[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: External SCSI disks with SS20

Hartwig Atrops wrote:

Hi all.

I try to run an external SCSI disk box (3 disks) with a Sparcstation 20 (Debian 3.0 Woody, kernel 2.4.18 SMP, 2 x SuperSparc II 75 MHz, no CDROM). Disk array: Sun P/N 595-3769-01, three 2GB disks (one slot is empty), narrow SCSI (50 pin).

The SS20 starts booting, linux bootup recognizes 4 SCSI disks: three times ST32550N (external) and a MAB3045S internal 4GB disk. External: SCSI ID 0-2, internal 3. This one is the boot disk. When I boot with the disk array powered down, the SS20 boots without problems.

With external disks on it ends up with:

  Kernel panic: VFS: Unable to mount root fs on 08:01
  Press L1-A to return to the boot prom
  spin_lock(f0209000) CPU#0 stuck at f00599a8, owner PC(f0027124):CPU(1)
  ... ( last line periodically repeated) ...

What is going wrong? Does my SS20 try to use an external disk as root filesystem? Which device is 08:01?

Thanks in advance,

Hi Hartwig,

as my first reply was send as a direct mail to you instead to this list I'll summarize my answers again in this email to keep the list in the loop.

it seems that your problem is related to the scan sequence of your SCSI drives.
See http://www.tldp.org/HOWTO/SCSI-Programming-HOWTO-4.html#ss4.3

Aber irgendwas gefaellt mir da nicht. Suns booten per default von SCSI-ID 3,
Suns will booten 'by default' from the device set in the 'boot-device' parameter of the NVRAM ;-) Whereby I currently don't know the behaviour of OpenBoot when the 'boot-device' parameter is set to only 'disk' instead e.g. 'disk3'.

darauf nimmt Linux wohl keine Ruecksicht. Ich hatte da schon oefter Aerger, hatte aber noch nicht verstanden warum.

As the boot process is initiated by OpenBoot (i.e. loading the MBR of the device set in the boot-device parameter of the NVRAM), the selection of the boot device schould behave the same regardless of SUN OS or Linux as a operating system.

Unfortunally, as layed out in the above link, Linux does assign the device reference as the SCSI devices are found during the scan of the SCSI-bus(es) the relationship between SCSI-id and device reference could be changed by adding devices. (which happens in your case)

While this has no effect on loading the MBR and the second stage boot loader from the boot device, it will create chaos as soon as the operating system referes to the device reference.

In your case as soon as the boot process will mount sdaX as /. As it seems that your Linux was installed and configured without the external disks it will expect the root partition on 'sdaX' (SCSI-id 3). Therefore when you boot your system with the external disks, the boot process expects the root partition again on 'sdaX' which now references to SCSI-id 0.

It furthermore seems that there is no suitable partion on SCSI-id 0 and therefore you'll get the kernel panic mentioned in your first mail.


Reply to: