doc/README.pxe

/*
 * Copyright 2010-2011 Calxeda, Inc.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the Free
 * Software Foundation; either version 2 of the License, or (at your option)
 * any later version.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program.  If not, see <http://www.gnu.org/licenses/>.
 */

The 'pxe' commands provide a near subset of the functionality provided by
the PXELINUX boot loader. This allows U-boot based systems to be controlled
remotely using the same PXE based techniques that many non U-boot based servers
use.

Commands
========

pxe get
-------
     syntax: pxe get

     follows PXELINUX's rules for retrieving configuration files from a tftp
     server, and supports a subset of PXELINUX's config file syntax.

     Environment
     -----------
     'pxe get' requires two environment variables to be set:

     pxefile_addr_r - should be set to a location in RAM large enough to hold
     pxe files while they're being processed. Up to 16 config files may be
     held in memory at once. The exact number and size of the files varies with
     how the system is being used. A typical config file is a few hundred bytes
     long.

     bootfile,serverip - these two are typically set in the DHCP response
     handler, and correspond to fields in the DHCP response.

     'pxe get' optionally supports these two environment variables being set:

     ethaddr - this is the standard MAC address for the ethernet adapter in use.
     'pxe get' uses it to look for a configuration file specific to a system's
     MAC address.

     pxeuuid - this is a UUID in standard form using lower case hexadecimal
     digits, for example, 550e8400-e29b-41d4-a716-446655440000. 'pxe get' uses
     it to look for a configuration file based on the system's UUID.

     File Paths
     ----------
     'pxe get' repeatedly tries to download config files until it either
     successfully downloads one or runs out of paths to try. The order and
     contents of paths it tries mirrors exactly that of PXELINUX - you can
     read in more detail about it at:

     http://syslinux.zytor.com/wiki/index.php/Doc/pxelinux

pxe boot
--------
     syntax: pxe boot [pxefile_addr_r]

     Interprets a pxe file stored in memory.

     pxefile_addr_r is an optional argument giving the location of the pxe file.
     The file must be terminated with a NUL byte.

     Environment
     -----------
     There are some environment variables that may need to be set, depending
     on conditions.

     pxefile_addr_r - if the optional argument pxefile_addr_r is not supplied,
     an environment variable named pxefile_addr_r must be supplied. This is
     typically the same value as is used for the 'pxe get' command.

     bootfile - typically set in the DHCP response handler based on the
     same field in the DHCP respone, this path is used to generate the base
     directory that all other paths to files retrieved by 'pxe boot' will use.
     If no bootfile is specified, paths used in pxe files will be used as is.

     serverip - typically set in the DHCP response handler, this is the IP
     address of the tftp server from which other files will be retrieved.

     kernel_addr_r, initrd_addr_r - locations in RAM at which 'pxe boot' will
     store the kernel and initrd it retrieves from tftp. These locations will
     be passed to the bootm command to boot the kernel. These environment
     variables are required to be set.

     fdt_addr - the location of a fdt blob. If this is set, it will be passed
     to bootm when booting a kernel.

pxe file format
===============
The pxe file format is nearly a subset of the PXELINUX file format; see
http://syslinux.zytor.com/wiki/index.php/PXELINUX. It's composed of one line
commands - global commands, and commands specific to labels. Lines begining
with # are treated as comments. White space between and at the beginning of
lines is ignored.

The size of pxe files and the number of labels is only limited by the amount
of RAM available to U-boot. Memory for labels is dynamically allocated as
they're parsed, and memory for pxe files is statically allocated, and its
location is given by the pxefile_addr_r environment variable. The pxe code is
not aware of the size of the pxefile memory and will outgrow it if pxe files
are too large.

Supported global commands
-------------------------
Unrecognized commands are ignored.

default <label>     - the label named here is treated as the default and is
                      the first label 'pxe boot' attempts to boot.

menu title <string> - sets a title for the menu of labels being displayed.

menu include <path> - use tftp to retrieve the pxe file at <path>, which
                      is then immediately parsed as if the start of its
                      contents were the next line in the current file. nesting
                      of include up to 16 files deep is supported.

prompt <flag>       - if 1, always prompt the user to enter a label to boot
                      from. if 0, only prompt the user if timeout expires.

timeout <num>	    - wait for user input for <num>/10 seconds before
                      auto-booting a node.

label <name>        - begin a label definition. labels continue until
                      a command not recognized as a label command is seen,
                      or EOF is reached.

Supported label commands
------------------------
labels end when a command not recognized as a label command is reached, or EOF.

menu default        - set this label as the default label to boot; this is
                      the same behavior as the global default command but
                      specified in a different way

kernel <path>       - if this label is chosen, use tftp to retrieve the kernel
                      at <path>. it will be stored at the address indicated in
                      the kernel_addr_r environment variable, and that address
                      will be passed to bootm to boot this kernel.

append <string>     - use <string> as the kernel command line when booting this
                      label.

initrd <path>       - if this label is chosen, use tftp to retrieve the initrd
                      at <path>. it will be stored at the address indicated in
                      the initrd_addr_r environment variable, and that address
                      will be passed to bootm.

localboot <flag>    - Run the command defined by "localcmd" in the environment.
                      <flag> is ignored and is only here to match the syntax of
                      PXELINUX config files.

Example
-------
Here's a couple of example files to show how this works.

------------/tftpboot/pxelinux.cfg/menus/linux.list----------
menu title Linux selections

# This is the default label
label install
	menu label Default Install Image
	kernel kernels/install.bin
	append console=ttyAMA0,38400 debug earlyprintk
	initrd initrds/uzInitrdDebInstall

# Just another label
label linux-2.6.38
	kernel kernels/linux-2.6.38.bin
	append root=/dev/sdb1

# The locally installed kernel
label local
	menu label Locally installed kernel
	append root=/dev/sdb1
	localboot 1
-------------------------------------------------------------

------------/tftpboot/pxelinux.cfg/default-------------------
menu include pxelinux.cfg/menus/base.menu
timeout 500

default linux-2.6.38
-------------------------------------------------------------

When a pxe client retrieves and boots the default pxe file,
'pxe boot' will wait for user input for 5 seconds before booting
the linux-2.6.38 label, which will cause /tftpboot/kernels/linux-2.6.38.bin
to be downloaded, and boot with the command line "root=/dev/sdb1"

Differences with PXELINUX
=========================
The biggest difference between U-boot's pxe and PXELINUX is that since
U-boot's pxe support is written entirely in C, it can run on any platform
with network support in U-boot. Here are some other differences between
PXELINUX and U-boot's pxe support.

- U-boot's pxe does not support the PXELINUX DHCP option codes specified
  in RFC 5071, but could be extended to do so.

- when U-boot's pxe fails to boot, it will return control to U-boot,
  allowing another command to run, other U-boot command, instead of resetting
  the machine like PXELINUX.

- U-boot's pxe doesn't rely on or provide an UNDI/PXE stack in memory, it
  only uses U-boot.

- U-boot's pxe doesn't provide the full menu implementation that PXELINUX
  does, only a simple text based menu using the commands described in
  this README.  With PXELINUX, it's possible to have a graphical boot
  menu, submenus, passwords, etc. U-boot's pxe could be extended to support
  a more robust menuing system like that of PXELINUX's.

- U-boot's pxe expects U-boot uimg's as kernels.  Anything that would work
  with the 'bootm' command in U-boot could work with the 'pxe boot' command.

- U-boot's pxe doesn't recognize initrd options in the append command - you
  must specify initrd files using the initrd command.

- U-boot's pxe only recognizes a single file on the initrd command line.  It
  could be extended to support multiple.

- in U-boot's pxe, the localboot command doesn't necessarily cause a local
  disk boot - it will do whatever is defined in the 'localcmd' env
  variable. And since it doesn't support a full UNDI/PXE stack, the
  type field is ignored.

- the interactive prompt in U-boot's pxe only allows you to choose a label
  from the menu.  If you want to boot something not listed, you can ctrl+c
  out of 'pxe boot' and use existing U-boot commands to accomplish it.