2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* net/sched/sch_api.c Packet scheduler API.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
* as published by the Free Software Foundation; either version
|
|
|
|
* 2 of the License, or (at your option) any later version.
|
|
|
|
*
|
|
|
|
* Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
|
|
|
|
*
|
|
|
|
* Fixes:
|
|
|
|
*
|
|
|
|
* Rani Assaf <rani@magic.metawire.com> :980802: JIFFIES and CPU clock sources are repaired.
|
|
|
|
* Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support
|
|
|
|
* Jamal Hadi Salim <hadi@nortelnetworks.com>: 990601: ingress support
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/types.h>
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/socket.h>
|
|
|
|
#include <linux/sockios.h>
|
|
|
|
#include <linux/in.h>
|
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/netdevice.h>
|
|
|
|
#include <linux/skbuff.h>
|
|
|
|
#include <linux/rtnetlink.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/proc_fs.h>
|
|
|
|
#include <linux/seq_file.h>
|
|
|
|
#include <linux/kmod.h>
|
|
|
|
#include <linux/list.h>
|
|
|
|
#include <linux/bitops.h>
|
|
|
|
|
|
|
|
#include <net/sock.h>
|
|
|
|
#include <net/pkt_sched.h>
|
|
|
|
|
|
|
|
#include <asm/processor.h>
|
|
|
|
#include <asm/uaccess.h>
|
|
|
|
#include <asm/system.h>
|
|
|
|
|
|
|
|
static int qdisc_notify(struct sk_buff *oskb, struct nlmsghdr *n, u32 clid,
|
|
|
|
struct Qdisc *old, struct Qdisc *new);
|
|
|
|
static int tclass_notify(struct sk_buff *oskb, struct nlmsghdr *n,
|
|
|
|
struct Qdisc *q, unsigned long cl, int event);
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
|
|
|
Short review.
|
|
|
|
-------------
|
|
|
|
|
|
|
|
This file consists of two interrelated parts:
|
|
|
|
|
|
|
|
1. queueing disciplines manager frontend.
|
|
|
|
2. traffic classes manager frontend.
|
|
|
|
|
|
|
|
Generally, queueing discipline ("qdisc") is a black box,
|
|
|
|
which is able to enqueue packets and to dequeue them (when
|
|
|
|
device is ready to send something) in order and at times
|
|
|
|
determined by algorithm hidden in it.
|
|
|
|
|
|
|
|
qdisc's are divided to two categories:
|
|
|
|
- "queues", which have no internal structure visible from outside.
|
|
|
|
- "schedulers", which split all the packets to "traffic classes",
|
|
|
|
using "packet classifiers" (look at cls_api.c)
|
|
|
|
|
|
|
|
In turn, classes may have child qdiscs (as rule, queues)
|
|
|
|
attached to them etc. etc. etc.
|
|
|
|
|
|
|
|
The goal of the routines in this file is to translate
|
|
|
|
information supplied by user in the form of handles
|
|
|
|
to more intelligible for kernel form, to make some sanity
|
|
|
|
checks and part of work, which is common to all qdiscs
|
|
|
|
and to provide rtnetlink notifications.
|
|
|
|
|
|
|
|
All real intelligent work is done inside qdisc modules.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Every discipline has two major routines: enqueue and dequeue.
|
|
|
|
|
|
|
|
---dequeue
|
|
|
|
|
|
|
|
dequeue usually returns a skb to send. It is allowed to return NULL,
|
|
|
|
but it does not mean that queue is empty, it just means that
|
|
|
|
discipline does not want to send anything this time.
|
|
|
|
Queue is really empty if q->q.qlen == 0.
|
|
|
|
For complicated disciplines with multiple queues q->q is not
|
|
|
|
real packet queue, but however q->q.qlen must be valid.
|
|
|
|
|
|
|
|
---enqueue
|
|
|
|
|
|
|
|
enqueue returns 0, if packet was enqueued successfully.
|
|
|
|
If packet (this one or another one) was dropped, it returns
|
|
|
|
not zero error code.
|
|
|
|
NET_XMIT_DROP - this packet dropped
|
|
|
|
Expected action: do not backoff, but wait until queue will clear.
|
|
|
|
NET_XMIT_CN - probably this packet enqueued, but another one dropped.
|
|
|
|
Expected action: backoff or ignore
|
|
|
|
NET_XMIT_POLICED - dropped by police.
|
|
|
|
Expected action: backoff or error to real-time apps.
|
|
|
|
|
|
|
|
Auxiliary routines:
|
|
|
|
|
|
|
|
---requeue
|
|
|
|
|
|
|
|
requeues once dequeued packet. It is used for non-standard or
|
|
|
|
just buggy devices, which can defer output even if dev->tbusy=0.
|
|
|
|
|
|
|
|
---reset
|
|
|
|
|
|
|
|
returns qdisc to initial state: purge all buffers, clear all
|
|
|
|
timers, counters (except for statistics) etc.
|
|
|
|
|
|
|
|
---init
|
|
|
|
|
|
|
|
initializes newly created qdisc.
|
|
|
|
|
|
|
|
---destroy
|
|
|
|
|
|
|
|
destroys resources allocated by init and during lifetime of qdisc.
|
|
|
|
|
|
|
|
---change
|
|
|
|
|
|
|
|
changes qdisc parameters.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Protects list of registered TC modules. It is pure SMP lock. */
|
|
|
|
static DEFINE_RWLOCK(qdisc_mod_lock);
|
|
|
|
|
|
|
|
|
|
|
|
/************************************************
|
|
|
|
* Queueing disciplines manipulation. *
|
|
|
|
************************************************/
|
|
|
|
|
|
|
|
|
|
|
|
/* The list of all installed queueing disciplines. */
|
|
|
|
|
|
|
|
static struct Qdisc_ops *qdisc_base;
|
|
|
|
|
|
|
|
/* Register/uregister queueing discipline */
|
|
|
|
|
|
|
|
int register_qdisc(struct Qdisc_ops *qops)
|
|
|
|
{
|
|
|
|
struct Qdisc_ops *q, **qp;
|
|
|
|
int rc = -EEXIST;
|
|
|
|
|
|
|
|
write_lock(&qdisc_mod_lock);
|
|
|
|
for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next)
|
|
|
|
if (!strcmp(qops->id, q->id))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (qops->enqueue == NULL)
|
|
|
|
qops->enqueue = noop_qdisc_ops.enqueue;
|
|
|
|
if (qops->requeue == NULL)
|
|
|
|
qops->requeue = noop_qdisc_ops.requeue;
|
|
|
|
if (qops->dequeue == NULL)
|
|
|
|
qops->dequeue = noop_qdisc_ops.dequeue;
|
|
|
|
|
|
|
|
qops->next = NULL;
|
|
|
|
*qp = qops;
|
|
|
|
rc = 0;
|
|
|
|
out:
|
|
|
|
write_unlock(&qdisc_mod_lock);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
int unregister_qdisc(struct Qdisc_ops *qops)
|
|
|
|
{
|
|
|
|
struct Qdisc_ops *q, **qp;
|
|
|
|
int err = -ENOENT;
|
|
|
|
|
|
|
|
write_lock(&qdisc_mod_lock);
|
|
|
|
for (qp = &qdisc_base; (q=*qp)!=NULL; qp = &q->next)
|
|
|
|
if (q == qops)
|
|
|
|
break;
|
|
|
|
if (q) {
|
|
|
|
*qp = q->next;
|
|
|
|
q->next = NULL;
|
|
|
|
err = 0;
|
|
|
|
}
|
|
|
|
write_unlock(&qdisc_mod_lock);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We know handle. Find qdisc among all qdisc's attached to device
|
|
|
|
(root qdisc, all its children, children of children etc.)
