2019-05-27 08:55:01 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-05-05 16:16:16 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-06-06 10:43:46 +00:00
|
|
|
|
2007-08-28 15:50:33 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
memcg: decrement static keys at real destroy time
We call the destroy function when a cgroup starts to be removed, such as
by a rmdir event.
However, because of our reference counters, some objects are still
inflight. Right now, we are decrementing the static_keys at destroy()
time, meaning that if we get rid of the last static_key reference, some
objects will still have charges, but the code to properly uncharge them
won't be run.
This becomes a problem specially if it is ever enabled again, because now
new charges will be added to the staled charges making keeping it pretty
much impossible.
We just need to be careful with the static branch activation: since there
is no particular preferred order of their activation, we need to make sure
that we only start using it after all call sites are active. This is
achieved by having a per-memcg flag that is only updated after
static_key_slow_inc() returns. At this time, we are sure all sites are
active.
This is made per-memcg, not global, for a reason: it also has the effect
of making socket accounting more consistent. The first memcg to be
limited will trigger static_key() activation, therefore, accounting. But
all the others will then be accounted no matter what. After this patch,
only limited memcgs will have its sockets accounted.
[akpm@linux-foundation.org: move enum sock_flag_bits into sock.h,
document enum sock_flag_bits,
convert memcg_proto_active() and memcg_proto_activated() to test_bit(),
redo tcp_update_limit() comment to 80 cols]
Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-05-29 15:07:11 -07:00
|
|
|
|
2006-07-03 00:25:35 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2006-12-03 23:15:30 -05:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
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 17:04:11 +09:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
mm: memcontrol: lockless page counters
Memory is internally accounted in bytes, using spinlock-protected 64-bit
counters, even though the smallest accounting delta is a page. The
counter interface is also convoluted and does too many things.
Introduce a new lockless word-sized page counter API, then change all
memory accounting over to it. The translation from and to bytes then only
happens when interfacing with userspace.
The removed locking overhead is noticable when scaling beyond the per-cpu
charge caches - on a 4-socket machine with 144-threads, the following test
shows the performance differences of 288 memcgs concurrently running a
page fault benchmark:
vanilla:
18631648.500498 task-clock (msec) # 140.643 CPUs utilized ( +- 0.33% )
1,380,638 context-switches # 0.074 K/sec ( +- 0.75% )
24,390 cpu-migrations # 0.001 K/sec ( +- 8.44% )
1,843,305,768 page-faults # 0.099 M/sec ( +- 0.00% )
50,134,994,088,218 cycles # 2.691 GHz ( +- 0.33% )
<not supported> stalled-cycles-frontend
<not supported> stalled-cycles-backend
8,049,712,224,651 instructions # 0.16 insns per cycle ( +- 0.04% )
1,586,970,584,979 branches # 85.176 M/sec ( +- 0.05% )
1,724,989,949 branch-misses # 0.11% of all branches ( +- 0.48% )
132.474343877 seconds time elapsed ( +- 0.21% )
lockless:
12195979.037525 task-clock (msec) # 133.480 CPUs utilized ( +- 0.18% )
832,850 context-switches # 0.068 K/sec ( +- 0.54% )
15,624 cpu-migrations # 0.001 K/sec ( +- 10.17% )
1,843,304,774 page-faults # 0.151 M/sec ( +- 0.00% )
32,811,216,801,141 cycles # 2.690 GHz ( +- 0.18% )
<not supported> stalled-cycles-frontend
<not supported> stalled-cycles-backend
9,999,265,091,727 instructions # 0.30 insns per cycle ( +- 0.10% )
2,076,759,325,203 branches # 170.282 M/sec ( +- 0.12% )
1,656,917,214 branch-misses # 0.08% of all branches ( +- 0.55% )
91.369330729 seconds time elapsed ( +- 0.45% )
On top of improved scalability, this also gets rid of the icky long long
types in the very heart of memcg, which is great for 32 bit and also makes
the code a lot more readable.
Notable differences between the old and new API:
- res_counter_charge() and res_counter_charge_nofail() become
page_counter_try_charge() and page_counter_charge() resp. to match
the more common kernel naming scheme of try_do()/do()
- res_counter_uncharge_until() is only ever used to cancel a local
counter and never to uncharge bigger segments of a hierarchy, so
it's replaced by the simpler page_counter_cancel()
- res_counter_set_limit() is replaced by page_counter_limit(), which
expects its callers to serialize against themselves
- res_counter_memparse_write_strategy() is replaced by
page_counter_limit(), which rounds down to the nearest page size -
rather than up. This is more reasonable for explicitely requested
hard upper limits.
- to keep charging light-weight, page_counter_try_charge() charges
speculatively, only to roll back if the result exceeds the limit.
Because of this, a failing bigger charge can temporarily lock out
smaller charges that would otherwise succeed. The error is bounded
to the difference between the smallest and the biggest possible
charge size, so for memcg, this means that a failing THP charge can
send base page charges into reclaim upto 2MB (4MB) before the limit
would have been reached. This should be acceptable.
[akpm@linux-foundation.org: add includes for WARN_ON_ONCE and memparse]
[akpm@linux-foundation.org: add includes for WARN_ON_ONCE, memparse, strncmp, and PAGE_SIZE]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-10 15:42:31 -08:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
2012-02-24 08:31:31 +01:00
|
|
|
|
2012-02-13 03:58:52 +00:00
|
|
|
|
2015-11-26 13:55:39 +08:00
|
|
|
|
2015-12-07 17:38:52 -05:00
|
|
|
|
2017-10-05 22:21:27 -07:00
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2020-07-23 08:08:50 +02:00
|
|
|
|
2020-11-13 07:08:09 -08:00
|
|
|
|
2010-11-15 19:58:26 +00:00
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
tcp: defer skb freeing after socket lock is released
tcp recvmsg() (or rx zerocopy) spends a fair amount of time
freeing skbs after their payload has been consumed.
A typical ~64KB GRO packet has to release ~45 page
references, eventually going to page allocator
for each of them.
Currently, this freeing is performed while socket lock
is held, meaning that there is a high chance that
BH handler has to queue incoming packets to tcp socket backlog.
This can cause additional latencies, because the user
thread has to process the backlog at release_sock() time,
and while doing so, additional frames can be added
by BH handler.
This patch adds logic to defer these frees after socket
lock is released, or directly from BH handler if possible.
Being able to free these skbs from BH handler helps a lot,
because this avoids the usual alloc/free assymetry,
when BH handler and user thread do not run on same cpu or
NUMA node.
One cpu can now be fully utilized for the kernel->user copy,
and another cpu is handling BH processing and skb/page
allocs/frees (assuming RFS is not forcing use of a single CPU)
Tested:
100Gbit NIC
Max throughput for one TCP_STREAM flow, over 10 runs
MTU : 1500
Before: 55 Gbit
After: 66 Gbit
MTU : 4096+(headers)
Before: 82 Gbit
After: 95 Gbit
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2021-11-15 11:02:46 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2015-03-15 21:12:12 -07:00
|
|
|
|
2014-08-04 22:11:46 -04:00
|
|
|
|
2018-01-04 14:03:54 -08:00
|
|
|
|
2021-08-04 10:55:56 +03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-03-21 15:54:53 -07:00
|
|
|
|
2011-10-31 17:11:33 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2008-03-21 15:54:53 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-09-12 10:44:19 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2006-07-03 00:25:35 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2007-12-04 01:15:45 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-12-02 07:33:10 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
2009-11-08 10:17:30 +00:00
|
|
|
|
2012-11-30 09:49:27 +00:00
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-22 09:49:50 +00:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
|
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
2009-11-08 10:17:58 +00:00
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2007-09-12 11:58:02 +02:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
2015-10-08 19:33:22 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 19:33:21 -07:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-11-30 09:49:27 +00:00
|
|
|
|
2012-12-02 07:33:10 +00:00
|
|
|
|
2012-11-30 09:49:27 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-08 10:17:30 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-11-30 09:49:27 +00:00
|
|
|
|
|
|
|
|
|
2012-12-02 07:33:10 +00:00
|
|
|
|
2012-11-30 09:49:27 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
|
|
|
|
|
2013-01-22 09:49:50 +00:00
|
|
|
|
2014-06-27 08:36:16 -07:00
|
|
|
|
|
|
|
|
|
2015-05-08 21:10:31 -05:00
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
2009-11-08 10:17:58 +00:00
|
|
|
|
|
|
|
|
|
2016-04-01 08:52:13 -07:00
|
|
|
|
2009-11-08 10:17:58 +00:00
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2015-03-11 23:06:44 -05:00
|
|
|
|
ipv6: make lookups simpler and faster
TCP listener refactoring, part 4 :
To speed up inet lookups, we moved IPv4 addresses from inet to struct
sock_common
Now is time to do the same for IPv6, because it permits us to have fast
lookups for all kind of sockets, including upcoming SYN_RECV.
Getting IPv6 addresses in TCP lookups currently requires two extra cache
lines, plus a dereference (and memory stall).
inet6_sk(sk) does the dereference of inet_sk(__sk)->pinet6
This patch is way bigger than its IPv4 counter part, because for IPv4,
we could add aliases (inet_daddr, inet_rcv_saddr), while on IPv6,
it's not doable easily.
inet6_sk(sk)->daddr becomes sk->sk_v6_daddr
inet6_sk(sk)->rcv_saddr becomes sk->sk_v6_rcv_saddr
And timewait socket also have tw->tw_v6_daddr & tw->tw_v6_rcv_saddr
at the same offset.
We get rid of INET6_TW_MATCH() as INET6_MATCH() is now the generic
macro.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 15:42:29 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-11 18:53:14 -07:00
|
|
|
|
|
|
|
|
|
2015-10-08 19:33:22 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-01-08 17:39:21 +00:00
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
2011-01-08 17:39:21 +00:00
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:51 -07:00
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
2015-10-08 19:33:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 19:33:24 -07:00
|
|
|
|
2015-10-08 19:33:23 -07:00
|
|
|
|
2015-10-08 19:33:21 -07:00
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
2011-01-08 17:39:21 +00:00
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
2015-10-08 19:33:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 19:33:24 -07:00
|
|
|
|
2015-10-08 19:33:23 -07:00
|
|
|
|
2011-01-08 17:39:21 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2020-08-25 20:29:13 +02:00
|
|
|
|
2021-12-28 16:49:13 -08:00
|
|
|
|
bpf: Introduce bpf sk local storage
After allowing a bpf prog to
- directly read the skb->sk ptr
- get the fullsock bpf_sock by "bpf_sk_fullsock()"
- get the bpf_tcp_sock by "bpf_tcp_sock()"
- get the listener sock by "bpf_get_listener_sock()"
- avoid duplicating the fields of "(bpf_)sock" and "(bpf_)tcp_sock"
into different bpf running context.
this patch is another effort to make bpf's network programming
more intuitive to do (together with memory and performance benefit).
When bpf prog needs to store data for a sk, the current practice is to
define a map with the usual 4-tuples (src/dst ip/port) as the key.
If multiple bpf progs require to store different sk data, multiple maps
have to be defined. Hence, wasting memory to store the duplicated
keys (i.e. 4 tuples here) in each of the bpf map.
[ The smallest key could be the sk pointer itself which requires
some enhancement in the verifier and it is a separate topic. ]
Also, the bpf prog needs to clean up the elem when sk is freed.
Otherwise, the bpf map will become full and un-usable quickly.
The sk-free tracking currently could be done during sk state
transition (e.g. BPF_SOCK_OPS_STATE_CB).
The size of the map needs to be predefined which then usually ended-up
with an over-provisioned map in production. Even the map was re-sizable,
while the sk naturally come and go away already, this potential re-size
operation is arguably redundant if the data can be directly connected
to the sk itself instead of proxy-ing through a bpf map.
This patch introduces sk->sk_bpf_storage to provide local storage space
at sk for bpf prog to use. The space will be allocated when the first bpf
prog has created data for this particular sk.
The design optimizes the bpf prog's lookup (and then optionally followed by
an inline update). bpf_spin_lock should be used if the inline update needs
to be protected.
BPF_MAP_TYPE_SK_STORAGE:
-----------------------
To define a bpf "sk-local-storage", a BPF_MAP_TYPE_SK_STORAGE map (new in
this patch) needs to be created. Multiple BPF_MAP_TYPE_SK_STORAGE maps can
be created to fit different bpf progs' needs. The map enforces
BTF to allow printing the sk-local-storage during a system-wise
sk dump (e.g. "ss -ta") in the future.
The purpose of a BPF_MAP_TYPE_SK_STORAGE map is not for lookup/update/delete
a "sk-local-storage" data from a particular sk.
Think of the map as a meta-data (or "type") of a "sk-local-storage". This
particular "type" of "sk-local-storage" data can then be stored in any sk.
The main purposes of this map are mostly:
1. Define the size of a "sk-local-storage" type.
2. Provide a similar syscall userspace API as the map (e.g. lookup/update,
map-id, map-btf...etc.)
3. Keep track of all sk's storages of this "type" and clean them up
when the map is freed.
sk->sk_bpf_storage:
------------------
The main lookup/update/delete is done on sk->sk_bpf_storage (which
is a "struct bpf_sk_storage"). When doing a lookup,
the "map" pointer is now used as the "key" to search on the
sk_storage->list. The "map" pointer is actually serving
as the "type" of the "sk-local-storage" that is being
requested.
To allow very fast lookup, it should be as fast as looking up an
array at a stable-offset. At the same time, it is not ideal to
set a hard limit on the number of sk-local-storage "type" that the
system can have. Hence, this patch takes a cache approach.
The last search result from sk_storage->list is cached in
sk_storage->cache[] which is a stable sized array. Each
"sk-local-storage" type has a stable offset to the cache[] array.
In the future, a map's flag could be introduced to do cache
opt-out/enforcement if it became necessary.
The cache size is 16 (i.e. 16 types of "sk-local-storage").
Programs can share map. On the program side, having a few bpf_progs
running in the networking hotpath is already a lot. The bpf_prog
should have already consolidated the existing sock-key-ed map usage
to minimize the map lookup penalty. 16 has enough runway to grow.
All sk-local-storage data will be removed from sk->sk_bpf_storage
during sk destruction.
bpf_sk_storage_get() and bpf_sk_storage_delete():
------------------------------------------------
Instead of using bpf_map_(lookup|update|delete)_elem(),
the bpf prog needs to use the new helper bpf_sk_storage_get() and
bpf_sk_storage_delete(). The verifier can then enforce the
ARG_PTR_TO_SOCKET argument. The bpf_sk_storage_get() also allows to
"create" new elem if one does not exist in the sk. It is done by
the new BPF_SK_STORAGE_GET_F_CREATE flag. An optional value can also be
provided as the initial value during BPF_SK_STORAGE_GET_F_CREATE.
The BPF_MAP_TYPE_SK_STORAGE also supports bpf_spin_lock. Together,
it has eliminated the potential use cases for an equivalent
bpf_map_update_elem() API (for bpf_prog) in this patch.
Misc notes:
----------
1. map_get_next_key is not supported. From the userspace syscall
perspective, the map has the socket fd as the key while the map
can be shared by pinned-file or map-id.
Since btf is enforced, the existing "ss" could be enhanced to pretty
print the local-storage.
Supporting a kernel defined btf with 4 tuples as the return key could
be explored later also.
2. The sk->sk_lock cannot be acquired. Atomic operations is used instead.
e.g. cmpxchg is done on the sk->sk_bpf_storage ptr.
Please refer to the source code comments for the details in
synchronization cases and considerations.
3. The mem is charged to the sk->sk_omem_alloc as the sk filter does.
Benchmark:
---------
Here is the benchmark data collected by turning on
the "kernel.bpf_stats_enabled" sysctl.
Two bpf progs are tested:
One bpf prog with the usual bpf hashmap (max_entries = 8192) with the
sk ptr as the key. (verifier is modified to support sk ptr as the key
That should have shortened the key lookup time.)
Another bpf prog is with the new BPF_MAP_TYPE_SK_STORAGE.
Both are storing a "u32 cnt", do a lookup on "egress_skb/cgroup" for
each egress skb and then bump the cnt. netperf is used to drive
data with 4096 connected UDP sockets.
BPF_MAP_TYPE_HASH with a modifier verifier (152ns per bpf run)
27: cgroup_skb name egress_sk_map tag 74f56e832918070b run_time_ns 58280107540 run_cnt 381347633
loaded_at 2019-04-15T13:46:39-0700 uid 0
xlated 344B jited 258B memlock 4096B map_ids 16
btf_id 5
BPF_MAP_TYPE_SK_STORAGE in this patch (66ns per bpf run)
30: cgroup_skb name egress_sk_stora tag d4aa70984cc7bbf6 run_time_ns 25617093319 run_cnt 390989739
loaded_at 2019-04-15T13:47:54-0700 uid 0
xlated 168B jited 156B memlock 4096B map_ids 17
btf_id 6
Here is a high-level picture on how are the objects organized:
sk
┌──────┐
│ │
│ │
│ │
│*sk_bpf_storage─────▶ bpf_sk_storage
└──────┘ ┌───────┐
┌───────────┤ list │
│ │ │
│ │ │
│ │ │
│ └───────┘
│
│ elem
│ ┌────────┐
├─▶│ snode │
│ ├────────┤
│ │ data │ bpf_map
│ ├────────┤ ┌─────────┐
│ │map_node│◀─┬─────┤ list │
│ └────────┘ │ │ │
│ │ │ │
│ elem │ │ │
│ ┌────────┐ │ └─────────┘
└─▶│ snode │ │
├────────┤ │
bpf_map │ data │ │
┌─────────┐ ├────────┤ │
│ list ├───────▶│map_node│ │
│ │ └────────┘ │
│ │ │
│ │ elem │
└─────────┘ ┌────────┐ │
┌─▶│ snode │ │
│ ├────────┤ │
│ │ data │ │
│ ├────────┤ │
│ │map_node│◀─┘
│ └────────┘
│
│
│ ┌───────┐
sk └──────────│ list │
┌──────┐ │ │
│ │ │ │
│ │ │ │
│ │ └───────┘
│*sk_bpf_storage───────▶bpf_sk_storage
└──────┘
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-26 16:39:39 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-09 08:09:05 +00:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2010-04-29 11:01:49 +00:00
|
|
|
|
2013-10-07 11:01:39 -05:00
|
|
|
|
2021-10-25 09:48:16 -07:00
|
|
|
|
2021-10-25 09:48:17 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-12 09:29:06 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
2013-06-10 11:39:50 +03:00
|
|
|
|
2013-06-14 16:33:57 +03:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
tcp: TSO packets automatic sizing
After hearing many people over past years complaining against TSO being
bursty or even buggy, we are proud to present automatic sizing of TSO
packets.
One part of the problem is that tcp_tso_should_defer() uses an heuristic
relying on upcoming ACKS instead of a timer, but more generally, having
big TSO packets makes little sense for low rates, as it tends to create
micro bursts on the network, and general consensus is to reduce the
buffering amount.
This patch introduces a per socket sk_pacing_rate, that approximates
the current sending rate, and allows us to size the TSO packets so
that we try to send one packet every ms.
This field could be set by other transports.
Patch has no impact for high speed flows, where having large TSO packets
makes sense to reach line rate.
For other flows, this helps better packet scheduling and ACK clocking.
This patch increases performance of TCP flows in lossy environments.
A new sysctl (tcp_min_tso_segs) is added, to specify the
minimal size of a TSO packet (default being 2).
A follow-up patch will provide a new packet scheduler (FQ), using
sk_pacing_rate as an input to perform optional per flow pacing.
This explains why we chose to set sk_pacing_rate to twice the current
rate, allowing 'slow start' ramp up.
sk_pacing_rate = 2 * cwnd * mss / srtt
v2: Neal Cardwell reported a suspect deferring of last two segments on
initial write of 10 MSS, I had to change tcp_tso_should_defer() to take
into account tp->xmit_size_goal_segs
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Van Jacobson <vanj@google.com>
Cc: Tom Herbert <therbert@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-08-27 05:46:32 -07:00
|
|
|
|
2017-05-16 04:24:36 -07:00
|
|
|
|
2013-09-29 01:12:40 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2017-07-12 09:29:06 -07:00
|
|
|
|
2016-10-23 09:28:29 -07:00
|
|
|
|
2014-05-23 08:47:19 -07:00
|
|
|
|
|
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2021-11-15 11:02:35 -08:00
|
|
|
|
2006-06-30 13:36:35 -07:00
|
|
|
|
[NET]: Add per-connection option to set max TSO frame size
Update: My mailer ate one of Jarek's feedback mails... Fixed the
parameter in netif_set_gso_max_size() to be u32, not u16. Fixed the
whitespace issue due to a patch import botch. Changed the types from
u32 to unsigned int to be more consistent with other variables in the
area. Also brought the patch up to the latest net-2.6.26 tree.
Update: Made gso_max_size container 32 bits, not 16. Moved the
location of gso_max_size within netdev to be less hotpath. Made more
consistent names between the sock and netdev layers, and added a
define for the max GSO size.
Update: Respun for net-2.6.26 tree.
Update: changed max_gso_frame_size and sk_gso_max_size from signed to
unsigned - thanks Stephen!
This patch adds the ability for device drivers to control the size of
the TSO frames being sent to them, per TCP connection. By setting the
netdevice's gso_max_size value, the socket layer will set the GSO
frame size based on that value. This will propogate into the TCP
layer, and send TSO's of that size to the hardware.
This can be desirable to help tune the bursty nature of TSO on a
per-adapter basis, where one may have 1 GbE and 10 GbE devices
coexisting in a system, one running multiqueue and the other not, etc.
This can also be desirable for devices that cannot support full 64 KB
TSO's, but still want to benefit from some level of segmentation
offloading.
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-03-21 03:43:19 -07:00
|
|
|
|
2012-07-30 16:11:42 +00:00
|
|
|
|
2017-11-11 15:54:12 -08:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-11-13 20:30:01 -08:00
|
|
|
|
|
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2007-11-13 20:30:01 -08:00
|
|
|
|
|
|
|
|
|
2008-06-17 21:04:56 -07:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
2017-07-12 09:29:06 -07:00
|
|
|
|
net: Introduce preferred busy-polling
The existing busy-polling mode, enabled by the SO_BUSY_POLL socket
option or system-wide using the /proc/sys/net/core/busy_read knob, is
an opportunistic. That means that if the NAPI context is not
scheduled, it will poll it. If, after busy-polling, the budget is
exceeded the busy-polling logic will schedule the NAPI onto the
regular softirq handling.
One implication of the behavior above is that a busy/heavy loaded NAPI
context will never enter/allow for busy-polling. Some applications
prefer that most NAPI processing would be done by busy-polling.
This series adds a new socket option, SO_PREFER_BUSY_POLL, that works
in concert with the napi_defer_hard_irqs and gro_flush_timeout
knobs. The napi_defer_hard_irqs and gro_flush_timeout knobs were
introduced in commit 6f8b12d661d0 ("net: napi: add hard irqs deferral
feature"), and allows for a user to defer interrupts to be enabled and
instead schedule the NAPI context from a watchdog timer. When a user
enables the SO_PREFER_BUSY_POLL, again with the other knobs enabled,
and the NAPI context is being processed by a softirq, the softirq NAPI
processing will exit early to allow the busy-polling to be performed.
If the application stops performing busy-polling via a system call,
the watchdog timer defined by gro_flush_timeout will timeout, and
regular softirq handling will resume.
In summary; Heavy traffic applications that prefer busy-polling over
softirq processing should use this option.
Example usage:
$ echo 2 | sudo tee /sys/class/net/ens785f1/napi_defer_hard_irqs
$ echo 200000 | sudo tee /sys/class/net/ens785f1/gro_flush_timeout
Note that the timeout should be larger than the userspace processing
window, otherwise the watchdog will timeout and fall back to regular
softirq processing.
Enable the SO_BUSY_POLL/SO_PREFER_BUSY_POLL options on your socket.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/20201130185205.196029-2-bjorn.topel@gmail.com
2020-11-30 19:51:56 +01:00
|
|
|
|
2020-11-30 19:51:57 +01:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-01 09:46:22 -07:00
|
|
|
|
2010-08-09 13:41:07 +00:00
|
|
|
|
|
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-01 21:32:17 -07:00
|
|
|
|
2022-01-31 16:31:22 +03:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
2021-06-30 16:11:59 +08:00
|
|
|
|
net: Introduce sk_use_task_frag in struct sock.
Sockets that can be used while recursing into memory reclaim, like
those used by network block devices and file systems, mustn't use
current->task_frag: if the current process is already using it, then
the inner memory reclaim call would corrupt the task_frag structure.
To avoid this, sk_page_frag() uses ->sk_allocation to detect sockets
that mustn't use current->task_frag, assuming that those used during
memory reclaim had their allocation constraints reflected in
->sk_allocation.
This unfortunately doesn't cover all cases: in an attempt to remove all
usage of GFP_NOFS and GFP_NOIO, sunrpc stopped setting these flags in
->sk_allocation, and used memalloc_nofs critical sections instead.
This breaks the sk_page_frag() heuristic since the allocation
constraints are now stored in current->flags, which sk_page_frag()
can't read without risking triggering a cache miss and slowing down
TCP's fast path.
This patch creates a new field in struct sock, named sk_use_task_frag,
which sockets with memory reclaim constraints can set to false if they
can't safely use current->task_frag. In such cases, sk_page_frag() now
always returns the socket's page_frag (->sk_frag). The first user is
sunrpc, which needs to avoid using current->task_frag but can keep
->sk_allocation set to GFP_KERNEL otherwise.
Eventually, it might be possible to simplify sk_page_frag() by only
testing ->sk_use_task_frag and avoid relying on the ->sk_allocation
heuristic entirely (assuming other sockets will set ->sk_use_task_frag
according to their constraints in the future).
The new ->sk_use_task_frag field is placed in a hole in struct sock and
belongs to a cache line shared with ->sk_shutdown. Therefore it should
be hot and shouldn't have negative performance impacts on TCP's fast
path (sk_shutdown is tested just before the while() loop in
tcp_sendmsg_locked()).
Link: https://lore.kernel.org/netdev/b4d8cb09c913d3e34f853736f3f5628abfd7f4b6.1656699567.git.gnault@redhat.com/
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-12-16 07:45:26 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-06-30 16:11:59 +08:00
|
|
|
|
|
|
|
|
|
2014-08-04 22:11:47 -04:00
|
|
|
|
2017-08-03 16:29:39 -04:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2022-11-14 20:16:19 +01:00
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
2012-04-17 14:03:53 +00:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2005-05-01 08:59:26 -07:00
|
|
|
|
2008-02-18 20:52:13 -08:00
|
|
|
|
2015-12-07 17:38:52 -05:00
|
|
|
|
2016-01-14 15:21:17 -08:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2023-05-26 16:34:58 +00:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
2016-01-04 17:41:45 -05:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2016-10-23 09:28:29 -07:00
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
2021-12-09 23:44:22 -08:00
|
|
|
|
net: Add a bhash2 table hashed by port and address
The current bind hashtable (bhash) is hashed by port only.
In the socket bind path, we have to check for bind conflicts by
traversing the specified port's inet_bind_bucket while holding the
hashbucket's spinlock (see inet_csk_get_port() and
inet_csk_bind_conflict()). In instances where there are tons of
sockets hashed to the same port at different addresses, the bind
conflict check is time-intensive and can cause softirq cpu lockups,
as well as stops new tcp connections since __inet_inherit_port()
also contests for the spinlock.
This patch adds a second bind table, bhash2, that hashes by
port and sk->sk_rcv_saddr (ipv4) and sk->sk_v6_rcv_saddr (ipv6).
Searching the bhash2 table leads to significantly faster conflict
resolution and less time holding the hashbucket spinlock.
Please note a few things:
* There can be the case where the a socket's address changes after it
has been bound. There are two cases where this happens:
1) The case where there is a bind() call on INADDR_ANY (ipv4) or
IPV6_ADDR_ANY (ipv6) and then a connect() call. The kernel will
assign the socket an address when it handles the connect()
2) In inet_sk_reselect_saddr(), which is called when rebuilding the
sk header and a few pre-conditions are met (eg rerouting fails).
In these two cases, we need to update the bhash2 table by removing the
entry for the old address, and add a new entry reflecting the updated
address.
* The bhash2 table must have its own lock, even though concurrent
accesses on the same port are protected by the bhash lock. Bhash2 must
have its own lock to protect against cases where sockets on different
ports hash to different bhash hashbuckets but to the same bhash2
hashbucket.
This brings up a few stipulations:
1) When acquiring both the bhash and the bhash2 lock, the bhash2 lock
will always be acquired after the bhash lock and released before the
bhash lock is released.
2) There are no nested bhash2 hashbucket locks. A bhash2 lock is always
acquired+released before another bhash2 lock is acquired+released.
* The bhash table cannot be superseded by the bhash2 table because for
bind requests on INADDR_ANY (ipv4) or IPV6_ADDR_ANY (ipv6), every socket
bound to that port must be checked for a potential conflict. The bhash
table is the only source of port->socket associations.
Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-22 11:10:21 -07:00
|
|
|
|
2016-10-23 09:28:29 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
net: optimize INET input path further
Followup of commit b178bb3dfc30 (net: reorder struct sock fields)
Optimize INET input path a bit further, by :
1) moving sk_refcnt close to sk_lock.
This reduces number of dirtied cache lines by one on 64bit arches (and
64 bytes cache line size).
2) moving inet_daddr & inet_rcv_saddr at the beginning of sk
(same cache line than hash / family / bound_dev_if / nulls_node)
This reduces number of accessed cache lines in lookups by one, and dont
increase size of inet and timewait socks.
inet and tw sockets now share same place-holder for these fields.
Before patch :
offsetof(struct sock, sk_refcnt) = 0x10
offsetof(struct sock, sk_lock) = 0x40
offsetof(struct sock, sk_receive_queue) = 0x60
offsetof(struct inet_sock, inet_daddr) = 0x270
offsetof(struct inet_sock, inet_rcv_saddr) = 0x274
After patch :
offsetof(struct sock, sk_refcnt) = 0x44
offsetof(struct sock, sk_lock) = 0x48
offsetof(struct sock, sk_receive_queue) = 0x68
offsetof(struct inet_sock, inet_daddr) = 0x0
offsetof(struct inet_sock, inet_rcv_saddr) = 0x4
compute_score() (udp or tcp) now use a single cache line per ignored
item, instead of two.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-30 19:04:07 +00:00
|
|
|
|
|
|
|
|
|
2009-07-15 23:13:10 +00:00
|
|
|
|
inet: consolidate INET_TW_MATCH
TCP listener refactoring, part 2 :
We can use a generic lookup, sockets being in whatever state, if
we are sure all relevant fields are at the same place in all socket
types (ESTABLISH, TIME_WAIT, SYN_RECV)
This patch removes these macros :
inet_addrpair, inet_addrpair, tw_addrpair, tw_portpair
And adds :
sk_portpair, sk_addrpair, sk_daddr, sk_rcv_saddr
Then, INET_TW_MATCH() is really the same than INET_MATCH()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-02 04:29:50 -07:00
|
|
|
|
tcp/dccp: remove twchain
TCP listener refactoring, part 3 :
Our goal is to hash SYN_RECV sockets into main ehash for fast lookup,
and parallel SYN processing.
Current inet_ehash_bucket contains two chains, one for ESTABLISH (and
friend states) sockets, another for TIME_WAIT sockets only.
As the hash table is sized to get at most one socket per bucket, it
makes little sense to have separate twchain, as it makes the lookup
slightly more complicated, and doubles hash table memory usage.
If we make sure all socket types have the lookup keys at the same
offsets, we can use a generic and faster lookup. It turns out TIME_WAIT
and ESTABLISHED sockets already have common lookup fields for IPv4.
[ INET_TW_MATCH() is no longer needed ]
I'll provide a follow-up to factorize IPv6 lookup as well, to remove
INET6_TW_MATCH()
This way, SYN_RECV pseudo sockets will be supported the same.
A new sock_gen_put() helper is added, doing either a sock_put() or
inet_twsk_put() [ and will support SYN_RECV later ].
Note this helper should only be called in real slow path, when rcu
lookup found a socket that was moved to another identity (freed/reused
immediately), but could eventually be used in other contexts, like
sock_edemux()
Before patch :
dmesg | grep "TCP established"
TCP established hash table entries: 524288 (order: 11, 8388608 bytes)
After patch :
TCP established hash table entries: 524288 (order: 10, 4194304 bytes)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 00:22:02 -07:00
|
|
|
|
|
|
|
|
|
inet: consolidate INET_TW_MATCH
TCP listener refactoring, part 2 :
We can use a generic lookup, sockets being in whatever state, if
we are sure all relevant fields are at the same place in all socket
types (ESTABLISH, TIME_WAIT, SYN_RECV)
This patch removes these macros :
inet_addrpair, inet_addrpair, tw_addrpair, tw_portpair
And adds :
sk_portpair, sk_addrpair, sk_daddr, sk_rcv_saddr
Then, INET_TW_MATCH() is really the same than INET_MATCH()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-02 04:29:50 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-22 09:49:50 +00:00
|
|
|
|
2014-06-27 08:36:16 -07:00
|
|
|
|
2015-05-08 21:10:31 -05:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2007-09-12 11:58:02 +02:00
|
|
|
|
ipv6: make lookups simpler and faster
TCP listener refactoring, part 4 :
To speed up inet lookups, we moved IPv4 addresses from inet to struct
sock_common
Now is time to do the same for IPv6, because it permits us to have fast
lookups for all kind of sockets, including upcoming SYN_RECV.
Getting IPv6 addresses in TCP lookups currently requires two extra cache
lines, plus a dereference (and memory stall).
inet6_sk(sk) does the dereference of inet_sk(__sk)->pinet6
This patch is way bigger than its IPv4 counter part, because for IPv4,
we could add aliases (inet_daddr, inet_rcv_saddr), while on IPv6,
it's not doable easily.
inet6_sk(sk)->daddr becomes sk->sk_v6_daddr
inet6_sk(sk)->rcv_saddr becomes sk->sk_v6_rcv_saddr
And timewait socket also have tw->tw_v6_daddr & tw->tw_v6_rcv_saddr
at the same offset.
We get rid of INET6_TW_MATCH() as INET6_MATCH() is now the generic
macro.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 15:42:29 -07:00
|
|
|
|
|
|
|
|
|
2015-03-11 18:53:14 -07:00
|
|
|
|
2015-10-08 19:33:21 -07:00
|
|
|
|
2015-10-08 19:33:22 -07:00
|
|
|
|
2015-10-08 19:33:23 -07:00
|
|
|
|
ipv6: make lookups simpler and faster
TCP listener refactoring, part 4 :
To speed up inet lookups, we moved IPv4 addresses from inet to struct
sock_common
Now is time to do the same for IPv6, because it permits us to have fast
lookups for all kind of sockets, including upcoming SYN_RECV.
Getting IPv6 addresses in TCP lookups currently requires two extra cache
lines, plus a dereference (and memory stall).
inet6_sk(sk) does the dereference of inet_sk(__sk)->pinet6
This patch is way bigger than its IPv4 counter part, because for IPv4,
we could add aliases (inet_daddr, inet_rcv_saddr), while on IPv6,
it's not doable easily.
inet6_sk(sk)->daddr becomes sk->sk_v6_daddr
inet6_sk(sk)->rcv_saddr becomes sk->sk_v6_rcv_saddr
And timewait socket also have tw->tw_v6_daddr & tw->tw_v6_rcv_saddr
at the same offset.
We get rid of INET6_TW_MATCH() as INET6_MATCH() is now the generic
macro.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 15:42:29 -07:00
|
|
|
|
2021-11-15 11:02:49 -08:00
|
|
|
|
2021-12-23 16:09:58 -08:00
|
|
|
|
2021-11-15 11:02:49 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tcp: defer skb freeing after socket lock is released
tcp recvmsg() (or rx zerocopy) spends a fair amount of time
freeing skbs after their payload has been consumed.
A typical ~64KB GRO packet has to release ~45 page
references, eventually going to page allocator
for each of them.
Currently, this freeing is performed while socket lock
is held, meaning that there is a high chance that
BH handler has to queue incoming packets to tcp socket backlog.
This can cause additional latencies, because the user
thread has to process the backlog at release_sock() time,
and while doing so, additional frames can be added
by BH handler.
This patch adds logic to defer these frees after socket
lock is released, or directly from BH handler if possible.
Being able to free these skbs from BH handler helps a lot,
because this avoids the usual alloc/free assymetry,
when BH handler and user thread do not run on same cpu or
NUMA node.
One cpu can now be fully utilized for the kernel->user copy,
and another cpu is handling BH processing and skb/page
allocs/frees (assuming RFS is not forcing use of a single CPU)
Tested:
100Gbit NIC
Max throughput for one TCP_STREAM flow, over 10 runs
MTU : 1500
Before: 55 Gbit
After: 66 Gbit
MTU : 4096+(headers)
Before: 82 Gbit
After: 95 Gbit
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2021-11-15 11:02:46 -08:00
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
net: introduce SO_INCOMING_CPU
Alternative to RPS/RFS is to use hardware support for multiple
queues.
Then split a set of million of sockets into worker threads, each
one using epoll() to manage events on its own socket pool.
Ideally, we want one thread per RX/TX queue/cpu, but we have no way to
know after accept() or connect() on which queue/cpu a socket is managed.
We normally use one cpu per RX queue (IRQ smp_affinity being properly
set), so remembering on socket structure which cpu delivered last packet
is enough to solve the problem.
After accept(), connect(), or even file descriptor passing around
processes, applications can use :
int cpu;
socklen_t len = sizeof(cpu);
getsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);
And use this information to put the socket into the right silo
for optimal performance, as all networking stack should run
on the appropriate cpu, without need to send IPI (RPS/RFS).
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-11-11 05:54:28 -08:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
2013-08-01 11:10:25 +08:00
|
|
|
|
2013-06-14 16:33:57 +03:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
|
|
|
|
|
2023-05-26 16:34:58 +00:00
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
2015-12-08 07:22:02 -08:00
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
2021-10-25 09:48:16 -07:00
|
|
|
|
2013-01-22 21:09:51 +00:00
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
|
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:00 +03:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
2017-10-05 22:21:27 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: reorder struct sock fields
Right now, fields in struct sock are not optimally ordered, because each
path (RX softirq, TX completion, RX user, TX user) has to touch fields
that are contained in many different cache lines.
The really critical thing is to shrink number of cache lines that are
used at RX softirq time : CPU handling softirqs for a device can receive
many frames per second for many sockets. If load is too big, we can drop
frames at NIC level. RPS or multiqueue cards can help, but better reduce
latency if possible.
This patch starts with UDP protocol, then additional patches will try to
reduce latencies of other ones as well.
At RX softirq time, fields of interest for UDP protocol are :
(not counting ones in inet struct for the lookup)
Read/Written:
sk_refcnt (atomic increment/decrement)
sk_rmem_alloc & sk_backlog.len (to check if there is room in queues)
sk_receive_queue
sk_backlog (if socket locked by user program)
sk_rxhash
sk_forward_alloc
sk_drops
Read only:
sk_rcvbuf (sk_rcvqueues_full())
sk_filter
sk_wq
sk_policy[0]
sk_flags
Additional notes :
- sk_backlog has one hole on 64bit arches. We can fill it to save 8
bytes.
- sk_backlog is used only if RX sofirq handler finds the socket while
locked by user.
- sk_rxhash is written only once per flow.
- sk_drops is written only if queues are full
Final layout :
[1] One section grouping all read/write fields, but placing rxhash and
sk_backlog at the end of this section.
[2] One section grouping all read fields in RX handler
(sk_filter, sk_rcv_buf, sk_wq)
[3] Section used by other paths
I'll post a patch on its own to put sk_refcnt at the end of struct
sock_common so that it shares same cache line than section [1]
New offsets on 64bit arch :
sizeof(struct sock)=0x268
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x48
offsetof(struct sock, sk_receive_queue)=0x68
offsetof(struct sock, sk_backlog)=0x80
offsetof(struct sock, sk_rmem_alloc)=0x80
offsetof(struct sock, sk_forward_alloc)=0x98
offsetof(struct sock, sk_rxhash)=0x9c
offsetof(struct sock, sk_rcvbuf)=0xa4
offsetof(struct sock, sk_drops) =0xa0
offsetof(struct sock, sk_filter)=0xa8
offsetof(struct sock, sk_wq)=0xb0
offsetof(struct sock, sk_policy)=0xd0
offsetof(struct sock, sk_flags) =0xe0
Instead of :
sizeof(struct sock)=0x270
offsetof(struct sock, sk_refcnt) =0x10
offsetof(struct sock, sk_lock) =0x50
offsetof(struct sock, sk_receive_queue)=0xc0
offsetof(struct sock, sk_backlog)=0x70
offsetof(struct sock, sk_rmem_alloc)=0xac
offsetof(struct sock, sk_forward_alloc)=0x10c
offsetof(struct sock, sk_rxhash)=0x128
offsetof(struct sock, sk_rcvbuf)=0x4c
offsetof(struct sock, sk_drops) =0x16c
offsetof(struct sock, sk_filter)=0x198
offsetof(struct sock, sk_wq)=0x88
offsetof(struct sock, sk_policy)=0x98
offsetof(struct sock, sk_flags) =0x130
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-11-16 05:56:04 +00:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
2017-05-16 04:24:36 -07:00
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: extend sk_pacing_rate to unsigned long
sk_pacing_rate has beed introduced as a u32 field in 2013,
effectively limiting per flow pacing to 34Gbit.
We believe it is time to allow TCP to pace high speed flows
on 64bit hosts, as we now can reach 100Gbit on one TCP flow.
This patch adds no cost for 32bit kernels.
The tcpi_pacing_rate and tcpi_max_pacing_rate were already
exported as 64bit, so iproute2/ss command require no changes.
Unfortunately the SO_MAX_PACING_RATE socket option will stay
32bit and we will need to add a new option to let applications
control high pacing rates.
State Recv-Q Send-Q Local Address:Port Peer Address:Port
ESTAB 0 1787144 10.246.9.76:49992 10.246.9.77:36741
timer:(on,003ms,0) ino:91863 sk:2 <->
skmem:(r0,rb540000,t66440,tb2363904,f605944,w1822984,o0,bl0,d0)
ts sack bbr wscale:8,8 rto:201 rtt:0.057/0.006 mss:1448
rcvmss:536 advmss:1448
cwnd:138 ssthresh:178 bytes_acked:256699822585 segs_out:177279177
segs_in:3916318 data_segs_out:177279175
bbr:(bw:31276.8Mbps,mrtt:0,pacing_gain:1.25,cwnd_gain:2)
send 28045.5Mbps lastrcv:73333
pacing_rate 38705.0Mbps delivery_rate 22997.6Mbps
busy:73333ms unacked:135 retrans:0/157 rcv_space:14480
notsent:2085120 minrtt:0.013
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-10-15 09:37:53 -07:00
|
|
|
|
|
|
|
|
|
2016-12-03 11:14:56 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-18 19:19:27 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-15 11:02:35 -08:00
|
|
|
|
2017-03-09 08:09:05 +00:00
|
|
|
|
2014-05-23 08:47:19 -07:00
|
|
|
|
|
|
|
|
|
2020-01-09 07:59:15 -08:00
|
|
|
|
2017-11-11 15:54:12 -08:00
|
|
|
|
2020-01-09 07:59:15 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2005-05-05 13:35:15 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-19 19:04:21 -07:00
|
|
|
|
|
|
|
|
|
net: core: Add a UID field to struct sock.
Protocol sockets (struct sock) don't have UIDs, but most of the
time, they map 1:1 to userspace sockets (struct socket) which do.
Various operations such as the iptables xt_owner match need
access to the "UID of a socket", and do so by following the
backpointer to the struct socket. This involves taking
sk_callback_lock and doesn't work when there is no socket
because userspace has already called close().
Simplify this by adding a sk_uid field to struct sock whose value
matches the UID of the corresponding struct socket. The semantics
are as follows:
1. Whenever sk_socket is non-null: sk_uid is the same as the UID
in sk_socket, i.e., matches the return value of sock_i_uid.
Specifically, the UID is set when userspace calls socket(),
fchown(), or accept().
2. When sk_socket is NULL, sk_uid is defined as follows:
- For a socket that no longer has a sk_socket because
userspace has called close(): the previous UID.
- For a cloned socket (e.g., an incoming connection that is
established but on which userspace has not yet called
accept): the UID of the socket it was cloned from.
- For a socket that has never had an sk_socket: UID 0 inside
the user namespace corresponding to the network namespace
the socket belongs to.
Kernel sockets created by sock_create_kern are a special case
of #1 and sk_uid is the user that created them. For kernel
sockets created at network namespace creation time, such as the
per-processor ICMP and TCP sockets, this is the user that created
the network namespace.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-04 02:23:41 +09:00
|
|
|
|
2022-01-31 16:31:22 +03:00
|
|
|
|
net: Introduce preferred busy-polling
The existing busy-polling mode, enabled by the SO_BUSY_POLL socket
option or system-wide using the /proc/sys/net/core/busy_read knob, is
an opportunistic. That means that if the NAPI context is not
scheduled, it will poll it. If, after busy-polling, the budget is
exceeded the busy-polling logic will schedule the NAPI onto the
regular softirq handling.
One implication of the behavior above is that a busy/heavy loaded NAPI
context will never enter/allow for busy-polling. Some applications
prefer that most NAPI processing would be done by busy-polling.
This series adds a new socket option, SO_PREFER_BUSY_POLL, that works
in concert with the napi_defer_hard_irqs and gro_flush_timeout
knobs. The napi_defer_hard_irqs and gro_flush_timeout knobs were
introduced in commit 6f8b12d661d0 ("net: napi: add hard irqs deferral
feature"), and allows for a user to defer interrupts to be enabled and
instead schedule the NAPI context from a watchdog timer. When a user
enables the SO_PREFER_BUSY_POLL, again with the other knobs enabled,
and the NAPI context is being processed by a softirq, the softirq NAPI
processing will exit early to allow the busy-polling to be performed.
If the application stops performing busy-polling via a system call,
the watchdog timer defined by gro_flush_timeout will timeout, and
regular softirq handling will resume.
In summary; Heavy traffic applications that prefer busy-polling over
softirq processing should use this option.
Example usage:
$ echo 2 | sudo tee /sys/class/net/ens785f1/napi_defer_hard_irqs
$ echo 200000 | sudo tee /sys/class/net/ens785f1/gro_flush_timeout
Note that the timeout should be larger than the userspace processing
window, otherwise the watchdog will timeout and fall back to regular
softirq processing.
Enable the SO_BUSY_POLL/SO_PREFER_BUSY_POLL options on your socket.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/20201130185205.196029-2-bjorn.topel@gmail.com
2020-11-30 19:51:56 +01:00
|
|
|
|
|
|
|
|
|
2020-11-30 19:51:57 +01:00
|
|
|
|
net: Introduce preferred busy-polling
The existing busy-polling mode, enabled by the SO_BUSY_POLL socket
option or system-wide using the /proc/sys/net/core/busy_read knob, is
an opportunistic. That means that if the NAPI context is not
scheduled, it will poll it. If, after busy-polling, the budget is
exceeded the busy-polling logic will schedule the NAPI onto the
regular softirq handling.
One implication of the behavior above is that a busy/heavy loaded NAPI
context will never enter/allow for busy-polling. Some applications
prefer that most NAPI processing would be done by busy-polling.
This series adds a new socket option, SO_PREFER_BUSY_POLL, that works
in concert with the napi_defer_hard_irqs and gro_flush_timeout
knobs. The napi_defer_hard_irqs and gro_flush_timeout knobs were
introduced in commit 6f8b12d661d0 ("net: napi: add hard irqs deferral
feature"), and allows for a user to defer interrupts to be enabled and
instead schedule the NAPI context from a watchdog timer. When a user
enables the SO_PREFER_BUSY_POLL, again with the other knobs enabled,
and the NAPI context is being processed by a softirq, the softirq NAPI
processing will exit early to allow the busy-polling to be performed.
If the application stops performing busy-polling via a system call,
the watchdog timer defined by gro_flush_timeout will timeout, and
regular softirq handling will resume.
In summary; Heavy traffic applications that prefer busy-polling over
softirq processing should use this option.
Example usage:
$ echo 2 | sudo tee /sys/class/net/ens785f1/napi_defer_hard_irqs
$ echo 200000 | sudo tee /sys/class/net/ens785f1/gro_flush_timeout
Note that the timeout should be larger than the userspace processing
window, otherwise the watchdog will timeout and fall back to regular
softirq processing.
