[PATCHv2 0/3] nilfs2 add feature flags

[PATCHv2 0/3] nilfs2 add feature flags

[Ryusuke Konishi]

This is a revised patchset that introduces feature flags and makes
nilfs refuse to mount future incompatibile filesystems.

The first patch is just for improving readability of super block
format.  The second patch extends super block and adds three flags to
record feature sets (i.e. "compat features", "read-only compat
features", and "incompat features").  The semantics of these flags is
mimic of ext3.  At present, no feature flags are defined.

The third patch makes nilfs reject filesystems with unknown features.
This serves as a safety for future disk format change which is not
forward compatible.

* Change from the v1 patchset is as follows:

 - insert the same test of feature flags also before proceed with
   recovery for read-only mount.

Thanks,
Ryusuke Konishi
--
 fs/nilfs2/nilfs.h         |    2 ++
 fs/nilfs2/super.c         |   39 +++++++++++++++++++++++++++++++++++++++
 fs/nilfs2/the_nilfs.c     |   20 ++++++++++++++++++++
 include/linux/nilfs2_fs.h |   41 +++++++++++++++++++++++++++--------------
 4 files changed, 88 insertions(+), 14 deletions(-)


--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCHv2 1/3] nilfs2: clarify byte offset in super block format

[Ryusuke Konishi]

This inserts comments indicating hexadecimal offset in declaration of
nilfs_super_block structure so that people can know offset of its
fields without counting from the head.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke-Zyj7fXuS5i5L9jVzuh4AOg@public.gmane.org>
---
 include/linux/nilfs2_fs.h |   26 +++++++++++++-------------
 1 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/include/linux/nilfs2_fs.h b/include/linux/nilfs2_fs.h
index 8c2c611..cc3465e 100644
--- a/include/linux/nilfs2_fs.h
+++ b/include/linux/nilfs2_fs.h
@@ -160,7 +160,7 @@ struct nilfs_super_root {
  * struct nilfs_super_block - structure of super block on disk
  */
 struct nilfs_super_block {
-	__le32	s_rev_level;		/* Revision level */
+/*00*/	__le32	s_rev_level;		/* Revision level */
 	__le16	s_minor_rev_level;	/* minor revision level */
 	__le16	s_magic;		/* Magic signature */
 
@@ -169,47 +169,47 @@ struct nilfs_super_block {
 					   is excluded. */
 	__le16  s_flags;		/* flags */
 	__le32  s_crc_seed;		/* Seed value of CRC calculation */
-	__le32	s_sum;			/* Check sum of super block */
+/*10*/	__le32	s_sum;			/* Check sum of super block */
 
 	__le32	s_log_block_size;	/* Block size represented as follows
 					   blocksize =
 					       1 << (s_log_block_size + 10) */
 	__le64  s_nsegments;		/* Number of segments in filesystem */
-	__le64  s_dev_size;		/* block device size in bytes */
+/*20*/	__le64  s_dev_size;		/* block device size in bytes */
 	__le64	s_first_data_block;	/* 1st seg disk block number */
-	__le32  s_blocks_per_segment;   /* number of blocks per full segment */
+/*30*/	__le32  s_blocks_per_segment;   /* number of blocks per full segment */
 	__le32	s_r_segments_percentage; /* Reserved segments percentage */
 
 	__le64  s_last_cno;		/* Last checkpoint number */
-	__le64  s_last_pseg;		/* disk block addr pseg written last */
+/*40*/	__le64  s_last_pseg;		/* disk block addr pseg written last */
 	__le64  s_last_seq;             /* seq. number of seg written last */
-	__le64	s_free_blocks_count;	/* Free blocks count */
+/*50*/	__le64	s_free_blocks_count;	/* Free blocks count */
 
 	__le64	s_ctime;		/* Creation time (execution time of
 					   newfs) */
-	__le64	s_mtime;		/* Mount time */
+/*60*/	__le64	s_mtime;		/* Mount time */
 	__le64	s_wtime;		/* Write time */
-	__le16	s_mnt_count;		/* Mount count */
+/*70*/	__le16	s_mnt_count;		/* Mount count */
 	__le16	s_max_mnt_count;	/* Maximal mount count */
 	__le16	s_state;		/* File system state */
 	__le16	s_errors;		/* Behaviour when detecting errors */
 	__le64	s_lastcheck;		/* time of last check */
 
