Re: [PATCH 7/9] signal: Make individual tasks exiting a first class concept.

[Eric W. Biederman]

Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> writes:

> On Thu, Jun 24, 2021 at 12:03 PM Eric W. Biederman
> <ebiederm@xxxxxxxxxxxx> wrote:
>>
>> Implement start_task_exit_locked and rewrite the de_thread logic
>> in exec using it.
>>
>> Calling start_task_exit_locked is equivalent to asyncrhonously
>> calling exit(2) aka pthread_exit on a task.
>
> Ok, so this is the patch that makes me go "Yeah, this seems to all go together".
>
> The whole "start_exit()" thing seemed fairly sane as an interesting
> concept to the whole ptrace notification thing, but this one actually
> made me think it makes conceptual sense and how we had exactly that
> "start exit asynchronously" case already in zap_other_threads().
>
> So doing that zap_other_threads() as that async exit makes me just
> thin kthat yes, this series is the right thing, because it not only
> cleans up the ptrace condition, it makes sense in this entirely
> unrelated area too.
>
> So I think I'm convinced. I'd like Oleg in particular to Ack this
> series, and Al to look it over, but I do think this is the right
> direction.

Thanks.

It took a bit of exploration and playing with things to get here,
but I had the same sense.

Next round I will see if I can clean up the patches a bit more.

Eric