Re: [patch] cpusets: do not allow TIF_MEMDIE tasks to allocate globally

[David Rientjes <rientjes@google.com>]

On Tue, 5 Jun 2007, Paul Jackson wrote:

> I'm a little surprised at this suggested change -- I'd have thought
> that it was a good idea to let tasks marked for extinction get memory
> anywhere, as they were going to use that memory to exit, and free up
> lots more memory.
> 

The intended purpose of TIF_MEMDIE was to allocate pages without being 
bound by the watermarks so that they have access to memory reserves on the 
per-zone level.  If the cpuset doesn't have access to a zone, whether it's 
memory reserve or not, it shouldn't allocate there.

> I'm pretty sure we have this same policy in other places in the
> kernel, besides cpusets.  Did you intend to change them too?
> 

You'd have to cite them first.

> If a MEMDIE task is taking enough memory to OOM other tasks anywhere
> in the system, then doesn't that mean your entire system was in deep
> yogurt, and we're just haggling over who to blame for the upcoming
> crash?
> 

No, it means that it can allocate anywhere based on the zonelist ordering 
and then can OOM a very small exclusive cpuset that would never have had 
any memory pressure if it wasn't violated.

		David
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/