Re: [RFC PATCH] gfs2: Fix missing cleanup quota in gfs2_put_super()

From: Wangshaobo (bobo)
Date: Sat Dec 03 2022 - 02:31:12 EST



在 2022/12/1 19:50, Wang ShaoBo 写道:
syzbot has reported an interesting issue:

INFO: task syz-executor888:3126 blocked for more than 143 seconds.
Not tainted 6.1.0-rc6-syzkaller-32662-g6d464646530f #0
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
task:syz-executor888 state:D stack:0 pid:3126 ppid:3124 flags:0x00000000
Call trace:
__switch_to+0x180/0x298 arch/arm64/kernel/process.c:553
context_switch kernel/sched/core.c:5209 [inline]
__schedule+0x408/0x594 kernel/sched/core.c:6521
schedule+0x64/0xa4 kernel/sched/core.c:6597
schedule_timeout+0x108/0x1b4 kernel/time/timer.c:1935
gfs2_gl_hash_clear+0xd4/0x1b0 fs/gfs2/glock.c:2263
gfs2_put_super+0x318/0x390 fs/gfs2/super.c:620
generic_shutdown_super+0x94/0x198 fs/super.c:492
kill_block_super+0x30/0x78 fs/super.c:1428
gfs2_kill_sb+0x68/0x78
deactivate_locked_super+0x70/0xe8 fs/super.c:332
deactivate_super+0xd0/0xd4 fs/super.c:363
cleanup_mnt+0x184/0x1c0 fs/namespace.c:1186
__cleanup_mnt+0x20/0x30 fs/namespace.c:1193
task_work_run+0x100/0x148 kernel/task_work.c:179
resume_user_mode_work include/linux/resume_user_mode.h:49 [inline]
do_notify_resume+0x174/0x1f0 arch/arm64/kernel/signal.c:1127
prepare_exit_to_user_mode arch/arm64/kernel/entry-common.c:137 [inline]
exit_to_user_mode arch/arm64/kernel/entry-common.c:142 [inline]
el0_svc+0x9c/0x150 arch/arm64/kernel/entry-common.c:638
el0t_64_sync_handler+0x84/0xf0 arch/arm64/kernel/entry-common.c:655
el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:584

Showing all locks held in the system:
1 lock held by rcu_tasks_kthre/11:
#0: ffff80000d4a4768 (rcu_tasks.tasks_gp_mutex){+.+.}-{3:3}, at: rcu_tasks_one_gp+0x3c/0x450 kernel/rcu/tasks.h:507
1 lock held by rcu_tasks_trace/12:
#0: ffff80000d4a4db8 (rcu_tasks_trace.tasks_gp_mutex){+.+.}-{3:3}, at: rcu_tasks_one_gp+0x3c/0x450 kernel/rcu/tasks.h:507
1 lock held by khungtaskd/27:
#0: ffff80000d4a4640 (rcu_read_lock){....}-{1:2}, at: rcu_lock_acquire+0x4/0x48 include/linux/rcupdate.h:303
2 locks held by getty/2759:
#0: ffff0000c7d7e098 (&tty->ldisc_sem){++++}-{0:0}, at: tty_ldisc_ref_wait+0x28/0x58 drivers/tty/tty_ldisc.c:244
#1: ffff80000f6be2f0 (&ldata->atomic_read_lock){+.+.}-{3:3}, at: n_tty_read+0x19c/0x89c drivers/tty/n_tty.c:2177
1 lock held by syz-executor888/3126:
#0: ffff0000cae680e0 (&type->s_umount_key#41){+.+.}-{3:3}, at: deactivate_super+0xc8/0xd4 fs/super.c:362

=============================================

One task blocked after executing gfs2_gl_hash_clear(), it looks
like go to sleep for waiting the condition (&sdp->sd_glock_disposal == 0),
this field sd_glock_disposal is used to count the getting times
that calling gfs2_glock_get() success, In most cases, when we operate
the mount interface, we will finally call this function:

do_mount()
path_mount()
vfs_get_tree()
gfs2_get_tree()
get_tree_bdev()
gfs2_fill_super()
________ init_inodes()_____________
V V
gfs2_lookup_simple() gfs2_rindex_update()
gfs2_lookupi() gfs2_ri_update()
gfs2_dir_search() -> read_rindex_entry() //*gfs2_rgrpd->rd_gl
-> gfs2_inode_lookup() //*gfs2_inode->i_gl |__ gfs2_glock_get()
|__gfs2_glock_get()

Correspondingly, release this glock's reference and decrease
sdp->sd_glock_disposal through this path when umount:

cleanup_mnt()
deactivate_super()
deactivate_locked_super()
gfs2_kill_sb()
kill_block_super()
generic_shutdown_super()
_gfs2_rgrpd->rd_gl__gfs2_put_super()__gfs2_inode->i_gl__
V V
gfs2_clear_rgrpd() gfs2_put_super()
gfs2_glock_put() iput()
gfs2_glock_free() evict()
gfs2_evict_inode()
gfs2_glock_put_eventually()
gfs2_glock_put()
gfs2_glock_free()

But if calling reconfigure_super() from sysfs between mount and umount,
this path will also take the glock lock and increase sdp->sd_glock_disposal
as it also call gfs2_glock_get():

reconfigure_super()
gfs2_reconfigure()
gfs2_make_fs_rw() //remount filesystem read-write
gfs2_quota_init()
-> qd_alloc() //*gfs2_quota_data->qd_gl
|___ gfs2_glock_get()

This relies on gfs2_quota_cleanup() that is eventually called by
gfs2_make_fs_ro() to clean up but unfortunately the file system
has not been remounted at this time, the condition (!sb_rdonly(sb))
is false and gfs2_make_fs_ro() can not be called, ultimately makes
sdp->sd_glock_disposal never be 0.

This add calling gfs2_quota_cleanup() when this filesystem remains
readOnly, in this case we don't need sync quota or other operation
otherwise the filesystem was really remounted.

But this modification will trigger this warning:

WARNING: CPU: 7 PID: 2595 at include/linux/backing-dev.h:246 __folio_mark_dirty+0x559/0x7b0
I paste wrong, the warning is:
gfs2: fsid=mounts.0: warning: assertion "!qd->qd_change" failed at function = gfs2_quota_cleanup, file = fs/gfs2/quota.c, line = 1474

qd->qd_change has been changed in gfs2_quota_init(), but it looks
we don't need to do anything before remounting the file system as read-write.

We can filter this warning with adding a parameter as it doesn't
really make sense here.

Fixes: 9e6e0a128bca ("GFS2: Merge mount.c and ops_super.c into super.c")
Reported-by: syzbot+ed7d0f71a89e28557a77@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Wang ShaoBo <bobo.shaobowang@xxxxxxxxxx>
---
fs/gfs2/super.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index b018957a1bb2..9a3c6da8db55 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -586,7 +586,10 @@ static void gfs2_put_super(struct super_block *sb)
if (!sb_rdonly(sb)) {
gfs2_make_fs_ro(sdp);
+ } else {
+ gfs2_quota_cleanup(sdp);
}
+
WARN_ON(gfs2_withdrawing(sdp));
/* At this point, we're through modifying the disk */