Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4

Pull ext4 bugfixes from Ted Ts'o: "Fixes for 3.11-rc2, sent at 5pm, in the professoinal style. :-)" I'm not sure I like this new level of "professionalism". 9-5, people, 9-5. * tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: ext4: call ext4_es_lru_add() after handling cache miss ext4: yield during large unlinks ext4: make the extent_status code more robust against ENOMEM failures ext4: simplify calculation of blocks to free on error ext4: fix error handling in ext4_ext_truncate()

Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Pull ext4 bugfixes from Ted Ts'o: "Fixes for 3.11-rc2, sent at 5pm, in the professoinal style. :-)" I'm not sure I like this new level of "professionalism". 9-5, people, 9-5. * tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: ext4: call ext4_es_lru_add() after handling cache miss ext4: yield during large unlinks ext4: make the extent_status code more robust against ENOMEM failures ext4: simplify calculation of blocks to free on error ext4: fix error handling in ext4_ext_truncate()
Linus Torvalds
2 parents 3be542d464 63b999685c
Showing 3 changed files Side-by-side Diff
fs/ext4/extents.c
fs/ext4/extents_status.c
fs/ext4/inode.c
@@ -2835,6 +2835,9 @@
 				err = -EIO;
 				break;
 			}
+			/* Yield here to deal with large extent trees.
+			 * Should be a no-op if we did IO above. */
+			cond_resched();
 			if (WARN_ON(i + 1 > depth)) {
 				err = -EIO;
 				break;
@@ -4261,8 +4264,8 @@
 		/* not a good idea to call discard here directly,
 		 * but otherwise we'd need to call it every free() */
 		ext4_discard_preallocations(inode);
-		ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex),
-				 ext4_ext_get_actual_len(&newex), fb_flags);
+		ext4_free_blocks(handle, inode, NULL, newblock,
+				 EXT4_C2B(sbi, allocated_clusters), fb_flags);
 		goto out2;
 	}
  
@@ -4382,8 +4385,9 @@
 	}
  
 out3:
-	trace_ext4_ext_map_blocks_exit(inode, flags, map, err ? err : allocated);
-
+	trace_ext4_ext_map_blocks_exit(inode, flags, map,
+				       err ? err : allocated);
+	ext4_es_lru_add(inode);
 	return err ? err : allocated;
 }
  
  
  
@@ -4405,9 +4409,20 @@
  
 	last_block = (inode->i_size + sb->s_blocksize - 1)
 			>> EXT4_BLOCK_SIZE_BITS(sb);
+retry:
 	err = ext4_es_remove_extent(inode, last_block,
 				    EXT_MAX_BLOCKS - last_block);
+	if (err == ENOMEM) {
+		cond_resched();
+		congestion_wait(BLK_RW_ASYNC, HZ/50);
+		goto retry;
+	}
+	if (err) {
+		ext4_std_error(inode->i_sb, err);
+		return;
+	}
 	err = ext4_ext_remove_space(inode, last_block, EXT_MAX_BLOCKS - 1);
+	ext4_std_error(inode->i_sb, err);
 }
  
 static void ext4_falloc_update_inode(struct inode *inode,
@@ -148,6 +148,8 @@
 			      ext4_lblk_t end);
 static int __es_try_to_reclaim_extents(struct ext4_inode_info *ei,
 				       int nr_to_scan);
+static int __ext4_es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
+			    struct ext4_inode_info *locked_ei);
  
 int __init ext4_init_es(void)
 {
  
@@ -665,7 +667,13 @@
 	err = __es_remove_extent(inode, lblk, end);
 	if (err != 0)
 		goto error;
+retry:
 	err = __es_insert_extent(inode, &newes);
+	if (err == -ENOMEM && __ext4_es_shrink(EXT4_SB(inode->i_sb), 1,
+					       EXT4_I(inode)))
+		goto retry;
+	if (err == -ENOMEM && !ext4_es_is_delayed(&newes))
+		err = 0;
  
 error:
 	write_unlock(&EXT4_I(inode)->i_es_lock);
  
