]> git.baikalelectronics.ru Git - kernel.git/commitdiff
thp: make split_queue per-node
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Wed, 3 Feb 2016 00:57:08 +0000 (16:57 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 3 Feb 2016 16:28:43 +0000 (08:28 -0800)
Andrea Arcangeli suggested to make split queue per-node to improve
scalability.  Let's do it.

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Suggested-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Jerome Marchand <jmarchan@redhat.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/mmzone.h
mm/huge_memory.c
mm/page_alloc.c

index 33bb1b19273e3ad165382f4c7f0f3cd8f8c0b407..7b6c2cfee390e939e13b0f14c92224bcb3aafcf4 100644 (file)
@@ -682,6 +682,12 @@ typedef struct pglist_data {
         */
        unsigned long first_deferred_pfn;
 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+       spinlock_t split_queue_lock;
+       struct list_head split_queue;
+       unsigned long split_queue_len;
+#endif
 } pg_data_t;
 
 #define node_present_pages(nid)        (NODE_DATA(nid)->node_present_pages)
index fd3a07b3e6f4e086b2111c312a78512bed927f9b..253a25e007d71eb9052e06c7383768f12247728b 100644 (file)
@@ -138,9 +138,6 @@ static struct khugepaged_scan khugepaged_scan = {
        .mm_head = LIST_HEAD_INIT(khugepaged_scan.mm_head),
 };
 
-static DEFINE_SPINLOCK(split_queue_lock);
-static LIST_HEAD(split_queue);
-static unsigned long split_queue_len;
 static struct shrinker deferred_split_shrinker;
 
 static void set_recommended_min_free_kbytes(void)
@@ -3358,6 +3355,7 @@ int total_mapcount(struct page *page)
 int split_huge_page_to_list(struct page *page, struct list_head *list)
 {
        struct page *head = compound_head(page);
+       struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
        struct anon_vma *anon_vma;
        int count, mapcount, ret;
        bool mlocked;
@@ -3401,19 +3399,19 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
                lru_add_drain();
 
        /* Prevent deferred_split_scan() touching ->_count */
-       spin_lock_irqsave(&split_queue_lock, flags);
+       spin_lock_irqsave(&pgdata->split_queue_lock, flags);
        count = page_count(head);
        mapcount = total_mapcount(head);
        if (!mapcount && count == 1) {
                if (!list_empty(page_deferred_list(head))) {
-                       split_queue_len--;
+                       pgdata->split_queue_len--;
                        list_del(page_deferred_list(head));
                }
-               spin_unlock_irqrestore(&split_queue_lock, flags);
+               spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
                __split_huge_page(page, list);
                ret = 0;
        } else if (IS_ENABLED(CONFIG_DEBUG_VM) && mapcount) {
-               spin_unlock_irqrestore(&split_queue_lock, flags);
+               spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
                pr_alert("total_mapcount: %u, page_count(): %u\n",
                                mapcount, count);
                if (PageTail(page))
@@ -3421,7 +3419,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
                dump_page(page, "total_mapcount(head) > 0");
                BUG();
        } else {
-               spin_unlock_irqrestore(&split_queue_lock, flags);
+               spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
                unfreeze_page(anon_vma, head);
                ret = -EBUSY;
        }
@@ -3436,52 +3434,56 @@ out:
 
 void free_transhuge_page(struct page *page)
 {
+       struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
        unsigned long flags;
 
-       spin_lock_irqsave(&split_queue_lock, flags);
+       spin_lock_irqsave(&pgdata->split_queue_lock, flags);
        if (!list_empty(page_deferred_list(page))) {
-               split_queue_len--;
+               pgdata->split_queue_len--;
                list_del(page_deferred_list(page));
        }
-       spin_unlock_irqrestore(&split_queue_lock, flags);
+       spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
        free_compound_page(page);
 }
 
 void deferred_split_huge_page(struct page *page)
 {
+       struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
        unsigned long flags;
 
        VM_BUG_ON_PAGE(!PageTransHuge(page), page);
 
