Skip to contentSkip to conversation

btrfs: preserve the compression property when other inode flags change

PATCH Superseded Sam Ho 2 participants 4 messages Aug 14 Btrfs file system

lore
Share & tools
lore

Sam Ho

author
#0 lore raw

PATCH 1/1 btrfs: preserve the compression property when other inode flags change

Setting the compression property on an inode also sets BTRFS_INODE_COMPRESS on it, and btrfs_inode_flags_to_fsflags() reports that back as FS_COMPR_FL to FS_IOC_GETFLAGS. chattr(1), like any other FS_IOC_SETFLAGS caller, reads the current flags, flips only the bit the user asked for and writes the whole set back, so a request as unrelated as "chattr +i" reaches btrfs_fileattr_set() with FS_COMPR_FL set.

btrfs_fileattr_set() takes that as a request to enable compression and overwrites the compression property with the algorithm from the mount options, falling back to zlib when the filesystem was not mounted with -o compress. The algorithm the user selected is silently replaced:

  # btrfs property set /mnt/foo compression zstd
  # btrfs property get /mnt/foo compression
  compression=zstd
  # chattr +i /mnt/foo
  # btrfs property get /mnt/foo compression
  compression=zlib

Every chattr operation triggers this, not just +i, and directories are affected as well, so files created afterwards inherit the wrong algorithm too. On a filesystem mounted with -o compress=lzo the property is replaced with lzo instead. Recovering needs a chattr -i first, because the immutable flag rejects the setxattr that "btrfs property set" issues.

Only pick the default algorithm when compression is actually being enabled by this call, that is when FS_COMPR_FL was not set before, and otherwise keep the algorithm recorded in the property. Inodes that have the compress flag set but no property still get the default, so they behave as before.

---
 fs/btrfs/ioctl.c | 22 +++++++++++++++++++---
 1 file changed, 19 insertions(+), 3 deletions(-)
fs/btrfs/ioctl.c +19 -3
index baa645e98812..2e54694f06f7 100644
@@ -384,9 +384,25 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
 		inode_flags |= BTRFS_INODE_COMPRESS;
 		inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 
-		comp = btrfs_compress_type2str(fs_info->compress_type);
-		if (!comp || comp[0] == 0)
-			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		/*
+		 * If compression was already enabled, keep the algorithm that
+		 * is recorded in the compression property. Otherwise changing
+		 * an unrelated attribute would reset it to the mount default,
+		 * since FS_IOC_SETFLAGS callers pass back the whole flag set
+		 * they got from FS_IOC_GETFLAGS.
+		 *
+		 * Fall back to the default when compression is being enabled
+		 * by this call, or when the inode has the compress flag set
+		 * but no property, which is possible on filesystems touched by
+		 * kernels that did not keep the two in sync.
+		 */
+		if (old_fsflags & FS_COMPR_FL)
+			comp = btrfs_compress_type2str(inode->prop_compress);
+		if (!comp || comp[0] == 0) {
+			comp = btrfs_compress_type2str(fs_info->compress_type);
+			if (!comp || comp[0] == 0)
+				comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		}
 	} else {
 		inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 	}
Signature · 4 lines
2.34.1


Disclaimer: The contents of this e-mail message and any attachments are confidential and are intended solely for addressee. The information may also be legally privileged. This transmission is sent in trust, for the sole purpose of delivery to the intended recipient. If you have received this transmission in error, any use, reproduction or dissemination of this transmission is strictly prohibited. If you are not the intended recipient, please immediately notify the sender by reply e-mail or phone and delete this message and its attachments, if any.
#1 lore raw
Quoted 25 lines · Sam Ho

Setting the compression property on an inode also sets BTRFS_INODE_COMPRESS on it, and btrfs_inode_flags_to_fsflags() reports that back as FS_COMPR_FL to FS_IOC_GETFLAGS. chattr(1), like any other FS_IOC_SETFLAGS caller, reads the current flags, flips only the bit the user asked for and writes the whole set back, so a request as unrelated as "chattr +i" reaches btrfs_fileattr_set() with FS_COMPR_FL set.

