Re: [PATCH 1/2] exec: Don't set group_exit_task during a coredump

[Eric W. Biederman]

Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> writes:

> On Fri, Jun 19, 2020 at 11:36 AM Eric W. Biederman
> <ebiederm@xxxxxxxxxxxx> wrote:
>>
>> Instead test SIGNAL_GROUP_COREDUMP in signal_group_exit().
>
> You say "instead", but the patch itself doesn't agree:
>
>>  static inline int signal_group_exit(const struct signal_struct *sig)
>>  {
>> -       return  (sig->flags & SIGNAL_GROUP_EXIT) ||
>> +       return  (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) ||
>>                 (sig->group_exit_task != NULL);
>>  }
>
> it does it _in_addition_to_.

Hmm.  I think I can change that line to:
>> Instead add a test for SIGNAL_GROUP_COREDUMP in signal_group_exit().

Does that read better?

> I think the whole test for "sig->group_exit_task != NULL" should be
> removed for this commit to make sense.

The code change is designed not to have a behavioral change in
signal_group_exit().  As de_thread also sets sig->group_exit_task
the test for sig->group_exit_task needs to remain in signal_group_exit()
for the behavior of signal_group_exit() to remain unchanged.



Why do you think the test sig->group_exit_task != NULL should be removed
for the commit to make sense?

Eric