-	__le32	s_checkinterval;	/* max. time between checks */
+/*80*/	__le32	s_checkinterval;	/* max. time between checks */
 	__le32	s_creator_os;		/* OS */
 	__le16	s_def_resuid;		/* Default uid for reserved blocks */
 	__le16	s_def_resgid;		/* Default gid for reserved blocks */
 	__le32	s_first_ino;		/* First non-reserved inode */
 
-	__le16  s_inode_size;		/* Size of an inode */
+/*90*/	__le16  s_inode_size;		/* Size of an inode */
 	__le16  s_dat_entry_size;       /* Size of a dat entry */
 	__le16  s_checkpoint_size;      /* Size of a checkpoint */
 	__le16	s_segment_usage_size;	/* Size of a segment usage */
 
-	__u8	s_uuid[16];		/* 128-bit uuid for volume */
-	char	s_volume_name[80];	/* volume name */
+/*98*/	__u8	s_uuid[16];		/* 128-bit uuid for volume */
+/*A8*/	char	s_volume_name[80];	/* volume name */
 
-	__le32  s_c_interval;           /* Commit interval of segment */
+/*F8*/	__le32  s_c_interval;           /* Commit interval of segment */
 	__le32  s_c_block_max;          /* Threshold of data amount for
 					   the segment construction */
 	__u32	s_reserved[192];	/* padding to the end of the block */
-- 
1.6.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCHv2 2/3] nilfs2: add feature set fields to super block

[Ryusuke Konishi]

This adds three new fields to nilfs_super_block structure, compatible
feature set, readonly-compatible feature set, and incompatible feature
set in order to prepare for future disk format modifications.

The role of these fields conforms to those of ext3 or other
filesystems.  Most important flags are the incompatible feature set;
it is used to refuse to mount the filesystem which sets an
incompatible feature the kernel doesn't know about.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke-Zyj7fXuS5i5L9jVzuh4AOg@public.gmane.org>
---
 include/linux/nilfs2_fs.h |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/include/linux/nilfs2_fs.h b/include/linux/nilfs2_fs.h
index cc3465e..7dd4cd4 100644
--- a/include/linux/nilfs2_fs.h
+++ b/include/linux/nilfs2_fs.h
@@ -212,7 +212,10 @@ struct nilfs_super_block {
 /*F8*/	__le32  s_c_interval;           /* Commit interval of segment */
 	__le32  s_c_block_max;          /* Threshold of data amount for
 					   the segment construction */
-	__u32	s_reserved[192];	/* padding to the end of the block */
+/*100*/	__le64  s_feature_compat;	/* Compatible feature set */
+	__le64  s_feature_compat_ro;	/* Read-only compatible feature set */
+	__le64  s_feature_incompat;	/* Incompatible feature set */
+	__u32	s_reserved[186];	/* padding to the end of the block */
 };
 
 /*
@@ -228,6 +231,16 @@ struct nilfs_super_block {
 #define NILFS_MINOR_REV		0	/* minor revision */
 
 /*
+ * Feature set definitions
+ *
+ * If there is a bit set in the incompatible feature set that the kernel
+ * doesn't know about, it should refuse to mount the filesystem.
+ */
+#define NILFS_FEATURE_COMPAT_SUPP	0ULL
+#define NILFS_FEATURE_COMPAT_RO_SUPP	0ULL
+#define NILFS_FEATURE_INCOMPAT_SUPP	0ULL
+
+/*
  * Bytes count of super_block for CRC-calculation
  */
 #define NILFS_SB_BYTES  \
-- 
1.6.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCHv2 3/3] nilfs2: reject incompatible filesystem

[Ryusuke Konishi]

