common implementation of dirent file types

RFC PATCH In review Phillip Potter 4 participants 7 messages Oct 23, 2018 – Oct 25, 2018

Subsystems (2)

VFS Btrfs file system

lore

Share & tools

lore

Phillip Potter

author

2018-10-23 20:19 UTC #0 lore raw

cover common implementation of dirent file types

This cleanup series is a respin of Amir Goldstein's work, created in late 2016. It removes several instances of duplicated code. Most of the duplication dates back to git pre-historic era.

The controversial aspect of this cleanup is that it uses common code to handle file system specific on-disk format bits. All 7 file systems use a single byte to store dirent file type on-disk and all of them use the same conversion routines from i_mode to 4bits DT_* constants to 3bits on-disk FT_* constants.

Patch 1 places a common implementation in file_type.h and add some useful conversion helpers.

Patches 2-3 make use of some helpers in ufs and hfsplus without any on-disk implications.

Patches 4-10 replace the specific implementations in ext2, exofs, ext4, ocfs2, f2fs, nilfs and btrfs with the common implementation. These patches also now include compile-time checks to ensure that the file system specific on-disk bits are equivalent to the common implementation FT_* bits. These compile-time checks are only included once per file system, as my reasoning is that regardless of their location, the build will fail/succeed anyway.

In addition, where needed (for patches which no longer apply), I've rebased them to apply to the newest 4.19 kernel sources. Each patch is independent of the others, except for the common implementation itself which they all depend on.

I would love feedback as newish kernel contributor, particularly from the original author Amir who suggested this task for me. I hope the various subsystem/fs maintainers will see fit to accept this work also.

Phillip Potter (10):
fs: common implementation of file type conversions
ufs: use fs_umode_to_dtype() helper
hfsplus: use fs_umode_to_dtype() helper
ext2: use common file type conversion
exofs: use common file type conversion
ext4: use common file type conversion
ocfs2: use common file type conversion
f2fs: use common file type conversion
nilfs2: use common file type conversion
btrfs: use common file type conversion

---
 MAINTAINERS                        |   1 +
 fs/btrfs/btrfs_inode.h             |   2 -
 fs/btrfs/delayed-inode.c           |   2 +-
 fs/btrfs/inode.c                   |  35 +++++-----
 fs/exofs/dir.c                     |  49 +++++--------
 fs/ext2/dir.c                      |  51 ++++++--------
 fs/ext4/ext4.h                     |  35 +++++-----
 fs/f2fs/dir.c                      |  43 +++++-------
 fs/f2fs/inline.c                   |   2 +-
 fs/hfsplus/dir.c                   |  16 +----
 fs/nilfs2/dir.c                    |  54 +++++----------
 fs/ocfs2/dir.c                     |  32 +++++----
 fs/ocfs2/ocfs2_fs.h                |  13 +---
 fs/ufs/util.h                      |  29 +-------
 include/linux/f2fs_fs.h            |   8 ++-
 include/linux/file_type.h          | 108 +++++++++++++++++++++++++++++
 include/linux/fs.h                 |  17 +----
 include/uapi/linux/btrfs_tree.h    |   2 +
 include/uapi/linux/nilfs2_ondisk.h |   1 +
 19 files changed, 256 insertions(+), 244 deletions(-)
 create mode 100644 include/linux/file_type.h

--
2.17.2

Phillip Potter

author

2018-10-23 21:17 UTC #4 lore raw

RFC PATCH v2 10/10 btrfs: use common file type conversion

Deduplicate the btrfs file type conversion implementation.

Original patch by Amir Goldstein.

v2:
- Rebased against Linux 4.19 by Phillip Potter
- Compile-time checks added by Phillip Potter to make sure the BTRFS_FT_x enum values stay same as FT_x values

v1:
- Initial implementation

Signed-off-by Phillip Potter <[email protected]>

---
 fs/btrfs/btrfs_inode.h          |  2 --
 fs/btrfs/delayed-inode.c        |  2 +-
 fs/btrfs/inode.c                | 35 +++++++++++++++++----------------
 include/uapi/linux/btrfs_tree.h |  2 ++
 4 files changed, 21 insertions(+), 20 deletions(-)

fs/btrfs/btrfs_inode.h +0 -2

index 1343ac57b438..c7c6db6b4a35 100644
@@ -196,8 +196,6 @@ struct btrfs_inode {
 	struct inode vfs_inode;
 };
 