-       spin_lock_irqsave(&split_queue_lock, flags);
+       spin_lock_irqsave(&pgdata->split_queue_lock, flags);
        if (list_empty(page_deferred_list(page))) {
-               list_add_tail(page_deferred_list(page), &split_queue);
-               split_queue_len++;
+               list_add_tail(page_deferred_list(page), &pgdata->split_queue);
+               pgdata->split_queue_len++;
        }
-       spin_unlock_irqrestore(&split_queue_lock, flags);
+       spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
 }
 
 static unsigned long deferred_split_count(struct shrinker *shrink,
                struct shrink_control *sc)
 {
+       struct pglist_data *pgdata = NODE_DATA(sc->nid);
        /*
         * Split a page from split_queue will free up at least one page,
         * at most HPAGE_PMD_NR - 1. We don't track exact number.
         * Let's use HPAGE_PMD_NR / 2 as ballpark.
         */
-       return ACCESS_ONCE(split_queue_len) * HPAGE_PMD_NR / 2;
+       return ACCESS_ONCE(pgdata->split_queue_len) * HPAGE_PMD_NR / 2;
 }
 
 static unsigned long deferred_split_scan(struct shrinker *shrink,
                struct shrink_control *sc)
 {
+       struct pglist_data *pgdata = NODE_DATA(sc->nid);
        unsigned long flags;
        LIST_HEAD(list), *pos, *next;
        struct page *page;
        int split = 0;
 
-       spin_lock_irqsave(&split_queue_lock, flags);
-       list_splice_init(&split_queue, &list);
+       spin_lock_irqsave(&pgdata->split_queue_lock, flags);
+       list_splice_init(&pgdata->split_queue, &list);
 
        /* Take pin on all head pages to avoid freeing them under us */
        list_for_each_safe(pos, next, &list) {
@@ -3490,10 +3492,10 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
                /* race with put_compound_page() */
                if (!get_page_unless_zero(page)) {
                        list_del_init(page_deferred_list(page));
-                       split_queue_len--;
+                       pgdata->split_queue_len--;
                }
        }
-       spin_unlock_irqrestore(&split_queue_lock, flags);
+       spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
 
        list_for_each_safe(pos, next, &list) {
                page = list_entry((void *)pos, struct page, mapping);
@@ -3505,9 +3507,9 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
                put_page(page);
        }
 
-       spin_lock_irqsave(&split_queue_lock, flags);
-       list_splice_tail(&list, &split_queue);
-       spin_unlock_irqrestore(&split_queue_lock, flags);
+       spin_lock_irqsave(&pgdata->split_queue_lock, flags);
+       list_splice_tail(&list, &pgdata->split_queue);
+       spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
 
        return split * HPAGE_PMD_NR / 2;
 }
@@ -3516,6 +3518,7 @@ static struct shrinker deferred_split_shrinker = {
        .count_objects = deferred_split_count,
        .scan_objects = deferred_split_scan,
        .seeks = DEFAULT_SEEKS,
+       .flags = SHRINKER_NUMA_AWARE,
 };
 
 #ifdef CONFIG_DEBUG_FS
index 63358d9f9aa98eff0848879b0503b5d80b9ea8d0..ea2c4d3e0c0354d241a7cd60ae81b2a9f95cdf93 100644 (file)
@@ -5209,6 +5209,11 @@ static void __paginginit free_area_init_core(struct pglist_data *pgdat)
        spin_lock_init(&pgdat->numabalancing_migrate_lock);
        pgdat->numabalancing_migrate_nr_pages = 0;
        pgdat->numabalancing_migrate_next_window = jiffies;
+#endif
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+       spin_lock_init(&pgdat->split_queue_lock);
+       INIT_LIST_HEAD(&pgdat->split_queue);
+       pgdat->split_queue_len = 0;
 #endif
        init_waitqueue_head(&pgdat->kswapd_wait);
        init_waitqueue_head(&pgdat->pfmemalloc_wait);