Skip to contentSkip to conversation

erofs: code cleanup by removing ifdef macro surrounding

PATCH In review Chengguang Xu 3 participants 8 messages May 26, 2020 – May 27, 2020 EROFS file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 erofs: code cleanup by removing ifdef macro surrounding

Define erofs_listxattr and erofs_xattr_handlers to NULL when CONFIG_EROFS_FS_XATTR is not enabled, then we can remove many ugly ifdef macros in the code.

---
Only compile tested.

 fs/erofs/inode.c | 6 ------
 fs/erofs/namei.c | 2 --
 fs/erofs/super.c | 4 +---
 fs/erofs/xattr.h | 7 ++-----
 4 files changed, 3 insertions(+), 16 deletions(-)
fs/erofs/inode.c +0 -6
index 3350ab65d892..7dd4bbe9674f 100644
@@ -311,27 +311,21 @@ int erofs_getattr(const struct path *path, struct kstat *stat,
 
 const struct inode_operations erofs_generic_iops = {
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
 const struct inode_operations erofs_symlink_iops = {
 	.get_link = page_get_link,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
 const struct inode_operations erofs_fast_symlink_iops = {
 	.get_link = simple_get_link,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
fs/erofs/namei.c +0 -2
index 3abbecbf73de..52f201e03c62 100644
@@ -244,9 +244,7 @@ static struct dentry *erofs_lookup(struct inode *dir,
 const struct inode_operations erofs_dir_iops = {
 	.lookup = erofs_lookup,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
fs/erofs/super.c +1 -3
index b514c67e5fc2..8e46d204a0c2 100644
@@ -408,10 +408,8 @@ static int erofs_fill_super(struct super_block *sb, void *data, int silent)
 	sb->s_time_gran = 1;
 
 	sb->s_op = &erofs_sops;
-
-#ifdef CONFIG_EROFS_FS_XATTR
 	sb->s_xattr = erofs_xattr_handlers;
-#endif
+
 	/* set erofs default mount options */
 	erofs_default_options(sbi);
 
fs/erofs/xattr.h +2 -5
index 50966f1c676e..e4e5093f012c 100644
@@ -76,11 +76,8 @@ static inline int erofs_getxattr(struct inode *inode, int index,
 	return -EOPNOTSUPP;
 }
 
-static inline ssize_t erofs_listxattr(struct dentry *dentry,
-				      char *buffer, size_t buffer_size)
-{
-	return -EOPNOTSUPP;
-}
+#define erofs_listxattr (NULL)
+#define erofs_xattr_handlers (NULL)
 #endif	/* !CONFIG_EROFS_FS_XATTR */
 
 #ifdef CONFIG_EROFS_FS_POSIX_ACL
Signature · 1 line
2.20.1
Reviewed-by on v1 1/1
#1 lore raw

Hi Chengguang,

Quoted 19 lines · Chengguang Xu

Define erofs_listxattr and erofs_xattr_handlers to NULL when CONFIG_EROFS_FS_XATTR is not enabled, then we can remove many ugly ifdef macros in the code.

---
Only compile tested.

 fs/erofs/inode.c | 6 ------
 fs/erofs/namei.c | 2 --
 fs/erofs/super.c | 4 +---
 fs/erofs/xattr.h | 7 ++-----
 4 files changed, 3 insertions(+), 16 deletions(-)
fs/erofs/inode.c +0 -0
index 3350ab65d892..7dd4bbe9674f 100644

fs/erofs/inode.c int erofs_getattr(const struct path *path, struct kstat *stat, · line 316

Quoted 6 lines
fs/erofs/inode.c +0 -2
@@ -311,27 +311,21 @@ int erofs_getattr(const struct path *path, struct kstat *stat,
 
 const struct inode_operations erofs_generic_iops = {
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif

It seems equivalent. And it seems ext2 and f2fs behave in the same way... But I'm not sure whether we'd return 0 (if I didn't see fs/xattr.c by mistake) or -EOPNOTSUPP here... Some thoughts about this?

Anyway, I'm fine with that if return 0 is okay here, but I'd like to know your and Chao's thoughts about this... I will play with it later as well.

