[PATCH] mm: ALLOC_HIGHATOMIC flag allocation issue

[Zhiguo Jiang]

In case of alloc_flags contain ALLOC_HIGHATOMIC and alloc order
is order1/2/3/10 in rmqueue(), if pages are alloced successfully
from pcplist cache, and move a free pageblock from the alloced
migratetype freelist to MIGRATE_HIGHATOMIC freelist, rather than
alloc from MIGRATE_HIGHATOMIC freelist firstly, so this will result
in an increasing number of pages on the MIGRATE_HIGHATOMIC freelist,
pages in other migratetype freelist are reduced and more likely to
allocation failure.

Currently the sequence of ALLOC_HIGHATOMIC allocation is:
pcplist cache --> buddy (batch >> order) allocation migratetype
freelist --> buddy MIGRATE_HIGHATOMIC freelist --> buddy allocation
migratetype freelist.

Due to the fact that requesting pages from the pcplist cache is faster
than buddy, the sequence of modifying the ALLOC_HIGHATOMIC allocation is:
pcplist --> buddy MIGRATE_HIGHATOMIC freelist --> buddy allocation
migrate freelist.

This patch can solve the allocation failure of the Non-ALLOC_HIGHATOMIC
alloc_flag due to excessive pages reservations in MIGRATE_HIGHATOMIC
freelists.

Signed-off-by: Zhiguo Jiang <justinjiang@xxxxxxxx>
---
 mm/internal.h   |  1 +
 mm/page_alloc.c | 22 ++++++++++++++--------
 2 files changed, 15 insertions(+), 8 deletions(-)
 mode change 100644 => 100755 mm/internal.h
 mode change 100644 => 100755 mm/page_alloc.c

diff --git a/mm/internal.h b/mm/internal.h
index 7499b5ea1cf6..a3b3fdda8031
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -844,6 +844,7 @@ unsigned int reclaim_clean_pages_from_list(struct zone *zone,
 #endif
 #define ALLOC_HIGHATOMIC	0x200 /* Allows access to MIGRATE_HIGHATOMIC */
 #define ALLOC_KSWAPD		0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
+#define ALLOC_PCPLIST		0x1000 /* page alloced from pcplist */
 
 /* Flags that allow allocations below the min watermark. */
 #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 442c1b3480aa..afcd0f875c92
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2722,6 +2722,8 @@ struct page *__rmqueue_pcplist(struct zone *zone, unsigned int order,
 			int batch = READ_ONCE(pcp->batch);
 			int alloced;
 
+			if (alloc_flags & ALLOC_HIGHATOMIC)
+				goto out;
 			/*
 			 * Scale batch relative to order if batch implies
 			 * free pages can be stored on the PCP. Batch can
@@ -2736,6 +2738,7 @@ struct page *__rmqueue_pcplist(struct zone *zone, unsigned int order,
 					migratetype, alloc_flags);
 
 			pcp->count += alloced << order;
+out:
 			if (unlikely(list_empty(list)))
 				return NULL;
 		}
@@ -2798,7 +2801,7 @@ __no_sanitize_memory
 static inline
 struct page *rmqueue(struct zone *preferred_zone,
 			struct zone *zone, unsigned int order,
-			gfp_t gfp_flags, unsigned int alloc_flags,
+			gfp_t gfp_flags, unsigned int *alloc_flags,
 			int migratetype)
 {
 	struct page *page;
@@ -2814,21 +2817,23 @@ struct page *rmqueue(struct zone *preferred_zone,
 		 * MIGRATE_MOVABLE pcplist could have the pages on CMA area and
 		 * we need to skip it when CMA area isn't allowed.
 		 */
-		if (!IS_ENABLED(CONFIG_CMA) || alloc_flags & ALLOC_CMA ||
+		if (!IS_ENABLED(CONFIG_CMA) || *alloc_flags & ALLOC_CMA ||
 				migratetype != MIGRATE_MOVABLE) {
 			page = rmqueue_pcplist(preferred_zone, zone, order,
-					migratetype, alloc_flags);
-			if (likely(page))
+					migratetype, *alloc_flags);
+			if (likely(page)) {
+				*alloc_flags |= ALLOC_PCPLIST;
 				goto out;
+			}
 		}
 	}
 
-	page = rmqueue_buddy(preferred_zone, zone, order, alloc_flags,
+	page = rmqueue_buddy(preferred_zone, zone, order, *alloc_flags,
 							migratetype);
 
 out:
 	/* Separate test+clear to avoid unnecessary atomics */
-	if ((alloc_flags & ALLOC_KSWAPD) &&
+	if ((*alloc_flags & ALLOC_KSWAPD) &&
 	    unlikely(test_bit(ZONE_BOOSTED_WATERMARK, &zone->flags))) {
 		clear_bit(ZONE_BOOSTED_WATERMARK, &zone->flags);
 		wakeup_kswapd(zone, 0, 0, zone_idx(zone));
@@ -3208,7 +3213,7 @@ get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags,
 
 try_this_zone:
 		page = rmqueue(ac->preferred_zoneref->zone, zone, order,
-				gfp_mask, alloc_flags, ac->migratetype);
+				gfp_mask, &alloc_flags, ac->migratetype);
 		if (page) {
 			prep_new_page(page, order, gfp_mask, alloc_flags);
 
@@ -3216,7 +3221,8 @@ get_page_from_freelist(gfp_t gfp_mask, unsigned int order, int alloc_flags,
 			 * If this is a high-order atomic allocation then check
 			 * if the pageblock should be reserved for the future
 			 */
-			if (unlikely(alloc_flags & ALLOC_HIGHATOMIC))
+			if (unlikely(alloc_flags & ALLOC_HIGHATOMIC)
+				&& unlikely(!(alloc_flags & ALLOC_PCPLIST)))
 				reserve_highatomic_pageblock(page, zone);
 
 			return page;
-- 
2.39.0