[vox-tech] need to debug boot crash

[ME]

On Sun, 24 Mar 2002, eric nelson wrote:
> First, something about mount program didn't pass correct address, then
> RPC: sendmsg returned error 101
> nfs: RPC call returned error 101
> .... over and over
> 
> There are so many errors, that I can't scroll back.  I'll need to redo
> the kernel w/ the option Peter Jay Salzman mentioned.
> 
> I'm not doing the kind of mount straight from the bios, but I want to
> learn how to do that one, later.  I have a boot floppy which loads a
> kernel, then gets an address from dhcp server, then mounts on nfs.  
> I'm sure the problem is in init scripts, or fstab or something.
> 
> It's good to know someone is doing this, it's a great approach.

First, check out the netboot howto/docs.

Second, make sure the server is exporting the filesystems in question on a
non-netbooting box/session with normal
# mount -t nfs host.name:/export/path /local/mount/point

Why? You can make sure the server's /etc/hosts.[allow||deny] is set up in
such a way to allow portmap and nfs stuff from a client's IP address to
work.

If that works, then try to test the next step. Start up a netbootable
kernel with loadlin or lilo (special entry on a disk-booting system) to
tell it to netboot instead of use the local disk. Certainly, it will still
grab a kernel from the local disk but shyould do the rest over the network
like it was diskless.

Checkout /usr/src/linux/Documentation/nfsroot.txt

You should be able to add an entry to lilo.conf (or at the lilo
prompt) like:

(Use IP addresses to eliminate DNS as yet another piece to work out.)

LILO:
Boot: mykernel root=/dev/nfs nfsroot=IP.Addr.Of.Srvr:/path/to/root/export
ip=<client-ip>:<server-ip>:<gw-ip>:<netmask>:<hostname>:<device>:<autoconf>*

*= See the above mentioned linux kernel doc for this line.

It is a good idea to test with a hand-enetered IP address for client and
server as well as all other info to eliminate bootp/dhcp from the list of
possible problems.

And you could, of course, have added those items into a separate
lilo.conf entry to save re-entry of those keystrokes every single time.

If that works, then remove the client ip and let everything else be
determined except for server ip,

next drop server IP and let it all be dynamic, and then try to shift to
let the special bootp/dhcp response include the nfsroot. 

(At this point, if all else works, then you would only be passing the:
root=/dev/nfs
)

Next, if you want it to be true network booting (bootp/dhcp then tftp of
kernel, and finally boting kernel get nfsroot and goes) then you will
likely need some sort of modification to your final compiled kernel that
would be dl via tftp (a boot strapper of sorts.) I use the netboot stuff
with programmed EPROMS dropped into the ethernet cards. (
http://sourceforge.net/projects/netboot )

I have found testing each part, one-at-a-time save troubleshooting and
leads to a steady advance to solutions. 

Of course, there is a great sense of accomplishment when you take a big
project with lots of pieces, throw it all together and note that it all
works the first time too. ]:>

-ME

-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GCS/CM$/IT$/LS$/S/O$ !d--(++) !s !a+++(-----) C++$(++++) U++++$(+$) P+$>+++ 
L+++$(++) E W+++$(+) N+ o K w+$>++>+++ O-@ M+$ V-$>- !PS !PE Y+ !PGP
t@-(++) 5+@ X@ R- tv- b++ DI+++ D+ G--@ e+>++>++++ h(++)>+ r*>? z?
------END GEEK CODE BLOCK------
decode: http://www.ebb.org/ungeek/ about: http://www.geekcode.com/geek.html