Re: [tip: sched/core] sched: Fix performance regression introduced by mm_cid

From: Mathieu Desnoyers
Date: Wed Jun 21 2023 - 14:51:30 EST

On 6/21/23 12:36, Swapnil Sapkal wrote:
Hello Mathieu,

[...]

I suspect the regression is caused by the mm_count cache line bouncing.

Please try with this additional patch applied:

https://lore.kernel.org/lkml/20230515143536.114960-1-mathieu.desnoyers@xxxxxxxxxxxx/

Thanks for the suggestion. I tried out with the patch you suggested. I am seeing
improvement in hackbench numbers with mm_count padding. But this is not matching
with what we achieved through reverting the new mm_cid patch.

Below are the results on the 1 Socket 4th Generation EPYC Processor (1 x 96C/192T):

Threads:

Test:              Base (v6.4-rc1)   Base + new_mmcid_reverted  Base + mm_count_padding
 1-groups:         5.23 (0.00 pct)         4.61 (11.85 pct)        5.11 (2.29 pct)
 2-groups:         4.99 (0.00 pct)         4.72 (5.41 pct)         5.00 (-0.20 pct)
 4-groups:         5.96 (0.00 pct)         4.87 (18.28 pct)        5.86 (1.67 pct)
 8-groups:         6.58 (0.00 pct)         5.44 (17.32 pct)        6.20 (5.77 pct)
16-groups:        11.48 (0.00 pct)         8.07 (29.70 pct)       10.68 (6.96 pct)

Processes:

Test:              Base (v6.4-rc1)  Base + new_mmcid_reverted   Base + mm_count_padding
 1-groups:         5.19 (0.00 pct)         4.90 (5.58 pct)         5.19 (0.00 pct)
 2-groups:         5.44 (0.00 pct)         5.39 (0.91 pct)         5.39 (0.91 pct)
 4-groups:         5.69 (0.00 pct)         5.64 (0.87 pct)         5.64 (0.87 pct)
 8-groups:         6.08 (0.00 pct)         6.01 (1.15 pct)         6.04 (0.65 pct)
16-groups:        10.87 (0.00 pct)        10.83 (0.36 pct)        10.93 (-0.55 pct)

The ibs profile shows that function __switch_to_asm() is coming at top in baseline
run and is not seen with mm_count padding patch. Will be attaching full ibs profile
data for all the 3 runs:

# Base (v6.4-rc1)
Threads:
Total time: 11.486 [sec]

   5.15%  sched-messaging  [kernel.vmlinux]      [k] __switch_to_asm
   4.31%  sched-messaging  [kernel.vmlinux]      [k] copyout
   4.29%  sched-messaging  [kernel.vmlinux]      [k] native_queued_spin_lock_slowpath
   4.22%  sched-messaging  [kernel.vmlinux]      [k] copyin
   3.92%  sched-messaging  [kernel.vmlinux]      [k] apparmor_file_permission
   2.91%  sched-messaging  [kernel.vmlinux]      [k] __schedule
   2.34%  swapper          [kernel.vmlinux]      [k] __switch_to_asm
   2.10%  sched-messaging  [kernel.vmlinux]      [k] prepare_to_wait_event
   2.10%  sched-messaging  [kernel.vmlinux]      [k] try_to_wake_up
   2.07%  sched-messaging  [kernel.vmlinux]      [k] finish_task_switch.isra.0
   2.00%  sched-messaging  [kernel.vmlinux]      [k] pipe_write
   1.82%  sched-messaging  [kernel.vmlinux]      [k] check_preemption_disabled
   1.73%  sched-messaging  [kernel.vmlinux]      [k] exit_to_user_mode_prepare
   1.52%  sched-messaging  [kernel.vmlinux]      [k] __entry_text_start
   1.49%  sched-messaging  [kernel.vmlinux]      [k] osq_lock
   1.45%  sched-messaging  libc.so.6             [.] write
   1.44%  swapper          [kernel.vmlinux]      [k] native_sched_clock
   1.38%  sched-messaging  [kernel.vmlinux]      [k] psi_group_change
   1.38%  sched-messaging  [kernel.vmlinux]      [k] pipe_read
   1.37%  sched-messaging  libc.so.6             [.] read
   1.06%  sched-messaging  [kernel.vmlinux]      [k] vfs_read
   1.01%  swapper          [kernel.vmlinux]      [k] psi_group_change
   1.00%  sched-messaging  [kernel.vmlinux]      [k] update_curr