btrfs_fileattr_set() takes that as a request to enable compression and overwrites the compression property with the algorithm from the mount options, falling back to zlib when the filesystem was not mounted with -o compress. The algorithm the user selected is silently replaced:

   # btrfs property set /mnt/foo compression zstd
   # btrfs property get /mnt/foo compression
   compression=zstd
   # chattr +i /mnt/foo
   # btrfs property get /mnt/foo compression
   compression=zlib

Every chattr operation triggers this, not just +i, and directories are affected as well, so files created afterwards inherit the wrong algorithm too. On a filesystem mounted with -o compress=lzo the property is replaced with lzo instead. Recovering needs a chattr -i first, because the immutable flag rejects the setxattr that "btrfs property set" issues.

Quoted 6 lines

Only pick the default algorithm when compression is actually being enabled by this call, that is when FS_COMPR_FL was not set before, and otherwise keep the algorithm recorded in the property. Inodes that have the compress flag set but no property still get the default, so they behave as before.

The analyze looks good to me.

Although a minor nitpick related to the compression checks.

Quoted 19 lines
---
  fs/btrfs/ioctl.c | 22 +++++++++++++++++++---
  1 file changed, 19 insertions(+), 3 deletions(-)
fs/btrfs/ioctl.c +4 -3
index baa645e98812..2e54694f06f7 100644
@@ -384,9 +384,25 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
  		inode_flags |= BTRFS_INODE_COMPRESS;
  		inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 
-		comp = btrfs_compress_type2str(fs_info->compress_type);
-		if (!comp || comp[0] == 0)
-			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		/*
+		 * If compression was already enabled, keep the algorithm that
+		 * is recorded in the compression property. Otherwise changing
+		 * an unrelated attribute would reset it to the mount default,

fs/btrfs/ioctl.c int btrfs_fileattr_set(struct mnt_idmap *idmap, · line 400

Quoted 10 lines
fs/btrfs/ioctl.c +10 -0
+		 * since FS_IOC_SETFLAGS callers pass back the whole flag set
+		 * they got from FS_IOC_GETFLAGS.
+		 *
+		 * Fall back to the default when compression is being enabled
+		 * by this call, or when the inode has the compress flag set
+		 * but no property, which is possible on filesystems touched by
+		 * kernels that did not keep the two in sync.
+		 */
+		if (old_fsflags & FS_COMPR_FL)
+			comp = btrfs_compress_type2str(inode->prop_compress);

I do not think we need to always use the string.

We can directly use the compression type and convert it to string at the last second.

And we can skip the old_fsflags check and directly check inode->prop_compress.

E.g. something like the following will be a little easier to read:

fs/btrfs/ioctl.c +8 -3
index ebfb258161c8..befc0df0d5ab 100644
@@ -384,6 +384,7 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
                inode_flags &= ~BTRFS_INODE_COMPRESS;
                inode_flags |= BTRFS_INODE_NOCOMPRESS;
        } else if (fsflags & FS_COMPR_FL) {
+               enum btrfs_compression_type comp_type;
 
                if (IS_SWAPFILE(&inode->vfs_inode))
                        return -ETXTBSY;
@@ -391,9 +392,13 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
                inode_flags |= BTRFS_INODE_COMPRESS;
                inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 
-               comp = btrfs_compress_type2str(fs_info->compress_type);
-               if (!comp || comp[0] == 0)
-                       comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+               if (inode->prop_compress)
+                       comp_type = inode->prop_compress;
+               else if (fs_info->compress_type)
+                       comp_type = fs_info->compress_type;
+               else
+                       comp_type = BTRFS_COMPRESS_ZLIB;
+               comp = btrfs_compress_type2str(comp_type);
        } else {
                inode_flags &= ~(BTRFS_INODE_COMPRESS | 
BTRFS_INODE_NOCOMPRESS);
        }

