[PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[Qu Wenruo]

'Norecovery' mount option is almost a standard for all other
filesystems, including f2fs, ext4, gfs2, nilfs2, and xfs.
And since current 'recovery' mount option for btrfs is too generic, it's
better to change btrfs 'recovery' to 'usebackuproot' and then we can
safely add 'norecovery' and 'nologreplay' mount option for btrfs.

The patch itself is not huge and already tested by others since v2.
Only mount option name is changing during all these version changes.

Thanks,
Qu

Qu Wenruo (2):
  btrfs: Introduce new mount option usebackuproot to replace recovery
  btrfs: Introduce new mount option to disable tree log replay

 Documentation/filesystems/btrfs.txt | 18 +++++++++++---
 fs/btrfs/ctree.h                    |  6 +++--
 fs/btrfs/disk-io.c                  | 15 ++++++++----
 fs/btrfs/super.c                    | 47 ++++++++++++++++++++++++++++++-------
 4 files changed, 68 insertions(+), 18 deletions(-)

-- 
2.7.0

[PATCH v5 1/2] btrfs: Introduce new mount option usebackuproot to replace recovery

[Qu Wenruo]

Current "recovery" mount option will only try to use backup root.
However the word "recovery" is too generic and may be confusing for some
users.

Here introduce a new and more specific mount option, "backuproot" to
replace "recovery" mount option.
"Recovery" will be kept for compatibility reason, but will be
deprecated.

Also, since "backuproot" will only affect mount behavior and after
open_ctree() it has nothing to do with the filesystem, so clear the flag
after mount succeeded.

This provides the basis for later unified "norecovery" mount option.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
v4:
  Newly introduced to avoid confusion with later 'norecovery' patch.
v5:
  Use 'usebackuproot' to replace 'backuproot' mount option.
---
 Documentation/filesystems/btrfs.txt |  9 ++++++---
 fs/btrfs/ctree.h                    |  2 +-
 fs/btrfs/disk-io.c                  |  8 +++++++-
 fs/btrfs/super.c                    | 17 +++++++++++------
 4 files changed, 25 insertions(+), 11 deletions(-)

diff --git a/Documentation/filesystems/btrfs.txt b/Documentation/filesystems/btrfs.txt
index c772b47..8e22d92 100644
--- a/Documentation/filesystems/btrfs.txt
+++ b/Documentation/filesystems/btrfs.txt
@@ -168,10 +168,13 @@ Options with (*) are default options and will not show in the mount options.
   notreelog
 	Enable/disable the tree logging used for fsync and O_SYNC writes.
 
-  recovery
-	Enable autorecovery attempts if a bad tree root is found at mount time.
-	Currently this scans a list of several previous tree roots and tries to
+  usebackuproot
+	Enable attempts to use backup tree roots if a bad tree root is found at
+	mount time.
+	Currently this scans a list of 4 previous tree roots and tries to
 	use the first readable.
+	And since the mount option doesn't affect any behavior after mount,
+	it won't be shown in mount info.
 
   rescan_uuid_tree
 	Force check and rebuild procedure of the UUID tree. This should not
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index c5f40dc..47a96a7 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2241,7 +2241,7 @@ struct btrfs_ioctl_defrag_range_args {
 #define BTRFS_MOUNT_ENOSPC_DEBUG	 (1 << 15)
 #define BTRFS_MOUNT_AUTO_DEFRAG		(1 << 16)
 #define BTRFS_MOUNT_INODE_MAP_CACHE	(1 << 17)
-#define BTRFS_MOUNT_RECOVERY		(1 << 18)
+#define BTRFS_MOUNT_USEBACKUPROOT	(1 << 18)
 #define BTRFS_MOUNT_SKIP_BALANCE	(1 << 19)
 #define BTRFS_MOUNT_CHECK_INTEGRITY	(1 << 20)
 #define BTRFS_MOUNT_CHECK_INTEGRITY_INCLUDING_EXTENT_DATA (1 << 21)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c67c129..6f16557 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -3141,6 +3141,12 @@ retry_root_backup:
 
 	fs_info->open = 1;
 
