[Patch net] inet: fix sleeping inside inet_wait_for_connect()
Subsystems:
networking [general] , the rest
3563d
5 messages,
4 authors,
2016-11-04 · open the first message on its own page
Andrey reported this kernel warning:
WARNING: CPU: 0 PID: 4608 at kernel/sched/core.c:7724
__might_sleep+0x14c/0x1a0 kernel/sched/core.c:7719
do not call blocking ops when !TASK_RUNNING; state=1 set at
[<ffffffff811f5a5c>] prepare_to_wait+0xbc/0x210
kernel/sched/wait.c:178
Modules linked in:
CPU: 0 PID: 4608 Comm: syz-executor Not tainted 4.9.0-rc2+ #320
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
ffff88006625f7a0 ffffffff81b46914 ffff88006625f818 0000000000000000
ffffffff84052960 0000000000000000 ffff88006625f7e8 ffffffff81111237
ffff88006aceac00 ffffffff00001e2c ffffed000cc4beff ffffffff84052960
Call Trace:
[< inline >] __dump_stack lib/dump_stack.c:15
[<ffffffff81b46914>] dump_stack+0xb3/0x10f lib/dump_stack.c:51
[<ffffffff81111237>] __warn+0x1a7/0x1f0 kernel/panic.c:550
[<ffffffff8111132c>] warn_slowpath_fmt+0xac/0xd0 kernel/panic.c:565
[<ffffffff811922fc>] __might_sleep+0x14c/0x1a0 kernel/sched/core.c:7719
[< inline >] slab_pre_alloc_hook mm/slab.h:393
[< inline >] slab_alloc_node mm/slub.c:2634
[< inline >] slab_alloc mm/slub.c:2716
[<ffffffff81508da0>] __kmalloc_track_caller+0x150/0x2a0 mm/slub.c:4240
[<ffffffff8146be14>] kmemdup+0x24/0x50 mm/util.c:113
[<ffffffff8388b2cf>] dccp_feat_clone_sp_val.part.5+0x4f/0xe0 net/dccp/feat.c:374
[< inline >] dccp_feat_clone_sp_val net/dccp/feat.c:1141
[< inline >] dccp_feat_change_recv net/dccp/feat.c:1141
[<ffffffff8388d491>] dccp_feat_parse_options+0xaa1/0x13d0 net/dccp/feat.c:1411
[<ffffffff83894f01>] dccp_parse_options+0x721/0x1010 net/dccp/options.c:128
[<ffffffff83891280>] dccp_rcv_state_process+0x200/0x15b0 net/dccp/input.c:644
[<ffffffff838b8a94>] dccp_v4_do_rcv+0xf4/0x1a0 net/dccp/ipv4.c:681
[< inline >] sk_backlog_rcv ./include/net/sock.h:872
[<ffffffff82b7ceb6>] __release_sock+0x126/0x3a0 net/core/sock.c:2044
[<ffffffff82b7d189>] release_sock+0x59/0x1c0 net/core/sock.c:2502
[< inline >] inet_wait_for_connect net/ipv4/af_inet.c:547
[<ffffffff8316b2a2>] __inet_stream_connect+0x5d2/0xbb0 net/ipv4/af_inet.c:617
[<ffffffff8316b8d5>] inet_stream_connect+0x55/0xa0 net/ipv4/af_inet.c:656
[<ffffffff82b705e4>] SYSC_connect+0x244/0x2f0 net/socket.c:1533
[<ffffffff82b72dd4>] SyS_connect+0x24/0x30 net/socket.c:1514
[<ffffffff83fbf701>] entry_SYSCALL_64_fastpath+0x1f/0xc2
arch/x86/entry/entry_64.S:209
Unlike commit 26cabd31259ba43f68026ce3f62b78094124333f
("sched, net: Clean up sk_wait_event() vs. might_sleep()"), the
sleeping function is called before schedule_timeout(), this is indeed
a bug. Fix this by moving the wait logic to the new API, it is similar
to commit ff960a731788a7408b6f66ec4fd772ff18833211
("netdev, sched/wait: Fix sleeping inside wait event").
