| Message ID | 6c508b165cc727f07137f0208ce68ce3ce70c2c9.1530262400.git.dsterba@suse.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Jun 29, 2018 at 10:57:07AM +0200, David Sterba wrote: > This is called either at the end of the mount or if the mount fails. > In both cases, there's nothing running that can use the table so the > lock is pointless. And then lockdep says no. The umount path frees the table but there's some unfinished bio that wants to use the table from the interrupt context. And this is puzzling, there should be no IO in flight, all workers should be stopped. btrfs/011: [ 1339.169842] ================================ [ 1339.171891] WARNING: inconsistent lock state [ 1339.173724] 4.17.0-rc7-default+ #168 Not tainted [ 1339.175661] -------------------------------- [ 1339.177479] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage. [ 1339.179758] umount/4029 [HC0[0]:SC0[0]:HE1:SE1] takes: [ 1339.182008] 0000000096eee2cd (&(&cur->lock)->rlock){?.-.}, at: __remove_rbio_from_cache+0x5f/0x140 [btrfs] [ 1339.183982] {IN-HARDIRQ-W} state was registered at: [ 1339.184819] _raw_spin_lock_irqsave+0x43/0x52 [ 1339.185433] unlock_stripe+0x78/0x3d0 [btrfs] [ 1339.186041] rbio_orig_end_io+0x41/0xd0 [btrfs] [ 1339.186648] blk_update_request+0xd7/0x330 [ 1339.187205] blk_mq_end_request+0x18/0x70 [ 1339.187752] flush_smp_call_function_queue+0x83/0x120 [ 1339.188405] smp_call_function_single_interrupt+0x43/0x270 [ 1339.189094] call_function_single_interrupt+0xf/0x20 [ 1339.189733] native_safe_halt+0x2/0x10 [ 1339.190255] default_idle+0x1f/0x190 [ 1339.190759] do_idle+0x217/0x240 [ 1339.191229] cpu_startup_entry+0x6f/0x80 [ 1339.191768] start_secondary+0x192/0x1e0 [ 1339.192385] secondary_startup_64+0xa5/0xb0 [ 1339.193145] irq event stamp: 783817 [ 1339.193701] hardirqs last enabled at (783817): [<ffffffff8267234d>] _raw_spin_unlock_irqrestore+0x4d/0x60 [ 1339.194896] hardirqs last disabled at (783816): [<ffffffff82672a10>] _raw_spin_lock_irqsave+0x20/0x52 [ 1339.196010] softirqs last enabled at (777902): [<ffffffff82a00397>] __do_softirq+0x397/0x502 [ 1339.197435] softirqs last disabled at (777879): [<ffffffff82066311>] irq_exit+0xc1/0xd0 [ 1339.198797] [ 1339.198797] other info that might help us debug this: [ 1339.200011] Possible unsafe locking scenario: [ 1339.200011] [ 1339.201066] CPU0 [ 1339.201464] ---- [ 1339.201845] lock(&(&cur->lock)->rlock); [ 1339.202372] <Interrupt> [ 1339.202768] lock(&(&cur->lock)->rlock); [ 1339.203313] [ 1339.203313] *** DEADLOCK *** [ 1339.203313] [ 1339.204215] 1 lock held by umount/4029: [ 1339.204727] #0: 000000007c3dd992 (&type->s_umount_key#26){++++}, at: deactivate_super+0x43/0x50 [ 1339.205822] [ 1339.205822] stack backtrace: [ 1339.206660] CPU: 2 PID: 4029 Comm: umount Not tainted 4.17.0-rc7-default+ #168 [ 1339.207875] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014 [ 1339.209357] Call Trace: [ 1339.209742] dump_stack+0x85/0xc0 [ 1339.210204] print_usage_bug.cold.57+0x1aa/0x1e4 [ 1339.210790] ? print_shortest_lock_dependencies+0x40/0x40 [ 1339.211680] mark_lock+0x530/0x630 [ 1339.212291] ? print_shortest_lock_dependencies+0x40/0x40 [ 1339.213180] __lock_acquire+0x549/0xf60 [ 1339.213855] ? flush_work+0x24a/0x280 [ 1339.214485] ? __lock_is_held+0x4f/0x90 [ 1339.215176] lock_acquire+0x9e/0x1d0 [ 1339.215850] ? __remove_rbio_from_cache+0x5f/0x140 [btrfs] [ 1339.216755] _raw_spin_lock+0x2c/0x40 [ 1339.217303] ? __remove_rbio_from_cache+0x5f/0x140 [btrfs] [ 1339.218005] __remove_rbio_from_cache+0x5f/0x140 [btrfs] [ 1339.218687] btrfs_free_stripe_hash_table+0x2a/0x50 [btrfs] [ 1339.219393] close_ctree+0x1d7/0x330 [btrfs] [ 1339.219961] generic_shutdown_super+0x64/0x100 [ 1339.220659] kill_anon_super+0xe/0x20 [ 1339.221245] btrfs_kill_super+0x12/0xa0 [btrfs] [ 1339.221840] deactivate_locked_super+0x29/0x60 [ 1339.222416] cleanup_mnt+0x3b/0x70 [ 1339.222892] task_work_run+0x9b/0xd0 [ 1339.223388] exit_to_usermode_loop+0x99/0xa0 [ 1339.223949] do_syscall_64+0x16c/0x170 [ 1339.224603] entry_SYSCALL_64_after_hwframe+0x49/0xbe [ 1339.225457] RIP: 0033:0x7ffb9180d017 [ 1339.226063] RSP: 002b:00007fff9cb72a28 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6 [ 1339.227289] RAX: 0000000000000000 RBX: 000055b804721970 RCX: 00007ffb9180d017 [ 1339.228373] RDX: 0000000000000001 RSI: 0000000000000000 RDI: 000055b804721b50 [ 1339.229451] RBP: 0000000000000000 R08: 000055b804721b70 R09: 00007fff9cb71290 [ 1339.230524] R10: 0000000000000000 R11: 0000000000000246 R12: 00007ffb91d291c4 [ 1339.231604] R13: 000055b804721b50 R14: 0000000000000000 R15: 00007fff9cb72c98 -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c index 272acd9b1192..0840b054e4b7 100644 --- a/fs/btrfs/raid56.c +++ b/fs/btrfs/raid56.c @@ -407,19 +407,16 @@ static void remove_rbio_from_cache(struct btrfs_raid_bio *rbio) static void btrfs_clear_rbio_cache(struct btrfs_fs_info *info) { struct btrfs_stripe_hash_table *table; - unsigned long flags; struct btrfs_raid_bio *rbio; table = info->stripe_hash_table; - spin_lock_irqsave(&table->cache_lock, flags); while (!list_empty(&table->stripe_cache)) { rbio = list_entry(table->stripe_cache.next, struct btrfs_raid_bio, stripe_cache); __remove_rbio_from_cache(rbio); } - spin_unlock_irqrestore(&table->cache_lock, flags); } /*
This is called either at the end of the mount or if the mount fails. In both cases, there's nothing running that can use the table so the lock is pointless. Signed-off-by: David Sterba <dsterba@suse.com> --- fs/btrfs/raid56.c | 3 --- 1 file changed, 3 deletions(-)