f2fs: support RENAME_WHITEOUT

As the description of rename in manual, RENAME_WHITEOUT is a special operation that only makes sense for overlay/union type filesystem.

When performing rename with RENAME_WHITEOUT, dst will be replace with src, and meanwhile, a 'whiteout' will be create with name of src.

A "whiteout" is designed to be a char device with 0,0 device number, it has specially meaning for stackable filesystem. In these filesystems, there are multiple layers exist, and only top of these can be modified. So a whiteout in top layer is used to hide a corresponding file in lower layer, as well removal of whiteout will make the file appear.

Now in overlayfs, when we rename a file which is exist in lower layer, it will be copied up to upper if it is not on upper layer yet, and then rename it on upper layer, source file will be whiteouted to hide corresponding file in lower layer at the same time.

So in upper layer filesystem, implementation of RENAME_WHITEOUT provide a atomic operation for stackable filesystem to support rename operation.

There are multiple ways to implement RENAME_WHITEOUT in log of this commit:
7dcf5c3e4527 ("xfs: add RENAME_WHITEOUT support") which pointed out by Dave Chinner.

For now, we just try to follow the way that xfs/ext4 use.

Signed-off-by Chao Yu <[email protected]>

---
 fs/f2fs/namei.c | 140 ++++++++++++++++++++++++++++++++++++--------------------
 1 file changed, 90 insertions(+), 50 deletions(-)

fs/f2fs/namei.c +30 -0

index 16b74da..bed0cb0 100644
@@ -510,14 +510,80 @@ out:
 	return err;
 }
 
+static int __f2fs_tmpfile(struct inode *dir, struct dentry *dentry,
+					umode_t mode, struct inode **whiteout)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
+	struct inode *inode;
+	int err;
+
+	inode = f2fs_new_inode(dir, mode);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	if (whiteout) {
+		init_special_inode(inode, inode->i_mode, WHITEOUT_DEV);
+		inode->i_op = &f2fs_special_inode_operations;
+	} else {
+		inode->i_op = &f2fs_file_inode_operations;
+		inode->i_fop = &f2fs_file_operations;
+		inode->i_mapping->a_ops = &f2fs_dblock_aops;
+	}
+
+	f2fs_lock_op(sbi);
+	err = acquire_orphan_inode(sbi);
+	if (err)
+		goto out;
+
+	err = f2fs_do_tmpfile(inode, dir);
+	if (err)
+		goto release_out;
+
+	/*

fs/f2fs/namei.c +50 -50

+		*whiteout = inode;
+	} else {
+		d_tmpfile(dentry, inode);
+	}
+	unlock_new_inode(inode);
+	return 0;
+
+release_out:
+	release_orphan_inode(sbi);
+out:
+	handle_failed_inode(inode);
+	return err;
+}
+
+static int f2fs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	return __f2fs_tmpfile(dir, dentry, mode, NULL);
+}
+
+static int f2fs_create_whiteout(struct inode *dir, struct inode **whiteout)
+{
+	return __f2fs_tmpfile(dir, NULL, S_IFCHR | WHITEOUT_MODE, whiteout);
+}
+
 static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
-			struct inode *new_dir, struct dentry *new_dentry)
+			struct inode *new_dir, struct dentry *new_dentry,
+			unsigned int flags)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(old_dir);
 	struct inode *old_inode = d_inode(old_dentry);
 	struct inode *new_inode = d_inode(new_dentry);
+	struct inode *whiteout = NULL;
 	struct page *old_dir_page;
-	struct page *old_page, *new_page;
+	struct page *old_page, *new_page = NULL;
 	struct f2fs_dir_entry *old_dir_entry = NULL;
 	struct f2fs_dir_entry *old_entry;
 	struct f2fs_dir_entry *new_entry;

@@ -543,6 +609,12 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			goto out_old;
 	}
 
+	if (flags & RENAME_WHITEOUT) {
+		err = f2fs_create_whiteout(old_dir, &whiteout);
+		if (err)
+			goto out_dir;
+	}
+
 	if (new_inode) {
 
 		err = -ENOTEMPTY;

@@ -611,8 +683,17 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 
 	f2fs_delete_entry(old_entry, old_page, old_dir, NULL);
 
