[PATCH 1/2] Add __GFP_FAIL

Use __GFP_FAIL to signal to the page allocator that an allocation attempt
should fail and not attempt any form of reclaim. This is of interest to the
implementation of GFP_THISNODE and memory pools.

Memorypools currently use __GFP_WAIT in order to cause an immediate failure.
However, __GFP_WAIT is typically only cleared for atomic allocations and even
if it is cleared, reclaim may still occur. So use __GFP_FAIL there.

If we have __GFP_FAIL then the special casing for GFP_THISNODE can be removed
from the page allocator. Do that.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
---
 include/linux/gfp.h |    6 ++++--
 mm/mempool.c        |    3 ++-
 mm/page_alloc.c     |   10 +---------
 3 files changed, 7 insertions(+), 12 deletions(-)

Index: linux-2.6.23-rc1-mm2/include/linux/gfp.h
===================================================================
--- linux-2.6.23-rc1-mm2.orig/include/linux/gfp.h	2007-08-03 16:35:33.000000000 -0700
+++ linux-2.6.23-rc1-mm2/include/linux/gfp.h	2007-08-03 16:38:02.000000000 -0700
@@ -43,6 +43,7 @@ struct vm_area_struct;
 #define __GFP_REPEAT	((__force gfp_t)0x400u)	/* Retry the allocation.  Might fail */
 #define __GFP_NOFAIL	((__force gfp_t)0x800u)	/* Retry for ever.  Cannot fail */
 #define __GFP_NORETRY	((__force gfp_t)0x1000u)/* Do not retry.  Might fail */
+#define __GFP_FAIL	((__force gfp_t)0x2000u)/* Fail immediately if there is a problem */
 #define __GFP_COMP	((__force gfp_t)0x4000u)/* Add compound page metadata */
 #define __GFP_ZERO	((__force gfp_t)0x8000u)/* Return zeroed page on success */
 #define __GFP_NOMEMALLOC ((__force gfp_t)0x10000u) /* Don't use emergency reserves */
@@ -59,7 +60,7 @@ struct vm_area_struct;
 			__GFP_COLD|__GFP_NOWARN|__GFP_REPEAT| \
 			__GFP_NOFAIL|__GFP_NORETRY|__GFP_COMP| \
 			__GFP_NOMEMALLOC|__GFP_HARDWALL|__GFP_THISNODE| \
-			__GFP_RECLAIMABLE|__GFP_MOVABLE)
+			__GFP_RECLAIMABLE|__GFP_MOVABLE|__GFP_FAIL)
 
 /* This mask makes up all the page movable related flags */
 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
@@ -87,7 +88,8 @@ struct vm_area_struct;
 				 __GFP_MOVABLE)
 
 #ifdef CONFIG_NUMA
-#define GFP_THISNODE	(__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)
+#define GFP_THISNODE	(__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY |\
+				__GFP_FAIL)
 #else
 #define GFP_THISNODE	((__force gfp_t)0)
 #endif
Index: linux-2.6.23-rc1-mm2/mm/mempool.c
===================================================================
--- linux-2.6.23-rc1-mm2.orig/mm/mempool.c	2007-07-22 13:41:00.000000000 -0700
+++ linux-2.6.23-rc1-mm2/mm/mempool.c	2007-08-03 16:37:38.000000000 -0700
@@ -211,8 +211,9 @@ void * mempool_alloc(mempool_t *pool, gf
 	gfp_mask |= __GFP_NOMEMALLOC;	/* don't allocate emergency reserves */
 	gfp_mask |= __GFP_NORETRY;	/* don't loop in __alloc_pages */
 	gfp_mask |= __GFP_NOWARN;	/* failures are OK */
+	gfp_mask |= __GFP_FAIL;		/* Fail instead of reclaim */
 
-	gfp_temp = gfp_mask & ~(__GFP_WAIT|__GFP_IO);
+	gfp_temp = gfp_mask & ~__GFP_IO;
 
 repeat_alloc:
 
Index: linux-2.6.23-rc1-mm2/mm/page_alloc.c
===================================================================
--- linux-2.6.23-rc1-mm2.orig/mm/page_alloc.c	2007-08-03 16:35:33.000000000 -0700
+++ linux-2.6.23-rc1-mm2/mm/page_alloc.c	2007-08-03 16:37:38.000000000 -0700
@@ -1569,15 +1569,7 @@ restart:
 	if (page)
 		goto got_pg;
 
-	/*
-	 * GFP_THISNODE (meaning __GFP_THISNODE, __GFP_NORETRY and
-	 * __GFP_NOWARN set) should not cause reclaim since the subsystem
-	 * (f.e. slab) using GFP_THISNODE may choose to trigger reclaim
-	 * using a larger set of nodes after it has established that the
-	 * allowed per node queues are empty and that nodes are
-	 * over allocated.
-	 */
-	if (NUMA_BUILD && (gfp_mask & GFP_THISNODE) == GFP_THISNODE)
+	if (gfp_mask & __GFP_FAIL)
 		goto nopage;
 
 	for (z = zonelist->zones; *z; z++)