[PATCH 0/4] ext3[34] barrier changes

[PATCH 0/4] ext3[34] barrier changes

[Eric Sandeen]

A collection of patches to make ext3 & 4 use barriers by
default, and to call blkdev_issue_flush on fsync if they
are enabled.

[PATCH 1/4] ext3: enable barriers by default

[Eric Sandeen]

I can't think of any valid reason for ext3 to not use barriers when
they are available;  I believe this is necessary for filesystem
integrity in the face of a volatile write cache on storage.

An administrator who trusts that the cache is sufficiently battery-
backed (and power supplies are sufficiently redundant, etc...)
can always turn it back off again.

SuSE has carried such a patch for quite some time now.

Also document the mount option while we're at it.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Acked-by: Jan Kara <jack@suse.cz>
---
 Documentation/filesystems/ext3.txt |   12 ++++++++++--
 fs/ext3/super.c                    |   11 +++++++++--
 2 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/Documentation/filesystems/ext3.txt b/Documentation/filesystems/ext3.txt
index b45f3c1..daab1f5 100644
--- a/Documentation/filesystems/ext3.txt
+++ b/Documentation/filesystems/ext3.txt
@@ -52,8 +52,16 @@ commit=nrsec	(*)	Ext3 can be told to sync all its data and metadata
 			Setting it to very large values will improve
 			performance.
 
-barrier=1		This enables/disables barriers.  barrier=0 disables
-			it, barrier=1 enables it.
+barrier=<0|1(*)>	This enables/disables the use of write barriers in
+			the jbd code.  barrier=0 disables, barrier=1 enables.
+			This also requires an IO stack which can support
+			barriers, and if jbd gets an error on a barrier
+			write, it will disable again with a warning.
+			Write barriers enforce proper on-disk ordering
+			of journal commits, making volatile disk write caches
+			safe to use, at some performance penalty.  If
+			your disks are battery-backed in one way or another,
+			disabling barriers may safely improve performance.
 
 orlov		(*)	This enables the new Orlov block allocator. It is
 			enabled by default.
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index fe3119a..9c30dc7 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -555,6 +555,7 @@ static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
 	struct super_block *sb = vfs->mnt_sb;
 	struct ext3_sb_info *sbi = EXT3_SB(sb);
 	struct ext3_super_block *es = sbi->s_es;
+	journal_t *journal = sbi->s_journal;
 	unsigned long def_mount_opts;
 
 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
@@ -613,8 +614,13 @@ static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_printf(seq, ",commit=%u",
 			   (unsigned) (sbi->s_commit_interval / HZ));
 	}
-	if (test_opt(sb, BARRIER))
-		seq_puts(seq, ",barrier=1");
+	/*
+ 	 * jbd inherits the barrier flag from ext3, and may actually
+ 	 * turn off barriers if a write fails, so it's the real test.
+ 	 */
+	if (!test_opt(sb, BARRIER) ||
+	    (journal && !(journal->j_flags & JFS_BARRIER)))
+		seq_puts(seq, ",barrier=0");
 	if (test_opt(sb, NOBH))
 		seq_puts(seq, ",nobh");
 
@@ -1589,6 +1595,7 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 
 	set_opt(sbi->s_mount_opt, RESERVATION);
+	set_opt(sbi->s_mount_opt, BARRIER);
 
 	if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
 			    NULL, 0))
-- 
1.5.3.6

[PATCH 2/4] ext3: call blkdev_issue_flush on fsync

[Eric Sandeen]

To ensure that bits are truly on-disk after an fsync,
ext3 should call blkdev_issue_flush if barriers are supported.

Inspired by an old thread on barriers, by reiserfs & xfs
which do the same, and by a patch SuSE ships with their kernel

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---
 fs/ext3/fsync.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/fs/ext3/fsync.c b/fs/ext3/fsync.c
index d336341..f6167ec 100644
--- a/fs/ext3/fsync.c
+++ b/fs/ext3/fsync.c
@@ -27,6 +27,7 @@
 #include <linux/sched.h>
 #include <linux/writeback.h>
 #include <linux/jbd.h>
+#include <linux/blkdev.h>
 #include <linux/ext3_fs.h>
 #include <linux/ext3_jbd.h>
 
@@ -45,6 +46,7 @@
 int ext3_sync_file(struct file * file, struct dentry *dentry, int datasync)
 {
 	struct inode *inode = dentry->d_inode;
+	journal_t *journal = EXT3_SB(inode->i_sb)->s_journal;
 	int ret = 0;
 
 	J_ASSERT(ext3_journal_current_handle() == NULL);
@@ -85,6 +87,9 @@ int ext3_sync_file(struct file * file, struct dentry *dentry, int datasync)
 			.nr_to_write = 0, /* sys_fsync did this */
 		};
 		ret = sync_inode(inode, &wbc);
+
+		if (journal && (journal->j_flags & JFS_BARRIER))
+			blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
 	}
 out:
 	return ret;
-- 
1.5.3.6

[PATCH 3/4] ext4: enable barriers by default

[Eric Sandeen]

I can't think of any valid reason for ext4 to not use barriers when
they are available;  I believe this is necessary for filesystem
integrity in the face of a volatile write cache on storage.