Enable the SO_BUSY_POLL/SO_PREFER_BUSY_POLL options on your socket.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/20201130185205.196029-2-bjorn.topel@gmail.com
2020-11-30 19:51:56 +01:00
|
|
|
|
2021-09-29 15:57:50 -07:00
|
|
|
|
2021-11-15 11:02:37 -08:00
|
|
|
|
2010-06-13 03:30:14 +00:00
|
|
|
|
|
|
|
|
|
2021-09-29 15:57:50 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2007-04-19 16:16:32 -07:00
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-02-17 09:05:02 -08:00
|
|
|
|
2017-08-03 16:29:39 -04:00
|
|
|
|
2022-12-07 09:37:01 -05:00
|
|
|
|
|
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-03 15:43:00 -07:00
|
|
|
|
|
|
|
|
|
net: Introduce sk_use_task_frag in struct sock.
Sockets that can be used while recursing into memory reclaim, like
those used by network block devices and file systems, mustn't use
current->task_frag: if the current process is already using it, then
the inner memory reclaim call would corrupt the task_frag structure.
To avoid this, sk_page_frag() uses ->sk_allocation to detect sockets
that mustn't use current->task_frag, assuming that those used during
memory reclaim had their allocation constraints reflected in
->sk_allocation.
This unfortunately doesn't cover all cases: in an attempt to remove all
usage of GFP_NOFS and GFP_NOIO, sunrpc stopped setting these flags in
->sk_allocation, and used memalloc_nofs critical sections instead.
This breaks the sk_page_frag() heuristic since the allocation
constraints are now stored in current->flags, which sk_page_frag()
can't read without risking triggering a cache miss and slowing down
TCP's fast path.
This patch creates a new field in struct sock, named sk_use_task_frag,
which sockets with memory reclaim constraints can set to false if they
can't safely use current->task_frag. In such cases, sk_page_frag() now
always returns the socket's page_frag (->sk_frag). The first user is
sunrpc, which needs to avoid using current->task_frag but can keep
->sk_allocation set to GFP_KERNEL otherwise.
Eventually, it might be possible to simplify sk_page_frag() by only
testing ->sk_use_task_frag and avoid relying on the ->sk_allocation
heuristic entirely (assuming other sockets will set ->sk_use_task_frag
according to their constraints in the future).
The new ->sk_use_task_frag field is placed in a hole in struct sock and
belongs to a cache line shared with ->sk_shutdown. Therefore it should
be hot and shouldn't have negative performance impacts on TCP's fast
path (sk_shutdown is tested just before the while() loop in
tcp_sendmsg_locked()).
Link: https://lore.kernel.org/netdev/b4d8cb09c913d3e34f853736f3f5628abfd7f4b6.1656699567.git.gnault@redhat.com/
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-12-16 07:45:26 -05:00
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2008-11-04 14:45:58 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2008-11-04 14:45:58 -08:00
|
|
|
|
2015-12-07 17:38:52 -05:00
|
|
|
|
2016-01-14 15:21:17 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2014-04-11 16:15:36 -04:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
|
|
|
|
|
2018-04-30 10:16:12 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2016-01-04 17:41:45 -05:00
|
|
|
|
bpf: Introduce bpf sk local storage
After allowing a bpf prog to
- directly read the skb->sk ptr
- get the fullsock bpf_sock by "bpf_sk_fullsock()"
- get the bpf_tcp_sock by "bpf_tcp_sock()"
- get the listener sock by "bpf_get_listener_sock()"
- avoid duplicating the fields of "(bpf_)sock" and "(bpf_)tcp_sock"
into different bpf running context.
this patch is another effort to make bpf's network programming
more intuitive to do (together with memory and performance benefit).
When bpf prog needs to store data for a sk, the current practice is to
define a map with the usual 4-tuples (src/dst ip/port) as the key.
If multiple bpf progs require to store different sk data, multiple maps
have to be defined. Hence, wasting memory to store the duplicated
keys (i.e. 4 tuples here) in each of the bpf map.
[ The smallest key could be the sk pointer itself which requires
some enhancement in the verifier and it is a separate topic. ]
Also, the bpf prog needs to clean up the elem when sk is freed.
Otherwise, the bpf map will become full and un-usable quickly.
The sk-free tracking currently could be done during sk state
transition (e.g. BPF_SOCK_OPS_STATE_CB).
The size of the map needs to be predefined which then usually ended-up
with an over-provisioned map in production. Even the map was re-sizable,
while the sk naturally come and go away already, this potential re-size
operation is arguably redundant if the data can be directly connected
to the sk itself instead of proxy-ing through a bpf map.
This patch introduces sk->sk_bpf_storage to provide local storage space
at sk for bpf prog to use. The space will be allocated when the first bpf
prog has created data for this particular sk.
The design optimizes the bpf prog's lookup (and then optionally followed by
an inline update). bpf_spin_lock should be used if the inline update needs
to be protected.
BPF_MAP_TYPE_SK_STORAGE:
-----------------------
To define a bpf "sk-local-storage", a BPF_MAP_TYPE_SK_STORAGE map (new in
this patch) needs to be created. Multiple BPF_MAP_TYPE_SK_STORAGE maps can
be created to fit different bpf progs' needs. The map enforces
BTF to allow printing the sk-local-storage during a system-wise
sk dump (e.g. "ss -ta") in the future.
The purpose of a BPF_MAP_TYPE_SK_STORAGE map is not for lookup/update/delete
a "sk-local-storage" data from a particular sk.
Think of the map as a meta-data (or "type") of a "sk-local-storage". This
particular "type" of "sk-local-storage" data can then be stored in any sk.
The main purposes of this map are mostly:
1. Define the size of a "sk-local-storage" type.
2. Provide a similar syscall userspace API as the map (e.g. lookup/update,
map-id, map-btf...etc.)
3. Keep track of all sk's storages of this "type" and clean them up
when the map is freed.
sk->sk_bpf_storage:
------------------
The main lookup/update/delete is done on sk->sk_bpf_storage (which
is a "struct bpf_sk_storage"). When doing a lookup,
the "map" pointer is now used as the "key" to search on the
sk_storage->list. The "map" pointer is actually serving
as the "type" of the "sk-local-storage" that is being
requested.
To allow very fast lookup, it should be as fast as looking up an
array at a stable-offset. At the same time, it is not ideal to
set a hard limit on the number of sk-local-storage "type" that the
system can have. Hence, this patch takes a cache approach.
The last search result from sk_storage->list is cached in
sk_storage->cache[] which is a stable sized array. Each
"sk-local-storage" type has a stable offset to the cache[] array.
In the future, a map's flag could be introduced to do cache
opt-out/enforcement if it became necessary.
The cache size is 16 (i.e. 16 types of "sk-local-storage").
Programs can share map. On the program side, having a few bpf_progs
running in the networking hotpath is already a lot. The bpf_prog
should have already consolidated the existing sock-key-ed map usage
to minimize the map lookup penalty. 16 has enough runway to grow.
All sk-local-storage data will be removed from sk->sk_bpf_storage
during sk destruction.
bpf_sk_storage_get() and bpf_sk_storage_delete():
------------------------------------------------
Instead of using bpf_map_(lookup|update|delete)_elem(),
the bpf prog needs to use the new helper bpf_sk_storage_get() and
bpf_sk_storage_delete(). The verifier can then enforce the
ARG_PTR_TO_SOCKET argument. The bpf_sk_storage_get() also allows to
"create" new elem if one does not exist in the sk. It is done by
the new BPF_SK_STORAGE_GET_F_CREATE flag. An optional value can also be
provided as the initial value during BPF_SK_STORAGE_GET_F_CREATE.
The BPF_MAP_TYPE_SK_STORAGE also supports bpf_spin_lock. Together,
it has eliminated the potential use cases for an equivalent
bpf_map_update_elem() API (for bpf_prog) in this patch.
Misc notes:
----------
1. map_get_next_key is not supported. From the userspace syscall
perspective, the map has the socket fd as the key while the map
can be shared by pinned-file or map-id.
Since btf is enforced, the existing "ss" could be enhanced to pretty
print the local-storage.
Supporting a kernel defined btf with 4 tuples as the return key could
be explored later also.
2. The sk->sk_lock cannot be acquired. Atomic operations is used instead.
e.g. cmpxchg is done on the sk->sk_bpf_storage ptr.
Please refer to the source code comments for the details in
synchronization cases and considerations.
3. The mem is charged to the sk->sk_omem_alloc as the sk filter does.
Benchmark:
---------
Here is the benchmark data collected by turning on
the "kernel.bpf_stats_enabled" sysctl.
Two bpf progs are tested:
One bpf prog with the usual bpf hashmap (max_entries = 8192) with the
sk ptr as the key. (verifier is modified to support sk ptr as the key
That should have shortened the key lookup time.)
Another bpf prog is with the new BPF_MAP_TYPE_SK_STORAGE.
Both are storing a "u32 cnt", do a lookup on "egress_skb/cgroup" for
each egress skb and then bump the cnt. netperf is used to drive
data with 4096 connected UDP sockets.
BPF_MAP_TYPE_HASH with a modifier verifier (152ns per bpf run)
27: cgroup_skb name egress_sk_map tag 74f56e832918070b run_time_ns 58280107540 run_cnt 381347633
loaded_at 2019-04-15T13:46:39-0700 uid 0
xlated 344B jited 258B memlock 4096B map_ids 16
btf_id 5
BPF_MAP_TYPE_SK_STORAGE in this patch (66ns per bpf run)
30: cgroup_skb name egress_sk_stora tag d4aa70984cc7bbf6 run_time_ns 25617093319 run_cnt 390989739
loaded_at 2019-04-15T13:47:54-0700 uid 0
xlated 168B jited 156B memlock 4096B map_ids 17
btf_id 6
Here is a high-level picture on how are the objects organized:
sk
┌──────┐
│ │
│ │
│ │
│*sk_bpf_storage─────▶ bpf_sk_storage
└──────┘ ┌───────┐
┌───────────┤ list │
│ │ │
│ │ │
│ │ │
│ └───────┘
│
│ elem
│ ┌────────┐
├─▶│ snode │
│ ├────────┤
│ │ data │ bpf_map
│ ├────────┤ ┌─────────┐
│ │map_node│◀─┬─────┤ list │
│ └────────┘ │ │ │
│ │ │ │
│ elem │ │ │
│ ┌────────┐ │ └─────────┘
└─▶│ snode │ │
├────────┤ │
bpf_map │ data │ │
┌─────────┐ ├────────┤ │
│ list ├───────▶│map_node│ │
│ │ └────────┘ │
│ │ │
│ │ elem │
└─────────┘ ┌────────┐ │
┌─▶│ snode │ │
│ ├────────┤ │
│ │ data │ │
│ ├────────┤ │
│ │map_node│◀─┘
│ └────────┘
│
│
│ ┌───────┐
sk └──────────│ list │
┌──────┐ │ │
│ │ │ │
│ │ │ │
│ │ └───────┘
│*sk_bpf_storage───────▶bpf_sk_storage
└──────┘
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-26 16:39:39 -07:00
|
|
|
|
2020-08-25 20:29:13 +02:00
|
|
|
|
bpf: Introduce bpf sk local storage
After allowing a bpf prog to
- directly read the skb->sk ptr
- get the fullsock bpf_sock by "bpf_sk_fullsock()"
- get the bpf_tcp_sock by "bpf_tcp_sock()"
- get the listener sock by "bpf_get_listener_sock()"
- avoid duplicating the fields of "(bpf_)sock" and "(bpf_)tcp_sock"
into different bpf running context.
this patch is another effort to make bpf's network programming
more intuitive to do (together with memory and performance benefit).
When bpf prog needs to store data for a sk, the current practice is to
define a map with the usual 4-tuples (src/dst ip/port) as the key.
If multiple bpf progs require to store different sk data, multiple maps
have to be defined. Hence, wasting memory to store the duplicated
keys (i.e. 4 tuples here) in each of the bpf map.
[ The smallest key could be the sk pointer itself which requires
some enhancement in the verifier and it is a separate topic. ]
Also, the bpf prog needs to clean up the elem when sk is freed.
Otherwise, the bpf map will become full and un-usable quickly.
The sk-free tracking currently could be done during sk state
transition (e.g. BPF_SOCK_OPS_STATE_CB).
The size of the map needs to be predefined which then usually ended-up
with an over-provisioned map in production. Even the map was re-sizable,
while the sk naturally come and go away already, this potential re-size
operation is arguably redundant if the data can be directly connected
to the sk itself instead of proxy-ing through a bpf map.
This patch introduces sk->sk_bpf_storage to provide local storage space
at sk for bpf prog to use. The space will be allocated when the first bpf
prog has created data for this particular sk.
The design optimizes the bpf prog's lookup (and then optionally followed by
an inline update). bpf_spin_lock should be used if the inline update needs
to be protected.
BPF_MAP_TYPE_SK_STORAGE:
-----------------------
To define a bpf "sk-local-storage", a BPF_MAP_TYPE_SK_STORAGE map (new in
this patch) needs to be created. Multiple BPF_MAP_TYPE_SK_STORAGE maps can
be created to fit different bpf progs' needs. The map enforces
BTF to allow printing the sk-local-storage during a system-wise
sk dump (e.g. "ss -ta") in the future.
The purpose of a BPF_MAP_TYPE_SK_STORAGE map is not for lookup/update/delete
a "sk-local-storage" data from a particular sk.
Think of the map as a meta-data (or "type") of a "sk-local-storage". This
particular "type" of "sk-local-storage" data can then be stored in any sk.
The main purposes of this map are mostly:
1. Define the size of a "sk-local-storage" type.
2. Provide a similar syscall userspace API as the map (e.g. lookup/update,
map-id, map-btf...etc.)
3. Keep track of all sk's storages of this "type" and clean them up
when the map is freed.
sk->sk_bpf_storage:
------------------
The main lookup/update/delete is done on sk->sk_bpf_storage (which
is a "struct bpf_sk_storage"). When doing a lookup,
the "map" pointer is now used as the "key" to search on the
sk_storage->list. The "map" pointer is actually serving
as the "type" of the "sk-local-storage" that is being
requested.
To allow very fast lookup, it should be as fast as looking up an
array at a stable-offset. At the same time, it is not ideal to
set a hard limit on the number of sk-local-storage "type" that the
system can have. Hence, this patch takes a cache approach.
The last search result from sk_storage->list is cached in
sk_storage->cache[] which is a stable sized array. Each
"sk-local-storage" type has a stable offset to the cache[] array.
In the future, a map's flag could be introduced to do cache
opt-out/enforcement if it became necessary.
The cache size is 16 (i.e. 16 types of "sk-local-storage").
Programs can share map. On the program side, having a few bpf_progs
running in the networking hotpath is already a lot. The bpf_prog
should have already consolidated the existing sock-key-ed map usage
to minimize the map lookup penalty. 16 has enough runway to grow.
All sk-local-storage data will be removed from sk->sk_bpf_storage
during sk destruction.
bpf_sk_storage_get() and bpf_sk_storage_delete():
------------------------------------------------
Instead of using bpf_map_(lookup|update|delete)_elem(),
the bpf prog needs to use the new helper bpf_sk_storage_get() and
bpf_sk_storage_delete(). The verifier can then enforce the
ARG_PTR_TO_SOCKET argument. The bpf_sk_storage_get() also allows to
"create" new elem if one does not exist in the sk. It is done by
the new BPF_SK_STORAGE_GET_F_CREATE flag. An optional value can also be
provided as the initial value during BPF_SK_STORAGE_GET_F_CREATE.
The BPF_MAP_TYPE_SK_STORAGE also supports bpf_spin_lock. Together,
it has eliminated the potential use cases for an equivalent
bpf_map_update_elem() API (for bpf_prog) in this patch.
Misc notes:
----------
1. map_get_next_key is not supported. From the userspace syscall
perspective, the map has the socket fd as the key while the map
can be shared by pinned-file or map-id.
Since btf is enforced, the existing "ss" could be enhanced to pretty
print the local-storage.
Supporting a kernel defined btf with 4 tuples as the return key could
be explored later also.
2. The sk->sk_lock cannot be acquired. Atomic operations is used instead.
e.g. cmpxchg is done on the sk->sk_bpf_storage ptr.
Please refer to the source code comments for the details in
synchronization cases and considerations.
3. The mem is charged to the sk->sk_omem_alloc as the sk filter does.
Benchmark:
---------
Here is the benchmark data collected by turning on
the "kernel.bpf_stats_enabled" sysctl.
Two bpf progs are tested:
One bpf prog with the usual bpf hashmap (max_entries = 8192) with the
sk ptr as the key. (verifier is modified to support sk ptr as the key
That should have shortened the key lookup time.)
Another bpf prog is with the new BPF_MAP_TYPE_SK_STORAGE.
Both are storing a "u32 cnt", do a lookup on "egress_skb/cgroup" for
each egress skb and then bump the cnt. netperf is used to drive
data with 4096 connected UDP sockets.
BPF_MAP_TYPE_HASH with a modifier verifier (152ns per bpf run)
27: cgroup_skb name egress_sk_map tag 74f56e832918070b run_time_ns 58280107540 run_cnt 381347633
loaded_at 2019-04-15T13:46:39-0700 uid 0
xlated 344B jited 258B memlock 4096B map_ids 16
btf_id 5
BPF_MAP_TYPE_SK_STORAGE in this patch (66ns per bpf run)
30: cgroup_skb name egress_sk_stora tag d4aa70984cc7bbf6 run_time_ns 25617093319 run_cnt 390989739
loaded_at 2019-04-15T13:47:54-0700 uid 0
xlated 168B jited 156B memlock 4096B map_ids 17
btf_id 6
Here is a high-level picture on how are the objects organized:
sk
┌──────┐
│ │
│ │
│ │
│*sk_bpf_storage─────▶ bpf_sk_storage
└──────┘ ┌───────┐
┌───────────┤ list │
│ │ │
│ │ │
│ │ │
│ └───────┘
│
│ elem
│ ┌────────┐
├─▶│ snode │
│ ├────────┤
│ │ data │ bpf_map
│ ├────────┤ ┌─────────┐
│ │map_node│◀─┬─────┤ list │
│ └────────┘ │ │ │
│ │ │ │
│ elem │ │ │
│ ┌────────┐ │ └─────────┘
└─▶│ snode │ │
├────────┤ │
bpf_map │ data │ │
┌─────────┐ ├────────┤ │
│ list ├───────▶│map_node│ │
│ │ └────────┘ │
│ │ │
│ │ elem │
└─────────┘ ┌────────┐ │
┌─▶│ snode │ │
│ ├────────┤ │
│ │ data │ │
│ ├────────┤ │
│ │map_node│◀─┘
│ └────────┘
│
│
│ ┌───────┐
sk └──────────│ list │
┌──────┐ │ │
│ │ │ │
│ │ │ │
│ │ └───────┘
│*sk_bpf_storage───────▶bpf_sk_storage
└──────┘
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2019-04-26 16:39:39 -07:00
|
|
|
|
2016-04-01 08:52:12 -07:00
|
|
|
|
2021-12-09 23:44:22 -08:00
|
|
|
|
net: Add a bhash2 table hashed by port and address
The current bind hashtable (bhash) is hashed by port only.
In the socket bind path, we have to check for bind conflicts by
traversing the specified port's inet_bind_bucket while holding the
hashbucket's spinlock (see inet_csk_get_port() and
inet_csk_bind_conflict()). In instances where there are tons of
sockets hashed to the same port at different addresses, the bind
conflict check is time-intensive and can cause softirq cpu lockups,
as well as stops new tcp connections since __inet_inherit_port()
also contests for the spinlock.
This patch adds a second bind table, bhash2, that hashes by
port and sk->sk_rcv_saddr (ipv4) and sk->sk_v6_rcv_saddr (ipv6).
Searching the bhash2 table leads to significantly faster conflict
resolution and less time holding the hashbucket spinlock.
Please note a few things:
* There can be the case where the a socket's address changes after it
has been bound. There are two cases where this happens:
1) The case where there is a bind() call on INADDR_ANY (ipv4) or
IPV6_ADDR_ANY (ipv6) and then a connect() call. The kernel will
assign the socket an address when it handles the connect()
2) In inet_sk_reselect_saddr(), which is called when rebuilding the
sk header and a few pre-conditions are met (eg rerouting fails).
In these two cases, we need to update the bhash2 table by removing the
entry for the old address, and add a new entry reflecting the updated
address.
* The bhash2 table must have its own lock, even though concurrent
accesses on the same port are protected by the bhash lock. Bhash2 must
have its own lock to protect against cases where sockets on different
ports hash to different bhash hashbuckets but to the same bhash2
hashbucket.
This brings up a few stipulations:
1) When acquiring both the bhash and the bhash2 lock, the bhash2 lock
will always be acquired after the bhash lock and released before the
bhash lock is released.
2) There are no nested bhash2 hashbucket locks. A bhash2 lock is always
acquired+released before another bhash2 lock is acquired+released.
* The bhash table cannot be superseded by the bhash2 table because for
bind requests on INADDR_ANY (ipv4) or IPV6_ADDR_ANY (ipv6), every socket
bound to that port must be checked for a potential conflict. The bhash
table is the only source of port->socket associations.
Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-22 11:10:21 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2017-05-16 04:24:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net, sk_msg: Clear sk_user_data pointer on clone if tagged
sk_user_data can hold a pointer to an object that is not intended to be
shared between the parent socket and the child that gets a pointer copy on
clone. This is the case when sk_user_data points at reference-counted
object, like struct sk_psock.
One way to resolve it is to tag the pointer with a no-copy flag by
repurposing its lowest bit. Based on the bit-flag value we clear the child
sk_user_data pointer after cloning the parent socket.
The no-copy flag is stored in the pointer itself as opposed to externally,
say in socket flags, to guarantee that the pointer and the flag are copied
from parent to child socket in an atomic fashion. Parent socket state is
subject to change while copying, we don't hold any locks at that time.
This approach relies on an assumption that sk_user_data holds a pointer to
an object aligned at least 2 bytes. A manual audit of existing users of
rcu_dereference_sk_user_data helper confirms our assumption.
Also, an RCU-protected sk_user_data is not likely to hold a pointer to a
char value or a pathological case of "struct { char c; }". To be safe, warn
when the flag-bit is set when setting sk_user_data to catch any future
misuses.
It is worth considering why clearing sk_user_data unconditionally is not an
option. There exist users, DRBD, NVMe, and Xen drivers being among them,
that rely on the pointer being copied when cloning the listening socket.
Potentially we could distinguish these users by checking if the listening
socket has been created in kernel-space via sock_create_kern, and hence has
sk_kern_sock flag set. However, this is not the case for NVMe and Xen
drivers, which create sockets without marking them as belonging to the
kernel.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-3-jakub@cloudflare.com
2020-02-18 17:10:14 +00:00
|
|
|
|
|
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net, sk_msg: Clear sk_user_data pointer on clone if tagged
sk_user_data can hold a pointer to an object that is not intended to be
shared between the parent socket and the child that gets a pointer copy on
clone. This is the case when sk_user_data points at reference-counted
object, like struct sk_psock.
One way to resolve it is to tag the pointer with a no-copy flag by
repurposing its lowest bit. Based on the bit-flag value we clear the child
sk_user_data pointer after cloning the parent socket.
The no-copy flag is stored in the pointer itself as opposed to externally,
say in socket flags, to guarantee that the pointer and the flag are copied
from parent to child socket in an atomic fashion. Parent socket state is
subject to change while copying, we don't hold any locks at that time.
This approach relies on an assumption that sk_user_data holds a pointer to
an object aligned at least 2 bytes. A manual audit of existing users of
rcu_dereference_sk_user_data helper confirms our assumption.
Also, an RCU-protected sk_user_data is not likely to hold a pointer to a
char value or a pathological case of "struct { char c; }". To be safe, warn
when the flag-bit is set when setting sk_user_data to catch any future
misuses.
It is worth considering why clearing sk_user_data unconditionally is not an
option. There exist users, DRBD, NVMe, and Xen drivers being among them,
that rely on the pointer being copied when cloning the listening socket.
Potentially we could distinguish these users by checking if the listening
socket has been created in kernel-space via sock_create_kern, and hence has
sk_kern_sock flag set. However, this is not the case for NVMe and Xen
drivers, which create sockets without marking them as belonging to the
kernel.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-3-jakub@cloudflare.com
2020-02-18 17:10:14 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-24 10:25:40 -07:00
|
|
|
|
|
|
|
|
|
net: Fix suspicious RCU usage in bpf_sk_reuseport_detach()
bpf_sk_reuseport_detach() calls __rcu_dereference_sk_user_data_with_flags()
to obtain the value of sk->sk_user_data, but that function is only usable
if the RCU read lock is held, and neither that function nor any of its
callers hold it.