-extern unsigned char btrfs_filetype_table[];
-
 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 {
 	return container_of(inode, struct btrfs_inode, vfs_inode);

fs/btrfs/delayed-inode.c +1 -1

index f51b509f2d9b..c1da34e3a775 100644
@@ -1691,7 +1691,7 @@ int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
 		name = (char *)(di + 1);
 		name_len = btrfs_stack_dir_name_len(di);
 
-		d_type = btrfs_filetype_table[di->type];
+		d_type = fs_dtype(di->type);
 		btrfs_disk_key_to_cpu(&location, &di->location);
 
 		over = !dir_emit(ctx, name, name_len,

fs/btrfs/inode.c +18 ~~-17~~

index 3ea5339603cf..4cbdcba4799a 100644
@@ -73,17 +73,6 @@ struct kmem_cache *btrfs_trans_handle_cachep;
 struct kmem_cache *btrfs_path_cachep;
 struct kmem_cache *btrfs_free_space_cachep;
 
-#define S_SHIFT 12
-static const unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
-	[S_IFREG >> S_SHIFT]	= BTRFS_FT_REG_FILE,
-	[S_IFDIR >> S_SHIFT]	= BTRFS_FT_DIR,
-	[S_IFCHR >> S_SHIFT]	= BTRFS_FT_CHRDEV,
-	[S_IFBLK >> S_SHIFT]	= BTRFS_FT_BLKDEV,
-	[S_IFIFO >> S_SHIFT]	= BTRFS_FT_FIFO,
-	[S_IFSOCK >> S_SHIFT]	= BTRFS_FT_SOCK,
-	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
-};
-
 static int btrfs_setsize(struct inode *inode, struct iattr *attr);
 static int btrfs_truncate(struct inode *inode, bool skip_writeback);
 static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);

@@ -5803,10 +5792,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 	return d_splice_alias(inode, dentry);
 }
 
-unsigned char btrfs_filetype_table[] = {
-	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
-};
-
 /*
  * All this infrastructure exists because dir_emit can fault, and we are holding
  * the tree lock when doing readdir.  For now just allocate a buffer and copy

@@ -5945,7 +5930,23 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 		name_ptr = (char *)(entry + 1);
 		read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1),
 				   name_len);
-		put_unaligned(btrfs_filetype_table[btrfs_dir_type(leaf, di)],
+
+		/*
+		 * compile-time asserts that generic FT_x types
+		 * still match BTRFS_FT_x types - no need to list
+		 * in other functions as well as build will
+		 * fail either way
+		 */
+		BUILD_BUG_ON(BTRFS_FT_UNKNOWN != FT_UNKNOWN);
+		BUILD_BUG_ON(BTRFS_FT_REG_FILE != FT_REG_FILE);
+		BUILD_BUG_ON(BTRFS_FT_DIR != FT_DIR);
+		BUILD_BUG_ON(BTRFS_FT_CHRDEV != FT_CHRDEV);
+		BUILD_BUG_ON(BTRFS_FT_BLKDEV != FT_BLKDEV);
+		BUILD_BUG_ON(BTRFS_FT_FIFO != FT_FIFO);
+		BUILD_BUG_ON(BTRFS_FT_SOCK != FT_SOCK);
+		BUILD_BUG_ON(BTRFS_FT_SYMLINK != FT_SYMLINK);
+
+		put_unaligned(fs_dtype(btrfs_dir_type(leaf, di)),
 				&entry->type);
 		btrfs_dir_item_key_to_cpu(leaf, di, &location);
 		put_unaligned(location.objectid, &entry->ino);

@@ -6350,7 +6351,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 
 static inline u8 btrfs_inode_type(struct inode *inode)
 {
-	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
+	return fs_umode_to_ftype(inode->i_mode);
 }
 
 /*

include/uapi/linux/btrfs_tree.h +2 -0

index aff1356c2bb8..5747cffa09fa 100644
@@ -307,6 +307,8 @@
  *
  * Used by:
  * struct btrfs_dir_item.type
+ *
+ * Values 0..7 should match common file type values in file_type.h.
  */
 #define BTRFS_FT_UNKNOWN	0
 #define BTRFS_FT_REG_FILE	1

Signature · 1 line

2.17.2

Amir Goldstein

2018-10-24 06:43 UTC #1 lore raw

In reply to Phillip Potter

Quoted 7 lines · Phillip Potter

This cleanup series is a respin of Amir Goldstein's work, created in late 2016. It removes several instances of duplicated code. Most of the duplication dates back to git pre-historic era.
The controversial aspect of this cleanup is that it uses common code to handle file system specific on-disk format bits.

