PXE-Boot HOWTO-Pitfalls, Hints and Troubleshooting

Note: It may be quite helpfull consulting the respective log files.
I.e. /var/log/messages - present on most distribution around - is a main target.
In case where it is not yet, just insert this line in /etc/syslog.conf:

*.*                                     /var/log/messages 
        

(note the whitespace between "*.*" and "/var/log/messages" as these _have_ to be tabs!) and issue a SIGHUP signal to syslogd:

[root@server:~ ]$ kill -SIGHUP `pidof syslogd`
        

After this treatment many debug and other information will appear in /var/log/messages. So we should comment out and re-SIGHUP syslogd again if do not need this source of information any more or this file will grow fast and the mass of information will hide important single line messages.

1. Again: read carefully and (at least try to) understand the log files (c.f. )
2. It is not seldomly the case your NIC's PXE software implementation is quite buggy. So try to get a new PXE software (firmware) and update your NIC if necessary (c.f. )
3. Another informative source are man/info- pages : c.f.

   [radtke@server:~ ]$ man mknbi
   [radtke@server:~ ]$ info make
           
   

4. fixme: * Disable path MTU discovery on the boot servers. Some versions of * these stacks are known to crash if the boot server supports path * MTU discovery on UDP. For a Linux server, path MTU discovery can * be disabled by setting the sysctl entry net.ipv4.ip_no_pmtu_disc * to 1: echo 1 > /proc/sys/net/ipv4/ip_no_pmtu_disc
5. Incomplete kernel download(PXE) : see ( )
6. Inetd won't start(services): The protocol names have to be resolved to numbers. The lookup file is /etc/services and the entries required for bootpd and tftpd are:

   ### Bootp
   bootps          67/tcp                          # BOOTP server
   bootps          67/udp
   bootpc          68/tcp                          # BOOTP client
   bootpc          68/udp
   
   ### TFTP
   tftp            69/tcp
   tftp            69/udp
                   
   

7. Bootptest is a tool that can be used to check whether your formerly started bootpd is working as intended. For more information please consult the .
8. PMTU: As some (older) PXE implementations have limits(often 32k or less packets) regarding the packet count, it may be necessary to transmit as many bytes as possible in one single packet(i.e. if your kernel, initrd, .. is too large to be transmitted with the default packet count).
   This may be the only way to increase the maximum transmittable file size and not all tftpds allow change of this parameter.

   In a more abstract consideration:

   As you pay more money for a postal message if you send it via many small letters, you try to minimize the financial aspect by increasing the count of messages transmitted by one single letter.

9. tftpd-hpa can be really gabby if you tell him to do so with -vvv ;-)

   tftp  dgram  udp  wait  root  /usr/sbin/tftpd in.tftpd -vvv \
                           -s /tftpboot -r blksize -r timeout
                   
   

This is what has been logged in /var/log/messages on the server during a boot of a single client.
bootpd calls this client "sweepie".
Here is what the bootpd has had to say:

Dec 17 18:38:45 ant bootpd[2750]: recvd pkt from IP addr 0.0.0.0
Dec 17 18:38:45 ant bootpd[2750]: bootptab mtime: Mon Dec 17 18:38:11 2001
Dec 17 18:38:45 ant bootpd[2750]: reading new "/etc/bootptab"
Dec 17 18:38:45 ant bootpd[2750]: read 5 entries (4 hosts) from "/etc/bootptab"
Dec 17 18:38:45 ant bootpd[2750]: request from Ethernet address 00:50:DA:E3:16:2A
Dec 17 18:38:45 ant bootpd[2750]: found 10.10.10.10 (sweepie)
Dec 17 18:38:45 ant bootpd[2750]: bootfile="/pxelinux.0"
Dec 17 18:38:45 ant bootpd[2750]: vendor magic field is 99.130.83.99
Dec 17 18:38:45 ant bootpd[2750]: request message length=548
Dec 17 18:38:45 ant bootpd[2750]: request has DHCP msglen=1260
Dec 17 18:38:45 ant bootpd[2750]: extended reply, length=1260, options=1024
Dec 17 18:38:45 ant bootpd[2750]: Received: DHCPDISCOVER
Dec 17 18:38:45 ant bootpd[2750]: Sent: DHCPOFFER
Dec 17 18:38:45 ant bootpd[2750]: sending reply (with RFC1048 options)
Dec 17 18:38:45 ant bootpd[2750]: setarp 10.10.10.10 - 00:50:DA:E3:16:2A
Dec 17 18:38:47 ant bootpd[2750]: recvd pkt from IP addr 0.0.0.0
Dec 17 18:38:47 ant bootpd[2750]: bootptab mtime: Mon Dec 17 18:38:11 2001
Dec 17 18:38:47 ant bootpd[2750]: request from Ethernet address 00:50:DA:E3:16:2A
Dec 17 18:38:47 ant bootpd[2750]: found 10.10.10.10 (sweepie)
Dec 17 18:38:47 ant bootpd[2750]: bootfile="/pxelinux.0"
Dec 17 18:38:47 ant bootpd[2750]: vendor magic field is 99.130.83.99
Dec 17 18:38:47 ant bootpd[2750]: request message length=548
Dec 17 18:38:47 ant bootpd[2750]: request has DHCP msglen=1260
Dec 17 18:38:47 ant bootpd[2750]: extended reply, length=1260, options=1024
Dec 17 18:38:47 ant bootpd[2750]: Received: DHCPREQUEST
Dec 17 18:38:47 ant bootpd[2750]: Sent: DHCPACK
Dec 17 18:38:47 ant bootpd[2750]: sending reply (with RFC1048 options)
Dec 17 18:38:47 ant bootpd[2750]: setarp 10.10.10.10 - 00:50:DA:E3:16:2A
        

Our in.tftpd says hereupon:

Dec 17 17:38:47 ant in.tftpd[4191]: RRQ from 10.10.10.10 filename /pxelinux.0
Dec 17 17:38:47 ant in.tftpd[4191]: tftp: client does not accept options
Dec 17 17:38:47 ant in.tftpd[4192]: RRQ from 10.10.10.10 filename /pxelinux.0
Dec 17 17:38:47 ant in.tftpd[4193]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31EC6B0
Dec 17 17:38:47 ant in.tftpd[4193]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4194]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31EC6B
Dec 17 17:38:47 ant in.tftpd[4194]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4195]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31EC6
Dec 17 17:38:47 ant in.tftpd[4195]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4196]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31EC
Dec 17 17:38:47 ant in.tftpd[4196]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4197]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31E
Dec 17 17:38:47 ant in.tftpd[4197]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4198]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C31
Dec 17 17:38:47 ant in.tftpd[4198]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4199]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C3
Dec 17 17:38:47 ant in.tftpd[4199]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4200]: RRQ from 10.10.10.10 filename /pxelinux.cfg/C
Dec 17 17:38:47 ant in.tftpd[4200]: sending NAK (1, File not found) to 10.10.10.10
Dec 17 17:38:47 ant in.tftpd[4201]: RRQ from 10.10.10.10 filename /pxelinux.cfg/default
Dec 17 17:38:50 ant in.tftpd[4202]: RRQ from 10.10.10.10 filename /vmlinuz
Dec 17 17:38:51 ant in.tftpd[4203]: RRQ from 10.10.10.10 filename //initrd