Re: [PATCH 03/34] mm: Reduce the amount of work done when updatingmin_free_kbytes

From: Mel Gorman
Date: Wed Jul 25 2012 - 03:57:36 EST


On Tue, Jul 24, 2012 at 03:47:12PM -0700, Greg KH wrote:
> On Mon, Jul 23, 2012 at 02:38:16PM +0100, Mel Gorman wrote:
> > commit 938929f14cb595f43cd1a4e63e22d36cab1e4a1f upstream.
> >
> > Stable note: Fixes https://bugzilla.novell.com/show_bug.cgi?id=726210 .
> > Large machines with 1TB or more of RAM take a long time to boot
> > without this patch and may spew out soft lockup warnings.
>
> In comparing this with the upstream version, you have a few different
> coding style differences, but no real content difference. Why?
>

This was a mistake in my workflow that needs a bit of ironing out.

The mistake is that I took the patch from the distribution kernel which was
merged at a time before the coding style fixes were applied. The upstream
commit and signed-off lines were "fixed" but I failed to refresh the
patch and missed that it differed from upstream. Thanks for catching this.
I'll adjust my workflow and assistant scripts to watch for this sort of
problem in the future.

> >
> > When min_free_kbytes is updated blocks marked MIGRATE_RESERVE are
> > updated. Ordinarily, this work is unnoticable as it happens early
> > in boot. However, on large machines with 1TB of memory, this can take
> > a considerable time when NUMA distances are taken into account. The bulk
> > of the work is done by pageblock_is_reserved() which examines the
> > metadata for almost every page in the system. Currently, we are doing
> > this far more than necessary as it is only required while there are
> > still blocks to be marked MIGRATE_RESERVE. This patch significantly
> > reduces the amount of work done by setup_zone_migrate_reserve()
> > improving boot times on 1TB machines.
> >
> > [akpm@xxxxxxxxxxxxxxxxxxxx: coding-style fixes]
>
> I'm guessing you didn't pick these up?
>

Correct but due to a mistake, not for any good reason.

> Anyway, I've taken it now as the original one from Linus's tree,
> hopefully this doesn't burn me later in the series...
>

I hope it didn't.

Thanks Greg.

--
Mel Gorman
SUSE Labs
--
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/