["Followup-To:" header set to comp.sys.sgi.admin.]
On 11 Aug 2004 08:39:12 -0700, Ian Mapleson wrote:
> So far, when I run the script on the 6.5 disk, it correctly sets
> SystemPartition and OSLoadPartition, but after rebooting the
> 5.3 startup gives the normal copyright message followed by this error:

Are these all NVRAM variables that are set?

> WARNING: initial mount of root device 0x2000010 failed with errno 22
>
> A few seconds later the screen clears and then this error is displayed:
>
> PANIC: vfs_mountroot: no root found

Maybe it is still trying to mount the root from the 6.5 disk
which is XFS that might not be readable for 5.3.

> Does anyone know what's happening? What is vfs_mountroot? I can't
> find any reference to it on techpubs, and there's no man page for it
> on either installation. I tried using grep to see if any of the scripts
> in /etc would be responsible for printout out either message, but none
> of them are. I'm going through bcheckrc and brc at the moment, trying
> to see how things are initially mounted, but it's a bit of a shell
> script decyphering job. Any ideas?

I guess it's happening way before anything from /etc is executed.
After all... no root filesystem, no "/etc", right?



Gerhard

On Wed, 11 Aug 2004 18:05:22 +0200, Gerhard Lenerz wrote:
> ["Followup-To:" header set to comp.sys.sgi.admin.]
> On 11 Aug 2004 08:39:12 -0700, Ian Mapleson wrote:
>> So far, when I run the script on the 6.5 disk, it correctly sets
>> SystemPartition and OSLoadPartition, but after rebooting the
>> 5.3 startup gives the normal copyright message followed by this error:
>
> Are these all NVRAM variables that are set?

Try also setting root to an appropriate value for both cases. As
I have seen it isn't required with one disk installed (even if the
boot disk is on #2 and another one on #1 exists), but maybe IRIX
gets confused somehow with two system disks.

I just tried setting root to the wrong value on my Octane so that
it loads the kernel from disk #2 and is asked to mount root from
the wrong disk (#1). The kernel comes up, then complains that it
cannot mount it's root. It then continues with it's default overriding
the wrong NVRAM setting. The message looks similar to what you get.


Gerhard

Ian Mapleson wrote:
> Dear All,
>
> I'm trying to setup an Indigo2 so that it can boot with either
> 6.5 or 5.3. Not wishing to make things too complicated, all I
> want is a script for each installation which one can run as root
> that will change the appropriate nvram settings and then reboot.
>
> The disk on SCSI ID 1 has a 6.5.15m installation. The 5.3 disk
> is on ID 2 (5.3 with XFS to be precise).
>
> So far, when I run the script on the 6.5 disk, it correctly sets
> SystemPartition and OSLoadPartition, but after rebooting the
> 5.3 startup gives the normal copyright message followed by this error:
>
> WARNING: initial mount of root device 0x2000010 failed with errno 22
>
> A few seconds later the screen clears and then this error is displayed:
>
> PANIC: vfs_mountroot: no root found
>
> Does anyone know what's happening? What is vfs_mountroot? I can't

Check that following params are set in the environment:

OSLoader=sash
OSLoadFilename=/unix

I once had a similar problem (dont remember the exact error messages,
but I belief they were different) trying to boot from a disk connected
to the external SCSI. These two params solved it. Seem to be the
default, but required when booting from other disk

Rainer

> find any reference to it on techpubs, and there's no man page for it
> on either installation. I tried using grep to see if any of the scripts
> in /etc would be responsible for printout out either message, but none
> of them are. I'm going through bcheckrc and brc at the moment, trying
> to see how things are initially mounted, but it's a bit of a shell
> script decyphering job. Any ideas?
>
> If I restart and do a resetenv in Command Monitor, all returns to normal
> and 6.5 boots fine. If I manually put the 5.3 disk back to SCSI ID 1, it
> also boots ok, so whatever is going wrong at least I know it's not
> harming the two installations.
>
> Thanks for any help!
>
> Cheers!
>
> Ian.

Brent Casavant wrote:
> Another one to check is:

Yep, tried clearing that. No effect.


> Oh, and vfs_mountroot is just the name of the kernel routine that
> tries to mount the root filesystem.

As near as I can make out, consensus in past posts (thanks to Tony
for the search suggestions!) is that the kernel has some record
built into it of where to boot from and this needs to be changed.
Mark Mitchell said in 1997 on .admin:

***************************
From: Mark Mitchell (mark@k-par.co.uk)
Subject: Re: Dual boot capability
View: Complete Thread (3 articles)
Original Format
Newsgroups: comp.sys.sgi.admin
Date: 1997/02/25

> I have a need to be able to boot an Indigo2 system under either 5.3 or 6.2.
> I have two drives (one with either operating system), and I can mount the
> 5.3 drive under 6.2, but the need is to be able to boot under the OS of
> choice without having to physically swap drives. So far, this capability is
> completely eluding me. Any pointers would be greatly appreciated. Thanks.

