
|
|
*/
|
|
|
|
#if ARCH_SUPPORTS_FTRACE_OPS
|
|
|
|
static void ftrace_ops_list_func(unsigned long ip, unsigned long parent_ip,
|
2011-08-09 18:50:46 +02:00
|
|
|
struct ftrace_ops *op, struct pt_regs *regs)
|
2011-08-08 22:57:47 +02:00
|
|
|
{
|
2011-08-09 18:50:46 +02:00
|
|
|
__ftrace_ops_list_func(ip, parent_ip, NULL, regs);
|
2011-08-08 22:57:47 +02:00
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void ftrace_ops_no_ops(unsigned long ip, unsigned long parent_ip)
|
|
|
|
{
|
2011-08-09 18:50:46 +02:00
|
|
|
__ftrace_ops_list_func(ip, parent_ip, NULL, NULL);
|
2011-08-08 22:57:47 +02:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2014-07-23 02:16:57 +02:00
|
|
|
/*
|
|
|
|
* If there's only one function registered but it does not support
|
|
|
|
* recursion, this function will be called by the mcount trampoline.
|
|
|
|
* This function will handle recursion protection.
|
|
|
|
*/
|
|
|
|
static void ftrace_ops_recurs_func(unsigned long ip, unsigned long parent_ip,
|
|
|
|
struct ftrace_ops *op, struct pt_regs *regs)
|
|
|
|
{
|
|
|
|
int bit;
|
|
|
|
|
|
|
|
bit = trace_test_and_set_recursion(TRACE_LIST_START, TRACE_LIST_MAX);
|
|
|
|
if (bit < 0)
|
|
|
|
return;
|
|
|
|
|
|
|
|
op->func(ip, parent_ip, op, regs);
|
|
|
|
|
|
|
|
trace_clear_recursion(bit);
|
|
|
|
}
|
|
|
|
|
2014-07-23 02:41:42 +02:00
|
|
|
/**
|
|
|
|
* ftrace_ops_get_func - get the function a trampoline should call
|
|
|
|
* @ops: the ops to get the function for
|
|
|
|
*
|
|
|
|
* Normally the mcount trampoline will call the ops->func, but there
|
|
|
|
* are times that it should not. For example, if the ops does not
|
|
|
|
* have its own recursion protection, then it should call the
|
|
|
|
* ftrace_ops_recurs_func() instead.
|
|
|
|
*
|
|
|
|
* Returns the function that the trampoline should call for @ops.
|
|
|
|
*/
|
|
|
|
ftrace_func_t ftrace_ops_get_func(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* If the func handles its own recursion, call it directly.
|
|
|
|
* Otherwise call the recursion protected function that
|
|
|
|
* will call the ftrace ops function.
|
|
|
|
*/
|
|
|
|
if (!(ops->flags & FTRACE_OPS_FL_RECURSION_SAFE))
|
|
|
|
return ftrace_ops_recurs_func;
|
|
|
|
|
|
|
|
return ops->func;
|
|
|
|
}
|
|
|
|
|
2008-12-04 06:26:41 +01:00
|
|
|
static void clear_ftrace_swapper(void)
|
2008-12-04 06:26:40 +01:00
|
|
|
{
|
|
|
|
struct task_struct *p;
|
2008-12-04 06:26:41 +01:00
|
|
|
int cpu;
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2008-12-04 06:26:41 +01:00
|
|
|
get_online_cpus();
|
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
p = idle_task(cpu);
|
2008-12-04 06:26:40 +01:00
|
|
|
clear_tsk_trace_trace(p);
|
2008-12-04 06:26:41 +01:00
|
|
|
}
|
|
|
|
put_online_cpus();
|
|
|
|
}
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2008-12-04 06:26:41 +01:00
|
|
|
static void set_ftrace_swapper(void)
|
|
|
|
{
|
|
|
|
struct task_struct *p;
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
get_online_cpus();
|
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
p = idle_task(cpu);
|
|
|
|
set_tsk_trace_trace(p);
|
|
|
|
}
|
|
|
|
put_online_cpus();
|
2008-12-04 06:26:40 +01:00
|
|
|
}
|
|
|
|
|
2008-12-04 06:26:41 +01:00
|
|
|
static void clear_ftrace_pid(struct pid *pid)
|
|
|
|
{
|
|
|
|
struct task_struct *p;
|
|
|
|
|
2009-02-03 20:39:04 +01:00
|
|
|
rcu_read_lock();
|
2008-12-04 06:26:41 +01:00
|
|
|
do_each_pid_task(pid, PIDTYPE_PID, p) {
|
|
|
|
clear_tsk_trace_trace(p);
|
|
|
|
} while_each_pid_task(pid, PIDTYPE_PID, p);
|
2009-02-03 20:39:04 +01:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
2008-12-04 06:26:41 +01:00
|
|
|
put_pid(pid);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void set_ftrace_pid(struct pid *pid)
|
2008-12-04 06:26:40 +01:00
|
|
|
{
|
|
|
|
struct task_struct *p;
|
|
|
|
|
2009-02-03 20:39:04 +01:00
|
|
|
rcu_read_lock();
|
2008-12-04 06:26:40 +01:00
|
|
|
do_each_pid_task(pid, PIDTYPE_PID, p) {
|
|
|
|
set_tsk_trace_trace(p);
|
|
|
|
} while_each_pid_task(pid, PIDTYPE_PID, p);
|
2009-02-03 20:39:04 +01:00
|
|
|
rcu_read_unlock();
|
2008-12-04 06:26:40 +01:00
|
|
|
}
|
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
static void clear_ftrace_pid_task(struct pid *pid)
|
2008-12-04 06:26:41 +01:00
|
|
|
{
|
2009-10-13 22:33:52 +02:00
|
|
|
if (pid == ftrace_swapper_pid)
|
2008-12-04 06:26:41 +01:00
|
|
|
clear_ftrace_swapper();
|
|
|
|
else
|
2009-10-13 22:33:52 +02:00
|
|
|
clear_ftrace_pid(pid);
|
2008-12-04 06:26:41 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void set_ftrace_pid_task(struct pid *pid)
|
|
|
|
{
|
|
|
|
if (pid == ftrace_swapper_pid)
|
|
|
|
set_ftrace_swapper();
|
|
|
|
else
|
|
|
|
set_ftrace_pid(pid);
|
|
|
|
}
|
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
static int ftrace_pid_add(int p)
|
2008-11-26 06:16:23 +01:00
|
|
|
{
|
2008-12-04 06:26:40 +01:00
|
|
|
struct pid *pid;
|
2009-10-13 22:33:52 +02:00
|
|
|
struct ftrace_pid *fpid;
|
|
|
|
int ret = -EINVAL;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
if (!p)
|
|
|
|
pid = ftrace_swapper_pid;
|
|
|
|
else
|
|
|
|
pid = find_get_pid(p);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
if (!pid)
|
|
|
|
goto out;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
ret = 0;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
list_for_each_entry(fpid, &ftrace_pids, list)
|
|
|
|
if (fpid->pid == pid)
|
|
|
|
goto out_put;
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
ret = -ENOMEM;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
fpid = kmalloc(sizeof(*fpid), GFP_KERNEL);
|
|
|
|
if (!fpid)
|
|
|
|
goto out_put;
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
list_add(&fpid->list, &ftrace_pids);
|
|
|
|
fpid->pid = pid;
|
2008-12-03 21:36:58 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
set_ftrace_pid_task(pid);
