memory: merge modification of tao.zeng [3/9]

mm: fix lowmem issue

PD#157252: mm: fix lowmem issue

1. add statistics for CMA pages;
2. reduce cma print and not protect cma unless driver called;
3. fix cma usage policy in alloc path;
4. increase file scan ratio in kswapd;
5. using NOOP for default IO-scheduler;
6. change alloc flags of ZRAM to retry harder.

Change-Id: If7b0363da03da1682efe3996c69bb9d511299209
Signed-off-by: tao zeng <tao.zeng@amlogic.com>

12 files changed, 124 insertions, 17 deletions

diff --git a/arch/arm/configs/meson32_defconfig b/arch/arm/configs/meson32_defconfig
index 6ebf43d..993a05c 100644
--- a/arch/arm/configs/meson32_defconfig
+++ b/arch/arm/configs/meson32_defconfig
@@ -18,6 +18,7 @@ CONFIG_MODULES=y
 CONFIG_MODULE_UNLOAD=y
 CONFIG_PARTITION_ADVANCED=y
 CONFIG_CMDLINE_PARTITION=y
+CONFIG_DEFAULT_NOOP=y
 CONFIG_ARCH_VIRT=y
 CONFIG_ARCH_MESON=y
 # CONFIG_MACH_MESON6 is not set
@@ -227,7 +228,7 @@ CONFIG_AMLOGIC_M8B_SM=y
 CONFIG_DEVTMPFS=y
 CONFIG_DEVTMPFS_MOUNT=y
 CONFIG_DMA_CMA=y
-CONFIG_CMA_SIZE_MBYTES=64
+CONFIG_CMA_SIZE_MBYTES=8
 CONFIG_MTD=y
 CONFIG_MTD_BLOCK=y
 CONFIG_MTD_NAND=y
diff --git a/arch/arm64/boot/dts/amlogic/gxl_p241_1g.dts b/arch/arm64/boot/dts/amlogic/gxl_p241_1g.dts
index c4175c7..5dc7a2e2 100644
--- a/arch/arm64/boot/dts/amlogic/gxl_p241_1g.dts
+++ b/arch/arm64/boot/dts/amlogic/gxl_p241_1g.dts
@@ -91,7 +91,7 @@
 		ion_reserved:linux,ion-dev {
 			compatible = "shared-dma-pool";
 			reusable;
-			size = <0x0 0x5C00000>;
+			size = <0x0 0x3000000>;
 			alignment = <0x0 0x400000>;
 		};
 
@@ -123,7 +123,7 @@
 			compatible = "shared-dma-pool";
 			reusable;
 			/* ion_codec_mm max can alloc size 80M*/
-			size = <0x0 0x13400000>;
+			size = <0x0 0x7800000>;
 			alignment = <0x0 0x400000>;
 			linux,contiguous-region;
 		};
diff --git a/arch/arm64/boot/dts/amlogic/gxl_p241_1g_buildroot.dts b/arch/arm64/boot/dts/amlogic/gxl_p241_1g_buildroot.dts
index 77fe520..6e2ea5f 100644
--- a/arch/arm64/boot/dts/amlogic/gxl_p241_1g_buildroot.dts
+++ b/arch/arm64/boot/dts/amlogic/gxl_p241_1g_buildroot.dts
@@ -121,7 +121,7 @@
 		codec_mm_cma:linux,codec_mm_cma {
 			compatible = "shared-dma-pool";
 			reusable;
-			size = <0x0 0xb400000>;
+			size = <0x0 0x7800000>;
 			alignment = <0x0 0x400000>;
 			linux,contiguous-region;
 		};
diff --git a/arch/arm64/boot/dts/amlogic/gxl_p241_v2-1g.dts b/arch/arm64/boot/dts/amlogic/gxl_p241_v2-1g.dts
index 3e10d3e..09efec4 100644
--- a/arch/arm64/boot/dts/amlogic/gxl_p241_v2-1g.dts
+++ b/arch/arm64/boot/dts/amlogic/gxl_p241_v2-1g.dts
@@ -91,7 +91,7 @@
 		ion_reserved:linux,ion-dev {
 			compatible = "shared-dma-pool";
 			reusable;
-			size = <0x0 0x5C00000>;
+			size = <0x0 0x3000000>;
 			alignment = <0x0 0x400000>;
 		};
 
