Metadata corruption detected at xfs_attr3_leaf_write_verify / Internal error xfs_attr3_leaf_write_verify
Eric Sandeen
sandeen at sandeen.net
Tue Dec 8 11:01:44 CST 2015
On 12/8/15 9:20 AM, Dietmar Putz wrote:
> Hello all,
...
> [73715.250762] XFS (sdc1): Metadata corruption detected at xfs_attr3_leaf_write_verify+0xe5/0x100 [xfs], block 0x44482d730
> [73715.250803] XFS (sdc1): Unmount and run xfs_repair
> [73715.250816] XFS (sdc1): First 64 bytes of corrupted metadata buffer:
> [73715.250832] ffff8807f7b07000: 00 00 00 00 00 00 00 00 fb ee 00 00 00 00 00 00 ................
> [73715.250856] ffff8807f7b07010: 10 00 00 00 00 20 0f e0 00 00 00 00 00 00 00 00 ..... ..........
> [73715.250876] ffff8807f7b07020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
> [73715.250907] ffff8807f7b07030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
...
> [175683.930045] ffff8807a8f59000: 00 00 00 00 00 00 00 00 fb ee 00 00 00 00 00 00 ................
^forward ^back ^magic^pad ^count^usedbytes
> [175683.932979] ffff8807a8f59010: 10 00 00 00 00 20 0f e0 00 00 00 00 00 00 00 00 ..... ..........
^firstused
> [175683.935812] ffff8807a8f59020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
> [175683.938644] ffff8807a8f59030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
> [175683.940870] XFS (sdc1): Internal error xfs_attr3_leaf_write_verify at line 216 of file /build/linux-XHaR1x/linux-3.13.0/fs/xfs/xfs_attr_leaf.c. Caller 0xffffffffa01936f0
Line 216 of xfs_attr_leaf.c: in 3.13 is:
if (!xfs_attr3_leaf_verify(bp)) {
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr);
>From your xfs_info output, this is not a crc-enabled filesystem.
So xfs_attr3_leaf_verify() only checks these 2 things:
if (ichdr.magic != XFS_ATTR_LEAF_MAGIC)
return false;
that magic is 0xfbee, so good there... and:
if (ichdr.count == 0)
return false;
so that's the problem, apparently. If I'm reading it correctly, it's trying
to write an empty attribute block to disk; this shouldn't happen.
If you have traces, perhaps we can see which inode it's working on, and learn something
from examining it in xfs_db...
-Eric
More information about the xfs
mailing list