Skip to contentSkip to conversation

btrfs: do not allow compression on nocow files

PATCH In review Chung-Chiang Cheng 4 participants 5 messages Apr 9, 2022 – Apr 14, 2022 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btrfs: do not allow compression on nocow files

Compression and nocow are mutually exclusive. Besides ioctl, there is another way to enable/disable/reset compression directly via xattr. The following steps will result in a invalid combination.

  $ touch bar
  $ lsattr bar
  ---------------C-- bar
  $ setfattr -n btrfs.compression -v zstd bar
  $ lsattr bar
  --------c------C-- bar
---
 fs/btrfs/props.c | 20 ++++++++++++--------
 fs/btrfs/props.h |  3 ++-
 fs/btrfs/xattr.c |  2 +-
 3 files changed, 15 insertions(+), 10 deletions(-)
fs/btrfs/props.c +12 -8
index 1a6d2d5b4b33..fde2bf069b03 100644
@@ -17,7 +17,7 @@ static DEFINE_HASHTABLE(prop_handlers_ht, BTRFS_PROP_HANDLERS_HT_BITS);
 struct prop_handler {
 	struct hlist_node node;
 	const char *xattr_name;
-	int (*validate)(const char *value, size_t len);
+	int (*validate)(struct inode *inode, const char *value, size_t len);
 	int (*apply)(struct inode *inode, const char *value, size_t len);
 	const char *(*extract)(struct inode *inode);
 	int inheritable;
@@ -55,7 +55,8 @@ find_prop_handler(const char *name,
 	return NULL;
 }
 
-int btrfs_validate_prop(const char *name, const char *value, size_t value_len)
+int btrfs_validate_prop(struct inode *inode, const char *name,
+			const char *value, size_t value_len)
 {
 	const struct prop_handler *handler;
 
@@ -69,7 +70,7 @@ int btrfs_validate_prop(const char *name, const char *value, size_t value_len)
 	if (value_len == 0)
 		return 0;
 
-	return handler->validate(value, value_len);
+	return handler->validate(inode, value, value_len);
 }
 
 int btrfs_set_prop(struct btrfs_trans_handle *trans, struct inode *inode,
@@ -252,18 +253,21 @@ int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path)
 	return ret;
 }
 
-static int prop_compression_validate(const char *value, size_t len)
+static int prop_compression_validate(struct inode *inode, const char *value, size_t len)
 {
 	if (!value)
 		return 0;
 
-	if (btrfs_compress_is_valid_type(value, len))
-		return 0;
-
 	if ((len == 2 && strncmp("no", value, 2) == 0) ||
 	    (len == 4 && strncmp("none", value, 4) == 0))
 		return 0;
 
+	if (!inode || BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
+		return -EINVAL;
+
+	if (btrfs_compress_is_valid_type(value, len))
+		return 0;
+
 	return -EINVAL;
 }
 
@@ -364,7 +368,7 @@ static int inherit_props(struct btrfs_trans_handle *trans,
 		 * This is not strictly necessary as the property should be
 		 * valid, but in case it isn't, don't propagate it further.
 		 */
-		ret = h->validate(value, strlen(value));
+		ret = h->validate(inode, value, strlen(value));
 		if (ret)
 			continue;
 
fs/btrfs/props.h +2 -1
index 40b2c65b518c..0504c03177e3 100644
@@ -13,7 +13,8 @@ void __init btrfs_props_init(void);
 int btrfs_set_prop(struct btrfs_trans_handle *trans, struct inode *inode,
 		   const char *name, const char *value, size_t value_len,
 		   int flags);
-int btrfs_validate_prop(const char *name, const char *value, size_t value_len);
+int btrfs_validate_prop(struct inode *inode, const char *name,
+			const char *value, size_t value_len);
 
 int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path);
 
fs/btrfs/xattr.c +1 -1
index 99abf41b89b9..9aceae07a02b 100644
@@ -403,7 +403,7 @@ static int btrfs_xattr_handler_set_prop(const struct xattr_handler *handler,
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
 	name = xattr_full_name(handler, name);
-	ret = btrfs_validate_prop(name, value, size);
+	ret = btrfs_validate_prop(inode, name, value, size);
 	if (ret)
 		return ret;
 
Signature · 1 line
2.34.1
#1 lore raw
Quoted 8 lines · Chung-Chiang Cheng

Compression and nocow are mutually exclusive. Besides ioctl, there is another way to enable/disable/reset compression directly via xattr. The following steps will result in a invalid combination.

