linux-kernel - [RFC PATCH] mm: exclude memory less nodes from zone

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-Id: <1392829383-4125-1-git-send-email-mhocko@suse.cz>
Date:	Wed, 19 Feb 2014 18:03:03 +0100
From:	Michal Hocko <mhocko@...e.cz>
To:	<linux-mm@...ck.org>
Cc:	David Rientjes <rientjes@...gle.com>,
	Nishanth Aravamudan <nacc@...ux.vnet.ibm.com>,
	LKML <linux-kernel@...r.kernel.org>
Subject: [RFC PATCH] mm: exclude memory less nodes from zone_reclaim

We had a report about strange OOM killer strikes on a PPC machine
although there was a lot of swap free and a tons of anonymous memory
which could be swapped out. In the end it turned out that the OOM was
a side effect of zone reclaim which wasn't doesn't unmap and swapp out
and so the system was pushed to the OOM. Although this sounds like a bug
somewhere in the kswapd vs. zone reclaim vs. direct reclaim interaction
numactl on the said hardware suggests that the zone reclaim should
have been set in the first place:
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
node 0 size: 0 MB
node 0 free: 0 MB
node 2 cpus:
node 2 size: 7168 MB
node 2 free: 6019 MB
node distances:
node   0   2
0:  10  40
2:  40  10

So all the CPUs are associated with Node0 which doesn't have any memory
while Node2 contains all the available memory. Node distances cause an
automatic zone_reclaim_mode enabling.

Zone reclaim is intended to keep the allocations local but this doesn't
make any sense on the memory less nodes. So let's exlcude such nodes
for init_zone_allows_reclaim which evaluates zone reclaim behavior and
suitable reclaim_nodes.

Signed-off-by: Michal Hocko <mhocko@...e.cz>
---
I haven't got to testing this so I am sending this as an RFC for now.
But does this look reasonable?

 mm/page_alloc.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 3e953f07edb0..4a44bdc7a8cf 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1855,7 +1855,7 @@ static void __paginginit init_zone_allows_reclaim(int nid)
 {
 	int i;

-	for_each_online_node(i)
+	for_each_node_state(i, N_HIGH_MEMORY)
 		if (node_distance(nid, i) <= RECLAIM_DISTANCE)
 			node_set(i, NODE_DATA(nid)->reclaim_nodes);
 		else
@@ -4901,7 +4901,8 @@ void __paginginit free_area_init_node(int nid, unsigned long *zones_size,

 	pgdat->node_id = nid;
 	pgdat->node_start_pfn = node_start_pfn;
-	init_zone_allows_reclaim(nid);
+	if (node_state(nid, N_HIGH_MEMORY))
+		init_zone_allows_reclaim(nid);
 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
 	get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
 #endif
-- 
1.9.0.rc3

--
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/