From: Andrew Perepechko <andrew.perepechko@xxxxxxx> When changing a large xattr value to a different large xattr value, the old xattr inode is freed. Truncate during the final iput causes current transaction restart. Eventually, parent inode bh is marked dirty and kernel panic happens when jbd2 figures out that this bh belongs to the committed transaction. Here is a reproducer #!/bin/bash dd if=/dev/zero of=/tmp/ldiskfs bs=1M count=100 mkfs.ext4 -O ea_inode /tmp/ldiskfs -J size=16 -I 512 mkdir -p /tmp/ldiskfs_m mount -t ext4 /tmp/ldiskfs /tmp/ldiskfs_m -o loop,commit=600,no_mbcache touch /tmp/ldiskfs_m/file{1..1024} V=$(for i in `seq 60000`; do echo -n x ; done) V1="1$V" V2="2$V" while true; do setfattr -n user.xattr -v $V /tmp/ldiskfs_m/file{1..1024} setfattr -n user.xattr -v $V1 /tmp/ldiskfs_m/file{1..1024} & setfattr -n user.xattr -v $V2 /tmp/ldiskfs_m/file{1024..1} & wait done A possible fix is to call this final iput in a separate thread. This way, setxattr transactions will never be split into two. Since the setxattr code adds xattr inodes with nlink=0 into the orphan list, old xattr inodes will be properly cleaned up in any case. Signed-off-by: Andrew Perepechko <andrew.perepechko@xxxxxxx> Signed-off-by: Artem Blagodarenko <artem.blagodarenko@xxxxxxxxx> HPE-bug-id: LUS-10534 --- fs/ext4/super.c | 1 + fs/ext4/xattr.c | 34 ++++++++++++++++++++++++++++++++-- 2 files changed, 33 insertions(+), 2 deletions(-) diff --git a/fs/ext4/super.c b/fs/ext4/super.c index c5021ca0a28a..8c04c19fa4b8 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -1199,6 +1199,7 @@ static void ext4_put_super(struct super_block *sb) int aborted = 0; int i, err; + flush_scheduled_work(); ext4_unregister_li_request(sb); ext4_quota_off_umount(sb); diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c index 042325349098..0cadbf4a9f2b 100644 --- a/fs/ext4/xattr.c +++ b/fs/ext4/xattr.c @@ -1544,6 +1544,31 @@ static int ext4_xattr_inode_lookup_create(handle_t *handle, struct inode *inode, return 0; } +struct delayed_iput_work { + struct work_struct work; + struct inode *inode; +}; + +static void delayed_iput_fn(struct work_struct *work) +{ + struct delayed_iput_work *diwork; + + diwork = container_of(work, struct delayed_iput_work, work); + iput(diwork->inode); + kfree(diwork); +} + +static void delayed_iput(struct inode *inode, struct delayed_iput_work *work) +{ + if (!work) { + iput(inode); + } else { + INIT_WORK(&work->work, delayed_iput_fn); + work->inode = inode; + schedule_work(&work->work); + } +} + /* * Reserve min(block_size/8, 1024) bytes for xattr entries/names if ea_inode * feature is enabled. @@ -1561,6 +1586,7 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i, int in_inode = i->in_inode; struct inode *old_ea_inode = NULL; struct inode *new_ea_inode = NULL; + struct delayed_iput_work *diwork = NULL; size_t old_size, new_size; int ret; @@ -1637,7 +1663,11 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i, * Finish that work before doing any modifications to the xattr data. */ if (!s->not_found && here->e_value_inum) { - ret = ext4_xattr_inode_iget(inode, + diwork = kmalloc(sizeof(*diwork), GFP_NOFS); + if (!diwork) + ret = -ENOMEM; + else + ret = ext4_xattr_inode_iget(inode, le32_to_cpu(here->e_value_inum), le32_to_cpu(here->e_hash), &old_ea_inode); @@ -1791,7 +1821,7 @@ static int ext4_xattr_set_entry(struct ext4_xattr_info *i, ret = 0; out: iput(old_ea_inode); - iput(new_ea_inode); + delayed_iput(old_ea_inode, diwork); return ret; } -- 2.31.1