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
ftrace_update_pid_func();
|
2014-08-05 23:19:38 +02:00
|
|
|
|
|
|
|
ftrace_startup_all(0);
|
2009-10-13 22:33:52 +02:00
|
|
|
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_put:
|
|
|
|
if (pid != ftrace_swapper_pid)
|
|
|
|
put_pid(pid);
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
out:
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ftrace_pid_reset(void)
|
|
|
|
{
|
|
|
|
struct ftrace_pid *fpid, *safe;
|
2008-12-04 06:26:40 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
list_for_each_entry_safe(fpid, safe, &ftrace_pids, list) {
|
|
|
|
struct pid *pid = fpid->pid;
|
|
|
|
|
|
|
|
clear_ftrace_pid_task(pid);
|
|
|
|
|
|
|
|
list_del(&fpid->list);
|
|
|
|
kfree(fpid);
|
2008-11-26 06:16:23 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
ftrace_update_pid_func();
|
2014-08-05 23:19:38 +02:00
|
|
|
ftrace_startup_all(0);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2009-10-13 22:33:52 +02:00
|
|
|
}
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
static void *fpid_start(struct seq_file *m, loff_t *pos)
|
|
|
|
{
|
|
|
|
mutex_lock(&ftrace_lock);
|
|
|
|
|
2015-07-24 16:38:12 +02:00
|
|
|
if (!ftrace_pids_enabled() && (!*pos))
|
2009-10-13 22:33:52 +02:00
|
|
|
return (void *) 1;
|
|
|
|
|
|
|
|
return seq_list_start(&ftrace_pids, *pos);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *fpid_next(struct seq_file *m, void *v, loff_t *pos)
|
|
|
|
{
|
|
|
|
if (v == (void *)1)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return seq_list_next(v, &ftrace_pids, pos);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void fpid_stop(struct seq_file *m, void *p)
|
|
|
|
{
|
|
|
|
mutex_unlock(&ftrace_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int fpid_show(struct seq_file *m, void *v)
|
|
|
|
{
|
|
|
|
const struct ftrace_pid *fpid = list_entry(v, struct ftrace_pid, list);
|
|
|
|
|
|
|
|
if (v == (void *)1) {
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "no pid\n");
|
2009-10-13 22:33:52 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fpid->pid == ftrace_swapper_pid)
|
2014-11-08 21:42:10 +01:00
|
|
|
seq_puts(m, "swapper tasks\n");
|
2009-10-13 22:33:52 +02:00
|
|
|
else
|
|
|
|
seq_printf(m, "%u\n", pid_vnr(fpid->pid));
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct seq_operations ftrace_pid_sops = {
|
|
|
|
.start = fpid_start,
|
|
|
|
.next = fpid_next,
|
|
|
|
.stop = fpid_stop,
|
|
|
|
.show = fpid_show,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int
|
|
|
|
ftrace_pid_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
if ((file->f_mode & FMODE_WRITE) &&
|
|
|
|
(file->f_flags & O_TRUNC))
|
|
|
|
ftrace_pid_reset();
|
|
|
|
|
|
|
|
if (file->f_mode & FMODE_READ)
|
|
|
|
ret = seq_open(file, &ftrace_pid_sops);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-11-26 06:16:23 +01:00
|
|
|
static ssize_t
|
|
|
|
ftrace_pid_write(struct file *filp, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
2009-11-23 11:03:28 +01:00
|
|
|
char buf[64], *tmp;
|
2008-11-26 06:16:23 +01:00
|
|
|
long val;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (cnt >= sizeof(buf))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (copy_from_user(&buf, ubuf, cnt))
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
buf[cnt] = 0;
|
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
/*
|
|
|
|
* Allow "echo > set_ftrace_pid" or "echo -n '' > set_ftrace_pid"
|
|
|
|
* to clean the filter quietly.
|
|
|
|
*/
|
2009-11-23 11:03:28 +01:00
|
|
|
tmp = strstrip(buf);
|
|
|
|
if (strlen(tmp) == 0)
|
2009-10-13 22:33:52 +02:00
|
|
|
return 1;
|
|
|
|
|
2012-09-26 22:08:38 +02:00
|
|
|
ret = kstrtol(tmp, 10, &val);
|
2008-11-26 06:16:23 +01:00
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
ret = ftrace_pid_add(val);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
return ret ? ret : cnt;
|
|
|
|
}
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
static int
|
|
|
|
ftrace_pid_release(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
if (file->f_mode & FMODE_READ)
|
|
|
|
seq_release(inode, file);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-10-13 22:33:52 +02:00
|
|
|
return 0;
|
2008-11-26 06:16:23 +01:00
|
|
|
}
|
|
|
|
|
2009-03-06 03:44:55 +01:00
|
|
|
static const struct file_operations ftrace_pid_fops = {
|
2009-10-13 22:33:52 +02:00
|
|
|
.open = ftrace_pid_open,
|
|
|
|
.write = ftrace_pid_write,
|
|
|
|
.read = seq_read,
|
2013-12-21 23:39:40 +01:00
|
|
|
.llseek = tracing_lseek,
|
2009-10-13 22:33:52 +02:00
|
|
|
.release = ftrace_pid_release,
|
2008-11-26 06:16:23 +01:00
|
|
|
};
|
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
static __init int ftrace_init_tracefs(void)
|
2008-11-26 06:16:23 +01:00
|
|
|
{
|
|
|
|
struct dentry *d_tracer;
|
|
|
|
|
|
|
|
d_tracer = tracing_init_dentry();
|
2015-01-20 17:14:16 +01:00
|
|
|
if (IS_ERR(d_tracer))
|
2008-11-26 06:16:23 +01:00
|
|
|
return 0;
|
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
ftrace_init_dyn_tracefs(d_tracer);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2009-03-27 00:25:38 +01:00
|
|
|
trace_create_file("set_ftrace_pid", 0644, d_tracer,
|
|
|
|
NULL, &ftrace_pid_fops);
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2015-01-20 18:13:40 +01:00
|
|
|
ftrace_profile_tracefs(d_tracer);
|
2009-03-23 22:12:36 +01:00
|
|
|
|
2008-11-26 06:16:23 +01:00
|
|
|
return 0;
|
|
|
|
}
|
2015-01-20 18:13:40 +01:00
|
|
|
fs_initcall(ftrace_init_tracefs);
|
2008-11-26 06:16:23 +01:00
|
|
|
|
2008-07-11 02:58:15 +02:00
|
|
|
/**
|
2008-10-23 15:33:02 +02:00
|
|
|
* ftrace_kill - kill ftrace
|
2008-07-11 02:58:15 +02:00
|
|
|
*
|
|
|
|
* This function should be used by panic code. It stops ftrace
|
|
|
|
* but in a not so nice way. If you need to simply kill ftrace
|
|
|
|
* from a non-atomic section, use ftrace_kill.
