Re: [PATCH v2 1/1] hugepages: Fix race between hugetlbfs umount andquota update.

[David Gibson]

On Fri, Oct 14, 2011 at 01:59:48PM -0700, Andrew Morton wrote:
> On Wed, 12 Oct 2011 15:43:17 +1100
> Paul Mackerras <paulus@xxxxxxxxx> wrote:
> 
> > In the meantime we have a user-triggerable kernel crash.  As far as I
> > can see, if we did what you suggest, we would end up with a situation
> > where we could run out of huge pages even though everyone was within
> > quota.  Which is arguably better than a kernel crash, but still less
> > than ideal.  What do you suggest?
> 
> My issue with the patch is that it's rather horrible.  We have a layer
> of separation between core hugetlb pages and hugetlbfs.  That layering
> has already been mucked up in various places and this patch mucks it up
> further, and quite severely.
> 
> So I believe we should rethink the patch.  Either a) get the layering
> correct by not poking into hugetlbfs internals from within hugetlb core
> via one of the usual techniques or

Which usual techniques did you have in mind?

> b) make a deliberate decision to
> just give up on that layering: state that hugetlb and hugetlbfs are now
> part of the same subsystem.  Make the necessaary Kconfig changes,
> remove ifdefs, move code around, etc.

Well, that might have something to be said for it, the distinction has
always been tenuous at best.

> If we go ahead with the proposed patch-n-run bugfix, the bad code will
> be there permanently - nobody will go in and clean this mess up and the
> kernel is permanently worsened.

Hrm, as opposed to leaving the crash bug there until someone has time
to do the requested cleanup.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
--
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/