Thread (14 messages) 14 messages, 5 authors, 2012-09-24

Re: [PATCH v4 4/4] cma: fix watermark checking

From: Bartlomiej Zolnierkiewicz <hidden>
Date: 2012-09-24 09:31:52

On Wednesday 19 September 2012 21:51:02 Andrew Morton wrote:
On Fri, 14 Sep 2012 16:29:34 +0200
Bartlomiej Zolnierkiewicz [off-list ref] wrote:
quoted
* Add ALLOC_CMA alloc flag and pass it to [__]zone_watermark_ok()
  (from Minchan Kim).
What is its meaning and why was it added.
ALLOC_CMA flag means that allocation can use CMA areas to get free pages
from.  Free CMA pages are accounted by system as normal free pages
(NR_FREE_PAGES) but they can be used only by movable type allocations
(otherwise they cannot be migrated once we want to do CMA allocation)
so we have to fix free_pages in __zone_watermark_ok() or the watermark
check will be too optimistic for unmovable allocations.
quoted
* During watermark check decrease available free pages number by
  free CMA pages number if necessary (unmovable allocations cannot
  use pages from CMA areas).

...
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -231,6 +231,21 @@ enum zone_watermarks {
 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
 
+/* The ALLOC_WMARK bits are used as an index to zone->watermark */
+#define ALLOC_WMARK_MIN		WMARK_MIN
+#define ALLOC_WMARK_LOW		WMARK_LOW
+#define ALLOC_WMARK_HIGH	WMARK_HIGH
+#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
+
+/* Mask to get the watermark bits */
+#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
+
+#define ALLOC_HARDER		0x10 /* try to alloc harder */
+#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
+#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
+
Unneeded newline.
quoted
+#define ALLOC_CMA		0x80
All the other enumerations were documented.  ALLOC_CMA was left
undocumented, despite sorely needing documentation.
quoted
 struct per_cpu_pages {
 	int count;		/* number of pages in the list */
 	int high;		/* high watermark, emptying needed */
diff --git a/mm/compaction.c b/mm/compaction.c
index 4b902aa..36d79ea 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -868,6 +868,7 @@ unsigned long try_to_compact_pages(struct zonelist *zonelist,
 	struct zoneref *z;
 	struct zone *zone;
 	int rc = COMPACT_SKIPPED;
+	int alloc_flags = 0;
 
 	/*
 	 * Check whether it is worth even starting compaction. The order check is
@@ -879,6 +880,10 @@ unsigned long try_to_compact_pages(struct zonelist *zonelist,
 
 	count_vm_event(COMPACTSTALL);
 
+#ifdef CONFIG_CMA
+	if (allocflags_to_migratetype(gfp_mask) == MIGRATE_MOVABLE)
+		alloc_flags |= ALLOC_CMA;
I find this rather obscure.  What is the significance of
MIGRATE_MOVABLE here?  If it had been 

:	if (allocflags_to_migratetype(gfp_mask) == MIGRATE_CMA)
:		alloc_flags |= ALLOC_CMA;

then I'd have read straight past it.  But it's unclear what's happening
here.  If we didn't have to resort to telepathy to understand the
meaning of ALLOC_CMA, this wouldn't be so hard.
quoted
+#endif
 	/* Compact each zone in the list */
 	for_each_zone_zonelist_nodemask(zone, z, zonelist, high_zoneidx,
 								nodemask) {
@@ -889,7 +894,8 @@ unsigned long try_to_compact_pages(struct zonelist *zonelist,
 		rc = max(status, rc);
 
 		/* If a normal allocation would succeed, stop compacting */
-		if (zone_watermark_ok(zone, order, low_wmark_pages(zone), 0, 0))
+		if (zone_watermark_ok(zone, order, low_wmark_pages(zone), 0,
+				      alloc_flags))
 			break;
 	}
 
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 287f79d..5985cbf 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1519,19 +1519,6 @@ failed:
 	return NULL;
 }
 