|
|
|
|
*/
|
|
|
|
|
2006-11-30 02:35:48 +01:00
|
|
|
static struct Qdisc *__qdisc_lookup(struct net_device *dev, u32 handle)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct Qdisc *q;
|
|
|
|
|
|
|
|
list_for_each_entry(q, &dev->qdisc_list, list) {
|
2006-11-30 02:35:48 +01:00
|
|
|
if (q->handle == handle)
|
2005-04-17 00:20:36 +02:00
|
|
|
return q;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2006-11-30 02:35:48 +01:00
|
|
|
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
|
|
|
|
{
|
|
|
|
struct Qdisc *q;
|
|
|
|
|
|
|
|
read_lock(&qdisc_tree_lock);
|
|
|
|
q = __qdisc_lookup(dev, handle);
|
|
|
|
read_unlock(&qdisc_tree_lock);
|
|
|
|
return q;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid)
|
|
|
|
{
|
|
|
|
unsigned long cl;
|
|
|
|
struct Qdisc *leaf;
|
|
|
|
struct Qdisc_class_ops *cops = p->ops->cl_ops;
|
|
|
|
|
|
|
|
if (cops == NULL)
|
|
|
|
return NULL;
|
|
|
|
cl = cops->get(p, classid);
|
|
|
|
|
|
|
|
if (cl == 0)
|
|
|
|
return NULL;
|
|
|
|
leaf = cops->leaf(p, cl);
|
|
|
|
cops->put(p, cl);
|
|
|
|
return leaf;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Find queueing discipline by name */
|
|
|
|
|
|
|
|
static struct Qdisc_ops *qdisc_lookup_ops(struct rtattr *kind)
|
|
|
|
{
|
|
|
|
struct Qdisc_ops *q = NULL;
|
|
|
|
|
|
|
|
if (kind) {
|
|
|
|
read_lock(&qdisc_mod_lock);
|
|
|
|
for (q = qdisc_base; q; q = q->next) {
|
|
|
|
if (rtattr_strcmp(kind, q->id) == 0) {
|
|
|
|
if (!try_module_get(q->owner))
|
|
|
|
q = NULL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
read_unlock(&qdisc_mod_lock);
|
|
|
|
}
|
|
|
|
return q;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct qdisc_rate_table *qdisc_rtab_list;
|
|
|
|
|
|
|
|
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r, struct rtattr *tab)
|
|
|
|
{
|
|
|
|
struct qdisc_rate_table *rtab;
|
|
|
|
|
|
|
|
for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) {
|
|
|
|
if (memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) == 0) {
|
|
|
|
rtab->refcnt++;
|
|
|
|
return rtab;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (tab == NULL || r->rate == 0 || r->cell_log == 0 || RTA_PAYLOAD(tab) != 1024)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
rtab = kmalloc(sizeof(*rtab), GFP_KERNEL);
|
|
|
|
if (rtab) {
|
|
|
|
rtab->rate = *r;
|
|
|
|
rtab->refcnt = 1;
|
|
|
|
memcpy(rtab->data, RTA_DATA(tab), 1024);
|
|
|
|
rtab->next = qdisc_rtab_list;
|
|
|
|
qdisc_rtab_list = rtab;
|
|
|
|
}
|
|
|
|
return rtab;
|
|
|
|
}
|
|
|
|
|
|
|
|
void qdisc_put_rtab(struct qdisc_rate_table *tab)
|
|
|
|
{
|
|
|
|
struct qdisc_rate_table *rtab, **rtabp;
|
|
|
|
|
|
|
|
if (!tab || --tab->refcnt)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for (rtabp = &qdisc_rtab_list; (rtab=*rtabp) != NULL; rtabp = &rtab->next) {
|
|
|
|
if (rtab == tab) {
|
|
|
|
*rtabp = rtab->next;
|
|
|
|
kfree(rtab);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Allocate an unique handle from space managed by kernel */
|
|
|
|
|
|
|
|
static u32 qdisc_alloc_handle(struct net_device *dev)
|
|
|
|
{
|
|
|
|
int i = 0x10000;
|
|
|
|
static u32 autohandle = TC_H_MAKE(0x80000000U, 0);
|
|
|
|
|
|
|
|
do {
|
|
|
|
autohandle += TC_H_MAKE(0x10000U, 0);
|
|
|
|
if (autohandle == TC_H_MAKE(TC_H_ROOT, 0))
|
|
|
|
autohandle = TC_H_MAKE(0x80000000U, 0);
|
|
|
|
} while (qdisc_lookup(dev, autohandle) && --i > 0);
|
|
|
|
|
|
|
|
return i>0 ? autohandle : 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Attach toplevel qdisc to device dev */
|
|
|
|
|
|
|
|
static struct Qdisc *
|
|
|
|
dev_graft_qdisc(struct net_device *dev, struct Qdisc *qdisc)
|
|
|
|
{
|
|
|
|
struct Qdisc *oqdisc;
|
|
|
|
|
|
|
|
if (dev->flags & IFF_UP)
|
|
|
|
dev_deactivate(dev);
|
|
|
|
|
|
|
|
qdisc_lock_tree(dev);
|
|
|
|
if (qdisc && qdisc->flags&TCQ_F_INGRESS) {
|
|
|
|
oqdisc = dev->qdisc_ingress;
|
|
|
|
/* Prune old scheduler */
|
|
|
|
if (oqdisc && atomic_read(&oqdisc->refcnt) <= 1) {
|
|
|
|
/* delete */
|
|
|
|
qdisc_reset(oqdisc);
|
|
|
|
dev->qdisc_ingress = NULL;
|
|
|
|
} else { /* new */
|
|
|
|
dev->qdisc_ingress = qdisc;
|
|
|
|
}
|
|
|
|
|
|
|
|
} else {
|
|
|
|
|
|
|
|
oqdisc = dev->qdisc_sleeping;
|
|
|
|
|
|
|
|
/* Prune old scheduler */
|
|
|
|
if (oqdisc && atomic_read(&oqdisc->refcnt) <= 1)
|
|
|
|
qdisc_reset(oqdisc);
|
|
|
|
|
|
|
|
/* ... and graft new one */
|
|
|
|
if (qdisc == NULL)
|
|
|
|
qdisc = &noop_qdisc;
|
|
|
|
dev->qdisc_sleeping = qdisc;
|
|
|
|
dev->qdisc = &noop_qdisc;
|
|
|
|
}
|
|
|
|
|
|
|
|
qdisc_unlock_tree(dev);
|
|
|
|
|
|
|
|
if (dev->flags & IFF_UP)
|
|
|
|
dev_activate(dev);
|
|
|
|
|
|
|
|
return oqdisc;
|
|
|
|
}
|
|
|
|
|
2006-11-30 02:35:48 +01:00
|
|
|
void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n)
|
|
|
|
{
|
|
|
|
struct Qdisc_class_ops *cops;
|
|
|
|
unsigned long cl;
|
|
|
|
u32 parentid;
|
|
|
|
|
|
|
|
if (n == 0)
|
|
|
|
return;
|
|
|
|
while ((parentid = sch->parent)) {
|
|
|
|
sch = __qdisc_lookup(sch->dev, TC_H_MAJ(parentid));
|
|
|
|
cops = sch->ops->cl_ops;
|
|
|
|
if (cops->qlen_notify) {
|
|
|
|
cl = cops->get(sch, parentid);
|
|
|
|
cops->qlen_notify(sch, cl);
|
|
|
|
cops->put(sch, cl);
|
|
|
|
}
|
|
|
|
sch->q.qlen -= n;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(qdisc_tree_decrease_qlen);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* Graft qdisc "new" to class "classid" of qdisc "parent" or
|
|
|
|
to device "dev".
