Hi Paul -
First I'd make sure you've got the latest xfsprogs, what version
are you running?
We've recently done some work with large block device kernels,
and I have used mkfs.xfs on up to 3T filesystems on ia32. I did
notice that your failing sectors do have the 32nd bit set, any driver
down the line could have problems with sign extensions - but if mkfs.ext3
works, that probably rules out underlying drivers.
I'm sure folks on the list are tired of our "try newer code" answers,
but it's a good place to start, to make sure we're not chasing old
bugs. You might even try a newer kernel - if you don't want to expose
your data to a different kernel, you could just use it to test mkfs
without mounting your existing filesystems on the raid.
Thanks,
-Eric
|