From: Shrikanth Hegde <sshegde@linux.ibm.com>
To: Ankur Arora <ankur.a.arora@oracle.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>,
tglx@linutronix.de, peterz@infradead.org,
torvalds@linux-foundation.org, paulmck@kernel.org,
rostedt@goodmis.org, mark.rutland@arm.com, juri.lelli@redhat.com,
joel@joelfernandes.org, raghavendra.kt@amd.com,
boris.ostrovsky@oracle.com, konrad.wilk@oracle.com,
LKML <linux-kernel@vger.kernel.org>,
Nicholas Piggin <npiggin@gmail.com>
Subject: Re: [PATCH v2 00/35] PREEMPT_AUTO: support lazy rescheduling
Date: Mon, 12 Aug 2024 23:02:19 +0530 [thread overview]
Message-ID: <1a973dda-c79e-4d95-935b-e4b93eb077b8@linux.ibm.com> (raw)
In-Reply-To: <87sewr6o5w.fsf@oracle.com>
On 7/3/24 10:57, Ankur Arora wrote:
>
> Shrikanth Hegde <sshegde@linux.ibm.com> writes:
>
Hi.
Sorry for the delayed response.
I could see this hackbench pipe regression with preempt=full kernel on 6.10-rc kernel. i.e without PREEMPT_AUTO as well.
There seems to more wakeups in read path, implies pipe was more often empty. Correspondingly more contention
is there on the mutex pipe lock in preempt=full. But why, not sure. One difference in powerpc is page size. but
here pipe isn't getting full. Its not the write side that is blocked.
preempt=none: Time taken for 20 groups in seconds : 25.70
preempt=full: Time taken for 20 groups in seconds : 54.56
----------------
hackbench (pipe)
----------------
top 3 callstacks of __schedule collected with bpftrace.
preempt=none preempt=full
__schedule+12 |@[
schedule+64 | __schedule+12
interrupt_exit_user_prepare_main+600 | preempt_schedule+84
interrupt_exit_user_prepare+88 | _raw_spin_unlock_irqrestore+124
interrupt_return_srr_user+8 | __wake_up_sync_key+108
, hackbench]: 482228 | pipe_write+1772
@[ | vfs_write+1052
__schedule+12 | ksys_write+248
schedule+64 | system_call_exception+296
pipe_write+1452 | system_call_vectored_common+348
vfs_write+940 |, hackbench]: 538591
ksys_write+248 |@[
system_call_exception+292 | __schedule+12
system_call_vectored_common+348 | schedule+76
, hackbench]: 1427161 | schedule_preempt_disabled+52
@[ | __mutex_lock.constprop.0+1748
__schedule+12 | pipe_write+132
schedule+64 | vfs_write+1052
interrupt_exit_user_prepare_main+600 | ksys_write+248
syscall_exit_prepare+336 | system_call_exception+296
system_call_vectored_common+360 | system_call_vectored_common+348
, hackbench]: 8151309 |, hackbench]: 5388301
@[ |@[
__schedule+12 | __schedule+12
schedule+64 | schedule+76
pipe_read+1100 | pipe_read+1100
vfs_read+716 | vfs_read+716
ksys_read+252 | ksys_read+252
system_call_exception+292 | system_call_exception+296
system_call_vectored_common+348 | system_call_vectored_common+348
, hackbench]: 18132753 |, hackbench]: 64424110
--------------------------------------------
hackbench (messaging) - one that uses sockets
--------------------------------------------
Here there is no regression with preempt=full.
preempt=none: Time taken for 20 groups in seconds : 55.51
preempt=full: Time taken for 20 groups in seconds : 55.10
Similar bpftrace collected for socket based hackbench. highest caller of __schedule doesn't change much.
