ext4: Fix deadlock with fs freezing and EA inodes
Subsystems (2)
Jan Kara
reviewerPATCH 1/1 ext4: Fix deadlock with fs freezing and EA inodes
Xattr code using inodes with large xattr data can end up dropping last inode reference (and thus deleting the inode) from places like ext4_xattr_set_entry(). That function is called with transaction started and so ext4_evict_inode() can deadlock against fs freezing like:
CPU1 CPU2
removexattr() freeze_super()
vfs_removexattr()
ext4_xattr_set()
handle = ext4_journal_start()
...
ext4_xattr_set_entry()
iput(old_ea_inode)
ext4_evict_inode(old_ea_inode)
sb->s_writers.frozen = SB_FREEZE_FS;
sb_wait_write(sb, SB_FREEZE_FS);
ext4_freeze()
jbd2_journal_lock_updates()
-> blocks waiting for all
handles to stop
sb_start_intwrite()
-> blocks as sb is already in SB_FREEZE_FS stateGenerally it is advisable to delete inodes from a separate transaction as it can consume quite some credits however in this case it would be quite clumsy and furthermore the credits for inode deletion are quite limited and already accounted for. So just tweak ext4_evict_inode() to avoid freeze protection if we have transaction already started and thus it is not really needed anyway.
--- fs/ext4/inode.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-)
fs/ext4/inode.c +14 -5
@@ -175,6 +175,7 @@ void ext4_evict_inode(struct inode *inode)*/ int extra_credits =6 ;struct ext4_xattr_inode_array *ea_inode_array =NULL ; +bool freeze_protected = false;trace_ext4_evict_inode (inode);
@@ -232,9 +233,14 @@ void ext4_evict_inode(struct inode *inode)/* * Protect us against freezing - iput() caller didn't have to have any -* protection against it +* protection against it. When we are in a running transaction though, +* we are already protected against freezing and we cannot grab further +* protection due to lock ordering constraints. */ -sb_start_intwrite (inode->i_sb); +if (!ext4_journal_current_handle ()) { +sb_start_intwrite (inode->i_sb); + freeze_protected = true; + }if (!IS_NOQUOTA (inode)) extra_credits +=EXT4_MAXQUOTAS_DEL_BLOCKS (inode->i_sb);
@@ -253,7 +259,8 @@ void ext4_evict_inode(struct inode *inode)* cleaned up. */ ext4_orphan_del (NULL , inode); -sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb); goto no_delete; }
@@ -294,7 +301,8 @@ void ext4_evict_inode(struct inode *inode) stop_handle:ext4_journal_stop (handle);ext4_orphan_del (NULL , inode); -sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb);ext4_xattr_inode_array_free (ea_inode_array); goto no_delete; }
@@ -323,7 +331,8 @@ void ext4_evict_inode(struct inode *inode)else ext4_free_inode (handle, inode);ext4_journal_stop (handle); -sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb);ext4_xattr_inode_array_free (ea_inode_array);return ; no_delete:
Signature · 1 line
2.16.4
Quoted 32 lines · Jan Kara
Xattr code using inodes with large xattr data can end up dropping last inode reference (and thus deleting the inode) from places like
ext4_xattr_set_entry(). That function is called with transaction started and soext4_evict_inode()can deadlock against fs freezing like:CPU1 CPU2
removexattr()freeze_super()
vfs_removexattr()
ext4_xattr_set()handle = ext4_journal_start() ... ext4_xattr_set_entry() iput(old_ea_inode) ext4_evict_inode(old_ea_inode) sb->s_writers.frozen = SB_FREEZE_FS; sb_wait_write(sb, SB_FREEZE_FS); ext4_freeze() jbd2_journal_lock_updates() -> blocks waiting for all handles to stop sb_start_intwrite() -> blocks as sb is already in SB_FREEZE_FS stateGenerally it is advisable to delete inodes from a separate transaction as it can consume quite some credits however in this case it would be quite clumsy and furthermore the credits for inode deletion are quite limited and already accounted for. So just tweak
ext4_evict_inode()to avoid freeze protection if we have transaction already started and thus it is not really needed anyway.
Quoted 4 lines
Fixes dec214d00e0d ("ext4: xattr inode deduplication")CC Tahsin Erdogan <[email protected]>Signed-off-by Jan Kara <[email protected]>
Quoted 62 lines
---
fs/ext4/inode.c| 19 ++++++++++++++-----1 file changed, 14 insertions(+), 5 deletions(-)fs/ext4/inode.c +10
-3@@ -175,6 +175,7 @@ void ext4_evict_inode(struct inode *inode)*/ int extra_credits =6 ;struct ext4_xattr_inode_array *ea_inode_array =NULL ; +bool freeze_protected = false;trace_ext4_evict_inode (inode);@@ -232,9 +233,14 @@ void ext4_evict_inode(struct inode *inode)/* * Protect us against freezing - iput() caller didn't have to have any -* protection against it +* protection against it. When we are in a running transaction though, +* we are already protected against freezing and we cannot grab further +* protection due to lock ordering constraints. */ -sb_start_intwrite (inode->i_sb); +if (!ext4_journal_current_handle ()) { +sb_start_intwrite (inode->i_sb); + freeze_protected = true; + }if (!IS_NOQUOTA (inode)) extra_credits +=EXT4_MAXQUOTAS_DEL_BLOCKS (inode->i_sb);@@ -253,7 +259,8 @@ void ext4_evict_inode(struct inode *inode)* cleaned up. */ ext4_orphan_del (NULL , inode); -sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb); goto no_delete; }@@ -294,7 +301,8 @@ void ext4_evict_inode(struct inode *inode)
stop_handle:ext4_journal_stop(handle); ext4_orphan_del(NULL, inode);fs/ext4/inode.c +2
-1-sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb);ext4_xattr_inode_array_free(ea_inode_array); goto no_delete; }fs/ext4/inode.c +2
-1@@ -323,7 +331,8 @@ void ext4_evict_inode(struct inode *inode)else ext4_free_inode (handle, inode);ext4_journal_stop (handle); -sb_end_intwrite (inode->i_sb); +if (freeze_protected) +sb_end_intwrite (inode->i_sb);ext4_xattr_inode_array_free (ea_inode_array);return ;
Quoted 4 lines
no_delete:
--
2.16.4
Cheers, Andreas
Quoted 31 lines · Jan Kara
Xattr code using inodes with large xattr data can end up dropping last inode reference (and thus deleting the inode) from places like
ext4_xattr_set_entry(). That function is called with transaction started and soext4_evict_inode()can deadlock against fs freezing like:CPU1 CPU2
removexattr()freeze_super()
vfs_removexattr()ext4_xattr_set() handle = ext4_journal_start() ... ext4_xattr_set_entry() iput(old_ea_inode) ext4_evict_inode(old_ea_inode) sb->s_writers.frozen = SB_FREEZE_FS; sb_wait_write(sb, SB_FREEZE_FS); ext4_freeze() jbd2_journal_lock_updates() -> blocks waiting for all handles to stop sb_start_intwrite() -> blocks as sb is already in SB_FREEZE_FS stateGenerally it is advisable to delete inodes from a separate transaction as it can consume quite some credits however in this case it would be quite clumsy and furthermore the credits for inode deletion are quite limited and already accounted for. So just tweak
ext4_evict_inode()to avoid freeze protection if we have transaction already started and thus it is not really needed anyway.
Quoted 4 lines
Fixes dec214d00e0d ("ext4: xattr inode deduplication")CC Tahsin Erdogan <[email protected]>Signed-off-by Jan Kara <[email protected]>
thanks, applied.
- Ted