|
|
|
|
|
|
|
|
Old qdisc is not destroyed but returned in *old.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int qdisc_graft(struct net_device *dev, struct Qdisc *parent,
|
|
|
|
u32 classid,
|
|
|
|
struct Qdisc *new, struct Qdisc **old)
|
|
|
|
{
|
|
|
|
int err = 0;
|
|
|
|
struct Qdisc *q = *old;
|
|
|
|
|
|
|
|
|
2007-02-09 15:25:16 +01:00
|
|
|
if (parent == NULL) {
|
2005-04-17 00:20:36 +02:00
|
|
|
if (q && q->flags&TCQ_F_INGRESS) {
|
|
|
|
*old = dev_graft_qdisc(dev, q);
|
|
|
|
} else {
|
|
|
|
*old = dev_graft_qdisc(dev, new);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
struct Qdisc_class_ops *cops = parent->ops->cl_ops;
|
|
|
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
|
|
|
if (cops) {
|
|
|
|
unsigned long cl = cops->get(parent, classid);
|
|
|
|
if (cl) {
|
|
|
|
err = cops->graft(parent, cl, new, old);
|
|
|
|
if (new)
|
|
|
|
new->parent = classid;
|
|
|
|
cops->put(parent, cl);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
Allocate and initialize new qdisc.
|
|
|
|
|
|
|
|
Parameters are passed via opt.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static struct Qdisc *
|
|
|
|
qdisc_create(struct net_device *dev, u32 handle, struct rtattr **tca, int *errp)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct rtattr *kind = tca[TCA_KIND-1];
|
|
|
|
struct Qdisc *sch;
|
|
|
|
struct Qdisc_ops *ops;
|
|
|
|
|
|
|
|
ops = qdisc_lookup_ops(kind);
|
|
|
|
#ifdef CONFIG_KMOD
|
|
|
|
if (ops == NULL && kind != NULL) {
|
|
|
|
char name[IFNAMSIZ];
|
|
|
|
if (rtattr_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) {
|
|
|
|
/* We dropped the RTNL semaphore in order to
|
|
|
|
* perform the module load. So, even if we
|
|
|
|
* succeeded in loading the module we have to
|
|
|
|
* tell the caller to replay the request. We
|
|
|
|
* indicate this using -EAGAIN.
|
|
|
|
* We replay the request because the device may
|
|
|
|
* go away in the mean time.
|
|
|
|
*/
|
|
|
|
rtnl_unlock();
|
|
|
|
request_module("sch_%s", name);
|
|
|
|
rtnl_lock();
|
|
|
|
ops = qdisc_lookup_ops(kind);
|
|
|
|
if (ops != NULL) {
|
|
|
|
/* We will try again qdisc_lookup_ops,
|
|
|
|
* so don't keep a reference.
|
|
|
|
*/
|
|
|
|
module_put(ops->owner);
|
|
|
|
err = -EAGAIN;
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2006-08-04 01:36:51 +02:00
|
|
|
err = -ENOENT;
|
2005-04-17 00:20:36 +02:00
|
|
|
if (ops == NULL)
|
|
|
|
goto err_out;
|
|
|
|
|
2005-07-05 23:15:09 +02:00
|
|
|
sch = qdisc_alloc(dev, ops);
|
|
|
|
if (IS_ERR(sch)) {
|
|
|
|
err = PTR_ERR(sch);
|
2005-04-17 00:20:36 +02:00
|
|
|
goto err_out2;
|
2005-07-05 23:15:09 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-07-05 23:15:09 +02:00
|
|
|
if (handle == TC_H_INGRESS) {
|
2005-04-17 00:20:36 +02:00
|
|
|
sch->flags |= TCQ_F_INGRESS;
|
2005-07-05 23:15:09 +02:00
|
|
|
handle = TC_H_MAKE(TC_H_INGRESS, 0);
|
|
|
|
} else if (handle == 0) {
|
2005-04-17 00:20:36 +02:00
|
|
|
handle = qdisc_alloc_handle(dev);
|
|
|
|
err = -ENOMEM;
|
|
|
|
if (handle == 0)
|
|
|
|
goto err_out3;
|
|
|
|
}
|
|
|
|
|
2005-07-05 23:15:09 +02:00
|
|
|
sch->handle = handle;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS-1])) == 0) {
|
2005-07-05 23:15:53 +02:00
|
|
|
#ifdef CONFIG_NET_ESTIMATOR
|
|
|
|
if (tca[TCA_RATE-1]) {
|
|
|
|
err = gen_new_estimator(&sch->bstats, &sch->rate_est,
|
|
|
|
sch->stats_lock,
|
|
|
|
tca[TCA_RATE-1]);
|
|
|
|
if (err) {
|
|
|
|
/*
|
|
|
|
* Any broken qdiscs that would require
|
|
|
|
* a ops->reset() here? The qdisc was never
|
|
|
|
* in action so it shouldn't be necessary.
