Skip to contentSkip to conversation

ext4: fix reserved space counter leakage

PATCH Applied Jeffle Xu 4 participants 6 messages Aug 23, 2021 – Sep 16, 2021 ext4 file system

lore
Share & tools
lore

Jeffle Xu

author
#0 lore raw

PATCH v2 1/1 ext4: fix reserved space counter leakage

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

Another effect of this bug is that /sys/fs/ext4/<dev>/delayed_allocation_blocks can remain non-zero even after syncfs has been executed on the filesystem.

Besides, add check for s_dirtyclusters_counter when inode is going to be evicted and freed. s_dirtyclusters_counter can still keep non-zero until inode is written back in .evict_inode(), and thus the check is delayed to .destroy_inode().

Fixes 51865fda28e5 ("ext4: let ext4 maintain extent status tree")

---
changes since v1:
- improve commit log suggested by Eric Whitney
- update "Suggested-by" title for Gao Xian, who actually found this bug code
- add check for s_dirtyclusters_counter in .destroy_inode()

---
 fs/ext4/inode.c | 5 +++++
 fs/ext4/super.c | 6 ++++++
 2 files changed, 11 insertions(+)
fs/ext4/inode.c +5 -0
index d8de607849df..73daf9443e5e 100644
@@ -1640,6 +1640,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	int ret;
 	bool allocated = false;
+	bool reserved = false;
 
 	/*
 	 * If the cluster containing lblk is shared with a delayed,
@@ -1656,6 +1657,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 		ret = ext4_da_reserve_space(inode);
 		if (ret != 0)   /* ENOSPC */
 			goto errout;
+		reserved = true;
 	} else {   /* bigalloc */
 		if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
 			if (!ext4_es_scan_clu(inode,
@@ -1668,6 +1670,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 					ret = ext4_da_reserve_space(inode);
 					if (ret != 0)   /* ENOSPC */
 						goto errout;
+					reserved = true;
 				} else {
 					allocated = true;
 				}
@@ -1678,6 +1681,8 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	}
 
 	ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
+	if (ret && reserved)
+		ext4_da_release_space(inode, 1);
 
 errout:
 	return ret;
fs/ext4/super.c +6 -0
index dfa09a277b56..61bf52b58fca 100644
@@ -1351,6 +1351,12 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
+
+	if (EXT4_I(inode)->i_reserved_data_blocks)
+		ext4_msg(inode->i_sb, KERN_ERR,
+			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
+			 inode->i_ino, EXT4_I(inode),
+			 EXT4_I(inode)->i_reserved_data_blocks);
 }
 
 static void init_once(void *foo)
Signature · 1 line
2.27.0
Reviewed-by on v2 1/1
#1 lore raw

* Jeffle Xu <[email protected]>:

Quoted 89 lines

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

Another effect of this bug is that /sys/fs/ext4/<dev>/delayed_allocation_blocks can remain non-zero even after syncfs has been executed on the filesystem.

Besides, add check for s_dirtyclusters_counter when inode is going to be evicted and freed. s_dirtyclusters_counter can still keep non-zero until inode is written back in .evict_inode(), and thus the check is delayed to .destroy_inode().

Fixes 51865fda28e5 ("ext4: let ext4 maintain extent status tree")

---
changes since v1:
- improve commit log suggested by Eric Whitney
- update "Suggested-by" title for Gao Xian, who actually found this bug code
- add check for s_dirtyclusters_counter in .destroy_inode()

---
 fs/ext4/inode.c | 5 +++++
 fs/ext4/super.c | 6 ++++++
 2 files changed, 11 insertions(+)
fs/ext4/inode.c +5 -0
index d8de607849df..73daf9443e5e 100644
@@ -1640,6 +1640,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	int ret;
 	bool allocated = false;
+	bool reserved = false;
 
 	/*
 	 * If the cluster containing lblk is shared with a delayed,
@@ -1656,6 +1657,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 		ret = ext4_da_reserve_space(inode);
 		if (ret != 0)   /* ENOSPC */
 			goto errout;
+		reserved = true;
 	} else {   /* bigalloc */
 		if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
 			if (!ext4_es_scan_clu(inode,
@@ -1668,6 +1670,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 					ret = ext4_da_reserve_space(inode);
 					if (ret != 0)   /* ENOSPC */
 						goto errout;
+					reserved = true;
 				} else {
 					allocated = true;
 				}
@@ -1678,6 +1681,8 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	}
 
 	ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
+	if (ret && reserved)
+		ext4_da_release_space(inode, 1);
 
 errout:
 	return ret;
fs/ext4/super.c +6 -0
index dfa09a277b56..61bf52b58fca 100644
@@ -1351,6 +1351,12 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
+
+	if (EXT4_I(inode)->i_reserved_data_blocks)
+		ext4_msg(inode->i_sb, KERN_ERR,
+			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
+			 inode->i_ino, EXT4_I(inode),
+			 EXT4_I(inode)->i_reserved_data_blocks);
 }
 
 static void init_once(void *foo)
Signature · 1 line
2.27.0

Looks good, passed 4k xfstests-bld regression. Feel free to add:

Eric

author
#2 lore raw
Quoted 86 lines · Eric Whitney

* Jeffle Xu <[email protected]>:

Quoted 85 lines

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

Another effect of this bug is that /sys/fs/ext4/<dev>/delayed_allocation_blocks can remain non-zero even after syncfs has been executed on the filesystem.

