ext4: Make fsync to sync parent dir in no-journal for real this time

PATCH Applied Lukas Czerner 3 participants 4 messages Feb 23, 2015 – Apr 3, 2015 ext4 file system

Share & tools

Lukas Czerner

author

2015-02-23 13:16 UTC #0 lore raw

PATCH 1/1 ext4: Make fsync to sync parent dir in no-journal for real this time

Previously commit 14ece1028b3ed53ffec1b1213ffc6acaf79ad77c added a support for for syncing parent directory of newly created inodes to make sure that the inode is not lost after a power failure in no-journal mode.

However this does not work in majority of cases, namely:
- if the directory has inline data
- if the directory is already indexed
- if the directory already has at least one block and:
- the new entry fits into it
- or we've successfully converted it to indexed

So in those cases we might lose the inode entirely even after fsync in the no-journal mode. This also includes ext2 default mode obviously.

I've noticed this while running xfstest generic/321 and even though the test should fail (we need to run fsck after a crash in no-journal mode) I could not find a newly created entries even when if it was fsynced before.

Fix this by adjusting the ext4_add_entry() successful exit paths to set the inode EXT4_STATE_NEWENTRY so that fsync has the chance to fsync the parent directory as well.

Signed-off-by Lukas Czerner <[email protected]>

Cc Frank Mayhar <[email protected]>

---
 fs/ext4/namei.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

fs/ext4/namei.c +10 -9

index 2291923..25e1e03 100644
@@ -1865,7 +1865,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			  struct inode *inode)
 {
 	struct inode *dir = dentry->d_parent->d_inode;
-	struct buffer_head *bh;
+	struct buffer_head *bh = NULL;
 	struct ext4_dir_entry_2 *de;
 	struct ext4_dir_entry_tail *t;
 	struct super_block *sb;

@@ -1889,14 +1889,14 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return retval;
 		if (retval == 1) {
 			retval = 0;
-			return retval;
+			goto out;
 		}
 	}
 
 	if (is_dx(dir)) {
 		retval = ext4_dx_add_entry(handle, dentry, inode);
 		if (!retval || (retval != ERR_BAD_DX_DIR))
-			return retval;
+			goto out;
 		ext4_clear_inode_flag(dir, EXT4_INODE_INDEX);
 		dx_fallback++;
 		ext4_mark_inode_dirty(handle, dir);

@@ -1908,14 +1908,14 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return PTR_ERR(bh);
 
 		retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
-		if (retval != -ENOSPC) {
-			brelse(bh);
-			return retval;
-		}
+		if (retval != -ENOSPC)
+			goto out;
 
 		if (blocks == 1 && !dx_fallback &&
-		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX))
-			return make_indexed_dir(handle, dentry, inode, bh);
+		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
+			retval = make_indexed_dir(handle, dentry, inode, bh);
+			goto out;
+		}
 		brelse(bh);
 	}
 	bh = ext4_append(handle, dir, &block);

@@ -1931,6 +1931,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 	}
 
 	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
+out:
 	brelse(bh);
 	if (retval == 0)
 		ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY);

Signature · 1 line

1.8.3.1

Jan Kara

reviewer Reviewed-by on v1 1/1

2015-03-10 18:29 UTC #1 lore raw

Quoted 20 lines · Lukas Czerner

Previously commit 14ece1028b3ed53ffec1b1213ffc6acaf79ad77c added a support for for syncing parent directory of newly created inodes to make sure that the inode is not lost after a power failure in no-journal mode.
However this does not work in majority of cases, namely:
- if the directory has inline data
- if the directory is already indexed
- if the directory already has at least one block and:
- the new entry fits into it
- or we've successfully converted it to indexed
So in those cases we might lose the inode entirely even after fsync in the no-journal mode. This also includes ext2 default mode obviously.
I've noticed this while running xfstest generic/321 and even though the test should fail (we need to run fsck after a crash in no-journal mode) I could not find a newly created entries even when if it was fsynced before.

Quoted 6 lines

Fix this by adjusting the ext4_add_entry() successful exit paths to set the inode EXT4_STATE_NEWENTRY so that fsync has the chance to fsync the parent directory as well.
Signed-off-by Lukas Czerner <[email protected]>
Cc Frank Mayhar <[email protected]>

Looks good. You can add:

Reviewed-by Jan Kara <[email protected]>

Honza

Quoted 70 lines

---
 fs/ext4/namei.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

fs/ext4/namei.c +10 -9

index 2291923..25e1e03 100644
@@ -1865,7 +1865,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			  struct inode *inode)
 {
 	struct inode *dir = dentry->d_parent->d_inode;
-	struct buffer_head *bh;
+	struct buffer_head *bh = NULL;
 	struct ext4_dir_entry_2 *de;
 	struct ext4_dir_entry_tail *t;
 	struct super_block *sb;

@@ -1889,14 +1889,14 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return retval;
 		if (retval == 1) {
 			retval = 0;
-			return retval;
+			goto out;
 		}
 	}
 
 	if (is_dx(dir)) {
 		retval = ext4_dx_add_entry(handle, dentry, inode);
 		if (!retval || (retval != ERR_BAD_DX_DIR))
-			return retval;
+			goto out;
 		ext4_clear_inode_flag(dir, EXT4_INODE_INDEX);
 		dx_fallback++;
 		ext4_mark_inode_dirty(handle, dir);

