Problems using etherboot 3.1 - long

[Harold Baur <hb@hoyte.nl>]

[Problems using etherboot 3.1 - long]

Hello,

I'm trying to get a netboot going using etherboot 3.1. I'm running into
a lot of problems, which are documented below.  I'd like to get some help
on how to solve these. If you're a busy person you better stop reading.

SYNOPSIS:
BootServer: HP 9000/720 runing HP-UX
Diskless client: PC running Linux 2.0.29
Problems:
1. can't make a bootrom image.
2. can't boot of of NFS mounted fs due to problems
   with exported block/character special files
3. can't boot of of ramdisk image


Ad 1.
	The configure script generates errros. As a result of this
I cannot do a "make bootrom" as suggested in the INSTALL file.
Here's the output:

checking whether char is unsigned... no
checking definition of __uxx in types.h... yes
checking whether byte ordering is bigendian... no
updating cache ./config.cache
creating ./config.status
creating Makefile
creating mknbi-dos/Makefile
creating mknbi-linux/Makefile
creating bootrom/make.defs
sed: can't read /home/bo/etherboot-3.1/netboot-0.5.3/./bootrom/make.defs.in: No such file or directory
mkdir: cannot make directory `bootrom/utils/utilsrc': No such file or directory
creating bootrom/utils/utilsrc/Makefile
./config.status: bootrom/utils/utilsrc/Makefile: No such file or directory
creating config.h

Therefor, I'm using a bootrom image (image.flo) from an older package,
etherboot 2.0. Hope that's okay.


Ad 2.

	After compiling a suitable kernel, running

mknbi-linux -x -d rom -i rom -k zImage -o bootImage

Kernel image file name  = zImage
Output file name        = bootImage
Kernel command line     = "auto rw root=/dev/nfs nfsroot=rom nfsaddrs=rom"

on it, placing this on server, and configuring all the neccessary stuff
(/etc/bootptab, /etc/exports), setting up an appropriate fs, I boot from
the bootrom flop. The kernel gets loaded just fine. Then the filesystem
gets mounted and I get the message "unable to open initial console".

I figured this is a device problem, and indeed when I look at the device
numbers on the NFS mounted system I see that all major numbers are 0.

Example:
ls -l /dev/udp on linux :
crw-------   1 root     sys       18,   4 Feb 19  1994 udp

cp them to server (HP-UX):
crw-------   1 root     root      18 0x000004 Feb 19  1994 udp

All seems okay, now NFS mount:
crw-------   1 root     root       0,   4 Feb 19  1994 udp

Oops. Major number is 0.  This appears to be a problem with nfsd
on the HP-UX, but I am not sure. The documentation doesn't mention
it at all.


Ad 3.
	Since I can't do an NFS mount I'll do it with a ramdisk.
After preparing a compressed filesystem image, I run

mknbi -v -d ram -i rom -k zImage -r fsZimage -o bootImage

Kernel image file name  = zImage
Output file name        = bootImage
Ramdisk image file name = fszImage
Kernel command line     = "auto rw root=/dev/ram nfsroot=kernel nfsaddrs=rom"

on it, put on server, there we go again. Kernel loads just fine, starts up
and then gives me:

Kernel panic: VFS: Unable to mount root fs on 01:00

and there is silence.

I'm guessing 01:00 are the major and minor numbers of root device.
In my case that's the ramdisk device, which makes sense. In the
prepared filesystem I use, I have:

lrwxrwxrwx   1 root     root            4 Jun 12 17:27 ram -> ram0
brw-r--r--   1 root     root       1,   0 Jun 12 17:27 ram0
brw-r-----   1 root     root       1,   1 Jun 12 17:26 ram1

Looks okay to me.



CONLUSION:

	I think I did RTFM. If I am missing something really obvious
please point it out anyway. I would like to get this going really bad.
Any solution is greatly appreciated, for any of the above problems.

Best regards,
		Bo.