Re: [PATCH 0/3] Reduce amount of time kswapd sleeps prematurely

From: Vlastimil Babka
Date: Wed Feb 15 2017 - 17:01:03 EST


On 15.2.2017 22:29, Mel Gorman wrote:
> On Wed, Feb 15, 2017 at 12:30:55PM -0800, Andrew Morton wrote:
>> On Wed, 15 Feb 2017 09:22:44 +0000 Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx> wrote:
>>
>>> This patchset is based on mmots as of Feb 9th, 2016. The baseline is
>>> important as there are a number of kswapd-related fixes in that tree and
>>> a comparison against v4.10-rc7 would be almost meaningless as a result.
>>
>> It's very late to squeeze this into 4.10. We can make it 4.11 material
>> and perhaps tag it for backporting into 4.10.1?
>
> It would be important that Johannes's patches go along with then because
> I'm relied on Johannes' fixes to deal with pages being inappropriately
> written back from reclaim context when I was analysing the workload.
> I'm thinking specifically about these patches
>
> mm-vmscan-scan-dirty-pages-even-in-laptop-mode.patch
> mm-vmscan-kick-flushers-when-we-encounter-dirty-pages-on-the-lru.patch
> mm-vmscan-kick-flushers-when-we-encounter-dirty-pages-on-the-lru-fix.patch
> mm-vmscan-remove-old-flusher-wakeup-from-direct-reclaim-path.patch
> mm-vmscan-only-write-dirty-pages-that-the-scanner-has-seen-twice.patch
> mm-vmscan-move-dirty-pages-out-of-the-way-until-theyre-flushed.patch
> mm-vmscan-move-dirty-pages-out-of-the-way-until-theyre-flushed-fix.patch
>
> This is 4.11 material for sure but I would not automatically try merging
> them to 4.10 unless those patches were also included, ideally with a rerun
> of just those patches against 4.10 to make sure there are no surprises
> lurking in there.

I wonder if we should also care about 4.9 which will be LTS, if we decide to
look at stable at all. IIUC at least the problem that patch 1/3 fixes (wrt
kcompactd not being woken up) is there since 4.8?