Reported-by: Andrey Konovalov <redacted>
Cc: Andrey Konovalov <redacted>
Cc: Eric Dumazet <redacted>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Cong Wang <redacted>
---
net/ipv4/af_inet.c | 9 ++++-----
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 9648c97..5ddf5cd 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c @@ -533,9 +533,9 @@ EXPORT_SYMBOL(inet_dgram_connect);
static long inet_wait_for_connect ( struct sock * sk , long timeo , int writebias )
{
- DEFINE_WAIT ( wait ); + DEFINE_WAIT_FUNC ( wait , woken_wake_function );
- prepare_to_wait ( sk_sleep ( sk ), & wait , TASK_INTERRUPTIBLE ); + add_wait_queue ( sk_sleep ( sk ), & wait ); sk -> sk_write_pending += writebias ;
/* Basic assumption: if someone sets sk->sk_err, he _must_ @@ -545,13 +545,12 @@ static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
*/
while (( 1 << sk -> sk_state ) & ( TCPF_SYN_SENT | TCPF_SYN_RECV )) {
release_sock ( sk );
- timeo = schedule_timeout ( timeo ); + timeo = wait_woken ( & wait , TASK_INTERRUPTIBLE , timeo ); lock_sock ( sk );
if ( signal_pending ( current ) || ! timeo )
break ;
- prepare_to_wait ( sk_sleep ( sk ), & wait , TASK_INTERRUPTIBLE ); }
- finish_wait ( sk_sleep ( sk ), & wait ); + remove_wait_queue ( sk_sleep ( sk ), & wait ); sk -> sk_write_pending -= writebias ;
return timeo ;
} --
2.1.0
On Tue, 2016-11-01 at 16:04 -0700, Cong Wang wrote: Andrey reported this kernel warning: Unlike commit 26cabd31259ba43f68026ce3f62b78094124333f
("sched, net: Clean up sk_wait_event() vs. might_sleep()"), the
sleeping function is called before schedule_timeout(), this is indeed
a bug. Fix this by moving the wait logic to the new API, it is similar
to commit ff960a731788a7408b6f66ec4fd772ff18833211
("netdev, sched/wait: Fix sleeping inside wait event").
Reported-by: Andrey Konovalov <redacted>
Cc: Andrey Konovalov <redacted>
Cc: Eric Dumazet <redacted>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Cong Wang <redacted>
---
Excellent.
I guess we could also define sk_wait_event_woken()
and use it instead of sk_wait_event(), and also in
inet_wait_for_connect()
+#define sk_wait_event_woken(__sk, __timeo, __condition, __wait) \
+ ({ int __rc; \
+ release_sock(__sk); \
+ __rc = __condition; \
+ if (!__rc) { \
+ *(__timeo) = wait_woken(__wait, TASK_INTERRUPTIBLE, \
+ *(__timeo)); \
+ } \
+ lock_sock(__sk); \
+ __rc = __condition; \
+ __rc; \
+ })
sk_wait_data() would need :
@@ -2078,14 +2080,14 @@ void __sk_flush_backlog(struct sock *sk)
*/
int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
{
+ DEFINE_WAIT_FUNC(wait, woken_wake_function); int rc;
- DEFINE_WAIT(wait);
- prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);+ add_wait_queue(sk_sleep(sk), &wait); sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
- rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);+ rc = sk_wait_event_woken(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait); sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
- finish_wait(sk_sleep(sk), &wait);+ remove_wait_queue(sk_sleep(sk), &wait); return rc;
}
EXPORT_SYMBOL(sk_wait_data);
On Tue, Nov 1, 2016 at 6:54 PM, Eric Dumazet [off-list ref] wrote: On Tue, 2016-11-01 at 16:04 -0700, Cong Wang wrote: quoted Andrey reported this kernel warning: quoted Unlike commit 26cabd31259ba43f68026ce3f62b78094124333f
("sched, net: Clean up sk_wait_event() vs. might_sleep()"), the
sleeping function is called before schedule_timeout(), this is indeed
a bug. Fix this by moving the wait logic to the new API, it is similar
to commit ff960a731788a7408b6f66ec4fd772ff18833211
("netdev, sched/wait: Fix sleeping inside wait event").
Reported-by: Andrey Konovalov <redacted>
Cc: Andrey Konovalov <redacted>
Cc: Eric Dumazet <redacted>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Cong Wang <redacted>
---
Excellent.
I guess we could also define sk_wait_event_woken()
and use it instead of sk_wait_event(), and also in
inet_wait_for_connect()
Agreed, I will send some followup patches to address this,
probably all release_sock() before a schedule_*() need
to fix.
Thanks!
From: Cong Wang <redacted>
Date: Tue, 1 Nov 2016 16:04:36 -0700
Andrey reported this kernel warning: ... Unlike commit 26cabd31259ba43f68026ce3f62b78094124333f
("sched, net: Clean up sk_wait_event() vs. might_sleep()"), the
sleeping function is called before schedule_timeout(), this is indeed
a bug. Fix this by moving the wait logic to the new API, it is similar
to commit ff960a731788a7408b6f66ec4fd772ff18833211
("netdev, sched/wait: Fix sleeping inside wait event").
Reported-by: Andrey Konovalov <redacted>
Cc: Andrey Konovalov <redacted>
Cc: Eric Dumazet <redacted>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Cong Wang <redacted>
Applied, thanks.
Greeting,
FYI, we noticed a 4.2% improvement of netperf.Throughput_tps due to commit:
commit 34630bc07341e30e42a974840a0d622b31f1845c ("inet: fix sleeping inside inet_wait_for_connect()")
https://github.com/0day-ci/linux Cong-Wang/inet-fix-sleeping-inside-inet_wait_for_connect/20161102-070925
in testcase: netperf
on test machine: 8 threads Intel(R) Atom(TM) CPU C2750 @ 2.40GHz with 16G memory
with following parameters:
ip: ipv4
runtime: 300s
nr_threads: 200%
cluster: cs-localhost
test: TCP_CRR
cpufreq_governor: performance
Netperf is a benchmark that can be use to measure various aspect of networking performance.
Disclaimer:
Results have been estimated based on internal Intel analysis and are provided
for informational purposes only. Any difference in system hardware or software
design or configuration may affect actual performance.
Details are as below:
-------------------------------------------------------------------------------------------------->
To reproduce:
git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
cd lkp-tests
bin/lkp install job.yaml # job file is attached in this email
bin/lkp run job.yaml
=========================================================================================
cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
cs-localhost/gcc-6/performance/ipv4/x86_64-rhel-7.2/200%/debian-x86_64-2016-08-31.cgz/300s/lkp-avoton3/TCP_CRR/netperf
commit:
b9b84fc07d ("net: mv643xx_eth: ensure coalesce settings survive read-modify-write")
34630bc073 ("inet: fix sleeping inside inet_wait_for_connect()")
b9b84fc07dbc6ac7 34630bc07341e30e42a974840a
---------------- --------------------------
%stddev %change %stddev
\ | \
2915 ± 0% +4.2% 3039 ± 0% netperf.Throughput_tps
20900055 ± 0% -97.4% 540516 ± 8% netperf.time.involuntary_context_switches
517.25 ± 0% +13.8% 588.50 ± 0% netperf.time.percent_of_cpu_this_job_got
1536 ± 0% +13.9% 1750 ± 0% netperf.time.system_time
7199745 ± 1% +100.6% 14442880 ± 0% netperf.time.voluntary_context_switches
18454 ± 1% +28.3% 23680 ± 4% softirqs.SCHED
0.08 ± 0% +34.4% 0.11 ± 7% turbostat.CPU%c1
187728 ± 0% -44.8% 103607 ± 0% vmstat.system.cs
9065 ± 1% +24.6% 11296 ± 4% vmstat.system.in
118058 ± 13% +582.9% 806220 ± 17% cpuidle.C1-AVN.time
2948 ± 28% +822.4% 27193 ± 17% cpuidle.C1-AVN.usage
24.75 ± 16% +238.4% 83.75 ± 20% cpuidle.POLL.usage
2121 ± 6% -16.2% 1778 ± 6% slabinfo.cred_jar.num_objs
5215 ± 23% +42.3% 7423 ± 3% slabinfo.kmalloc-32.active_objs
5215 ± 23% +42.3% 7423 ± 3% slabinfo.kmalloc-32.num_objs
2.561e+10 ± 0% -1.4% 2.526e+10 ± 0% perf-stat.branch-misses
1.27 ± 3% +14.3% 1.45 ± 2% perf-stat.cache-miss-rate%
2.882e+09 ± 3% +13.4% 3.269e+09 ± 2% perf-stat.cache-misses
57816048 ± 0% -45.0% 31803360 ± 0% perf-stat.context-switches
8221 ± 2% +24.9% 10268 ± 5% perf-stat.cpu-migrations
2.315e+10 ± 0% -1.1% 2.289e+10 ± 0% perf-stat.iTLB-load-misses
285020 ± 0% +4.2% 296904 ± 0% perf-stat.minor-faults
285022 ± 0% +4.2% 296904 ± 0% perf-stat.page-faults