We handled this aspect with care by placing compile time checks.

Quoted 18 lines

All 7 file systems use a single byte to store dirent file type on-disk and all of them use the same conversion routines from i_mode to 4bits DT_* constants to 3bits on-disk FT_* constants.
Patch 1 places a common implementation in file_type.h and add some useful conversion helpers.
Patches 2-3 make use of some helpers in ufs and hfsplus without any on-disk implications.
Patches 4-10 replace the specific implementations in ext2, exofs, ext4, ocfs2, f2fs, nilfs and btrfs with the common implementation. These patches also now include compile-time checks to ensure that the file system specific on-disk bits are equivalent to the common implementation FT_* bits. These compile-time checks are only included once per file system, as my reasoning is that regardless of their location, the build will fail/succeed anyway.

Quoted 5 lines

In addition, where needed (for patches which no longer apply), I've rebased them to apply to the newest 4.19 kernel sources. Each patch is independent of the others, except for the common implementation itself which they all depend on.

This is still RFC, but since this is an exercise let's talk about social engineering. First (now) you want to gather ACKs and review comments from fs developers/maintainers.

Once all comments are addressed and assuming no objections, you have two possible ways forward:

1. Wait for VFS maintainer to take the common code and patches ACKed by individual maintainers (Al practically maintains some of the small fs himself). Al's backlog is long, so this may be a long wait.

2. Ask a filesystem maintainer to carry the common patch along with his specific fs patch through his tree and let other maintainers apply the patch at their own will on the following development cycle.
The ext2 tree would be a good candidate for such a move, if Jan accepts this change, considering that all other fs copied the buggy code from ext2.

Quoted 5 lines

I would love feedback as newish kernel contributor, particularly from the original author Amir who suggested this task for me. I hope the various subsystem/fs maintainers will see fit to accept this work also.

I usually prefer to have the revisions log in the cover letter for a 10 patch series, much less work and more appropriate when revision log looks mostly the same on all patches, e.g.:

v2:
- Rebased against Linux 4.19 by Phillip Potter
- This version does not remove filesystem specific XXX_FT_x definitions, as these values are now used in compile-time checks added by Phillip Potter to make sure they remain the same as FT_x values
- Removed xfs patch (a variant of original patch has already been applied)
- Anything else relevant to the context of the series...

v1:
- Initial implementation by Amir Goldstein:
https://marc.info/?l=linux-fsdevel&m=148217829301701&w=2

Quoted 37 lines

---
 MAINTAINERS                        |   1 +
 fs/btrfs/btrfs_inode.h             |   2 -
 fs/btrfs/delayed-inode.c           |   2 +-
 fs/btrfs/inode.c                   |  35 +++++-----
 fs/exofs/dir.c                     |  49 +++++--------
 fs/ext2/dir.c                      |  51 ++++++--------
 fs/ext4/ext4.h                     |  35 +++++-----
 fs/f2fs/dir.c                      |  43 +++++-------
 fs/f2fs/inline.c                   |   2 +-
 fs/hfsplus/dir.c                   |  16 +----
 fs/nilfs2/dir.c                    |  54 +++++----------
 fs/ocfs2/dir.c                     |  32 +++++----
 fs/ocfs2/ocfs2_fs.h                |  13 +---
 fs/ufs/util.h                      |  29 +-------
 include/linux/f2fs_fs.h            |   8 ++-
 include/linux/file_type.h          | 108 +++++++++++++++++++++++++++++
 include/linux/fs.h                 |  17 +----
 include/uapi/linux/btrfs_tree.h    |   2 +
 include/uapi/linux/nilfs2_ondisk.h |   1 +
 19 files changed, 256 insertions(+), 244 deletions(-)
 create mode 100644 include/linux/file_type.h

--
2.17.2

David Sterba

2018-10-24 10:11 UTC #5 lore raw

Re: [RFC][PATCH v2 10/10] btrfs: use common file type conversion

In reply to Phillip Potter

Quoted 1 line · Phillip Potter

Deduplicate the btrfs file type conversion implementation.

The per-filesystem changelogs need a brief explanation why this is done, like "Filesystems that use the same filetypes as defined by POSIX do not need to define their own versions and use the common defines in file_type.h", rephrase at you like.