preempt=none preempt=full
| __schedule+12
| preempt_schedule+84
| _raw_spin_unlock+108
@[ | unix_stream_sendmsg+660
__schedule+12 | sock_write_iter+372
schedule+64 | vfs_write+1052
schedule_timeout+412 | ksys_write+248
sock_alloc_send_pskb+684 | system_call_exception+296
unix_stream_sendmsg+448 | system_call_vectored_common+348
sock_write_iter+372 |, hackbench]: 819290
vfs_write+940 |@[
ksys_write+248 | __schedule+12
system_call_exception+292 | schedule+76
system_call_vectored_common+348 | schedule_timeout+476
, hackbench]: 3424197 | sock_alloc_send_pskb+684
@[ | unix_stream_sendmsg+444
__schedule+12 | sock_write_iter+372
schedule+64 | vfs_write+1052
interrupt_exit_user_prepare_main+600 | ksys_write+248
syscall_exit_prepare+336 | system_call_exception+296
system_call_vectored_common+360 | system_call_vectored_common+348
, hackbench]: 9800144 |, hackbench]: 3386594
@[ |@[
__schedule+12 | __schedule+12
schedule+64 | schedule+76
schedule_timeout+412 | schedule_timeout+476
unix_stream_data_wait+528 | unix_stream_data_wait+468
unix_stream_read_generic+872 | unix_stream_read_generic+804
unix_stream_recvmsg+196 | unix_stream_recvmsg+196
sock_recvmsg+164 | sock_recvmsg+156
sock_read_iter+200 | sock_read_iter+200
vfs_read+716 | vfs_read+716
ksys_read+252 | ksys_read+252
system_call_exception+292 | system_call_exception+296
system_call_vectored_common+348 | system_call_vectored_common+348
, hackbench]: 25375142 |, hackbench]: 27275685
next prev parent reply other threads:[~2024-08-12 17:33 UTC|newest]
Thread overview: 95+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-28 0:34 [PATCH v2 00/35] PREEMPT_AUTO: support lazy rescheduling Ankur Arora
2024-05-28 0:34 ` [PATCH v2 01/35] sched/core: Move preempt_model_*() helpers from sched.h to preempt.h Ankur Arora
2024-06-06 17:45 ` [tip: sched/core] " tip-bot2 for Sean Christopherson
2024-05-28 0:34 ` [PATCH v2 02/35] sched/core: Drop spinlocks on contention iff kernel is preemptible Ankur Arora
2024-05-28 0:34 ` [PATCH v2 03/35] sched: make test_*_tsk_thread_flag() return bool Ankur Arora
2024-05-28 0:34 ` [PATCH v2 04/35] preempt: introduce CONFIG_PREEMPT_AUTO Ankur Arora
2024-06-03 15:04 ` Shrikanth Hegde
2024-06-04 17:52 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 05/35] thread_info: selector for TIF_NEED_RESCHED[_LAZY] Ankur Arora
2024-05-28 15:55 ` Peter Zijlstra
2024-05-30 9:07 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 06/35] thread_info: define __tif_need_resched(resched_t) Ankur Arora
2024-05-28 16:03 ` Peter Zijlstra
2024-05-28 0:34 ` [PATCH v2 07/35] sched: define *_tsk_need_resched_lazy() helpers Ankur Arora
2024-05-28 16:09 ` Peter Zijlstra
2024-05-30 9:02 ` Ankur Arora
2024-05-29 8:25 ` Peter Zijlstra
2024-05-30 9:08 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 08/35] entry: handle lazy rescheduling at user-exit Ankur Arora
2024-05-28 16:12 ` Peter Zijlstra
2024-05-28 0:34 ` [PATCH v2 09/35] entry/kvm: handle lazy rescheduling at guest-entry Ankur Arora
2024-05-28 16:13 ` Peter Zijlstra
2024-05-30 9:04 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 10/35] entry: irqentry_exit only preempts for TIF_NEED_RESCHED Ankur Arora
2024-05-28 16:18 ` Peter Zijlstra
2024-05-30 9:03 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 11/35] sched: __schedule_loop() doesn't need to check for need_resched_lazy() Ankur Arora
2024-05-28 0:34 ` [PATCH v2 12/35] sched: separate PREEMPT_DYNAMIC config logic Ankur Arora
2024-05-28 16:25 ` Peter Zijlstra
2024-05-30 9:30 ` Ankur Arora
2024-05-28 0:34 ` [PATCH v2 13/35] sched: allow runtime config for PREEMPT_AUTO Ankur Arora
2024-05-28 16:27 ` Peter Zijlstra
2024-05-30 9:29 ` Ankur Arora
2024-06-06 11:51 ` Peter Zijlstra
2024-06-06 15:11 ` Ankur Arora
2024-06-06 17:32 ` Peter Zijlstra
2024-06-09 0:46 ` Ankur Arora