+	if (whiteout) {
+		whiteout->i_state |= I_LINKABLE;
+		set_inode_flag(F2FS_I(whiteout), FI_INC_LINK);
+		err = f2fs_add_link(old_dentry, whiteout);
+		if (err)
+			goto put_out_dir;
+		whiteout->i_state &= ~I_LINKABLE;
+	}
+
 	if (old_dir_entry) {
-		if (old_dir != new_dir) {
+		if (old_dir != new_dir && !whiteout) {
 			f2fs_set_link(old_inode, old_dir_entry,
 						old_dir_page, new_dir);
 			update_inode_page(old_inode);

@@ -633,8 +714,10 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 
 put_out_dir:
 	f2fs_unlock_op(sbi);
-	f2fs_dentry_kunmap(new_dir, new_page);
-	f2fs_put_page(new_page, 0);
+	if (new_page) {
+		f2fs_dentry_kunmap(new_dir, new_page);
+		f2fs_put_page(new_page, 0);
+	}
 out_dir:
 	if (old_dir_entry) {
 		f2fs_dentry_kunmap(old_inode, old_dir_page);

@@ -805,7 +888,7 @@ static int f2fs_rename2(struct inode *old_dir, struct dentry *old_dentry,
 			struct inode *new_dir, struct dentry *new_dentry,
 			unsigned int flags)
 {
-	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
+	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
 		return -EINVAL;
 
 	if (flags & RENAME_EXCHANGE) {

@@ -816,50 +899,7 @@ static int f2fs_rename2(struct inode *old_dir, struct dentry *old_dentry,
 	 * VFS has already handled the new dentry existence case,
 	 * here, we just deal with "RENAME_NOREPLACE" as regular rename.
 	 */
-	return f2fs_rename(old_dir, old_dentry, new_dir, new_dentry);
-}
-
-static int f2fs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
-{
-	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
-	struct inode *inode;
-	int err;
-
-	inode = f2fs_new_inode(dir, mode);
-	if (IS_ERR(inode))
-		return PTR_ERR(inode);
-
-	inode->i_op = &f2fs_file_inode_operations;
-	inode->i_fop = &f2fs_file_operations;
-	inode->i_mapping->a_ops = &f2fs_dblock_aops;
-
-	f2fs_lock_op(sbi);
-	err = acquire_orphan_inode(sbi);
-	if (err)
-		goto out;
-
-	err = f2fs_do_tmpfile(inode, dir);
-	if (err)
-		goto release_out;
-
-	/*
-	 * add this non-linked tmpfile to orphan list, in this way we could
-	 * remove all unused data of tmpfile after abnormal power-off.
-	 */
-	add_orphan_inode(sbi, inode->i_ino);
-	f2fs_unlock_op(sbi);
-
-	alloc_nid_done(sbi, inode->i_ino);
-
-	d_tmpfile(dentry, inode);
-	unlock_new_inode(inode);
-	return 0;
-
-release_out:
-	release_orphan_inode(sbi);
-out:
-	handle_failed_inode(inode);
-	return err;
+	return f2fs_rename(old_dir, old_dentry, new_dir, new_dentry, flags);
 }
 
 #ifdef CONFIG_F2FS_FS_ENCRYPTION

Signature · 1 line

2.3.3

-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, May 19, 2015 1:51 PM
To: Chao Yu
Cc: Changman Lee; [email protected]; [email protected]
Subject: Re: [PATCH 1/2] f2fs: support RENAME_WHITEOUT

Hi Chao,

Quoted 20 lines · Chao Yu

As the description of rename in manual, RENAME_WHITEOUT is a special operation that only makes sense for overlay/union type filesystem.

When performing rename with RENAME_WHITEOUT, dst will be replace with src, and meanwhile, a 'whiteout' will be create with name of src.

A "whiteout" is designed to be a char device with 0,0 device number, it has specially meaning for stackable filesystem. In these filesystems, there are multiple layers exist, and only top of these can be modified. So a whiteout in top layer is used to hide a corresponding file in lower layer, as well removal of whiteout will make the file appear.

Now in overlayfs, when we rename a file which is exist in lower layer, it will be copied up to upper if it is not on upper layer yet, and then rename it on upper layer, source file will be whiteouted to hide corresponding file in lower layer at the same time.

So in upper layer filesystem, implementation of RENAME_WHITEOUT provide a atomic operation for stackable filesystem to support rename operation.

Quoted 5 lines

There are multiple ways to implement RENAME_WHITEOUT in log of this commit:
7dcf5c3e4527 ("xfs: add RENAME_WHITEOUT support") which pointed out by Dave Chinner.

For now, we just try to follow the way that xfs/ext4 use.

Could you merge the two patches into one?

And, after finishing xfstests, kernel reports missing inode objects.

Could you check it out?