Quoted 79 lines

Original patch by Amir Goldstein.

v2:
- Rebased against Linux 4.19 by Phillip Potter
- Compile-time checks added by Phillip Potter to make sure the BTRFS_FT_x enum values stay same as FT_x values

v1:
- Initial implementation

Signed-off-by Phillip Potter <[email protected]>

---
 fs/btrfs/btrfs_inode.h          |  2 --
 fs/btrfs/delayed-inode.c        |  2 +-
 fs/btrfs/inode.c                | 35 +++++++++++++++++----------------
 include/uapi/linux/btrfs_tree.h |  2 ++
 4 files changed, 21 insertions(+), 20 deletions(-)

fs/btrfs/btrfs_inode.h +0 -2

index 1343ac57b438..c7c6db6b4a35 100644
@@ -196,8 +196,6 @@ struct btrfs_inode {
 	struct inode vfs_inode;
 };
 
-extern unsigned char btrfs_filetype_table[];
-
 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 {
 	return container_of(inode, struct btrfs_inode, vfs_inode);

fs/btrfs/delayed-inode.c +1 -1

index f51b509f2d9b..c1da34e3a775 100644
@@ -1691,7 +1691,7 @@ int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
 		name = (char *)(di + 1);
 		name_len = btrfs_stack_dir_name_len(di);
 
-		d_type = btrfs_filetype_table[di->type];
+		d_type = fs_dtype(di->type);
 		btrfs_disk_key_to_cpu(&location, &di->location);
 
 		over = !dir_emit(ctx, name, name_len,

fs/btrfs/inode.c +0 ~~-15~~

index 3ea5339603cf..4cbdcba4799a 100644
@@ -73,17 +73,6 @@ struct kmem_cache *btrfs_trans_handle_cachep;
 struct kmem_cache *btrfs_path_cachep;
 struct kmem_cache *btrfs_free_space_cachep;
 
-#define S_SHIFT 12
-static const unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
-	[S_IFREG >> S_SHIFT]	= BTRFS_FT_REG_FILE,
-	[S_IFDIR >> S_SHIFT]	= BTRFS_FT_DIR,
-	[S_IFCHR >> S_SHIFT]	= BTRFS_FT_CHRDEV,
-	[S_IFBLK >> S_SHIFT]	= BTRFS_FT_BLKDEV,
-	[S_IFIFO >> S_SHIFT]	= BTRFS_FT_FIFO,
-	[S_IFSOCK >> S_SHIFT]	= BTRFS_FT_SOCK,
-	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
-};
-
 static int btrfs_setsize(struct inode *inode, struct iattr *attr);
 static int btrfs_truncate(struct inode *inode, bool skip_writeback);
 static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);

@@ -5803,10 +5792,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 	return d_splice_alias(inode, dentry);
 }
 
-unsigned char btrfs_filetype_table[] = {
-	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
-};
-
 /*
  * All this infrastructure exists because dir_emit can fault, and we are holding
  * the tree lock when doing readdir.  For now just allocate a buffer and copy

@@ -5945,7 +5930,23 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 		name_ptr = (char *)(entry + 1);

fs/btrfs/inode.c static int btrfs_real_readdir(struct file *file, struct dir_context *ctx) · line 5939

Quoted 10 lines

fs/btrfs/inode.c +7 -1

 		read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1),
 				   name_len);
-		put_unaligned(btrfs_filetype_table[btrfs_dir_type(leaf, di)],
+
+		/*
+		 * compile-time asserts that generic FT_x types
+		 * still match BTRFS_FT_x types - no need to list
+		 * in other functions as well as build will
+		 * fail either way
+		 */

Please format comments to 80 columns.

Quoted 27 lines

fs/btrfs/inode.c +11 -1

+		BUILD_BUG_ON(BTRFS_FT_UNKNOWN != FT_UNKNOWN);
+		BUILD_BUG_ON(BTRFS_FT_REG_FILE != FT_REG_FILE);
+		BUILD_BUG_ON(BTRFS_FT_DIR != FT_DIR);
+		BUILD_BUG_ON(BTRFS_FT_CHRDEV != FT_CHRDEV);
+		BUILD_BUG_ON(BTRFS_FT_BLKDEV != FT_BLKDEV);
+		BUILD_BUG_ON(BTRFS_FT_FIFO != FT_FIFO);
+		BUILD_BUG_ON(BTRFS_FT_SOCK != FT_SOCK);
+		BUILD_BUG_ON(BTRFS_FT_SYMLINK != FT_SYMLINK);
+
+		put_unaligned(fs_dtype(btrfs_dir_type(leaf, di)),
 				&entry->type);
 		btrfs_dir_item_key_to_cpu(leaf, di, &location);
 		put_unaligned(location.objectid, &entry->ino);