# Base + mm_count_padding
Threads:
Total time: 11.384 [sec]

   4.43%  sched-messaging  [kernel.vmlinux]         [k] copyin
   4.39%  sched-messaging  [kernel.vmlinux]         [k] native_queued_spin_lock_slowpath
   4.07%  sched-messaging  [kernel.vmlinux]         [k] apparmor_file_permission
   4.07%  sched-messaging  [kernel.vmlinux]         [k] copyout
   2.49%  sched-messaging  [kernel.vmlinux]         [k] entry_SYSCALL_64
   2.37%  sched-messaging  [kernel.vmlinux]         [k] update_cfs_group
   2.19%  sched-messaging  [kernel.vmlinux]         [k] pipe_write
   2.00%  sched-messaging  [kernel.vmlinux]         [k] check_preemption_disabled
   1.93%  swapper          [kernel.vmlinux]         [k] update_load_avg
   1.81%  sched-messaging  [kernel.vmlinux]         [k] exit_to_user_mode_prepare
   1.69%  sched-messaging  [kernel.vmlinux]         [k] try_to_wake_up
   1.58%  sched-messaging  libc.so.6                [.] write
   1.53%  sched-messaging  [kernel.vmlinux]         [k] psi_group_change
   1.50%  sched-messaging  libc.so.6                [.] read
   1.50%  sched-messaging  [kernel.vmlinux]         [k] pipe_read
   1.39%  sched-messaging  [kernel.vmlinux]         [k] update_load_avg
   1.39%  sched-messaging  [kernel.vmlinux]         [k] osq_lock
   1.30%  sched-messaging  [kernel.vmlinux]         [k] update_curr
   1.28%  swapper          [kernel.vmlinux]         [k] psi_group_change
   1.16%  sched-messaging  [kernel.vmlinux]         [k] vfs_read
   1.12%  sched-messaging  [kernel.vmlinux]         [k] vfs_write
   1.10%  sched-messaging  [kernel.vmlinux]         [k] entry_SYSRETQ_unsafe_stack
   1.09%  sched-messaging  [kernel.vmlinux]         [k] __switch_to_asm
   1.08%  sched-messaging  [kernel.vmlinux]         [k] do_syscall_64
   1.06%  sched-messaging  [kernel.vmlinux]         [k] select_task_rq_fair
   1.03%  swapper          [kernel.vmlinux]         [k] update_cfs_group
   1.00%  swapper          [kernel.vmlinux]         [k] rb_insert_color

# Base + reverted_new_mm_cid
Threads:
Total time: 7.847 [sec]

  12.14%  sched-messaging  [kernel.vmlinux]      [k] native_queued_spin_lock_slowpath
   8.86%  swapper          [kernel.vmlinux]      [k] native_queued_spin_lock_slowpath
   6.13%  sched-messaging  [kernel.vmlinux]      [k] copyin
   5.54%  sched-messaging  [kernel.vmlinux]      [k] apparmor_file_permission
   3.59%  sched-messaging  [kernel.vmlinux]      [k] copyout
   2.61%  sched-messaging  [kernel.vmlinux]      [k] osq_lock
   2.48%  sched-messaging  [kernel.vmlinux]      [k] pipe_write
   2.33%  sched-messaging  [kernel.vmlinux]      [k] exit_to_user_mode_prepare
   2.01%  sched-messaging  [kernel.vmlinux]      [k] check_preemption_disabled
   1.96%  sched-messaging  [kernel.vmlinux]      [k] __entry_text_start
   1.91%  sched-messaging  libc.so.6             [.] write
   1.77%  sched-messaging  libc.so.6             [.] read
   1.64%  sched-messaging  [kernel.vmlinux]      [k] mutex_spin_on_owner
   1.58%  sched-messaging  [kernel.vmlinux]      [k] pipe_read
   1.52%  sched-messaging  [kernel.vmlinux]      [k] try_to_wake_up
   1.38%  sched-messaging  [kernel.vmlinux]      [k] ktime_get_coarse_real_ts64
   1.35%  sched-messaging  [kernel.vmlinux]      [k] vfs_write
   1.28%  sched-messaging  [kernel.vmlinux]      [k] entry_SYSRETQ_unsafe_stack
   1.28%  sched-messaging  [kernel.vmlinux]      [k] vfs_read
   1.25%  sched-messaging  [kernel.vmlinux]      [k] do_syscall_64
   1.22%  sched-messaging  [kernel.vmlinux]      [k] __fget_light
   1.18%  sched-messaging  [kernel.vmlinux]      [k] mutex_lock
   1.12%  sched-messaging  [kernel.vmlinux]      [k] file_update_time
   1.04%  sched-messaging  [kernel.vmlinux]      [k] _copy_from_iter
   1.01%  sched-messaging  [kernel.vmlinux]      [k] current_time

So with the reverted new_mm_cid patch, we are seeing a lot of time being spent in
native_queued_spin_lock_slowpath and yet, hackbench finishes faster.

I keep further digging into this please let me know if you have any pointers for me.

Do you have CONFIG_SECURITY_APPARMOR=y ? Can you try without ?

I notice that apparmor_file_permission appears near the top of your
profiles, and apparmor uses an internal aa_buffers_lock spinlock,
which could possibly explain the top hits for
native_queued_spin_lock_slowpath. My current suspicion is that
the raw spinlock that was taken by "Base + reverted_new_mm_cid"
changed the contention pattern on the apparmor lock enough to
speed things up by pure accident.

Thanks,

Mathieu




This patch has recently been merged into the mm tree.

Thanks,

Mathieu

--
Thanks and Regards,
Swapnil

--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com