231706 ± 10% +143.8% 564885 ± 13% sched_debug.cfs_rq:/.MIN_vruntime.avg
645748 ± 13% +49.9% 967850 ± 0% sched_debug.cfs_rq:/.MIN_vruntime.max
297027 ± 10% +52.4% 452531 ± 3% sched_debug.cfs_rq:/.MIN_vruntime.stddev
231706 ± 10% +143.8% 564885 ± 13% sched_debug.cfs_rq:/.max_vruntime.avg
645748 ± 13% +49.9% 967851 ± 0% sched_debug.cfs_rq:/.max_vruntime.max
297027 ± 10% +52.4% 452532 ± 3% sched_debug.cfs_rq:/.max_vruntime.stddev
769174 ± 0% +24.3% 955802 ± 0% sched_debug.cfs_rq:/.min_vruntime.avg
786747 ± 1% +23.4% 970602 ± 0% sched_debug.cfs_rq:/.min_vruntime.max
758458 ± 0% +24.1% 941243 ± 0% sched_debug.cfs_rq:/.min_vruntime.min
1.23 ± 3% +24.9% 1.54 ± 2% sched_debug.cfs_rq:/.nr_running.avg
1.79 ± 4% +18.6% 2.12 ± 6% sched_debug.cfs_rq:/.nr_running.max
0.41 ± 8% +21.7% 0.50 ± 5% sched_debug.cfs_rq:/.nr_running.stddev
107.00 ± 2% -20.2% 85.38 ± 10% sched_debug.cfs_rq:/.runnable_load_avg.min
11.17 ± 14% +66.3% 18.58 ± 23% sched_debug.cfs_rq:/.util_avg.stddev
390011 ± 29% -55.3% 174458 ± 12% sched_debug.cpu.avg_idle.avg
813413 ± 18% -65.5% 280413 ± 22% sched_debug.cpu.avg_idle.max
257758 ± 22% -77.5% 58027 ± 35% sched_debug.cpu.avg_idle.stddev
3.38 ± 25% -45.6% 1.84 ± 20% sched_debug.cpu.clock.stddev
3.38 ± 25% -45.6% 1.84 ± 20% sched_debug.cpu.clock_task.stddev
115535 ± 4% +16.2% 134278 ± 7% sched_debug.cpu.load.min
0.58 ± 3% +22.7% 0.71 ± 8% sched_debug.cpu.nr_running.stddev
3579054 ± 0% -45.0% 1966781 ± 0% sched_debug.cpu.nr_switches.avg
3645312 ± 0% -43.6% 2056851 ± 0% sched_debug.cpu.nr_switches.max
3525172 ± 0% -45.9% 1908138 ± 0% sched_debug.cpu.nr_switches.min
0.00 ± 18% -88.7% 0.00 ± 58% sched_debug.rt_rq:/.rt_time.min
15.67 ± 0% -17.4% 12.94 ± 1% perf-profile.calltrace.cycles-pp.SYSC_recvfrom.sys_recvfrom.entry_SYSCALL_64_fastpath
16.23 ± 0% +19.3% 19.36 ± 0% perf-profile.calltrace.cycles-pp.SYSC_sendto.sys_sendto.entry_SYSCALL_64_fastpath
2.41 ± 3% +10.7% 2.67 ± 1% perf-profile.calltrace.cycles-pp.__release_sock.release_sock.tcp_close.inet_release.sock_release
0.69 ± 2% +137.3% 1.64 ± 3% perf-profile.calltrace.cycles-pp.__release_sock.release_sock.tcp_sendmsg.inet_sendmsg.sock_sendmsg
1.17 ± 4% -100.0% 0.00 ± -1% perf-profile.calltrace.cycles-pp.__schedule.schedule.schedule_timeout.__inet_stream_connect.inet_stream_connect
2.06 ± 1% -13.5% 1.78 ± 2% perf-profile.calltrace.cycles-pp.__schedule.schedule.schedule_timeout.sk_wait_data.tcp_recvmsg
9.07 ± 0% -28.3% 6.51 ± 2% perf-profile.calltrace.cycles-pp.__tcp_ack_snd_check.tcp_rcv_established.tcp_v4_do_rcv.tcp_prequeue_process.tcp_recvmsg
0.82 ± 1% +109.5% 1.71 ± 4% perf-profile.calltrace.cycles-pp.__tcp_ack_snd_check.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
12.40 ± 0% +16.7% 14.46 ± 1% perf-profile.calltrace.cycles-pp.__tcp_push_pending_frames.tcp_push.tcp_sendmsg.inet_sendmsg.sock_sendmsg
1.94 ± 3% -25.4% 1.45 ± 1% perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_sync_key.tcp_prequeue.tcp_v4_rcv.ip_local_deliver_finish
2.04 ± 3% -25.8% 1.51 ± 2% perf-profile.calltrace.cycles-pp.__wake_up_sync_key.tcp_prequeue.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
