Hi,
while playing with lockedp, I started to hit consistent crashes with
xfstests/013 (fsstress):
[ 537.284234] ------------[ cut here ]------------
[ 537.288040] kernel BUG at fs/btrfs/tree-log.c:1792!
[ 537.288040] invalid opcode: 0000 [#1] SMP
[ 537.288040] last sysfs file: /sys/devices/virtual/bdi/btrfs-10/uevent
[ 537.288040] CPU 1
[ 537.288040] Modules linked in: btrfs
[ 537.288040]
[ 537.288040] Pid: 12069, comm: fsstress Not tainted 2.6.39-rc7-default+ #24 Intel
[ 537.288040] RIP: 0010:[<ffffffffa0063c79>] [<ffffffffa0063c79>] walk_up_log_tree+0x2f9/0x330 [btrfs]
[ 537.288040] RSP: 0018:ffff880058191c08 EFLAGS: 00010282
[ 537.288040] RAX: 00000000ffffffa1 RBX: ffff88004f863910 RCX: 0000000000000001
[ 537.288040] RDX: 0000000000000f0e RSI: 0000000000000001 RDI: ffff880078bd8600
[ 537.288040] RBP: ffff880058191c88 R08: 0000000000000000 R09: 0000000000000001
[ 537.288040] R10: 0000000000000000 R11: 0000000000006945 R12: ffff880058191cb4
[ 537.288040] R13: 0000000000000000 R14: ffff880000000000 R15: ffff880058191cf8
[ 537.288040] FS: 00007f469ba3d700(0000) GS:ffff88007e600000(0000) knlGS:0000000000000000
[ 537.288040] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 537.288040] CR2: 00007f469af27430 CR3: 000000007ba3c000 CR4: 00000000000006e0
[ 537.288040] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 537.288040] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 537.288040] Process fsstress (pid: 12069, threadinfo ffff880058190000, task ffff8800782f1140)
[ 537.288040] Stack:
[ 537.288040] ffff880058191fd8 ffff88005e326f20 ffff880058191fd8 fffffffffffffffa
[ 537.288040] ffff880058191fd8 ffff880058191fd8 ffff880058191fd8 ffff880058191fd8
[ 537.288040] ffff88004eca4598 ffff8800580a6000 ffff880058191da8 0000000000000000
[ 537.288040] Call Trace:
[ 537.288040] [<ffffffffa0064348>] walk_log_tree+0x108/0x2a0 [btrfs]
[ 537.288040] [<ffffffffa0064523>] free_log_tree+0x43/0xc0 [btrfs]
[ 537.288040] [<ffffffffa0061950>] ? insert_one_name+0xc0/0xc0 [btrfs]
[ 537.288040] [<ffffffffa006663f>] btrfs_free_log+0x1f/0x30 [btrfs]
[ 537.288040] [<ffffffffa002bb61>] commit_fs_roots+0x91/0x150 [btrfs]
[ 537.288040] [<ffffffffa002cc70>] ? btrfs_commit_transaction+0x340/0x7d0 [btrfs]
[ 537.288040] [<ffffffffa002cc70>] ? btrfs_commit_transaction+0x340/0x7d0 [btrfs]
[ 537.288040] [<ffffffffa002cc7b>] btrfs_commit_transaction+0x34b/0x7d0 [btrfs]
[ 537.288040] [<ffffffff810b8c80>] ? wake_up_bit+0x40/0x40
[ 537.288040] [<ffffffff811a8cb0>] ? __sync_filesystem+0x90/0x90
[ 537.288040] [<ffffffffa0003210>] btrfs_sync_fs+0x60/0xe0 [btrfs]
[ 537.288040] [<ffffffff811a8c7e>] __sync_filesystem+0x5e/0x90
[ 537.288040] [<ffffffff811a8ccf>] sync_one_sb+0x1f/0x30
[ 537.288040] [<ffffffff8117f287>] iterate_supers+0x77/0xe0
[ 537.288040] [<ffffffff811a8bd0>] sync_filesystems+0x20/0x30
[ 537.616540] [<ffffffff811a8d71>] sys_sync+0x21/0x40
[ 537.616540] [<ffffffff81e94f42>] system_call_fastpath+0x16/0x1b
[ 537.616540] Code: 00 00 48 83 c4 58 b8 01 00 00 00 5b 41 5c 41 5d 41 5e 41 5f c9 c3 0f 1f 40 00 48 83 c4 58 b0 01 5b 41 5c 41 5d 41 5e 41 5f c9 c3 <0f> 0b be fc 06 00 00 48 c7 c7 c8 f4 07 a0 e8 34 08 03 e1 e9 c0
[ 537.616540] RIP [<ffffffffa0063c79>] walk_up_log_tree+0x2f9/0x330 [btrfs]
[ 537.616540] RSP <ffff880058191c08>
[ 537.619306] ---[ end trace 5f9597f2f24fa95c ]---
it's applied on top of today's linus c1d10d18c5. Mount options are
compress and discard.
fsstress process is in D-state with
# cat /proc/12068/stack
[<ffffffffa002d419>] start_transaction+0x79/0x2f0 [btrfs]
[<ffffffffa002da13>] btrfs_start_transaction+0x13/0x20 [btrfs]
[<ffffffffa002e1aa>] __unlink_start_trans+0x4a/0x400 [btrfs]
[<ffffffffa0035997>] btrfs_unlink+0x37/0xe0 [btrfs]
[<ffffffff8118ab6a>] vfs_unlink+0x8a/0xf0
[<ffffffff8118ad53>] do_unlinkat+0x183/0x1c0
[<ffffffff8118bbc6>] sys_unlink+0x16/0x20
[<ffffffff81e94f42>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
shutdown is stuck, sync of btrfs does not proceed due to blocked
transaction commit due to previsou failures.
crash site:
1789 ret = btrfs_free_reserved_extent(root,
1790 path->nodes[*level]->start,
1791 path->nodes[*level]->len);
1792 BUG_ON(ret);
if EAX contains ret, 0xFFFFFFa1, which seems to be EOPNOTSUPP comming from
discard handling in btrfs_free_reserved_extent().
walk_up_log_tree
btrfs_free_reserved_extent
btrfs_discard_extent
return -EOPNOTSUPP
BUG_ON ret
I suspect line 1793 in btrfs_discard_extent():
1766 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1767 u64 num_bytes, u64 *actual_bytes)
1768 {
...
1793 if (discarded_bytes && ret == -EOPNOTSUPP)
1794 ret = 0;
...
1800 return ret;
1801 }
discarded_bytes is initially 0, ret is EOPNOTSUPP from not-done
btrfs_issue_discard, true branch is not taken and nonzero retcode is
propagated.
I think the "discarded_bytes != 0" check should be dropped, I do not see why
it's there. Going to test now.
david
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
