During SACK processing, most of the benefits of TSO are eaten by
the SACK blocks that one-by-one fragment SKBs to MSS sized chunks.
Then we're in problems when cleanup work for them has to be done
when a large cumulative ACK comes. Try to return back to pre-split
state already while more and more SACK info gets discovered by
combining newly discovered SACK areas with the previous skb if
that's SACKed as well.
This approach has a number of benefits:
1) The processing overhead is spread more equally over the RTT
2) Write queue has less skbs to process (affect everything
which has to walk in the queue past the sacked areas)
3) Write queue is consistent whole the time, so no other parts
of TCP has to be aware of this (this was not the case with
some other approach that was, well, quite intrusive all
around).
4) Clean_rtx_queue can release most of the pages using single
put_page instead of previous PAGE_SIZE/mss+1 calls
In case a hole is fully filled by the new SACK block, we attempt
to combine the next skb too which allows construction of skbs
that are even larger than what tso split them to and it handles
hole per on every nth patterns that often occur during slow start
overshoot pretty nicely. Though this to be really useful also
a retransmission would have to get lost since cumulative ACKs
advance one hole at a time in the most typical case.
TODO: handle upwards only merging. That should be rather easy
when segment is fully sacked but I'm leaving that as future
work item (it won't make very large difference anyway since
this current approach already covers quite a lot of normal
cases).
I was earlier thinking of some sophisticated way of tracking
timestamps of the first and the last segment but later on
realized that it won't be that necessary at all to store the
timestamp of the last segment. The cases that can occur are
basically either:
1) ambiguous => no sensible measurement can be taken anyway
2) non-ambiguous is due to reordering => having the timestamp
of the last segment there is just skewing things more off
than does some good since the ack got triggered by one of
the holes (besides some substle issues that would make
determining right hole/skb even harder problem). Anyway,
it has nothing to do with this change then.
I choose to route some abnormal looking cases with goto noop,
some could be handled differently (eg., by stopping the
walking at that skb but again). In general, they either
shouldn't happen at all or are rare enough to make no difference
in practice.
In theory this change (as whole) could cause some macroscale
regression (global) because of cache misses that are taken over
the round-trip time but it gets very likely better because of much
less (local) cache misses per other write queue walkers and the
big recovery clearing cumulative ack.
Worth to note that these benefits would be very easy to get also
without TSO/GSO being on as long as the data is in pages so that
we can merge them. Currently I won't let that happen because
DSACK splitting at fragment that would mess up pcounts due to
sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets
avoided, we have some conditions that can be made less strict.
TODO: I will probably have to convert the excessive pointer
passing to struct sacktag_state... :-)
My testing revealed that considerable amount of skbs couldn't
be shifted because they were cloned (most likely still awaiting
tx reclaim)...
[The rest is considering future work instead since I got
repeatably EFAULT to tcpdump's recvfrom when I added
pskb_expand_head to deal with clones, so I separated that
into another, later patch]
...To counter that, I gave up on the fifth advantage:
5) When growing previous SACK block, less allocs for new skbs
are done, basically a new alloc is needed only when new hole
is detected and when the previous skb runs out of frags space
...which now only happens of if reclaim is fast enough to dispose
the clone before the SACK block comes in (the window is RTT long),
otherwise we'll have to alloc some.
With clones being handled I got these numbers (will be somewhat
worse without that), taken with fine-grained mibs:
TCPSackShifted 398
TCPSackMerged 877
TCPSackShiftFallback 320
TCPSACKCOLLAPSEFALLBACKGSO 0
TCPSACKCOLLAPSEFALLBACKSKBBITS 0
TCPSACKCOLLAPSEFALLBACKSKBDATA 0
TCPSACKCOLLAPSEFALLBACKBELOW 0
TCPSACKCOLLAPSEFALLBACKFIRST 1
TCPSACKCOLLAPSEFALLBACKPREVBITS 318
TCPSACKCOLLAPSEFALLBACKMSS 1
TCPSACKCOLLAPSEFALLBACKNOHEAD 0
TCPSACKCOLLAPSEFALLBACKSHIFT 0
TCPSACKCOLLAPSENOOPSEQ 0
TCPSACKCOLLAPSENOOPSMALLPCOUNT 0
TCPSACKCOLLAPSENOOPSMALLLEN 0
TCPSACKCOLLAPSEHOLE 12
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since skb_reset_tail_pointer() reads skb->data, we need to set
skb->data before calling skb_reset_tail_pointer(). This was causing
spurious skb_over_panic()s from skb_put() being called on a recycled
skb that had its skb->tail set to beyond where it should have been.