+	/*
+	 * backuproot only affect mount behavior, and if open_ctree succeeded,
+	 * no need to keep the flag
+	 */
+	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
+
 	return 0;
 
 fail_qgroup:
@@ -3195,7 +3201,7 @@ fail:
 	return err;
 
 recovery_tree_root:
-	if (!btrfs_test_opt(tree_root, RECOVERY))
+	if (!btrfs_test_opt(tree_root, USEBACKUPROOT))
 		goto fail_tree_roots;
 
 	free_root_pointers(fs_info, 0);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 86f7fdc..0842456 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -303,7 +303,7 @@ enum {
 	Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
 	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
-	Opt_datasum, Opt_treelog, Opt_noinode_cache,
+	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
 #ifdef CONFIG_BTRFS_DEBUG
 	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 #endif
@@ -352,7 +352,8 @@ static const match_table_t tokens = {
 	{Opt_inode_cache, "inode_cache"},
 	{Opt_noinode_cache, "noinode_cache"},
 	{Opt_no_space_cache, "nospace_cache"},
-	{Opt_recovery, "recovery"},
+	{Opt_recovery, "recovery"}, /* deprecated */
+	{Opt_usebackuproot, "usebackuproot"},
 	{Opt_skip_balance, "skip_balance"},
 	{Opt_check_integrity, "check_int"},
 	{Opt_check_integrity_including_extent_data, "check_int_data"},
@@ -681,8 +682,12 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 					     "disabling auto defrag");
 			break;
 		case Opt_recovery:
-			btrfs_info(root->fs_info, "enabling auto recovery");
-			btrfs_set_opt(info->mount_opt, RECOVERY);
+			btrfs_warn(root->fs_info,
+				   "'recovery' is deprecated, use 'usebackuproot' instead");
+		case Opt_usebackuproot:
+			btrfs_info(root->fs_info,
+				   "trying to use backup root at mount time");
+			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 			break;
 		case Opt_skip_balance:
 			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
@@ -1213,8 +1218,8 @@ static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 		seq_puts(seq, ",inode_cache");
 	if (btrfs_test_opt(root, SKIP_BALANCE))
 		seq_puts(seq, ",skip_balance");
-	if (btrfs_test_opt(root, RECOVERY))
-		seq_puts(seq, ",recovery");
+	if (btrfs_test_opt(root, USEBACKUPROOT))
+		seq_puts(seq, ",usebackuproot");
 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 	if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
 		seq_puts(seq, ",check_int_data");
-- 
2.7.0

[PATCH v5 2/2] btrfs: Introduce new mount option to disable tree log replay

[Qu Wenruo]

Introduce a new mount option "nologreplay" to co-operate with "ro" mount
option to get real readonly mount, like "norecovery" in ext* and xfs.

Since the new parse_options() need to check new flags at remount time,
so add a new parameter for parse_options().

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Tested-by: Austin S. Hemmelgarn <ahferroin7@gmail.com>
---
v2:
  Make RO check mandatory for btrfs_parse_options().
  Add btrfs_show_options() support for nologreplay.

  Document for btrfs-mount(5) will follow after the patch being merged.
v3:
  Apply the document suggestion from  Chandan Rajendra.
v4:
  Add 'norecovery' alias for 'nologreplay'.
v5:
  None
---
 Documentation/filesystems/btrfs.txt |  9 +++++++++
 fs/btrfs/ctree.h                    |  4 +++-
 fs/btrfs/disk-io.c                  |  7 ++++---
 fs/btrfs/super.c                    | 30 +++++++++++++++++++++++++++---
 4 files changed, 43 insertions(+), 7 deletions(-)

diff --git a/Documentation/filesystems/btrfs.txt b/Documentation/filesystems/btrfs.txt
index 8e22d92..a4dab53 100644
--- a/Documentation/filesystems/btrfs.txt
+++ b/Documentation/filesystems/btrfs.txt
@@ -168,6 +168,15 @@ Options with (*) are default options and will not show in the mount options.
   notreelog
 	Enable/disable the tree logging used for fsync and O_SYNC writes.
 
