Syslinux - Aug 2008 - pxelinux stopped working

Hello.

Some time ago, everything worked for me.
But now it is very strange, it does no more work (using exactly the same setup)
Do you have any idea how to find out what's wrong?

Here is my situation:
I am trying to boot over PXE.
TFTP server is running. DHCP server is running from the same machine. Both by
dnsmasq.
TFTP server has the following files and directories:
/pxelinux.cfg/pxelinux.0
/pxelinux.cfg/default
I can use 'tftp localhost' (using tftp-hpa) and type 'get
/pxelinux.cfg/default' command which downloads the default config file.
Seems there is problem here on server.

The client initiates DHCP requests. Client receives response from the server,
assigns a new (unique) IP address to itself and it downloads pxelinux.0 file.
Then it starts that file. Then pxelinux tries to download
pxelinux.cfg/123456789ABCD and so on, and also it tries pxelinux.cfg/default and
then it dies with a message "Unable to locate configuration file". But
it's there! :)

Syslog on the server shows a strange message from dnsmasq:
TFTP error 0 TFTP Aborted received from 172.16.73.122
FTFP failed sending /boot/pxelinux.cfg/pxelinux.0 to 172.16.73.122

Whats up? Anybody have any idea? :)
I know pxelinux.0 has been sended to the client for sure, because it is executed
there, and it searches for the default config at the first place ...

Hope somebody could suggest how to track down where the problem is.

Thank you

Tomas M

Selon Tomas M <tomas at slax.org>:

...
> Do you have any idea how to find out what's wrong?
>
> Here is my situation:
> I am trying to boot over PXE.
> TFTP server is running. DHCP server is running from the same machine. Both
by
> dnsmasq.
> TFTP server has the following files and directories:
> /pxelinux.cfg/pxelinux.0
> /pxelinux.cfg/default
> I can use 'tftp localhost' (using tftp-hpa) and type 'get
> /pxelinux.cfg/default' command which downloads the default config file.
Seems
> there is problem here on server.
>
> The client initiates DHCP requests. Client receives response from the
server,
> assigns a new (unique) IP address to itself and it downloads pxelinux.0
file.
> Then it starts that file. Then pxelinux tries to download
> pxelinux.cfg/123456789ABCD and so on, and also it tries
pxelinux.cfg/default
> and then it dies with a message "Unable to locate configuration
file". But
> it's there! :)
>
> Syslog on the server shows a strange message from dnsmasq:
> TFTP error 0 TFTP Aborted received from 172.16.73.122
> FTFP failed sending /boot/pxelinux.cfg/pxelinux.0 to 172.16.73.122
>
This suggest you set on dnsmasq, tftp-root=/boot/
and it does not look to match how files are set on the tftp server as you say
it's on /pxelinux.cfg/
Set tftp-root=/var/lib/tftpboot and move pxelinux.cfg directory inside
/var/lib/tftpboot and restart dnsmasq.


Gilles

Op 20080821 om 16:39 schreef Tomas M:
> Hello.
>
> Some time ago, everything worked for me.  But now it is very strange,
> it does no more work (using exactly the same setup) Do you have any
> idea how to find out what's wrong?
>
> Here is my situation:
> I am trying to boot over PXE.
> TFTP server is running. DHCP server is running from the same
> machine. Both by dnsmasq.
> TFTP server has the following files and directories:
> /pxelinux.cfg/pxelinux.0
> /pxelinux.cfg/default
> I can use 'tftp localhost' (using tftp-hpa) and type 'get
> /pxelinux.cfg/default' command which downloads the default config
> file. Seems there is problem here on server.
>
> The client initiates DHCP requests. Client receives response from the
> server, assigns a new (unique) IP address to itself and it downloads
> pxelinux.0 file. Then it starts that file. Then pxelinux tries to
> download pxelinux.cfg/123456789ABCD and so on, and also it tries
> pxelinux.cfg/default and then it dies with a message "Unable to locate
> configuration file". But it's there! :)
>
> Syslog on the server shows a strange message from dnsmasq:
> TFTP error 0 TFTP Aborted received from 172.16.73.122
> FTFP failed sending /boot/pxelinux.cfg/pxelinux.0 to 172.16.73.122
>
> Whats up? Anybody have any idea? :)
> I know pxelinux.0 has been sended to the client for sure, because it
> is executed there, and it searches for the default config at the first
> place ...
>
> Hope somebody could suggest how to track down where the problem is.
My suggestion is to check firewall settings and name resolving settings.


I think firewall rules are blocking the TFTP transmission. ( TFTP has a
somewhat odd port usage ( opens are port number 69 and continues on a
hard to predict port )).  The test `tftp localhost` is done at the "good
side" of the FW, it is done _behind_ the firewall. 

Another reason for breakage on a "working setup" is that previous
(working) clients are known to the server. New clients are unknown to
the server. Unknown in the meaning of no name to an IP-address.
Solution is to make name resolving of the new clients possible.
( add then to domain name server setup  (or to /etc/hosts ))
> Thank you
Euh, it only are some suggestions.
Mostly to say: "I'm curious what did break an working setup"
> Tomas M
Cheers
Geert Stappers

Tomas M wrote:
> 
> Syslog on the server shows a strange message from dnsmasq:
> TFTP error 0 TFTP Aborted received from 172.16.73.122
> FTFP failed sending /boot/pxelinux.cfg/pxelinux.0 to 172.16.73.122
> 
This is normal behaviour for most Intel-derived PXE stacks.
> Whats up? Anybody have any idea? :)
> I know pxelinux.0 has been sended to the client for sure, because it is
executed there, and it searches for the default config at the first place ...
> 
> Hope somebody could suggest how to track down where the problem is.
Try getting a packet trace using Wireshark.

	-hpa