Besides, add check for s_dirtyclusters_counter when inode is going to be evicted and freed. s_dirtyclusters_counter can still keep non-zero until inode is written back in .evict_inode(), and thus the check is delayed to .destroy_inode().

Fixes 51865fda28e5 ("ext4: let ext4 maintain extent status tree")

---
changes since v1:
- improve commit log suggested by Eric Whitney
- update "Suggested-by" title for Gao Xian, who actually found this bug code
- add check for s_dirtyclusters_counter in .destroy_inode()

---
 fs/ext4/inode.c | 5 +++++
 fs/ext4/super.c | 6 ++++++
 2 files changed, 11 insertions(+)
fs/ext4/inode.c +5 -0
index d8de607849df..73daf9443e5e 100644
@@ -1640,6 +1640,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	int ret;
 	bool allocated = false;
+	bool reserved = false;
 
 	/*
 	 * If the cluster containing lblk is shared with a delayed,
@@ -1656,6 +1657,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 		ret = ext4_da_reserve_space(inode);
 		if (ret != 0)   /* ENOSPC */
 			goto errout;
+		reserved = true;
 	} else {   /* bigalloc */
 		if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
 			if (!ext4_es_scan_clu(inode,
@@ -1668,6 +1670,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 					ret = ext4_da_reserve_space(inode);
 					if (ret != 0)   /* ENOSPC */
 						goto errout;
+					reserved = true;
 				} else {
 					allocated = true;
 				}
@@ -1678,6 +1681,8 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	}
 
 	ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
+	if (ret && reserved)
+		ext4_da_release_space(inode, 1);
 
 errout:
 	return ret;
fs/ext4/super.c +6 -0
index dfa09a277b56..61bf52b58fca 100644
@@ -1351,6 +1351,12 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
+
+	if (EXT4_I(inode)->i_reserved_data_blocks)
+		ext4_msg(inode->i_sb, KERN_ERR,
+			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
+			 inode->i_ino, EXT4_I(inode),
+			 EXT4_I(inode)->i_reserved_data_blocks);
 }
 
Quoted 8 lines
Quoted 4 lines
fs/ext4/super.c +0 -0
 static void init_once(void *foo)
Signature · 1 line
2.27.0

Looks good, passed 4k xfstests-bld regression. Feel free to add:

Hi tytso, it's a bug fix and it would be great if it could be merged to 5.15.

Signature · 2 lines
Thanks,
Jeffle
author
#3 lore raw
Quoted 96 lines · JeffleXu
Quoted 93 lines · Eric Whitney

* Jeffle Xu <[email protected]>:

Quoted 89 lines

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

Another effect of this bug is that /sys/fs/ext4/<dev>/delayed_allocation_blocks can remain non-zero even after syncfs has been executed on the filesystem.

Besides, add check for s_dirtyclusters_counter when inode is going to be evicted and freed. s_dirtyclusters_counter can still keep non-zero until inode is written back in .evict_inode(), and thus the check is delayed to .destroy_inode().

Fixes 51865fda28e5 ("ext4: let ext4 maintain extent status tree")

---
changes since v1:
- improve commit log suggested by Eric Whitney
- update "Suggested-by" title for Gao Xian, who actually found this bug code
- add check for s_dirtyclusters_counter in .destroy_inode()

---
 fs/ext4/inode.c | 5 +++++
 fs/ext4/super.c | 6 ++++++
 2 files changed, 11 insertions(+)
fs/ext4/inode.c +5 -0
index d8de607849df..73daf9443e5e 100644
@@ -1640,6 +1640,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	int ret;
 	bool allocated = false;
+	bool reserved = false;
 
 	/*
 	 * If the cluster containing lblk is shared with a delayed,
@@ -1656,6 +1657,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 		ret = ext4_da_reserve_space(inode);
 		if (ret != 0)   /* ENOSPC */
 			goto errout;
+		reserved = true;
 	} else {   /* bigalloc */
 		if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
 			if (!ext4_es_scan_clu(inode,
@@ -1668,6 +1670,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 					ret = ext4_da_reserve_space(inode);
 					if (ret != 0)   /* ENOSPC */
 						goto errout;
+					reserved = true;
 				} else {
 					allocated = true;
 				}
@@ -1678,6 +1681,8 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	}
 
