Re: [PATCH] oom_pardon, aka don't kill my xlock

[William Lee Irwin III]

On Thu, Sep 23, 2004 at 01:23:08AM +0200, Thomas Habets wrote:
> How about a sysctl that does "for the love of kbaek, don't ever kill these 
> processes when OOM. If nothing else can be killed, I'd rather you panic"?
> Examples for this list would be /usr/bin/vlock and /usr/X11R6/bin/xlock. 
> I just got a very uncomfortable surprise when found my box unlocked thanks to 
> this.
> After playing around a bit, I made the patch below, but it's almost
> completely untested. I'm not even sure I take the binaries name from
> the right place. And I don't know if the locking can race. If it's
> too ugly then it'd be great if someone implemented it the right way.
> (iow: huge fucking disclaimer)
> echo "/usr/bin/vlock /usr/X11R6/bin/xlock" > /proc/sys/vm/oom_pardon

Assuming this is desirable (otherwise, why would you have written it?)

(1) uts_sem isn't the right lock.
(2) You acquire uts_sem under tasklist_lock, a deadlock.
(3) It would probably make more sense to dynamically register and
	unregister the various criteria for exempt processes than mess
	with space-separated fields of a single string.


-- wli
-
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/