Re: [RFC] decrease tsk->signal->live before profile_task_exit

From: liuchao (CR)
Date: Mon Jun 08 2020 - 22:32:09 EST


Eric W. Biederman <mailto:ebiederm@xxxxxxxxxxxx> writes:

> liuchao <liuchao173@xxxxxxxxxx> writes:
>
> > I want to dermine which thread is the last one to enter do_exit in
> > profile_task_exit. But when a lot of threads exit, tsk->signal->live
> > is not correct since it decrease after profile_task_exit.
>
> I don't think that would be wise.
>
> Any additional code before the sanity checks at the start of do_exit seems
> like a bad idea.
>
> We could probably move the decrement of tsk->signal->live a little earlier,
> but not that much earlier in the function.
>
> Does profile_task_exit even make sense that early in the code? If the code
> is doing much of anything that is a completely inappopriate placement of
> profile_task_exit.

I think so too.

Move the decrement of tsk->signal->live after the sanity checks, then
profile_task_exit and kcov_task_exit make more sense.

>
> Eric
>
>
> > Signed-off-by: liuchao <liuchao173@xxxxxxxxxx>
> > ---
> > kernel/exit.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/kernel/exit.c b/kernel/exit.c index
> > ce2a75bc0ade..1693764bc356 100644
> > --- a/kernel/exit.c
> > +++ b/kernel/exit.c
> > @@ -708,6 +708,7 @@ void __noreturn do_exit(long code)
> > struct task_struct *tsk = current;
> > int group_dead;
> >
> > + group_dead = atomic_dec_and_test(&tsk->signal->live);
> > profile_task_exit(tsk);
> > kcov_task_exit(tsk);
> >
> > @@ -755,7 +756,6 @@ void __noreturn do_exit(long code)
> > if (tsk->mm)
> > sync_mm_rss(tsk->mm);
> > acct_update_integrals(tsk);
> > - group_dead = atomic_dec_and_test(&tsk->signal->live);
> > if (group_dead) {
> > /*
> > * If the last thread of global init has exited, panic