Bug report from Peter van Valderen <linux@ddcrew.com>.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 04a4bb55bc ("net: add
skb_recycle_check() to enable netdriver skb recycling") added a
method for network drivers to recycle skbuffs, but while use of
this mechanism was documented in the commit message, it should
really have been added as a docbook comment as well -- this
patch does that.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wireless HW without any dedicated queues for aggregation
do not need the ampdu_queues mechanism present right now
in mac80211. Since mac80211 is still incomplete wrt TX MQ
changes, do not allow aggregation sessions for drivers that
set ampdu_queues.
This is only an interim hack until Intel fixes the requeue issue.
Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: Luis Rodriguez <Luis.Rodriguez@Atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Clean up the various different email addresses of mine listed in the code
to a single current and valid address. As Dave says his network merges
for 2.6.28 are now done this seems a good point to send them in where
they won't risk disrupting real changes.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add some packet-split receive hooks.
For one this allows to do NUMA node affine page allocs. Later on these
hooks will be extended to do emergency reserve allocations for
fragments.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the caller of pskb_expand_head specifies a negative nhead
we'll silently overwrite other people's memory. This patch
makes it BUG instead.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds skb_recycle_check(), which can be used by a network
driver after transmitting an skb to check whether this skb can be
recycled as a receive buffer.
skb_recycle_check() checks that the skb is not shared or cloned, and
that it is linear and its head portion large enough (as determined by
the driver) to be recycled as a receive buffer. If these conditions
are met, it does any necessary reference count dropping and cleans
up the skbuff as if it just came from __alloc_skb().
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_gso_segment didn't preserve some attributes in the original skb
such as the netfilter fields. This was harmless until they were used
which is the case for packets going through lo.
This patch makes it call __copy_skb_header which also picks up some
other missing attributes.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes mac80211 to not use the skb->cb over the queue step
from virtual interfaces to the master. The patch also, for now,
disables aggregation because that would still require requeuing,
will fix that in a separate patch. There are two other places (software
requeue and powersaving stations) where requeue can happen, but that is
not currently used by any drivers/not possible to use respectively.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Removes legacy reinvent-the-wheel type thing. The generic
machinery integrates much better to automated debugging aids
such as kerneloops.org (and others), and is unambiguous due to
better naming. Non-intuively BUG_TRAP() is actually equal to
WARN_ON() rather than BUG_ON() though some might actually be
promoted to BUG_ON() but I left that to future.
I could make at least one BUILD_BUG_ON conversion.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/core/skbuff.c:1335:5: warning: symbol '__skb_splice_bits' was not declared. Should it be static?
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- move all of the details on offsets, lengths and buffers into a
single function instead of doing these operation from multiple places
- use a bottom up approach: try to avoid details in the high level
functions, introduce them gradually as we go deeper in the function
call stack
With helpful feedback from Jarek Poplawski.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use a real skb member to store the skb to avoid clashes with qdiscs,
which are allowed to use the cb area themselves. As currently only real
devices that consume the skb set the NETIF_F_HW_VLAN_TX flag, no explicit
invalidation is neccessary.