@@ -1908,14 +1908,14 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return PTR_ERR(bh);
 
 		retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
-		if (retval != -ENOSPC) {
-			brelse(bh);
-			return retval;
-		}
+		if (retval != -ENOSPC)
+			goto out;
 
 		if (blocks == 1 && !dx_fallback &&
-		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX))
-			return make_indexed_dir(handle, dentry, inode, bh);
+		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
+			retval = make_indexed_dir(handle, dentry, inode, bh);
+			goto out;
+		}
 		brelse(bh);
 	}
 	bh = ext4_append(handle, dir, &block);

@@ -1931,6 +1931,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 	}
 
 	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
+out:
 	brelse(bh);
 	if (retval == 0)
 		ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY);

Signature · 6 lines

1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Theodore Ts'o

maintainer Applied

2015-04-03 03:45 UTC #2 lore raw

In reply to Lukas Czerner

Quoted 20 lines · Lukas Czerner

Previously commit 14ece1028b3ed53ffec1b1213ffc6acaf79ad77c added a support for for syncing parent directory of newly created inodes to make sure that the inode is not lost after a power failure in no-journal mode.
However this does not work in majority of cases, namely:
- if the directory has inline data
- if the directory is already indexed
- if the directory already has at least one block and:
- the new entry fits into it
- or we've successfully converted it to indexed
So in those cases we might lose the inode entirely even after fsync in the no-journal mode. This also includes ext2 default mode obviously.
I've noticed this while running xfstest generic/321 and even though the test should fail (we need to run fsck after a crash in no-journal mode) I could not find a newly created entries even when if it was fsynced before.

Quoted 6 lines

Fix this by adjusting the ext4_add_entry() successful exit paths to set the inode EXT4_STATE_NEWENTRY so that fsync has the chance to fsync the parent directory as well.
Signed-off-by Lukas Czerner <[email protected]>
Cc Frank Mayhar <[email protected]>

Thanks, applied.

- Ted

Theodore Ts'o

maintainer

2015-04-03 15:16 UTC #3 lore raw

PATCH 1/1 ext4: make fsync to sync parent dir in no-journal for real this time

In reply to Lukas Czerner

It turns out this patch was buggy; make_indexed_dir() releases the buffer head, so it was missing a "bh = NULL" in a critical place. (Note: please try running at least "kvm-xfstests smoke" before submitting a patch; it only takes about 30 minutes, and in this case generic/006 would have failed very quickly.)

- Ted

Quoted 1 line

From e12fb97222fc41e8442896934f76d39ef99b590a Mon Sep 17 00:00:00 2001

From: Lukas Czerner <[email protected]>
Date: Fri, 3 Apr 2015 10:46:58 -0400
Subject: [PATCH] ext4: make fsync to sync parent dir in no-journal for real
this time

So in those cases we might lose the inode entirely even after fsync in the no-journal mode. This also includes ext2 default mode obviously.

Fix this by adjusting the ext4_add_entry() successful exit paths to set the inode EXT4_STATE_NEWENTRY so that fsync has the chance to fsync the parent directory as well.

Signed-off-by Lukas Czerner <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Reviewed-by Jan Kara <[email protected]>

Cc Frank Mayhar <[email protected]>

Cc [email protected]

---
 fs/ext4/namei.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)

fs/ext4/namei.c +11 -9

index efb64ae..23a0b9b 100644
@@ -1864,7 +1864,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			  struct inode *inode)
 {
 	struct inode *dir = dentry->d_parent->d_inode;
-	struct buffer_head *bh;
+	struct buffer_head *bh = NULL;
 	struct ext4_dir_entry_2 *de;
 	struct ext4_dir_entry_tail *t;
 	struct super_block *sb;

@@ -1888,14 +1888,14 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return retval;
 		if (retval == 1) {
 			retval = 0;
-			return retval;
+			goto out;
 		}
 	}
 
 	if (is_dx(dir)) {
 		retval = ext4_dx_add_entry(handle, dentry, inode);
 		if (!retval || (retval != ERR_BAD_DX_DIR))
-			return retval;
+			goto out;
 		ext4_clear_inode_flag(dir, EXT4_INODE_INDEX);
 		dx_fallback++;
 		ext4_mark_inode_dirty(handle, dir);

@@ -1907,14 +1907,15 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 			return PTR_ERR(bh);
 
 		retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh);
-		if (retval != -ENOSPC) {
-			brelse(bh);
-			return retval;
-		}
+		if (retval != -ENOSPC)
+			goto out;
 
 		if (blocks == 1 && !dx_fallback &&
-		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX))
-			return make_indexed_dir(handle, dentry, inode, bh);
+		    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
+			retval = make_indexed_dir(handle, dentry, inode, bh);
+			bh = NULL; /* make_indexed_dir releases bh */
+			goto out;
+		}
 		brelse(bh);
 	}
 	bh = ext4_append(handle, dir, &block);

@@ -1930,6 +1931,7 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 	}
 
 	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
+out:
 	brelse(bh);
 	if (retval == 0)
 		ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY);

Signature · 1 line

2.3.0