Re: Terminate process that fails on a constrained allocation

[Andrew Morton]

Andi Kleen <ak@xxxxxxx> wrote:
>
> 
> Unfortunately Andrew's point with the GFP_NOFS still applies :/
> But I would consider any caller of this not handling NULL be broken.

Sure.

> Andrew do you have any stronger evidence it's a real problem?

No.  About a million years ago I had a make-alloc_pages-fail-at-1%-rate
debug patch.  Doing that again would be an interesting exercise.

Another option is to only fail userspace allocations (__GFP_HIGHMEM is set,
or a new flag in GFP_HIGHUSER).  For the workloads which the NUMA guys care
about I expect this is a 99% solution and the amount of code which it puts
at risk is vastly less.

-
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/