 	ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
+	if (ret && reserved)
+		ext4_da_release_space(inode, 1);
 
 errout:
 	return ret;
fs/ext4/super.c +6 -0
index dfa09a277b56..61bf52b58fca 100644
@@ -1351,6 +1351,12 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
+
+	if (EXT4_I(inode)->i_reserved_data_blocks)
+		ext4_msg(inode->i_sb, KERN_ERR,
+			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
+			 inode->i_ino, EXT4_I(inode),
+			 EXT4_I(inode)->i_reserved_data_blocks);
 }
 
 static void init_once(void *foo)
Signature · 1 line
2.27.0

Looks good, passed 4k xfstests-bld regression. Feel free to add:

Quoted 6 lines
Quoted 1 line

Hi tytso, it's a bug fix and it would be great if it could be merged to 5.15.

ping ...

Signature · 2 lines
Thanks,
Jeffle
#4 lore raw

Hi Ted,

Quoted 99 lines · JeffleXu
Quoted 96 lines · JeffleXu
Quoted 93 lines · Eric Whitney

* Jeffle Xu <[email protected]>:

Quoted 89 lines

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

Another effect of this bug is that /sys/fs/ext4/<dev>/delayed_allocation_blocks can remain non-zero even after syncfs has been executed on the filesystem.

Besides, add check for s_dirtyclusters_counter when inode is going to be evicted and freed. s_dirtyclusters_counter can still keep non-zero until inode is written back in .evict_inode(), and thus the check is delayed to .destroy_inode().

Fixes 51865fda28e5 ("ext4: let ext4 maintain extent status tree")

---
changes since v1:
- improve commit log suggested by Eric Whitney
- update "Suggested-by" title for Gao Xian, who actually found this bug code
- add check for s_dirtyclusters_counter in .destroy_inode()

---
 fs/ext4/inode.c | 5 +++++
 fs/ext4/super.c | 6 ++++++
 2 files changed, 11 insertions(+)
fs/ext4/inode.c +5 -0
index d8de607849df..73daf9443e5e 100644
@@ -1640,6 +1640,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	int ret;
 	bool allocated = false;
+	bool reserved = false;
 
 	/*
 	 * If the cluster containing lblk is shared with a delayed,
@@ -1656,6 +1657,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 		ret = ext4_da_reserve_space(inode);
 		if (ret != 0)   /* ENOSPC */
 			goto errout;
+		reserved = true;
 	} else {   /* bigalloc */
 		if (!ext4_es_scan_clu(inode, &ext4_es_is_delonly, lblk)) {
 			if (!ext4_es_scan_clu(inode,
@@ -1668,6 +1670,7 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 					ret = ext4_da_reserve_space(inode);
 					if (ret != 0)   /* ENOSPC */
 						goto errout;
+					reserved = true;
 				} else {
 					allocated = true;
 				}
@@ -1678,6 +1681,8 @@ static int ext4_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk)
 	}
 
 	ret = ext4_es_insert_delayed_block(inode, lblk, allocated);
+	if (ret && reserved)
+		ext4_da_release_space(inode, 1);
 
 errout:
 	return ret;
fs/ext4/super.c +6 -0
index dfa09a277b56..61bf52b58fca 100644
@@ -1351,6 +1351,12 @@ static void ext4_destroy_inode(struct inode *inode)
 				true);
 		dump_stack();
 	}
+
+	if (EXT4_I(inode)->i_reserved_data_blocks)
+		ext4_msg(inode->i_sb, KERN_ERR,
+			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
+			 inode->i_ino, EXT4_I(inode),
+			 EXT4_I(inode)->i_reserved_data_blocks);
 }
 
 static void init_once(void *foo)
Signature · 1 line
2.27.0

Looks good, passed 4k xfstests-bld regression. Feel free to add:

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

Hi tytso, it's a bug fix and it would be great if it could be merged to 5.15.

ping ...

It seems somewhat a real issue and has some impact in our production. Would you mind giving some hints if it could be resolved in this way?

Or if some other concerns / solution about this?

Thanks,
Gao Xiang

Quoted 4 lines
Signature · 3 lines
-- 
Thanks,
Jeffle
maintainer Applied
#5 lore raw
In reply to Jeffle Xu
Quoted 9 lines · Jeffle Xu

When ext4_insert_delayed block receives and recovers from an error from ext4_es_insert_delayed_block(), e.g., ENOMEM, it does not release the space it has reserved for that block insertion as it should. One effect of this bug is that s_dirtyclusters_counter is not decremented and remains incorrectly elevated until the file system has been unmounted. This can result in premature ENOSPC returns and apparent loss of free space.

[...]

Applied, thanks!

[1/1] ext4: fix reserved space counter leakage

      commit: 8ec5146a46609a89658e5dc9bc40b70ce70ef43b

Best regards,

Signature · 1 line
Theodore Ts'o <[email protected]>