Re: [PATCH 2/3 for 2.6.38] oom: select_bad_process: ignore TIF_MEMDIEzombies

[David Rientjes]

On Mon, 14 Mar 2011, Oleg Nesterov wrote:

> select_bad_process() assumes that a TIF_MEMDIE process should go away.
> But it can only go away it its parent does wait(). Change this check to
> ignore the TIF_MEMDIE zombies.
> 

The equivalent of this change would be to set TIF_MEMDIE for all threads 
in a thread group when choosing a process to kill; as we've already 
discussed in your first series of patches, that has the risk of fully 
depleting memory reserves and causing the kernel the deadlock.  We want to 
limit TIF_MEMDIE to an oom killed task or to current when it is responding 
to a SIGKILL or already in the exit path because we know it's exiting and 
without memory reserves it may never exit.

This patch is even more concerning, however, because select_bad_process() 
isn't even guaranteed to select a thread from the same thread group this 
time.

> Note: this is _not_ enough. Just a minimal fix.
> 
> Signed-off-by: Oleg Nesterov <oleg@xxxxxxxxxx>
> ---
> 
>  mm/oom_kill.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> --- 38/mm/oom_kill.c~2_tif_memdie_zombie	2011-03-14 18:51:49.000000000 +0100
> +++ 38/mm/oom_kill.c	2011-03-14 18:52:39.000000000 +0100
> @@ -311,7 +311,8 @@ static struct task_struct *select_bad_pr
>  		 * blocked waiting for another task which itself is waiting
>  		 * for memory. Is there a better alternative?
>  		 */
> -		if (test_tsk_thread_flag(p, TIF_MEMDIE))
> +		if (test_tsk_thread_flag(p, TIF_MEMDIE) &&
> +		    !p->exit_state && thread_group_empty(p))
>  			return ERR_PTR(-1UL);
>  
>  		/*
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/