Fix this by adding a new helper, __locked_read_sk_user_data_with_flags()
that checks to see if sk->sk_callback_lock() is held and use that here
instead.
Alternatively, making __rcu_dereference_sk_user_data_with_flags() use
rcu_dereference_checked() might suffice.
Without this, the following warning can be occasionally observed:
=============================
WARNING: suspicious RCU usage
6.0.0-rc1-build2+ #563 Not tainted
-----------------------------
include/net/sock.h:592 suspicious rcu_dereference_check() usage!
other info that might help us debug this:
rcu_scheduler_active = 2, debug_locks = 1
5 locks held by locktest/29873:
#0: ffff88812734b550 (&sb->s_type->i_mutex_key#9){+.+.}-{3:3}, at: __sock_release+0x77/0x121
#1: ffff88812f5621b0 (sk_lock-AF_INET){+.+.}-{0:0}, at: tcp_close+0x1c/0x70
#2: ffff88810312f5c8 (&h->lhash2[i].lock){+.+.}-{2:2}, at: inet_unhash+0x76/0x1c0
#3: ffffffff83768bb8 (reuseport_lock){+...}-{2:2}, at: reuseport_detach_sock+0x18/0xdd
#4: ffff88812f562438 (clock-AF_INET){++..}-{2:2}, at: bpf_sk_reuseport_detach+0x24/0xa4
stack backtrace:
CPU: 1 PID: 29873 Comm: locktest Not tainted 6.0.0-rc1-build2+ #563
Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014
Call Trace:
<TASK>
dump_stack_lvl+0x4c/0x5f
bpf_sk_reuseport_detach+0x6d/0xa4
reuseport_detach_sock+0x75/0xdd
inet_unhash+0xa5/0x1c0
tcp_set_state+0x169/0x20f
? lockdep_sock_is_held+0x3a/0x3a
? __lock_release.isra.0+0x13e/0x220
? reacquire_held_locks+0x1bb/0x1bb
? hlock_class+0x31/0x96
? mark_lock+0x9e/0x1af
__tcp_close+0x50/0x4b6
tcp_close+0x28/0x70
inet_release+0x8e/0xa7
__sock_release+0x95/0x121
sock_close+0x14/0x17
__fput+0x20f/0x36a
task_work_run+0xa3/0xcc
exit_to_user_mode_prepare+0x9c/0x14d
syscall_exit_to_user_mode+0x18/0x44
entry_SYSCALL_64_after_hwframe+0x63/0xcd
Fixes: cf8c1e967224 ("net: refactor bpf_sk_reuseport_detach()")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Hawkins Jiawei <yin31149@gmail.com>
Link: https://lore.kernel.org/r/166064248071.3502205.10036394558814861778.stgit@warthog.procyon.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-16 10:34:40 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net, sk_msg: Clear sk_user_data pointer on clone if tagged
sk_user_data can hold a pointer to an object that is not intended to be
shared between the parent socket and the child that gets a pointer copy on
clone. This is the case when sk_user_data points at reference-counted
object, like struct sk_psock.
One way to resolve it is to tag the pointer with a no-copy flag by
repurposing its lowest bit. Based on the bit-flag value we clear the child
sk_user_data pointer after cloning the parent socket.
The no-copy flag is stored in the pointer itself as opposed to externally,
say in socket flags, to guarantee that the pointer and the flag are copied
from parent to child socket in an atomic fashion. Parent socket state is
subject to change while copying, we don't hold any locks at that time.
This approach relies on an assumption that sk_user_data holds a pointer to
an object aligned at least 2 bytes. A manual audit of existing users of
rcu_dereference_sk_user_data helper confirms our assumption.
Also, an RCU-protected sk_user_data is not likely to hold a pointer to a
char value or a pathological case of "struct { char c; }". To be safe, warn
when the flag-bit is set when setting sk_user_data to catch any future
misuses.
It is worth considering why clearing sk_user_data unconditionally is not an
option. There exist users, DRBD, NVMe, and Xen drivers being among them,
that rely on the pointer being copied when cloning the listening socket.
Potentially we could distinguish these users by checking if the listening
socket has been created in kernel-space via sock_create_kern, and hence has
sk_kern_sock flag set. However, this is not the case for NVMe and Xen
drivers, which create sockets without marking them as belonging to the
kernel.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-3-jakub@cloudflare.com
2020-02-18 17:10:14 +00:00
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
|
|
|
|
|
net, sk_msg: Clear sk_user_data pointer on clone if tagged
sk_user_data can hold a pointer to an object that is not intended to be
shared between the parent socket and the child that gets a pointer copy on
clone. This is the case when sk_user_data points at reference-counted
object, like struct sk_psock.
One way to resolve it is to tag the pointer with a no-copy flag by
repurposing its lowest bit. Based on the bit-flag value we clear the child
sk_user_data pointer after cloning the parent socket.
The no-copy flag is stored in the pointer itself as opposed to externally,
say in socket flags, to guarantee that the pointer and the flag are copied
from parent to child socket in an atomic fashion. Parent socket state is
subject to change while copying, we don't hold any locks at that time.
This approach relies on an assumption that sk_user_data holds a pointer to
an object aligned at least 2 bytes. A manual audit of existing users of
rcu_dereference_sk_user_data helper confirms our assumption.
Also, an RCU-protected sk_user_data is not likely to hold a pointer to a
char value or a pathological case of "struct { char c; }". To be safe, warn
when the flag-bit is set when setting sk_user_data to catch any future
misuses.
It is worth considering why clearing sk_user_data unconditionally is not an
option. There exist users, DRBD, NVMe, and Xen drivers being among them,
that rely on the pointer being copied when cloning the listening socket.
Potentially we could distinguish these users by checking if the listening
socket has been created in kernel-space via sock_create_kern, and hence has
sk_kern_sock flag set. However, this is not the case for NVMe and Xen
drivers, which create sockets without marking them as belonging to the
kernel.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-3-jakub@cloudflare.com
2020-02-18 17:10:14 +00:00
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net, sk_msg: Clear sk_user_data pointer on clone if tagged
sk_user_data can hold a pointer to an object that is not intended to be
shared between the parent socket and the child that gets a pointer copy on
clone. This is the case when sk_user_data points at reference-counted
object, like struct sk_psock.
One way to resolve it is to tag the pointer with a no-copy flag by
repurposing its lowest bit. Based on the bit-flag value we clear the child
sk_user_data pointer after cloning the parent socket.
The no-copy flag is stored in the pointer itself as opposed to externally,
say in socket flags, to guarantee that the pointer and the flag are copied
from parent to child socket in an atomic fashion. Parent socket state is
subject to change while copying, we don't hold any locks at that time.
This approach relies on an assumption that sk_user_data holds a pointer to
an object aligned at least 2 bytes. A manual audit of existing users of
rcu_dereference_sk_user_data helper confirms our assumption.
Also, an RCU-protected sk_user_data is not likely to hold a pointer to a
char value or a pathological case of "struct { char c; }". To be safe, warn
when the flag-bit is set when setting sk_user_data to catch any future
misuses.
It is worth considering why clearing sk_user_data unconditionally is not an
option. There exist users, DRBD, NVMe, and Xen drivers being among them,
that rely on the pointer being copied when cloning the listening socket.
Potentially we could distinguish these users by checking if the listening
socket has been created in kernel-space via sock_create_kern, and hence has
sk_kern_sock flag set. However, this is not the case for NVMe and Xen
drivers, which create sockets without marking them as belonging to the
kernel.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20200218171023.844439-3-jakub@cloudflare.com
2020-02-18 17:10:14 +00:00
|
|
|
|
|
|
|
|
|
net: fix refcount bug in sk_psock_get (2)
Syzkaller reports refcount bug as follows:
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19
Modules linked in:
CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0
<TASK>
__refcount_add_not_zero include/linux/refcount.h:163 [inline]
__refcount_inc_not_zero include/linux/refcount.h:227 [inline]
refcount_inc_not_zero include/linux/refcount.h:245 [inline]
sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439
tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091
tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983
tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057
tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659
tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682
sk_backlog_rcv include/net/sock.h:1061 [inline]
__release_sock+0x134/0x3b0 net/core/sock.c:2849
release_sock+0x54/0x1b0 net/core/sock.c:3404
inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909
__sys_shutdown_sock net/socket.c:2331 [inline]
__sys_shutdown_sock net/socket.c:2325 [inline]
__sys_shutdown+0xf1/0x1b0 net/socket.c:2343
__do_sys_shutdown net/socket.c:2351 [inline]
__se_sys_shutdown net/socket.c:2349 [inline]
__x64_sys_shutdown+0x50/0x70 net/socket.c:2349
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
</TASK>
During SMC fallback process in connect syscall, kernel will
replaces TCP with SMC. In order to forward wakeup
smc socket waitqueue after fallback, kernel will sets
clcsk->sk_user_data to origin smc socket in
smc_fback_replace_callbacks().
Later, in shutdown syscall, kernel will calls
sk_psock_get(), which treats the clcsk->sk_user_data
as psock type, triggering the refcnt warning.
So, the root cause is that smc and psock, both will use
sk_user_data field. So they will mismatch this field
easily.
This patch solves it by using another bit(defined as
SK_USER_DATA_PSOCK) in PTRMASK, to mark whether
sk_user_data points to a psock object or not.
This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e
("net, sk_msg: Clear sk_user_data pointer on clone if tagged").
For there will possibly be more flags in the sk_user_data field,
this patch also refactor sk_user_data flags code to be more generic
to improve its maintainability.
Reported-and-tested-by: syzbot+5f26f85569bd179c18ce@syzkaller.appspotmail.com
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-05 15:48:34 +08:00
|
|
|
|
2013-09-24 10:25:40 -07:00
|
|
|
|
2022-01-31 16:31:21 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-04-19 03:39:36 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-05 12:41:16 -04:00
|
|
|
|
|
|
|
|
|
2022-06-06 07:34:58 -04:00
|
|
|
|
2012-02-21 07:31:34 +00:00
|
|
|
|
2016-04-05 12:41:14 -04:00
|
|
|
|
datagram: When peeking datagrams with offset < 0 don't skip empty skbs
Due to commit e6afc8ace6dd5cef5e812f26c72579da8806f5ac ("udp: remove
headers from UDP packets before queueing"), when udp packets are being
peeked the requested extra offset is always 0 as there is no need to skip
the udp header. However, when the offset is 0 and the next skb is
of length 0, it is only returned once. The behaviour can be seen with
the following python script:
from socket import *;
f=socket(AF_INET6, SOCK_DGRAM | SOCK_NONBLOCK, 0);
g=socket(AF_INET6, SOCK_DGRAM | SOCK_NONBLOCK, 0);
f.bind(('::', 0));
addr=('::1', f.getsockname()[1]);
g.sendto(b'', addr)
g.sendto(b'b', addr)
print(f.recvfrom(10, MSG_PEEK));
print(f.recvfrom(10, MSG_PEEK));
Where the expected output should be the empty string twice.
Instead, make sk_peek_offset return negative values, and pass those values
to __skb_try_recv_datagram/__skb_try_recv_from_queue. If the passed offset
to __skb_try_recv_from_queue is negative, the checked skb is never skipped.
__skb_try_recv_from_queue will then ensure the offset is reset back to 0
if a peek is requested without an offset, unless no packets are found.
Also simplify the if condition in __skb_try_recv_from_queue. If _off is
greater then 0, and off is greater then or equal to skb->len, then
(_off || skb->len) must always be true assuming skb->len >= 0 is always
true.
Also remove a redundant check around a call to sk_peek_offset in af_unix.c,
as it double checked if MSG_PEEK was set in the flags.
V2:
- Moved the negative fixup into __skb_try_recv_from_queue, and remove now
redundant checks
- Fix peeking in udp{,v6}_recvmsg to report the right value when the
offset is 0
V3:
- Marked new branch in __skb_try_recv_from_queue as unlikely.
Signed-off-by: Matthew Dawson <matthew@mjdsystems.ca>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-08-18 15:04:54 -04:00
|
|
|
|
2016-04-05 12:41:14 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-02-21 07:31:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-05 12:41:14 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-02-21 07:31:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-05 12:41:14 -04:00
|
|
|
|
2012-02-21 07:31:34 +00:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-08 23:18:45 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:46 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:46 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-08-09 20:09:46 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2017-01-20 22:27:04 +08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2006-04-28 15:21:23 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-22 07:57:18 +00:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: force inlining of netif_tx_start/stop_queue, sock_hold, __sock_put
Sometimes gcc mysteriously doesn't inline
very small functions we expect to be inlined. See
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66122
Arguably, gcc should do better, but gcc people aren't willing
to invest time into it, asking to use __always_inline instead.
With this .config:
http://busybox.net/~vda/kernel_config_OPTIMIZE_INLINING_and_Os,
the following functions get deinlined many times.
netif_tx_stop_queue: 207 copies, 590 calls:
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 80 8f e0 01 00 00 01 lock orb $0x1,0x1e0(%rdi)
5d pop %rbp
c3 retq
netif_tx_start_queue: 47 copies, 111 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 80 a7 e0 01 00 00 fe lock andb $0xfe,0x1e0(%rdi)
5d pop %rbp
c3 retq
sock_hold: 39 copies, 124 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 ff 87 80 00 00 00 lock incl 0x80(%rdi)
5d pop %rbp
c3 retq
__sock_put: 6 copies, 13 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 ff 8f 80 00 00 00 lock decl 0x80(%rdi)
5d pop %rbp
c3 retq
This patch fixes this via s/inline/__always_inline/.
Code size decrease after the patch is ~2.5k:
text data bss dec hex filename
56719876 56364551 36196352 149280779 8e5d80b vmlinux_before
56717440 56364551 36196352 149278343 8e5ce87 vmlinux
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
CC: David S. Miller <davem@davemloft.net>
CC: linux-kernel@vger.kernel.org
CC: netdev@vger.kernel.org
CC: netfilter-devel@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-08 17:51:54 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: force inlining of netif_tx_start/stop_queue, sock_hold, __sock_put
Sometimes gcc mysteriously doesn't inline
very small functions we expect to be inlined. See
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66122
Arguably, gcc should do better, but gcc people aren't willing
to invest time into it, asking to use __always_inline instead.
With this .config:
http://busybox.net/~vda/kernel_config_OPTIMIZE_INLINING_and_Os,
the following functions get deinlined many times.
netif_tx_stop_queue: 207 copies, 590 calls:
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 80 8f e0 01 00 00 01 lock orb $0x1,0x1e0(%rdi)
5d pop %rbp
c3 retq
netif_tx_start_queue: 47 copies, 111 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 80 a7 e0 01 00 00 fe lock andb $0xfe,0x1e0(%rdi)
5d pop %rbp
c3 retq
sock_hold: 39 copies, 124 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 ff 87 80 00 00 00 lock incl 0x80(%rdi)
5d pop %rbp
c3 retq
__sock_put: 6 copies, 13 calls
55 push %rbp
48 89 e5 mov %rsp,%rbp
f0 ff 8f 80 00 00 00 lock decl 0x80(%rdi)
5d pop %rbp
c3 retq
This patch fixes this via s/inline/__always_inline/.
Code size decrease after the patch is ~2.5k:
text data bss dec hex filename
56719876 56364551 36196352 149280779 8e5d80b vmlinux_before
56717440 56364551 36196352 149278343 8e5ce87 vmlinux
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
CC: David S. Miller <davem@davemloft.net>
CC: linux-kernel@vger.kernel.org
CC: netdev@vger.kernel.org
CC: netfilter-devel@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-08 17:51:54 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-22 07:57:18 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2010-02-22 07:57:18 +00:00
|
|
|
|
|
|
|
|
|
2016-04-25 10:42:12 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-22 07:57:18 +00:00
|
|
|
|
|
|
|
|
|
2019-03-16 14:41:30 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
2017-12-05 12:45:56 -08:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
2019-12-13 18:20:41 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: Add a bhash2 table hashed by port and address
The current bind hashtable (bhash) is hashed by port only.
In the socket bind path, we have to check for bind conflicts by
traversing the specified port's inet_bind_bucket while holding the
hashbucket's spinlock (see inet_csk_get_port() and
inet_csk_bind_conflict()). In instances where there are tons of
sockets hashed to the same port at different addresses, the bind
conflict check is time-intensive and can cause softirq cpu lockups,
as well as stops new tcp connections since __inet_inherit_port()
also contests for the spinlock.
This patch adds a second bind table, bhash2, that hashes by
port and sk->sk_rcv_saddr (ipv4) and sk->sk_v6_rcv_saddr (ipv6).
Searching the bhash2 table leads to significantly faster conflict
resolution and less time holding the hashbucket spinlock.
Please note a few things:
* There can be the case where the a socket's address changes after it
has been bound. There are two cases where this happens:
1) The case where there is a bind() call on INADDR_ANY (ipv4) or
IPV6_ADDR_ANY (ipv6) and then a connect() call. The kernel will
assign the socket an address when it handles the connect()
2) In inet_sk_reselect_saddr(), which is called when rebuilding the
sk header and a few pre-conditions are met (eg rerouting fails).
In these two cases, we need to update the bhash2 table by removing the
entry for the old address, and add a new entry reflecting the updated
address.
* The bhash2 table must have its own lock, even though concurrent
accesses on the same port are protected by the bhash lock. Bhash2 must
have its own lock to protect against cases where sockets on different
ports hash to different bhash hashbuckets but to the same bhash2
hashbucket.
This brings up a few stipulations:
1) When acquiring both the bhash and the bhash2 lock, the bhash2 lock
will always be acquired after the bhash lock and released before the
bhash lock is released.
2) There are no nested bhash2 hashbucket locks. A bhash2 lock is always
acquired+released before another bhash2 lock is acquired+released.
* The bhash table cannot be superseded by the bhash2 table because for
bind requests on INADDR_ANY (ipv4) or IPV6_ADDR_ANY (ipv6), every socket
bound to that port must be checked for a potential conflict. The bhash
table is the only source of port->socket associations.
Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-22 11:10:21 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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>
2013-02-27 17:06:00 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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>
2013-02-27 17:06:00 -08:00
|
|
|
|
|
|
|
|
|
2008-11-16 19:39:21 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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>
2013-02-27 17:06:00 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: Add a bhash2 table hashed by port and address
The current bind hashtable (bhash) is hashed by port only.
In the socket bind path, we have to check for bind conflicts by
traversing the specified port's inet_bind_bucket while holding the
hashbucket's spinlock (see inet_csk_get_port() and
inet_csk_bind_conflict()). In instances where there are tons of
sockets hashed to the same port at different addresses, the bind
conflict check is time-intensive and can cause softirq cpu lockups,
as well as stops new tcp connections since __inet_inherit_port()
also contests for the spinlock.
This patch adds a second bind table, bhash2, that hashes by
port and sk->sk_rcv_saddr (ipv4) and sk->sk_v6_rcv_saddr (ipv6).
Searching the bhash2 table leads to significantly faster conflict
resolution and less time holding the hashbucket spinlock.
Please note a few things:
* There can be the case where the a socket's address changes after it
has been bound. There are two cases where this happens:
1) The case where there is a bind() call on INADDR_ANY (ipv4) or
IPV6_ADDR_ANY (ipv6) and then a connect() call. The kernel will
assign the socket an address when it handles the connect()
2) In inet_sk_reselect_saddr(), which is called when rebuilding the
sk header and a few pre-conditions are met (eg rerouting fails).
In these two cases, we need to update the bhash2 table by removing the
entry for the old address, and add a new entry reflecting the updated
address.
* The bhash2 table must have its own lock, even though concurrent
accesses on the same port are protected by the bhash lock. Bhash2 must
have its own lock to protect against cases where sockets on different
ports hash to different bhash hashbuckets but to the same bhash2
hashbucket.
This brings up a few stipulations:
1) When acquiring both the bhash and the bhash2 lock, the bhash2 lock
will always be acquired after the bhash lock and released before the
bhash lock is released.
2) There are no nested bhash2 hashbucket locks. A bhash2 lock is always
acquired+released before another bhash2 lock is acquired+released.
* The bhash table cannot be superseded by the bhash2 table because for
bind requests on INADDR_ANY (ipv4) or IPV6_ADDR_ANY (ipv6), every socket
bound to that port must be checked for a potential conflict. The bhash
table is the only source of port->socket associations.
Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-08-22 11:10:21 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2014-07-15 23:28:32 -04:00
|
|
|
|
2016-04-01 08:52:13 -07:00
|
|
|
|
2014-07-15 23:28:32 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-01 08:52:13 -07:00
|
|
|
|
2017-10-23 15:35:58 -04:00
|
|
|
|
2016-04-01 08:52:13 -07:00
|
|
|
|
2014-07-15 23:28:32 -04:00
|
|
|
|
2017-10-23 15:35:58 -04:00
|
|
|
|
2014-07-15 23:28:32 -04:00
|
|
|
|
2022-06-06 07:34:58 -04:00
|
|
|
|
2012-05-24 17:56:43 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-03-25 22:14:49 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-07-31 16:44:16 -07:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
net: speedup sk_wake_async()
An incoming datagram must bring into cpu cache *lot* of cache lines,
in particular : (other parts omitted (hash chains, ip route cache...))
On 32bit arches :
offsetof(struct sock, sk_rcvbuf) =0x30 (read)
offsetof(struct sock, sk_lock) =0x34 (rw)
offsetof(struct sock, sk_sleep) =0x50 (read)
offsetof(struct sock, sk_rmem_alloc) =0x64 (rw)
offsetof(struct sock, sk_receive_queue)=0x74 (rw)
offsetof(struct sock, sk_forward_alloc)=0x98 (rw)
offsetof(struct sock, sk_callback_lock)=0xcc (rw)
offsetof(struct sock, sk_drops) =0xd8 (read if we add dropcount support, rw if frame dropped)
offsetof(struct sock, sk_filter) =0xf8 (read)
offsetof(struct sock, sk_socket) =0x138 (read)
offsetof(struct sock, sk_data_ready) =0x15c (read)
We can avoid sk->sk_socket and socket->fasync_list referencing on sockets
with no fasync() structures. (socket->fasync_list ptr is probably already in cache
because it shares a cache line with socket->wait, ie location pointed by sk->sk_sleep)
This avoids one cache line load per incoming packet for common cases (no fasync())
We can leave (or even move in a future patch) sk->sk_socket in a cold location
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-10-06 17:28:29 -07:00
|
|
|
|
net: Generalize socket rx gap / receive queue overflow cmsg
Create a new socket level option to report number of queue overflows
Recently I augmented the AF_PACKET protocol to report the number of frames lost
on the socket receive queue between any two enqueued frames. This value was
exported via a SOL_PACKET level cmsg. AFter I completed that work it was
requested that this feature be generalized so that any datagram oriented socket
could make use of this option. As such I've created this patch, It creates a
new SOL_SOCKET level option called SO_RXQ_OVFL, which when enabled exports a
SOL_SOCKET level cmsg that reports the nubmer of times the sk_receive_queue
overflowed between any two given frames. It also augments the AF_PACKET
protocol to take advantage of this new feature (as it previously did not touch
sk->sk_drops, which this patch uses to record the overflow count). Tested
successfully by me.
Notes:
1) Unlike my previous patch, this patch simply records the sk_drops value, which
is not a number of drops between packets, but rather a total number of drops.
Deltas must be computed in user space.
2) While this patch currently works with datagram oriented protocols, it will
also be accepted by non-datagram oriented protocols. I'm not sure if thats
agreeable to everyone, but my argument in favor of doing so is that, for those
protocols which aren't applicable to this option, sk_drops will always be zero,
and reporting no drops on a receive queue that isn't used for those
non-participating protocols seems reasonable to me. This also saves us having
to code in a per-protocol opt in mechanism.