1.78 ± 3% -14.8% 1.52 ± 1% perf-profile.calltrace.cycles-pp.activate_task.ttwu_do_activate.try_to_wake_up.default_wake_function.autoremove_wake_function
1.87 ± 3% -26.0% 1.39 ± 1% perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_sync_key.tcp_prequeue.tcp_v4_rcv
1.82 ± 3% -25.9% 1.35 ± 1% perf-profile.calltrace.cycles-pp.default_wake_function.autoremove_wake_function.__wake_up_common.__wake_up_sync_key.tcp_prequeue
2.00 ± 3% -24.8% 1.51 ± 3% perf-profile.calltrace.cycles-pp.dev_hard_start_xmit.__dev_queue_xmit.dev_queue_xmit.ip_finish_output2.ip_finish_output
1.15 ± 5% -65.9% 0.39 ± 57% perf-profile.calltrace.cycles-pp.enqueue_entity.enqueue_task_fair.activate_task.ttwu_do_activate.try_to_wake_up
1.57 ± 3% -14.8% 1.34 ± 1% perf-profile.calltrace.cycles-pp.enqueue_task_fair.activate_task.ttwu_do_activate.try_to_wake_up.default_wake_function
14.95 ± 0% -18.8% 12.14 ± 1% perf-profile.calltrace.cycles-pp.inet_recvmsg.sock_recvmsg.SYSC_recvfrom.sys_recvfrom.entry_SYSCALL_64_fastpath
15.71 ± 0% +20.0% 18.84 ± 0% perf-profile.calltrace.cycles-pp.inet_sendmsg.sock_sendmsg.SYSC_sendto.sys_sendto.entry_SYSCALL_64_fastpath
7.37 ± 0% -17.4% 6.08 ± 1% perf-profile.calltrace.cycles-pp.ip_local_out.ip_queue_xmit.tcp_transmit_skb.tcp_send_ack.__tcp_ack_snd_check
7.59 ± 0% -16.9% 6.31 ± 1% perf-profile.calltrace.cycles-pp.ip_queue_xmit.tcp_transmit_skb.tcp_send_ack.__tcp_ack_snd_check.tcp_rcv_established
9.50 ± 0% +13.6% 10.79 ± 1% perf-profile.calltrace.cycles-pp.ip_queue_xmit.tcp_transmit_skb.tcp_write_xmit.__tcp_push_pending_frames.tcp_push
2.53 ± 2% +10.4% 2.79 ± 1% perf-profile.calltrace.cycles-pp.release_sock.tcp_close.inet_release.sock_release.sock_close
0.82 ± 2% +117.7% 1.79 ± 2% perf-profile.calltrace.cycles-pp.release_sock.tcp_sendmsg.inet_sendmsg.sock_sendmsg.SYSC_sendto
1.24 ± 3% -100.0% 0.00 ± -1% perf-profile.calltrace.cycles-pp.schedule.schedule_timeout.__inet_stream_connect.inet_stream_connect.SYSC_connect
2.17 ± 1% -12.8% 1.89 ± 2% perf-profile.calltrace.cycles-pp.schedule.schedule_timeout.sk_wait_data.tcp_recvmsg.inet_recvmsg
1.28 ± 3% -100.0% 0.00 ± -1% perf-profile.calltrace.cycles-pp.schedule_timeout.__inet_stream_connect.inet_stream_connect.SYSC_connect.sys_connect
2.21 ± 1% -13.0% 1.93 ± 2% perf-profile.calltrace.cycles-pp.schedule_timeout.sk_wait_data.tcp_recvmsg.inet_recvmsg.sock_recvmsg
0.81 ± 25% +54.9% 1.25 ± 2% perf-profile.calltrace.cycles-pp.security_sock_rcv_skb.sk_filter_trim_cap.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
1.19 ± 23% +53.6% 1.82 ± 14% perf-profile.calltrace.cycles-pp.sk_filter_trim_cap.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv_finish
2.67 ± 0% -15.5% 2.26 ± 2% perf-profile.calltrace.cycles-pp.sk_wait_data.tcp_recvmsg.inet_recvmsg.sock_recvmsg.SYSC_recvfrom
15.30 ± 0% -18.1% 12.53 ± 1% perf-profile.calltrace.cycles-pp.sock_recvmsg.SYSC_recvfrom.sys_recvfrom.entry_SYSCALL_64_fastpath
15.96 ± 0% +19.6% 19.08 ± 0% perf-profile.calltrace.cycles-pp.sock_sendmsg.SYSC_sendto.sys_sendto.entry_SYSCALL_64_fastpath
15.77 ± 0% -17.4% 13.02 ± 1% perf-profile.calltrace.cycles-pp.sys_recvfrom.entry_SYSCALL_64_fastpath
