Re: [Patch v4 1/2] freezer: check OOM kill while being frozen

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Monday, September 08, 2014 04:00:41 PM Cong Wang wrote:
> On Mon, Sep 8, 2014 at 4:15 PM, Rafael J. Wysocki <rjw@rjwysocki.net> wrote:
> > On Tuesday, September 09, 2014 07:50:12 AM Tejun Heo wrote:
> >> Hello,
> >>
> >> On Tue, Sep 09, 2014 at 12:48:28AM +0200, Rafael J. Wysocki wrote:
> >> > Well, not exactly from anywhere.  Just from where try_to_freeze() is called
> >> > I suppose?
> >>
> >> Yeap, anywhere try_to_freeze() may be called.
> >>
> >> > Which means that if this is a user space task, it won't to a lot before dying,
> >> > will it?
> >>
> >> Userland tasks aren't likely to a lot of damages before dying but then
> >> again there's no guarantee.  As it currently stands, try_to_freeze()
> >> can be anywhere.
> >
> > Well, in that case the TIF_MEMDIE test alone is not sufficient in my opinion,
> > because we can't guarantee that the task will not do something it was frozen
> > to prevent it from doing.
> >
> >
> 
> That is true, there is a small window between TIF_MEMDIE is set
> and SIGKILL is handled (aka process is killed), but this seems to
> be a general signal handling issue not related with freezer after all
> OOM relies on SIGKILL to function. No?

The reason why it matters for the suspend-time freezing is that we freeze tasks
to take them out of the picture entirely until they are thawed.  Therefore we
can't allow them to go back to the picture just for a while until they are
killed.  Frozen tasks are not supposed to get back to the picture at all.

Rafael