|
|
|
|
*/
|
2008-10-23 15:33:02 +02:00
|
|
|
void ftrace_kill(void)
|
2008-07-11 02:58:15 +02:00
|
|
|
{
|
|
|
|
ftrace_disabled = 1;
|
|
|
|
ftrace_enabled = 0;
|
|
|
|
clear_ftrace_function();
|
|
|
|
}
|
|
|
|
|
2011-09-30 03:26:16 +02:00
|
|
|
/**
|
|
|
|
* Test if ftrace is dead or not.
|
|
|
|
*/
|
|
|
|
int ftrace_is_dead(void)
|
|
|
|
{
|
|
|
|
return ftrace_disabled;
|
|
|
|
}
|
|
|
|
|
2008-05-12 21:20:42 +02:00
|
|
|
/**
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* register_ftrace_function - register a function for profiling
|
|
|
|
* @ops - ops structure that holds the function for profiling.
|
2008-05-12 21:20:42 +02:00
|
|
|
*
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* Register a function to be called by all functions in the
|
|
|
|
* kernel.
|
|
|
|
*
|
|
|
|
* Note: @ops->func and all the functions it calls must be labeled
|
|
|
|
* with "notrace", otherwise it will go into a
|
|
|
|
* recursive loop.
|
2008-05-12 21:20:42 +02:00
|
|
|
*/
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
int register_ftrace_function(struct ftrace_ops *ops)
|
2008-05-12 21:20:42 +02:00
|
|
|
{
|
2011-04-22 05:16:46 +02:00
|
|
|
int ret = -1;
|
2008-05-12 21:20:48 +02:00
|
|
|
|
2013-05-09 07:44:17 +02:00
|
|
|
ftrace_ops_init(ops);
|
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2008-11-16 06:02:06 +01:00
|
|
|
|
2013-11-26 02:59:46 +01:00
|
|
|
ret = ftrace_startup(ops, 0);
|
2011-05-04 15:27:52 +02:00
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2012-03-29 19:11:40 +02:00
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
return ret;
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
}
|
2011-05-06 03:14:55 +02:00
|
|
|
EXPORT_SYMBOL_GPL(register_ftrace_function);
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
|
|
|
|
/**
|
2009-01-12 23:35:50 +01:00
|
|
|
* unregister_ftrace_function - unregister a function for profiling.
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
* @ops - ops structure that holds the function to unregister
|
|
|
|
*
|
|
|
|
* Unregister a function that was added to be called by ftrace profiling.
|
|
|
|
*/
|
|
|
|
int unregister_ftrace_function(struct ftrace_ops *ops)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2013-11-26 02:59:46 +01:00
|
|
|
ret = ftrace_shutdown(ops, 0);
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
2011-05-06 03:14:55 +02:00
|
|
|
EXPORT_SYMBOL_GPL(unregister_ftrace_function);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2008-05-12 21:20:51 +02:00
|
|
|
int
|
2008-05-12 21:20:43 +02:00
|
|
|
ftrace_enable_sysctl(struct ctl_table *table, int write,
|
2009-09-24 00:57:19 +02:00
|
|
|
void __user *buffer, size_t *lenp,
|
2008-05-12 21:20:43 +02:00
|
|
|
loff_t *ppos)
|
|
|
|
{
|
2011-04-22 05:16:46 +02:00
|
|
|
int ret = -ENODEV;
|
2008-05-12 21:20:48 +02:00
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2011-04-22 05:16:46 +02:00
|
|
|
if (unlikely(ftrace_disabled))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
ret = proc_dointvec(table, write, buffer, lenp, ppos);
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2009-06-26 10:55:51 +02:00
|
|
|
if (ret || !write || (last_ftrace_enabled == !!ftrace_enabled))
|
2008-05-12 21:20:43 +02:00
|
|
|
goto out;
|
|
|
|
|
2009-06-26 10:55:51 +02:00
|
|
|
last_ftrace_enabled = !!ftrace_enabled;
|
2008-05-12 21:20:43 +02:00
|
|
|
|
|
|
|
if (ftrace_enabled) {
|
|
|
|
|
|
|
|
/* we are starting ftrace again */
|
2013-03-26 17:53:03 +01:00
|
|
|
if (ftrace_ops_list != &ftrace_list_end)
|
|
|
|
update_ftrace_function();
|
2008-05-12 21:20:43 +02:00
|
|
|
|
2015-03-07 01:55:13 +01:00
|
|
|
ftrace_startup_sysctl();
|
|
|
|
|
2008-05-12 21:20:43 +02:00
|
|
|
} else {
|
|
|
|
/* stopping ftrace calls (just send to ftrace_stub) */
|
|
|
|
ftrace_trace_function = ftrace_stub;
|
|
|
|
|
|
|
|
ftrace_shutdown_sysctl();
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
ftrace: dynamic enabling/disabling of function calls
This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.
The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.
Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.
e.g.
call ftrace /* 5 bytes */
is replaced with
jmp 3f /* jmp is 2 bytes and we jump 3 forward */
3:
When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace. When it is disabled,
we replace the code back to the jmp.
Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls. A large batch is allocated at
boot up to get most of the calls there.
Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-05-12 21:20:42 +02:00
|
|
|
return ret;
|
2008-05-12 21:20:42 +02:00
|
|
|
}
|
2008-10-24 12:47:10 +02:00
|
|
|
|
2008-11-25 21:07:04 +01:00
|
|
|
#ifdef CONFIG_FUNCTION_GRAPH_TRACER
|
2008-11-16 06:02:06 +01:00
|
|
|
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
static struct ftrace_ops graph_ops = {
|
|
|
|
.func = ftrace_stub,
|
|
|
|
.flags = FTRACE_OPS_FL_RECURSION_SAFE |
|
|
|
|
FTRACE_OPS_FL_INITIALIZED |
|
2015-07-24 16:38:12 +02:00
|
|
|
FTRACE_OPS_FL_PID |
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
FTRACE_OPS_FL_STUB,
|
|
|
|
#ifdef FTRACE_GRAPH_TRAMP_ADDR
|
|
|
|
.trampoline = FTRACE_GRAPH_TRAMP_ADDR,
|
2014-11-19 03:14:11 +01:00
|
|
|
/* trampoline_size is only needed for dynamically allocated tramps */
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
#endif
|
|
|
|
ASSIGN_OPS_HASH(graph_ops, &global_ops.local_hash)
|
|
|
|
};
|
|
|
|
|
2015-09-30 01:06:50 +02:00
|
|
|
void ftrace_graph_sleep_time_control(bool enable)
|
|
|
|
{
|
|
|
|
fgraph_sleep_time = enable;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ftrace_graph_graph_time_control(bool enable)
|
|
|
|
{
|
|
|
|
fgraph_graph_time = enable;
|
|
|
|
}
|
|
|
|
|
2008-12-03 05:50:05 +01:00
|
|
|
int ftrace_graph_entry_stub(struct ftrace_graph_ent *trace)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-26 00:57:25 +01:00
|
|
|
/* The callbacks that hook a function */
|
|
|
|
trace_func_graph_ret_t ftrace_graph_return =
|
|
|
|
(trace_func_graph_ret_t)ftrace_stub;
|
2008-12-03 05:50:05 +01:00
|
|
|
trace_func_graph_ent_t ftrace_graph_entry = ftrace_graph_entry_stub;
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
static trace_func_graph_ent_t __ftrace_graph_entry = ftrace_graph_entry_stub;
|
2008-11-23 06:22:56 +01:00
|
|
|
|
|
|
|
/* Try to assign a return stack array on FTRACE_RETSTACK_ALLOC_SIZE tasks. */
|
|
|
|
static int alloc_retstack_tasklist(struct ftrace_ret_stack **ret_stack_list)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int ret = 0;
|
|
|
|
unsigned long flags;
|
|
|
|
int start = 0, end = FTRACE_RETSTACK_ALLOC_SIZE;
|
|
|
|
struct task_struct *g, *t;
|
|
|
|
|
|
|
|
for (i = 0; i < FTRACE_RETSTACK_ALLOC_SIZE; i++) {
|
|
|
|
ret_stack_list[i] = kmalloc(FTRACE_RETFUNC_DEPTH
|
|
|
|
* sizeof(struct ftrace_ret_stack),
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!ret_stack_list[i]) {
|
|
|
|
start = 0;
|
|
|
|
end = i;
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto free;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
read_lock_irqsave(&tasklist_lock, flags);
|
|
|
|
do_each_thread(g, t) {
|
|
|
|
if (start == end) {
|
|
|
|
ret = -EAGAIN;
|
|
|
|
goto unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (t->ret_stack == NULL) {
|
2008-12-06 03:43:41 +01:00
|
|
|
atomic_set(&t->tracing_graph_pause, 0);
|
2008-11-23 06:22:56 +01:00
|
|
|
atomic_set(&t->trace_overrun, 0);
|
2009-06-02 20:01:19 +02:00
|
|
|
t->curr_ret_stack = -1;
|
|
|
|
/* Make sure the tasks see the -1 first: */
|
|
|
|
smp_wmb();
|
|
|
|
t->ret_stack = ret_stack_list[start++];
|
2008-11-23 06:22:56 +01:00
|
|
|
}
|
|
|
|
} while_each_thread(g, t);
|
|
|
|
|
|
|
|
unlock:
|
|
|
|
read_unlock_irqrestore(&tasklist_lock, flags);
|
|
|
|
free:
|
|
|
|
for (i = start; i < end; i++)
|
|
|
|
kfree(ret_stack_list[i]);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-03-24 06:10:15 +01:00
|
|
|
static void
|
2015-09-28 18:06:56 +02:00
|
|
|
ftrace_graph_probe_sched_switch(void *ignore, bool preempt,
|
tracing: Let tracepoints have data passed to tracepoint callbacks
This patch adds data to be passed to tracepoint callbacks.
The created functions from DECLARE_TRACE() now need a mandatory data
parameter. For example:
DECLARE_TRACE(mytracepoint, int value, value)
Will create the register function:
int register_trace_mytracepoint((void(*)(void *data, int value))probe,
void *data);
As the first argument, all callbacks (probes) must take a (void *data)
parameter. So a callback for the above tracepoint will look like:
void myprobe(void *data, int value)
{
}
The callback may choose to ignore the data parameter.
This change allows callbacks to register a private data pointer along
with the function probe.
void mycallback(void *data, int value);
register_trace_mytracepoint(mycallback, mydata);
Then the mycallback() will receive the "mydata" as the first parameter
before the args.
A more detailed example:
DECLARE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
/* In the C file */
DEFINE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
[...]
trace_mytracepoint(status);
/* In a file registering this tracepoint */
int my_callback(void *data, int status)
{
struct my_struct my_data = data;
[...]
}
[...]
my_data = kmalloc(sizeof(*my_data), GFP_KERNEL);
init_my_data(my_data);
register_trace_mytracepoint(my_callback, my_data);
The same callback can also be registered to the same tracepoint as long
as the data registered is different. Note, the data must also be used
to unregister the callback:
unregister_trace_mytracepoint(my_callback, my_data);
Because of the data parameter, tracepoints declared this way can not have
no args. That is:
DECLARE_TRACE(mytracepoint, TP_PROTO(void), TP_ARGS());
will cause an error.
If no arguments are needed, a new macro can be used instead:
DECLARE_TRACE_NOARGS(mytracepoint);
Since there are no arguments, the proto and args fields are left out.
