linux-kernel - [PATCH] mm: vmscan: do not throttle based on pfmemalloc reserves if node has no reclaimable zones

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20150327192850.GA18701@linux.vnet.ibm.com>
Date:	Fri, 27 Mar 2015 12:28:50 -0700
From:	Nishanth Aravamudan <nacc@...ux.vnet.ibm.com>
To:	Mel Gorman <mgorman@...e.de>
Cc:	anton@...bar.org, linuxppc-dev@...ts.ozlabs.org,
	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	Andrew Morton <akpm@...ux-foundation.org>,
	Johannes Weiner <hannes@...xchg.org>,
	Michal Hocko <mhocko@...e.cz>, Rik van Riel <riel@...hat.com>,
	Dan Streetman <ddstreet@...e.org>
Subject: [PATCH] mm: vmscan: do not throttle based on pfmemalloc reserves if
 node has no reclaimable zones

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:

(0) root @ br30p03: /root
# 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:

echo 37 > /proc/sys/vm/nr_hugepages

The echo enver 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 and see
if the 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
recliamable). 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 succeeds
and correctly round-robins between Nodes 1 and 3.

Signed-off-by: Nishanth Aravamudan <nacc@...ux.vnet.ibm.com>

diff --git a/mm/vmscan.c b/mm/vmscan.c
index dcd90c8..033c2b7 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2585,7 +2585,7 @@ static bool pfmemalloc_watermark_ok(pg_data_t *pgdat)

        for (i = 0; i <= ZONE_NORMAL; i++) {
                zone = &pgdat->node_zones[i];
-               if (!populated_zone(zone))
+               if (!populated_zone(zone) || !zone_reclaimable(zone))
                        continue;

                pfmemalloc_reserve += min_wmark_pages(zone);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/