|
|
|
|
*/
|
|
|
|
if (ops->destroy)
|
|
|
|
ops->destroy(sch);
|
|
|
|
goto err_out3;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
2005-04-17 00:20:36 +02:00
|
|
|
qdisc_lock_tree(dev);
|
|
|
|
list_add_tail(&sch->list, &dev->qdisc_list);
|
|
|
|
qdisc_unlock_tree(dev);
|
|
|
|
|
|
|
|
return sch;
|
|
|
|
}
|
|
|
|
err_out3:
|
|
|
|
dev_put(dev);
|
2005-07-05 23:15:09 +02:00
|
|
|
kfree((char *) sch - sch->padded);
|
2005-04-17 00:20:36 +02:00
|
|
|
err_out2:
|
|
|
|
module_put(ops->owner);
|
|
|
|
err_out:
|
|
|
|
*errp = err;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qdisc_change(struct Qdisc *sch, struct rtattr **tca)
|
|
|
|
{
|
|
|
|
if (tca[TCA_OPTIONS-1]) {
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (sch->ops->change == NULL)
|
|
|
|
return -EINVAL;
|
|
|
|
err = sch->ops->change(sch, tca[TCA_OPTIONS-1]);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
#ifdef CONFIG_NET_ESTIMATOR
|
|
|
|
if (tca[TCA_RATE-1])
|
|
|
|
gen_replace_estimator(&sch->bstats, &sch->rate_est,
|
|
|
|
sch->stats_lock, tca[TCA_RATE-1]);
|
|
|
|
#endif
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct check_loop_arg
|
|
|
|
{
|
|
|
|
struct qdisc_walker w;
|
|
|
|
struct Qdisc *p;
|
|
|
|
int depth;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w);
|
|
|
|
|
|
|
|
static int check_loop(struct Qdisc *q, struct Qdisc *p, int depth)
|
|
|
|
{
|
|
|
|
struct check_loop_arg arg;
|
|
|
|
|
|
|
|
if (q->ops->cl_ops == NULL)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
arg.w.stop = arg.w.skip = arg.w.count = 0;
|
|
|
|
arg.w.fn = check_loop_fn;
|
|
|
|
arg.depth = depth;
|
|
|
|
arg.p = p;
|
|
|
|
q->ops->cl_ops->walk(q, &arg.w);
|
|
|
|
return arg.w.stop ? -ELOOP : 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w)
|
|
|
|
{
|
|
|
|
struct Qdisc *leaf;
|
|
|
|
struct Qdisc_class_ops *cops = q->ops->cl_ops;
|
|
|
|
struct check_loop_arg *arg = (struct check_loop_arg *)w;
|
|
|
|
|
|
|
|
leaf = cops->leaf(q, cl);
|
|
|
|
if (leaf) {
|
|
|
|
if (leaf == arg->p || arg->depth > 7)
|
|
|
|
return -ELOOP;
|
|
|
|
return check_loop(leaf, arg->p, arg->depth + 1);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Delete/get qdisc.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
|
|
|
|
{
|
|
|
|
struct tcmsg *tcm = NLMSG_DATA(n);
|
|
|
|
struct rtattr **tca = arg;
|
|
|
|
struct net_device *dev;
|
|
|
|
u32 clid = tcm->tcm_parent;
|
|
|
|
struct Qdisc *q = NULL;
|
|
|
|
struct Qdisc *p = NULL;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if ((dev = __dev_get_by_index(tcm->tcm_ifindex)) == NULL)
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
if (clid) {
|
|
|
|
if (clid != TC_H_ROOT) {
|
|
|
|
if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) {
|
|
|
|
if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL)
|
|
|
|
return -ENOENT;
|
|
|
|
q = qdisc_leaf(p, clid);
|
|
|
|
} else { /* ingress */
|
|
|
|
q = dev->qdisc_ingress;
|
2007-02-09 15:25:16 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
} else {
|
|
|
|
q = dev->qdisc_sleeping;
|
|
|
|
}
|
|
|
|
if (!q)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
if (tcm->tcm_handle && q->handle != tcm->tcm_handle)
|
|
|
|
return -EINVAL;
|
|
|
|
} else {
|
|
|
|
if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL)
|
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (tca[TCA_KIND-1] && rtattr_strcmp(tca[TCA_KIND-1], q->ops->id))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (n->nlmsg_type == RTM_DELQDISC) {
|
|
|
|
if (!clid)
|
|
|
|
return -EINVAL;
|
|
|
|
if (q->handle == 0)
|
|
|
|
return -ENOENT;
|
|
|
|
if ((err = qdisc_graft(dev, p, clid, NULL, &q)) != 0)
|
|
|
|
return err;
|
|
|
|
if (q) {
|
|
|
|
qdisc_notify(skb, n, clid, q, NULL);
|
|
|
|
spin_lock_bh(&dev->queue_lock);
|
|
|
|
qdisc_destroy(q);
|
|
|
|
spin_unlock_bh(&dev->queue_lock);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
qdisc_notify(skb, n, clid, NULL, q);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
Create/change qdisc.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
|
|
|
|
{
|
|
|
|
struct tcmsg *tcm;
|
|
|
|
struct rtattr **tca;
|
|
|
|
struct net_device *dev;
|
|
|
|
u32 clid;
|
|
|
|
struct Qdisc *q, *p;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
replay:
|
|
|
|
/* Reinit, just in case something touches this. */
|
|
|
|
tcm = NLMSG_DATA(n);
|
|
|
|
tca = arg;
|
|
|
|
clid = tcm->tcm_parent;
|
|
|
|
q = p = NULL;
|
|
|
|
|
|
|
|
if ((dev = __dev_get_by_index(tcm->tcm_ifindex)) == NULL)
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
if (clid) {
|
|
|
|
if (clid != TC_H_ROOT) {
|
|
|
|
if (clid != TC_H_INGRESS) {
|
|
|
|
if ((p = qdisc_lookup(dev, TC_H_MAJ(clid))) == NULL)
|
|
|
|
return -ENOENT;
|
|
|
|
q = qdisc_leaf(p, clid);
|
|
|
|
} else { /*ingress */
|
|
|
|
q = dev->qdisc_ingress;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
q = dev->qdisc_sleeping;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* It may be default qdisc, ignore it */
|
|
|
|
if (q && q->handle == 0)
|
|
|
|
q = NULL;
|
|
|
|
|
|
|
|
if (!q || !tcm->tcm_handle || q->handle != tcm->tcm_handle) {
|
|
|
|
if (tcm->tcm_handle) {
|
|
|
|
if (q && !(n->nlmsg_flags&NLM_F_REPLACE))
|
|
|
|
return -EEXIST;
|
|
|
|
if (TC_H_MIN(tcm->tcm_handle))
|
|
|
|
return -EINVAL;
|
|
|
|
if ((q = qdisc_lookup(dev, tcm->tcm_handle)) == NULL)
|
|
|
|
goto create_n_graft;
|
|
|
|
if (n->nlmsg_flags&NLM_F_EXCL)
|
|
|
|
return -EEXIST;
|
|
|
|
if (tca[TCA_KIND-1] && rtattr_strcmp(tca[TCA_KIND-1], q->ops->id))
|
|
|
|
return -EINVAL;
|
|
|
|
if (q == p ||
|
|
|
|
(p && check_loop(q, p, 0)))
|
|
|
|
return -ELOOP;
|
|
|
|
atomic_inc(&q->refcnt);
|
|
|
|
goto graft;
|
|
|
|
} else {
|
|
|
|
if (q == NULL)
|
|
|
|
goto create_n_graft;
|
|
|
|
|
|
|
|
/* This magic test requires explanation.
