Re: [PATCH] kernel/exit: do panic earlier to get coredump if global init task exit

[Christian Brauner]

On Thu, Dec 12, 2019 at 11:08:38AM +0100, Oleg Nesterov wrote:
> can't you use is_global_init() && group_dead ?

Seems reasonable.
Looks like we can move
group_dead = atomic_dec_and_test(&tsk->signal->live);
further up...

(Ideally I'd like to have a test for this to ensure that this lets you
capture a global init coredump but that might be tricky. But since you've
seem to have run into this case maybe you even have something that could
be turned into a test? (Similar to how we already have a purely opt-in
test for pstore.))

Christian