3) This applies cleanly to net-next assuming that commit
977750076d98c7ff6cbda51858bb5a5894a9d9ab (my af packet cmsg patch) is reverted
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-10-12 13:26:31 -07:00
|
|
|
|
2011-07-06 12:17:30 +00:00
|
|
|
|
2011-11-09 10:15:42 +01:00
|
|
|
|
2012-02-11 15:39:30 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-16 22:55:49 +01:00
|
|
|
|
2013-03-28 11:19:25 +00:00
|
|
|
|
2016-04-01 08:52:12 -07:00
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
2018-09-12 11:16:59 +08:00
|
|
|
|
2019-02-02 07:34:50 -08:00
|
|
|
|
2022-04-27 16:02:37 -04:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2015-12-04 15:14:04 -02:00
|
|
|
|
|
|
|
|
|
2022-06-06 07:34:58 -04:00
|
|
|
|
2005-08-23 10:11:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-20 18:30:50 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 05:57:07 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-07-31 16:44:19 -07:00
|
|
|
|
2018-05-08 09:06:59 -07:00
|
|
|
|
2012-07-31 16:44:19 -07:00
|
|
|
|
|
|
|
|
|
2018-05-08 09:06:59 -07:00
|
|
|
|
2012-07-31 16:44:19 -07:00
|
|
|
|
2020-06-09 16:11:29 -07:00
|
|
|
|
|
|
|
|
|
2012-07-31 16:44:19 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-09 16:11:29 -07:00
|
|
|
|
|
|
|
|
|
2012-07-31 16:44:19 -07:00
|
|
|
|
|
|
|
|
|
2015-11-30 08:57:28 -08:00
|
|
|
|
2012-07-31 16:44:14 -07:00
|
|
|
|
2015-11-30 08:57:28 -08:00
|
|
|
|
2012-07-31 16:44:14 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2019-11-05 14:11:53 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-05 14:11:53 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2021-03-31 09:35:12 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2021-03-31 09:35:12 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2019-10-10 20:17:46 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2019-10-10 20:17:46 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2010-03-04 18:01:40 +00:00
|
|
|
|
2010-03-04 18:01:47 +00:00
|
|
|
|
2005-11-08 09:39:42 -08:00
|
|
|
|
2010-05-11 23:19:48 +00:00
|
|
|
|
2017-09-21 09:15:46 -07:00
|
|
|
|
2010-05-11 23:19:48 +00:00
|
|
|
|
|
|
|
|
|
2019-11-06 10:04:11 -08:00
|
|
|
|
2010-05-11 23:19:48 +00:00
|
|
|
|
2005-11-08 09:39:42 -08:00
|
|
|
|
2010-05-11 23:19:48 +00:00
|
|
|
|
2019-11-06 10:04:11 -08:00
|
|
|
|
2005-11-08 09:39:42 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2010-04-27 15:13:20 -07:00
|
|
|
|
|
|
|
|
|
2011-12-21 07:11:44 +00:00
|
|
|
|
|
|
|
|
|
2010-04-27 15:13:20 -07:00
|
|
|
|
2014-07-22 21:16:51 +03:00
|
|
|
|
2010-04-27 15:13:20 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-04-22 23:34:26 +00:00
|
|
|
|
2010-04-27 15:13:20 -07:00
|
|
|
|
|
|
|
|
|
2010-03-04 18:01:40 +00:00
|
|
|
|
2012-04-22 23:34:26 +00:00
|
|
|
|
|
|
|
|
|
2010-03-04 18:01:40 +00:00
|
|
|
|
2014-07-22 21:16:51 +03:00
|
|
|
|
2010-03-04 18:01:40 +00:00
|
|
|
|
|
|
|
|
|
2015-09-29 18:52:25 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-03-04 18:01:47 +00:00
|
|
|
|
2010-03-04 18:01:40 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2012-07-31 16:44:26 -07:00
|
|
|
|
2021-11-15 11:02:41 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-10-07 14:18:42 -07:00
|
|
|
|
|
|
|
|
|
2012-07-31 16:44:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-15 11:02:41 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-10-07 14:18:42 -07:00
|
|
|
|
|
|
|
|
|
net: introduce SO_INCOMING_CPU
Alternative to RPS/RFS is to use hardware support for multiple
queues.
Then split a set of million of sockets into worker threads, each
one using epoll() to manage events on its own socket pool.
Ideally, we want one thread per RX/TX queue/cpu, but we have no way to
know after accept() or connect() on which queue/cpu a socket is managed.
We normally use one cpu per RX queue (IRQ smp_affinity being properly
set), so remembering on socket structure which cpu delivered last packet
is enough to solve the problem.
After accept(), connect(), or even file descriptor passing around
processes, applications can use :
int cpu;
socklen_t len = sizeof(cpu);
getsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);
And use this information to put the socket into the right silo
for optimal performance, as all networking stack should run
on the appropriate cpu, without need to send IPI (RPS/RFS).
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-11-11 05:54:28 -08:00
|
|
|
|
|
|
|
|
|
2017-06-21 11:45:31 +02:00
|
|
|
|
|
|
|
|
|
2019-10-30 13:00:04 -07:00
|
|
|
|
|
|
|
|
|
net: introduce SO_INCOMING_CPU
Alternative to RPS/RFS is to use hardware support for multiple
queues.
Then split a set of million of sockets into worker threads, each
one using epoll() to manage events on its own socket pool.
Ideally, we want one thread per RX/TX queue/cpu, but we have no way to
know after accept() or connect() on which queue/cpu a socket is managed.
We normally use one cpu per RX queue (IRQ smp_affinity being properly
set), so remembering on socket structure which cpu delivered last packet
is enough to solve the problem.
After accept(), connect(), or even file descriptor passing around
processes, applications can use :
int cpu;
socklen_t len = sizeof(cpu);
getsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);
And use this information to put the socket into the right silo
for optimal performance, as all networking stack should run
on the appropriate cpu, without need to send IPI (RPS/RFS).
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-11-11 05:54:28 -08:00
|
|
|
|
|
|
|
|
|
2013-12-22 18:54:31 +08:00
|
|
|
|
2010-04-27 15:05:31 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-22 18:54:31 +08:00
|
|
|
|
2010-04-27 15:05:31 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-22 18:54:31 +08:00
|
|
|
|
|
|
|
|
|
2014-01-01 04:31:01 +08:00
|
|
|
|
2019-03-22 08:56:38 -07:00
|
|
|
|
2016-12-07 08:29:10 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-06-06 07:41:14 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-12-07 08:29:10 -08:00
|
|
|
|
2014-01-01 04:31:01 +08:00
|
|
|
|
2013-12-22 18:54:31 +08:00
|
|
|
|
|
|
|
|
|
2011-08-14 19:45:55 +00:00
|
|
|
|
|
|
|
|
|
2010-04-27 15:05:31 -07:00
|
|
|
|
|
|
|
|
|
2023-06-06 07:41:14 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-27 15:05:31 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-08-14 19:45:55 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-06-06 07:41:14 +00:00
|
|
|
|
|
|
|
|
|
2011-08-14 19:45:55 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 10:20:50 -08:00
|
|
|
|
2007-10-09 01:59:42 -07:00
|
|
|
|
2023-05-26 16:34:58 +00:00
|
|
|
|
2007-10-09 01:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-11 10:20:50 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-10-09 01:59:42 -07:00
|
|
|
|
2016-11-11 10:20:50 -08:00
|
|
|
|
2007-10-09 01:59:42 -07:00
|
|
|
|
2023-05-26 16:34:58 +00:00
|
|
|
|
2007-10-09 01:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2016-04-29 14:16:53 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-24 18:19:25 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2005-06-18 22:47:21 -07:00
|
|
|
|
2005-12-13 23:25:19 -08:00
|
|
|
|
[SOCK] proto: Add hashinfo member to struct proto
This way we can remove TCP and DCCP specific versions of
sk->sk_prot->get_port: both v4 and v6 use inet_csk_get_port
sk->sk_prot->hash: inet_hash is directly used, only v6 need
a specific version to deal with mapped sockets
sk->sk_prot->unhash: both v4 and v6 use inet_hash directly
struct inet_connection_sock_af_ops also gets a new member, bind_conflict, so
that inet_csk_get_port can find the per family routine.
Now only the lookup routines receive as a parameter a struct inet_hashtable.
With this we further reuse code, reducing the difference among INET transport
protocols.
Eventually work has to be done on UDP and SCTP to make them share this
infrastructure and get as a bonus inet_diag interfaces so that iproute can be
used with these protocols.
net-2.6/net/ipv4/inet_hashtables.c:
struct proto | +8
struct inet_connection_sock_af_ops | +8
2 structs changed
__inet_hash_nolisten | +18
__inet_hash | -210
inet_put_port | +8
inet_bind_bucket_create | +1
__inet_hash_connect | -8
5 functions changed, 27 bytes added, 218 bytes removed, diff: -191
net-2.6/net/core/sock.c:
proto_seq_show | +3
1 function changed, 3 bytes added, diff: +3
net-2.6/net/ipv4/inet_connection_sock.c:
inet_csk_get_port | +15
1 function changed, 15 bytes added, diff: +15
net-2.6/net/ipv4/tcp.c:
tcp_set_state | -7
1 function changed, 7 bytes removed, diff: -7
net-2.6/net/ipv4/tcp_ipv4.c:
tcp_v4_get_port | -31
tcp_v4_hash | -48
tcp_v4_destroy_sock | -7
tcp_v4_syn_recv_sock | -2
tcp_unhash | -179
5 functions changed, 267 bytes removed, diff: -267
net-2.6/net/ipv6/inet6_hashtables.c:
__inet6_hash | +8
1 function changed, 8 bytes added, diff: +8
net-2.6/net/ipv4/inet_hashtables.c:
inet_unhash | +190
inet_hash | +242
2 functions changed, 432 bytes added, diff: +432
vmlinux:
16 functions changed, 485 bytes added, 492 bytes removed, diff: -7
/home/acme/git/net-2.6/net/ipv6/tcp_ipv6.c:
tcp_v6_get_port | -31
tcp_v6_hash | -7
tcp_v6_syn_recv_sock | -9
3 functions changed, 47 bytes removed, diff: -47
/home/acme/git/net-2.6/net/dccp/proto.c:
dccp_destroy_sock | -7
dccp_unhash | -179
dccp_hash | -49
dccp_set_state | -7
dccp_done | +1
5 functions changed, 1 bytes added, 242 bytes removed, diff: -241
/home/acme/git/net-2.6/net/dccp/ipv4.c:
dccp_v4_get_port | -31
dccp_v4_request_recv_sock | -2
2 functions changed, 33 bytes removed, diff: -33
/home/acme/git/net-2.6/net/dccp/ipv6.c:
dccp_v6_get_port | -31
dccp_v6_hash | -7
dccp_v6_request_recv_sock | +5
3 functions changed, 5 bytes added, 38 bytes removed, diff: -33
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-02-03 04:06:04 -08:00
|
|
|
|
2008-03-22 16:56:51 -07:00
|
|
|
|
2017-01-09 16:55:26 +01:00
|
|
|
|
2011-05-26 13:46:22 -04:00
|
|
|
|
2021-04-06 20:21:11 -07:00
|
|
|
|
[NET] Generalise TCP's struct open_request minisock infrastructure
Kept this first changeset minimal, without changing existing names to
ease peer review.
Basicaly tcp_openreq_alloc now receives the or_calltable, that in turn
has two new members:
->slab, that replaces tcp_openreq_cachep
->obj_size, to inform the size of the openreq descendant for
a specific protocol
The protocol specific fields in struct open_request were moved to a
class hierarchy, with the things that are common to all connection
oriented PF_INET protocols in struct inet_request_sock, the TCP ones
in tcp_request_sock, that is an inet_request_sock, that is an
open_request.
I.e. this uses the same approach used for the struct sock class
hierarchy, with sk_prot indicating if the protocol wants to use the
open_request infrastructure by filling in sk_prot->rsk_prot with an
or_calltable.
Results? Performance is improved and TCP v4 now uses only 64 bytes per
open request minisock, down from 96 without this patch :-)
Next changeset will rename some of the structs, fields and functions
mentioned above, struct or_calltable is way unclear, better name it
struct request_sock_ops, s/struct open_request/struct request_sock/g,
etc.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-18 22:46:52 -07:00
|
|
|
|
2013-05-09 10:28:16 +00:00
|
|
|
|
2017-01-18 02:53:44 -08:00
|
|
|
|
2013-05-09 10:28:16 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2018-03-30 15:08:05 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-09 08:09:05 +00:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-06-14 17:04:49 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2020-07-23 08:09:07 +02:00
|
|
|
|
2009-09-30 16:12:20 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-01-09 16:55:12 +01:00
|
|
|
|
2008-08-28 02:53:51 -07:00
|
|
|
|
2011-01-29 16:15:56 +00:00
|
|
|
|
|
|
|
|
|
2008-08-28 02:53:51 -07:00
|
|
|
|
2015-03-02 15:37:48 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: remove noblock parameter from recvmsg() entities
The internal recvmsg() functions have two parameters 'flags' and 'noblock'
that were merged inside skb_recv_datagram(). As a follow up patch to commit
f4b41f062c42 ("net: remove noblock parameter from skb_recv_datagram()")
this patch removes the separate 'noblock' parameter for recvmsg().
Analogue to the referenced patch for skb_recv_datagram() the 'flags' and
'noblock' parameters are unnecessarily split up with e.g.
err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
flags & ~MSG_DONTWAIT, &addr_len);
or in
err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg,
sk, msg, size, flags & MSG_DONTWAIT,
flags & ~MSG_DONTWAIT, &addr_len);
instead of simply using only flags all the time and check for MSG_DONTWAIT
where needed (to preserve for the formerly separated no(n)block condition).
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Link: https://lore.kernel.org/r/20220411124955.154876-1-socketcan@hartkopp.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2022-04-11 14:49:55 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2020-05-29 14:09:42 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2021-01-15 08:34:59 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
tcp: TCP Small Queues
This introduce TSQ (TCP Small Queues)
TSQ goal is to reduce number of TCP packets in xmit queues (qdisc &
device queues), to reduce RTT and cwnd bias, part of the bufferbloat
problem.
sk->sk_wmem_alloc not allowed to grow above a given limit,
allowing no more than ~128KB [1] per tcp socket in qdisc/dev layers at a
given time.
TSO packets are sized/capped to half the limit, so that we have two
TSO packets in flight, allowing better bandwidth use.
As a side effect, setting the limit to 40000 automatically reduces the
standard gso max limit (65536) to 40000/2 : It can help to reduce
latencies of high prio packets, having smaller TSO packets.
This means we divert sock_wfree() to a tcp_wfree() handler, to
queue/send following frames when skb_orphan() [2] is called for the
already queued skbs.
Results on my dev machines (tg3/ixgbe nics) are really impressive,
using standard pfifo_fast, and with or without TSO/GSO.
Without reduction of nominal bandwidth, we have reduction of buffering
per bulk sender :
< 1ms on Gbit (instead of 50ms with TSO)
< 8ms on 100Mbit (instead of 132 ms)
I no longer have 4 MBytes backlogged in qdisc by a single netperf
session, and both side socket autotuning no longer use 4 Mbytes.
As skb destructor cannot restart xmit itself ( as qdisc lock might be
taken at this point ), we delegate the work to a tasklet. We use one
tasklest per cpu for performance reasons.
If tasklet finds a socket owned by the user, it sets TSQ_OWNED flag.
This flag is tested in a new protocol method called from release_sock(),
to eventually send new segments.
[1] New /proc/sys/net/ipv4/tcp_limit_output_bytes tunable
[2] skb_orphan() is usually called at TX completion time,
but some drivers call it in their start_xmit() handler.
These drivers should at least use BQL, or else a single TCP
session can still fill the whole NIC TX ring, since TSQ will
have no effect.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Dave Taht <dave.taht@bufferbloat.net>
Cc: Tom Herbert <therbert@google.com>
Cc: Matt Mathis <mattmathis@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-11 05:50:31 +00:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2016-02-10 11:50:35 -05:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
udp: add rehash on connect()
commit 30fff923 introduced in linux-2.6.33 (udp: bind() optimisation)
added a secondary hash on UDP, hashed on (local addr, local port).
Problem is that following sequence :
fd = socket(...)
connect(fd, &remote, ...)
not only selects remote end point (address and port), but also sets
local address, while UDP stack stored in secondary hash table the socket
while its local address was INADDR_ANY (or ipv6 equivalent)
Sequence is :
- autobind() : choose a random local port, insert socket in hash tables
[while local address is INADDR_ANY]
- connect() : set remote address and port, change local address to IP
given by a route lookup.
When an incoming UDP frame comes, if more than 10 sockets are found in
primary hash table, we switch to secondary table, and fail to find
socket because its local address changed.
One solution to this problem is to rehash datagram socket if needed.
We add a new rehash(struct socket *) method in "struct proto", and
implement this method for UDP v4 & v6, using a common helper.
This rehashing only takes care of secondary hash table, since primary
hash (based on local port only) is not changed.
Reported-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-08 05:08:44 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
net: bpf: Handle return value of BPF_CGROUP_RUN_PROG_INET{4,6}_POST_BIND()
The return value of BPF_CGROUP_RUN_PROG_INET{4,6}_POST_BIND() in
__inet_bind() is not handled properly. While the return value
is non-zero, it will set inet_saddr and inet_rcv_saddr to 0 and
exit:
err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk);
if (err) {
inet->inet_saddr = inet->inet_rcv_saddr = 0;
goto out_release_sock;
}
Let's take UDP for example and see what will happen. For UDP
socket, it will be added to 'udp_prot.h.udp_table->hash' and
'udp_prot.h.udp_table->hash2' after the sk->sk_prot->get_port()
called success. If 'inet->inet_rcv_saddr' is specified here,
then 'sk' will be in the 'hslot2' of 'hash2' that it don't belong
to (because inet_saddr is changed to 0), and UDP packet received
will not be passed to this sock. If 'inet->inet_rcv_saddr' is not
specified here, the sock will work fine, as it can receive packet
properly, which is wired, as the 'bind()' is already failed.
To undo the get_port() operation, introduce the 'put_port' field
for 'struct proto'. For TCP proto, it is inet_put_port(); For UDP
proto, it is udp_lib_unhash(); For icmp proto, it is
ping_unhash().
Therefore, after sys_bind() fail caused by
BPF_CGROUP_RUN_PROG_INET4_POST_BIND(), it will be unbinded, which
means that it can try to be binded to another port.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20220106132022.3470772-2-imagedong@tencent.com
2022-01-06 21:20:20 +08:00
|
|
|
|
2021-03-30 19:32:31 -07:00
|
|
|
|
2021-04-06 20:21:11 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-30 19:32:31 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
[NET]: Define infrastructure to keep 'inuse' changes in an efficent SMP/NUMA way.
"struct proto" currently uses an array stats[NR_CPUS] to track change on
'inuse' sockets per protocol.
If NR_CPUS is big, this means we use a big memory area for this.
Moreover, all this memory area is located on a single node on NUMA
machines, increasing memory pressure on the boot node.
In this patch, I tried to :
- Keep a fast !CONFIG_SMP implementation
- Keep a fast CONFIG_SMP implementation for often used protocols
(tcp,udp,raw,...)
- Introduce a NUMA efficient implementation
Some helper macros are defined in include/net/sock.h
These macros take into account CONFIG_SMP
If a "struct proto" is declared without using DEFINE_PROTO_INUSE /
REF_PROTO_INUSE
macros, it will automatically use a default implementation, using a
dynamically allocated percpu zone.
This default implementation will be NUMA efficient, but might use 32/64
bytes per possible cpu
because of current alloc_percpu() implementation.
However it still should be better than previous implementation based on
stats[NR_CPUS] field.
When a "struct proto" is changed to use the new macros, we use a single
static "int" percpu variable,
lowering the memory and cpu costs, still preserving NUMA efficiency.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2007-11-05 23:38:39 -08:00
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
2008-03-28 16:38:17 -07:00
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
2007-11-21 22:08:50 +08:00
|
|
|
|
2021-11-15 11:02:38 -08:00
|
|
|
|
2021-10-26 16:29:14 -07:00
|
|
|
|
2021-11-15 11:02:38 -08:00
|
|
|
|
2021-10-26 16:29:14 -07:00
|
|
|
|
2018-12-04 07:58:17 -08:00
|
|
|
|
2021-10-08 13:33:03 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2008-07-16 20:28:10 -07:00
|
|
|
|
2017-06-07 13:29:12 -07:00
|
|
|
|
2010-11-09 23:24:26 +00:00
|
|
|
|
2022-06-08 23:34:08 -07:00
|
|
|
|
2008-11-25 21:16:35 -08:00
|
|
|
|
2021-10-26 16:29:14 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-18 01:51:32 +00:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2017-06-07 13:29:12 -07:00
|
|
|
|
2010-11-09 23:24:26 +00:00
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2010-07-10 20:41:55 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
udp: RCU handling for Unicast packets.
Goals are :
1) Optimizing handling of incoming Unicast UDP frames, so that no memory
writes should happen in the fast path.
Note: Multicasts and broadcasts still will need to take a lock,
because doing a full lockless lookup in this case is difficult.
2) No expensive operations in the socket bind/unhash phases :
- No expensive synchronize_rcu() calls.
- No added rcu_head in socket structure, increasing memory needs,
but more important, forcing us to use call_rcu() calls,
that have the bad property of making sockets structure cold.
(rcu grace period between socket freeing and its potential reuse
make this socket being cold in CPU cache).
David did a previous patch using call_rcu() and noticed a 20%
impact on TCP connection rates.
Quoting Cristopher Lameter :
"Right. That results in cacheline cooldown. You'd want to recycle
the object as they are cache hot on a per cpu basis. That is screwed
up by the delayed regular rcu processing. We have seen multiple
regressions due to cacheline cooldown.
The only choice in cacheline hot sensitive areas is to deal with the
complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU."
- Because udp sockets are allocated from dedicated kmem_cache,
use of SLAB_DESTROY_BY_RCU can help here.
Theory of operation :
---------------------
As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),
special attention must be taken by readers and writers.
Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,
reused, inserted in a different chain or in worst case in the same chain
while readers could do lookups in the same time.
In order to avoid loops, a reader must check each socket found in a chain
really belongs to the chain the reader was traversing. If it finds a
mismatch, lookup must start again at the begining. This *restart* loop
is the reason we had to use rdlock for the multicast case, because
we dont want to send same message several times to the same socket.
We use RCU only for fast path.
Thus, /proc/net/udp still takes spinlocks.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-10-29 02:11:14 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2017-11-15 17:32:18 -08:00
|
|
|
|
2018-04-05 16:21:31 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2021-10-14 06:41:26 -07:00
|
|
|
|
2005-08-09 20:09:30 -07:00
|
|
|
|
2005-06-18 22:47:21 -07:00
|
|
|
|
2005-12-13 23:25:19 -08:00
|
|
|
|
[NET] Generalise TCP's struct open_request minisock infrastructure
Kept this first changeset minimal, without changing existing names to
ease peer review.
Basicaly tcp_openreq_alloc now receives the or_calltable, that in turn
has two new members:
->slab, that replaces tcp_openreq_cachep
->obj_size, to inform the size of the openreq descendant for
a specific protocol
The protocol specific fields in struct open_request were moved to a
class hierarchy, with the things that are common to all connection
oriented PF_INET protocols in struct inet_request_sock, the TCP ones
in tcp_request_sock, that is an inet_request_sock, that is an
open_request.
I.e. this uses the same approach used for the struct sock class
hierarchy, with sk_prot indicating if the protocol wants to use the
open_request infrastructure by filling in sk_prot->rsk_prot with an
or_calltable.
Results? Performance is improved and TCP v4 now uses only 64 bytes per
open request minisock, down from 96 without this patch :-)
Next changeset will rename some of the structs, fields and functions
mentioned above, struct or_calltable is way unclear, better name it
struct request_sock_ops, s/struct open_request/struct request_sock/g,
etc.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-18 22:46:52 -07:00
|
|
|
|
2008-03-22 16:50:58 -07:00
|
|
|
|
|
|
|
|
|
2008-10-29 01:41:45 -07:00
|
|
|
|
2008-03-22 16:56:51 -07:00
|
|
|
|
2017-01-09 16:55:26 +01:00
|
|
|
|
2008-03-22 16:50:58 -07:00
|
|
|
|
[SOCK] proto: Add hashinfo member to struct proto
This way we can remove TCP and DCCP specific versions of
sk->sk_prot->get_port: both v4 and v6 use inet_csk_get_port
sk->sk_prot->hash: inet_hash is directly used, only v6 need
a specific version to deal with mapped sockets
sk->sk_prot->unhash: both v4 and v6 use inet_hash directly
struct inet_connection_sock_af_ops also gets a new member, bind_conflict, so
that inet_csk_get_port can find the per family routine.
Now only the lookup routines receive as a parameter a struct inet_hashtable.
With this we further reuse code, reducing the difference among INET transport
protocols.