I didn't take a closer look at it.

Quoted 49 lines

Signed-off-by Chao Yu <[email protected]>

---
 fs/f2fs/namei.c | 140 ++++++++++++++++++++++++++++++++++++--------------------
 1 file changed, 90 insertions(+), 50 deletions(-)

fs/f2fs/namei.c +35 -0

index 16b74da..bed0cb0 100644
@@ -510,14 +510,80 @@ out:
 	return err;
 }
 
+static int __f2fs_tmpfile(struct inode *dir, struct dentry *dentry,
+					umode_t mode, struct inode **whiteout)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
+	struct inode *inode;
+	int err;
+
+	inode = f2fs_new_inode(dir, mode);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	if (whiteout) {
+		init_special_inode(inode, inode->i_mode, WHITEOUT_DEV);
+		inode->i_op = &f2fs_special_inode_operations;
+	} else {
+		inode->i_op = &f2fs_file_inode_operations;
+		inode->i_fop = &f2fs_file_operations;
+		inode->i_mapping->a_ops = &f2fs_dblock_aops;
+	}
+
+	f2fs_lock_op(sbi);
+	err = acquire_orphan_inode(sbi);
+	if (err)
+		goto out;
+
+	err = f2fs_do_tmpfile(inode, dir);
+	if (err)
+		goto release_out;
+
+	/*
+	 * add this non-linked tmpfile to orphan list, in this way we could
+	 * remove all unused data of tmpfile after abnormal power-off.
+	 */
+	add_orphan_inode(sbi, inode->i_ino);
+	f2fs_unlock_op(sbi);

fs/f2fs/namei.c out: · line 552

Quoted 5 lines

fs/f2fs/namei.c +5 -0

+
+	alloc_nid_done(sbi, inode->i_ino);
+
+	if (whiteout) {
+		inode_dec_link_count(inode);

Maybe due to this?
We don't need to decrease its link count?

Thanks,

Quoted 147 lines

fs/f2fs/namei.c +50 -51

+		*whiteout = inode;
+	} else {
+		d_tmpfile(dentry, inode);
+	}
+	unlock_new_inode(inode);
+	return 0;
+
+release_out:
+	release_orphan_inode(sbi);
+out:
+	handle_failed_inode(inode);
+	return err;
+}
+
+static int f2fs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	return __f2fs_tmpfile(dir, dentry, mode, NULL);
+}
+
+static int f2fs_create_whiteout(struct inode *dir, struct inode **whiteout)
+{
+	return __f2fs_tmpfile(dir, NULL, S_IFCHR | WHITEOUT_MODE, whiteout);
+}
+
 static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
-			struct inode *new_dir, struct dentry *new_dentry)
+			struct inode *new_dir, struct dentry *new_dentry,
+			unsigned int flags)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(old_dir);
 	struct inode *old_inode = d_inode(old_dentry);
 	struct inode *new_inode = d_inode(new_dentry);
+	struct inode *whiteout = NULL;
 	struct page *old_dir_page;
-	struct page *old_page, *new_page;
+	struct page *old_page, *new_page = NULL;
 	struct f2fs_dir_entry *old_dir_entry = NULL;
 	struct f2fs_dir_entry *old_entry;
 	struct f2fs_dir_entry *new_entry;

@@ -543,6 +609,12 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			goto out_old;
 	}
 
+	if (flags & RENAME_WHITEOUT) {
+		err = f2fs_create_whiteout(old_dir, &whiteout);
+		if (err)
+			goto out_dir;
+	}
+
 	if (new_inode) {
 
 		err = -ENOTEMPTY;

@@ -611,8 +683,17 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 
 	f2fs_delete_entry(old_entry, old_page, old_dir, NULL);
 
+	if (whiteout) {
+		whiteout->i_state |= I_LINKABLE;
+		set_inode_flag(F2FS_I(whiteout), FI_INC_LINK);
+		err = f2fs_add_link(old_dentry, whiteout);
+		if (err)
+			goto put_out_dir;
+		whiteout->i_state &= ~I_LINKABLE;
+	}
+
 	if (old_dir_entry) {
-		if (old_dir != new_dir) {
+		if (old_dir != new_dir && !whiteout) {
 			f2fs_set_link(old_inode, old_dir_entry,
 						old_dir_page, new_dir);
 			update_inode_page(old_inode);

