http://oss.sgi.com/bugzilla/show_bug.cgi?id=761
------- Additional Comments From nathans@xxxxxxxxxx 2007-07-12 15:46 CST
-------
On Wed, 2007-07-11 at 05:43 -0700, bugzilla-daemon@xxxxxxxxxxx wrote:
> ...
> We have 2.5 TB hardware RAID with one large XFS partition on it. Recently
> there
> was a short power outage which rendered filesystem unmountable. When I tried
> to
> use xfs_repair it seg. faulted. I tried to debug xfs_repair and I found out
> that
> pread64 gives error 33 (Invalid argument) (libxfs/rdwr.c, function
> libxfs_readbufr). I guess the cause is that this RAID has hardware sector
> size=1024 bytes and not the typical 512 bytes.
>
Did mkfs create this filesystem with 1K sector size? If not, why not?
(It should have done so automatically ... else your RAID driver is not
functioning correctly).
Still, the gdb stack trace from a failing xfs_repair run would be of
some interest, I'm sure. Can you do an xfs_metadump on this filesystem
for Barry? Although, without the 1k sector hardware, I guess this will
be hard to for him to reproduce - making a gdb trace even more valuable.
> I was able to recover the filesystem by copying it to another disk using dd
> and
> accessing it using loopback device. Nevertheless there should be possibility
> to
> repair filesystem in place on devices like ours.
>
libxfs should align to the filesystem sector size. If yours is set to
512 bytes, then the filesystem is not configured correctly. If its 1k,
then this indeed would be a libxfs problem. What does the output from
xfs_info report on this filesystem (needs to be mounted for that).
cheers.
--
Nathan
--
Configure bugmail: http://oss.sgi.com/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
|