|
|
|
|
*
|
|
|
|
* We know, that some child q is already
|
|
|
|
* attached to this parent and have choice:
|
|
|
|
* either to change it or to create/graft new one.
|
|
|
|
*
|
|
|
|
* 1. We are allowed to create/graft only
|
|
|
|
* if CREATE and REPLACE flags are set.
|
|
|
|
*
|
|
|
|
* 2. If EXCL is set, requestor wanted to say,
|
|
|
|
* that qdisc tcm_handle is not expected
|
|
|
|
* to exist, so that we choose create/graft too.
|
|
|
|
*
|
|
|
|
* 3. The last case is when no flags are set.
|
|
|
|
* Alas, it is sort of hole in API, we
|
|
|
|
* cannot decide what to do unambiguously.
|
|
|
|
* For now we select create/graft, if
|
|
|
|
* user gave KIND, which does not match existing.
|
|
|
|
*/
|
|
|
|
if ((n->nlmsg_flags&NLM_F_CREATE) &&
|
|
|
|
(n->nlmsg_flags&NLM_F_REPLACE) &&
|
|
|
|
((n->nlmsg_flags&NLM_F_EXCL) ||
|
|
|
|
(tca[TCA_KIND-1] &&
|
|
|
|
rtattr_strcmp(tca[TCA_KIND-1], q->ops->id))))
|
|
|
|
goto create_n_graft;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (!tcm->tcm_handle)
|
|
|
|
return -EINVAL;
|
|
|
|
q = qdisc_lookup(dev, tcm->tcm_handle);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Change qdisc parameters */
|
|
|
|
if (q == NULL)
|
|
|
|
return -ENOENT;
|
|
|
|
if (n->nlmsg_flags&NLM_F_EXCL)
|
|
|
|
return -EEXIST;
|
|
|
|
if (tca[TCA_KIND-1] && rtattr_strcmp(tca[TCA_KIND-1], q->ops->id))
|
|
|
|
return -EINVAL;
|
|
|
|
err = qdisc_change(q, tca);
|
|
|
|
if (err == 0)
|
|
|
|
qdisc_notify(skb, n, clid, NULL, q);
|
|
|
|
return err;
|
|
|
|
|
|
|
|
create_n_graft:
|
|
|
|
if (!(n->nlmsg_flags&NLM_F_CREATE))
|
|
|
|
return -ENOENT;
|
|
|
|
if (clid == TC_H_INGRESS)
|
|
|
|
q = qdisc_create(dev, tcm->tcm_parent, tca, &err);
|
2007-02-09 15:25:16 +01:00
|
|
|
else
|
2005-04-17 00:20:36 +02:00
|
|
|
q = qdisc_create(dev, tcm->tcm_handle, tca, &err);
|
|
|
|
if (q == NULL) {
|
|
|
|
if (err == -EAGAIN)
|
|
|
|
goto replay;
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
graft:
|
|
|
|
if (1) {
|
|
|
|
struct Qdisc *old_q = NULL;
|
|
|
|
err = qdisc_graft(dev, p, clid, q, &old_q);
|
|
|
|
if (err) {
|
|
|
|
if (q) {
|
|
|
|
spin_lock_bh(&dev->queue_lock);
|
|
|
|
qdisc_destroy(q);
|
|
|
|
spin_unlock_bh(&dev->queue_lock);
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
qdisc_notify(skb, n, clid, old_q, q);
|
|
|
|
if (old_q) {
|
|
|
|
spin_lock_bh(&dev->queue_lock);
|
|
|
|
qdisc_destroy(old_q);
|
|
|
|
spin_unlock_bh(&dev->queue_lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid,
|
2005-06-19 07:55:31 +02:00
|
|
|
u32 pid, u32 seq, u16 flags, int event)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct tcmsg *tcm;
|
|
|
|
struct nlmsghdr *nlh;
|
|
|
|
unsigned char *b = skb->tail;
|
|
|
|
struct gnet_dump d;
|
|
|
|
|
2005-06-19 07:55:31 +02:00
|
|
|
nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
tcm = NLMSG_DATA(nlh);
|
|
|
|
tcm->tcm_family = AF_UNSPEC;
|
2005-06-28 21:55:30 +02:00
|
|
|
tcm->tcm__pad1 = 0;
|
|
|
|
tcm->tcm__pad2 = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
tcm->tcm_ifindex = q->dev->ifindex;
|
|
|
|
tcm->tcm_parent = clid;
|
|
|
|
tcm->tcm_handle = q->handle;
|
|
|
|
tcm->tcm_info = atomic_read(&q->refcnt);
|
|
|
|
RTA_PUT(skb, TCA_KIND, IFNAMSIZ, q->ops->id);
|
|
|
|
if (q->ops->dump && q->ops->dump(q, skb) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
q->qstats.qlen = q->q.qlen;
|
|
|
|
|
|
|
|
if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS,
|
|
|
|
TCA_XSTATS, q->stats_lock, &d) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
if (gnet_stats_copy_basic(&d, &q->bstats) < 0 ||
|
|
|
|
#ifdef CONFIG_NET_ESTIMATOR
|
|
|
|
gnet_stats_copy_rate_est(&d, &q->rate_est) < 0 ||
|
|
|
|
#endif
|
|
|
|
gnet_stats_copy_queue(&d, &q->qstats) < 0)
|
|
|
|
goto rtattr_failure;
|
2007-02-09 15:25:16 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
if (gnet_stats_finish_copy(&d) < 0)
|
|
|
|
goto rtattr_failure;
|
2007-02-09 15:25:16 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
nlh->nlmsg_len = skb->tail - b;
|
|
|
|
return skb->len;
|
|
|
|
|
|
|
|
nlmsg_failure:
|
|
|
|
rtattr_failure:
|
|
|
|
skb_trim(skb, b - skb->data);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qdisc_notify(struct sk_buff *oskb, struct nlmsghdr *n,
|
|
|
|
u32 clid, struct Qdisc *old, struct Qdisc *new)
|
|
|
|
{
|
|
|
|
struct sk_buff *skb;
|
|
|
|
u32 pid = oskb ? NETLINK_CB(oskb).pid : 0;
|
|
|
|
|
|
|
|
skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
|
|
|
|
if (!skb)
|
|
|
|
return -ENOBUFS;
|
|
|
|
|
|
|
|
if (old && old->handle) {
|
|
|
|
if (tc_fill_qdisc(skb, old, clid, pid, n->nlmsg_seq, 0, RTM_DELQDISC) < 0)
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
if (new) {