Eventually work has to be done on UDP and SCTP to make them share this
infrastructure and get as a bonus inet_diag interfaces so that iproute can be
used with these protocols.
net-2.6/net/ipv4/inet_hashtables.c:
struct proto | +8
struct inet_connection_sock_af_ops | +8
2 structs changed
__inet_hash_nolisten | +18
__inet_hash | -210
inet_put_port | +8
inet_bind_bucket_create | +1
__inet_hash_connect | -8
5 functions changed, 27 bytes added, 218 bytes removed, diff: -191
net-2.6/net/core/sock.c:
proto_seq_show | +3
1 function changed, 3 bytes added, diff: +3
net-2.6/net/ipv4/inet_connection_sock.c:
inet_csk_get_port | +15
1 function changed, 15 bytes added, diff: +15
net-2.6/net/ipv4/tcp.c:
tcp_set_state | -7
1 function changed, 7 bytes removed, diff: -7
net-2.6/net/ipv4/tcp_ipv4.c:
tcp_v4_get_port | -31
tcp_v4_hash | -48
tcp_v4_destroy_sock | -7
tcp_v4_syn_recv_sock | -2
tcp_unhash | -179
5 functions changed, 267 bytes removed, diff: -267
net-2.6/net/ipv6/inet6_hashtables.c:
__inet6_hash | +8
1 function changed, 8 bytes added, diff: +8
net-2.6/net/ipv4/inet_hashtables.c:
inet_unhash | +190
inet_hash | +242
2 functions changed, 432 bytes added, diff: +432
vmlinux:
16 functions changed, 485 bytes added, 492 bytes removed, diff: -7
/home/acme/git/net-2.6/net/ipv6/tcp_ipv6.c:
tcp_v6_get_port | -31
tcp_v6_hash | -7
tcp_v6_syn_recv_sock | -9
3 functions changed, 47 bytes removed, diff: -47
/home/acme/git/net-2.6/net/dccp/proto.c:
dccp_destroy_sock | -7
dccp_unhash | -179
dccp_hash | -49
dccp_set_state | -7
dccp_done | +1
5 functions changed, 1 bytes added, 242 bytes removed, diff: -241
/home/acme/git/net-2.6/net/dccp/ipv4.c:
dccp_v4_get_port | -31
dccp_v4_request_recv_sock | -2
2 functions changed, 33 bytes removed, diff: -33
/home/acme/git/net-2.6/net/dccp/ipv6.c:
dccp_v6_get_port | -31
dccp_v6_hash | -7
dccp_v6_request_recv_sock | +5
3 functions changed, 5 bytes added, 38 bytes removed, diff: -33
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2008-02-03 04:06:04 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-16 12:30:03 +09:00
|
|
|
|
2016-10-28 01:22:25 -07:00
|
|
|
|
2011-12-11 21:47:03 +00:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
sock_diag: request _diag module only when the family or proto has been registered
Now when using 'ss' in iproute, kernel would try to load all _diag
modules, which also causes corresponding family and proto modules
to be loaded as well due to module dependencies.
Like after running 'ss', sctp, dccp, af_packet (if it works as a module)
would be loaded.
For example:
$ lsmod|grep sctp
$ ss
$ lsmod|grep sctp
sctp_diag 16384 0
sctp 323584 5 sctp_diag
inet_diag 24576 4 raw_diag,tcp_diag,sctp_diag,udp_diag
libcrc32c 16384 3 nf_conntrack,nf_nat,sctp
As these family and proto modules are loaded unintentionally, it
could cause some problems, like:
- Some debug tools use 'ss' to collect the socket info, which loads all
those diag and family and protocol modules. It's noisy for identifying
issues.
- Users usually expect to drop sctp init packet silently when they
have no sense of sctp protocol instead of sending abort back.
- It wastes resources (especially with multiple netns), and SCTP module
can't be unloaded once it's loaded.
...
In short, it's really inappropriate to have these family and proto
modules loaded unexpectedly when just doing debugging with inet_diag.
This patch is to introduce sock_load_diag_module() where it loads
the _diag module only when it's corresponding family or proto has
been already registered.
Note that we can't just load _diag module without the family or
proto loaded, as some symbols used in _diag module are from the
family or proto module.
v1->v2:
- move inet proto check to inet_diag to avoid a compiling err.
v2->v3:
- define sock_load_diag_module in sock.c and export one symbol
only.
- improve the changelog.
Reported-by: Sabrina Dubroca <sd@queasysnail.net>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Phil Sutter <phil@nwl.cc>
Acked-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-03-10 18:57:50 +08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2020-11-13 07:08:09 -08:00
|
|
|
|
|
|
|
|
|
2021-10-26 16:29:14 -07:00
|
|
|
|
|
|
|
|
|
2021-11-15 11:02:38 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-26 16:29:14 -07:00
|
|
|
|
|
|
|
|
|
2018-12-04 07:58:17 -08:00
|
|
|
|
tcp: TCP_NOTSENT_LOWAT socket option
Idea of this patch is to add optional limitation of number of
unsent bytes in TCP sockets, to reduce usage of kernel memory.
TCP receiver might announce a big window, and TCP sender autotuning
might allow a large amount of bytes in write queue, but this has little
performance impact if a large part of this buffering is wasted :
Write queue needs to be large only to deal with large BDP, not
necessarily to cope with scheduling delays (incoming ACKS make room
for the application to queue more bytes)
For most workloads, using a value of 128 KB or less is OK to give
applications enough time to react to POLLOUT events in time
(or being awaken in a blocking sendmsg())
This patch adds two ways to set the limit :
1) Per socket option TCP_NOTSENT_LOWAT
2) A sysctl (/proc/sys/net/ipv4/tcp_notsent_lowat) for sockets
not using TCP_NOTSENT_LOWAT socket option (or setting a zero value)
Default value being UINT_MAX (0xFFFFFFFF), meaning this has no effect.
This changes poll()/select()/epoll() to report POLLOUT
only if number of unsent bytes is below tp->nosent_lowat
Note this might increase number of sendmsg()/sendfile() calls
when using non blocking sockets,
and increase number of context switches for blocking sockets.
Note this is not related to SO_SNDLOWAT (as SO_SNDLOWAT is
defined as :
Specify the minimum number of bytes in the buffer until
the socket layer will pass the data to the protocol)
Tested:
netperf sessions, and watching /proc/net/protocols "memory" column for TCP
With 200 concurrent netperf -t TCP_STREAM sessions, amount of kernel memory
used by TCP buffers shrinks by ~55 % (20567 pages instead of 45458)
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 45458 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 45458 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 20567 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 20567 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
Using 128KB has no bad effect on the throughput or cpu usage
of a single flow, although there is an increase of context switches.
A bonus is that we hold socket lock for a shorter amount
of time and should improve latencies of ACK processing.
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1651584 6291456 16384 20.00 17447.90 10^6bits/s 3.13 S -1.00 U 0.353 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
412,514 context-switches
200.034645535 seconds time elapsed
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1593240 6291456 16384 20.00 17321.16 10^6bits/s 3.35 S -1.00 U 0.381 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
2,675,818 context-switches
200.029651391 seconds time elapsed
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-By: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-22 20:27:07 -07:00
|
|
|
|
2019-10-10 20:17:46 -07:00
|
|
|
|
tcp: TCP_NOTSENT_LOWAT socket option
Idea of this patch is to add optional limitation of number of
unsent bytes in TCP sockets, to reduce usage of kernel memory.
TCP receiver might announce a big window, and TCP sender autotuning
might allow a large amount of bytes in write queue, but this has little
performance impact if a large part of this buffering is wasted :
Write queue needs to be large only to deal with large BDP, not
necessarily to cope with scheduling delays (incoming ACKS make room
for the application to queue more bytes)
For most workloads, using a value of 128 KB or less is OK to give
applications enough time to react to POLLOUT events in time
(or being awaken in a blocking sendmsg())
This patch adds two ways to set the limit :
1) Per socket option TCP_NOTSENT_LOWAT
2) A sysctl (/proc/sys/net/ipv4/tcp_notsent_lowat) for sockets
not using TCP_NOTSENT_LOWAT socket option (or setting a zero value)
Default value being UINT_MAX (0xFFFFFFFF), meaning this has no effect.
This changes poll()/select()/epoll() to report POLLOUT
only if number of unsent bytes is below tp->nosent_lowat
Note this might increase number of sendmsg()/sendfile() calls
when using non blocking sockets,
and increase number of context switches for blocking sockets.
Note this is not related to SO_SNDLOWAT (as SO_SNDLOWAT is
defined as :
Specify the minimum number of bytes in the buffer until
the socket layer will pass the data to the protocol)
Tested:
netperf sessions, and watching /proc/net/protocols "memory" column for TCP
With 200 concurrent netperf -t TCP_STREAM sessions, amount of kernel memory
used by TCP buffers shrinks by ~55 % (20567 pages instead of 45458)
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 45458 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 45458 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 20567 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 20567 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
Using 128KB has no bad effect on the throughput or cpu usage
of a single flow, although there is an increase of context switches.
A bonus is that we hold socket lock for a shorter amount
of time and should improve latencies of ACK processing.
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1651584 6291456 16384 20.00 17447.90 10^6bits/s 3.13 S -1.00 U 0.353 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
412,514 context-switches
200.034645535 seconds time elapsed
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1593240 6291456 16384 20.00 17321.16 10^6bits/s 3.35 S -1.00 U 0.381 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
2,675,818 context-switches
200.029651391 seconds time elapsed
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-By: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-22 20:27:07 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-27 18:12:05 -07:00
|
|
|
|
|
|
|
|
|
2018-12-04 07:58:17 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tcp: TCP_NOTSENT_LOWAT socket option
Idea of this patch is to add optional limitation of number of
unsent bytes in TCP sockets, to reduce usage of kernel memory.
TCP receiver might announce a big window, and TCP sender autotuning
might allow a large amount of bytes in write queue, but this has little
performance impact if a large part of this buffering is wasted :
Write queue needs to be large only to deal with large BDP, not
necessarily to cope with scheduling delays (incoming ACKS make room
for the application to queue more bytes)
For most workloads, using a value of 128 KB or less is OK to give
applications enough time to react to POLLOUT events in time
(or being awaken in a blocking sendmsg())
This patch adds two ways to set the limit :
1) Per socket option TCP_NOTSENT_LOWAT
2) A sysctl (/proc/sys/net/ipv4/tcp_notsent_lowat) for sockets
not using TCP_NOTSENT_LOWAT socket option (or setting a zero value)
Default value being UINT_MAX (0xFFFFFFFF), meaning this has no effect.
This changes poll()/select()/epoll() to report POLLOUT
only if number of unsent bytes is below tp->nosent_lowat
Note this might increase number of sendmsg()/sendfile() calls
when using non blocking sockets,
and increase number of context switches for blocking sockets.
Note this is not related to SO_SNDLOWAT (as SO_SNDLOWAT is
defined as :
Specify the minimum number of bytes in the buffer until
the socket layer will pass the data to the protocol)
Tested:
netperf sessions, and watching /proc/net/protocols "memory" column for TCP
With 200 concurrent netperf -t TCP_STREAM sessions, amount of kernel memory
used by TCP buffers shrinks by ~55 % (20567 pages instead of 45458)
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 45458 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 45458 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 20567 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 20567 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
Using 128KB has no bad effect on the throughput or cpu usage
of a single flow, although there is an increase of context switches.
A bonus is that we hold socket lock for a shorter amount
of time and should improve latencies of ACK processing.
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1651584 6291456 16384 20.00 17447.90 10^6bits/s 3.13 S -1.00 U 0.353 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
412,514 context-switches
200.034645535 seconds time elapsed
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1593240 6291456 16384 20.00 17321.16 10^6bits/s 3.35 S -1.00 U 0.381 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
2,675,818 context-switches
200.029651391 seconds time elapsed
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-By: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-22 20:27:07 -07:00
|
|
|
|
|
|
|
|
|
2013-07-22 20:26:31 -07:00
|
|
|
|
|
|
|
|
|
2018-12-04 07:58:17 -08:00
|
|
|
|
2013-07-22 20:26:31 -07:00
|
|
|
|
2011-12-11 21:47:03 +00:00
|
|
|
|
2016-08-18 01:00:41 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tcp: TCP_NOTSENT_LOWAT socket option
Idea of this patch is to add optional limitation of number of
unsent bytes in TCP sockets, to reduce usage of kernel memory.
TCP receiver might announce a big window, and TCP sender autotuning
might allow a large amount of bytes in write queue, but this has little
performance impact if a large part of this buffering is wasted :
Write queue needs to be large only to deal with large BDP, not
necessarily to cope with scheduling delays (incoming ACKS make room
for the application to queue more bytes)
For most workloads, using a value of 128 KB or less is OK to give
applications enough time to react to POLLOUT events in time
(or being awaken in a blocking sendmsg())
This patch adds two ways to set the limit :
1) Per socket option TCP_NOTSENT_LOWAT
2) A sysctl (/proc/sys/net/ipv4/tcp_notsent_lowat) for sockets
not using TCP_NOTSENT_LOWAT socket option (or setting a zero value)
Default value being UINT_MAX (0xFFFFFFFF), meaning this has no effect.
This changes poll()/select()/epoll() to report POLLOUT
only if number of unsent bytes is below tp->nosent_lowat
Note this might increase number of sendmsg()/sendfile() calls
when using non blocking sockets,
and increase number of context switches for blocking sockets.
Note this is not related to SO_SNDLOWAT (as SO_SNDLOWAT is
defined as :
Specify the minimum number of bytes in the buffer until
the socket layer will pass the data to the protocol)
Tested:
netperf sessions, and watching /proc/net/protocols "memory" column for TCP
With 200 concurrent netperf -t TCP_STREAM sessions, amount of kernel memory
used by TCP buffers shrinks by ~55 % (20567 pages instead of 45458)
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 45458 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 45458 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6 1880 2 20567 no 208 yes ipv6 y y y y y y y y y y y y y n y y y y y
TCP 1696 508 20567 no 208 yes kernel y y y y y y y y y y y y y n y y y y y
Using 128KB has no bad effect on the throughput or cpu usage
of a single flow, although there is an increase of context switches.
A bonus is that we hold socket lock for a shorter amount
of time and should improve latencies of ACK processing.
lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1651584 6291456 16384 20.00 17447.90 10^6bits/s 3.13 S -1.00 U 0.353 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
412,514 context-switches
200.034645535 seconds time elapsed
lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local Remote Local Elapsed Throughput Throughput Local Local Remote Remote Local Remote Service
Send Socket Recv Socket Send Time Units CPU CPU CPU CPU Service Service Demand
Size Size Size (sec) Util Util Util Util Demand Demand Units
Final Final % Method % Method
1593240 6291456 16384 20.00 17321.16 10^6bits/s 3.35 S -1.00 U 0.381 -1.000 usec/KB
Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':
2,675,818 context-switches
200.029651391 seconds time elapsed
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-By: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-22 20:27:07 -07:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-16 17:12:22 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-18 01:51:32 +00:00
|
|
|
|
2023-08-16 17:12:22 +08:00
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:03 +00:00
|
|
|
|
2016-01-14 15:21:17 -08:00
|
|
|
|
|
|
|
|
|
net: tcp_memcontrol: sanitize tcp memory accounting callbacks
There won't be a tcp control soft limit, so integrating the memcg code
into the global skmem limiting scheme complicates things unnecessarily.
Replace this with simple and clear charge and uncharge calls--hidden
behind a jump label--to account skb memory.
Note that this is not purely aesthetic: as a result of shoehorning the
per-memcg code into the same memory accounting functions that handle the
global level, the old code would compare the per-memcg consumption
against the smaller of the per-memcg limit and the global limit. This
allowed the total consumption of multiple sockets to exceed the global
limit, as long as the individual sockets stayed within bounds. After
this change, the code will always compare the per-memcg consumption to
the per-memcg limit, and the global consumption to the global limit, and
thus close this loophole.
Without a soft limit, the per-memcg memory pressure state in sockets is
generally questionable. However, we did it until now, so we continue to
enter it when the hard limit is hit, and packets are dropped, to let
other sockets in the cgroup know that they shouldn't grow their transmit
windows, either. However, keep it simple in the new callback model and
leave memory pressure lazily when the next packet is accepted (as
opposed to doing it synchroneously when packets are processed). When
packets are dropped, network performance will already be in the toilet,
so that should be a reasonable trade-off.
As described above, consumption is now checked on the per-memcg level
and the global level separately. Likewise, memory pressure states are
maintained on both the per-memcg level and the global level, and a
socket is considered under pressure when either level asserts as much.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 15:21:14 -08:00
|
|
|
|
2011-12-11 21:47:03 +00:00
|
|
|
|
2023-08-18 01:51:32 +00:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2022-06-08 23:34:09 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-08 23:34:09 -07:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2022-06-08 23:34:09 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-10 20:30:16 -07:00
|
|
|
|
net: tcp_memcontrol: sanitize tcp memory accounting callbacks
There won't be a tcp control soft limit, so integrating the memcg code
into the global skmem limiting scheme complicates things unnecessarily.
Replace this with simple and clear charge and uncharge calls--hidden
behind a jump label--to account skb memory.
Note that this is not purely aesthetic: as a result of shoehorning the
per-memcg code into the same memory accounting functions that handle the
global level, the old code would compare the per-memcg consumption
against the smaller of the per-memcg limit and the global limit. This
allowed the total consumption of multiple sockets to exceed the global
limit, as long as the individual sockets stayed within bounds. After
this change, the code will always compare the per-memcg consumption to
the per-memcg limit, and the global consumption to the global limit, and
thus close this loophole.
Without a soft limit, the per-memcg memory pressure state in sockets is
generally questionable. However, we did it until now, so we continue to
enter it when the hard limit is hit, and packets are dropped, to let
other sockets in the cgroup know that they shouldn't grow their transmit
windows, either. However, keep it simple in the new callback model and
leave memory pressure lazily when the next packet is accepted (as
opposed to doing it synchroneously when packets are processed). When
packets are dropped, network performance will already be in the toilet,
so that should be a reasonable trade-off.
As described above, consumption is now checked on the per-memcg level
and the global level separately. Likewise, memory pressure states are
maintained on both the per-memcg level and the global level, and a
socket is considered under pressure when either level asserts as much.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-14 15:21:14 -08:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
2022-06-08 23:34:09 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: introduce res_counter_charge_nofail() for socket allocations
There is a case in __sk_mem_schedule(), where an allocation
is beyond the maximum, but yet we are allowed to proceed.
It happens under the following condition:
sk->sk_wmem_queued + size >= sk->sk_sndbuf
The network code won't revert the allocation in this case,
meaning that at some point later it'll try to do it. Since
this is never communicated to the underlying res_counter
code, there is an inbalance in res_counter uncharge operation.
I see two ways of fixing this:
1) storing the information about those allocations somewhere
in memcg, and then deducting from that first, before
we start draining the res_counter,
2) providing a slightly different allocation function for
the res_counter, that matches the original behavior of
the network code more closely.
I decided to go for #2 here, believing it to be more elegant,
since #1 would require us to do basically that, but in a more
obscure way.
Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
CC: Tejun Heo <tj@kernel.org>
CC: Li Zefan <lizf@cn.fujitsu.com>
CC: Laurent Chavey <chavey@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-01-20 04:57:16 +00:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
2022-06-08 23:34:09 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2021-02-02 11:34:08 -08:00
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2021-02-02 11:34:08 -08:00
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-02 11:34:08 -08:00
|
|
|
|
|
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2019-02-12 12:26:27 -08:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2016-01-14 15:21:08 -08:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-18 01:51:32 +00:00
|
|
|
|
2011-12-11 21:47:02 +00:00
|
|
|
|
|
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:47 -08:00
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:49 -08:00
|
|
|
|
2021-11-15 09:11:47 -08:00
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:50 -08:00
|
|
|
|
2021-11-15 09:11:47 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:50 -08:00
|
|
|
|
2021-11-15 09:11:47 -08:00
|
|
|
|
2021-11-15 09:11:48 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:49 -08:00
|
|
|
|
2021-11-15 09:11:48 -08:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2017-12-14 05:51:58 -08:00
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
2021-11-15 09:11:47 -08:00
|
|
|
|
|
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
|
|
|
|
|
2021-11-15 09:11:48 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-01-03 20:46:48 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2005-08-09 19:47:37 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-02-10 11:50:35 -05:00
|
|
|
|
2005-08-09 19:47:37 -07:00
|
|
|
|
|
|
|
|
|
2016-02-10 11:50:35 -05:00
|
|
|
|
2005-08-09 19:47:37 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-10-21 13:55:45 +02:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2016-10-21 13:55:45 +02:00
|
|
|
|
2015-05-15 12:39:25 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-06-08 23:34:06 -07:00
|
|
|
|
2016-10-31 13:32:55 -07:00
|
|
|
|
|
|
|
|
|
2022-07-14 14:19:42 -07:00
|
|
|
|
2016-10-31 13:32:55 -07:00
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-06-08 23:34:07 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-09-01 17:48:23 -07:00
|
|
|
|
2022-06-08 23:34:10 -07:00
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2022-06-08 23:34:10 -07:00
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
netvm: prevent a stream-specific deadlock
This patch series is based on top of "Swap-over-NBD without deadlocking
v15" as it depends on the same reservation of PF_MEMALLOC reserves logic.
When a user or administrator requires swap for their application, they
create a swap partition and file, format it with mkswap and activate it
with swapon. In diskless systems this is not an option so if swap if
required then swapping over the network is considered. The two likely
scenarios are when blade servers are used as part of a cluster where the
form factor or maintenance costs do not allow the use of disks and thin
clients.
The Linux Terminal Server Project recommends the use of the Network Block
Device (NBD) for swap but this is not always an option. There is no
guarantee that the network attached storage (NAS) device is running Linux
or supports NBD. However, it is likely that it supports NFS so there are
users that want support for swapping over NFS despite any performance
concern. Some distributions currently carry patches that support swapping
over NFS but it would be preferable to support it in the mainline kernel.
Patch 1 avoids a stream-specific deadlock that potentially affects TCP.
Patch 2 is a small modification to SELinux to avoid using PFMEMALLOC
reserves.
Patch 3 adds three helpers for filesystems to handle swap cache pages.
For example, page_file_mapping() returns page->mapping for
file-backed pages and the address_space of the underlying
swap file for swap cache pages.
Patch 4 adds two address_space_operations to allow a filesystem
to pin all metadata relevant to a swapfile in memory. Upon
successful activation, the swapfile is marked SWP_FILE and
the address space operation ->direct_IO is used for writing
and ->readpage for reading in swap pages.
Patch 5 notes that patch 3 is bolting
filesystem-specific-swapfile-support onto the side and that
the default handlers have different information to what
is available to the filesystem. This patch refactors the
code so that there are generic handlers for each of the new
address_space operations.
Patch 6 adds an API to allow a vector of kernel addresses to be
translated to struct pages and pinned for IO.
Patch 7 adds support for using highmem pages for swap by kmapping
the pages before calling the direct_IO handler.
Patch 8 updates NFS to use the helpers from patch 3 where necessary.
Patch 9 avoids setting PF_private on PG_swapcache pages within NFS.
Patch 10 implements the new swapfile-related address_space operations
for NFS and teaches the direct IO handler how to manage
kernel addresses.
Patch 11 prevents page allocator recursions in NFS by using GFP_NOIO
where appropriate.
Patch 12 fixes a NULL pointer dereference that occurs when using
swap-over-NFS.
With the patches applied, it is possible to mount a swapfile that is on an
NFS filesystem. Swap performance is not great with a swap stress test
taking roughly twice as long to complete than if the swap device was
backed by NBD.
This patch: netvm: prevent a stream-specific deadlock
It could happen that all !SOCK_MEMALLOC sockets have buffered so much data
that we're over the global rmem limit. This will prevent SOCK_MEMALLOC
buffers from receiving data, which will prevent userspace from running,
which is needed to reduce the buffered data.
Fix this by exempting the SOCK_MEMALLOC sockets from the rmem limit. Once
this change it applied, it is important that sockets that set
SOCK_MEMALLOC do not clear the flag until the socket is being torn down.
If this happens, a warning is generated and the tokens reclaimed to avoid
accounting errors until the bug is fixed.
[davem@davemloft.net: Warning about clearing SOCK_MEMALLOC]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Neil Brown <neilb@suse.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-07-31 16:44:41 -07:00
|
|
|
|
2012-09-17 14:09:11 -07:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2022-06-08 23:34:10 -07:00
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2022-06-08 23:34:10 -07:00
|
|
|
|
|
|
|
|
|
netvm: prevent a stream-specific deadlock
This patch series is based on top of "Swap-over-NBD without deadlocking
v15" as it depends on the same reservation of PF_MEMALLOC reserves logic.
When a user or administrator requires swap for their application, they
create a swap partition and file, format it with mkswap and activate it
with swapon. In diskless systems this is not an option so if swap if
required then swapping over the network is considered. The two likely
scenarios are when blade servers are used as part of a cluster where the
form factor or maintenance costs do not allow the use of disks and thin
clients.