@@ -744,8 +752,10 @@
 	struct extent_status orig_es;
 	ext4_lblk_t len1, len2;
 	ext4_fsblk_t block;
-	int err = 0;
+	int err;
  
+retry:
+	err = 0;
 	es = __es_tree_search(&tree->root, lblk);
 	if (!es)
 		goto out;
@@ -780,6 +790,10 @@
 			if (err) {
 				es->es_lblk = orig_es.es_lblk;
 				es->es_len = orig_es.es_len;
+				if ((err == -ENOMEM) &&
+				    __ext4_es_shrink(EXT4_SB(inode->i_sb), 1,
+						     EXT4_I(inode)))
+					goto retry;
 				goto out;
 			}
 		} else {
  
  
  
@@ -889,22 +903,14 @@
 		return -1;
 }
  
-static int ext4_es_shrink(struct shrinker *shrink, struct shrink_control *sc)
+static int __ext4_es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
+			    struct ext4_inode_info *locked_ei)
 {
-	struct ext4_sb_info *sbi = container_of(shrink,
-					struct ext4_sb_info, s_es_shrinker);
 	struct ext4_inode_info *ei;
 	struct list_head *cur, *tmp;
 	LIST_HEAD(skiped);
-	int nr_to_scan = sc->nr_to_scan;
 	int ret, nr_shrunk = 0;
  
-	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
-	trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan, ret);
-
-	if (!nr_to_scan)
-		return ret;
-
 	spin_lock(&sbi->s_es_lru_lock);
  
 	/*
@@ -933,7 +939,7 @@
 			continue;
 		}
  
-		if (ei->i_es_lru_nr == 0)
+		if (ei->i_es_lru_nr == 0 || ei == locked_ei)
 			continue;
  
 		write_lock(&ei->i_es_lock);
@@ -951,6 +957,27 @@
 	/* Move the newer inodes into the tail of the LRU list. */
 	list_splice_tail(&skiped, &sbi->s_es_lru);
 	spin_unlock(&sbi->s_es_lru_lock);
+
+	if (locked_ei && nr_shrunk == 0)
+		nr_shrunk = __es_try_to_reclaim_extents(ei, nr_to_scan);
+
+	return nr_shrunk;
+}
+
+static int ext4_es_shrink(struct shrinker *shrink, struct shrink_control *sc)
+{
+	struct ext4_sb_info *sbi = container_of(shrink,
+					struct ext4_sb_info, s_es_shrinker);
+	int nr_to_scan = sc->nr_to_scan;
+	int ret, nr_shrunk;
+
+	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
+	trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan, ret);
+
+	if (!nr_to_scan)
+		return ret;
+
+	nr_shrunk = __ext4_es_shrink(sbi, nr_to_scan, NULL);
  
 	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
 	trace_ext4_es_shrink_exit(sbi->s_sb, nr_shrunk, ret);
@@ -514,10 +514,9 @@
 		  "logical block %lu\n", inode->i_ino, flags, map->m_len,
 		  (unsigned long) map->m_lblk);
  
-	ext4_es_lru_add(inode);
-
 	/* Lookup extent status tree firstly */
 	if (ext4_es_lookup_extent(inode, map->m_lblk, &es)) {
+		ext4_es_lru_add(inode);
 		if (ext4_es_is_written(&es) || ext4_es_is_unwritten(&es)) {
 			map->m_pblk = ext4_es_pblock(&es) +
 					map->m_lblk - es.es_lblk;
  
@@ -1529,11 +1528,9 @@
 		  "logical block %lu\n", inode->i_ino, map->m_len,
 		  (unsigned long) map->m_lblk);
  
