[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <68a1f51b-50bf-0770-2367-c3e1b38be535@oracle.com>
Date: Fri, 19 Jun 2020 14:56:58 -0700
From: Junxiao Bi <junxiao.bi@...cle.com>
To: "Eric W. Biederman" <ebiederm@...ssion.com>
Cc: Matthew Wilcox <willy@...radead.org>, linux-kernel@...r.kernel.org,
linux-fsdevel@...r.kernel.org,
Matthew Wilcox <matthew.wilcox@...cle.com>,
Srinivas Eeda <SRINIVAS.EEDA@...cle.com>,
"joe.jin@...cle.com" <joe.jin@...cle.com>,
Wengang Wang <wen.gang.wang@...cle.com>
Subject: Re: [PATCH] proc: Avoid a thundering herd of threads freeing proc
dentries
On 6/19/20 10:24 AM, ebiederm@...ssion.com wrote:
> Junxiao Bi <junxiao.bi@...cle.com> writes:
>
>> Hi Eric,
>>
>> The patch didn't improve lock contention.
> Which raises the question where is the lock contention coming from.
>
> Especially with my first variant. Only the last thread to be reaped
> would free up anything in the cache.
>
> Can you comment out the call to proc_flush_pid entirely?
Still high lock contention. Collect the following hot path.
74.90% 0.01% proc_race
[kernel.kallsyms] [k]
entry_SYSCALL_64_after_hwframe
|
--74.89%--entry_SYSCALL_64_after_hwframe
|
--74.88%--do_syscall_64
|
|--69.70%--exit_to_usermode_loop
| |
| --69.70%--do_signal
| |
| --69.69%--get_signal
| |
| |--56.30%--do_group_exit
| | |
| | --56.30%--do_exit
| | |
| |
|--27.50%--_raw_write_lock_irq
| | | |
| | |
--27.47%--queued_write_lock_slowpath
| |
| |
| | |
--27.18%--native_queued_spin_lock_slowpath
| | |
| |
|--26.10%--release_task.part.20
| | | |
| | |
--25.60%--_raw_write_lock_irq
| |
| |
| | |
--25.56%--queued_write_lock_slowpath
| |
| |
| | |
--25.23%--native_queued_spin_lock_slowpath
| | |
| | --0.56%--mmput
| | |
| |
--0.55%--exit_mmap
| |
| --13.31%--_raw_spin_lock_irq
| |
| --13.28%--native_queued_spin_lock_slowpath
|
Thanks,
Junxiao.
>
> That will rule out the proc_flush_pid in d_invalidate entirely.
>
> The only candidate I can think of d_invalidate aka (proc_flush_pid) vs ps.
>
> Eric
Powered by blists - more mailing lists