This is part of a series to make the tracepoint footprint smaller:
text data bss dec hex filename
4913961 1088356 861512 6863829 68bbd5 vmlinux.orig
4914025 1088868 861512 6864405 68be15 vmlinux.class
4918492 1084612 861512 6864616 68bee8 vmlinux.tracepoint
Again, this patch also increases the size of the kernel, but
lays the ground work for decreasing it.
v5: Fixed net/core/drop_monitor.c to handle these updates.
v4: Moved the DECLARE_TRACE() DECLARE_TRACE_NOARGS out of the
#ifdef CONFIG_TRACE_POINTS, since the two are the same in both
cases. The __DECLARE_TRACE() is what changes.
Thanks to Frederic Weisbecker for pointing this out.
v3: Made all register_* functions require data to be passed and
all callbacks to take a void * parameter as its first argument.
This makes the calling functions comply with C standards.
Also added more comments to the modifications of DECLARE_TRACE().
v2: Made the DECLARE_TRACE() have the ability to pass arguments
and added a new DECLARE_TRACE_NOARGS() for tracepoints that
do not need any arguments.
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-20 23:04:50 +02:00
|
|
|
struct task_struct *prev, struct task_struct *next)
|
2009-03-24 06:10:15 +01:00
|
|
|
{
|
|
|
|
unsigned long long timestamp;
|
|
|
|
int index;
|
|
|
|
|
2009-03-24 16:06:24 +01:00
|
|
|
/*
|
|
|
|
* Does the user want to count the time a function was asleep.
|
|
|
|
* If so, do not update the time stamps.
|
|
|
|
*/
|
2015-09-30 01:06:50 +02:00
|
|
|
if (fgraph_sleep_time)
|
2009-03-24 16:06:24 +01:00
|
|
|
return;
|
|
|
|
|
2009-03-24 06:10:15 +01:00
|
|
|
timestamp = trace_clock_local();
|
|
|
|
|
|
|
|
prev->ftrace_timestamp = timestamp;
|
|
|
|
|
|
|
|
/* only process tasks that we timestamped */
|
|
|
|
if (!next->ftrace_timestamp)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update all the counters in next to make up for the
|
|
|
|
* time next was sleeping.
|
|
|
|
*/
|
|
|
|
timestamp -= next->ftrace_timestamp;
|
|
|
|
|
|
|
|
for (index = next->curr_ret_stack; index >= 0; index--)
|
|
|
|
next->ret_stack[index].calltime += timestamp;
|
|
|
|
}
|
|
|
|
|
2008-11-23 06:22:56 +01:00
|
|
|
/* Allocate a return stack for each task */
|
2008-11-25 21:07:04 +01:00
|
|
|
static int start_graph_tracing(void)
|
2008-11-23 06:22:56 +01:00
|
|
|
{
|
|
|
|
struct ftrace_ret_stack **ret_stack_list;
|
2009-02-17 18:35:34 +01:00
|
|
|
int ret, cpu;
|
2008-11-23 06:22:56 +01:00
|
|
|
|
|
|
|
ret_stack_list = kmalloc(FTRACE_RETSTACK_ALLOC_SIZE *
|
|
|
|
sizeof(struct ftrace_ret_stack *),
|
|
|
|
GFP_KERNEL);
|
|
|
|
|
|
|
|
if (!ret_stack_list)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2009-02-17 18:35:34 +01:00
|
|
|
/* The cpu_boot init_task->ret_stack will never be freed */
|
2009-06-02 18:03:19 +02:00
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
if (!idle_task(cpu)->ret_stack)
|
ftrace: Fix memory leak with function graph and cpu hotplug
When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.
On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.
ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.
The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-02-11 03:26:13 +01:00
|
|
|
ftrace_graph_init_idle_task(idle_task(cpu), cpu);
|
2009-06-02 18:03:19 +02:00
|
|
|
}
|
2009-02-17 18:35:34 +01:00
|
|
|
|
2008-11-23 06:22:56 +01:00
|
|
|
do {
|
|
|
|
ret = alloc_retstack_tasklist(ret_stack_list);
|
|
|
|
} while (ret == -EAGAIN);
|
|
|
|
|
2009-03-24 06:10:15 +01:00
|
|
|
if (!ret) {
|
tracing: Let tracepoints have data passed to tracepoint callbacks
This patch adds data to be passed to tracepoint callbacks.
The created functions from DECLARE_TRACE() now need a mandatory data
parameter. For example:
DECLARE_TRACE(mytracepoint, int value, value)
Will create the register function:
int register_trace_mytracepoint((void(*)(void *data, int value))probe,
void *data);
As the first argument, all callbacks (probes) must take a (void *data)
parameter. So a callback for the above tracepoint will look like:
void myprobe(void *data, int value)
{
}
The callback may choose to ignore the data parameter.
This change allows callbacks to register a private data pointer along
with the function probe.
void mycallback(void *data, int value);
register_trace_mytracepoint(mycallback, mydata);
Then the mycallback() will receive the "mydata" as the first parameter
before the args.
A more detailed example:
DECLARE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
/* In the C file */
DEFINE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
[...]
trace_mytracepoint(status);
/* In a file registering this tracepoint */
int my_callback(void *data, int status)
{
struct my_struct my_data = data;
[...]
}
[...]
my_data = kmalloc(sizeof(*my_data), GFP_KERNEL);
init_my_data(my_data);
register_trace_mytracepoint(my_callback, my_data);
The same callback can also be registered to the same tracepoint as long
as the data registered is different. Note, the data must also be used
to unregister the callback:
unregister_trace_mytracepoint(my_callback, my_data);
Because of the data parameter, tracepoints declared this way can not have
no args. That is:
DECLARE_TRACE(mytracepoint, TP_PROTO(void), TP_ARGS());
will cause an error.
If no arguments are needed, a new macro can be used instead:
DECLARE_TRACE_NOARGS(mytracepoint);
Since there are no arguments, the proto and args fields are left out.
This is part of a series to make the tracepoint footprint smaller:
text data bss dec hex filename
4913961 1088356 861512 6863829 68bbd5 vmlinux.orig
4914025 1088868 861512 6864405 68be15 vmlinux.class
4918492 1084612 861512 6864616 68bee8 vmlinux.tracepoint
Again, this patch also increases the size of the kernel, but
lays the ground work for decreasing it.
v5: Fixed net/core/drop_monitor.c to handle these updates.
v4: Moved the DECLARE_TRACE() DECLARE_TRACE_NOARGS out of the
#ifdef CONFIG_TRACE_POINTS, since the two are the same in both
cases. The __DECLARE_TRACE() is what changes.
Thanks to Frederic Weisbecker for pointing this out.
v3: Made all register_* functions require data to be passed and
all callbacks to take a void * parameter as its first argument.
This makes the calling functions comply with C standards.
Also added more comments to the modifications of DECLARE_TRACE().
v2: Made the DECLARE_TRACE() have the ability to pass arguments
and added a new DECLARE_TRACE_NOARGS() for tracepoints that
do not need any arguments.
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-20 23:04:50 +02:00
|
|
|
ret = register_trace_sched_switch(ftrace_graph_probe_sched_switch, NULL);
|
2009-03-24 06:10:15 +01:00
|
|
|
if (ret)
|
|
|
|
pr_info("ftrace_graph: Couldn't activate tracepoint"
|
|
|
|
" probe to kernel_sched_switch\n");
|
|
|
|
}
|
|
|
|
|
2008-11-23 06:22:56 +01:00
|
|
|
kfree(ret_stack_list);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-01-14 22:33:27 +01:00
|
|
|
/*
|
|
|
|
* Hibernation protection.
|
|
|
|
* The state of the current task is too much unstable during
|
|
|
|
* suspend/restore to disk. We want to protect against that.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
ftrace_suspend_notifier_call(struct notifier_block *bl, unsigned long state,
|
|
|
|
void *unused)
|
|
|
|
{
|
|
|
|
switch (state) {
|
|
|
|
case PM_HIBERNATION_PREPARE:
|
|
|
|
pause_graph_tracing();
|
|
|
|
break;
|
|
|
|
|
|
|
|
case PM_POST_HIBERNATION:
|
|
|
|
unpause_graph_tracing();
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return NOTIFY_DONE;
|
|
|
|
}
|
|
|
|
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
static int ftrace_graph_entry_test(struct ftrace_graph_ent *trace)
|
|
|
|
{
|
|
|
|
if (!ftrace_ops_test(&global_ops, trace->func, NULL))
|
|
|
|
return 0;
|
|
|
|
return __ftrace_graph_entry(trace);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The function graph tracer should only trace the functions defined
|
|
|
|
* by set_ftrace_filter and set_ftrace_notrace. If another function
|
|
|
|
* tracer ops is registered, the graph tracer requires testing the
|
|
|
|
* function against the global ops, and not just trace any function
|
|
|
|
* that any ftrace_ops registered.
|
|
|
|
*/
|
|
|
|
static void update_function_graph_func(void)
|
|
|
|
{
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
struct ftrace_ops *op;
|
|
|
|
bool do_test = false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The graph and global ops share the same set of functions
|
|
|
|
* to test. If any other ops is on the list, then
|
|
|
|
* the graph tracing needs to test if its the function
|
|
|
|
* it should call.
|
|
|
|
*/
|
|
|
|
do_for_each_ftrace_op(op, ftrace_ops_list) {
|
|
|
|
if (op != &global_ops && op != &graph_ops &&
|
|
|
|
op != &ftrace_list_end) {
|
|
|
|
do_test = true;
|
|
|
|
/* in double loop, break out with goto */
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
} while_for_each_ftrace_op(op);
|
|
|
|
out:
|
|
|
|
if (do_test)
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
ftrace_graph_entry = ftrace_graph_entry_test;
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
else
|
|
|
|
ftrace_graph_entry = __ftrace_graph_entry;
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
}
|
|
|
|
|
2014-03-30 15:31:50 +02:00
|
|
|
static struct notifier_block ftrace_suspend_notifier = {
|
|
|
|
.notifier_call = ftrace_suspend_notifier_call,
|
|
|
|
};
|
|
|
|
|
2008-11-26 00:57:25 +01:00
|
|
|
int register_ftrace_graph(trace_func_graph_ret_t retfunc,
|
|
|
|
trace_func_graph_ent_t entryfunc)
|
2008-11-11 07:14:25 +01:00
|
|
|
{
|
2008-11-16 06:02:06 +01:00
|
|
|
int ret = 0;
|
|
|
|
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2008-11-16 06:02:06 +01:00
|
|
|
|
2009-03-24 05:18:31 +01:00
|
|
|
/* we currently allow only one tracer registered at a time */
|
2009-04-03 21:24:12 +02:00
|
|
|
if (ftrace_graph_active) {
|
2009-03-24 05:18:31 +01:00
|
|
|
ret = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-01-14 22:33:27 +01:00
|
|
|
register_pm_notifier(&ftrace_suspend_notifier);
|
|
|
|
|
2009-04-03 21:24:12 +02:00
|
|
|
ftrace_graph_active++;
|
2008-11-25 21:07:04 +01:00
|
|
|
ret = start_graph_tracing();
|
2008-11-23 06:22:56 +01:00
|
|
|
if (ret) {
|
2009-04-03 21:24:12 +02:00
|
|
|
ftrace_graph_active--;
|
2008-11-23 06:22:56 +01:00
|
|
|
goto out;
|
|
|
|
}
|
2008-11-26 06:16:25 +01:00
|
|
|
|
2008-11-26 00:57:25 +01:00
|
|
|
ftrace_graph_return = retfunc;
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the indirect function to the entryfunc, and the
|
|
|
|
* function that gets called to the entry_test first. Then
|
|
|
|
* call the update fgraph entry function to determine if
|
|
|
|
* the entryfunc should be called directly or not.
|
|
|
|
*/
|
|
|
|
__ftrace_graph_entry = entryfunc;
|
|
|
|
ftrace_graph_entry = ftrace_graph_entry_test;
|
|
|
|
update_function_graph_func();
|
2008-11-26 06:16:25 +01:00
|
|
|
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
ret = ftrace_startup(&graph_ops, FTRACE_START_FUNC_RET);
|
2008-11-16 06:02:06 +01:00
|
|
|
out:
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-11-16 06:02:06 +01:00
|
|
|
return ret;
|
2008-11-11 07:14:25 +01:00
|
|
|
}
|
|
|
|
|
2008-11-25 21:07:04 +01:00
|
|
|
void unregister_ftrace_graph(void)
|
2008-11-11 07:14:25 +01:00
|
|
|
{
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_lock(&ftrace_lock);
|
2008-11-16 06:02:06 +01:00
|
|
|
|
2009-04-03 21:24:12 +02:00
|
|
|
if (unlikely(!ftrace_graph_active))
|
2009-03-30 17:11:28 +02:00
|
|
|
goto out;
|
|
|
|
|
2009-04-03 21:24:12 +02:00
|
|
|
ftrace_graph_active--;
|
2008-11-26 00:57:25 +01:00
|
|
|
ftrace_graph_return = (trace_func_graph_ret_t)ftrace_stub;
|
2008-12-03 05:50:05 +01:00
|
|
|
ftrace_graph_entry = ftrace_graph_entry_stub;
|
ftrace: Have function graph only trace based on global_ops filters
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
For example, one just needs to do the following:
# cd /sys/kernel/debug/tracing
# echo schedule > set_ftrace_filter
# echo function_graph > current_tracer
# cat trace
[..]
