Re: [PATCH 1/3 for 2.6.38] oom: oom_kill_process: don't set TIF_MEMDIE if !p->mm

[Oleg Nesterov <oleg@redhat.com>]

On 03/14, Linus Torvalds wrote:
>
> On Mon, Mar 14, 2011 at 12:04 PM, Oleg Nesterov <oleg@redhat.com> wrote:
> > oom_kill_process() simply sets TIF_MEMDIE and returns if PF_EXITING.
> > This is very wrong by many reasons. In particular, this thread can
> > be the dead group leader. Check p->mm != NULL.
>
> Explain more, please. Maybe I'm missing some context because I wasn't
> cc'd on the original thread, but PF_EXITING gets set by exit_signal(),
> and exit_mm() is called almost immediately afterwards which will set
> p->mm to NULL.
>
> So afaik, this will basically just remove the whole point of the code
> entirely - so why not remove it then?

I am afraid I am going to lie... But iirc I tried to remove this code
before. Can't find the previous discussion, probably I am wrong.

Anyway. I never understood why do we have this special case.

> The combination of testing PF_EXITING and p->mm just doesn't seem to
> make any sense.

To me, it doesn't make too much sense even if we do not check ->mm.

But. I _think_ the intent was to wait until this "exiting" process
does exit_mm() and frees the memory. This is like the
"the process of releasing memory " code in select_bad_process(). Once
again, this is only my speculation.

In any case, this patch doesn't pretend to be the right fix.

Oleg.