linux-kernel - Re: [PATCH 2/2] kernel: rerun task_work while freezing in get

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <933e7957-7a73-4c9a-87a7-c85b702a3a32@gmail.com>
Date: Wed, 10 Jul 2024 18:53:03 +0100
From: Pavel Begunkov <asml.silence@...il.com>
To: Tejun Heo <tj@...nel.org>
Cc: Oleg Nesterov <oleg@...hat.com>, io-uring@...r.kernel.org,
 Jens Axboe <axboe@...nel.dk>, Andrew Morton <akpm@...ux-foundation.org>,
 Christian Brauner <brauner@...nel.org>,
 Tycho Andersen <tandersen@...flix.com>, Thomas Gleixner
 <tglx@...utronix.de>, linux-kernel@...r.kernel.org,
 Julian Orth <ju.orth@...il.com>, Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH 2/2] kernel: rerun task_work while freezing in
 get_signal()

On 7/10/24 01:54, Tejun Heo wrote:
> Hello,
> 
> On Tue, Jul 09, 2024 at 08:55:43PM +0100, Pavel Begunkov wrote:
> ...
>>>> CRIU, I assume. I'll try it ...
>>>
>>> Than I think we can forget about task_works and this patch. CRIU dumps
>>> the tasks in TASK_TRACED state.
>>
>> And would be hard to test, io_uring (the main source of task_work)
>> is not supported
>>
>> (00.466022) Error (criu/proc_parse.c:477): Unknown shit 600 (anon_inode:[io_uring])
>> ...
>> (00.467642) Unfreezing tasks into 1
>> (00.467656)     Unseizing 15488 into 1
>> (00.468149) Error (criu/cr-dump.c:2111): Dumping FAILED.
> 
> Yeah, the question is: If CRIU is to use cgroup freezer to freeze the tasks
> and then go around tracing each to make dump, would the freezer be enough in
> avoiding interim state changes? Using CRIU implementation is a bit arbitrary
> but I think checkpoint-restart is a useful bar to measure what should stay
> stable while a cgroup is frozen.

Sounds like in the long run we might want to ignore task_work while
it's frozen, but hard to say for all task_work users.

-- 
Pavel Begunkov