16.29 ± 0% +19.2% 19.41 ± 0% perf-profile.calltrace.cycles-pp.sys_sendto.entry_SYSCALL_64_fastpath
1.11 ± 2% -100.0% 0.00 ± -1% perf-profile.calltrace.cycles-pp.tcp_ack.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
1.81 ± 2% +6.9% 1.93 ± 4% perf-profile.calltrace.cycles-pp.tcp_ack.tcp_rcv_state_process.tcp_v4_do_rcv.__release_sock.release_sock
1.86 ± 1% +34.3% 2.50 ± 1% perf-profile.calltrace.cycles-pp.tcp_ack.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
0.26 ±173% +211.4% 0.82 ± 37% perf-profile.calltrace.cycles-pp.tcp_clean_rtx_queue.tcp_ack.tcp_rcv_state_process.tcp_v4_do_rcv.__release_sock
0.55 ± 3% +106.3% 1.14 ± 1% perf-profile.calltrace.cycles-pp.tcp_clean_rtx_queue.tcp_ack.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv
1.12 ± 2% -19.6% 0.90 ± 6% perf-profile.calltrace.cycles-pp.tcp_finish_connect.tcp_rcv_state_process.tcp_v4_do_rcv.__release_sock.release_sock
2.58 ± 2% -26.3% 1.90 ± 4% perf-profile.calltrace.cycles-pp.tcp_prequeue.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv_finish
10.67 ± 0% -24.3% 8.08 ± 2% perf-profile.calltrace.cycles-pp.tcp_prequeue_process.tcp_recvmsg.inet_recvmsg.sock_recvmsg.SYSC_recvfrom
12.42 ± 0% +16.7% 14.49 ± 1% perf-profile.calltrace.cycles-pp.tcp_push.tcp_sendmsg.inet_sendmsg.sock_sendmsg.SYSC_sendto
0.57 ± 1% +141.7% 1.39 ± 3% perf-profile.calltrace.cycles-pp.tcp_rcv_established.tcp_v4_do_rcv.__release_sock.release_sock.tcp_sendmsg
8.66 ± 0% -11.5% 7.67 ± 1% perf-profile.calltrace.cycles-pp.tcp_rcv_established.tcp_v4_do_rcv.tcp_prequeue_process.tcp_recvmsg.inet_recvmsg
3.09 ± 1% -23.5% 2.36 ± 4% perf-profile.calltrace.cycles-pp.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
1.05 ± 2% +9.1% 1.14 ± 3% perf-profile.calltrace.cycles-pp.tcp_rcv_state_process.tcp_child_process.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
2.23 ± 3% +9.5% 2.44 ± 1% perf-profile.calltrace.cycles-pp.tcp_rcv_state_process.tcp_v4_do_rcv.__release_sock.release_sock.tcp_close
10.91 ± 1% +9.6% 11.96 ± 1% perf-profile.calltrace.cycles-pp.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
14.75 ± 0% -19.2% 11.91 ± 1% perf-profile.calltrace.cycles-pp.tcp_recvmsg.inet_recvmsg.sock_recvmsg.SYSC_recvfrom.sys_recvfrom
8.79 ± 0% -30.0% 6.16 ± 2% perf-profile.calltrace.cycles-pp.tcp_send_ack.__tcp_ack_snd_check.tcp_rcv_established.tcp_v4_do_rcv.tcp_prequeue_process
0.78 ± 1% +110.2% 1.65 ± 4% perf-profile.calltrace.cycles-pp.tcp_send_ack.__tcp_ack_snd_check.tcp_rcv_established.tcp_v4_do_rcv.tcp_v4_rcv
15.33 ± 0% +20.3% 18.45 ± 0% perf-profile.calltrace.cycles-pp.tcp_sendmsg.inet_sendmsg.sock_sendmsg.SYSC_sendto.sys_sendto
8.73 ± 0% -20.4% 6.95 ± 1% perf-profile.calltrace.cycles-pp.tcp_transmit_skb.tcp_send_ack.__tcp_ack_snd_check.tcp_rcv_established.tcp_v4_do_rcv
10.37 ± 0% +11.4% 11.56 ± 1% perf-profile.calltrace.cycles-pp.tcp_transmit_skb.tcp_write_xmit.__tcp_push_pending_frames.tcp_push.tcp_sendmsg
2.29 ± 3% +10.4% 2.53 ± 1% perf-profile.calltrace.cycles-pp.tcp_v4_do_rcv.__release_sock.release_sock.tcp_close.inet_release
0.61 ± 2% +140.6% 1.47 ± 3% perf-profile.calltrace.cycles-pp.tcp_v4_do_rcv.__release_sock.release_sock.tcp_sendmsg.inet_sendmsg