The Linux Terminal Server Project recommends the use of the Network Block
Device (NBD) for swap but this is not always an option. There is no
guarantee that the network attached storage (NAS) device is running Linux
or supports NBD. However, it is likely that it supports NFS so there are
users that want support for swapping over NFS despite any performance
concern. Some distributions currently carry patches that support swapping
over NFS but it would be preferable to support it in the mainline kernel.
Patch 1 avoids a stream-specific deadlock that potentially affects TCP.
Patch 2 is a small modification to SELinux to avoid using PFMEMALLOC
reserves.
Patch 3 adds three helpers for filesystems to handle swap cache pages.
For example, page_file_mapping() returns page->mapping for
file-backed pages and the address_space of the underlying
swap file for swap cache pages.
Patch 4 adds two address_space_operations to allow a filesystem
to pin all metadata relevant to a swapfile in memory. Upon
successful activation, the swapfile is marked SWP_FILE and
the address space operation ->direct_IO is used for writing
and ->readpage for reading in swap pages.
Patch 5 notes that patch 3 is bolting
filesystem-specific-swapfile-support onto the side and that
the default handlers have different information to what
is available to the filesystem. This patch refactors the
code so that there are generic handlers for each of the new
address_space operations.
Patch 6 adds an API to allow a vector of kernel addresses to be
translated to struct pages and pinned for IO.
Patch 7 adds support for using highmem pages for swap by kmapping
the pages before calling the direct_IO handler.
Patch 8 updates NFS to use the helpers from patch 3 where necessary.
Patch 9 avoids setting PF_private on PG_swapcache pages within NFS.
Patch 10 implements the new swapfile-related address_space operations
for NFS and teaches the direct IO handler how to manage
kernel addresses.
Patch 11 prevents page allocator recursions in NFS by using GFP_NOIO
where appropriate.
Patch 12 fixes a NULL pointer dereference that occurs when using
swap-over-NFS.
With the patches applied, it is possible to mount a swapfile that is on an
NFS filesystem. Swap performance is not great with a swap stress test
taking roughly twice as long to complete than if the swap device was
backed by NBD.
This patch: netvm: prevent a stream-specific deadlock
It could happen that all !SOCK_MEMALLOC sockets have buffered so much data
that we're over the global rmem limit. This will prevent SOCK_MEMALLOC
buffers from receiving data, which will prevent userspace from running,
which is needed to reduce the buffered data.
Fix this by exempting the SOCK_MEMALLOC sockets from the rmem limit. Once
this change it applied, it is important that sockets that set
SOCK_MEMALLOC do not clear the flag until the socket is being torn down.
If this happens, a warning is generated and the tokens reclaimed to avoid
accounting errors until the bug is fixed.
[davem@davemloft.net: Warning about clearing SOCK_MEMALLOC]
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Neil Brown <neilb@suse.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-07-31 16:44:41 -07:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-08 23:34:07 -07:00
|
|
|
|
2021-09-29 10:25:11 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-08 23:34:11 -07:00
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
|
|
|
|
|
2006-12-06 20:35:24 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2006-12-06 20:35:24 -08:00
|
|
|
|
2008-11-12 01:38:36 +00:00
|
|
|
|
2006-12-06 20:35:24 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2006-12-06 20:35:24 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-01-17 07:14:14 -08:00
|
|
|
|
2016-04-05 17:10:15 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2006-11-08 22:44:35 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-27 11:10:22 +01:00
|
|
|
|
2018-10-01 23:24:26 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-07-03 00:25:13 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
net: introduce and use lock_sock_fast_nested()
Syzkaller reported a false positive deadlock involving
the nl socket lock and the subflow socket lock:
MPTCP: kernel_bind error, err=-98
============================================
WARNING: possible recursive locking detected
5.15.0-rc1-syzkaller #0 Not tainted
--------------------------------------------
syz-executor998/6520 is trying to acquire lock:
ffff8880795718a0 (k-sk_lock-AF_INET){+.+.}-{0:0}, at: mptcp_close+0x267/0x7b0 net/mptcp/protocol.c:2738
but task is already holding lock:
ffff8880787c8c60 (k-sk_lock-AF_INET){+.+.}-{0:0}, at: lock_sock include/net/sock.h:1612 [inline]
ffff8880787c8c60 (k-sk_lock-AF_INET){+.+.}-{0:0}, at: mptcp_close+0x23/0x7b0 net/mptcp/protocol.c:2720
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(k-sk_lock-AF_INET);
lock(k-sk_lock-AF_INET);
*** DEADLOCK ***
May be due to missing lock nesting notation
3 locks held by syz-executor998/6520:
#0: ffffffff8d176c50 (cb_lock){++++}-{3:3}, at: genl_rcv+0x15/0x40 net/netlink/genetlink.c:802
#1: ffffffff8d176d08 (genl_mutex){+.+.}-{3:3}, at: genl_lock net/netlink/genetlink.c:33 [inline]
#1: ffffffff8d176d08 (genl_mutex){+.+.}-{3:3}, at: genl_rcv_msg+0x3e0/0x580 net/netlink/genetlink.c:790
#2: ffff8880787c8c60 (k-sk_lock-AF_INET){+.+.}-{0:0}, at: lock_sock include/net/sock.h:1612 [inline]
#2: ffff8880787c8c60 (k-sk_lock-AF_INET){+.+.}-{0:0}, at: mptcp_close+0x23/0x7b0 net/mptcp/protocol.c:2720
stack backtrace:
CPU: 1 PID: 6520 Comm: syz-executor998 Not tainted 5.15.0-rc1-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
print_deadlock_bug kernel/locking/lockdep.c:2944 [inline]
check_deadlock kernel/locking/lockdep.c:2987 [inline]
validate_chain kernel/locking/lockdep.c:3776 [inline]
__lock_acquire.cold+0x149/0x3ab kernel/locking/lockdep.c:5015
lock_acquire kernel/locking/lockdep.c:5625 [inline]
lock_acquire+0x1ab/0x510 kernel/locking/lockdep.c:5590
lock_sock_fast+0x36/0x100 net/core/sock.c:3229
mptcp_close+0x267/0x7b0 net/mptcp/protocol.c:2738
inet_release+0x12e/0x280 net/ipv4/af_inet.c:431
__sock_release net/socket.c:649 [inline]
sock_release+0x87/0x1b0 net/socket.c:677
mptcp_pm_nl_create_listen_socket+0x238/0x2c0 net/mptcp/pm_netlink.c:900
mptcp_nl_cmd_add_addr+0x359/0x930 net/mptcp/pm_netlink.c:1170
genl_family_rcv_msg_doit+0x228/0x320 net/netlink/genetlink.c:731
genl_family_rcv_msg net/netlink/genetlink.c:775 [inline]
genl_rcv_msg+0x328/0x580 net/netlink/genetlink.c:792
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504
genl_rcv+0x24/0x40 net/netlink/genetlink.c:803
netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline]
netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340
netlink_sendmsg+0x86d/0xdb0 net/netlink/af_netlink.c:1929
sock_sendmsg_nosec net/socket.c:704 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:724
sock_no_sendpage+0x101/0x150 net/core/sock.c:2980
kernel_sendpage.part.0+0x1a0/0x340 net/socket.c:3504
kernel_sendpage net/socket.c:3501 [inline]
sock_sendpage+0xe5/0x140 net/socket.c:1003
pipe_to_sendpage+0x2ad/0x380 fs/splice.c:364
splice_from_pipe_feed fs/splice.c:418 [inline]
__splice_from_pipe+0x43e/0x8a0 fs/splice.c:562
splice_from_pipe fs/splice.c:597 [inline]
generic_splice_sendpage+0xd4/0x140 fs/splice.c:746
do_splice_from fs/splice.c:767 [inline]
direct_splice_actor+0x110/0x180 fs/splice.c:936
splice_direct_to_actor+0x34b/0x8c0 fs/splice.c:891
do_splice_direct+0x1b3/0x280 fs/splice.c:979
do_sendfile+0xae9/0x1240 fs/read_write.c:1249
__do_sys_sendfile64 fs/read_write.c:1314 [inline]
__se_sys_sendfile64 fs/read_write.c:1300 [inline]
__x64_sys_sendfile64+0x1cc/0x210 fs/read_write.c:1300
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7f215cb69969
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 14 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 c0 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ffc96bb3868 EFLAGS: 00000246 ORIG_RAX: 0000000000000028
RAX: ffffffffffffffda RBX: 00007f215cbad072 RCX: 00007f215cb69969
RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000000000000005
RBP: 0000000000000000 R08: 00007ffc96bb3a08 R09: 00007ffc96bb3a08
R10: 0000000100000002 R11: 0000000000000246 R12: 00007ffc96bb387c
R13: 431bde82d7b634db R14: 0000000000000000 R15: 0000000000000000
the problem originates from uncorrect lock annotation in the mptcp
code and is only visible since commit 2dcb96bacce3 ("net: core: Correct
the sock::sk_lock.owned lockdep annotations"), but is present since
the port-based endpoint support initial implementation.
This patch addresses the issue introducing a nested variant of
lock_sock_fast() and using it in the relevant code path.
Fixes: 1729cf186d8a ("mptcp: create the listening socket for new port")
Fixes: 2dcb96bacce3 ("net: core: Correct the sock::sk_lock.owned lockdep annotations")
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Reported-and-tested-by: syzbot+1dd53f7a89b299d59eaf@syzkaller.appspotmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2021-09-29 11:59:17 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-17 19:43:49 +01:00
|
|
|
|
2010-05-26 19:20:18 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-17 19:43:49 +01:00
|
|
|
|
2010-04-28 14:35:48 -07:00
|
|
|
|
2020-11-17 19:43:49 +01:00
|
|
|
|
2010-05-26 19:20:18 +00:00
|
|
|
|
2020-11-17 19:43:49 +01:00
|
|
|
|
|
|
|
|
|
net: core: Correct the sock::sk_lock.owned lockdep annotations
lock_sock_fast() and lock_sock_nested() contain lockdep annotations for the
sock::sk_lock.owned 'mutex'. sock::sk_lock.owned is not a regular mutex. It
is just lockdep wise equivalent. In fact it's an open coded trivial mutex
implementation with some interesting features.
sock::sk_lock.slock is a regular spinlock protecting the 'mutex'
representation sock::sk_lock.owned which is a plain boolean. If 'owned' is
true, then some other task holds the 'mutex', otherwise it is uncontended.
As this locking construct is obviously endangered by lock ordering issues as
any other locking primitive it got lockdep annotated via a dedicated
dependency map sock::sk_lock.dep_map which has to be updated at the lock
and unlock sites.
lock_sock_nested() is a straight forward 'mutex' lock operation:
might_sleep();
spin_lock_bh(sock::sk_lock.slock)
while (!try_lock(sock::sk_lock.owned)) {
spin_unlock_bh(sock::sk_lock.slock);
wait_for_release();
spin_lock_bh(sock::sk_lock.slock);
}
The lockdep annotation for sock::sk_lock.owned is for unknown reasons
_after_ the lock has been acquired, i.e. after the code block above and
after releasing sock::sk_lock.slock, but inside the bottom halves disabled
region:
spin_unlock(sock::sk_lock.slock);
mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
local_bh_enable();
The placement after the unlock is obvious because otherwise the
mutex_acquire() would nest into the spin lock held region.
But that's from the lockdep perspective still the wrong place:
1) The mutex_acquire() is issued _after_ the successful acquisition which
is pointless because in a dead lock scenario this point is never
reached which means that if the deadlock is the first instance of
exposing the wrong lock order lockdep does not have a chance to detect
it.
2) It only works because lockdep is rather lax on the context from which
the mutex_acquire() is issued. Acquiring a mutex inside a bottom halves
and therefore non-preemptible region is obviously invalid, except for a
trylock which is clearly not the case here.
This 'works' stops working on RT enabled kernels where the bottom halves
serialization is done via a local lock, which exposes this misplacement
because the 'mutex' and the local lock nest the wrong way around and
lockdep complains rightfully about a lock inversion.
The placement is wrong since the initial commit a5b5bb9a053a ("[PATCH]
lockdep: annotate sk_locks") which introduced this.
Fix it by moving the mutex_acquire() in front of the actual lock
acquisition, which is what the regular mutex_lock() operation does as well.
lock_sock_fast() is not that straight forward. It looks at the first glance
like a convoluted trylock operation:
spin_lock_bh(sock::sk_lock.slock)
if (!sock::sk_lock.owned)
return false;
while (!try_lock(sock::sk_lock.owned)) {
spin_unlock_bh(sock::sk_lock.slock);
wait_for_release();
spin_lock_bh(sock::sk_lock.slock);
}
spin_unlock(sock::sk_lock.slock);
mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
local_bh_enable();
return true;
But that's not the case: lock_sock_fast() is an interesting optimization
for short critical sections which can run with bottom halves disabled and
sock::sk_lock.slock held. This allows to shortcut the 'mutex' operation in
the non contended case by preventing other lockers to acquire
sock::sk_lock.owned because they are blocked on sock::sk_lock.slock, which
in turn avoids the overhead of doing the heavy processing in release_sock()
including waking up wait queue waiters.
In the contended case, i.e. when sock::sk_lock.owned == true the behavior
is the same as lock_sock_nested().
Semantically this shortcut means, that the task acquired the 'mutex' even
if it does not touch the sock::sk_lock.owned field in the non-contended
case. Not telling lockdep about this shortcut acquisition is hiding
potential lock ordering violations in the fast path.
As a consequence the same reasoning as for the above lock_sock_nested()
case vs. the placement of the lockdep annotation applies.
The current placement of the lockdep annotation was just copied from
the original lock_sock(), now renamed to lock_sock_nested(),
implementation.
Fix this by moving the mutex_acquire() in front of the actual lock
acquisition and adding the corresponding mutex_release() into
unlock_sock_fast(). Also document the fast path return case with a comment.
Reported-by: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: netdev@vger.kernel.org
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2021-09-18 14:42:35 +02:00
|
|
|
|
2010-05-26 19:20:18 +00:00
|
|
|
|
2020-11-17 19:43:49 +01:00
|
|
|
|
2010-04-28 14:35:48 -07:00
|
|
|
|
|
|
|
|
|
2022-08-16 23:17:17 -07:00
|
|
|
|
|
|
|
|
|
2022-08-16 23:17:23 -07:00
|
|
|
|
|
|
|
|
|
2022-08-16 23:17:17 -07:00
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-03 16:56:03 -07:00
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
|
|
|
|
|
2016-04-25 06:34:09 -07:00
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
2016-05-03 16:56:03 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-12-28 11:00:43 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-12-08 15:21:58 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-12-08 15:21:58 +09:00
|
|
|
|
|
|
|
|
|
2016-04-08 15:11:27 +02:00
|
|
|
|
2010-04-28 14:35:48 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2015-05-08 21:09:13 -05:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2015-06-15 11:26:18 -04:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2017-03-01 16:35:08 -03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2016-05-02 10:56:27 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2017-08-03 16:29:37 -04:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2014-09-04 13:31:35 -04:00
|
|
|
|
2014-09-04 13:32:11 -04:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2020-03-29 15:53:38 -07:00
|
|
|
|
2014-09-04 13:32:11 -04:00
|
|
|
|
2017-01-27 07:11:27 -08:00
|
|
|
|
2014-09-04 13:32:11 -04:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-08-16 23:18:04 -07:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2020-07-23 08:08:50 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-09-01 17:29:12 -07:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2019-04-17 22:51:48 +02:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-28 11:58:44 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2014-11-19 17:13:11 +01:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-10-27 00:25:57 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 14:56:48 -07:00
|
|
|
|
2018-07-03 15:42:48 -07:00
|
|
|
|
2015-10-08 14:56:48 -07:00
|
|
|
|
2022-12-07 09:37:01 -05:00
|
|
|
|
2015-10-08 14:56:48 -07:00
|
|
|
|
|
|
|
|
|
2018-07-06 10:12:56 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-20 06:54:41 +00:00
|
|
|
|
2016-04-02 23:08:06 -04:00
|
|
|
|
2015-10-08 14:56:48 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-09 08:09:05 +00:00
|
|
|
|
2018-02-12 20:00:20 +01:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-02 15:37:48 +08:00
|
|
|
|
2017-07-28 16:22:41 -07:00
|
|
|
|
2015-03-02 15:37:48 +08:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-28 16:22:41 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2015-03-02 15:37:48 +08:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2020-07-23 08:09:07 +02:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2023-02-04 17:39:20 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:01 +03:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
tcp/dccp: remove twchain
TCP listener refactoring, part 3 :
Our goal is to hash SYN_RECV sockets into main ehash for fast lookup,
and parallel SYN processing.
Current inet_ehash_bucket contains two chains, one for ESTABLISH (and
friend states) sockets, another for TIME_WAIT sockets only.
As the hash table is sized to get at most one socket per bucket, it
makes little sense to have separate twchain, as it makes the lookup
slightly more complicated, and doubles hash table memory usage.
If we make sure all socket types have the lookup keys at the same
offsets, we can use a generic and faster lookup. It turns out TIME_WAIT
and ESTABLISHED sockets already have common lookup fields for IPv4.
[ INET_TW_MATCH() is no longer needed ]
I'll provide a follow-up to factorize IPv6 lookup as well, to remove
INET6_TW_MATCH()
This way, SYN_RECV pseudo sockets will be supported the same.
A new sock_gen_put() helper is added, doing either a sock_put() or
inet_twsk_put() [ and will support SYN_RECV later ].
Note this helper should only be called in real slow path, when rcu
lookup found a socket that was moved to another identity (freed/reused
immediately), but could eventually be used in other contexts, like
sock_edemux()
Before patch :
dmesg | grep "TCP established"
TCP established hash table entries: 524288 (order: 11, 8388608 bytes)
After patch :
TCP established hash table entries: 524288 (order: 10, 4194304 bytes)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 00:22:02 -07:00
|
|
|
|
2015-03-12 16:44:08 -07:00
|
|
|
|
tcp/dccp: remove twchain
TCP listener refactoring, part 3 :
Our goal is to hash SYN_RECV sockets into main ehash for fast lookup,
and parallel SYN processing.
Current inet_ehash_bucket contains two chains, one for ESTABLISH (and
friend states) sockets, another for TIME_WAIT sockets only.
As the hash table is sized to get at most one socket per bucket, it
makes little sense to have separate twchain, as it makes the lookup
slightly more complicated, and doubles hash table memory usage.
If we make sure all socket types have the lookup keys at the same
offsets, we can use a generic and faster lookup. It turns out TIME_WAIT
and ESTABLISHED sockets already have common lookup fields for IPv4.
[ INET_TW_MATCH() is no longer needed ]
I'll provide a follow-up to factorize IPv6 lookup as well, to remove
INET6_TW_MATCH()
This way, SYN_RECV pseudo sockets will be supported the same.
A new sock_gen_put() helper is added, doing either a sock_put() or
inet_twsk_put() [ and will support SYN_RECV later ].
Note this helper should only be called in real slow path, when rcu
lookup found a socket that was moved to another identity (freed/reused
immediately), but could eventually be used in other contexts, like
sock_edemux()
Before patch :
dmesg | grep "TCP established"
TCP established hash table entries: 524288 (order: 11, 8388608 bytes)
After patch :
TCP established hash table entries: 524288 (order: 10, 4194304 bytes)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-10-03 00:22:02 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2016-07-12 18:18:57 -04:00
|
|
|
|
2016-11-02 17:14:41 -07:00
|
|
|
|
2016-07-12 18:18:57 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-11-02 17:14:41 -07:00
|
|
|
|
2016-07-12 18:18:57 -04:00
|
|
|
|
2005-12-27 02:42:22 -02:00
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:51 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:51 -07:00
|
|
|
|
|
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:51 -07:00
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:51 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
|
|
|
|
|
2021-11-30 10:29:39 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-30 10:29:39 -08:00
|
|
|
|
|
|
|
|
|
2021-10-25 09:48:19 -07:00
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-30 10:29:39 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2021-10-25 09:48:20 -07:00
|
|
|
|
2018-06-29 21:26:57 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-06-29 21:27:02 -07:00
|
|
|
|
|
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2021-10-25 09:48:20 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-11 13:35:51 +02:00
|
|
|
|
2018-06-29 21:27:02 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-06-17 22:41:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-20 13:03:51 +00:00
|
|
|
|
|
|
|
|
|
2011-02-18 03:26:36 +00:00
|
|
|
|
|
|
|
|
|
2010-04-20 13:03:51 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-06-17 22:41:38 -07:00
|
|
|
|
2010-04-29 11:01:49 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-06 08:15:07 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2019-07-05 20:14:16 +01:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2008-06-17 22:41:38 -07:00
|
|
|
|
net: core: Add a UID field to struct sock.
Protocol sockets (struct sock) don't have UIDs, but most of the
time, they map 1:1 to userspace sockets (struct socket) which do.
Various operations such as the iptables xt_owner match need
access to the "UID of a socket", and do so by following the
backpointer to the struct socket. This involves taking
sk_callback_lock and doesn't work when there is no socket
because userspace has already called close().
Simplify this by adding a sk_uid field to struct sock whose value
matches the UID of the corresponding struct socket. The semantics
are as follows:
1. Whenever sk_socket is non-null: sk_uid is the same as the UID
in sk_socket, i.e., matches the return value of sock_i_uid.
Specifically, the UID is set when userspace calls socket(),
fchown(), or accept().
2. When sk_socket is NULL, sk_uid is defined as follows:
- For a socket that no longer has a sk_socket because
userspace has called close(): the previous UID.
- For a cloned socket (e.g., an incoming connection that is
established but on which userspace has not yet called
accept): the UID of the socket it was cloned from.
- For a socket that has never had an sk_socket: UID 0 inside
the user namespace corresponding to the network namespace
the socket belongs to.
Kernel sockets created by sock_create_kern are a special case
of #1 and sk_uid is the user that created them. For kernel
sockets created at network namespace creation time, such as the
per-processor ICMP and TCP sockets, this is the user that created
the network namespace.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-04 02:23:41 +09:00
|
|
|
|
2006-07-24 23:32:50 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2023-06-26 09:43:13 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
net: core: Add a UID field to struct sock.
Protocol sockets (struct sock) don't have UIDs, but most of the
time, they map 1:1 to userspace sockets (struct socket) which do.
Various operations such as the iptables xt_owner match need
access to the "UID of a socket", and do so by following the
backpointer to the struct socket. This involves taking
sk_callback_lock and doesn't work when there is no socket
because userspace has already called close().
Simplify this by adding a sk_uid field to struct sock whose value
matches the UID of the corresponding struct socket. The semantics
are as follows:
1. Whenever sk_socket is non-null: sk_uid is the same as the UID
in sk_socket, i.e., matches the return value of sock_i_uid.
Specifically, the UID is set when userspace calls socket(),
fchown(), or accept().
2. When sk_socket is NULL, sk_uid is defined as follows:
- For a socket that no longer has a sk_socket because
userspace has called close(): the previous UID.
- For a cloned socket (e.g., an incoming connection that is
established but on which userspace has not yet called
accept): the UID of the socket it was cloned from.
- For a socket that has never had an sk_socket: UID 0 inside
the user namespace corresponding to the network namespace
the socket belongs to.
