Hello,
> > Well, I have been installing different kind of Unices for about 10 years,
> > the slicing metaphores is not a problem for me.
> > I did indeed put the layout myself. I created: /,
> > swap, /tmp, /var /export/home, /opt and /usr. I usually gace 2Gb, 6Gb, 24
> > Gb etc... per slice. 1 Gb being for me 1024 Mb. However, the system did
> > some rounding for me, so 2Gb = 2048Mb became 2055Mb, for instance.
>
> While you should not put /tmp on a disk, and you really have a lot of
> unnecessary separate filesystems, there's nothing there which is
> likely to cause a write error. It might be worth trying with just /
> and swap to rule out any sector numbering problems, but that sort of
> bug seems unlikely.
I did that. No positive results.
> > I can try again with a new slicing though. I am confident in the HD,
> > the Linux fisk tested the HD for me and did not reveal any faults.
>
> Really? It did a destructive multi-pass write/read/write/read test
> looking for stuck 0s and stuck 1s? And you know it turns off the
> on-disk cache and does direct I/O (via a SCSI-generic layer or
> similar, not read/write/mmap)? And you actually trust Linux's error
> propagation enough to give you EIO if a SCSI or ATA error occurs?
> Unless both the test and the OS's I/O stack are very well constructed,
> it's easy to miss errors of this type. I put no stock in the Linux
> fdisk tests.
OK.... Well, that was a test I had in mind.
I did an other test: I replaced the HD with an other one of the same size,
same vendor, bought at the same time.
No positive results either. Howeve, I am now inclined to rule out any hardware
issue with the disk.
So, why on earth do I have this issue with a simple EIDE disk on an ASUS
P5-VD1 motherboard with a PT880U chipset?
> It might be worth dropping into a shell and trying to write and then
> read back the disk's contents. It's not a great test (you'd need a
> program written for that purpose, as described above) but it's
> something. My money's on the disk.
As I wrote, I think I can rule out the disk, considering that one disk giving
an error can be a problem with the disk but two disks giving me the same
error....
Daniel
--
**************************************************************************
Daniel TOURDE E-mail : daniel.tourde at foi.se
FOI, Swedish Defence Research Agency Tel : +46 (0)8-55 50 32 12
Defence & Security, Systems and Technology Fax : +46 (0)8-55 50 36 51
Department of Autonomous Systems Cellular : +46 (0)70-849 93 40
SE-164 90 Stockholm, Sweden
**************************************************************************
