Comments on: Network boot with Debian

By: Alessandro

Alessandro — Sat, 15 Aug 2009 18:46:14 +0000

Thankyou for this great howto. Debian wiki is fairly complete but missing some details in configuring PXE kernel boot params. Here I’ve found what I was looking for. Thankyou again.

d3p

By: Strange problem on network booting « Sickels blogg

Strange problem on network booting « Sickels blogg — Sat, 14 Feb 2009 11:35:43 +0000

[...] package replaced by nfs-user-server. I don’t know why, may be that I folloved Adam Pierces’ instructions too [...]

By: adam

adam — Tue, 16 Dec 2008 22:59:51 +0000

I think the key is “No init found”. Check that your path to the root file system is correct and that the root file system contains the file /sbin/init

By: Dafydd

Dafydd — Tue, 16 Dec 2008 13:20:56 +0000

nealy working but i get this error any ideas?
device=eth0, addr=10.10.10.15, mask=255.255.255.0, gw=10.10.10.1,
host=10.10.10.15, domain=, nis-domain=(none),
bootserver=10.10.10.254, rootserver=10.10.10.254, rootpath=
Looking up port of RPC 100003/2 on 10.10.10.254
Looking up port of RPC 100005/1 on 10.10.10.254
VFS: Mounted root (nfs filesystem).
Freeing unused kernel memory: 252k freed
Warning: unable to open an initial console.
Kernel panic – not syncing: No init found. Try passing init= option to kernel.

By: Tim

Tim — Mon, 16 Jun 2008 09:59:27 +0000

The problem was the kernel, it couldn’t find it… in the “/tftpboot/pxelinux.cfg/default” file I changed the kernel into the full kernel path and now it work fine…

But now i got the following error:
‘VFS: Cannot open root device “export/dvc/stbroot”
Please append a correct “root=” boot option
Kernel panic – not syncing: VFS: Unable to mount root fs on …’

Anyone help?

Greetz Tim

By: adam

adam — Sat, 07 Jun 2008 11:54:33 +0000

Check that your tftp service is working properly. You can test it by loading up a tftp client and trying to download the kernel with it.

By: Tim

Tim — Fri, 06 Jun 2008 11:51:47 +0000

I tried this too… but i got a problem… after loading the pxelinux.cfg it says: “boot failed: press a key to retry or wait for reboot”

Anyone has a solution?

By: Jeff

Jeff — Wed, 13 Feb 2008 23:10:48 +0000

I’ve got it all up and running.

I suggest anyone interested in this check out http://www.onesis.org for some handy utilities to copy root file systems and create symlinks to ram copies of files to allow errorless NFSRoot booting.

-Jeff

PS – More to come when I clean up my notes.

By: Jeff

Jeff — Fri, 08 Feb 2008 15:38:34 +0000

I was using a stock Debian 4.0 kernel (2.6.18). It didn’t appear to have NFSRoot support when I went in to rebuild the kernel. I had to enable that, the IP stuff, and in my case, it turns out I’ve got a Via Esther 1GHz processor which uses Via Rhine II NIC, so I added that in as well.

I’m just about to try the rebuilt kernel now…

-Jeff

By: adam

adam — Thu, 07 Feb 2008 22:25:57 +0000

Yes, good point. I was using a Via EDEN motherboard. The NIC driver for that is built in to the standard Debian kernel. NFSRoot is also enabled by default in Debian so it Just Worked for me.

But if you have different hardware, then yes you might need to build a new kernel.

The fstab you should change is the one in /export. Here’s mine:

none /proc proc defaults 1 0
/dev/ram0 /tmp ext2 defaults 0 2
203.35.65.33:/export/stbroot / nfs ro 0 1

By: Jeff

Jeff — Thu, 07 Feb 2008 21:29:47 +0000

I think I’ve got to go back and make sure my IP support, NFSRoot support and NIC driver are in my kernel… you may want to mention that. :)

-Jeff

By: Jeff

Jeff — Thu, 07 Feb 2008 21:19:15 +0000

Well I’m kind of stuck. I have the kernel loading but it fails on the NFS mount.

You mentioned I had to edit my /etc/fstab file to tell it to nfs mount the root parition. Which /etc/fstab do I edit, the one in the /export directory?

Right now it says “Cannot open root device “nfs” or unknown-block(0,255)” and then panics.

Through experimentation I discovered this is the /dev/nfs line in the pxelinux.cfg directory config for the machine (ie making it /dev/jeff alters that error).

It almost appears as if the kernel doesn’t have a valid IP so it can’t NFS mount the directory (which appears to be properly exported on my NFS host). Anyway I can confirm if the kernel has an IP that would allow it to talk? It isn’t responding to pings at the assigned boot address.

-Jeff

By: adam

adam — Thu, 07 Feb 2008 10:06:28 +0000

Yes, that should work just fine.

There are also more complicated ways of doing it.

I once had to set up a whole bunch of diskless clients and didn’t want to fill up the server’s disk with a dozen copies of the OS so I separated the files into stuff that could be shared (/bin, /boot, /usr and so on) and stuff which needed to be specific to each box (/etc, /var etc.) and wrote a complicated script to mount them based on the box’s address.

By: Jeff C

Jeff C — Thu, 07 Feb 2008 03:40:06 +0000

Thinking about this further, if I had a bunch of localized config information, I should be able to create /stbroot/IPOfMachine/localfiles on the NFS server and just mount each one of those unique directories based on the IP the box gets (which I’ll hard code in the DHCP conf). That should work, right?

-Jeff

By: Jeff C

Jeff C — Thu, 07 Feb 2008 03:33:41 +0000

Great, thanks!

In my case, the device has a CF to IDE adapter and I have a working version of exactly what I want to replicate, running off the CF card, I just want to remove the CF card and network boot the boxes.

So essentially copy the contents of “/” to my tftpboot/stbroot, as is. No changes at all? :) Cool.

Now what if I want to have 35 of these little beasts? I’m guessing some config files need to be truly localized and some network mounted?

-Jeff

By: adam

adam — Thu, 07 Feb 2008 03:10:11 +0000

The easiest way is to get an old hard disk, put it into your Via box and install Linux onto it. The contents of the hard disk are now your root file system.

Pull the hard drive out of your Via box and put it in your PC. Copy the data to the stbroot directory.

You’ll also need to edit /etc/fstab to mount root via NFS.

I have successfully done it this way.

Alternatively, there is a Debian package called fai-client ( http://packages.debian.org/etch/fai-client ) which supposedly automates the installation of diskless clients but I haven’t tried it.

By: Jeff C

Jeff C — Thu, 07 Feb 2008 02:49:29 +0000

Hi Adam,

Great article, but I’m kind of stuck without knowing how to prepare the root filesystem I want to use. Can you point me in the right direction? I want to PXE boot some Via boxes with PVR-150s, running Debian 4.0 and VLC. Thanks.

-Jeff