Re: [RFC PATCH] mm: exclude memory less nodes from zone_reclaim

[Michal Hocko <mhocko@suse.cz>]

On Wed 19-02-14 15:05:58, Nishanth Aravamudan wrote:
> On 19.02.2014 [13:56:00 -0800], David Rientjes wrote:
> > On Wed, 19 Feb 2014, Nishanth Aravamudan wrote:
> > 
> > > > 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]?
> > > 
> > 
> > The only time when pgdat->reclaim_nodes or zone_reclaim_mode matters is 
> > when iterating through a zonelist for page allocation and a memoryless 
> > node should never appear in a zonelist for page allocation, so this is 
> > just preventing setting zone_reclaim_mode unnecessarily because the only 
> > nodes with > RECLAIM_DISTANCE to another node are memoryless.  So this 
> > patch is fine as long as it gets s/N_HIGH_MEMORY/N_MEMORY/.
> 
> Ah yes, sorry, I've been looking at this code perhaps too much and going
> a bit cross-eyed!
> 
> I wonder if we should also put some comments in? But
> 
> Acked-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>
> Tested-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>

Thanks both of you for acks and testing. I will submit the updated patch
and include Andrew to pick it up.

-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>