|
|
|
|
if (tc_fill_qdisc(skb, new, clid, pid, n->nlmsg_seq, old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0)
|
|
|
|
goto err_out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (skb->len)
|
2005-08-15 04:29:52 +02:00
|
|
|
return rtnetlink_send(skb, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
err_out:
|
|
|
|
kfree_skb(skb);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb)
|
|
|
|
{
|
|
|
|
int idx, q_idx;
|
|
|
|
int s_idx, s_q_idx;
|
|
|
|
struct net_device *dev;
|
|
|
|
struct Qdisc *q;
|
|
|
|
|
|
|
|
s_idx = cb->args[0];
|
|
|
|
s_q_idx = q_idx = cb->args[1];
|
|
|
|
read_lock(&dev_base_lock);
|
|
|
|
for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
|
|
|
|
if (idx < s_idx)
|
|
|
|
continue;
|
|
|
|
if (idx > s_idx)
|
|
|
|
s_q_idx = 0;
|
[NET_SCHED]: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-09-28 01:45:45 +02:00
|
|
|
read_lock(&qdisc_tree_lock);
|
2005-04-17 00:20:36 +02:00
|
|
|
q_idx = 0;
|
|
|
|
list_for_each_entry(q, &dev->qdisc_list, list) {
|
|
|
|
if (q_idx < s_q_idx) {
|
|
|
|
q_idx++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).pid,
|
|
|
|
cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0) {
|
[NET_SCHED]: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-09-28 01:45:45 +02:00
|
|
|
read_unlock(&qdisc_tree_lock);
|
2005-04-17 00:20:36 +02:00
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
q_idx++;
|
|
|
|
}
|
[NET_SCHED]: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-09-28 01:45:45 +02:00
|
|
|
read_unlock(&qdisc_tree_lock);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
done:
|
|
|
|
read_unlock(&dev_base_lock);
|
|
|
|
|
|
|
|
cb->args[0] = idx;
|
|
|
|
cb->args[1] = q_idx;
|
|
|
|
|
|
|
|
return skb->len;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/************************************************
|
|
|
|
* Traffic classes manipulation. *
|
|
|
|
************************************************/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
|
|
|
|
{
|
|
|
|
struct tcmsg *tcm = NLMSG_DATA(n);
|
|
|
|
struct rtattr **tca = arg;
|
|
|
|
struct net_device *dev;
|
|
|
|
struct Qdisc *q = NULL;
|
|
|
|
struct Qdisc_class_ops *cops;
|
|
|
|
unsigned long cl = 0;
|
|
|
|
unsigned long new_cl;
|
|
|
|
u32 pid = tcm->tcm_parent;
|
|
|
|
u32 clid = tcm->tcm_handle;
|
|
|
|
u32 qid = TC_H_MAJ(clid);
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if ((dev = __dev_get_by_index(tcm->tcm_ifindex)) == NULL)
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
/*
|
|
|
|
parent == TC_H_UNSPEC - unspecified parent.
|
|
|
|
parent == TC_H_ROOT - class is root, which has no parent.
|
|
|
|
parent == X:0 - parent is root class.
|
|
|
|
parent == X:Y - parent is a node in hierarchy.
|
|
|
|
parent == 0:Y - parent is X:Y, where X:0 is qdisc.
|
|
|
|
|
|
|
|
handle == 0:0 - generate handle from kernel pool.
|
|
|
|
handle == 0:Y - class is X:Y, where X:0 is qdisc.
|
|
|
|
handle == X:Y - clear.
|
|
|
|
handle == X:0 - root class.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Step 1. Determine qdisc handle X:0 */
|
|
|
|
|
|
|
|
if (pid != TC_H_ROOT) {
|
|
|
|
u32 qid1 = TC_H_MAJ(pid);
|
|
|
|
|
|
|
|
if (qid && qid1) {
|
|
|
|
/* If both majors are known, they must be identical. */
|
|
|
|
if (qid != qid1)
|
|
|
|
return -EINVAL;
|
|
|
|
} else if (qid1) {
|
|
|
|
qid = qid1;
|
|
|
|
} else if (qid == 0)
|
|
|
|
qid = dev->qdisc_sleeping->handle;
|
|
|
|
|
|
|
|
/* Now qid is genuine qdisc handle consistent
|
|
|
|
both with parent and child.
|
|
|
|
|
|
|
|
TC_H_MAJ(pid) still may be unspecified, complete it now.
|
|
|
|
*/
|
|
|
|
if (pid)
|
|
|
|
pid = TC_H_MAKE(qid, pid);
|
|
|
|
} else {
|
|
|
|
if (qid == 0)
|
|
|
|
qid = dev->qdisc_sleeping->handle;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* OK. Locate qdisc */
|
2007-02-09 15:25:16 +01:00
|
|
|
if ((q = qdisc_lookup(dev, qid)) == NULL)
|
2005-04-17 00:20:36 +02:00
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
/* An check that it supports classes */
|
|
|
|
cops = q->ops->cl_ops;
|
|
|
|
if (cops == NULL)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* Now try to get class */
|
|
|
|
if (clid == 0) {
|
|
|
|
if (pid == TC_H_ROOT)
|
|
|
|
clid = qid;
|
|
|
|
} else
|
|
|
|
clid = TC_H_MAKE(qid, clid);
|
|
|
|
|
|
|
|
if (clid)
|
|
|
|
cl = cops->get(q, clid);
|
|
|
|
|
|
|
|
if (cl == 0) {
|
|
|
|
err = -ENOENT;
|
|
|
|
if (n->nlmsg_type != RTM_NEWTCLASS || !(n->nlmsg_flags&NLM_F_CREATE))
|
|
|
|
goto out;
|
|
|
|
} else {
|
|
|
|
switch (n->nlmsg_type) {
|
2007-02-09 15:25:16 +01:00
|
|
|
case RTM_NEWTCLASS:
|
2005-04-17 00:20:36 +02:00
|
|
|
err = -EEXIST;
|
|
|
|
if (n->nlmsg_flags&NLM_F_EXCL)