7.73 ± 0% -23.5% 5.91 ± 2% perf-profile.calltrace.cycles-pp.tcp_v4_do_rcv.tcp_prequeue_process.tcp_recvmsg.inet_recvmsg.sock_recvmsg
2.06 ± 1% +7.3% 2.21 ± 3% perf-profile.calltrace.cycles-pp.tcp_v4_syn_recv_sock.tcp_check_req.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
12.26 ± 0% +16.7% 14.31 ± 1% perf-profile.calltrace.cycles-pp.tcp_write_xmit.__tcp_push_pending_frames.tcp_push.tcp_sendmsg.inet_sendmsg
3.04 ± 1% -12.7% 2.65 ± 0% perf-profile.calltrace.cycles-pp.try_to_wake_up.default_wake_function.autoremove_wake_function.__wake_up_common.__wake_up_sync_key
2.36 ± 3% -14.1% 2.02 ± 2% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.default_wake_function.autoremove_wake_function.__wake_up_common
15.69 ± 0% -17.4% 12.96 ± 1% perf-profile.children.cycles-pp.SYSC_recvfrom
16.25 ± 0% +19.3% 19.38 ± 1% perf-profile.children.cycles-pp.SYSC_sendto
1.66 ± 4% +16.7% 1.93 ± 1% perf-profile.children.cycles-pp.__kfree_skb
15.12 ± 1% +10.6% 16.72 ± 0% perf-profile.children.cycles-pp.__release_sock
5.51 ± 1% -33.1% 3.69 ± 1% perf-profile.children.cycles-pp.__schedule
10.11 ± 0% -18.0% 8.29 ± 2% perf-profile.children.cycles-pp.__tcp_ack_snd_check
3.83 ± 1% -18.0% 3.14 ± 1% perf-profile.children.cycles-pp.__wake_up_common
3.63 ± 2% -12.2% 3.18 ± 1% perf-profile.children.cycles-pp.__wake_up_sync_key
1.86 ± 3% -11.7% 1.64 ± 1% perf-profile.children.cycles-pp.activate_task
3.66 ± 1% -20.5% 2.91 ± 1% perf-profile.children.cycles-pp.autoremove_wake_function
3.56 ± 1% -18.0% 2.92 ± 0% perf-profile.children.cycles-pp.default_wake_function
1.21 ± 5% -10.4% 1.08 ± 1% perf-profile.children.cycles-pp.enqueue_entity
1.64 ± 2% -10.8% 1.46 ± 2% perf-profile.children.cycles-pp.enqueue_task_fair
14.98 ± 0% -18.8% 12.17 ± 1% perf-profile.children.cycles-pp.inet_recvmsg
15.72 ± 0% +20.0% 18.87 ± 0% perf-profile.children.cycles-pp.inet_sendmsg
2.03 ± 2% -50.4% 1.01 ± 5% perf-profile.children.cycles-pp.pick_next_task_fair
15.94 ± 1% +10.2% 17.57 ± 1% perf-profile.children.cycles-pp.release_sock
5.86 ± 1% -33.6% 3.89 ± 0% perf-profile.children.cycles-pp.schedule
5.25 ± 1% -27.3% 3.81 ± 0% perf-profile.children.cycles-pp.schedule_timeout
2.68 ± 0% -15.5% 2.27 ± 2% perf-profile.children.cycles-pp.sk_wait_data
1.35 ± 0% +14.6% 1.55 ± 2% perf-profile.children.cycles-pp.skb_release_all
15.32 ± 0% -18.1% 12.54 ± 1% perf-profile.children.cycles-pp.sock_recvmsg
15.97 ± 0% +19.6% 19.10 ± 0% perf-profile.children.cycles-pp.sock_sendmsg
15.79 ± 0% -17.3% 13.05 ± 1% perf-profile.children.cycles-pp.sys_recvfrom
16.32 ± 0% +19.2% 19.46 ± 0% perf-profile.children.cycles-pp.sys_sendto
1.14 ± 2% -19.3% 0.92 ± 6% perf-profile.children.cycles-pp.tcp_finish_connect
2.75 ± 1% -24.1% 2.08 ± 4% perf-profile.children.cycles-pp.tcp_prequeue
10.69 ± 0% -24.3% 8.09 ± 2% perf-profile.children.cycles-pp.tcp_prequeue_process
12.50 ± 0% +16.6% 14.58 ± 1% perf-profile.children.cycles-pp.tcp_push
1.82 ± 3% +18.1% 2.15 ± 3% perf-profile.children.cycles-pp.tcp_rearm_rto
14.78 ± 0% -19.2% 11.95 ± 1% perf-profile.children.cycles-pp.tcp_recvmsg
15.36 ± 0% +20.3% 18.48 ± 0% perf-profile.children.cycles-pp.tcp_sendmsg