Thanks,
Qu

Quoted 8 lines
fs/btrfs/ioctl.c +5 -0
+		if (!comp || comp[0] == 0) {
+			comp = btrfs_compress_type2str(fs_info->compress_type);
+			if (!comp || comp[0] == 0)
+				comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		}
  	} else {
  		inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
  	}
author
#2 lore raw

PATCH v2 1/1 btrfs: preserve the compression property when other inode flags change

In reply to Sam Ho

Setting the compression property on an inode also sets BTRFS_INODE_COMPRESS on it, and btrfs_inode_flags_to_fsflags() reports that back as FS_COMPR_FL to FS_IOC_GETFLAGS. chattr(1), like any other FS_IOC_SETFLAGS caller, reads the current flags, flips only the bit the user asked for and writes the whole set back, so a request as unrelated as "chattr +i" reaches btrfs_fileattr_set() with FS_COMPR_FL set.

btrfs_fileattr_set() takes that as a request to enable compression and overwrites the compression property with the algorithm from the mount options, falling back to zlib when the filesystem was not mounted with -o compress. The algorithm the user selected is silently replaced:

  # btrfs property set /mnt/foo compression zstd
  # btrfs property get /mnt/foo compression
  compression=zstd
  # chattr +i /mnt/foo
  # btrfs property get /mnt/foo compression
  compression=zlib

Every chattr operation triggers this, not just +i, and directories are affected as well, so files created afterwards inherit the wrong algorithm too. On a filesystem mounted with -o compress=lzo the property is replaced with lzo instead. Recovering needs a chattr -i first, because the immutable flag rejects the setxattr that "btrfs property set" issues.

Prefer the algorithm recorded in the compression property and only fall back to the mount default when there is no property, so that unrelated flag changes no longer overwrite the user's choice. Inodes that have the compress flag set but no property still get the default, so they behave as before.

---
v2:
- Pick the compression type first and convert it to a string only once at the end, instead of going through btrfs_compress_type2str() for every candidate, and check inode->prop_compress directly rather than testing old_fsflags for FS_COMPR_FL. Suggested by Qu Wenruo; the two are equivalent, since prop_compression_apply() only leaves prop_compress set while BTRFS_INODE_COMPRESS is set.
- Reword the last changelog paragraph and the comment to match.

v1: https://lore.kernel.org/linux-btrfs/[email protected]/

 fs/btrfs/ioctl.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)
fs/btrfs/ioctl.c +18 -3
index baa645e98812..343d089aa5c3 100644
@@ -377,6 +377,7 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
 		inode_flags &= ~BTRFS_INODE_COMPRESS;
 		inode_flags |= BTRFS_INODE_NOCOMPRESS;
 	} else if (fsflags & FS_COMPR_FL) {
+		enum btrfs_compression_type comp_type;
 
 		if (IS_SWAPFILE(&inode->vfs_inode))
 			return -ETXTBSY;
@@ -384,9 +385,23 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
 		inode_flags |= BTRFS_INODE_COMPRESS;
 		inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 
-		comp = btrfs_compress_type2str(fs_info->compress_type);
-		if (!comp || comp[0] == 0)
-			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		/*
+		 * Keep the algorithm recorded in the compression property,
+		 * otherwise changing an unrelated attribute would reset it to
+		 * the mount default, since FS_IOC_SETFLAGS callers write back
+		 * the whole flag set they got from FS_IOC_GETFLAGS and that
+		 * includes FS_COMPR_FL for any inode carrying the property.
+		 *
+		 * Inodes with the compress flag set but no property keep using
+		 * the mount default, so they behave as before.
+		 */
+		if (inode->prop_compress)
+			comp_type = inode->prop_compress;
+		else if (fs_info->compress_type)
+			comp_type = fs_info->compress_type;
+		else
+			comp_type = BTRFS_COMPRESS_ZLIB;
+		comp = btrfs_compress_type2str(comp_type);
 	} else {
 		inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 	}
Signature · 4 lines
2.34.1


