On Fri, Oct 16, 2009 at 12:07:07PM -0700, David Rientjes wrote:

On Fri, 16 Oct 2009, Mel Gorman wrote:

quoted

ALLOC_NO_WATERMARKS should be cleared when trying to allocate from the
free-lists after a direct reclaim. If it's not, __GFP_NOFAIL allocations
from a process that is exiting can ignore watermarks. __GFP_NOFAIL is not
often used but the journal layer is one of those places. This is suspected of
causing an increase in the number of GFP_ATOMIC allocation failures reported.

Signed-off-by: Mel Gorman <redacted>
---
 mm/page_alloc.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index dfa4362..a3e5fed 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c

@@ -1860,7 +1860,8 @@ rebalance:
 	page = __alloc_pages_direct_reclaim(gfp_mask, order,
 					zonelist, high_zoneidx,
 					nodemask,
-					alloc_flags, preferred_zone,
+					alloc_flags & ~ALLOC_NO_WATERMARKS,
+					preferred_zone,
 					migratetype, &did_some_progress);
 	if (page)
 		goto got_pg;

I don't get it.  __alloc_pages_high_priority() will already loop 
indefinitely if ALLOC_NO_WATERMARKS is set and its a __GFP_NOFAIL 
allocation.  How do we even reach this code in such a condition?

Frans, you reported that both patches in combination reduced the number
of failures. Was it in fact just the kswapd change that made the
difference?

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
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>