The new member fills a hole on 64 bit, the skb layout changes from:
__u32 mark; /* 172 4 */
sk_buff_data_t transport_header; /* 176 4 */
sk_buff_data_t network_header; /* 180 4 */
sk_buff_data_t mac_header; /* 184 4 */
sk_buff_data_t tail; /* 188 4 */
/* --- cacheline 3 boundary (192 bytes) --- */
sk_buff_data_t end; /* 192 4 */
/* XXX 4 bytes hole, try to pack */
to
__u32 mark; /* 172 4 */
__u16 vlan_tci; /* 176 2 */
/* XXX 2 bytes hole, try to pack */
sk_buff_data_t transport_header; /* 180 4 */
sk_buff_data_t network_header; /* 184 4 */
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
If an skb has nr_frags set to zero but its frag_list is not empty (as
it can happen if software LRO is enabled), and a previous
tcp_read_sock has consumed the linear part of the skb, then
__skb_splice_bits:
(a) incorrectly reports an error and
(b) forgets to update the offset to account for the linear part
Any of the two problems will cause the subsequent __skb_splice_bits
call (the one that handles the frag_list skbs) to either skip data,
or, if the unadjusted offset is greater then the size of the next skb
in the frag_list, make tcp_splice_read loop forever.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add skb_warn_if_lro() to test whether an skb was received with LRO and
warn if so.
Change br_forward(), ip_forward() and ip6_forward() to call it) and
discard the skb if it returns true.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes CVS keywords that weren't updated for a long time
from comments.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_splice_bits temporary drops the socket lock while iterating over
the socket queue in order to break a reverse locking condition which
happens with sendfile. This, however, opens a window of opportunity
for tcp_collapse() to aggregate skbs and thus potentially free the
current skb used in skb_splice_bits and tcp_read_sock.
This patch fixes the problem by (re-)getting the same "logical skb"
after the lock has been temporary dropped.
Based on idea and initial patch from Evgeniy Polyakov.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Acked-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/skbuff.h says:
/* These elements must be at the end, see alloc_skb() for details. */
net/core/skbuff.c says:
* See comment in sk_buff definition, just before the 'tail' member
This patch contains my guess as to the actual reason rather than a
dead comment reference loop.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This expresses __skb_append in terms of __skb_queue_after, exploiting that
__skb_append(old, new, list) = __skb_queue_after(list, old, new).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
The kernel-doc comment for skb_segment is clearly wrong. This states
what it actually does.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (82 commits)
[NET]: Make sure sockets implement splice_read
netconsole: avoid null pointer dereference at show_local_mac()
[IPV6]: Fix reversed local_df test in ip6_fragment
[XFRM]: Avoid bogus BUG() when throwing new policy away.
[AF_KEY]: Fix bug in spdadd
[NETFILTER] nf_conntrack_proto_tcp.c: Mistyped state corrected.
net: xfrm statistics depend on INET
[NETFILTER]: make secmark_tg_destroy() static
[INET]: Unexport inet_listen_wlock
[INET]: Unexport __inet_hash_connect
[NET]: Improve cache line coherency of ingress qdisc
[NET]: Fix race in dev_close(). (Bug 9750)
[IPSEC]: Fix bogus usage of u64 on input sequence number
[RTNETLINK]: Send a single notification on device state changes.
[NETLABLE]: Hide netlbl_unlabel_audit_addr6 under ifdef CONFIG_IPV6.
[NETLABEL]: Don't produce unused variables when IPv6 is off.
[NETLABEL]: Compilation for CONFIG_AUDIT=n case.
[GENETLINK]: Relax dances with genl_lock.
[NETLABEL]: Fix lookup logic of netlbl_domhsh_search_def.
[IPV6]: remove unused method declaration (net/ndisc.h).
...
Move networking (core and drivers) docbook to its own networking book.
Fix a few kernel-doc errors in header and source files.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use it in virtio_net (replacing buggy version there), it's also going
to be used by TAP for partial csum support.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: David S. Miller <davem@davemloft.net>
Both NetLabel and SELinux (other LSMs may grow to use it as well) rely
on the 'iif' field to determine the receiving network interface of
inbound packets. Unfortunately, at present this field is not
preserved across a skb clone operation which can lead to garbage
values if the cloned skb is sent back through the network stack. This
patch corrects this problem by properly copying the 'iif' field in
__skb_clone() and removing the 'iif' field assignment from
skb_act_clone() since it is no longer needed.