Kernel sockets created by sock_create_kern are a special case
of #1 and sk_uid is the user that created them. For kernel
sockets created at network namespace creation time, such as the
per-processor ICMP and TCP sockets, this is the user that created
the network namespace.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-04 02:23:41 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-09-15 15:24:20 -07:00
|
|
|
|
|
|
|
|
|
2022-10-05 17:43:22 +02:00
|
|
|
|
2015-09-15 15:24:20 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-28 16:02:05 -07:00
|
|
|
|
|
|
|
|
|
2021-06-10 07:44:11 -07:00
|
|
|
|
|
|
|
|
|
2015-07-28 16:02:05 -07:00
|
|
|
|
|
|
|
|
|
2021-01-19 11:26:19 -08:00
|
|
|
|
2015-07-28 16:02:06 -07:00
|
|
|
|
2022-01-31 16:31:22 +03:00
|
|
|
|
2015-07-28 16:02:06 -07:00
|
|
|
|
2021-01-19 11:26:19 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-07-28 16:02:06 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-05 17:10:15 +02:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
|
|
|
|
|
2023-03-23 21:55:32 +01:00
|
|
|
|
2014-06-24 10:05:11 -07:00
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-01-19 11:26:19 -08:00
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-10-22 16:23:38 +08:00
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-01-19 11:26:19 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-10-19 23:46:20 +00:00
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
2017-03-06 11:23:55 -08:00
|
|
|
|
|
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-30 01:26:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
2014-07-02 02:39:38 -07:00
|
|
|
|
2014-06-30 01:26:23 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-08 23:03:29 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-30 01:26:23 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
|
|
|
|
|
2019-11-05 14:11:51 -08:00
|
|
|
|
|
|
|
|
|
2017-02-06 23:14:11 +02:00
|
|
|
|
|
|
|
|
|
2017-02-06 23:14:12 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-05 14:11:51 -08:00
|
|
|
|
|
|
|
|
|
2021-11-23 10:54:30 +08:00
|
|
|
|
2017-02-06 23:14:12 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-04-01 17:07:44 +02:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2006-06-30 13:36:35 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-08-09 19:49:02 -07:00
|
|
|
|
2021-11-15 11:02:35 -08:00
|
|
|
|
2010-05-16 00:36:33 -07:00
|
|
|
|
2021-11-15 11:02:35 -08:00
|
|
|
|
|
|
|
|
|
2010-05-16 00:36:33 -07:00
|
|
|
|
|
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2014-11-28 13:40:20 -05:00
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
2014-11-28 13:40:20 -05:00
|
|
|
|
2016-11-01 22:42:45 -04:00
|
|
|
|
2014-11-28 13:40:20 -05:00
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2016-11-01 22:42:45 -04:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2016-11-01 22:42:45 -04:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-11-28 13:40:20 -05:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
|
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-11-28 13:40:20 -05:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-04-06 18:40:12 +00:00
|
|
|
|
|
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-22 18:09:03 +02:00
|
|
|
|
2019-10-10 20:17:46 -07:00
|
|
|
|
2011-04-04 22:30:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:00 +03:00
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2009-06-16 10:12:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2015-11-26 13:55:39 +08:00
|
|
|
|
2010-05-24 23:54:18 -07:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2015-11-26 13:55:39 +08:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
2017-05-12 09:35:46 -03:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2017-05-12 09:35:46 -03:00
|
|
|
|
|
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-29 11:01:49 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-07-08 12:10:31 +00:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2015-11-26 13:55:39 +08:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2015-11-26 13:55:39 +08:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-10-23 13:40:39 +02:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
2010-04-29 11:01:49 +00:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2018-10-23 13:40:39 +02:00
|
|
|
|
|
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
2018-07-30 09:42:11 +02:00
|
|
|
|
2019-07-05 20:14:16 +01:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
2010-04-29 11:01:49 +00:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2009-07-08 12:09:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-01 21:32:17 -07:00
|
|
|
|
|
|
|
|
|
2021-06-10 07:44:11 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-01 21:32:17 -07:00
|
|
|
|
2021-06-10 07:44:11 -07:00
|
|
|
|
2014-07-01 21:32:17 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-11-01 15:36:55 -08:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-22 02:25:25 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-31 00:11:19 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2021-05-11 10:35:21 +02:00
|
|
|
|
2021-03-30 18:43:54 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-11 10:35:21 +02:00
|
|
|
|
2021-03-30 18:43:54 +02:00
|
|
|
|
2021-05-11 10:35:21 +02:00
|
|
|
|
2021-03-30 18:43:54 +02:00
|
|
|
|
|
|
|
|
|
2023-02-09 16:22:01 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-07-28 18:24:03 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2020-09-24 08:30:01 +08:00
|
|
|
|
|
|
|
|
|
2017-05-16 11:20:13 +02:00
|
|
|
|
|
|
|
|
|
2017-02-05 09:25:24 -08:00
|
|
|
|
|
|
|
|
|
2016-04-05 12:41:15 -04:00
|
|
|
|
2022-04-07 14:20:49 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
2014-08-31 21:30:27 -04:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-12-13 23:22:19 -08:00
|
|
|
|
2021-06-10 07:27:37 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-12-13 23:22:19 -08:00
|
|
|
|
2021-06-10 07:27:37 -07:00
|
|
|
|
2005-12-13 23:22:19 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-06-27 18:48:21 -04:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-06-30 13:08:00 +03:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:10 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2016-04-25 10:39:34 -07:00
|
|
|
|
|
|
|
|
|
2016-04-25 10:39:32 -07:00
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
2015-11-29 20:03:10 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-04-25 10:39:34 -07:00
|
|
|
|
|
|
|
|
|
2016-04-25 10:39:32 -07:00
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
2015-11-29 20:03:10 -08:00
|
|
|
|
|
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2015-11-29 20:03:11 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
net: sock: adapt SOCK_MIN_RCVBUF and SOCK_MIN_SNDBUF
The current situation is that SOCK_MIN_RCVBUF is 2048 + sizeof(struct sk_buff))
while SOCK_MIN_SNDBUF is 2048. Since in both cases, skb->truesize is used for
sk_{r,w}mem_alloc accounting, we should have both sizes adjusted via defining a
TCP_SKB_MIN_TRUESIZE.
Further, as Eric Dumazet points out, the minimal skb truesize in transmit path is
SKB_TRUESIZE(2048) after commit f07d960df33c5 ("tcp: avoid frag allocation for
small frames"), and tcp_sendmsg() tries to limit skb size to half the congestion
window, meaning we try to build two skbs at minimum. Thus, having SOCK_MIN_SNDBUF
as 2048 can hit a small regression for some applications setting to low
SO_SNDBUF / SO_RCVBUF. Note that we define a TCP_SKB_MIN_TRUESIZE, because
SKB_TRUESIZE(2048) adds SKB_DATA_ALIGN(sizeof(struct skb_shared_info)), but in
case of TCP skbs, the skb_shared_info is part of the 2048 bytes allocation for
skb->head.
The minor adaption in sk_stream_moderate_sndbuf() is to silence a warning by
using a typed max macro, as similarly done in SOCK_MIN_RCVBUF occurences, that
would appear otherwise.
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-19 12:51:20 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-09-26 18:53:07 -07:00
|
|
|
|
2013-07-03 05:02:22 -07:00
|
|
|
|
net: sock: adapt SOCK_MIN_RCVBUF and SOCK_MIN_SNDBUF
The current situation is that SOCK_MIN_RCVBUF is 2048 + sizeof(struct sk_buff))
while SOCK_MIN_SNDBUF is 2048. Since in both cases, skb->truesize is used for
sk_{r,w}mem_alloc accounting, we should have both sizes adjusted via defining a
TCP_SKB_MIN_TRUESIZE.
Further, as Eric Dumazet points out, the minimal skb truesize in transmit path is
SKB_TRUESIZE(2048) after commit f07d960df33c5 ("tcp: avoid frag allocation for
small frames"), and tcp_sendmsg() tries to limit skb size to half the congestion
window, meaning we try to build two skbs at minimum. Thus, having SOCK_MIN_SNDBUF
as 2048 can hit a small regression for some applications setting to low
SO_SNDBUF / SO_RCVBUF. Note that we define a TCP_SKB_MIN_TRUESIZE, because
SKB_TRUESIZE(2048) adds SKB_DATA_ALIGN(sizeof(struct skb_shared_info)), but in
case of TCP skbs, the skb_shared_info is part of the 2048 bytes allocation for
skb->head.
The minor adaption in sk_stream_moderate_sndbuf() is to silence a warning by
using a typed max macro, as similarly done in SOCK_MIN_RCVBUF occurences, that
would appear otherwise.
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-19 12:51:20 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-10 20:17:45 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-29 10:25:12 -07:00
|
|
|
|
2019-10-10 20:17:45 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: fix sk_page_frag() recursion from memory reclaim
sk_page_frag() optimizes skb_frag allocations by using per-task
skb_frag cache when it knows it's the only user. The condition is
determined by seeing whether the socket allocation mask allows
blocking - if the allocation may block, it obviously owns the task's
context and ergo exclusively owns current->task_frag.
Unfortunately, this misses recursion through memory reclaim path.
Please take a look at the following backtrace.
[2] RIP: 0010:tcp_sendmsg_locked+0xccf/0xe10
...
tcp_sendmsg+0x27/0x40
sock_sendmsg+0x30/0x40
sock_xmit.isra.24+0xa1/0x170 [nbd]
nbd_send_cmd+0x1d2/0x690 [nbd]
nbd_queue_rq+0x1b5/0x3b0 [nbd]
__blk_mq_try_issue_directly+0x108/0x1b0
blk_mq_request_issue_directly+0xbd/0xe0
blk_mq_try_issue_list_directly+0x41/0xb0
blk_mq_sched_insert_requests+0xa2/0xe0
blk_mq_flush_plug_list+0x205/0x2a0
blk_flush_plug_list+0xc3/0xf0
[1] blk_finish_plug+0x21/0x2e
_xfs_buf_ioapply+0x313/0x460
__xfs_buf_submit+0x67/0x220
xfs_buf_read_map+0x113/0x1a0
xfs_trans_read_buf_map+0xbf/0x330
xfs_btree_read_buf_block.constprop.42+0x95/0xd0
xfs_btree_lookup_get_block+0x95/0x170
xfs_btree_lookup+0xcc/0x470
xfs_bmap_del_extent_real+0x254/0x9a0
__xfs_bunmapi+0x45c/0xab0
xfs_bunmapi+0x15/0x30
xfs_itruncate_extents_flags+0xca/0x250
xfs_free_eofblocks+0x181/0x1e0
xfs_fs_destroy_inode+0xa8/0x1b0
destroy_inode+0x38/0x70
dispose_list+0x35/0x50
prune_icache_sb+0x52/0x70
super_cache_scan+0x120/0x1a0
do_shrink_slab+0x120/0x290
shrink_slab+0x216/0x2b0
shrink_node+0x1b6/0x4a0
do_try_to_free_pages+0xc6/0x370
try_to_free_mem_cgroup_pages+0xe3/0x1e0
try_charge+0x29e/0x790
mem_cgroup_charge_skmem+0x6a/0x100
__sk_mem_raise_allocated+0x18e/0x390
__sk_mem_schedule+0x2a/0x40
[0] tcp_sendmsg_locked+0x8eb/0xe10
tcp_sendmsg+0x27/0x40
sock_sendmsg+0x30/0x40
___sys_sendmsg+0x26d/0x2b0
__sys_sendmsg+0x57/0xa0
do_syscall_64+0x42/0x100
entry_SYSCALL_64_after_hwframe+0x44/0xa9
In [0], tcp_send_msg_locked() was using current->page_frag when it
called sk_wmem_schedule(). It already calculated how many bytes can
be fit into current->page_frag. Due to memory pressure,
sk_wmem_schedule() called into memory reclaim path which called into
xfs and then IO issue path. Because the filesystem in question is
backed by nbd, the control goes back into the tcp layer - back into
tcp_sendmsg_locked().
nbd sets sk_allocation to (GFP_NOIO | __GFP_MEMALLOC) which makes
sense - it's in the process of freeing memory and wants to be able to,
e.g., drop clean pages to make forward progress. However, this
confused sk_page_frag() called from [2]. Because it only tests
whether the allocation allows blocking which it does, it now thinks
current->page_frag can be used again although it already was being
used in [0].
After [2] used current->page_frag, the offset would be increased by
the used amount. When the control returns to [0],
current->page_frag's offset is increased and the previously calculated
number of bytes now may overrun the end of allocated memory leading to
silent memory corruptions.
Fix it by adding gfpflags_normal_context() which tests sleepable &&
!reclaim and use it to determine whether to use current->task_frag.
v2: Eric didn't like gfp flags being tested twice. Introduce a new
helper gfpflags_normal_context() and combine the two tests.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-10-24 13:50:27 -07:00
|
|
|
|
2021-11-26 19:34:21 +01:00
|
|
|
|
net: fix sk_page_frag() recursion from memory reclaim
sk_page_frag() optimizes skb_frag allocations by using per-task
skb_frag cache when it knows it's the only user. The condition is
determined by seeing whether the socket allocation mask allows
blocking - if the allocation may block, it obviously owns the task's
context and ergo exclusively owns current->task_frag.
Unfortunately, this misses recursion through memory reclaim path.
Please take a look at the following backtrace.
[2] RIP: 0010:tcp_sendmsg_locked+0xccf/0xe10
...
tcp_sendmsg+0x27/0x40
sock_sendmsg+0x30/0x40
sock_xmit.isra.24+0xa1/0x170 [nbd]
nbd_send_cmd+0x1d2/0x690 [nbd]
nbd_queue_rq+0x1b5/0x3b0 [nbd]
__blk_mq_try_issue_directly+0x108/0x1b0
blk_mq_request_issue_directly+0xbd/0xe0
blk_mq_try_issue_list_directly+0x41/0xb0
blk_mq_sched_insert_requests+0xa2/0xe0
blk_mq_flush_plug_list+0x205/0x2a0
blk_flush_plug_list+0xc3/0xf0
[1] blk_finish_plug+0x21/0x2e
_xfs_buf_ioapply+0x313/0x460
__xfs_buf_submit+0x67/0x220
xfs_buf_read_map+0x113/0x1a0
xfs_trans_read_buf_map+0xbf/0x330
xfs_btree_read_buf_block.constprop.42+0x95/0xd0
xfs_btree_lookup_get_block+0x95/0x170
xfs_btree_lookup+0xcc/0x470
xfs_bmap_del_extent_real+0x254/0x9a0
__xfs_bunmapi+0x45c/0xab0
xfs_bunmapi+0x15/0x30
xfs_itruncate_extents_flags+0xca/0x250
xfs_free_eofblocks+0x181/0x1e0
xfs_fs_destroy_inode+0xa8/0x1b0
destroy_inode+0x38/0x70
dispose_list+0x35/0x50
prune_icache_sb+0x52/0x70
super_cache_scan+0x120/0x1a0
do_shrink_slab+0x120/0x290
shrink_slab+0x216/0x2b0
shrink_node+0x1b6/0x4a0
do_try_to_free_pages+0xc6/0x370
try_to_free_mem_cgroup_pages+0xe3/0x1e0
try_charge+0x29e/0x790
mem_cgroup_charge_skmem+0x6a/0x100
__sk_mem_raise_allocated+0x18e/0x390
__sk_mem_schedule+0x2a/0x40
[0] tcp_sendmsg_locked+0x8eb/0xe10
tcp_sendmsg+0x27/0x40
sock_sendmsg+0x30/0x40
___sys_sendmsg+0x26d/0x2b0
__sys_sendmsg+0x57/0xa0
do_syscall_64+0x42/0x100
entry_SYSCALL_64_after_hwframe+0x44/0xa9
In [0], tcp_send_msg_locked() was using current->page_frag when it
called sk_wmem_schedule(). It already calculated how many bytes can
be fit into current->page_frag. Due to memory pressure,
sk_wmem_schedule() called into memory reclaim path which called into
xfs and then IO issue path. Because the filesystem in question is
backed by nbd, the control goes back into the tcp layer - back into
tcp_sendmsg_locked().
nbd sets sk_allocation to (GFP_NOIO | __GFP_MEMALLOC) which makes
sense - it's in the process of freeing memory and wants to be able to,
e.g., drop clean pages to make forward progress. However, this
confused sk_page_frag() called from [2]. Because it only tests
whether the allocation allows blocking which it does, it now thinks
current->page_frag can be used again although it already was being
used in [0].
After [2] used current->page_frag, the offset would be increased by
the used amount. When the control returns to [0],
current->page_frag's offset is increased and the previously calculated
number of bytes now may overrun the end of allocated memory leading to
silent memory corruptions.
Fix it by adding gfpflags_normal_context() which tests sleepable &&
!reclaim and use it to determine whether to use current->task_frag.
v2: Eric didn't like gfp flags being tested twice. Introduce a new
helper gfpflags_normal_context() and combine the two tests.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-10-24 13:50:27 -07:00
|
|
|
|
|
|
|
|
|
2021-11-26 19:34:21 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-16 07:45:28 -05:00
|
|
|
|
2020-02-15 11:42:37 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2022-12-16 07:45:28 -05:00
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
net: use a per task frag allocator
We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.
This page is used to build fragments for skbs.
Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)
But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page
Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.
This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.
(up to 32768 bytes per frag, thats order-3 pages on x86)
This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.
Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536
Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-23 23:04:42 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2019-10-10 20:17:45 -07:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-10-07 07:46:04 +01:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2009-02-12 16:43:17 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2021-08-17 12:40:03 -07:00
|
|
|
|
|
|
|
|
|
2022-10-21 09:03:04 -07:00
|
|
|
|
2021-08-17 12:40:03 -07:00
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-09 15:32:35 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-01 14:58:31 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-12-09 10:31:43 -08:00
|
|
|
|
2015-03-01 14:58:31 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-01 14:58:29 +02:00
|
|
|
|
2015-03-01 14:58:31 +02:00
|
|
|
|
2015-03-01 14:58:29 +02:00
|
|
|
|
2015-03-01 14:58:30 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-12-07 10:05:36 -08:00
|
|
|
|
|
|
|
|
|
2015-03-01 14:58:30 +02:00
|
|
|
|
|
|
|
|
|
2016-04-01 08:52:19 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-04 21:38:43 -08:00
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-04 21:38:43 -08:00
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-03-25 22:14:49 -07:00
|
|
|
|
2012-05-16 22:48:15 +00:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2007-04-19 16:16:32 -07:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
2005-08-14 17:24:31 -07:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
|
|
|
|
|
2014-08-04 22:11:46 -04:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-08-04 22:11:46 -04:00
|
|
|
|
2016-12-25 11:38:40 +01:00
|
|
|
|
|
|
|
|
|
2014-08-04 22:11:46 -04:00
|
|
|
|
2007-03-25 22:14:49 -07:00
|
|
|
|
|
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
2011-11-09 10:15:42 +01:00
|
|
|
|
2023-04-17 08:53:46 -07:00
|
|
|
|
2011-11-09 10:15:42 +01:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2022-04-27 16:02:37 -04:00
|
|
|
|
|
|
|
|
|
2010-04-28 19:14:43 +00:00
|
|
|
|
2017-03-30 14:03:06 +02:00
|
|
|
|
2022-04-27 16:02:37 -04:00
|
|
|
|
|
|
|
|
|
2010-04-28 19:14:43 +00:00
|
|
|
|
2022-04-27 16:02:37 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-08-04 22:11:46 -04:00
|
|
|
|
|
|
|
|
|
2010-04-28 19:14:43 +00:00
|
|
|
|
2022-04-27 16:02:37 -04:00
|
|
|
|
|
|
|
|
|
2017-03-31 14:59:25 -07:00
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
2023-05-08 10:55:43 -07:00
|
|
|
|
2018-12-27 18:55:09 -08:00
|
|
|
|
2010-04-28 19:14:43 +00:00
|
|
|
|
net: Generalize socket rx gap / receive queue overflow cmsg
Create a new socket level option to report number of queue overflows
Recently I augmented the AF_PACKET protocol to report the number of frames lost
on the socket receive queue between any two enqueued frames. This value was
exported via a SOL_PACKET level cmsg. AFter I completed that work it was
requested that this feature be generalized so that any datagram oriented socket
could make use of this option. As such I've created this patch, It creates a
new SOL_SOCKET level option called SO_RXQ_OVFL, which when enabled exports a
SOL_SOCKET level cmsg that reports the nubmer of times the sk_receive_queue
overflowed between any two given frames. It also augments the AF_PACKET
protocol to take advantage of this new feature (as it previously did not touch
sk->sk_drops, which this patch uses to record the overflow count). Tested
successfully by me.
Notes:
1) Unlike my previous patch, this patch simply records the sk_drops value, which
is not a number of drops between packets, but rather a total number of drops.
Deltas must be computed in user space.
2) While this patch currently works with datagram oriented protocols, it will
also be accepted by non-datagram oriented protocols. I'm not sure if thats
agreeable to everyone, but my argument in favor of doing so is that, for those
protocols which aren't applicable to this option, sk_drops will always be zero,
and reporting no drops on a receive queue that isn't used for those
non-participating protocols seems reasonable to me. This also saves us having
to code in a per-protocol opt in mechanism.
3) This applies cleanly to net-next assuming that commit
977750076d98c7ff6cbda51858bb5a5894a9d9ab (my af packet cmsg patch) is reverted
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-10-12 13:26:31 -07:00
|
|
|
|
2016-04-02 23:08:12 -04:00
|
|
|
|
2014-09-08 19:58:58 -04:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
2016-04-02 23:08:12 -04:00
|
|
|
|
2014-08-06 11:49:29 +02:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
2014-08-06 11:49:29 +02:00
|
|
|
|
2017-05-12 09:35:46 -03:00
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
|
|
|
|
|
2014-09-08 19:58:58 -04:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
2016-04-02 23:08:12 -04:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
|
|
|
|
|
2022-02-17 09:05:02 -08:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
2014-09-08 19:58:58 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-02-12 05:03:38 +00:00
|
|
|
|
2018-12-17 12:24:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-15 11:02:33 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
2005-05-01 08:59:25 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-12-30 12:37:29 -08:00
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:40 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:39 -07:00
|
|
|
|
2020-04-07 22:55:25 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:39 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-10-07 12:41:01 -07:00
|
|
|
|
2012-06-24 13:03:07 +00:00
|
|
|
|
2008-10-07 12:41:01 -07:00
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:39 -07:00
|
|
|
|
2020-03-29 15:53:40 -07:00
|
|
|
|
|
|
|
|
|
2008-10-07 12:41:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-29 15:53:39 -07:00
|
|
|
|
2008-10-07 12:41:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-04-30 10:16:12 +03:00
|
|
|
|
|
|
|
|
|
2019-08-07 17:03:59 -07:00
|
|
|
|
2018-04-30 10:16:12 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-08-07 17:03:59 -07:00
|
|
|
|
2018-04-30 10:16:12 +03:00
|
|
|
|
2019-08-07 17:03:59 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-04-30 10:16:12 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-10-08 05:01:55 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-03 23:56:37 +03:00
|
|
|
|
2013-09-22 10:32:26 -07:00
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
2014-04-23 14:26:56 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-20 15:22:03 -04:00
|
|
|
|
|
|
|
|
|
2017-08-29 15:16:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-30 13:29:32 -05:00
|
|
|
|
2005-09-05 18:14:11 -07:00
|
|
|
|
|
|
|
|
|
2005-08-16 02:18:02 -03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-26 10:49:47 +08:00
|
|
|
|
2019-06-14 16:22:21 -07:00
|
|
|
|
|
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-07-22 11:22:00 -07:00
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
2022-07-22 11:22:00 -07:00
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-07-22 11:22:00 -07:00
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
2022-07-22 11:22:00 -07:00
|
|
|
|
2017-11-07 00:29:27 -08:00
|
|
|
|
|
|
|
|
|
2017-12-12 06:34:19 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-12-16 18:51:03 -08:00
|
|
|
|
2017-12-12 06:34:19 -08:00
|
|
|
|
2019-12-16 18:51:03 -08:00
|
|
|
|
2017-12-12 06:34:19 -08:00
|
|
|
|
|
|
|
|
|
2018-01-04 14:03:54 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-13 11:55:41 -07:00
|
|
|
|
2018-01-04 14:03:54 -08:00
|
|
|
|
|
|
|
|
|
2022-05-13 11:55:41 -07:00
|
|
|
|
2018-01-04 14:03:54 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-13 11:55:41 -07:00
|
|
|
|
2018-01-04 14:03:54 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-01-20 10:29:17 +01:00
|
|
|
|
|
|
|
|
|
2020-05-30 23:09:00 +02:00
|
|
|
|
2021-06-03 16:24:27 -07:00
|
|
|
|
2021-06-30 16:11:59 +08:00
|
|
|
|
|
|
|
|
|
2021-06-03 16:24:28 -07:00
|
|
|
|
2020-05-28 07:12:14 +02:00
|
|
|
|
2020-05-28 07:12:10 +02:00
|
|
|
|
2020-05-28 07:12:15 +02:00
|
|
|
|
2020-05-28 07:12:11 +02:00
|
|
|
|
2020-05-28 07:12:16 +02:00
|
|
|
|
2020-06-26 13:26:48 -04:00
|
|
|
|
2020-05-28 07:12:09 +02:00
|
|
|
|
2020-05-28 07:12:17 +02:00
|
|
|
|
2020-05-28 07:12:12 +02:00
|
|
|
|
2020-05-28 07:12:09 +02:00
|
|
|
|
2020-05-29 14:09:42 +02:00
|
|
|
|
|
|
|
|
|
2021-10-08 11:00:53 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-08 13:33:03 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2005-04-16 15:20:36 -07:00
|
|
|
|