2.07 ± 1% +7.2% 2.22 ± 3% perf-profile.children.cycles-pp.tcp_v4_syn_recv_sock
3.43 ± 2% -17.5% 2.83 ± 0% perf-profile.children.cycles-pp.try_to_wake_up
2.46 ± 3% -11.6% 2.17 ± 1% perf-profile.children.cycles-pp.ttwu_do_activate
1.34 ± 3% +8.4% 1.45 ± 4% perf-profile.self.cycles-pp.__local_bh_enable_ip
0.81 ± 6% +18.8% 0.96 ± 3% perf-profile.self.cycles-pp.mod_timer
perf-stat.context-switches
6e+07 ++----------------------------------------------------------------+
| **.*.*.**.*.*.*.**.* **.*.* *.*.*.**.*.*.*.**.*.*.**.*.*
5.5e+07 ++ + + + + + |
*.* *.* ** |
| |
5e+07 ++ |
| |
4.5e+07 ++ |
| |
4e+07 ++ |
| |
| |
3.5e+07 ++ |
O O OO O O OO O O O OO O O O OO O O OO O O O |
3e+07 ++----------------------------------------------------------------+
perf-stat.cpu-migrations
11500 ++----O-------------------------------------------------------------+
| O OO |
11000 O+ O O O |
10500 ++O O O O O O O |
| O O O O |
10000 ++ O O O |
9500 ++ O O O |
| |
9000 ++ |
8500 ++ .* |
| .*. .*. .* .*. .*. .*. *.* +|
8000 ++ *.** *.* **.* * *.*.* : *.*.* *.**.* * * *
7500 ++ + + + + : : : |
*.* * *.* *.* |
7000 ++------------------------------------------------------------------+
netperf.Throughput_tps
3050 ++------------------------------O-OO---O-O-O-------------------------+
| O O O O O O OO O O O O O O |
3000 O+O O OO |
2950 ++ |
| .*. .*. .*.|
2900 ++ *.**.*. .*.*.*.*.**.* *.*.** *.*.*.** *.* *.*.**.* *
| : * : : : : |
2850 ++ : : : : : |
| : : : : : |
2800 ++ : : : : : |
2750 ++ : : : : : |
| : : : : : |
2700 ++: : : : : |
*.* *.* *.* |
2650 ++-------------------------------------------------------------------+
netperf.time.system_time
1800 ++-------------------------------------------------------------------+
| |
1750 O+O O OO O O O O O O OO O O O O O OO O O O O O |
| |
| |
1700 ++ |
| |
1650 ++ |
| |
1600 ++ |
| |
| |
1550 *+*.*.**.*.*.*.*.*.*.**.*.*.*.*.*.**.*.*.*.*.*.**.*.*.*.*.*.*.**.*.*.*
| |
1500 ++-------------------------------------------------------------------+
netperf.time.percent_of_cpu_this_job_got
590 O+O-O-O-OO-O-O-O-O-O-O-O-OO-O-O-O-O-O-O-O-O-OO------------------------+
| |
580 ++ |
570 ++ |
| |
560 ++ |
| |
550 ++ |
| |
540 ++ |
530 ++ |
| |
520 ++ .*.*. .*.*. |
*.*.*.*.** * *.*.**.*.*.*.*.*.*.*.*.**.*.*.*.*.*.*.*.**.*.*.*.*
510 ++--------------------------------------------------------------------+
netperf.time.voluntary_context_switches
1.5e+07 ++----------------------------------------------------------------+
O O OO O O OO O O O OO O O O OO O O OO O O O |
1.4e+07 ++ |
1.3e+07 ++ |
| |
1.2e+07 ++ |
1.1e+07 ++ |
| |
1e+07 ++ |
9e+06 ++ |
| |
8e+06 ++ |
7e+06 ++ .**.*.*.**.*.*.*.**.*. .**.*.*. .*.*.*.**.*.*.*.**.*.*.**.*.*
*.* *.* ** |
6e+06 ++----------------------------------------------------------------+
netperf.time.involuntary_context_switches
2.5e+07 ++----------------------------------------------------------------+
| |
| .**.*.*.**.*.*.*.**.*. .**.*.*. .*.*.*.**.*.*.*.**.*.*.**.*.*
2e+07 *+* *.* ** |
| |
| |
1.5e+07 ++ |
| |
1e+07 ++ |
| |
| |
5e+06 ++ |
| |
| |
0 O+O-OO-O-O-OO-O-O-O-OO-O-O-O-OO-O-O-OO-O-O-O----------------------+
[*] bisect-good sample
[O] bisect-bad sample
Thanks,
Xiaolong