Thanks,
Gao Xiang

Quoted 73 lines
fs/erofs/inode.c +0 -4
 	.get_acl = erofs_get_acl,
 };
 
 const struct inode_operations erofs_symlink_iops = {
 	.get_link = page_get_link,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
 const struct inode_operations erofs_fast_symlink_iops = {
 	.get_link = simple_get_link,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
fs/erofs/namei.c +0 -2
index 3abbecbf73de..52f201e03c62 100644
@@ -244,9 +244,7 @@ static struct dentry *erofs_lookup(struct inode *dir,
 const struct inode_operations erofs_dir_iops = {
 	.lookup = erofs_lookup,
 	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
 	.listxattr = erofs_listxattr,
-#endif
 	.get_acl = erofs_get_acl,
 };
 
fs/erofs/super.c +1 -3
index b514c67e5fc2..8e46d204a0c2 100644
@@ -408,10 +408,8 @@ static int erofs_fill_super(struct super_block *sb, void *data, int silent)
 	sb->s_time_gran = 1;
 
 	sb->s_op = &erofs_sops;
-
-#ifdef CONFIG_EROFS_FS_XATTR
 	sb->s_xattr = erofs_xattr_handlers;
-#endif
+
 	/* set erofs default mount options */
 	erofs_default_options(sbi);
 
fs/erofs/xattr.h +2 -5
index 50966f1c676e..e4e5093f012c 100644
@@ -76,11 +76,8 @@ static inline int erofs_getxattr(struct inode *inode, int index,
 	return -EOPNOTSUPP;
 }
 
-static inline ssize_t erofs_listxattr(struct dentry *dentry,
-				      char *buffer, size_t buffer_size)
-{
-	return -EOPNOTSUPP;
-}
+#define erofs_listxattr (NULL)
+#define erofs_xattr_handlers (NULL)
 #endif	/* !CONFIG_EROFS_FS_XATTR */
 
 #ifdef CONFIG_EROFS_FS_POSIX_ACL
Signature · 1 line
2.20.1
author
#2 lore raw
Quoted 23 lines · Gao Xiang

Hi Chengguang,

Quoted 20 lines · Chengguang Xu

Define erofs_listxattr and erofs_xattr_handlers to NULL when CONFIG_EROFS_FS_XATTR is not enabled, then we can remove many ugly ifdef macros in the code.

---
Only compile tested.

  fs/erofs/inode.c | 6 ------
  fs/erofs/namei.c | 2 --
  fs/erofs/super.c | 4 +---
  fs/erofs/xattr.h | 7 ++-----
  4 files changed, 3 insertions(+), 16 deletions(-)
fs/erofs/inode.c +0 -0
index 3350ab65d892..7dd4bbe9674f 100644
@@ -311,27 +311,21 @@ int erofs_getattr(const struct path *path, struct kstat *stat,
 
Quoted 7 lines

fs/erofs/inode.c int erofs_getattr(const struct path *path, struct kstat *stat, · line 316

Quoted 5 lines
fs/erofs/inode.c +0 -2
  const struct inode_operations erofs_generic_iops = {
  	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
  	.listxattr = erofs_listxattr,
-#endif

It seems equivalent. And it seems ext2 and f2fs behave in the same way...

I posted similar patch for ext2 and Jack merged to his tree the other day, though that series also included a real bugfix. I also posted similar patch to f2fs, so if erofs and f2fs merge these patches then all three will behave in the same way, ;-)

You may refer below link for the detail.

https://lore.kernel.org/linux-ext4/[email protected]/

Quoted 4 lines

But I'm not sure whether we'd return 0 (if I didn't see fs/xattr.c by mistake) or -EOPNOTSUPP here... Some thoughts about this? >
Anyway, I'm fine with that if return 0 is okay here, but I'd like to know your and Chao's thoughts about this... I will play with it later as well.

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Thanks,
cgxu

#3 lore raw
Quoted 37 lines · cgxu
Quoted 17 lines · Gao Xiang

Hi Chengguang,

Quoted 14 lines · Chengguang Xu

Define erofs_listxattr and erofs_xattr_handlers to NULL when CONFIG_EROFS_FS_XATTR is not enabled, then we can remove many ugly ifdef macros in the code.

