Re: [PATCH 11 of 11] not-wait-memdie

[Andrea Arcangeli <andrea@cpushare.com>]

On Mon, Jan 07, 2008 at 05:57:41PM -0800, David Rientjes wrote:
> That's only possible with my proposal of adding
> 
> 	unsigned long oom_kill_jiffies;
> 
> to struct task_struct.  We can't get away with a system-wide jiffies 

I already added it.

> variable, nor can we get away with per-cgroup, per-cpuset, or 
> per-mempolicy variable.  The only way to clear such a variable is in the 
> exit path (by checking test_thread_flag(tsk, TIF_MEMDIE) in do_exit()) and 
> fails miserably if there are simultaneous but zone-disjoint OOMs 
> occurring.

I don't see much issues with zone-disjoints oom with my current
patchset. The trouble is a new deadlock that I'm reproducing now, I
submit you privately a preview of the memdie_jiffies, did you see any
problem in my implementation? I guess I'll resubmit to linux-mm
too. The new deadlock I run into after adding memdie_jiffies is likely
unrelated to the memdie_jiffies.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>