[PATCH 3.10 152/319] ext4: use __GFP_NOFAIL in ext4_free_blocks()

[Willy Tarreau]

From: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>

commit adb7ef600cc9d9d15ecc934cc26af5c1379777df upstream.

This might be unexpected but pages allocated for sbi->s_buddy_cache are
charged to current memory cgroup. So, GFP_NOFS allocation could fail if
current task has been killed by OOM or if current memory cgroup has no
free memory left. Block allocator cannot handle such failures here yet.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx>
Signed-off-by: Theodore Ts'o <tytso@xxxxxxx>
Signed-off-by: Willy Tarreau <w@xxxxxx>
---
 fs/ext4/mballoc.c | 47 ++++++++++++++++++++++++++++-------------------
 1 file changed, 28 insertions(+), 19 deletions(-)

diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 08b4495..cb9eec0 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -808,7 +808,7 @@ static void mb_regenerate_buddy(struct ext4_buddy *e4b)
  * for this page; do not hold this lock when calling this routine!
  */
 
-static int ext4_mb_init_cache(struct page *page, char *incore)
+static int ext4_mb_init_cache(struct page *page, char *incore, gfp_t gfp)
 {
 	ext4_group_t ngroups;
 	int blocksize;
@@ -841,7 +841,7 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
 	/* allocate buffer_heads to read bitmaps */
 	if (groups_per_page > 1) {
 		i = sizeof(struct buffer_head *) * groups_per_page;
-		bh = kzalloc(i, GFP_NOFS);
+		bh = kzalloc(i, gfp);
 		if (bh == NULL) {
 			err = -ENOMEM;
 			goto out;
@@ -966,7 +966,7 @@ out:
  * are on the same page e4b->bd_buddy_page is NULL and return value is 0.
  */
 static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
-		ext4_group_t group, struct ext4_buddy *e4b)
+		ext4_group_t group, struct ext4_buddy *e4b, gfp_t gfp)
 {
 	struct inode *inode = EXT4_SB(sb)->s_buddy_cache;
 	int block, pnum, poff;
@@ -985,7 +985,7 @@ static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
 	block = group * 2;
 	pnum = block / blocks_per_page;
 	poff = block % blocks_per_page;
-	page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+	page = find_or_create_page(inode->i_mapping, pnum, gfp);
 	if (!page)
 		return -EIO;
 	BUG_ON(page->mapping != inode->i_mapping);
@@ -999,7 +999,7 @@ static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
 
 	block++;
 	pnum = block / blocks_per_page;
-	page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+	page = find_or_create_page(inode->i_mapping, pnum, gfp);
 	if (!page)
 		return -EIO;
 	BUG_ON(page->mapping != inode->i_mapping);
@@ -1025,7 +1025,7 @@ static void ext4_mb_put_buddy_page_lock(struct ext4_buddy *e4b)
  * calling this routine!
  */
 static noinline_for_stack
-int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
+int ext4_mb_init_group(struct super_block *sb, ext4_group_t group, gfp_t gfp)
 {
 
 	struct ext4_group_info *this_grp;
@@ -1043,7 +1043,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	 * have taken a reference using ext4_mb_load_buddy and that
 	 * would have pinned buddy page to page cache.
 	 */
-	ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b);
+	ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b, gfp);
 	if (ret || !EXT4_MB_GRP_NEED_INIT(this_grp)) {
 		/*
 		 * somebody initialized the group
@@ -1053,7 +1053,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	}
 
 	page = e4b.bd_bitmap_page;
-	ret = ext4_mb_init_cache(page, NULL);
+	ret = ext4_mb_init_cache(page, NULL, gfp);
 	if (ret)
 		goto err;
 	if (!PageUptodate(page)) {
@@ -1073,7 +1073,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	}
 	/* init buddy cache */
 	page = e4b.bd_buddy_page;
-	ret = ext4_mb_init_cache(page, e4b.bd_bitmap);
+	ret = ext4_mb_init_cache(page, e4b.bd_bitmap, gfp);
 	if (ret)
 		goto err;
 	if (!PageUptodate(page)) {
@@ -1092,8 +1092,8 @@ err:
  * calling this routine!
  */
 static noinline_for_stack int
-ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
-					struct ext4_buddy *e4b)
+ext4_mb_load_buddy_gfp(struct super_block *sb, ext4_group_t group,
+		       struct ext4_buddy *e4b, gfp_t gfp)
 {
 	int blocks_per_page;
 	int block;
@@ -1123,7 +1123,7 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 		 * we need full data about the group
 		 * to make a good selection
 		 */
-		ret = ext4_mb_init_group(sb, group);
+		ret = ext4_mb_init_group(sb, group, gfp);
 		if (ret)
 			return ret;
 	}
@@ -1151,11 +1151,11 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 			 * wait for it to initialize.
 			 */
 			page_cache_release(page);
-		page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+		page = find_or_create_page(inode->i_mapping, pnum, gfp);
 		if (page) {
 			BUG_ON(page->mapping != inode->i_mapping);
 			if (!PageUptodate(page)) {
-				ret = ext4_mb_init_cache(page, NULL);
+				ret = ext4_mb_init_cache(page, NULL, gfp);
 				if (ret) {
 					unlock_page(page);
 					goto err;
@@ -1182,11 +1182,12 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 	if (page == NULL || !PageUptodate(page)) {
 		if (page)
 			page_cache_release(page);
-		page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+		page = find_or_create_page(inode->i_mapping, pnum, gfp);
 		if (page) {
 			BUG_ON(page->mapping != inode->i_mapping);
 			if (!PageUptodate(page)) {
-				ret = ext4_mb_init_cache(page, e4b->bd_bitmap);
+				ret = ext4_mb_init_cache(page, e4b->bd_bitmap,
+							 gfp);
 				if (ret) {
 					unlock_page(page);
 					goto err;
@@ -1220,6 +1221,12 @@ err:
 	return ret;
 }
 
+static int ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
+			      struct ext4_buddy *e4b)
+{
+	return ext4_mb_load_buddy_gfp(sb, group, e4b, GFP_NOFS);
+}
+
 static void ext4_mb_unload_buddy(struct ext4_buddy *e4b)
 {
 	if (e4b->bd_bitmap_page)
@@ -1993,7 +2000,7 @@ static int ext4_mb_good_group(struct ext4_allocation_context *ac,
 
 	/* We only do this if the grp has never been initialized */
 	if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
-		int ret = ext4_mb_init_group(ac->ac_sb, group);
+		int ret = ext4_mb_init_group(ac->ac_sb, group, GFP_NOFS);
 		if (ret)
 			return 0;
 	}
@@ -4748,7 +4755,9 @@ do_more:
 #endif
 	trace_ext4_mballoc_free(sb, inode, block_group, bit, count_clusters);
 
-	err = ext4_mb_load_buddy(sb, block_group, &e4b);
+	/* __GFP_NOFAIL: retry infinitely, ignore TIF_MEMDIE and memcg limit. */
+	err = ext4_mb_load_buddy_gfp(sb, block_group, &e4b,
+				     GFP_NOFS|__GFP_NOFAIL);
 	if (err)
 		goto error_return;
 
@@ -5159,7 +5168,7 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
 		grp = ext4_get_group_info(sb, group);
 		/* We only do this if the grp has never been initialized */
 		if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
-			ret = ext4_mb_init_group(sb, group);
+			ret = ext4_mb_init_group(sb, group, GFP_NOFS);
 			if (ret)
 				break;
 		}
-- 
2.8.0.rc2.1.gbe9624a