This forces nilfs to check compatibility of feature flags so as to
reject a filesystem with unknown features when it mounts or remounts
the filesystem.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke-Zyj7fXuS5i5L9jVzuh4AOg@public.gmane.org>
---
 fs/nilfs2/nilfs.h     |    2 ++
 fs/nilfs2/super.c     |   39 +++++++++++++++++++++++++++++++++++++++
 fs/nilfs2/the_nilfs.c |   20 ++++++++++++++++++++
 3 files changed, 61 insertions(+), 0 deletions(-)

diff --git a/fs/nilfs2/nilfs.h b/fs/nilfs2/nilfs.h
index cfedc48..0842d77 100644
--- a/fs/nilfs2/nilfs.h
+++ b/fs/nilfs2/nilfs.h
@@ -275,6 +275,8 @@ extern struct nilfs_super_block *
 nilfs_read_super_block(struct super_block *, u64, int, struct buffer_head **);
 extern int nilfs_store_magic_and_option(struct super_block *,
 					struct nilfs_super_block *, char *);
+extern int nilfs_check_feature_compatibility(struct super_block *,
+					     struct nilfs_super_block *);
 extern void nilfs_set_log_cursor(struct nilfs_super_block *,
 				 struct the_nilfs *);
 extern struct nilfs_super_block **nilfs_prepare_super(struct nilfs_sb_info *,
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index 1644573..26078b3 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -790,6 +790,30 @@ int nilfs_store_magic_and_option(struct super_block *sb,
 	return !parse_options(data, sb, 0) ? -EINVAL : 0 ;
 }
 
+int nilfs_check_feature_compatibility(struct super_block *sb,
+				      struct nilfs_super_block *sbp)
+{
+	__u64 features;
+
+	features = le64_to_cpu(sbp->s_feature_incompat) &
+		~NILFS_FEATURE_INCOMPAT_SUPP;
+	if (features) {
+		printk(KERN_ERR "NILFS: couldn't mount because of unsupported "
+		       "optional features (%llx)\n",
+		       (unsigned long long)features);
+		return -EINVAL;
+	}
+	features = le64_to_cpu(sbp->s_feature_compat_ro) &
+		~NILFS_FEATURE_COMPAT_RO_SUPP;
+	if (!(sb->s_flags & MS_RDONLY) && features) {
+		printk(KERN_ERR "NILFS: couldn't mount RDWR because of "
+		       "unsupported optional features (%llx)\n",
+		       (unsigned long long)features);
+		return -EINVAL;
+	}
+	return 0;
+}
+
 /**
  * nilfs_fill_super() - initialize a super block instance
  * @sb: super_block
@@ -984,11 +1008,26 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 		nilfs_cleanup_super(sbi);
 		up_write(&nilfs->ns_sem);
 	} else {
+		__u64 features;
+
 		/*
 		 * Mounting a RDONLY partition read-write, so reread and
 		 * store the current valid flag.  (It may have been changed
 		 * by fsck since we originally mounted the partition.)
 		 */
+		down_read(&nilfs->ns_sem);
+		features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) &
+			~NILFS_FEATURE_COMPAT_RO_SUPP;
+		up_read(&nilfs->ns_sem);
+		if (features) {
+			printk(KERN_WARNING "NILFS (device %s): couldn't "
+			       "remount RDWR because of unsupported optional "
+			       "features (%llx)\n",
+			       sb->s_id, (unsigned long long)features);
+			err = -EROFS;
+			goto restore_opts;
+		}
+
 		sb->s_flags &= ~MS_RDONLY;
 
 		err = nilfs_attach_segment_constructor(sbi);
diff --git a/fs/nilfs2/the_nilfs.c b/fs/nilfs2/the_nilfs.c
index f2efc8c..da67b56 100644
--- a/fs/nilfs2/the_nilfs.c
+++ b/fs/nilfs2/the_nilfs.c
@@ -385,11 +385,23 @@ int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
 		goto skip_recovery;
 
 	if (s_flags & MS_RDONLY) {
+		__u64 features;
+
 		if (nilfs_test_opt(sbi, NORECOVERY)) {
 			printk(KERN_INFO "NILFS: norecovery option specified. "
 			       "skipping roll-forward recovery\n");
 			goto skip_recovery;
 		}
+		features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) &
+			~NILFS_FEATURE_COMPAT_RO_SUPP;
+		if (features) {
+			printk(KERN_ERR "NILFS: couldn't proceed with "
+			       "recovery because of unsupported optional "
+			       "features (%llx)\n",
+			       (unsigned long long)features);
+			err = -EROFS;
+			goto failed_unload;
+		}
 		if (really_read_only) {
 			printk(KERN_ERR "NILFS: write access "
 			       "unavailable, cannot proceed.\n");
@@ -644,6 +656,10 @@ int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
 		if (err)
 			goto out;
 