-/* The ALLOC_WMARK bits are used as an index to zone->watermark */
-#define ALLOC_WMARK_MIN		WMARK_MIN
-#define ALLOC_WMARK_LOW		WMARK_LOW
-#define ALLOC_WMARK_HIGH	WMARK_HIGH
-#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
-
-/* Mask to get the watermark bits */
-#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
-
-#define ALLOC_HARDER		0x10 /* try to alloc harder */
-#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
-#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
Perhaps mm/internal.h wouild have been a better place to move these.
quoted
 #ifdef CONFIG_FAIL_PAGE_ALLOC
 
 static struct {
@@ -1626,7 +1613,10 @@ static bool __zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 		min -= min / 2;
 	if (alloc_flags & ALLOC_HARDER)
 		min -= min / 4;
-
+#ifdef CONFIG_CMA
+	if (!(alloc_flags & ALLOC_CMA))
+		free_pages -= zone_page_state(z, NR_FREE_CMA_PAGES);
Again, the negated test looks weird or just wrong.



Please do something to make this code more understandable.
Here is an incremental patch:

From: Bartlomiej Zolnierkiewicz <redacted>
Subject: [PATCH] cma: fix watermark checking cleanup

Changes:
* document ALLOC_CMA
* add comment to __zone_watermark_ok()
* move ALLOC_* defines to mm/internal.h

Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Nazarewicz <redacted>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <redacted>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
---
 include/linux/mmzone.h |   15 ---------------
 mm/internal.h          |   14 ++++++++++++++
 mm/page_alloc.c        |    1 +
 3 files changed, 15 insertions(+), 15 deletions(-)

Index: b/include/linux/mmzone.h
===================================================================
--- a/include/linux/mmzone.h	2012-09-24 11:09:10.700992708 +0200
+++ b/include/linux/mmzone.h	2012-09-24 11:11:36.520992691 +0200
@@ -231,21 +231,6 @@ enum zone_watermarks {
 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
 
-/* The ALLOC_WMARK bits are used as an index to zone->watermark */
-#define ALLOC_WMARK_MIN		WMARK_MIN
-#define ALLOC_WMARK_LOW		WMARK_LOW
-#define ALLOC_WMARK_HIGH	WMARK_HIGH
-#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
-
-/* Mask to get the watermark bits */
-#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
-
-#define ALLOC_HARDER		0x10 /* try to alloc harder */
-#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
-#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
-
-#define ALLOC_CMA		0x80
-
 struct per_cpu_pages {
 	int count;		/* number of pages in the list */
 	int high;		/* high watermark, emptying needed */
Index: b/mm/internal.h
===================================================================
--- a/mm/internal.h	2012-09-24 11:10:04.848992709 +0200
+++ b/mm/internal.h	2012-09-24 11:11:39.972992691 +0200
@@ -356,3 +356,17 @@ extern unsigned long vm_mmap_pgoff(struc
         unsigned long, unsigned long);
 
 extern void set_pageblock_order(void);
+
+/* The ALLOC_WMARK bits are used as an index to zone->watermark */
+#define ALLOC_WMARK_MIN		WMARK_MIN
+#define ALLOC_WMARK_LOW		WMARK_LOW
+#define ALLOC_WMARK_HIGH	WMARK_HIGH
+#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
+
+/* Mask to get the watermark bits */
+#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
+
+#define ALLOC_HARDER		0x10 /* try to alloc harder */
+#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
+#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
+#define ALLOC_CMA		0x80 /* allow allocations from CMA areas */
Index: b/mm/page_alloc.c
===================================================================
--- a/mm/page_alloc.c	2012-09-24 11:12:22.212992717 +0200
+++ b/mm/page_alloc.c	2012-09-24 11:15:28.212992661 +0200
@@ -1614,6 +1614,7 @@ static bool __zone_watermark_ok(struct z
 	if (alloc_flags & ALLOC_HARDER)
 		min -= min / 4;
 #ifdef CONFIG_CMA
+	/* If allocation can't use CMA areas don't use free CMA pages */
 	if (!(alloc_flags & ALLOC_CMA))
 		free_pages -= zone_page_state(z, NR_FREE_CMA_PAGES);
 #endif

--
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>
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help