+  norecovery
+  nologreplay
+	Disable the log tree replay at mount time to prevent filesystem
+	from getting modified.
+	Must be used with 'ro' mount option.
+	A filesystem mounted with this option cannot transition to a
+	read-write mount via remount,rw - the filesystem must be unmounted
+	and mounted back again if read-write access is desired.
+
   usebackuproot
 	Enable attempts to use backup tree roots if a bad tree root is found at
 	mount time.
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 47a96a7..7ef9b2a 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2250,6 +2250,7 @@ struct btrfs_ioctl_defrag_range_args {
 #define BTRFS_MOUNT_FRAGMENT_DATA	(1 << 24)
 #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
 #define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
+#define BTRFS_MOUNT_NOLOGREPLAY		(1 << 27)
 
 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
 #define BTRFS_DEFAULT_MAX_INLINE	(8192)
@@ -4144,7 +4145,8 @@ void btrfs_sysfs_remove_mounted(struct btrfs_fs_info *fs_info);
 ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size);
 
 /* super.c */
-int btrfs_parse_options(struct btrfs_root *root, char *options);
+int btrfs_parse_options(struct btrfs_root *root, char *options,
+			unsigned long new_flags);
 int btrfs_sync_fs(struct super_block *sb, int wait);
 
 #ifdef CONFIG_PRINTK
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 6f16557..0ec4199 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2726,7 +2726,7 @@ int open_ctree(struct super_block *sb,
 	 */
 	fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
 
-	ret = btrfs_parse_options(tree_root, options);
+	ret = btrfs_parse_options(tree_root, options, sb->s_flags);
 	if (ret) {
 		err = ret;
 		goto fail_alloc;
@@ -3024,8 +3024,9 @@ retry_root_backup:
 	if (ret)
 		goto fail_trans_kthread;
 
-	/* do not make disk changes in broken FS */
-	if (btrfs_super_log_root(disk_super) != 0) {
+	/* do not make disk changes in broken FS or nologreplay is given */
+	if (btrfs_super_log_root(disk_super) != 0 &&
+	    !btrfs_test_opt(tree_root, NOLOGREPLAY)) {
 		ret = btrfs_replay_log(fs_info, fs_devices);
 		if (ret) {
 			err = ret;
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 0842456..df5534e 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -304,6 +304,7 @@ enum {
 	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
 	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
 	Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
+	Opt_norecovery, Opt_nologreplay,
 #ifdef CONFIG_BTRFS_DEBUG
 	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 #endif
@@ -335,6 +336,8 @@ static const match_table_t tokens = {
 	{Opt_noacl, "noacl"},
 	{Opt_notreelog, "notreelog"},
 	{Opt_treelog, "treelog"},
+	{Opt_norecovery, "norecovery"},
+	{Opt_nologreplay, "nologreplay"},
 	{Opt_flushoncommit, "flushoncommit"},
 	{Opt_noflushoncommit, "noflushoncommit"},
 	{Opt_ratio, "metadata_ratio=%d"},
@@ -374,7 +377,8 @@ static const match_table_t tokens = {
  * reading in a new superblock is parsed here.
  * XXX JDM: This needs to be cleaned up for remount.
  */
-int btrfs_parse_options(struct btrfs_root *root, char *options)
+int btrfs_parse_options(struct btrfs_root *root, char *options,
+			unsigned long new_flags)
 {
 	struct btrfs_fs_info *info = root->fs_info;
 	substring_t args[MAX_OPT_ARGS];
@@ -391,8 +395,12 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 	else if (cache_gen)
 		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 
+	/*
+	 * Even the options are empty, we still need to do extra check
+	 * against new flags
+	 */
 	if (!options)
-		goto out;
+		goto check;
 