+		err = nilfs_check_feature_compatibility(sb, sbp);
+		if (err)
+			goto out;
+
 		blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size);
 		if (sb->s_blocksize != blocksize &&
 		    !sb_set_blocksize(sb, blocksize)) {
@@ -669,6 +685,10 @@ int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
 	if (err)
 		goto failed_sbh;
 
+	err = nilfs_check_feature_compatibility(sb, sbp);
+	if (err)
+		goto failed_sbh;
+
 	blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size);
 	if (sb->s_blocksize != blocksize) {
 		int hw_blocksize = bdev_logical_block_size(sb->s_bdev);
-- 
1.6.6.2

--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv2 0/3] nilfs2 add feature flags

[Jiro SEKIBA]

Hi, 

At Thu, 22 Jul 2010 03:22:17 +0900,
Ryusuke Konishi wrote:
> 
> This is a revised patchset that introduces feature flags and makes
> nilfs refuse to mount future incompatibile filesystems.
> 
> The first patch is just for improving readability of super block
> format.  The second patch extends super block and adds three flags to
> record feature sets (i.e. "compat features", "read-only compat
> features", and "incompat features").  The semantics of these flags is
> mimic of ext3.  At present, no feature flags are defined.
> 
> The third patch makes nilfs reject filesystems with unknown features.
> This serves as a safety for future disk format change which is not
> forward compatible.
> 
> * Change from the v1 patchset is as follows:
> 
>  - insert the same test of feature flags also before proceed with
>    recovery for read-only mount.

This is, I think, one of the mandatory functionality to take
"experimental flag" away.  Fixing disk format is important, 
but any unexpected incompatible disk format change may happen in future.

At least, nilfs can protect its filesystem from destructing by itself.

Thank you for the improvement!


> Thanks,
> Ryusuke Konishi
> --
>  fs/nilfs2/nilfs.h         |    2 ++
>  fs/nilfs2/super.c         |   39 +++++++++++++++++++++++++++++++++++++++
>  fs/nilfs2/the_nilfs.c     |   20 ++++++++++++++++++++
>  include/linux/nilfs2_fs.h |   41 +++++++++++++++++++++++++++--------------
>  4 files changed, 88 insertions(+), 14 deletions(-)
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> 
> 



-- 
Jiro SEKIBA <jir-hfpbi5WX9J54Eiagz67IpQ@public.gmane.org>
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv2 0/3] nilfs2 add feature flags

[Ryusuke Konishi]

On Fri, 23 Jul 2010 13:30:05 +0900, Jiro SEKIBA wrote:
> Hi, 
> 
> At Thu, 22 Jul 2010 03:22:17 +0900,
> Ryusuke Konishi wrote:
> > 
> > This is a revised patchset that introduces feature flags and makes
> > nilfs refuse to mount future incompatibile filesystems.
> > 
> > The first patch is just for improving readability of super block
> > format.  The second patch extends super block and adds three flags to
> > record feature sets (i.e. "compat features", "read-only compat
> > features", and "incompat features").  The semantics of these flags is
> > mimic of ext3.  At present, no feature flags are defined.
> > 
> > The third patch makes nilfs reject filesystems with unknown features.
> > This serves as a safety for future disk format change which is not
> > forward compatible.
> > 
> > * Change from the v1 patchset is as follows:
> > 
> >  - insert the same test of feature flags also before proceed with
> >    recovery for read-only mount.
> 
> This is, I think, one of the mandatory functionality to take
> "experimental flag" away.  Fixing disk format is important, 
> but any unexpected incompatible disk format change may happen in future.
>
> At least, nilfs can protect its filesystem from destructing by itself.

Yes, that's the reason why I push it now.
All right, I'll queue it for the next merge window as planned.

Thanks,
Ryusuke Konishi


 

 
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html