Re: [PATCH] sched,numa,mm: revert to checking pmd/pte_write instead of VMA flags

[Mel Gorman]

On Mon, Sep 12, 2016 at 11:09:43AM -0400, Rik van Riel wrote:
> > Patch looks ok other than the comment above the second hunk being out
> > of
> > date. Out of curiousity, what workload benefitted from this? I saw a
> > mix
> > of marginal results when I ran this on a 2-socket and 4-socket box.
> 
> I did not performance test the change, because I believe
> the VM_WRITE test has a small logical error.
> 
> Specifically, VM_WRITE is also true for VMAs that are
> PROT_WRITE|MAP_PRIVATE, which we do NOT want to group
> on. Every shared library mapped on my system seems to
> have a (small) read-write VMA:
> 

Ok, while I agree with you, the patch is not a guaranteed win. However,
in the event it is not, I agree that the problem will be with the
grouping code. If the comment is updated then feel free to add my

Acked-by: Mel Gorman <mgorman@xxxxxxx>