|
|
|
|
goto out;
|
|
|
|
break;
|
|
|
|
case RTM_DELTCLASS:
|
|
|
|
err = cops->delete(q, cl);
|
|
|
|
if (err == 0)
|
|
|
|
tclass_notify(skb, n, q, cl, RTM_DELTCLASS);
|
|
|
|
goto out;
|
|
|
|
case RTM_GETTCLASS:
|
|
|
|
err = tclass_notify(skb, n, q, cl, RTM_NEWTCLASS);
|
|
|
|
goto out;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
new_cl = cl;
|
|
|
|
err = cops->change(q, clid, pid, tca, &new_cl);
|
|
|
|
if (err == 0)
|
|
|
|
tclass_notify(skb, n, q, new_cl, RTM_NEWTCLASS);
|
|
|
|
|
|
|
|
out:
|
|
|
|
if (cl)
|
|
|
|
cops->put(q, cl);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q,
|
|
|
|
unsigned long cl,
|
2005-06-19 07:55:31 +02:00
|
|
|
u32 pid, u32 seq, u16 flags, int event)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct tcmsg *tcm;
|
|
|
|
struct nlmsghdr *nlh;
|
|
|
|
unsigned char *b = skb->tail;
|
|
|
|
struct gnet_dump d;
|
|
|
|
struct Qdisc_class_ops *cl_ops = q->ops->cl_ops;
|
|
|
|
|
2005-06-19 07:55:31 +02:00
|
|
|
nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*tcm), flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
tcm = NLMSG_DATA(nlh);
|
|
|
|
tcm->tcm_family = AF_UNSPEC;
|
|
|
|
tcm->tcm_ifindex = q->dev->ifindex;
|
|
|
|
tcm->tcm_parent = q->handle;
|
|
|
|
tcm->tcm_handle = q->handle;
|
|
|
|
tcm->tcm_info = 0;
|
|
|
|
RTA_PUT(skb, TCA_KIND, IFNAMSIZ, q->ops->id);
|
|
|
|
if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS,
|
|
|
|
TCA_XSTATS, q->stats_lock, &d) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
if (gnet_stats_finish_copy(&d) < 0)
|
|
|
|
goto rtattr_failure;
|
|
|
|
|
|
|
|
nlh->nlmsg_len = skb->tail - b;
|
|
|
|
return skb->len;
|
|
|
|
|
|
|
|
nlmsg_failure:
|
|
|
|
rtattr_failure:
|
|
|
|
skb_trim(skb, b - skb->data);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tclass_notify(struct sk_buff *oskb, struct nlmsghdr *n,
|
|
|
|
struct Qdisc *q, unsigned long cl, int event)
|
|
|
|
{
|
|
|
|
struct sk_buff *skb;
|
|
|
|
u32 pid = oskb ? NETLINK_CB(oskb).pid : 0;
|
|
|
|
|
|
|
|
skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
|
|
|
|
if (!skb)
|
|
|
|
return -ENOBUFS;
|
|
|
|
|
|
|
|
if (tc_fill_tclass(skb, q, cl, pid, n->nlmsg_seq, 0, event) < 0) {
|
|
|
|
kfree_skb(skb);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2005-08-15 04:29:52 +02:00
|
|
|
return rtnetlink_send(skb, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
struct qdisc_dump_args
|
|
|
|
{
|
|
|
|
struct qdisc_walker w;
|
|
|
|
struct sk_buff *skb;
|
|
|
|
struct netlink_callback *cb;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int qdisc_class_dump(struct Qdisc *q, unsigned long cl, struct qdisc_walker *arg)
|
|
|
|
{
|
|
|
|
struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg;
|
|
|
|
|
|
|
|
return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).pid,
|
|
|
|
a->cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWTCLASS);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb)
|
|
|
|
{
|
|
|
|
int t;
|
|
|
|
int s_t;
|
|
|
|
struct net_device *dev;
|
|
|
|
struct Qdisc *q;
|
|
|
|
struct tcmsg *tcm = (struct tcmsg*)NLMSG_DATA(cb->nlh);
|
|
|
|
struct qdisc_dump_args arg;
|
|
|
|
|
|
|
|
if (cb->nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*tcm)))
|
|
|
|
return 0;
|
|
|
|
if ((dev = dev_get_by_index(tcm->tcm_ifindex)) == NULL)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
s_t = cb->args[0];
|
|
|
|
t = 0;
|
|
|
|
|
[NET_SCHED]: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-09-28 01:45:45 +02:00
|
|
|
read_lock(&qdisc_tree_lock);
|
2005-04-17 00:20:36 +02:00
|
|
|
list_for_each_entry(q, &dev->qdisc_list, list) {
|
|
|
|
if (t < s_t || !q->ops->cl_ops ||
|
|
|
|
(tcm->tcm_parent &&
|
|
|
|
TC_H_MAJ(tcm->tcm_parent) != q->handle)) {
|
|
|
|
t++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (t > s_t)
|
|
|
|
memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0]));
|
|
|
|
arg.w.fn = qdisc_class_dump;
|
|
|
|
arg.skb = skb;
|
|
|
|
arg.cb = cb;
|
|
|
|
arg.w.stop = 0;
|
|
|
|
arg.w.skip = cb->args[1];
|
|
|
|
arg.w.count = 0;
|
|
|
|
q->ops->cl_ops->walk(q, &arg.w);
|
|
|
|
cb->args[1] = arg.w.count;
|
|
|
|
if (arg.w.stop)
|
|
|
|
break;
|
|
|
|
t++;
|
|
|
|
}
|
[NET_SCHED]: Fix fallout from dev->qdisc RCU change
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2006-09-28 01:45:45 +02:00
|
|
|
read_unlock(&qdisc_tree_lock);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
cb->args[0] = t;
|
|
|
|
|
|
|
|
dev_put(dev);
|
|
|
|
return skb->len;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Main classifier routine: scans classifier chain attached
|
|
|
|
to this qdisc, (optionally) tests for protocol and asks
|
|
|
|
specific classifiers.