@@ -6350,7 +6351,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 
 static inline u8 btrfs_inode_type(struct inode *inode)
 {
-	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
+	return fs_umode_to_ftype(inode->i_mode);
 }
 
 /*

include/uapi/linux/btrfs_tree.h +0 -0

index aff1356c2bb8..5747cffa09fa 100644

include/uapi/linux/btrfs_tree.h line 311

Quoted 5 lines

include/uapi/linux/btrfs_tree.h +2 -0

@@ -307,6 +307,8 @@
  *
  * Used by:
  * struct btrfs_dir_item.type
+ *
+ * Values 0..7 should match common file type values in file_type.h.

I think it's rather 'must' than 'should' as there are the compile-time checks.

Phillip Potter

author

2018-10-24 13:28 UTC #6 lore raw

Re: [RFC][PATCH v2 10/10] btrfs: use common file type conversion

Quoted 121 lines · David Sterba

Quoted 1 line · Phillip Potter

Deduplicate the btrfs file type conversion implementation.

Quoted 79 lines

Original patch by Amir Goldstein.

v2:
- Rebased against Linux 4.19 by Phillip Potter
- Compile-time checks added by Phillip Potter to make sure the BTRFS_FT_x enum values stay same as FT_x values

v1:
- Initial implementation

Signed-off-by Phillip Potter <[email protected]>

---
 fs/btrfs/btrfs_inode.h          |  2 --
 fs/btrfs/delayed-inode.c        |  2 +-
 fs/btrfs/inode.c                | 35 +++++++++++++++++----------------
 include/uapi/linux/btrfs_tree.h |  2 ++
 4 files changed, 21 insertions(+), 20 deletions(-)

fs/btrfs/btrfs_inode.h +0 -2

index 1343ac57b438..c7c6db6b4a35 100644
@@ -196,8 +196,6 @@ struct btrfs_inode {
 	struct inode vfs_inode;
 };
 
-extern unsigned char btrfs_filetype_table[];
-
 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 {
 	return container_of(inode, struct btrfs_inode, vfs_inode);

fs/btrfs/delayed-inode.c +1 -1

index f51b509f2d9b..c1da34e3a775 100644
@@ -1691,7 +1691,7 @@ int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
 		name = (char *)(di + 1);
 		name_len = btrfs_stack_dir_name_len(di);
 
-		d_type = btrfs_filetype_table[di->type];
+		d_type = fs_dtype(di->type);
 		btrfs_disk_key_to_cpu(&location, &di->location);
 
 		over = !dir_emit(ctx, name, name_len,

fs/btrfs/inode.c +0 ~~-15~~

index 3ea5339603cf..4cbdcba4799a 100644
@@ -73,17 +73,6 @@ struct kmem_cache *btrfs_trans_handle_cachep;
 struct kmem_cache *btrfs_path_cachep;
 struct kmem_cache *btrfs_free_space_cachep;
 
-#define S_SHIFT 12
-static const unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
-	[S_IFREG >> S_SHIFT]	= BTRFS_FT_REG_FILE,
-	[S_IFDIR >> S_SHIFT]	= BTRFS_FT_DIR,
-	[S_IFCHR >> S_SHIFT]	= BTRFS_FT_CHRDEV,
-	[S_IFBLK >> S_SHIFT]	= BTRFS_FT_BLKDEV,
-	[S_IFIFO >> S_SHIFT]	= BTRFS_FT_FIFO,
-	[S_IFSOCK >> S_SHIFT]	= BTRFS_FT_SOCK,
-	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
-};
-
 static int btrfs_setsize(struct inode *inode, struct iattr *attr);
 static int btrfs_truncate(struct inode *inode, bool skip_writeback);
 static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);

@@ -5803,10 +5792,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 	return d_splice_alias(inode, dentry);
 }
 
-unsigned char btrfs_filetype_table[] = {
-	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
-};
-
 /*
  * All this infrastructure exists because dir_emit can fault, and we are holding
  * the tree lock when doing readdir.  For now just allocate a buffer and copy

@@ -5945,7 +5930,23 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 		name_ptr = (char *)(entry + 1);

fs/btrfs/inode.c static int btrfs_real_readdir(struct file *file, struct dir_context *ctx) · line 5939

Quoted 10 lines

fs/btrfs/inode.c +7 -1

 		read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1),
 				   name_len);
-		put_unaligned(btrfs_filetype_table[btrfs_dir_type(leaf, di)],
+
+		/*
+		 * compile-time asserts that generic FT_x types
+		 * still match BTRFS_FT_x types - no need to list
+		 * in other functions as well as build will
+		 * fail either way
+		 */

Please format comments to 80 columns.