  $ touch bar
  $ lsattr bar
  ---------------C-- bar

The example should show how the NOCOW attribute ended up in the file. Either the filesystem mounted with -o nodatacow or a chattr +C was done on the file bar before the lsattr command.

Quoted 6 lines
  $ setfattr -n btrfs.compression -v zstd bar
  $ lsattr bar
  --------c------C-- bar

The changelog should mention what happens if we end up both with compression and nodatacow set on a file.
Will all future writes be compressed and nodatacow have no effect? Or will compression have no effect and we nodatacow overrides it? Or something else? Some corruption, some crash, etc?

Quoted 20 lines
---
 fs/btrfs/props.c | 20 ++++++++++++--------
 fs/btrfs/props.h |  3 ++-
 fs/btrfs/xattr.c |  2 +-
 3 files changed, 15 insertions(+), 10 deletions(-)
fs/btrfs/props.c +1 -1
index 1a6d2d5b4b33..fde2bf069b03 100644
@@ -17,7 +17,7 @@ static DEFINE_HASHTABLE(prop_handlers_ht, BTRFS_PROP_HANDLERS_HT_BITS);
 struct prop_handler {
        struct hlist_node node;
        const char *xattr_name;
-       int (*validate)(const char *value, size_t len);
+       int (*validate)(struct inode *inode, const char *value, size_t len);
        int (*apply)(struct inode *inode, const char *value, size_t len);
        const char *(*extract)(struct inode *inode);
        int inheritable;

fs/btrfs/props.c find_prop_handler(const char *name, · line 58

Quoted 5 lines
fs/btrfs/props.c +1 -1
@@ -55,7 +55,8 @@ find_prop_handler(const char *name,
        return NULL;
 }
 
-int btrfs_validate_prop(const char *name, const char *value, size_t value_len)
+int btrfs_validate_prop(struct inode *inode, const char *name,

The inode can be made const.
The validation handler should never need to change anything in the inode.

Also, we can use struct btrfs_inode * instead of struct inode (preferred).

Quoted 14 lines
fs/btrfs/props.c +2 -1
+                       const char *value, size_t value_len)
 {
        const struct prop_handler *handler;
 
@@ -69,7 +70,7 @@ int btrfs_validate_prop(const char *name, const char *value, size_t value_len)
        if (value_len == 0)
                return 0;
 
-       return handler->validate(value, value_len);
+       return handler->validate(inode, value, value_len);
 }
 
 int btrfs_set_prop(struct btrfs_trans_handle *trans, struct inode *inode,

fs/btrfs/props.c int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path) · line 256

Quoted 5 lines
fs/btrfs/props.c +1 -1
@@ -252,18 +253,21 @@ int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path)
        return ret;
 }
 
-static int prop_compression_validate(const char *value, size_t len)
+static int prop_compression_validate(struct inode *inode, const char *value, size_t len)

Same here, about const and using struct btrfs_inode * instead.

fs/btrfs/props.c int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path) · line 265

Quoted 12 lines
fs/btrfs/props.c +1 -3
 {
        if (!value)
                return 0;
 
-       if (btrfs_compress_is_valid_type(value, len))
-               return 0;
-
        if ((len == 2 && strncmp("no", value, 2) == 0) ||
            (len == 4 && strncmp("none", value, 4) == 0))
                return 0;
 
+       if (!inode || BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)

How can inode ever be NULL?
Unless I missed something, it can't be NULL.

We should also get a test case for fstests, to help prevent regressions in the future.

Otherwise it looks fine.

Thanks.