Mike,
I've got an IRIX 5.3 disk on ID 1, and a 6.2 disk on ID 2. To boot
IRIX 5, I run the following script:

nvram -v SystemPartition "dksc(0,1,8)"
nvram -v OSLoadPartition "dksc(0,1,0)"
/etc/reboot

or to boot IRIX 6:

nvram -v SystemPartition "dksc(0,2,8)"
nvram -v OSLoadPartition "dksc(0,2,0)"
/etc/reboot

I hope this helps.

By the way - be careful if you're changing the scsi IDs of these disks. E.g. if
you've currently got both disks on ID 1. The scsi ID (or some reference to it from
/dev/root) is compiled into the kernel: if you change the scsi ID of the disk and
try to boot - it'll fail - or it'll load the kernel off the disk at ID 1.

If you do have this situation, (and you don't want to re-install):
- set your 6.2 disk to scsi ID 1
- boot
- change /dev/root and /dev/rroot so that the minor numbers are correct for the
new scsi ID: it'll be the same as the fs partition containing unix:
e.g. for ID 2, partition 0:
ls -l /dev/rdsk/dks0d2s0
brw------- 2 root sys 128, 32 Oct 8 11:49 dks0d2s0
giving a major number of 128 and a minor number of 32.
Create the new devices with:
mknod /dev/root b 128 32
mknod /dev/rroot c 128 32
- run /etc/autoconfig
- Shutdown, set the scsi ID to the new ID and boot.

This last section should probably have "use at your own risk" and "may be a bad
idea" plastered all over it.

Mark.
***************************

What do you reckon? Also, one other poster thought that the swap reference
may have to be changed in the same way; if so, how?

I'll setup a basic 5.3 disk to test this out, make sure I don't wreck
my main complete 5.3 disk.

Thanks for the replies so far!

Ian.

Gerhard Lenerz wrote:
> Are these all NVRAM variables that are set?

Yes. kernname is set automatically, as is root if it's cleared.


> Maybe it is still trying to mount the root from the 6.5 disk
> which is XFS that might not be readable for 5.3.

Due to the kernel issues mentioned in my other reply, I think this
is what's happening even though initially it must be grabbing the
unix file from the 5.3 disk (otherwise the 5.3 copyright wouldn't
appear I would have thought).

When installing the 6.5 disk, I forgot to fx the disk using a 6.2
boot CD, so yes 5.3 can't use it. My intention, once the dual-boot
was up & active, was to clone the 6.5 disk elsewhere, re-fx/mkfs
the disk with 6.2, then clone the 6.5 data back again. Both 5.3 with
XFS and 6.5.x can mount XFS disks which have been initialised with
6.2. Thus, the 5.3 disk will have the 6.5 disk mounted on /6.5
for access, and vice versa. The 6.5 disk already has the 5.3 disk
mounted on /5.3, works fine (the 5.3 disk is a 9GB, so 5.3 with
XFS is essential).

Cheers!

Ian.

>>>>> "mapesdhs" == Ian Mapleson writes:

mapesdhs> Brent Casavant wrote:
>> Another one to check is:

mapesdhs> Yep, tried clearing that. No effect.

Here is an example of script which we use to do "multiboot"

#! /bin/sh
/sbin/nvram OSLoadPartition 'dksc(0,1,11)'
/sbin/nvram root 'dks0d1s11'

/sbin/telinit 6

In this case, all partitions are on the same disk and all are variants
of 6.5, so we keep one SystemPartition and boot different releases of
6.5.

You could consider using 6.5 sash for both and just change to 5.3
OSLoadPartiton and root - AFAIK there is no requirement to have both
sash and root on the same disk and as long as both disks are
connected, it should "just work (tm)".

>> Oh, and vfs_mountroot is just the name of the kernel routine that
>> tries to mount the root filesystem.

mapesdhs> As near as I can make out, consensus in past posts (thanks to Tony
mapesdhs> for the search suggestions!) is that the kernel has some record
mapesdhs> built into it of where to boot from and this needs to be
mapesdhs> changed.

No, that is (was?) Linux way of doing things, Irix has no information
about root device in the kernel whatsoever: you can boot from
different disk, from NFS, from CD, all with the same kernel.

max

Max Matveev wrote:
> No, that is (was?) Linux way of doing things, Irix has no information
> about root device in the kernel whatsoever: you can boot from
> different disk, from NFS, from CD, all with the same kernel.

Either that or the devices mentioned in the post I quoted, so I'm
going to try that. I certainly know that settings I'm trying are as they
should be, but it still looks as if the system is trying to use
disk 1. Must create a fresh 5.3 first...

Cheers!

Ian.

All solved! Three device files must be changed: root, rroot and swap.
I've added section on how to do it to my misc problems page:

http://www.futuretech.blinkenlights....blems.html#008

Thanks for all the advice!

Cheers!

Ian.