|
|
|
|
*/
|
|
|
|
int tc_classify(struct sk_buff *skb, struct tcf_proto *tp,
|
|
|
|
struct tcf_result *res)
|
|
|
|
{
|
|
|
|
int err = 0;
|
2006-11-21 03:07:51 +01:00
|
|
|
__be16 protocol = skb->protocol;
|
2005-04-17 00:20:36 +02:00
|
|
|
#ifdef CONFIG_NET_CLS_ACT
|
|
|
|
struct tcf_proto *otp = tp;
|
|
|
|
reclassify:
|
|
|
|
#endif
|
|
|
|
protocol = skb->protocol;
|
|
|
|
|
|
|
|
for ( ; tp; tp = tp->next) {
|
|
|
|
if ((tp->protocol == protocol ||
|
|
|
|
tp->protocol == __constant_htons(ETH_P_ALL)) &&
|
|
|
|
(err = tp->classify(skb, tp, res)) >= 0) {
|
|
|
|
#ifdef CONFIG_NET_CLS_ACT
|
|
|
|
if ( TC_ACT_RECLASSIFY == err) {
|
|
|
|
__u32 verd = (__u32) G_TC_VERD(skb->tc_verd);
|
|
|
|
tp = otp;
|
|
|
|
|
|
|
|
if (MAX_REC_LOOP < verd++) {
|
|
|
|
printk("rule prio %d protocol %02x reclassify is buggy packet dropped\n",
|
|
|
|
tp->prio&0xffff, ntohs(tp->protocol));
|
|
|
|
return TC_ACT_SHOT;
|
|
|
|
}
|
|
|
|
skb->tc_verd = SET_TC_VERD(skb->tc_verd,verd);
|
|
|
|
goto reclassify;
|
|
|
|
} else {
|
2007-02-09 15:25:16 +01:00
|
|
|
if (skb->tc_verd)
|
2005-04-17 00:20:36 +02:00
|
|
|
skb->tc_verd = SET_TC_VERD(skb->tc_verd,0);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
|
|
|
|
return err;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int psched_us_per_tick = 1;
|
|
|
|
static int psched_tick_per_us = 1;
|
|
|
|
|
|
|
|
#ifdef CONFIG_PROC_FS
|
|
|
|
static int psched_show(struct seq_file *seq, void *v)
|
|
|
|
{
|
|
|
|
seq_printf(seq, "%08x %08x %08x %08x\n",
|
|
|
|
psched_tick_per_us, psched_us_per_tick,
|
|
|
|
1000000, HZ);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int psched_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
return single_open(file, psched_show, PDE(inode)->data);
|
|
|
|
}
|
|
|
|
|
2007-02-12 09:55:36 +01:00
|
|
|
static const struct file_operations psched_fops = {
|
2005-04-17 00:20:36 +02:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.open = psched_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
|
|
|
.release = single_release,
|
2007-02-09 15:25:16 +01:00
|
|
|
};
|
2005-04-17 00:20:36 +02:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_SCH_CLK_CPU
|
|
|
|
psched_tdiff_t psched_clock_per_hz;
|
|
|
|
int psched_clock_scale;
|
|
|
|
EXPORT_SYMBOL(psched_clock_per_hz);
|
|
|
|
EXPORT_SYMBOL(psched_clock_scale);
|
|
|
|
|
|
|
|
psched_time_t psched_time_base;
|
|
|
|
cycles_t psched_time_mark;
|
|
|
|
EXPORT_SYMBOL(psched_time_mark);
|
|
|
|
EXPORT_SYMBOL(psched_time_base);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Periodically adjust psched_time_base to avoid overflow
|
|
|
|
* with 32-bit get_cycles(). Safe up to 4GHz CPU.
|
|
|
|
*/
|
|
|
|
static void psched_tick(unsigned long);
|
2005-09-09 22:10:40 +02:00
|
|
|
static DEFINE_TIMER(psched_timer, psched_tick, 0, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
static void psched_tick(unsigned long dummy)
|
|
|
|
{
|
|
|
|
if (sizeof(cycles_t) == sizeof(u32)) {
|
|
|
|
psched_time_t dummy_stamp;
|
|
|
|
PSCHED_GET_TIME(dummy_stamp);
|
|
|
|
psched_timer.expires = jiffies + 1*HZ;
|
|
|
|
add_timer(&psched_timer);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
int __init psched_calibrate_clock(void)
|
|
|
|
{
|
|
|
|
psched_time_t stamp, stamp1;
|
|
|
|
struct timeval tv, tv1;
|
|
|
|
psched_tdiff_t delay;
|
|
|
|
long rdelay;
|
|
|
|
unsigned long stop;
|
|
|
|
|
|
|
|
psched_tick(0);
|
|
|
|
stop = jiffies + HZ/10;
|
|
|
|
PSCHED_GET_TIME(stamp);
|
|
|
|
do_gettimeofday(&tv);
|
|
|
|
while (time_before(jiffies, stop)) {
|
|
|
|
barrier();
|
|
|
|
cpu_relax();
|
|
|
|
}
|
|
|
|
PSCHED_GET_TIME(stamp1);
|
|
|
|
do_gettimeofday(&tv1);
|
|
|
|
|
|
|
|
delay = PSCHED_TDIFF(stamp1, stamp);
|
|
|
|
rdelay = tv1.tv_usec - tv.tv_usec;
|
|
|
|
rdelay += (tv1.tv_sec - tv.tv_sec)*1000000;
|
|
|
|
if (rdelay > delay)
|
|
|
|
return -1;
|
|
|
|
delay /= rdelay;
|
|
|
|
psched_tick_per_us = delay;
|
|
|
|
while ((delay>>=1) != 0)
|
|
|
|
psched_clock_scale++;
|
|
|
|
psched_us_per_tick = 1<<psched_clock_scale;
|
|
|
|
psched_clock_per_hz = (psched_tick_per_us*(1000000/HZ))>>psched_clock_scale;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
static int __init pktsched_init(void)
|
|
|
|
{
|
|
|
|
struct rtnetlink_link *link_p;
|
|
|
|
|
|
|
|
#ifdef CONFIG_NET_SCH_CLK_CPU
|
|
|
|
if (psched_calibrate_clock() < 0)
|
|
|
|
return -1;
|
|
|
|
#elif defined(CONFIG_NET_SCH_CLK_JIFFIES)
|
|
|
|
psched_tick_per_us = HZ<<PSCHED_JSCALE;
|
|
|
|
psched_us_per_tick = 1000000;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
link_p = rtnetlink_links[PF_UNSPEC];
|
|
|
|
|
|
|
|
/* Setup rtnetlink links. It is made here to avoid
|
|
|
|
exporting large number of public symbols.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (link_p) {
|
|
|
|
link_p[RTM_NEWQDISC-RTM_BASE].doit = tc_modify_qdisc;
|
|
|
|
link_p[RTM_DELQDISC-RTM_BASE].doit = tc_get_qdisc;
|
|
|
|
link_p[RTM_GETQDISC-RTM_BASE].doit = tc_get_qdisc;
|
|
|
|
link_p[RTM_GETQDISC-RTM_BASE].dumpit = tc_dump_qdisc;
|
|
|
|
link_p[RTM_NEWTCLASS-RTM_BASE].doit = tc_ctl_tclass;
|
|
|
|
link_p[RTM_DELTCLASS-RTM_BASE].doit = tc_ctl_tclass;
|
|
|
|
link_p[RTM_GETTCLASS-RTM_BASE].doit = tc_ctl_tclass;
|
|
|
|
link_p[RTM_GETTCLASS-RTM_BASE].dumpit = tc_dump_tclass;
|
|
|
|
}
|
|
|
|
|
|
|
|
register_qdisc(&pfifo_qdisc_ops);
|
|
|
|
register_qdisc(&bfifo_qdisc_ops);
|
|
|
|
proc_net_fops_create("psched", 0, &psched_fops);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
subsys_initcall(pktsched_init);
|
|
|
|
|
|
|
|
EXPORT_SYMBOL(qdisc_get_rtab);
|
|
|
|
EXPORT_SYMBOL(qdisc_put_rtab);
|
|
|
|
EXPORT_SYMBOL(register_qdisc);
|
|
|
|
EXPORT_SYMBOL(unregister_qdisc);
|
|
|
|
EXPORT_SYMBOL(tc_classify);
|