An administrator who trusts that the cache is sufficiently battery-
backed (and power supplies are sufficiently redundant, etc...)
can always turn it back off again.

SuSE has carried such a patch for ext3 for quite some time now.

Also document the mount option while we're at it.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---
 Documentation/filesystems/ext4.txt |   12 ++++++++++--
 fs/ext4/super.c                    |   11 +++++++++--
 2 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/Documentation/filesystems/ext4.txt b/Documentation/filesystems/ext4.txt
index 560f88d..0c5086d 100644
--- a/Documentation/filesystems/ext4.txt
+++ b/Documentation/filesystems/ext4.txt
@@ -139,8 +139,16 @@ commit=nrsec	(*)	Ext4 can be told to sync all its data and metadata
 			Setting it to very large values will improve
 			performance.
 
-barrier=1		This enables/disables barriers.  barrier=0 disables
-			it, barrier=1 enables it.
+barrier=<0|1(*)>	This enables/disables the use of write barriers in
+			the jbd code.  barrier=0 disables, barrier=1 enables.
+			This also requires an IO stack which can support
+			barriers, and if jbd gets an error on a barrier
+			write, it will disable again with a warning.
+			Write barriers enforce proper on-disk ordering
+			of journal commits, making volatile disk write caches
+			safe to use, at some performance penalty.  If
+			your disks are battery-backed in one way or another,
+			disabling barriers may safely improve performance.
 
 orlov		(*)	This enables the new Orlov block allocator. It is
 			enabled by default.
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 52dd067..77b036a 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -671,6 +671,7 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
 	unsigned long def_mount_opts;
 	struct super_block *sb = vfs->mnt_sb;
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	journal_t *journal = sbi->s_journal;
 	struct ext4_super_block *es = sbi->s_es;
 
 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
@@ -729,8 +730,13 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_printf(seq, ",commit=%u",
 			   (unsigned) (sbi->s_commit_interval / HZ));
 	}
-	if (test_opt(sb, BARRIER))
-		seq_puts(seq, ",barrier=1");
+	/*
+	 * jbd2 inherits the barrier flag from ext4, and may actually
+	 * turn off barriers if a write fails, so it's the real test.
+	 */
+	if (!test_opt(sb, BARRIER) ||
+	    (journal && !(journal->j_flags & JBD2_BARRIER)))
+		seq_puts(seq, ",barrier=0");
 	if (test_opt(sb, NOBH))
 		seq_puts(seq, ",nobh");
 	if (!test_opt(sb, EXTENTS))
@@ -1890,6 +1896,7 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
 	sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
 
 	set_opt(sbi->s_mount_opt, RESERVATION);
+	set_opt(sbi->s_mount_opt, BARRIER);
 
 	/*
 	 * turn on extents feature by default in ext4 filesystem
-- 
1.5.3.6

[PATCH 4/4] ext4: call blkdev_issue_flush on fsync

[Eric Sandeen]

To ensure that bits are truly on-disk after an fsync,
we should call blkdev_issue_flush if barriers are supported.

Inspired by an old thread on barriers, by reiserfs & xfs
which do the same, and by a patch SuSE ships with their kernel

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---
 fs/ext4/fsync.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
index 1c8ba48..a45c373 100644
--- a/fs/ext4/fsync.c
+++ b/fs/ext4/fsync.c
@@ -27,6 +27,7 @@
 #include <linux/sched.h>
 #include <linux/writeback.h>
 #include <linux/jbd2.h>
+#include <linux/blkdev.h>
 #include "ext4.h"
 #include "ext4_jbd2.h"
 
@@ -45,6 +46,7 @@
 int ext4_sync_file(struct file * file, struct dentry *dentry, int datasync)
 {
 	struct inode *inode = dentry->d_inode;
+	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
 	int ret = 0;
 
 	J_ASSERT(ext4_journal_current_handle() == NULL);
@@ -85,6 +87,8 @@ int ext4_sync_file(struct file * file, struct dentry *dentry, int datasync)
 			.nr_to_write = 0, /* sys_fsync did this */
 		};
 		ret = sync_inode(inode, &wbc);
+		if (journal && (journal->j_flags & JBD2_BARRIER))
+			blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
 	}
 out:
 	return ret;
-- 
1.5.3.6