Quoted 21 lines

fs/btrfs/inode.c +11 -1

+		BUILD_BUG_ON(BTRFS_FT_UNKNOWN != FT_UNKNOWN);
+		BUILD_BUG_ON(BTRFS_FT_REG_FILE != FT_REG_FILE);
+		BUILD_BUG_ON(BTRFS_FT_DIR != FT_DIR);
+		BUILD_BUG_ON(BTRFS_FT_CHRDEV != FT_CHRDEV);
+		BUILD_BUG_ON(BTRFS_FT_BLKDEV != FT_BLKDEV);
+		BUILD_BUG_ON(BTRFS_FT_FIFO != FT_FIFO);
+		BUILD_BUG_ON(BTRFS_FT_SOCK != FT_SOCK);
+		BUILD_BUG_ON(BTRFS_FT_SYMLINK != FT_SYMLINK);
+
+		put_unaligned(fs_dtype(btrfs_dir_type(leaf, di)),
 				&entry->type);
 		btrfs_dir_item_key_to_cpu(leaf, di, &location);
 		put_unaligned(location.objectid, &entry->ino);

@@ -6350,7 +6351,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 
 static inline u8 btrfs_inode_type(struct inode *inode)
 {
-	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
+	return fs_umode_to_ftype(inode->i_mode);
 }

Quoted 14 lines

include/uapi/linux/btrfs_tree.h line 311
Quoted 11 lines
fs/btrfs/inode.c +0 -0
 /*
include/uapi/linux/btrfs_tree.h +2 -0
index aff1356c2bb8..5747cffa09fa 100644
@@ -307,6 +307,8 @@
  *
  * Used by:
  * struct btrfs_dir_item.type
+ *
+ * Values 0..7 should match common file type values in file_type.h.
I think it's rather 'must' than 'should' as there are the compile-time checks.

Dear David,

Good points, thank you, I will make these changes and republish as part of a new series.

Regards,
Phil

Jan Kara

reviewer

2018-10-25 11:08 UTC #2 lore raw

In reply to Amir Goldstein

Quoted 7 lines · Amir Goldstein

Once all comments are addressed and assuming no objections, you have two possible ways forward:
1. Wait for VFS maintainer to take the common code and patches ACKed by individual maintainers (Al practically maintains some of the small fs himself). Al's backlog is long, so this may be a long wait.

Quoted 5 lines

2. Ask a filesystem maintainer to carry the common patch along with his specific fs patch through his tree and let other maintainers apply the patch at their own will on the following development cycle.
The ext2 tree would be a good candidate for such a move, if Jan accepts this change, considering that all other fs copied the buggy code from ext2.

I can take these patches through my tree if Al takes too long to respond (but he apparently already had a look at the patches which is a good sign ;) and if patches are reviewed by respective fs maintainers.

Honza

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Phillip Potter

author

2018-10-25 21:20 UTC #3 lore raw

Quoted 14 lines · Jan Kara

Quoted 7 lines · Amir Goldstein
Once all comments are addressed and assuming no objections, you have two possible ways forward:
1. Wait for VFS maintainer to take the common code and patches ACKed by individual maintainers (Al practically maintains some of the small fs himself). Al's backlog is long, so this may be a long wait.
Quoted 5 lines
2. Ask a filesystem maintainer to carry the common patch along with his specific fs patch through his tree and let other maintainers apply the patch at their own will on the following development cycle.
The ext2 tree would be a good candidate for such a move, if Jan accepts this change, considering that all other fs copied the buggy code from ext2.

Quoted 9 lines

I can take these patches through my tree if Al takes too long to respond (but he apparently already had a look at the patches which is a good sign ;) and if patches are reviewed by respective fs maintainers.
Honza
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR

Dear Jan,

Thank you - and thank you Al as well for taking a look :-) I will publish a new series later on (possibly tomorrow) with all the feedback I've gathered.

Regards,
Phil