 	/*
 	 * strsep changes the string, duplicate it because parse_options
@@ -592,6 +600,11 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			btrfs_clear_and_info(root, NOTREELOG,
 					     "enabling tree log");
 			break;
+		case Opt_nologreplay:
+		case Opt_norecovery:
+			btrfs_set_and_info(root, NOLOGREPLAY,
+					   "disabling log replay at mount time");
+			break;
 		case Opt_flushoncommit:
 			btrfs_set_and_info(root, FLUSHONCOMMIT,
 					   "turning on flush-on-commit");
@@ -782,6 +795,15 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			break;
 		}
 	}
+check:
+	/*
+	 * Extra check for current option against current flag
+	 */
+	if (btrfs_test_opt(root, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
+		btrfs_err(root->fs_info,
+			  "nologreplay must be used with ro mount option");
+		ret = -EINVAL;
+	}
 out:
 	if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
 	    !btrfs_test_opt(root, FREE_SPACE_TREE) &&
@@ -1192,6 +1214,8 @@ static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 		seq_puts(seq, ",ssd");
 	if (btrfs_test_opt(root, NOTREELOG))
 		seq_puts(seq, ",notreelog");
+	if (btrfs_test_opt(root, NOLOGREPLAY))
+		seq_puts(seq, ",nologreplay");
 	if (btrfs_test_opt(root, FLUSHONCOMMIT))
 		seq_puts(seq, ",flushoncommit");
 	if (btrfs_test_opt(root, DISCARD))
@@ -1677,7 +1701,7 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 		}
 	}
 
-	ret = btrfs_parse_options(root, data);
+	ret = btrfs_parse_options(root, data, *flags);
 	if (ret) {
 		ret = -EINVAL;
 		goto restore;
-- 
2.7.0

Re: [PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[David Sterba]

On Tue, Jan 12, 2016 at 10:05:11AM +0800, Qu Wenruo wrote:
> 'Norecovery' mount option is almost a standard for all other
> filesystems, including f2fs, ext4, gfs2, nilfs2, and xfs.
> And since current 'recovery' mount option for btrfs is too generic, it's
> better to change btrfs 'recovery' to 'usebackuproot' and then we can
> safely add 'norecovery' and 'nologreplay' mount option for btrfs.
> 
> The patch itself is not huge and already tested by others since v2.
> Only mount option name is changing during all these version changes.

So we have some kind of consensus about the options, but I think you
need to split the patches further.

* the new option 'nologreplay'
* keep 'usebackuproot' as a new alias for 'recovery' (with the
  deprecation warning)
* new 'norecovery' that switches on 'nologreplay' and 'nousebackuproot'
  (ie. the desired "no writes")

There is a case in remount, when we start with 'ro,nologreplay' then
remount with just 'ro', the log should be replayed. The other
combinations should be easy. I haven't looked at the patch if it's
implemented that way though.

Re: [PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[Qu Wenruo]

David Sterba wrote on 2016/01/14 17:00 +0100:
> On Tue, Jan 12, 2016 at 10:05:11AM +0800, Qu Wenruo wrote:
>> 'Norecovery' mount option is almost a standard for all other
>> filesystems, including f2fs, ext4, gfs2, nilfs2, and xfs.
>> And since current 'recovery' mount option for btrfs is too generic, it's
>> better to change btrfs 'recovery' to 'usebackuproot' and then we can
>> safely add 'norecovery' and 'nologreplay' mount option for btrfs.
>>
>> The patch itself is not huge and already tested by others since v2.
>> Only mount option name is changing during all these version changes.
>
> So we have some kind of consensus about the options, but I think you
> need to split the patches further.

It's never a bad idea to split patches :)

>
> * the new option 'nologreplay'
> * keep 'usebackuproot' as a new alias for 'recovery' (with the
>    deprecation warning)
> * new 'norecovery' that switches on 'nologreplay' and 'nousebackuproot'
>    (ie. the desired "no writes")
>
> There is a case in remount, when we start with 'ro,nologreplay' then
> remount with just 'ro', the log should be replayed. The other
> combinations should be easy. I haven't looked at the patch if it's
> implemented that way though.
>
>
For that case, even remount with -o remount,ro the nologreplay mount 
option won't be erased.
That's why we added a lot of "noXXXX" mount option, or we can't disable 
some feature by remount.

