Re: [PATCH 1/2] mm: page allocator: Adjust the per-cpu counterthreshold when memory is low

[Andrew Morton]

On Thu, 23 Dec 2010 15:07:02 -0800 (PST)
David Rientjes <rientjes@xxxxxxxxxx> wrote:

> On Thu, 23 Dec 2010, Andrew Morton wrote:
> 
> > > We had to pull aa454840 "mm: page allocator: calculate a better estimate 
> > > of NR_FREE_PAGES when memory is low and kswapd is awake" from 2.6.36 
> > > internally because tests showed that it would cause the machine to stall 
> > > as the result of heavy kswapd activity.  I merged it back with this fix as 
> > > it is pending in the -mm tree and it solves the issue we were seeing, so I 
> > > definitely think this should be pushed to -stable (and I would seriously 
> > > consider it for 2.6.37 inclusion even at this late date).
> > 
> > How's about I send
> > mm-page-allocator-adjust-the-per-cpu-counter-threshold-when-memory-is-low.patch
> > in for 2.6.38 and tag it for backporting into 2.6.37.1 and 2.6.36.x? 
> > That way it'll get a bit of 2.6.38-rc testing before being merged into
> > 2.6.37.x.
> > 
> 
> I don't think anyone would be able to answer that judgment call other than 
> you or Linus, it's a trade-off on whether 2.6.37 should be released with 
> the knowledge that it regresses just like 2.6.36 does (rendering both 
> unusable on some of our machines out of the box) because we're late in the 
> cycle.
> 
> I personally think the testing is already sufficient since it's been 
> sitting in -mm for two months, it's been suggested as stable material by a 
> couple different parties, it was a prerequisite for the transparent 
> hugepage series, and we've tested and merged it as fixing the regression 
> in 2.6.36 (as Fedora has, as far as I know).  We've already merged the fix 
> internally, though, so it's not for selfish reasons :)

Wibble, wobble.  It's good that the patch has been used in RH kernels. 
otoh, the patch is really quite big and the problem was present in
2.6.36 without a lot of complaints and we're very late in -rc and not
many people will be testing over xmas/newyear, and it would be most sad
to put badness into mainline at this time.

So I'm still inclined to go with (discretion > valour).
--
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/