Very Computer

Hi there!

I have really big problems. I try to set up a 45 GB Western Digital
Caviar 450AA (secondary IDE controller, master-drive). I patched my
machine with the recommended patch cluster which also contains patch no.
110202-01 for the ata-drivers. I am able to fdisk the drive with a
geometry file using fdisk with options. I am also able to partition the
disk with the native solaris 8 x86 format tool. I label the disk - fine
so far. But when I try to build a filesystem on it, newfs terminates wih
errors. I afterwards tried to set up a scsi-disk with 4,1 GB (IBM DCAS
34330). The same problem occured: newfs (as well as mkfs with options)
terminated with the following error message:

mkfs: bad value for size: 0 must be between 1024 and 0
mkfs: size reset to default 0
mkfs: bad value for ntrack: 6 must be between 1 and 0
mkfs: ntrack reset to default 16
seek error on sector -1: invalid argument

That's pretty strange, isn't it? I don't know wheter Solaris x86 is that
different from Solaris SPARC edition. Perhaps I'm too dumb to get this
machine working and should put my fingers off Solaris x86 edition.

Could anybody in this world help me? Or has anybody the same problems?
You know - it's just for my collegues: they are now calling me brad...

Greetings

SERGEJ

this is what i used on my 60gb ibm drive.
/J?rgen
--------------------------
* MAXCYL IN solaris 8 without ata patch is 16838
*
* BIOS reports 61496MB 7476 255 63
*                      CYL  HD  SEC
* recalc 7476*255*63 = 120101940 blocks (512byte)
*
* solaris 8 convert 120101940 /16383 = 7330.89 sectors per track
*
*PCYL  NCYL  ACYL BCYL HEADS NSECT SECTSIZE
16383  16383 2    0    1     7330  512
*
* layout syntax: fdisk -S "thisfile" -I /dev/rdsk/c0d1p0

I'm guessing that's supposed to mean you're running Solaris 8 x86.
You should state such things up front and not expect people to
be able to correctly guess what release you're running.

I haven't tried installing it, but AFAIK, the README for the latest
Solaris 8 x86 recommended patch cluster does *not* list 110202 (nor
109798).

Are you certain you've got 110202 and/or 109798 and the latest ata
driver? If so then you must either be running something more recent
than Solaris 8 x86 FCS and/or you must've meant you installed some MU
release, not a recommended patch cluster, or you might have meant you
installed 110202 on top of the latest recommended patch cluster.

Quote:> I am able to fdisk the drive with a
> geometry file using fdisk with options.

I don't think anything's changed in fdisk recently so I don't see
why you think a geometry file is necessary. You should just
run fdisk without any extraneous args. Re-run fdisk, delete the
partition you created, exit fdisk, then re-create your Solaris x86
partition without any extra command line args. If you do that, what's
the output of "fdisk -v -W - /dev/rdsk/c1d0p0"? Make certain you
quote the command line also, since for all I know you might not
be pointing fdisk at the right device.

Quote:> I am also able to partition the
> disk with the native solaris 8 x86 format tool. I label the disk - fine
> so far. But when I try to build a filesystem on it, newfs terminates wih
> errors.

Any special reason you didn't quote the exact newfs command you
ran and all of its output?

If that's all you did, then I'd guess that you labeled the disk but
you didn't create any slices. Or, you setup the VTOC and you didn't
write it to disk.

Create a s0 slice in your VTOC and write it and the label out to disk,
exit from format, and then what's the output of "prtvtoc /dev/rdsk/c1d0s2"?

Then exactly what happens when you do "newfs /dev/rdsk/c1d0s0"?

Quote:> I afterwards tried to set up a scsi-disk with 4,1 GB (IBM DCAS
> 34330). The same problem occured: newfs (as well as mkfs with options)
> terminated with the following error message:

Would it be too difficult for you to capture and post the exact
sequence of commands (and they're outputs) you ran? That's the
simplest way for someone to spot what you did wrong. If you just
summarize the commands and results there's no way to know what
you've left out or described incorrectly.