> At 12:05 23-10-2001, Steve Lord wrote:
> >Hi Arjen,
> >I suspect you are being hit by the scsi layer here - you are seeing two
> >error reports, the first one is the scsi layer reporting an error, the
> >second is xfs reporting it got an I/O error from the scsi layer.
> >Apart from hex vs decimal, they are both complaining about the same
> >disk block.
> >Have you tried some raw I/O tests to the same device without a filesystem
> >on it?
> I'll try and do some testing (specific suggestions are welcome); I am
> pretty sure this is a SCSI layer issue, especially since any kernel before
> 2.4.10 will not even see the partition on LUN 1; I was hoping maybe
> somebody had encountered the same problem.
> What In do find strange however, is that xfs_check/xfs_repair see no errors
> on the filesystem, considering the error message about errors writing metadat
Well, they are read only unless they find and error and the error you
got was a write. And just because the write returned and error it does
not mean it did not work.....
As for a test, try doing a dd down to the same disk addresses from
user space for starters.