Troubleshooting the problem "Terminal won't download the kernel and boot"

When troubleshooting the boot process, having a grasp of what is going on during boot is surely helpful. Etherboot and PXE might give you some hints.

We'll try to walk you through different common problems encountered during terminal boot.

Floppy disc unreadable

Etherboot floppy boot: After first accessing the floppy disk, the screen is filled with lines like 0200 0cfe 0000 0000

This means your floppy disc went bad. Try another one.

Etherboot F?F?F? error

Etherboot starts, prints "Searching for DHCP... F?F?F?

This means Etherboot receives DHCP offers that lack the "filename" field. So etherboot cannot determine which kernel file to load.

A typical entry would look like:

   option filename         "/lts/vmlinuz-2.4.26-ltsp-2";

Please read in DHCP for more info.

Etherboot "access denied", "file not found" errors

After printing the correct filename to download (check that for typos, just in case...) etherboot claims some problem like "access denied" or "file not found"

There can be several common mistakes that cause this behaviour. One certainly is to specify an incorrect filename. Check if your TFTP-Daemon is setup with the "-s" parameter (meaning pre-pending a given path to any file request).

Example: TFTP-Daemon is setup with "-s /tftpboot", Etherboot requests "/lts/vmlinuz-2.4.x" => File /tftpboot/lts/vmlinuz-2.4.x

The file to be downloaded should be at least r--r--r-- (0444 in unix slang), the directories structure containing it should be at least r-xr-xr-x (0555 in unix slang).

If all those hints didn't resolve your problem, you should try to download the kernel file from a running Linux box (e.g. from the server itself) with the command-line utility "tftp". ("cd /tmp" first to store the downloaded file in a place where it doesn't overwrite important files in any case).

tftp 192.168.0.1
get /lts/vmlinuz-2.4.x

would be the commands to run (replace 192.168.0.1 with your server's ip address and adjust the filename to what your dhcpd.conf is setup for).

Etherboot seems to start downloading, prints dots very slowly (every x seconds or so)

Etherboot sends file/read requests to the TFTP server but does not receive answers

This is often caused by the TFTP daemon not running at all. To find out wether a TFTP daemon is running at all, enter

netstat -lnp | grep ":69"

as "root" on your server. This should give you a line like the following:

udp 0 0 0.0.0.0:69 0.0.0.0:* 1927/inetd

If you don't get this line, TFTP is not running. You'll have to setup a TFTP daemon then.

Another common mistake is to not setup the firewall correctly. UDP traffic from and to port 69 has to be allowed for all thin clients.

Lastly ensure the BIOS on the thin client is up to date. The PXE boot code is part of the BIOS and may have bugs in it. If the dots stop at a random point along the kernel or initrd transfer that is a sign that the BIOS on the client needs updating.

PXE gives an error message "NBP is to big to fit into memory"

You are using PXE, not Etherboot, and get this error.

It means you are pointing the PXE to the wrong file. PXE can only load very small files, "2nd stage loaders", that in turn can load whole linux kernels. You most probably have to adjust the filename to point to "pxelinux.0" in the appropriate directory. http://www.ltsp.org/README.pxe might help you setting up everything correctly.

---

At this stage, the linux kernel should be loaded and start printing the well-known messages about hardware. If not (so if you encounter any different error before the kernel was loaded), feel free to enter information here.

-- AnselmMartinHoffmeister - 14 Dec 2004

Helpful Links

http://www.emboot.com/faq-pxe.shtml[PXE Client Status/Error Codes]

-- MohammedArafa - 25 Mar 2005