Message-ID: <f4fcb703-ba1d-4614-a411-eefbf04ddf45@kylinos.cn>
Date: Fri, 4 Jul 2025 09:45:16 +0800
From: Zihuan Zhang <zhangzihuan@...inos.cn>
To: "Rafael J. Wysocki" <rafael@...nel.org>,
 Peter Zijlstra <peterz@...radead.org>
Cc: pavel@...nel.org, len.brown@...el.com, linux-pm@...r.kernel.org,
 linux-kernel@...r.kernel.org, Oleg Nesterov <oleg@...hat.com>
Subject: Re: [PATCH v3 1/1] PM / Freezer: Skip zombie/dead processes to

Hi Rafael,

在 2025/7/4 01:15, Rafael J. Wysocki 写道:
> On Thu, Jul 3, 2025 at 6:40 PM Peter Zijlstra <peterz@...radead.org> wrote:
>> On Thu, Jul 03, 2025 at 04:15:10PM +0200, Rafael J. Wysocki wrote:
>>> The patch subject appears to be incomplete.
You’re right — the patch subject was indeed incomplete. That was an 
oversight on my part, and I’ll fix it in the next version.

Thanks for pointing it out.
>>> On Wed, Jun 11, 2025 at 12:13 PM Zihuan Zhang <zhangzihuan@...inos.cn> wrote:
>>>> When freezing user space during suspend or hibernation, the freezer
>>>> iterates over all tasks and attempts to freeze them via
>>>> try_to_freeze_tasks().
>>>>
>>>> However, zombie processes (i.e., tasks in EXIT_ZOMBIE state) are no
>>>> longer running and will never enter the refrigerator. Trying to freeze
>>>> them is meaningless and causes extra overhead, especially when there are
>>>> thousands of zombies created during stress conditions such as fork
>>>> storms.
>>>>
>>>> This patch skips zombie processes during the freezing phase.
>>>>
>>>> In our testing with ~30,000 user processes (including many zombies), the
>>>> average freeze time during suspend (S3) dropped from ~43 ms to ~16 ms:
>>>>
>>>>      - Without the patch: ~43 ms average freeze latency
>>>>      - With the patch:    ~16 ms average freeze latency
>>>>      - Improvement:       ~62%
>>> And what's the total suspend time on the system in question?
>>>
We used the sleepgraph tool to measure the full suspend-to-RAM (S3) 
latency on our test platform. The total suspend time was around 1859.055 
ms, so the optimization to skip zombie processes — reducing freeze time 
from ~43 ms to ~16 ms — accounts for roughly 1% of the total suspend 
latency.

While the absolute gain is relatively small, it helps reduce unnecessary 
overhead under stress conditions such as fork storms with many zombie 
tasks, and improves freeze-time predictability.

>>>> This confirms that skipping zombies significantly speeds up the freezing
>>>> process when the system is under heavy load with many short-lived tasks.
>>>>
>>>> Signed-off-by: Zihuan Zhang <zhangzihuan@...inos.cn>
>>>>
>>>> Changes in v3:
>>>> - Added performance test
>>>>
>>>> Changes in v2:
>>>> - Simplified code, added judgment of dead processes
>>>> - Rewrite changelog
>>>> ---
>>>>   kernel/power/process.c | 2 +-
>>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/kernel/power/process.c b/kernel/power/process.c
>>>> index a6f7ba2d283d..2bbe22610522 100644
>>>> --- a/kernel/power/process.c
>>>> +++ b/kernel/power/process.c
>>>> @@ -51,7 +51,7 @@ static int try_to_freeze_tasks(bool user_only)
>>>>                  todo = 0;
>>>>                  read_lock(&tasklist_lock);
>>>>                  for_each_process_thread(g, p) {
>>>> -                       if (p == current || !freeze_task(p))
>>>> +                       if (p == current || p->exit_state || !freeze_task(p))
>>>>                                  continue;
>>>>
>>>>                          todo++;
>>>> --
>>> This is basically fine by me, but I wonder what other people think.
>>>
>>> Peter?
>> How realistic is it to have a significant amount of zombies when
>> freezing? This seems like an artificial corner case at best.
>>
>> Zombie tasks are stuck waiting on their parent to consume their exit
>> state or something, right? And those parents being frozen, they pretty
>> much stay there.
>>
>> So I suppose the logic holds, but urgh, do we really need this?
> Unlikely in practice, but the code change is small and it would be
> prudent to get this addressed IMV (at least so we don't need to
> revisit it).
>
> But I would ask for a comment above this check to explain that zombies
> need not be frozen.
Thanks for the suggestion.

Yes, I’ll add a comment to clarify the rationale. Planning to add the 
following just above the check:
> /*
>   * Zombie and dead tasks are not running anymore and cannot enter
>   * the __refrigerator(). Skipping them avoids unnecessary freeze attempts.
>   */

I’ll include this in the next version of the patch.

Best regards,
Zihuan Zhang

Powered by blists - more mailing lists