-	ext4_es_lru_add(inode);
-
 	/* Lookup extent status tree firstly */
 	if (ext4_es_lookup_extent(inode, iblock, &es)) {
-
+		ext4_es_lru_add(inode);
 		if (ext4_es_is_hole(&es)) {
 			retval = 0;
 			down_read((&EXT4_I(inode)->i_data_sem));
...	...	@@ -2835,6 +2835,9 @@
2835	2835	err = -EIO;
2836	2836	break;
2837	2837	}
	2838	+ /* Yield here to deal with large extent trees.
	2839	+ * Should be a no-op if we did IO above. */
	2840	+ cond_resched();
2838	2841	if (WARN_ON(i + 1 > depth)) {
2839	2842	err = -EIO;
2840	2843	break;
...	...	@@ -4261,8 +4264,8 @@
4261	4264	/* not a good idea to call discard here directly,
4262	4265	* but otherwise we'd need to call it every free() */
4263	4266	ext4_discard_preallocations(inode);
4264		- ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex),
4265		- ext4_ext_get_actual_len(&newex), fb_flags);
	4267	+ ext4_free_blocks(handle, inode, NULL, newblock,
	4268	+ EXT4_C2B(sbi, allocated_clusters), fb_flags);
4266	4269	goto out2;
4267	4270	}
4268	4271
...	...	@@ -4382,8 +4385,9 @@
4382	4385	}
4383	4386
4384	4387	out3:
4385		- trace_ext4_ext_map_blocks_exit(inode, flags, map, err ? err : allocated);
4386		-
	4388	+ trace_ext4_ext_map_blocks_exit(inode, flags, map,
	4389	+ err ? err : allocated);
	4390	+ ext4_es_lru_add(inode);
4387	4391	return err ? err : allocated;
4388	4392	}
4389	4393
4390	4394
4391	4395
...	...	@@ -4405,9 +4409,20 @@
4405	4409
4406	4410	last_block = (inode->i_size + sb->s_blocksize - 1)
4407	4411	>> EXT4_BLOCK_SIZE_BITS(sb);
	4412	+retry:
4408	4413	err = ext4_es_remove_extent(inode, last_block,
4409	4414	EXT_MAX_BLOCKS - last_block);
	4415	+ if (err == ENOMEM) {
	4416	+ cond_resched();
	4417	+ congestion_wait(BLK_RW_ASYNC, HZ/50);
	4418	+ goto retry;
	4419	+ }
	4420	+ if (err) {
	4421	+ ext4_std_error(inode->i_sb, err);
	4422	+ return;
	4423	+ }
4410	4424	err = ext4_ext_remove_space(inode, last_block, EXT_MAX_BLOCKS - 1);
	4425	+ ext4_std_error(inode->i_sb, err);
4411	4426	}
4412	4427
4413	4428	static void ext4_falloc_update_inode(struct inode *inode,
...	...	@@ -148,6 +148,8 @@
148	148	ext4_lblk_t end);
149	149	static int __es_try_to_reclaim_extents(struct ext4_inode_info *ei,
150	150	int nr_to_scan);
	151	+static int __ext4_es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
	152	+ struct ext4_inode_info *locked_ei);
151	153
152	154	int __init ext4_init_es(void)
153	155	{
154	156
...	...	@@ -665,7 +667,13 @@
665	667	err = __es_remove_extent(inode, lblk, end);
666	668	if (err != 0)
667	669	goto error;
	670	+retry:
668	671	err = __es_insert_extent(inode, &newes);
	672	+ if (err == -ENOMEM && __ext4_es_shrink(EXT4_SB(inode->i_sb), 1,
	673	+ EXT4_I(inode)))
	674	+ goto retry;
	675	+ if (err == -ENOMEM && !ext4_es_is_delayed(&newes))
	676	+ err = 0;
669	677
670	678	error:
671	679	write_unlock(&EXT4_I(inode)->i_es_lock);
672	680
...	...	@@ -744,8 +752,10 @@
744	752	struct extent_status orig_es;
745	753	ext4_lblk_t len1, len2;
746	754	ext4_fsblk_t block;
747		- int err = 0;
	755	+ int err;
748	756
	757	+retry:
	758	+ err = 0;
749	759	es = __es_tree_search(&tree->root, lblk);
750	760	if (!es)
751	761	goto out;
...	...	@@ -780,6 +790,10 @@
780	790	if (err) {
