ext4: Don't allow lg prealloc list to be grow large.

Currently, the locality group prealloc list is freed only when there
is a block allocation failure. This can result in large number of
entries in the preallocation list making ext4_mb_use_preallocated()
expensive.

To fix this, we convert the locality group prealloc list to a hash
list. The hash index is the order of number of blocks in the prealloc
space with a max order of 9. When adding prealloc space to the list we
make sure total entries for each order does not exceed 8. If it is
more than 8 we discard few entries and make sure the we have only <= 5
entries.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

fs/ext4/mballoc.c

@@ -2480,7 +2480,7 @@ static int ext4_mb_init_backend(struct super_block *sb)
 int ext4_mb_init(struct super_block *sb, int needs_recovery)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
-	unsigned i;
+	unsigned i, j;
 	unsigned offset;
 	unsigned max;
 	int ret;
@@ -2552,7 +2552,8 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
 		struct ext4_locality_group *lg;
 		lg = &sbi->s_locality_groups[i];
 		mutex_init(&lg->lg_mutex);
-		INIT_LIST_HEAD(&lg->lg_prealloc_list);
+		for (j = 0; j < PREALLOC_TB_SIZE; j++)
+			INIT_LIST_HEAD(&lg->lg_prealloc_list[j]);
 		spin_lock_init(&lg->lg_prealloc_lock);
 	}
 
@@ -3263,6 +3264,7 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
 				struct ext4_prealloc_space *pa)
 {
 	unsigned int len = ac->ac_o_ex.fe_len;
+
 	ext4_get_group_no_and_offset(ac->ac_sb, pa->pa_pstart,
 					&ac->ac_b_ex.fe_group,
 					&ac->ac_b_ex.fe_start);
@@ -3285,6 +3287,7 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
 static noinline_for_stack int
 ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
 {
+	int order, i;
 	struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
 	struct ext4_locality_group *lg;
 	struct ext4_prealloc_space *pa;
@@ -3325,22 +3328,29 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
 	lg = ac->ac_lg;
 	if (lg == NULL)
 		return 0;
+	order  = fls(ac->ac_o_ex.fe_len) - 1;
+	if (order > PREALLOC_TB_SIZE - 1)
+		/* The max size of hash table is PREALLOC_TB_SIZE */
+		order = PREALLOC_TB_SIZE - 1;
 
-	rcu_read_lock();
+	for (i = order; i < PREALLOC_TB_SIZE; i++) {
-	list_for_each_entry_rcu(pa, &lg->lg_prealloc_list, pa_inode_list) {
+		rcu_read_lock();
-		spin_lock(&pa->pa_lock);
+		list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i],
-		if (pa->pa_deleted == 0 && pa->pa_free >= ac->ac_o_ex.fe_len) {
+					pa_inode_list) {
-			atomic_inc(&pa->pa_count);
+			spin_lock(&pa->pa_lock);
-			ext4_mb_use_group_pa(ac, pa);
+			if (pa->pa_deleted == 0 &&
+					pa->pa_free >= ac->ac_o_ex.fe_len) {
+				atomic_inc(&pa->pa_count);
+				ext4_mb_use_group_pa(ac, pa);
+				spin_unlock(&pa->pa_lock);
+				ac->ac_criteria = 20;
+				rcu_read_unlock();
+				return 1;
+			}
 			spin_unlock(&pa->pa_lock);
-			ac->ac_criteria = 20;
-			rcu_read_unlock();
-			return 1;
 		}
-		spin_unlock(&pa->pa_lock);
+		rcu_read_unlock();
 	}
-	rcu_read_unlock();
-
 	return 0;
 }
 
@@ -3563,6 +3573,7 @@ ext4_mb_new_group_pa(struct ext4_allocation_context *ac)
 	pa->pa_free = pa->pa_len;
 	atomic_set(&pa->pa_count, 1);
 	spin_lock_init(&pa->pa_lock);
+	INIT_LIST_HEAD(&pa->pa_inode_list);
 	pa->pa_deleted = 0;
 	pa->pa_linear = 1;
 
@@ -3583,10 +3594,10 @@ ext4_mb_new_group_pa(struct ext4_allocation_context *ac)
 	list_add(&pa->pa_group_list, &grp->bb_prealloc_list);
 	ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
 
