Messages in this thread | | | Date | Wed, 14 Mar 2012 21:08:13 +0800 | Subject | Re: [PATCH] mm: hugetlb: defer freeing pages when gathering surplus pages | From | Hillf Danton <> |
| |
On Wed, Mar 14, 2012 at 7:13 PM, Michal Hocko <mhocko@suse.cz> wrote: > [Sorry for the late reply but I was away from email for quite sometime] >
Nice to see you back:)
> On Tue 14-02-12 20:53:51, Hillf Danton wrote: >> When gathering surplus pages, the number of needed pages is recomputed after >> reacquiring hugetlb lock to catch changes in resv_huge_pages and >> free_huge_pages. Plus it is recomputed with the number of newly allocated >> pages involved. >> >> Thus freeing pages could be deferred a bit to see if the final page request is >> satisfied, though pages could be allocated less than needed. > > The patch looks OK but I am missing a word why we need it. I guess
False negative is removed as it should be.
> your primary motivation is that we want to reduce false positives when > we fail to allocate surplus pages while somebody freed some in the > background. > What is the workload that you observed such a behavior? Or is this just > from the code review? > The second.
-hd
| |