781	791	es->es_lblk = orig_es.es_lblk;
782	792	es->es_len = orig_es.es_len;
	793	+ if ((err == -ENOMEM) &&
	794	+ __ext4_es_shrink(EXT4_SB(inode->i_sb), 1,
	795	+ EXT4_I(inode)))
	796	+ goto retry;
783	797	goto out;
784	798	}
785	799	} else {
786	800
787	801
788	802
...	...	@@ -889,22 +903,14 @@
889	903	return -1;
890	904	}
891	905
892		-static int ext4_es_shrink(struct shrinker shrink, struct shrink_control sc)
	906	+static int __ext4_es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
	907	+ struct ext4_inode_info *locked_ei)
893	908	{
894		- struct ext4_sb_info *sbi = container_of(shrink,
895		- struct ext4_sb_info, s_es_shrinker);
896	909	struct ext4_inode_info *ei;
897	910	struct list_head cur, tmp;
898	911	LIST_HEAD(skiped);
899		- int nr_to_scan = sc->nr_to_scan;
900	912	int ret, nr_shrunk = 0;
901	913
902		- ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
903		- trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan, ret);
904		-
905		- if (!nr_to_scan)
906		- return ret;
907		-
908	914	spin_lock(&sbi->s_es_lru_lock);
909	915
910	916	/*
...	...	@@ -933,7 +939,7 @@
933	939	continue;
934	940	}
935	941
936		- if (ei->i_es_lru_nr == 0)
	942	+ if (ei->i_es_lru_nr == 0 \|\| ei == locked_ei)
937	943	continue;
938	944
939	945	write_lock(&ei->i_es_lock);
...	...	@@ -951,6 +957,27 @@
951	957	/* Move the newer inodes into the tail of the LRU list. */
952	958	list_splice_tail(&skiped, &sbi->s_es_lru);
953	959	spin_unlock(&sbi->s_es_lru_lock);
	960	+
	961	+ if (locked_ei && nr_shrunk == 0)
	962	+ nr_shrunk = __es_try_to_reclaim_extents(ei, nr_to_scan);
	963	+
	964	+ return nr_shrunk;
	965	+}
	966	+
	967	+static int ext4_es_shrink(struct shrinker shrink, struct shrink_control sc)
	968	+{
	969	+ struct ext4_sb_info *sbi = container_of(shrink,
	970	+ struct ext4_sb_info, s_es_shrinker);
	971	+ int nr_to_scan = sc->nr_to_scan;
	972	+ int ret, nr_shrunk;
	973	+
	974	+ ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
	975	+ trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan, ret);
	976	+
	977	+ if (!nr_to_scan)
	978	+ return ret;
	979	+
	980	+ nr_shrunk = __ext4_es_shrink(sbi, nr_to_scan, NULL);
954	981
955	982	ret = percpu_counter_read_positive(&sbi->s_extent_cache_cnt);
956	983	trace_ext4_es_shrink_exit(sbi->s_sb, nr_shrunk, ret);
...	...	@@ -514,10 +514,9 @@
514	514	"logical block %lu\n", inode->i_ino, flags, map->m_len,
515	515	(unsigned long) map->m_lblk);
516	516
517		- ext4_es_lru_add(inode);
518		-
519	517	/* Lookup extent status tree firstly */
520	518	if (ext4_es_lookup_extent(inode, map->m_lblk, &es)) {
	519	+ ext4_es_lru_add(inode);
521	520	if (ext4_es_is_written(&es) \|\| ext4_es_is_unwritten(&es)) {
522	521	map->m_pblk = ext4_es_pblock(&es) +
523	522	map->m_lblk - es.es_lblk;
524	523
...	...	@@ -1529,11 +1528,9 @@
1529	1528	"logical block %lu\n", inode->i_ino, map->m_len,
1530	1529	(unsigned long) map->m_lblk);
1531	1530
1532		- ext4_es_lru_add(inode);
1533		-
1534	1531	/* Lookup extent status tree firstly */
1535	1532	if (ext4_es_lookup_extent(inode, iblock, &es)) {
1536		-
	1533	+ ext4_es_lru_add(inode);
1537	1534	if (ext4_es_is_hole(&es)) {
1538	1535	retval = 0;
1539	1536	down_read((&EXT4_I(inode)->i_data_sem));