And since I didn't add such "logreplay" mount option intentionally, so 
there is no way to replay log unless umount the fs and mount it without 
"nologreplay" again.

I'll update the patchset soon and hopes it can catch the rc1 merge window.

Thanks,
Qu

Re: [PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[David Sterba]

On Fri, Jan 15, 2016 at 09:49:58AM +0800, Qu Wenruo wrote:
> David Sterba wrote on 2016/01/14 17:00 +0100:
> > On Tue, Jan 12, 2016 at 10:05:11AM +0800, Qu Wenruo wrote:
> >> 'Norecovery' mount option is almost a standard for all other
> >> filesystems, including f2fs, ext4, gfs2, nilfs2, and xfs.
> >> And since current 'recovery' mount option for btrfs is too generic, it's
> >> better to change btrfs 'recovery' to 'usebackuproot' and then we can
> >> safely add 'norecovery' and 'nologreplay' mount option for btrfs.
> >>
> >> The patch itself is not huge and already tested by others since v2.
> >> Only mount option name is changing during all these version changes.
> >
> > So we have some kind of consensus about the options, but I think you
> > need to split the patches further.
> 
> It's never a bad idea to split patches :)

My comment was motivated by the patch that introduced nologreplay and
norecovery in one go, although the functionality is logically separate
from the user's POV.

> > * the new option 'nologreplay'
> > * keep 'usebackuproot' as a new alias for 'recovery' (with the
> >    deprecation warning)
> > * new 'norecovery' that switches on 'nologreplay' and 'nousebackuproot'
> >    (ie. the desired "no writes")
> >
> > There is a case in remount, when we start with 'ro,nologreplay' then
> > remount with just 'ro', the log should be replayed. The other
> > combinations should be easy. I haven't looked at the patch if it's
> > implemented that way though.
> >
> >
> For that case, even remount with -o remount,ro the nologreplay mount 
> option won't be erased.
> That's why we added a lot of "noXXXX" mount option, or we can't disable 
> some feature by remount.

Yeah, I've realized that later that we need the 'logreplay' option to
achieve the readonly transition to the replayed state.

> And since I didn't add such "logreplay" mount option intentionally, so 
> there is no way to replay log unless umount the fs and mount it without 
> "nologreplay" again.
> 
> I'll update the patchset soon and hopes it can catch the rc1 merge window.

Yeah, I'd like to see it 4.5 as well, the topic has been discussed
enough and the implementation is simple.

Re: [PATCH v5 0/2][PULL REQUEST] Unify 'norecovery' bahavior with other filesystems

[Qu Wenruo]

David Sterba wrote on 2016/01/14 17:00 +0100:
> On Tue, Jan 12, 2016 at 10:05:11AM +0800, Qu Wenruo wrote:
>> 'Norecovery' mount option is almost a standard for all other
>> filesystems, including f2fs, ext4, gfs2, nilfs2, and xfs.
>> And since current 'recovery' mount option for btrfs is too generic, it's
>> better to change btrfs 'recovery' to 'usebackuproot' and then we can
>> safely add 'norecovery' and 'nologreplay' mount option for btrfs.
>>
>> The patch itself is not huge and already tested by others since v2.
>> Only mount option name is changing during all these version changes.
>
> So we have some kind of consensus about the options, but I think you
> need to split the patches further.
>
> * the new option 'nologreplay'
> * keep 'usebackuproot' as a new alias for 'recovery' (with the
>    deprecation warning)
> * new 'norecovery' that switches on 'nologreplay' and 'nousebackuproot'
>    (ie. the desired "no writes")

Hi David,

I'm curious about why 'norecovery' needs to imply 'nousebackuproot'.
As usebackuproot only affects where btrfs to read the tree/chunk roots.
It doesn't mean to change/write anything. RO will ensure superblock is 
not modified.

So I'll only add 'norecovery' as alias for 'nologreplay'.

Thanks,
Qu

>
> There is a case in remount, when we start with 'ro,nologreplay' then
> remount with just 'ro', the log should be replayed. The other
> combinations should be easy. I haven't looked at the patch if it's
> implemented that way though.
>
>