Quoted 47 lines
fs/btrfs/props.c +6 -1
+               return -EINVAL;
+
+       if (btrfs_compress_is_valid_type(value, len))
+               return 0;
+
        return -EINVAL;
 }
 
@@ -364,7 +368,7 @@ static int inherit_props(struct btrfs_trans_handle *trans,
                 * This is not strictly necessary as the property should be
                 * valid, but in case it isn't, don't propagate it further.
                 */
-               ret = h->validate(value, strlen(value));
+               ret = h->validate(inode, value, strlen(value));
                if (ret)
                        continue;
 
fs/btrfs/props.h +2 -1
index 40b2c65b518c..0504c03177e3 100644
@@ -13,7 +13,8 @@ void __init btrfs_props_init(void);
 int btrfs_set_prop(struct btrfs_trans_handle *trans, struct inode *inode,
                   const char *name, const char *value, size_t value_len,
                   int flags);
-int btrfs_validate_prop(const char *name, const char *value, size_t value_len);
+int btrfs_validate_prop(struct inode *inode, const char *name,
+                       const char *value, size_t value_len);
 
 int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path);
 
fs/btrfs/xattr.c +1 -2
index 99abf41b89b9..9aceae07a02b 100644
@@ -403,7 +403,7 @@ static int btrfs_xattr_handler_set_prop(const struct xattr_handler *handler,
        struct btrfs_root *root = BTRFS_I(inode)->root;
 
        name = xattr_full_name(handler, name);
-       ret = btrfs_validate_prop(name, value, size);
+       ret = btrfs_validate_prop(inode, name, value, size);
        if (ret)
                return ret;
 
--

2.34.1

#2 lore raw
Quoted 9 lines
Quoted 1 line

+ if (!inode || BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)

How can inode ever be NULL?
Unless I missed something, it can't be NULL.

We should also get a test case for fstests, to help prevent regressions in the future.

Otherwise it looks fine.

Thanks for your comments. I will send a v2 patch and a fstest item later. And as David mentioned, inode_can_compress() can be reused to reduce duplication.

Thanks.

#3 lore raw
Quoted 15 lines · David Sterba
Quoted 13 lines · Chung-Chiang Cheng
fs/btrfs/props.c +1 -4
-static int prop_compression_validate(const char *value, size_t len)
+static int prop_compression_validate(struct inode *inode, const char *value, size_t len)
 {
      if (!value)
              return 0;
 
-     if (btrfs_compress_is_valid_type(value, len))
-             return 0;
-
      if ((len == 2 && strncmp("no", value, 2) == 0) ||
          (len == 4 && strncmp("none", value, 4) == 0))
              return 0;
Quoted 6 lines

fs/btrfs/props.c int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path) · line 266

Quoted 2 lines
fs/btrfs/props.c +2 -0
+     if (!inode || BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
+             return -EINVAL;

I think the nodatacow check should be the first one, before the validation of value for "no" or "none", it's logically the same as the btrfs_compress_is_valid_type.

if this check is located before the validation of value for no/none, the following operation isn't allowed. is it ok? although it makes no sense, it doesn't produce any invalid combination.

      $ touch bar
      $ chattr +C bar
      $ lsattr bar
      ---------------C-- bar
      $ setfattr -n btrfs.compression -v no bar
      setfattr: bar: Invalid argument

Thanks.

#4 lore raw
In reply to Chung-Chiang Cheng
Quoted 5 lines · Chung-Chiang Cheng
fs/btrfs/props.c +1 -1
-static int prop_compression_validate(const char *value, size_t len)
+static int prop_compression_validate(struct inode *inode, const char *value, size_t len)
 {
 	if (!value)
 		return 0;

fs/btrfs/props.c int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path) · line 266

Quoted 10 lines
fs/btrfs/props.c +2 -3
 
-	if (btrfs_compress_is_valid_type(value, len))
-		return 0;
-
 	if ((len == 2 && strncmp("no", value, 2) == 0) ||
 	    (len == 4 && strncmp("none", value, 4) == 0))
 		return 0;
 
+	if (!inode || BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
+		return -EINVAL;

I think the nodatacow check should be the first one, before the validation of value for "no" or "none", it's logically the same as the btrfs_compress_is_valid_type.

Also, should there be a check for NODATASUM? The checks in the property should do the same as the normal compression code, see eg. inode_can_compress. It's an internal helper so it would be cleaner to have it exported and reuse here instead of duplicatin the conditions.

For a minimal fix I'd be OK with the code duplication as this would need to go to stable, so "fixes before cleanups".