-	spin_lock(pa->pa_obj_lock);
+	/*
-	list_add_tail_rcu(&pa->pa_inode_list, &lg->lg_prealloc_list);
+	 * We will later add the new pa to the right bucket
-	spin_unlock(pa->pa_obj_lock);
+	 * after updating the pa_free in ext4_mb_release_context
-
+	 */
 	return 0;
 }
 
@@ -4123,22 +4134,168 @@ ext4_mb_initialize_context(struct ext4_allocation_context *ac,
 
 }
 
+static noinline_for_stack void
+ext4_mb_discard_lg_preallocations(struct super_block *sb,
+					struct ext4_locality_group *lg,
+					int order, int total_entries)
+{
+	ext4_group_t group = 0;
+	struct ext4_buddy e4b;
+	struct list_head discard_list;
+	struct ext4_prealloc_space *pa, *tmp;
+	struct ext4_allocation_context *ac;
+
+	mb_debug("discard locality group preallocation\n");
+
+	INIT_LIST_HEAD(&discard_list);
+	ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
+
+	spin_lock(&lg->lg_prealloc_lock);
+	list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[order],
+						pa_inode_list) {
+		spin_lock(&pa->pa_lock);
+		if (atomic_read(&pa->pa_count)) {
+			/*
+			 * This is the pa that we just used
+			 * for block allocation. So don't
+			 * free that
+			 */
+			spin_unlock(&pa->pa_lock);
+			continue;
+		}
+		if (pa->pa_deleted) {
+			spin_unlock(&pa->pa_lock);
+			continue;
+		}
+		/* only lg prealloc space */
+		BUG_ON(!pa->pa_linear);
+
+		/* seems this one can be freed ... */
+		pa->pa_deleted = 1;
+		spin_unlock(&pa->pa_lock);
+
+		list_del_rcu(&pa->pa_inode_list);
+		list_add(&pa->u.pa_tmp_list, &discard_list);
+
+		total_entries--;
+		if (total_entries <= 5) {
+			/*
+			 * we want to keep only 5 entries
+			 * allowing it to grow to 8. This
+			 * mak sure we don't call discard
+			 * soon for this list.
+			 */
+			break;
+		}
+	}
+	spin_unlock(&lg->lg_prealloc_lock);
+
+	list_for_each_entry_safe(pa, tmp, &discard_list, u.pa_tmp_list) {
+
+		ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, NULL);
+		if (ext4_mb_load_buddy(sb, group, &e4b)) {
+			ext4_error(sb, __func__, "Error in loading buddy "
+					"information for %lu\n", group);
+			continue;
+		}
+		ext4_lock_group(sb, group);
+		list_del(&pa->pa_group_list);
+		ext4_mb_release_group_pa(&e4b, pa, ac);
+		ext4_unlock_group(sb, group);
+
+		ext4_mb_release_desc(&e4b);
+		list_del(&pa->u.pa_tmp_list);
+		call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
+	}
+	if (ac)
+		kmem_cache_free(ext4_ac_cachep, ac);
+}
+
+/*
+ * We have incremented pa_count. So it cannot be freed at this
+ * point. Also we hold lg_mutex. So no parallel allocation is
+ * possible from this lg. That means pa_free cannot be updated.
+ *
+ * A parallel ext4_mb_discard_group_preallocations is possible.
+ * which can cause the lg_prealloc_list to be updated.
+ */
+
+static void ext4_mb_add_n_trim(struct ext4_allocation_context *ac)
+{
+	int order, added = 0, lg_prealloc_count = 1;
+	struct super_block *sb = ac->ac_sb;
+	struct ext4_locality_group *lg = ac->ac_lg;
+	struct ext4_prealloc_space *tmp_pa, *pa = ac->ac_pa;
+
+	order = fls(pa->pa_free) - 1;
+	if (order > PREALLOC_TB_SIZE - 1)
