Re: [RFD] posix-timers: CRIU woes

[Thomas Gleixner]

Andrey!

On Thu, May 11 2023 at 18:21, Andrey Vagin wrote:
> On Thu, May 11, 2023 at 2:36 AM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>>
>> You know the UABI regression rules of the kernel...
>
> There is no rule without exceptions... With all pros and cons, we may
> consider this case as an exception. From our side, we will try to make
> everything to minimize the impact. Here are steps off the top of my
> head:
> * releasing the criu fix before the kernel release.
> * update packages in Linux distros (Debian, Ubuntu, Fedora, and
>   others that we will find).
> * send an announcement to the criu mailing list and to users that we know.
> * add the error to FAQ.
> * create a GitHub issue with a full description.

Thanks for this plan. After digging deeper I managed to resolve the
actual problem I was chasing without changing that ID generator at
all.

The main pain point of having to do that lookup from the signal delivery
path is gone, which made it trivial to do the fix for the SIG_IGN mess
w/o these global lookups too.

Addressing this global ID issues I pointed out becomes therefore an
orthogonal issue which we can handle completely independent of the
kernel internal problems I'm trying to address.

I still think we should do that for sanity sake, but we can stage that
properly without dependencies outside of this particular ABI
problem. That makes me way more comfortable as that's something which
can be in the worst case reverted without doing any other damage.

Thanks,

        tglx