Also, while we are here, put the assignments in the same order as the
offsets to reduce cacheline bounces.
Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The skb_morph function only freed the data part of the dst skb, but leaked
the auxiliary data such as the netfilter fields. This patch fixes this by
moving the relevant parts from __kfree_skb to skb_release_all and calling
it in skb_morph.
It also makes kfree_skbmem static since it's no longer called anywhere else
and it now no longer does skb_release_data.
Thanks to Yasuyuki KOZAKAI for finding this problem and posting a patch for
it.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
sg_mark_end() overwrites the page_link information, but all users want
__sg_mark_end() behaviour where we just set the end bit. That is the most
natural way to use the sg list, since you'll fill it in and then mark the
end point.
So change sg_mark_end() to only set the termination bit. Add a sg_magic
debug check as well, and clear a chain pointer if it is set.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
This fixes scatterlist corruptions added by
commit 68e3f5dd4d
[CRYPTO] users: Fix up scatterlist conversion errors
The issue is that the code calls sg_mark_end() which clobbers the
sg_page() pointer of the final scatterlist entry.
The first part fo the fix makes skb_to_sgvec() do __sg_mark_end().
After considering all skb_to_sgvec() call sites the most correct
solution is to call __sg_mark_end() in skb_to_sgvec() since that is
what all of the callers would end up doing anyways.
I suspect this might have fixed some problems in virtio_net which is
the sole non-crypto user of skb_to_sgvec().
Other similar sg_mark_end() cases were converted over to
__sg_mark_end() as well.
Arguably sg_mark_end() is a poorly named function because it doesn't
just "mark", it clears out the page pointer as a side effect, which is
what led to these bugs in the first place.
The one remaining plain sg_mark_end() call is in scsi_alloc_sgtable()
and arguably it could be converted to __sg_mark_end() if only so that
we can delete this confusing interface from linux/scatterlist.h
Signed-off-by: David S. Miller <davem@davemloft.net>
clean skb_clone of any signs of CONFIG_NET_CLS_ACT and
have mirred us skb_act_clone()
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most drivers need to set length and offset as well, so may as well fold
those three lines into one.
Add sg_assign_page() for those two locations that only needed to set
the page, where the offset/length is set outside of the function context.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
I got confused by the dual nature of the off variable in the
function pskb_expand_head. The csum_start offset should use
nhead instead of off which can change depending on whether we
are using offsets or pointers.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch creates a new function skb_morph that's just like skb_clone
except that it lets user provide the spare skb that will be overwritten
by the one that's to be cloned.
This will be used by IP fragment reassembly so that we get back the same
skb that went in last (rather than the head skb that we get now which
requires us to carry around double pointers all over the place).
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch creates a new function __copy_skb_header to merge the common
code between copy_skb_header and skb_clone. Having two functions which
are largely the same is a source of wasted labour as well as confusion.
In fact the tc_verd stuff is almost certainly a bug since it's treated
differently in skb_clone compared to the callers of copy_skb_header
(skb_copy/pskb_copy/skb_copy_expand).
I've kept that difference in tact with a comment added asking for
clarification.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Thanks for noticing the bug where csum_start is not updated
when the head room changes.
This patch fixes that. It also moves the csum/ip_summed
copying into copy_skb_header so that skb_copy_expand gets
it too. I've checked its callers and no one should be upset
by this.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Slab destructors were no longer supported after Christoph's
c59def9f22 change. They've been
BUGs for both slab and slub, and slob never supported them
either.
This rips out support for the dtor pointer from kmem_cache_create()
completely and fixes up every single callsite in the kernel (there were
about 224, not including the slab allocator definitions themselves,
or the documentation references).
Signed-off-by: Paul Mundt <lethal@linux-sh.org>