@@ -633,8 +714,10 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 
 put_out_dir:
 	f2fs_unlock_op(sbi);
-	f2fs_dentry_kunmap(new_dir, new_page);
-	f2fs_put_page(new_page, 0);
+	if (new_page) {
+		f2fs_dentry_kunmap(new_dir, new_page);
+		f2fs_put_page(new_page, 0);
+	}
 out_dir:
 	if (old_dir_entry) {
 		f2fs_dentry_kunmap(old_inode, old_dir_page);

@@ -805,7 +888,7 @@ static int f2fs_rename2(struct inode *old_dir, struct dentry *old_dentry,
 			struct inode *new_dir, struct dentry *new_dentry,
 			unsigned int flags)
 {
-	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
+	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
 		return -EINVAL;
 
 	if (flags & RENAME_EXCHANGE) {

@@ -816,50 +899,7 @@ static int f2fs_rename2(struct inode *old_dir, struct dentry *old_dentry,
 	 * VFS has already handled the new dentry existence case,
 	 * here, we just deal with "RENAME_NOREPLACE" as regular rename.
 	 */
-	return f2fs_rename(old_dir, old_dentry, new_dir, new_dentry);
-}
-
-static int f2fs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
-{
-	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
-	struct inode *inode;
-	int err;
-
-	inode = f2fs_new_inode(dir, mode);
-	if (IS_ERR(inode))
-		return PTR_ERR(inode);
-
-	inode->i_op = &f2fs_file_inode_operations;
-	inode->i_fop = &f2fs_file_operations;
-	inode->i_mapping->a_ops = &f2fs_dblock_aops;
-
-	f2fs_lock_op(sbi);
-	err = acquire_orphan_inode(sbi);
-	if (err)
-		goto out;
-
-	err = f2fs_do_tmpfile(inode, dir);
-	if (err)
-		goto release_out;
-
-	/*
-	 * add this non-linked tmpfile to orphan list, in this way we could
-	 * remove all unused data of tmpfile after abnormal power-off.
-	 */
-	add_orphan_inode(sbi, inode->i_ino);
-	f2fs_unlock_op(sbi);
-
-	alloc_nid_done(sbi, inode->i_ino);
-
-	d_tmpfile(dentry, inode);
-	unlock_new_inode(inode);
-	return 0;
-
-release_out:
-	release_orphan_inode(sbi);
-out:
-	handle_failed_inode(inode);
-	return err;
+	return f2fs_rename(old_dir, old_dentry, new_dir, new_dentry, flags);
 }
 
 #ifdef CONFIG_F2FS_FS_ENCRYPTION
--

2.3.3

Hi Jaegeuk,

Quoted 30 lines

-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, May 19, 2015 1:51 PM
To: Chao Yu
Cc: Changman Lee; [email protected]; [email protected]
Subject: Re: [PATCH 1/2] f2fs: support RENAME_WHITEOUT

Hi Chao,

Quoted 20 lines · Chao Yu

As the description of rename in manual, RENAME_WHITEOUT is a special operation that only makes sense for overlay/union type filesystem.

When performing rename with RENAME_WHITEOUT, dst will be replace with src, and meanwhile, a 'whiteout' will be create with name of src.

A "whiteout" is designed to be a char device with 0,0 device number, it has specially meaning for stackable filesystem. In these filesystems, there are multiple layers exist, and only top of these can be modified. So a whiteout in top layer is used to hide a corresponding file in lower layer, as well removal of whiteout will make the file appear.

Now in overlayfs, when we rename a file which is exist in lower layer, it will be copied up to upper if it is not on upper layer yet, and then rename it on upper layer, source file will be whiteouted to hide corresponding file in lower layer at the same time.

So in upper layer filesystem, implementation of RENAME_WHITEOUT provide a atomic operation for stackable filesystem to support rename operation.

Quoted 7 lines

Quoted 5 lines

There are multiple ways to implement RENAME_WHITEOUT in log of this commit:
7dcf5c3e4527 ("xfs: add RENAME_WHITEOUT support") which pointed out by Dave Chinner.

For now, we just try to follow the way that xfs/ext4 use.

Could you merge the two patches into one?

OK.

Quoted 1 line

And, after finishing xfstests, kernel reports missing inode objects.

Do you mean this?

BUG f2fs_inode_cache (Tainted: G B O ):
Objects remaining in f2fs_inode_cache on kmem_cache_close()

Whenever rmmod f2fs module after tests/generic/078, system report this.

Quoted 2 lines

Could you check it out?

I find that the whiteout file is not be iput() after being linkated to an entry. After fixing this issue, no exceptional report appears again.

So can you please help to test the v2 patch?