+		/* The max size of hash table is PREALLOC_TB_SIZE */
+		order = PREALLOC_TB_SIZE - 1;
+	/* Add the prealloc space to lg */
+	rcu_read_lock();
+	list_for_each_entry_rcu(tmp_pa, &lg->lg_prealloc_list[order],
+						pa_inode_list) {
+		spin_lock(&tmp_pa->pa_lock);
+		if (tmp_pa->pa_deleted) {
+			spin_unlock(&pa->pa_lock);
+			continue;
+		}
+		if (!added && pa->pa_free < tmp_pa->pa_free) {
+			/* Add to the tail of the previous entry */
+			list_add_tail_rcu(&pa->pa_inode_list,
+						&tmp_pa->pa_inode_list);
+			added = 1;
+			/*
+			 * we want to count the total
+			 * number of entries in the list
+			 */
+		}
+		spin_unlock(&tmp_pa->pa_lock);
+		lg_prealloc_count++;
+	}
+	if (!added)
+		list_add_tail_rcu(&pa->pa_inode_list,
+					&lg->lg_prealloc_list[order]);
+	rcu_read_unlock();
+
+	/* Now trim the list to be not more than 8 elements */
+	if (lg_prealloc_count > 8) {
+		ext4_mb_discard_lg_preallocations(sb, lg,
+						order, lg_prealloc_count);
+		return;
+	}
+	return ;
+}
+
 /*
  * release all resource we used in allocation
  */
 static int ext4_mb_release_context(struct ext4_allocation_context *ac)
 {
-	if (ac->ac_pa) {
+	struct ext4_prealloc_space *pa = ac->ac_pa;
-		if (ac->ac_pa->pa_linear) {
+	if (pa) {
+		if (pa->pa_linear) {
 			/* see comment in ext4_mb_use_group_pa() */
-			spin_lock(&ac->ac_pa->pa_lock);
+			spin_lock(&pa->pa_lock);
-			ac->ac_pa->pa_pstart += ac->ac_b_ex.fe_len;
+			pa->pa_pstart += ac->ac_b_ex.fe_len;
-			ac->ac_pa->pa_lstart += ac->ac_b_ex.fe_len;
+			pa->pa_lstart += ac->ac_b_ex.fe_len;
-			ac->ac_pa->pa_free -= ac->ac_b_ex.fe_len;
+			pa->pa_free -= ac->ac_b_ex.fe_len;
-			ac->ac_pa->pa_len -= ac->ac_b_ex.fe_len;
+			pa->pa_len -= ac->ac_b_ex.fe_len;
-			spin_unlock(&ac->ac_pa->pa_lock);
+			spin_unlock(&pa->pa_lock);
+			/*
+			 * We want to add the pa to the right bucket.
+			 * Remove it from the list and while adding
+			 * make sure the list to which we are adding
+			 * doesn't grow big.
+			 */
+			if (likely(pa->pa_free)) {
+				spin_lock(pa->pa_obj_lock);
+				list_del_rcu(&pa->pa_inode_list);
+				spin_unlock(pa->pa_obj_lock);
+				ext4_mb_add_n_trim(ac);
+			}
 		}
-		ext4_mb_put_pa(ac, ac->ac_sb, ac->ac_pa);
+		ext4_mb_put_pa(ac, ac->ac_sb, pa);
 	}
 	if (ac->ac_bitmap_page)
 		page_cache_release(ac->ac_bitmap_page);

fs/ext4/mballoc.h

@@ -164,11 +164,17 @@ struct ext4_free_extent {
  * Locality group:
  *   we try to group all related changes together
  *   so that writeback can flush/allocate them together as well
+ *   Size of lg_prealloc_list hash is determined by MB_DEFAULT_GROUP_PREALLOC
+ *   (512). We store prealloc space into the hash based on the pa_free blocks
+ *   order value.ie, fls(pa_free)-1;
  */
+#define PREALLOC_TB_SIZE 10
 struct ext4_locality_group {
 	/* for allocator */
-	struct mutex		lg_mutex;	/* to serialize allocates */
+	/* to serialize allocates */
-	struct list_head	lg_prealloc_list;/* list of preallocations */
+	struct mutex		lg_mutex;
+	/* list of preallocations */
+	struct list_head	lg_prealloc_list[PREALLOC_TB_SIZE];
 	spinlock_t		lg_prealloc_lock;
 };