Re: [PATCH 1/1] exit: change the release_task() paths to call flush_sigqueue() lockless

[Thomas Gleixner]

Oleg!

On Thu, Feb 06 2025 at 17:23, Oleg Nesterov wrote:
> On 02/06, Thomas Gleixner wrote:
>> > 	- do_sigaction() can hit the similar problem
>>
>> Indeed, but that's a tough on to solve.
>
> Yeah... Although I have to admit that yesterday I had a very simple
> (and wrong) solution in mind ;)

I can relate. My initial "brilliant" idea to solve that turned out to be
broken after less than 10 seconds... :)

That needs some more thought to the whole signal handling business and
unfortunately the recent posix-timer changes do not help much in that
context. Let me think about it some more.

Thanks,

        tglx