linux-kernel - Re: [PATCH 3/3] fork, vhost: Use CLONE_THREAD to fix freezer/ps regression

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: <022f4de6-9eae-0a94-0f55-b84be4982fc3@oracle.com>
Date:   Mon, 29 May 2023 11:09:27 -0500
From:   michael.christie@...cle.com
To:     Oleg Nesterov <oleg@...hat.com>,
        "Eric W. Biederman" <ebiederm@...ssion.com>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        linux@...mhuis.info, nicolas.dichtel@...nd.com, axboe@...nel.dk,
        linux-kernel@...r.kernel.org,
        virtualization@...ts.linux-foundation.org, mst@...hat.com,
        sgarzare@...hat.com, jasowang@...hat.com, stefanha@...hat.com,
        brauner@...nel.org
Subject: Re: [PATCH 3/3] fork, vhost: Use CLONE_THREAD to fix freezer/ps
 regression

On 5/29/23 6:19 AM, Oleg Nesterov wrote:
> On 05/27, Eric W. Biederman wrote:
>>
>> Looking forward I don't see not asking the worker threads to stop
>> for the coredump right now causing any problems in the future.
>> So I think we can use this to resolve the coredump issue I spotted.
> 
> But we have almost the same problem with exec.
> 
> Execing thread will wait for vhost_worker() while vhost_worker will wait for
> .release -> vhost_task_stop().

For this type of case, what is the goal or correct behavior in the end?

When get_signal returns true we can code things like you mention below and
clean up the task_struct. However, we now have a non-functioning vhost device
open and just sitting around taking up memory and it can't do any IO.

For this type of case, do we expect just not to crash/hang, or was this new
exec'd thread suppose to be able to use the vhost device?

I would normally say it probably wants to use the vhost device still. However,
I don't think this comes up so just not hanging might be ok. Before 6.4-rc1,
we ignored signals so it would have worked if we are concerned about a possible
regression if this was a common thing.

> 
> And even O_CLOEXEC won't help, do_close_on_exec() is called after de_thread().
> 
> Or suppose that vhost_worker's sub-thread forks a child with CLONE_FILES...

You mean the vhost_task's task/thread doing a function that does a copy_process
right? That type of thing is not needed. I can add a check in vhost_task_create
for this so new code doesn't try to do it. I don't think it will come up that some
code vhost is using will call kernel_thread/copy_process directly since those
calls are so rare and the functions are not exported to modules.

> 
> If we want CLONE_THREAD, I think vhost_worker() should exit after get_signal()
> returns SIGKILL. Perhaps it should "disable" vhost_work_queue() somehow and
> flush the pending works on ->work_list before exit, I dunno. But imo it should
> not wait for the final fput().
> 
> Oleg.
>