On Mon, 2012-07-23 at 14:38 +0100, Mel Gorman wrote:
> Changelog since V1
>   o Expand some of the notes					(jrnieder)
>   o Correct upstream commit SHA1				(hugh)
> 
> This series is related to the new addition to stable_kernel_rules.txt
> 
>  - Serious issues as reported by a user of a distribution kernel may also
>    be considered if they fix a notable performance or interactivity issue.
>    As these fixes are not as obvious and have a higher risk of a subtle
>    regression they should only be submitted by a distribution kernel
>    maintainer and include an addendum linking to a bugzilla entry if it
>    exists and additional information on the user-visible impact.
> 
> All of these patches have been backported to a distribution kernel and
> address some sort of performance issue in the VM. As they are not all
> obvious, I've added a "Stable note" to the top of each patch giving
> additional information on why the patch was backported. Lets see where
> the boundaries lie on how this new rule is interpreted in practice :).
>
> Patch 1	Performance fix for tmpfs
> Patch 2 Memory hotadd fix
> Patch 3 Reduce boot time on large machines
> Patches 4-5 Reduce stalls for wait_iff_congested
> Patches 6-8 Reduce excessive reclaim of slab objects which for some workloads
> 	will reduce the amount of IO required
> Patches 9-10 limits the amount of page reclaim when THP/Compaction is active.
> 	Excessive reclaim in low memory situations can lead to stalls some
> 	of which are user visible.
> Patches 11-19 reduce the amount of churn of the LRU lists. Poor reclaim
> 	decisions can impair workloads in different ways and there have
> 	been complaints recently the reclaim decisions of modern kernels
> 	are worse than older ones.
> Patches 20-21 reduce the amount of CPU kswapd uses in some cases. This
> 	is harder to trigger but were developed due to bug reports about
> 	100% CPU usage from kswapd.
> Patches 22-25 are mostly related to interactivity when THP is enabled.
> Patches 26-30 are also related to page reclaim decisions, particularly
> 	the residency of mapped pages.
> Patches 31-34 fix a major page allocator performance regression
[...]
> The patches are based on 3.0.36 but there should not be problems applying
> the series to later stable releases.
[...]

Patches 1-2, 4-15, 20-21, 31-32 correspond to commits included in Linux
3.2.  I've added the rest to the queue for 3.2.y, generally using the
versions Greg has queued for 3.0.39.

Patch 30 'mm: vmscan: convert global reclaim to per-memcg LRU lists'
needed a further context change.

For patch 33 'cpuset: mm: reduce large amounts of memory barrier related
damage v3' I folded in the two fixes Herton pointed out and you
acknowledged, and took the upstream version of the changes to
get_any_partial() in slub.c.

Ben.

-- 
Ben Hutchings
It is impossible to make anything foolproof because fools are so ingenious.