0) | schedule() {
------------------------------------------
0) <idle>-0 => rcu_pre-7
------------------------------------------
0) ! 2980.314 us | }
0) | schedule() {
------------------------------------------
0) rcu_pre-7 => <idle>-0
------------------------------------------
0) + 20.701 us | }
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Cc: stable@vger.kernel.org # 3.3+
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-01-13 16:30:23 +01:00
|
|
|
__ftrace_graph_entry = ftrace_graph_entry_stub;
|
ftrace: Fix function_profiler and function tracer together
The latest rewrite of ftrace removed the separate ftrace_ops of
the function tracer and the function graph tracer and had them
share the same ftrace_ops. This simplified the accounting by removing
the multiple layers of functions called, where the global_ops func
would call a special list that would iterate over the other ops that
were registered within it (like function and function graph), which
itself was registered to the ftrace ops list of all functions
currently active. If that sounds confusing, the code that implemented
it was also confusing and its removal is a good thing.
The problem with this change was that it assumed that the function
and function graph tracer can never be used at the same time.
This is mostly true, but there is an exception. That is when the
function profiler uses the function graph tracer to profile.
The function profiler can be activated the same time as the function
tracer, and this breaks the assumption and the result is that ftrace
will crash (it detects the error and shuts itself down, it does not
cause a kernel oops).
To solve this issue, a previous change allowed the hash tables
for the functions traced by a ftrace_ops to be a pointer and let
multiple ftrace_ops share the same hash. This allows the function
and function_graph tracer to have separate ftrace_ops, but still
share the hash, which is what is done.
Now the function and function graph tracers have separate ftrace_ops
again, and the function tracer can be run while the function_profile
is active.
Cc: stable@vger.kernel.org # 3.16 (apply after 3.17-rc4 is out)
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-08-15 23:18:46 +02:00
|
|
|
ftrace_shutdown(&graph_ops, FTRACE_STOP_FUNC_RET);
|
2009-01-14 22:33:27 +01:00
|
|
|
unregister_pm_notifier(&ftrace_suspend_notifier);
|
tracing: Let tracepoints have data passed to tracepoint callbacks
This patch adds data to be passed to tracepoint callbacks.
The created functions from DECLARE_TRACE() now need a mandatory data
parameter. For example:
DECLARE_TRACE(mytracepoint, int value, value)
Will create the register function:
int register_trace_mytracepoint((void(*)(void *data, int value))probe,
void *data);
As the first argument, all callbacks (probes) must take a (void *data)
parameter. So a callback for the above tracepoint will look like:
void myprobe(void *data, int value)
{
}
The callback may choose to ignore the data parameter.
This change allows callbacks to register a private data pointer along
with the function probe.
void mycallback(void *data, int value);
register_trace_mytracepoint(mycallback, mydata);
Then the mycallback() will receive the "mydata" as the first parameter
before the args.
A more detailed example:
DECLARE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
/* In the C file */
DEFINE_TRACE(mytracepoint, TP_PROTO(int status), TP_ARGS(status));
[...]
trace_mytracepoint(status);
/* In a file registering this tracepoint */
int my_callback(void *data, int status)
{
struct my_struct my_data = data;
[...]
}
[...]
my_data = kmalloc(sizeof(*my_data), GFP_KERNEL);
init_my_data(my_data);
register_trace_mytracepoint(my_callback, my_data);
The same callback can also be registered to the same tracepoint as long
as the data registered is different. Note, the data must also be used
to unregister the callback:
unregister_trace_mytracepoint(my_callback, my_data);
Because of the data parameter, tracepoints declared this way can not have
no args. That is:
DECLARE_TRACE(mytracepoint, TP_PROTO(void), TP_ARGS());
will cause an error.
If no arguments are needed, a new macro can be used instead:
DECLARE_TRACE_NOARGS(mytracepoint);
Since there are no arguments, the proto and args fields are left out.
This is part of a series to make the tracepoint footprint smaller:
text data bss dec hex filename
4913961 1088356 861512 6863829 68bbd5 vmlinux.orig
4914025 1088868 861512 6864405 68be15 vmlinux.class
4918492 1084612 861512 6864616 68bee8 vmlinux.tracepoint
Again, this patch also increases the size of the kernel, but
lays the ground work for decreasing it.
v5: Fixed net/core/drop_monitor.c to handle these updates.
v4: Moved the DECLARE_TRACE() DECLARE_TRACE_NOARGS out of the
#ifdef CONFIG_TRACE_POINTS, since the two are the same in both
cases. The __DECLARE_TRACE() is what changes.
Thanks to Frederic Weisbecker for pointing this out.
v3: Made all register_* functions require data to be passed and
all callbacks to take a void * parameter as its first argument.
This makes the calling functions comply with C standards.
Also added more comments to the modifications of DECLARE_TRACE().
v2: Made the DECLARE_TRACE() have the ability to pass arguments
and added a new DECLARE_TRACE_NOARGS() for tracepoints that
do not need any arguments.
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-20 23:04:50 +02:00
|
|
|
unregister_trace_sched_switch(ftrace_graph_probe_sched_switch, NULL);
|
2008-11-16 06:02:06 +01:00
|
|
|
|
ftrace/x86: Add dynamic allocated trampoline for ftrace_ops
The current method of handling multiple function callbacks is to register
a list function callback that calls all the other callbacks based on
their hash tables and compare it to the function that the callback was
called on. But this is very inefficient.
For example, if you are tracing all functions in the kernel and then
add a kprobe to a function such that the kprobe uses ftrace, the
mcount trampoline will switch from calling the function trace callback
to calling the list callback that will iterate over all registered
ftrace_ops (in this case, the function tracer and the kprobes callback).
That means for every function being traced it checks the hash of the
ftrace_ops for function tracing and kprobes, even though the kprobes
is only set at a single function. The kprobes ftrace_ops is checked
for every function being traced!