@@ -123,7 +123,7 @@
 			compatible = "shared-dma-pool";
 			reusable;
 			/* ion_codec_mm max can alloc size 80M*/
-			size = <0x0 0x13400000>;
+			size = <0x0 0x7800000>;
 			alignment = <0x0 0x400000>;
 			linux,contiguous-region;
 		};
diff --git a/arch/arm64/boot/dts/amlogic/gxl_p241_v2_1g_buildroot.dts b/arch/arm64/boot/dts/amlogic/gxl_p241_v2_1g_buildroot.dts
index 1e6f560..302b770 100644
--- a/arch/arm64/boot/dts/amlogic/gxl_p241_v2_1g_buildroot.dts
+++ b/arch/arm64/boot/dts/amlogic/gxl_p241_v2_1g_buildroot.dts
@@ -121,7 +121,7 @@
 		codec_mm_cma:linux,codec_mm_cma {
 			compatible = "shared-dma-pool";
 			reusable;
-			size = <0x0 0xb400000>;
+			size = <0x0 0x7800000>;
 			alignment = <0x0 0x400000>;
 			linux,contiguous-region;
 		};
diff --git a/arch/arm64/configs/meson64_defconfig b/arch/arm64/configs/meson64_defconfig
index 1395f8b..0ce1cc5 100644
--- a/arch/arm64/configs/meson64_defconfig
+++ b/arch/arm64/configs/meson64_defconfig
@@ -29,6 +29,7 @@ CONFIG_CC_STACKPROTECTOR_STRONG=y
 CONFIG_MODULES=y
 CONFIG_MODULE_UNLOAD=y
 CONFIG_MODVERSIONS=y
+CONFIG_DEFAULT_NOOP=y
 CONFIG_PCI=y
 CONFIG_PCIE_DW_PLAT=y
 CONFIG_SCHED_MC=y
@@ -326,6 +327,7 @@ CONFIG_UEVENT_HELPER_PATH="/sbin/hotplug"
 CONFIG_DEVTMPFS=y
 CONFIG_DEVTMPFS_MOUNT=y
 CONFIG_DMA_CMA=y
+CONFIG_CMA_SIZE_MBYTES=8
 CONFIG_CONNECTOR=y
 CONFIG_MTD=y
 CONFIG_MTD_CMDLINE_PARTS=y
diff --git a/arch/arm64/configs/meson64_smarthome_defconfig b/arch/arm64/configs/meson64_smarthome_defconfig
index 8762164..2c9189a 100644
--- a/arch/arm64/configs/meson64_smarthome_defconfig
+++ b/arch/arm64/configs/meson64_smarthome_defconfig
@@ -23,6 +23,7 @@ CONFIG_PROFILING=y
 CONFIG_JUMP_LABEL=y
 CONFIG_MODULES=y
 CONFIG_MODULE_UNLOAD=y
+CONFIG_DEFAULT_NOOP=y
 CONFIG_PCI=y
 CONFIG_PCIE_DW_PLAT=y
 CONFIG_SCHED_MC=y
@@ -305,6 +306,7 @@ CONFIG_UEVENT_HELPER_PATH="/sbin/hotplug"
 CONFIG_DEVTMPFS=y
 CONFIG_DEVTMPFS_MOUNT=y
 CONFIG_DMA_CMA=y
+CONFIG_CMA_SIZE_MBYTES=8
 CONFIG_CONNECTOR=y
 CONFIG_MTD=y
 CONFIG_MTD_CMDLINE_PARTS=y
diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index c9914d6..896c688 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -755,9 +755,18 @@ compress_again:
 
 		atomic64_inc(&zram->stats.writestall);
 
+	#ifdef CONFIG_AMLOGIC_MODIFY
+		handle = zs_malloc(meta->mem_pool, clen,
+				GFP_NOIO |
+				__GFP_HIGHMEM |
+				__GFP_NOWARN |
+				__GFP_REPEAT |
+				__GFP_MOVABLE);
+	#else
 		handle = zs_malloc(meta->mem_pool, clen,
 				GFP_NOIO | __GFP_HIGHMEM |
 				__GFP_MOVABLE);
+	#endif
 		if (handle)
 			goto compress_again;
 
diff --git a/include/linux/mm_inline.h b/include/linux/mm_inline.h
index 41d376e..2b91387 100644
--- a/include/linux/mm_inline.h
+++ b/include/linux/mm_inline.h
@@ -3,6 +3,9 @@
 
 #include <linux/huge_mm.h>
 #include <linux/swap.h>
+#ifdef CONFIG_AMLOGIC_MODIFY
+#include <linux/page-isolation.h>
+#endif /* CONFIG_AMLOGIC_MODIFY */
 
