lkml.org 
[lkml]   [2011]   [Jul]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 4/8] coredump_wait: don't call complete_vfork_done()
On 07/29, Matt Fleming wrote:
>
> On Wed, 2011-07-27 at 18:33 +0200, Oleg Nesterov wrote:
> > Now that CLONE_VFORK is killable, coredump_wait() no longer needs
> > complete_vfork_done(). zap_threads() should find and kill all tasks
> > with the same ->mm, this includes our parent if ->vfork_done is set.
> >
> > mm_release() becomes the only caller, unexport complete_vfork_done().
> >
> > Signed-off-by: Oleg Nesterov <oleg@redhat.com>
>
> Isn't there a subtle change in user-visible behaviour regarding wait()
> with this patch?
>
> Before the patch, if a child dumps its core it will wakeup the parent
> which can read the status of the child via wait(), whereas with this
> patch applied the parent will actually be killed along with the child.

No.

Please note that if ->vfork_done != NULL, then ->real_parent shares
->mm with us, by definition of CLONE_VFORK.

In this case, with or without this patch, the parent was already
killed by zap_threads(). It can never do wait() or something else.

However. before 3/8, it was necessary to wakeup the TASK_UNINTERRUPTIBLE
parent, otherwise we deadlock. Once again, it can't do anything,
it will die immediately because of fatal_signal_pending().

After 3/8, zap_process()->signal_wake_up(1) wakes up the KILLABLE
parent, no need to do complete().

Oleg.



\
 
 \ /
  Last update: 2011-07-29 16:31    [W:0.053 / U:0.836 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site