2024-06-12 18:10 ` Paul E. McKenney
2024-05-28 0:35 ` [PATCH v2 14/35] rcu: limit PREEMPT_RCU to full preemption under PREEMPT_AUTO Ankur Arora
2024-05-28 0:35 ` [PATCH v2 15/35] rcu: fix header guard for rcu_all_qs() Ankur Arora
2024-05-28 0:35 ` [PATCH v2 16/35] preempt,rcu: warn on PREEMPT_RCU=n, preempt=full Ankur Arora
2024-05-29 8:14 ` Peter Zijlstra
2024-05-30 18:32 ` Paul E. McKenney
2024-05-30 23:05 ` Ankur Arora
2024-05-30 23:15 ` Paul E. McKenney
2024-05-30 23:04 ` Ankur Arora
2024-05-30 23:20 ` Paul E. McKenney
2024-06-06 11:53 ` Peter Zijlstra
2024-06-06 13:38 ` Paul E. McKenney
2024-06-17 15:54 ` Paul E. McKenney
2024-06-18 16:29 ` Paul E. McKenney
2024-05-28 0:35 ` [PATCH v2 17/35] rcu: handle quiescent states for PREEMPT_RCU=n, PREEMPT_COUNT=y Ankur Arora
2024-05-28 0:35 ` [PATCH v2 18/35] rcu: force context-switch " Ankur Arora
2024-05-28 0:35 ` [PATCH v2 19/35] x86/thread_info: define TIF_NEED_RESCHED_LAZY Ankur Arora
2024-05-28 0:35 ` [PATCH v2 20/35] powerpc: add support for PREEMPT_AUTO Ankur Arora
2024-05-28 0:35 ` [PATCH v2 21/35] sched: prepare for lazy rescheduling in resched_curr() Ankur Arora
2024-05-29 9:32 ` Peter Zijlstra
2024-05-28 0:35 ` [PATCH v2 22/35] sched: default preemption policy for PREEMPT_AUTO Ankur Arora
2024-05-28 0:35 ` [PATCH v2 23/35] sched: handle idle preemption " Ankur Arora
2024-05-28 0:35 ` [PATCH v2 24/35] sched: schedule eagerly in resched_cpu() Ankur Arora
2024-05-28 0:35 ` [PATCH v2 25/35] sched/fair: refactor update_curr(), entity_tick() Ankur Arora
2024-05-28 0:35 ` [PATCH v2 26/35] sched/fair: handle tick expiry under lazy preemption Ankur Arora
2024-05-28 0:35 ` [PATCH v2 27/35] sched: support preempt=none under PREEMPT_AUTO Ankur Arora
2024-05-28 0:35 ` [PATCH v2 28/35] sched: support preempt=full " Ankur Arora
2024-05-28 0:35 ` [PATCH v2 29/35] sched: handle preempt=voluntary " Ankur Arora
2024-06-17 3:20 ` Tianchen Ding
2024-06-21 18:58 ` Ankur Arora
2024-06-24 2:35 ` Tianchen Ding
2024-06-25 1:12 ` Ankur Arora
2024-06-26 2:43 ` Tianchen Ding
2024-05-28 0:35 ` [PATCH v2 30/35] sched: latency warn for TIF_NEED_RESCHED_LAZY Ankur Arora
2024-05-28 0:35 ` [PATCH v2 31/35] tracing: support lazy resched Ankur Arora
2024-05-28 0:35 ` [PATCH v2 32/35] Documentation: tracing: add TIF_NEED_RESCHED_LAZY Ankur Arora
2024-05-28 0:35 ` [PATCH v2 33/35] osnoise: handle quiescent states for PREEMPT_RCU=n, PREEMPTION=y Ankur Arora
2024-05-28 13:12 ` Daniel Bristot de Oliveira
2024-05-28 0:35 ` [PATCH v2 34/35] kconfig: decompose ARCH_NO_PREEMPT Ankur Arora
2024-05-28 0:35 ` [PATCH v2 35/35] arch: " Ankur Arora
2024-05-29 6:16 ` [PATCH v2 00/35] PREEMPT_AUTO: support lazy rescheduling Shrikanth Hegde
2024-06-01 11:47 ` Ankur Arora
2024-06-04 7:32 ` Shrikanth Hegde
2024-06-07 16:48 ` Shrikanth Hegde
2024-06-10 7:23 ` Ankur Arora
2024-06-15 15:04 ` Shrikanth Hegde
2024-06-18 18:27 ` Shrikanth Hegde
2024-06-19 2:40 ` Ankur Arora
2024-06-24 18:37 ` Shrikanth Hegde
2024-06-27 2:50 ` Ankur Arora
2024-06-27 5:56 ` Michael Ellerman
2024-06-27 15:44 ` Shrikanth Hegde
2024-07-03 5:27 ` Ankur Arora
2024-08-12 17:32 ` Shrikanth Hegde [this message]
2024-08-12 21:07 ` Linus Torvalds
2024-08-13 5:40 ` Ankur Arora
2024-06-05 15:44 ` Sean Christopherson
2024-06-05 17:45 ` Peter Zijlstra
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1a973dda-c79e-4d95-935b-e4b93eb077b8@linux.ibm.com \
--to=sshegde@linux.ibm.com \
--cc=ankur.a.arora@oracle.com \
--cc=boris.ostrovsky@oracle.com \
--cc=joel@joelfernandes.org \
--cc=juri.lelli@redhat.com \
--cc=konrad.wilk@oracle.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mpe@ellerman.id.au \
--cc=npiggin@gmail.com \
--cc=paulmck@kernel.org \
--cc=peterz@infradead.org \
--cc=raghavendra.kt@amd.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox