On Mon, 2 Jun 2014, Davidlohr Bueso wrote:

On Mon, 2014-06-02 at 13:08 -0700, Andrew Morton wrote:

quoted

On Thu, 29 May 2014 19:20:15 -0700 Davidlohr Bueso [off-list ref] wrote:

quoted

On Thu, 2014-05-22 at 20:33 -0700, Davidlohr Bueso wrote:

quoted

This patchset extends the work started by Ingo Molnar in late 2012,
optimizing the anon-vma mutex lock, converting it from a exclusive mutex
to a rwsem, and sharing the lock for read-only paths when walking the
the vma-interval tree. More specifically commits 5a505085 and 4fc3f1d6.

The i_mmap_mutex has similar responsibilities with the anon-vma, protecting
file backed pages. Therefore we can use similar locking techniques: covert
the mutex to a rwsem and share the lock when possible.

With the new optimistic spinning property we have in rwsems, we no longer
take a hit in performance when using this lock, and we can therefore
safely do the conversion. Tests show no throughput regressions in aim7 or
pgbench runs, and we can see gains from sharing the lock, in disk workloads
~+15% for over 1000 users on a 8-socket Westmere system.

This patchset applies on linux-next-20140522.

ping? Andrew any chance of getting this in -next?

(top-posting repaired)

It was a bit late for 3.16 back on May 26, when you said "I will dig
deeper (probably for 3.17 now)".  So, please take another look at the
patch factoring and let's get this underway for -rc1.

Ok, so I meant that I'd dig deeper for the additional sharing
opportunities (which I've found a few as Hugh correctly suggested). So
those eventual patches could come later. 

But I see no reason for *this* patchset to be delayed, as even if it
gets to be 3.17 material, I'd still very much want to have the same
patch factoring I have now. I think its the correct way to handle lock
transitioning for both correctness and bisectability.

I'd be glad to see it go into 3.16 if it works as well as advertized.
And if you're attached to your current 2/5, fine, do stick with that.
But please do a proper job on your 3/5, instead of just aping how the
anon case worked out.

Hugh

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>