hello, my xfs has a probelm. does anybody know this?thank you

[Eric Sandeen]

On 8/25/15 4:08 AM, zhengbin.08747 at h3c.com wrote:
> My OS is redhat7.1, the version of Linux kernel is “Linux version 3.10.0-229.el7.x86_64”
> 
> Sometime xfs give a message like this
> 
> Aug 11 15:51:05 localhost kernel: XFS (dm-4): metadata I/O error: block 0x7d9a010 ("xfs_trans_read_buf_map") error 117 numblks 16
> Aug 11 15:51:05 localhost kernel: XFS (dm-4): xfs_imap_to_bp: xfs_trans_read_buf() returned error 117. 
> Aug 11 15:31:08 localhost kernel: XFS (dm-4): Metadata corruption detected at xfs_inode_buf_verify+0x75/0xd0 [xfs], block 0x7d9a010
> Aug 11 15:31:08 localhost kernel: XFS (dm-4): Unmount and run xfs_repair

Did you try running xfs_repair (or possibly better for the first run, do xfs_repair -n, to do a check-only run?)

> Aug 11 15:31:08 localhost kernel: XFS (dm-4): First 64 bytes of corrupted metadata buffer:
> Aug 11 15:31:08 localhost kernel: ffff88089144a000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
> Aug 11 15:31:08 localhost kernel: ffff88089144a010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
> Aug 11 15:31:08 localhost kernel: ffff88089144a020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
> Aug 11 15:31:08 localhost kernel: ffff88089144a030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................

So it seems to have read a completely zeroed-out block.

Isn't there a stack trace after this part of the message?
Hm I guess not; if you can hit it again, try 

# sysctl -w fs.xfs.error_level=11

to turn up the error reporting level.


Any chance that this is a thinly provisioned device?  Or what type of dm device is it?

> So is this a bug of xfs? or is the device’s error(the device actually did not save the data, but give xfs success message)?

Hard to say at this point.

-Eric