The patch titled
Subject: mm: vmscan: do not throttle based on pfmemalloc reserves if node has no reclaimable pages
has been added to the -mm tree. Its filename is
mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages.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/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Nishanth Aravamudan <nacc@xxxxxxxxxxxxxxxxxx>
Subject: mm: vmscan: do not throttle based on pfmemalloc reserves if node has no reclaimable pages
Based upon 675becce15 ("mm: vmscan: do not throttle based on pfmemalloc
reserves if node has no ZONE_NORMAL") from Mel.
We have a system with the following topology:
# numactl -H
available: 3 nodes (0,2-3)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
23 24 25 26 27 28 29 30 31
node 0 size: 28273 MB
node 0 free: 27323 MB
node 2 cpus:
node 2 size: 16384 MB
node 2 free: 0 MB
node 3 cpus: 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
node 3 size: 30533 MB
node 3 free: 13273 MB
node distances:
node 0 2 3
0: 10 20 20
2: 20 10 20
3: 20 20 10
Node 2 has no free memory, because:
# cat /sys/devices/system/node/node2/hugepages/hugepages-16777216kB/nr_hugepages
1
This leads to the following zoneinfo:
Node 2, zone DMA
pages free 0
min 1840
low 2300
high 2760
scanned 0
spanned 262144
present 262144
managed 262144
...
all_unreclaimable: 1
If one then attempts to allocate some normal 16M hugepages via
echo 37 > /proc/sys/vm/nr_hugepages
The echo never returns and kswapd2 consumes CPU cycles.
This is because throttle_direct_reclaim ends up calling
wait_event(pfmemalloc_wait, pfmemalloc_watermark_ok...).
pfmemalloc_watermark_ok() in turn checks all zones on the node if there
are any reserves, and if so, then indicates the watermarks are ok, by
seeing if there are sufficient free pages.
675becce15 added a condition already for memoryless nodes. In this case,
though, the node has memory, it is just all consumed (and not
reclaimable). Effectively, though, the result is the same on this call to
pfmemalloc_watermark_ok() and thus seems like a reasonable additional
condition.
With this change, the afore-mentioned 16M hugepage allocation attempt
succeeds and correctly round-robins between Nodes 1 and 3.
Signed-off-by: Nishanth Aravamudan <nacc@xxxxxxxxxxxxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxx>
Cc: Mel Gorman <mgorman@xxxxxxx>
Cc: Anton Blanchard <anton@xxxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Cc: Michal Hocko <mhocko@xxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxx>
Cc: Dan Streetman <ddstreet@xxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/vmscan.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff -puN mm/vmscan.c~mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages mm/vmscan.c
--- a/mm/vmscan.c~mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages
+++ a/mm/vmscan.c
@@ -2646,7 +2646,8 @@ static bool pfmemalloc_watermark_ok(pg_d
for (i = 0; i <= ZONE_NORMAL; i++) {
zone = &pgdat->node_zones[i];
- if (!populated_zone(zone))
+ if (!populated_zone(zone) ||
+ zone_reclaimable_pages(zone) == 0)
continue;
pfmemalloc_reserve += min_wmark_pages(zone);
_
Patches currently in -mm which might be from nacc@xxxxxxxxxxxxxxxxxx are
mm-vmscan-do-not-throttle-based-on-pfmemalloc-reserves-if-node-has-no-reclaimable-pages.patch
linux-next.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
