The patch titled
Subject: mm,vmscan: accumulated slab reclaim pressure fixes
has been added to the -mm tree. Its filename is
mmslabvmscan-accumulate-gradual-pressure-on-small-slabs-fix-2.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mmslabvmscan-accumulate-gradual-pressure-on-small-slabs-fix-2.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mmslabvmscan-accumulate-gradual-pressure-on-small-slabs-fix-2.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Rik van Riel <riel@xxxxxxxxxxx>
Subject: mm,vmscan: accumulated slab reclaim pressure fixes
Jonathan asked about a divide by zero if a slab with shrinker->seeks==0
comes into the new code path, which appears to be possible because the
shrinker list is locked with a read-write semaphore, that allows multiple
parallel reclaimers.
This led me to take another closer look at the code, and find a
possibility for a small slab with a shrinker->seeks > 4 to not have any
items reclaimed from it at all.
This patch fixes both of these bugs.
Link: http://lkml.kernel.org/r/20190129142831.6a373403@xxxxxxxxxxxxxxxxxxxx
Signed-off-by: Rik van Riel <riel@xxxxxxxxxxx>
Suggested-by: Jonathan Lemon <bsd@xxxxxx>
Cc: Chris Mason <clm@xxxxxx>
Cc: Roman Gushchin <guro@xxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/vmscan.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
--- a/mm/vmscan.c~mmslabvmscan-accumulate-gradual-pressure-on-small-slabs-fix-2
+++ a/mm/vmscan.c
@@ -500,14 +500,17 @@ static unsigned long do_shrink_slab(stru
* corresponding structures like per-cpu stats and kmem caches
* can be really big, so it may lead to a significant waste of memory.
*/
- if (!delta) {
- shrinker->small_scan += freeable;
+ if (!delta && shrinker->seeks) {
+ unsigned long nr_considered;
- delta = shrinker->small_scan >> priority;
- shrinker->small_scan -= delta << priority;
+ shrinker->small_scan += freeable;
+ nr_considered = shrinker->small_scan >> priority;
- delta *= 4;
+ delta = 4 * nr_considered;
do_div(delta, shrinker->seeks);
+
+ if (delta)
+ shrinker->small_scan -= nr_considered << priority;
}
total_scan += delta;
_
Patches currently in -mm which might be from riel@xxxxxxxxxxx are
mmslabvmscan-accumulate-gradual-pressure-on-small-slabs.patch
mmslabvmscan-accumulate-gradual-pressure-on-small-slabs-fix-2.patch
