On 05/31, Michal Hocko wrote:
>
> On Mon 30-05-16 19:35:05, Oleg Nesterov wrote:
> >
> > Well, let me suggest this again. I think it should do
> >
> >
> > if (SIGNAL_GROUP_COREDUMP)
> > return false;
> >
> > if (SIGNAL_GROUP_EXIT)
> > return true;
> >
> > if (thread_group_empty() && PF_EXITING)
> > return true;
> >
> > return false;
> >
> > we do not need fatal_signal_pending(), in this case SIGNAL_GROUP_EXIT should
> > be set (ignoring some bugs with sub-namespaces which we need to fix anyway).
>
> OK, so we shouldn't care about race when the fatal_signal is set on the
> task until it reaches do_group_exit?
if fatal_signal() is true then (ignoring exec and coredump) SIGNAL_GROUP_EXIT
is already set (again, ignoring the bugs with sub-namespace inits).
At the same time, SIGKILL can be already dequeued when the task exits, so
fatal_signal_pending() can be "false negative".
> > And. I think this needs smp_rmb() at the end of the loop (assuming we have
> > the
> > process_shares_mm() check here). We need it to ensure that we read p->mm
> > before
> > we read next_task(), to avoid the race with exit() + clone(CLONE_VM).
>
> Why don't we need the same barrier in oom_kill_process?
Because it calls do_send_sig_info() which takes ->siglock and copy_process()
takes the same lock. Not a barrier, but acts the same way.
> Which barrier it
> would pair with?
With the barrier implied by list_add_tail_rcu(&p->tasks, &init_task.tasks).
> Anyway I think this would deserve it's own patch.
> Barriers are always tricky and it is better to have them in a small
> patch with a full explanation.
OK, agreed.
I am not sure I can read the new patch correctly, it depends on the previous
changes... but afaics it looks good.
Cosmetic/subjective nit, feel free to ignore,
> +bool task_will_free_mem(struct task_struct *task)
> +{
> + struct mm_struct *mm = NULL;
unnecessary initialization ;)
> + struct task_struct *p;
> + bool ret;
> +
> + /*
> + * If the process has passed exit_mm we have to skip it because
> + * we have lost a link to other tasks sharing this mm, we do not
> + * have anything to reap and the task might then get stuck waiting
> + * for parent as zombie and we do not want it to hold TIF_MEMDIE
> + */
> + p = find_lock_task_mm(task);
> + if (!p)
> + return false;
> +
> + if (!__task_will_free_mem(p)) {
> + task_unlock(p);
> + return false;
> + }
We can call the 1st __task_will_free_mem(p) before find_lock_task_mm(). In the
likely case (I think) it should return false.
And since __task_will_free_mem() has no other callers perhaps it should go into
oom_kill.c too.
Oleg.