 /**
  * page_is_file_cache - should the page be on a file LRU or anon LRU?
@@ -46,15 +49,41 @@ static __always_inline void update_lru_size(struct lruvec *lruvec,
 static __always_inline void add_page_to_lru_list(struct page *page,
 				struct lruvec *lruvec, enum lru_list lru)
 {
+#ifdef CONFIG_AMLOGIC_MODIFY
+	int nr_pages = hpage_nr_pages(page);
+	int num = NR_INACTIVE_ANON_CMA - NR_INACTIVE_ANON;
+	int migrate_type = 0;
+#endif /* CONFIG_AMLOGIC_MODIFY */
+
 	update_lru_size(lruvec, lru, page_zonenum(page), hpage_nr_pages(page));
 	list_add(&page->lru, &lruvec->lists[lru]);
+
+#ifdef CONFIG_AMLOGIC_MODIFY
+	migrate_type = get_pageblock_migratetype(page);
+	if (is_migrate_cma(migrate_type) || is_migrate_isolate(migrate_type))
+		__mod_zone_page_state(page_zone(page),
+				      NR_LRU_BASE + lru + num, nr_pages);
+#endif /* CONFIG_AMLOGIC_MODIFY */
 }
 
 static __always_inline void del_page_from_lru_list(struct page *page,
 				struct lruvec *lruvec, enum lru_list lru)
 {
+#ifdef CONFIG_AMLOGIC_MODIFY
+	int nr_pages = hpage_nr_pages(page);
+	int num = NR_INACTIVE_ANON_CMA - NR_INACTIVE_ANON;
+	int migrate_type = 0;
+#endif /* CONFIG_AMLOGIC_MODIFY */
+
 	list_del(&page->lru);
 	update_lru_size(lruvec, lru, page_zonenum(page), -hpage_nr_pages(page));
+
+#ifdef CONFIG_AMLOGIC_MODIFY
+	migrate_type = get_pageblock_migratetype(page);
+	if (is_migrate_cma(migrate_type) || is_migrate_isolate(migrate_type))
+		__mod_zone_page_state(page_zone(page),
+				      NR_LRU_BASE + lru + num, -nr_pages);
+#endif /* CONFIG_AMLOGIC_MODIFY */
 }
 
 /**
diff --git a/mm/cma.c b/mm/cma.c
index a07c7ba..a51103a 100644
--- a/mm/cma.c
+++ b/mm/cma.c
@@ -416,9 +416,6 @@ struct page *cma_alloc(struct cma *cma, size_t count, unsigned int align)
 	if (bitmap_count > bitmap_maxno)
 		return NULL;
 
-#ifdef CONFIG_AMLOGIC_MODIFY
-	atomic_inc(&cma_alloc_ref);
-#endif
 	for (;;) {
 		mutex_lock(&cma->lock);
 		bitmap_no = bitmap_find_next_zero_area_off(cma->bitmap,
@@ -465,13 +462,10 @@ struct page *cma_alloc(struct cma *cma, size_t count, unsigned int align)
 	trace_cma_alloc(pfn, page, count, align);
 
 #ifdef CONFIG_AMLOGIC_MODIFY
-	atomic_dec(&cma_alloc_ref);
 	if (page) {
 		atomic_long_add(count, &driver_alloc_cma);
 		update_cma_page_trace(page, count);
 	}
-	WARN_ONCE(!page, "can't alloc from %lx with size:%ld, ret:%d\n",
-		  cma->base_pfn, count, ret);
 #endif /* CONFIG_AMLOGIC_MODIFY */
 	pr_debug("%s(): returned %p\n", __func__, page);
 	return page;
@@ -522,7 +516,7 @@ bool cma_suitable(gfp_t gfp_mask)
 		return false;
 
 	/* try to reduce page lock wait for read */
-	if (atomic_read(&cma_alloc_ref) && (gfp_mask & __GFP_COLD))
+	if (atomic_read(&cma_alloc_ref))
 		return false;
 
 	return true;
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index ecb2cf4..dc0ff7e 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2314,14 +2314,19 @@ __rmqueue_fallback(struct zone *zone, unsigned int order, int start_migratetype)
  * Do the hard work of removing an element from the buddy allocator.
  * Call me with the zone->lock already held.
  */
+#ifdef CONFIG_AMLOGIC_MODIFY
+static struct page *__rmqueue(struct zone *zone, unsigned int order,
+				int migratetype, gfp_t gfp_flags)
+#else
 static struct page *__rmqueue(struct zone *zone, unsigned int order,
 				int migratetype)
+#endif /* CONFIG_AMLOGIC_MODIFY */
 {
 	struct page *page;
 