Instead of calling the list function for functions that are only being
traced by a single callback, we can call a dynamically allocated
trampoline that calls the callback directly. The function graph tracer
already uses a direct call trampoline when it is being traced by itself
but it is not dynamically allocated. It's trampoline is static in the
kernel core. The infrastructure that called the function graph trampoline
can also be used to call a dynamically allocated one.
For now, only ftrace_ops that are not dynamically allocated can have
a trampoline. That is, users such as function tracer or stack tracer.
kprobes and perf allocate their ftrace_ops, and until there's a safe
way to free the trampoline, it can not be used. The dynamically allocated
ftrace_ops may, although, use the trampoline if the kernel is not
compiled with CONFIG_PREEMPT. But that will come later.
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2014-07-03 05:23:31 +02:00
|
|
|
#ifdef CONFIG_DYNAMIC_FTRACE
|
|
|
|
/*
|
|
|
|
* Function graph does not allocate the trampoline, but
|
|
|
|
* other global_ops do. We need to reset the ALLOC_TRAMP flag
|
|
|
|
* if one was used.
|
|
|
|
*/
|
|
|
|
global_ops.trampoline = save_global_trampoline;
|
|
|
|
if (save_global_flags & FTRACE_OPS_FL_ALLOC_TRAMP)
|
|
|
|
global_ops.flags |= FTRACE_OPS_FL_ALLOC_TRAMP;
|
|
|
|
#endif
|
|
|
|
|
2009-03-30 17:11:28 +02:00
|
|
|
out:
|
2009-02-14 07:42:44 +01:00
|
|
|
mutex_unlock(&ftrace_lock);
|
2008-11-11 07:14:25 +01:00
|
|
|
}
|
2008-11-23 06:22:56 +01:00
|
|
|
|
ftrace: Fix memory leak with function graph and cpu hotplug
When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.
On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.
ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.
The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-02-11 03:26:13 +01:00
|
|
|
static DEFINE_PER_CPU(struct ftrace_ret_stack *, idle_ret_stack);
|
|
|
|
|
|
|
|
static void
|
|
|
|
graph_init_task(struct task_struct *t, struct ftrace_ret_stack *ret_stack)
|
|
|
|
{
|
|
|
|
atomic_set(&t->tracing_graph_pause, 0);
|
|
|
|
atomic_set(&t->trace_overrun, 0);
|
|
|
|
t->ftrace_timestamp = 0;
|
2011-03-31 03:57:33 +02:00
|
|
|
/* make curr_ret_stack visible before we add the ret_stack */
|
ftrace: Fix memory leak with function graph and cpu hotplug
When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.
On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.
ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.
The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-02-11 03:26:13 +01:00
|
|
|
smp_wmb();
|
|
|
|
t->ret_stack = ret_stack;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate a return stack for the idle task. May be the first
|
|
|
|
* time through, or it may be done by CPU hotplug online.
|
|
|
|
*/
|
|
|
|
void ftrace_graph_init_idle_task(struct task_struct *t, int cpu)
|
|
|
|
{
|
|
|
|
t->curr_ret_stack = -1;
|
|
|
|
/*
|
|
|
|
* The idle task has no parent, it either has its own
|
|
|
|
* stack or no stack at all.
|
|
|
|
*/
|
|
|
|
if (t->ret_stack)
|
|
|
|
WARN_ON(t->ret_stack != per_cpu(idle_ret_stack, cpu));
|
|
|
|
|
|
|
|
if (ftrace_graph_active) {
|
|
|
|
struct ftrace_ret_stack *ret_stack;
|
|
|
|
|
|
|
|
ret_stack = per_cpu(idle_ret_stack, cpu);
|
|
|
|
if (!ret_stack) {
|
|
|
|
ret_stack = kmalloc(FTRACE_RETFUNC_DEPTH
|
|
|
|
* sizeof(struct ftrace_ret_stack),
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!ret_stack)
|
|
|
|
return;
|
|
|
|
per_cpu(idle_ret_stack, cpu) = ret_stack;
|
|
|
|
}
|
|
|
|
graph_init_task(t, ret_stack);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-11-23 06:22:56 +01:00
|
|
|
/* Allocate a return stack for newly created task */
|
2008-11-25 21:07:04 +01:00
|
|
|
void ftrace_graph_init_task(struct task_struct *t)
|
2008-11-23 06:22:56 +01:00
|
|
|
{
|
2009-06-02 22:51:55 +02:00
|
|
|
/* Make sure we do not use the parent ret_stack */
|
|
|
|
t->ret_stack = NULL;
|
2010-03-13 01:41:23 +01:00
|
|
|
t->curr_ret_stack = -1;
|
2009-06-02 22:51:55 +02:00
|
|
|
|
2009-04-03 21:24:12 +02:00
|
|
|
if (ftrace_graph_active) {
|
2009-06-02 18:26:07 +02:00
|
|
|
struct ftrace_ret_stack *ret_stack;
|
|
|
|
|
|
|
|
ret_stack = kmalloc(FTRACE_RETFUNC_DEPTH
|
2008-11-23 06:22:56 +01:00
|
|
|
* sizeof(struct ftrace_ret_stack),
|
|
|
|
GFP_KERNEL);
|
2009-06-02 18:26:07 +02:00
|
|
|
if (!ret_stack)
|
2008-11-23 06:22:56 +01:00
|
|
|
return;
|
ftrace: Fix memory leak with function graph and cpu hotplug
When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.
On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.
ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.
The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2011-02-11 03:26:13 +01:00
|
|
|
graph_init_task(t, ret_stack);
|
2009-06-02 22:51:55 +02:00
|
|
|
}
|
2008-11-23 06:22:56 +01:00
|
|
|
}
|
|
|
|
|
2008-11-25 21:07:04 +01:00
|
|
|
void ftrace_graph_exit_task(struct task_struct *t)
|
2008-11-23 06:22:56 +01:00
|
|
|
{
|
2008-11-23 17:33:12 +01:00
|
|
|
struct ftrace_ret_stack *ret_stack = t->ret_stack;
|
|
|
|
|
2008-11-23 06:22:56 +01:00
|
|
|
t->ret_stack = NULL;
|
2008-11-23 17:33:12 +01:00
|
|
|
/* NULL must become visible to IRQs before we free it: */
|
|
|
|
barrier();
|
|
|
|
|
|
|
|
kfree(ret_stack);
|
2008-11-23 06:22:56 +01:00
|
|
|
}
|
2008-11-11 07:14:25 +01:00
|
|
|
#endif
|