---
Only compile tested.

  fs/erofs/inode.c | 6 ------
  fs/erofs/namei.c | 2 --
  fs/erofs/super.c | 4 +---
  fs/erofs/xattr.h | 7 ++-----
  4 files changed, 3 insertions(+), 16 deletions(-)
Quoted 12 lines

fs/erofs/inode.c int erofs_getattr(const struct path *path, struct kstat *stat, · line 315

Quoted 10 lines
fs/erofs/inode.c +0 -2
index 3350ab65d892..7dd4bbe9674f 100644
@@ -311,27 +311,21 @@ int erofs_getattr(const struct path *path, struct kstat *stat,
  const struct inode_operations erofs_generic_iops = {
  	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
  	.listxattr = erofs_listxattr,
-#endif

It seems equivalent. And it seems ext2 and f2fs behave in the same way...

I posted similar patch for ext2 and Jack merged to his tree the other day, though that series also included a real bugfix. I also posted similar patch to f2fs, so if erofs and f2fs merge these patches then all three will behave in the same way, ;-)

Quoted 3 lines

You may refer below link for the detail.

https://lore.kernel.org/linux-ext4/[email protected]/

Thanks for your link...

Quoted 7 lines
Quoted 6 lines

But I'm not sure whether we'd return 0 (if I didn't see fs/xattr.c by mistake) or -EOPNOTSUPP here... Some thoughts about this? >
Anyway, I'm fine with that if return 0 is okay here, but I'd like to know your and Chao's thoughts about this... I will play with it later as well.

Quoted 5 lines

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Yeah, I agree there is no logic change, so I'm fine with the patch. But I'm little worry about if return 0 is actually wrong here...

see the return value at:
http://man7.org/linux/man-pages/man2/listxattr.2.html

Thanks,
Gao Xiang

Quoted 5 lines

Thanks,
cgxu

Reviewed-by
#4 lore raw
Quoted 38 lines · Gao Xiang
Quoted 37 lines · cgxu
Quoted 17 lines · Gao Xiang

Hi Chengguang,

Quoted 14 lines · Chengguang Xu

Define erofs_listxattr and erofs_xattr_handlers to NULL when CONFIG_EROFS_FS_XATTR is not enabled, then we can remove many ugly ifdef macros in the code.

---
Only compile tested.

  fs/erofs/inode.c | 6 ------
  fs/erofs/namei.c | 2 --
  fs/erofs/super.c | 4 +---
  fs/erofs/xattr.h | 7 ++-----
  4 files changed, 3 insertions(+), 16 deletions(-)
Quoted 12 lines

fs/erofs/inode.c int erofs_getattr(const struct path *path, struct kstat *stat, · line 315

Quoted 10 lines
fs/erofs/inode.c +0 -2
index 3350ab65d892..7dd4bbe9674f 100644
@@ -311,27 +311,21 @@ int erofs_getattr(const struct path *path, struct kstat *stat,
  const struct inode_operations erofs_generic_iops = {
  	.getattr = erofs_getattr,
-#ifdef CONFIG_EROFS_FS_XATTR
  	.listxattr = erofs_listxattr,
-#endif

It seems equivalent. And it seems ext2 and f2fs behave in the same way...

I posted similar patch for ext2 and Jack merged to his tree the other day, though that series also included a real bugfix. I also posted similar patch to f2fs, so if erofs and f2fs merge these patches then all three will behave in the same way, ;-)

Quoted 12 lines
Quoted 3 lines

You may refer below link for the detail.

https://lore.kernel.org/linux-ext4/[email protected]/

Thanks for your link...

Quoted 6 lines
Quoted 6 lines

But I'm not sure whether we'd return 0 (if I didn't see fs/xattr.c by mistake) or -EOPNOTSUPP here... Some thoughts about this? >
Anyway, I'm fine with that if return 0 is okay here, but I'd like to know your and Chao's thoughts about this... I will play with it later as well.

I'm okay with this change, please feel free to add:

Quoted 7 lines
Quoted 6 lines

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Quoted 5 lines

Yeah, I agree there is no logic change, so I'm fine with the patch. But I'm little worry about if return 0 is actually wrong here...

see the return value at:
http://man7.org/linux/man-pages/man2/listxattr.2.html

Yeah, I guess vfs should check that whether lower filesystem has set .listxattr callback function to decide to return that value, something like:

static ssize_t
ecryptfs_listxattr(struct dentry *dentry, char *list, size_t size) {
...

	if (!d_inode(lower_dentry)->i_op->listxattr) {
		rc = -EOPNOTSUPP;
		goto out;
	}
...
	rc = d_inode(lower_dentry)->i_op->listxattr(lower_dentry, list, size);

...
}

Quoted 12 lines

Thanks,
Gao Xiang

Quoted 5 lines

Thanks,
cgxu

.