 #ifdef CONFIG_AMLOGIC_MODIFY
 	/* use CMA first */
-	if (migratetype == MIGRATE_MOVABLE) {
+	if (migratetype == MIGRATE_MOVABLE && cma_suitable(gfp_flags)) {
 		page = __rmqueue_cma_fallback(zone, order);
 		if (page) {
 			trace_mm_page_alloc_zone_locked(page, order,
@@ -2372,15 +2377,25 @@ static struct page *rmqueue_no_cma(struct zone *zone, unsigned int order,
  * a single hold of the lock, for efficiency.  Add them to the supplied list.
  * Returns the number of new pages which were placed at *list.
  */
+#ifdef CONFIG_AMLOGIC_MODIFY
+static int rmqueue_bulk(struct zone *zone, unsigned int order,
+			unsigned long count, struct list_head *list,
+			int migratetype, bool cold, gfp_t flags)
+#else
 static int rmqueue_bulk(struct zone *zone, unsigned int order,
 			unsigned long count, struct list_head *list,
 			int migratetype, bool cold)
+#endif /* CONFIG_AMLOGIC_MODIFY */
 {
 	int i, alloced = 0;
 
 	spin_lock(&zone->lock);
 	for (i = 0; i < count; ++i) {
+	#ifdef CONFIG_AMLOGIC_MODIFY
+		struct page *page = __rmqueue(zone, order, migratetype, flags);
+	#else
 		struct page *page = __rmqueue(zone, order, migratetype);
+	#endif /* CONFIG_AMLOGIC_MODIFY */
 		if (unlikely(page == NULL))
 			break;
 
@@ -2803,9 +2818,16 @@ struct page *buffered_rmqueue(struct zone *preferred_zone,
 			pcp = &this_cpu_ptr(zone->pageset)->pcp;
 			list = &pcp->lists[migratetype];
 			if (list_empty(list)) {
+			#ifdef CONFIG_AMLOGIC_MODIFY
+				pcp->count += rmqueue_bulk(zone, 0,
+						pcp->batch, list,
+						migratetype, cold,
+						gfp_flags);
+			#else
 				pcp->count += rmqueue_bulk(zone, 0,
 						pcp->batch, list,
 						migratetype, cold);
+			#endif /* CONFIG_AMLOGIC_MODIFY */
 				if (unlikely(list_empty(list)))
 					goto failed;
 			}
@@ -2875,7 +2897,12 @@ use_pcp:
 					trace_mm_page_alloc_zone_locked(page, order, migratetype);
 			}
 			if (!page)
+			#ifdef CONFIG_AMLOGIC_MODIFY
+				page = __rmqueue(zone, order,
+						 migratetype, gfp_flags);
+			#else
 				page = __rmqueue(zone, order, migratetype);
+			#endif /* CONFIG_AMLOGIC_MODIFY */
 		} while (page && check_new_pages(page, order));
 		spin_unlock(&zone->lock);
 		if (!page)
@@ -3986,7 +4013,7 @@ got_pg:
 static inline void should_wakeup_kswap(gfp_t gfp_mask, int order,
 				       struct alloc_context *ac)
 {
-	unsigned long free_pages;
+	unsigned long free_pages, free_cma = 0;
 	struct zoneref *z = ac->preferred_zoneref;
 	struct zone *zone;
 
