linux-kernel - Re: [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]

Date:	Wed, 19 Feb 2014 09:53:39 -0800
From:	Nishanth Aravamudan <nacc@...ux.vnet.ibm.com>
To:	Michal Hocko <mhocko@...e.cz>
Cc:	linux-mm@...ck.org, David Rientjes <rientjes@...gle.com>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [RFC PATCH] mm: exclude memory less nodes from zone_reclaim

On 19.02.2014 [18:03:03 +0100], Michal Hocko wrote:
> 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);

I'm still new to this code, but isn't this saying that if a node has no
memory, then it shouldn't reclaim from any node? But, for a memoryless
node to ensure progress later if reclaim is necessary, it *must* reclaim
from other nodes? So wouldn't we want to set reclaim_nodes() in that
case to node_states[N_MEMORY]?

Thanks,
Nish

--
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/