From: Liu Bo <bo.li.liu@xxxxxxxxxx>
I found this while testing xfstests 068, the story is
t1 t2
sys_sync thaw_super
iterate_supers
down_read(sb->s_umount) down_write(sb->s_umount) --->wait for t1
sync_fs (with wait mode)
start_transaction
sb_start_intwrite --------------------> wait for t2 to set s_writers.frozen to SB_UNFROZEN
In this patch, I add an helper sb_start_intwrite_trylock() and use it before we
start_transaction in sync_fs() with wait mode so that we won't hit the deadlock.
Signed-off-by: Liu Bo <bo.li.liu@xxxxxxxxxx>
---
fs/btrfs/super.c | 15 +++++++++++++++
include/linux/fs.h | 5 +++++
2 files changed, 20 insertions(+), 0 deletions(-)
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index f2eb24c..1e04b41 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -847,6 +847,21 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
return 0;
}
+ /*
+ * sys_sync can cause an ABBA deadlock with freeze/thaw
+ * o freeze_super() grabs s_umount lock and set sb to SB_FREEZE_FS.
+ * o thaw_super() grabs s_umount lock and set sb to SB_UNFROZEN.
+ * o iterate_supers() grabs s_umount lock, and sync fs, during which
+ * we need to do sb_start_intwrite() in starting a
+ * new transaction.
+ * so iterate_supers() will wait for thaw_super() to reset sb's frozen
+ * state, while thaw_super() will wait for iterate_supers() to drop the
+ * s_umount lock. This is an ABBA deadlock.
+ */
+ if (!sb_start_intwrite_trylock(sb))
+ return 0;
+ sb_end_intwrite(sb);
+
btrfs_wait_ordered_extents(root, 0, 0);
trans = btrfs_start_transaction(root, 0);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index aa11047..8a3efd0 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1700,6 +1700,11 @@ static inline void sb_start_intwrite(struct super_block *sb)
__sb_start_write(sb, SB_FREEZE_FS, true);
}
+static inline int sb_start_intwrite_trylock(struct super_block *sb)
+{
+ return __sb_start_write(sb, SB_FREEZE_FS, false);
+}
+
extern bool inode_owner_or_capable(const struct inode *inode);
--
1.7.7.6
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