@@ -3996,6 +4023,11 @@ static inline void should_wakeup_kswap(gfp_t gfp_mask, int order,
 	for_next_zone_zonelist_nodemask(zone, z, ac->zonelist, ac->high_zoneidx,
 								ac->nodemask) {
 		free_pages = zone_page_state(zone, NR_FREE_PAGES);
+	#ifdef CONFIG_CMA
+		if (cma_suitable(gfp_mask))
+			free_cma = zone_page_state(zone, NR_FREE_CMA_PAGES);
+	#endif
+		free_pages -= free_cma;
 		/*
 		 * wake up kswapd before get pages from buddy, this help to
 		 * fast reclaim process and can avoid memory become too low
@@ -4584,7 +4616,7 @@ void show_free_areas(unsigned int filter)
 		" slab_reclaimable:%lu slab_unreclaimable:%lu\n"
 		" mapped:%lu shmem:%lu pagetables:%lu bounce:%lu\n"
 	#ifdef CONFIG_AMLOGIC_MODIFY
-		" driver_cma:%lu"
+		" [cma] driver:%lu anon:%lu file:%lu isolate:%lu total:%lu\n"
 	#endif /* CONFIG_AMLOGIC_MODIFY */
 		" free:%lu free_pcp:%lu free_cma:%lu\n",
 		global_node_page_state(NR_ACTIVE_ANON),
@@ -4605,6 +4637,12 @@ void show_free_areas(unsigned int filter)
 		global_page_state(NR_BOUNCE),
 	#ifdef CONFIG_AMLOGIC_MODIFY
 		get_driver_alloc_cma(),
+		global_page_state(NR_INACTIVE_ANON_CMA) +
+		global_page_state(NR_ACTIVE_ANON_CMA),
+		global_page_state(NR_INACTIVE_FILE_CMA) +
+		global_page_state(NR_ACTIVE_FILE_CMA),
+		global_page_state(NR_CMA_ISOLATED),
+		totalcma_pages,
 	#endif /* CONFIG_AMLOGIC_MODIFY */
 		global_page_state(NR_FREE_PAGES),
 		free_pcp,
diff --git a/mm/vmscan.c b/mm/vmscan.c
index c4ae05f..f2ca309 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1456,6 +1456,10 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 	unsigned long nr_skipped[MAX_NR_ZONES] = { 0, };
 	unsigned long scan, nr_pages;
 	LIST_HEAD(pages_skipped);
+#ifdef CONFIG_AMLOGIC_MODIFY
+	int num = NR_INACTIVE_ANON_CMA - NR_INACTIVE_ANON;
+	int migrate_type = 0;
+#endif /* CONFIG_AMLOGIC_MODIFY */
 
 	for (scan = 0; scan < nr_to_scan && nr_taken < nr_to_scan &&
 					!list_empty(src);) {
@@ -1484,6 +1488,14 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 			nr_taken += nr_pages;
 			nr_zone_taken[page_zonenum(page)] += nr_pages;
 			list_move(&page->lru, dst);
+		#ifdef CONFIG_AMLOGIC_MODIFY
+			migrate_type = get_pageblock_migratetype(page);
+			if (is_migrate_cma(migrate_type) ||
+			    is_migrate_isolate(migrate_type))
+				__mod_zone_page_state(page_zone(page),
+					NR_LRU_BASE + lru + num,
+					-nr_pages);
+		#endif /* CONFIG_AMLOGIC_MODIFY */
 			break;
 
 		case -EBUSY:
@@ -1896,6 +1908,10 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 	unsigned long pgmoved = 0;
 	struct page *page;
 	int nr_pages;
+#ifdef CONFIG_AMLOGIC_MODIFY
+	int num = NR_INACTIVE_ANON_CMA - NR_INACTIVE_ANON;
+	int migrate_type = 0;
+#endif /* CONFIG_AMLOGIC_MODIFY */
 
 	while (!list_empty(list)) {
 		page = lru_to_page(list);
@@ -1908,6 +1924,14 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
 		update_lru_size(lruvec, lru, page_zonenum(page), nr_pages);
 		list_move(&page->lru, &lruvec->lists[lru]);
 		pgmoved += nr_pages;
+	#ifdef CONFIG_AMLOGIC_MODIFY
+		migrate_type = get_pageblock_migratetype(page);
+		if (is_migrate_cma(migrate_type) ||
+		    is_migrate_isolate(migrate_type))
+			__mod_zone_page_state(page_zone(page),
+					      NR_LRU_BASE + lru + num,
+					      nr_pages);
+	#endif /* CONFIG_AMLOGIC_MODIFY */
 
 		if (put_page_testzero(page)) {
 			__ClearPageLRU(page);
@@ -2076,6 +2100,10 @@ static bool inactive_list_is_low(struct lruvec *lruvec, bool file,
 	gb = (inactive + active) >> (30 - PAGE_SHIFT);
 	if (gb)
 		inactive_ratio = int_sqrt(10 * gb);
+#ifdef CONFIG_AMLOGIC_MODIFY
+	else if (!file && (totalram_pages >> (20 - PAGE_SHIFT)) >= 512)
+		inactive_ratio = 2;
+#endif /* CONFIG_AMLOGIC_MODIFY */
 	else
 		inactive_ratio = 1;
 
@@ -2230,6 +2258,10 @@ static void get_scan_count(struct lruvec *lruvec, struct mem_cgroup *memcg,
 	 * This scanning priority is essentially the inverse of IO cost.
 	 */
 	anon_prio = swappiness;
+#ifdef CONFIG_AMLOGIC_MODIFY
+	if (get_nr_swap_pages() * 3 < total_swap_pages)
+		anon_prio >>= 1;
+#endif /* CONFIG_AMLOGIC_MODIFY */
 	file_prio = 200 - anon_prio;
 
 	/*