#5 lore raw

On Wed, May 27, 2020 at 09:55:17AM +0800, Chao Yu wrote:

...

Quoted 20 lines
Quoted 7 lines
Quoted 6 lines

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Quoted 5 lines

Yeah, I agree there is no logic change, so I'm fine with the patch. But I'm little worry about if return 0 is actually wrong here...

see the return value at:
http://man7.org/linux/man-pages/man2/listxattr.2.html

Yeah, I guess vfs should check that whether lower filesystem has set .listxattr callback function to decide to return that value, something like:

static ssize_t
ecryptfs_listxattr(struct dentry *dentry, char *list, size_t size) {
...

Quoted 8 lines
	if (!d_inode(lower_dentry)->i_op->listxattr) {
		rc = -EOPNOTSUPP;
		goto out;
	}
...
	rc = d_inode(lower_dentry)->i_op->listxattr(lower_dentry, list, size);

...
}

This approach seems better. ;) Let me recheck all of this. Maybe we could submit a patch to fsdevel for some further advice...

Thanks,
Gao Xiang

Quoted 15 lines
Quoted 14 lines

Thanks,
Gao Xiang

Quoted 5 lines

Thanks,
cgxu

.

author
#6 lore raw
Quoted 33 lines · Gao Xiang

On Wed, May 27, 2020 at 09:55:17AM +0800, Chao Yu wrote:

...

Quoted 20 lines
Quoted 7 lines
Quoted 6 lines

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Quoted 5 lines

Yeah, I agree there is no logic change, so I'm fine with the patch. But I'm little worry about if return 0 is actually wrong here...

see the return value at:
http://man7.org/linux/man-pages/man2/listxattr.2.html

Yeah, I guess vfs should check that whether lower filesystem has set .listxattr callback function to decide to return that value, something like:

static ssize_t
ecryptfs_listxattr(struct dentry *dentry, char *list, size_t size) {
...

Quoted 8 lines
	if (!d_inode(lower_dentry)->i_op->listxattr) {
		rc = -EOPNOTSUPP;
		goto out;
	}
...
	rc = d_inode(lower_dentry)->i_op->listxattr(lower_dentry, list, size);

...
}

Quoted 3 lines

This approach seems better. ;) Let me recheck all of this. Maybe we could submit a patch to fsdevel for some further advice...

I agree that doing the check in vfs layer looks tidy and unified. However, we should be aware this change may break user space tools.

Thanks,
cgxu

#7 lore raw
Quoted 34 lines · cgxu
Quoted 33 lines · Gao Xiang

On Wed, May 27, 2020 at 09:55:17AM +0800, Chao Yu wrote:

...

Quoted 28 lines
Quoted 7 lines
Quoted 6 lines

Originally, we set erofs_listxattr to ->listxattr only when the config macro CONFIG_EROFS_FS_XATTR is enabled, it means that erofs_listxattr() never returns -EOPNOTSUPP in any case, so actually there is no logic change here, right?

Quoted 5 lines

Yeah, I agree there is no logic change, so I'm fine with the patch. But I'm little worry about if return 0 is actually wrong here...

see the return value at:
http://man7.org/linux/man-pages/man2/listxattr.2.html

Yeah, I guess vfs should check that whether lower filesystem has set .listxattr callback function to decide to return that value, something like:

static ssize_t
ecryptfs_listxattr(struct dentry *dentry, char *list, size_t size) {
...

	if (!d_inode(lower_dentry)->i_op->listxattr) {
		rc = -EOPNOTSUPP;
		goto out;
	}
...
	rc = d_inode(lower_dentry)->i_op->listxattr(lower_dentry, list, size);

...
}

Quoted 6 lines
Quoted 3 lines

This approach seems better. ;) Let me recheck all of this. Maybe we could submit a patch to fsdevel for some further advice...

I agree that doing the check in vfs layer looks tidy and unified. However, we should be aware this change may break user space tools.

I think I already sorted out the reason, it actually seems a regression due to multiple commits, let me try to submit a patch for some advice...

Thanks,
Gao Xiang

Quoted 5 lines

Thanks,
cgxu