Re: [GIT PULL] core/urgent for v5.16-rc6

[Thomas Gleixner]

On Mon, Dec 20 2021 at 08:25, Linus Torvalds wrote:
> On Mon, Dec 20, 2021 at 8:20 AM Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>>
>> Whatever. I don't care that much, but this all smells like you just
>> dug your own hole for very questionable causes, and instead of a
>> "don't do that then" this all is doubling down on a bad idea.
>
> It further looks like it's really only the sas_ss_size that is
> checked, so if people wan tto have a lock, make it clear that's the
> only thing that the lock is about.
>
> So the actual "do I even need to lock" condition should likely just be
>
>         if (ss_size < t->sas_ss_size)
>                 .. don't bother locking ..
>
> but as mentioned, I don't really see much of a point in being so
> careful even about the growing case.
>
> If somebody is changing xstate features concurrently with another
> thread setting up their altstack, they can keep both pieces.

In principle I agree, but the whole signal stack business is a nightmare
and the way how a program ends up using some xfeature is hideous at
best.

An application does not necessarily know about it at all because the
usage is hidden in random library code. So there is a chance to run into
concurrency issues for real.

Let me grab your (Dave's) patch and rework the whole thing into
something sensible. I had a patch around which replaced sighand lock
with an explicit lock for that purpose. Let me dig that out and polish
it all up.

Thanks,

        tglx