Disclaimer: The contents of this e-mail message and any attachments are confidential and are intended solely for addressee. The information may also be legally privileged. This transmission is sent in trust, for the sole purpose of delivery to the intended recipient. If you have received this transmission in error, any use, reproduction or dissemination of this transmission is strictly prohibited. If you are not the intended recipient, please immediately notify the sender by reply e-mail or phone and delete this message and its attachments, if any.
Reviewed-by on v2 1/1
#3 lore raw
Quoted 25 lines · Sam Ho

Setting the compression property on an inode also sets BTRFS_INODE_COMPRESS on it, and btrfs_inode_flags_to_fsflags() reports that back as FS_COMPR_FL to FS_IOC_GETFLAGS. chattr(1), like any other FS_IOC_SETFLAGS caller, reads the current flags, flips only the bit the user asked for and writes the whole set back, so a request as unrelated as "chattr +i" reaches btrfs_fileattr_set() with FS_COMPR_FL set.

btrfs_fileattr_set() takes that as a request to enable compression and overwrites the compression property with the algorithm from the mount options, falling back to zlib when the filesystem was not mounted with -o compress. The algorithm the user selected is silently replaced:

   # btrfs property set /mnt/foo compression zstd
   # btrfs property get /mnt/foo compression
   compression=zstd
   # chattr +i /mnt/foo
   # btrfs property get /mnt/foo compression
   compression=zlib

Every chattr operation triggers this, not just +i, and directories are affected as well, so files created afterwards inherit the wrong algorithm too. On a filesystem mounted with -o compress=lzo the property is replaced with lzo instead. Recovering needs a chattr -i first, because the immutable flag rejects the setxattr that "btrfs property set" issues.

Quoted 7 lines

Prefer the algorithm recorded in the compression property and only fall back to the mount default when there is no property, so that unrelated flag changes no longer overwrite the user's choice. Inodes that have the compress flag set but no property still get the default, so they behave as before.

Thanks,
Qu

Quoted 54 lines

---
v2:
- Pick the compression type first and convert it to a string only once at the end, instead of going through btrfs_compress_type2str() for every candidate, and check inode->prop_compress directly rather than testing old_fsflags for FS_COMPR_FL. Suggested by Qu Wenruo; the two are equivalent, since prop_compression_apply() only leaves prop_compress set while BTRFS_INODE_COMPRESS is set.
- Reword the last changelog paragraph and the comment to match.

v1: https://lore.kernel.org/linux-btrfs/[email protected]/

  fs/btrfs/ioctl.c | 21 ++++++++++++++++++---
  1 file changed, 18 insertions(+), 3 deletions(-)
fs/btrfs/ioctl.c +18 -3
index baa645e98812..343d089aa5c3 100644
@@ -377,6 +377,7 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
  		inode_flags &= ~BTRFS_INODE_COMPRESS;
  		inode_flags |= BTRFS_INODE_NOCOMPRESS;
  	} else if (fsflags & FS_COMPR_FL) {
+		enum btrfs_compression_type comp_type;
 
  		if (IS_SWAPFILE(&inode->vfs_inode))
  			return -ETXTBSY;
@@ -384,9 +385,23 @@ int btrfs_fileattr_set(struct mnt_idmap *idmap,
  		inode_flags |= BTRFS_INODE_COMPRESS;
  		inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 
-		comp = btrfs_compress_type2str(fs_info->compress_type);
-		if (!comp || comp[0] == 0)
-			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
+		/*
+		 * Keep the algorithm recorded in the compression property,
+		 * otherwise changing an unrelated attribute would reset it to
+		 * the mount default, since FS_IOC_SETFLAGS callers write back
+		 * the whole flag set they got from FS_IOC_GETFLAGS and that
+		 * includes FS_COMPR_FL for any inode carrying the property.
+		 *
+		 * Inodes with the compress flag set but no property keep using
+		 * the mount default, so they behave as before.
+		 */
+		if (inode->prop_compress)
+			comp_type = inode->prop_compress;
+		else if (fs_info->compress_type)
+			comp_type = fs_info->compress_type;
+		else
+			comp_type = BTRFS_COMPRESS_ZLIB;
+		comp = btrfs_compress_type2str(comp_type);
  	} else {
  		inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
  	}