✎ soc:2010:peper:journal:week0 - Etherboot/gPXE Wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.

====== Piotr Jaroszyński: Usermode debugging under Linux ======

===== Previous work [ Prior to 24 May 2010 ] =====

==== Unrelated work ====

Before I even started thinking about a proposal for my project I decided to do some pretty much random unrelated work to get a better feeling of gPXE in general. This resulted in a few trivial patches:

  *[[http://git.etherboot.org/?p=gpxe.git;a=commitdiff;h=46d6ec7d77a041d8266d0d9811f57ba92e86599e|[build] Add support for local configuration files]]
  *[[http://git.etherboot.org/?p=gpxe.git;a=commitdiff;h=9feb7b38a5d38fd1dcf431b8ff2139d5d7afd90f|[build] Look for isolinux.bin in more places]]
  *[[http://git.etherboot.org/?p=gpxe.git;a=commitdiff;h=70537cbe3504b90f3b524cec1bb3867b04514c37|[util] Hide an expected error from the 'which' command]]

And then I decided to tackle something involving proper coding - adding base64 support to iSCSI. This resulted in not yet merged changes:

  *[[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/base64|base64 branch]]

==== Related work ====

Even after doing some work on gPXE earlier I didn't know where to start the project I wanted to work on during summer. Josh helped me a lot with that.

Firstly I tried an efi build to see how the different ''ARCH''s and ''PLATFORM''s work. Unfortunately that resulted in some compilation as well as linking problems. I have managed to fix them:

  *[[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/efi64|efi64 branch]]

With that behind me, I added a new ''linux'' ''PLATFORM'', which involved the following:

  * Remapping stdlib's symbols to avoid collisions with gPXE (see [[..:notes/implementation_details#kernel_interface|Kernel Interface]])
  * Adding linker scripts (for ''i386'' and ''x86_64'') just to handle tables (see ''include/gpxe/tables.h'')
  * Adding an empty ''linuxprefix''

Having that base I implemented the very first linux subsytem: ''CONSOLE_LINUX'', which got usermode gPXE as far as:

  $ bin-x86_64-linux/rtl8139.linux
  
  gPXE initialising devices...
   
  Segmentation fault

That work can be seen on [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/linux|linux branch]] (up to the "[linux] Add console" commit).

After that I decided to add a network driver based on tun/tap and all the subsystems necessary for that:
  * ''TIMER_LINUX''
  * ''UACCESS_LINUX''
  * ''UMALLOC_LINUX''
  * ''NAP_LINUX''

It's in an ugly state, but seems to be working:
  $ ./bin-x86_64-linux/virtual.linux
  gPXE initialising devices...
  
  
  
  gPXE 1.0.0+ -- Open Source Boot Firmware -- http://etherboot.org
  Features: HTTP DNS TFTP
  net0: 52:54:00:12:34:56 on  (open)
    [Link:up, TX:0 TXE:0 RX:0 RXE:0]
  DHCP (net0 52:54:00:12:34:56).... ok
  net0: 192.168.1.201/255.255.255.0 gw 192.168.1.1
  Booting from filename "http://boot.kernel.org/bko/pxelinux.0"
  http://boot.kernel.org/bko/pxelinux.0. Not an executable image (0x2e008001)
  Could not boot from filename "http://boot.kernel.org/bko/pxelinux.0": Not an executable image (0x2e008001)
  No more network devices

The changes can be seen currently on the [[http://git.etherboot.org/?p=people/peper/gpxe.git;a=shortlog;h=refs/heads/linux-ugly|linux-ugly branch]] with just one monolithic commit.

During that work usermode gPXE helped me track down and fix two bugs:
  *[[http://git.etherboot.org/?p=gpxe.git;a=commitdiff;h=019327eca42c76e1135154c4d002859d84edc5c0|[uri] Fix NULL dereference in parse_uri()]]
  *[[http://git.etherboot.org/?p=gpxe.git;a=commitdiff;h=79d9b2861545837ef92f259326c78ef12f8d2e3f|[uri] Special case NULL in churi()]]

Trace: • week3 • week5 • week0 • week2 • pictures • week5 • week4 • week4 • week4 • week3 • week2 • week5 • week7 • week1 • week7 • week3 • week8 • week8 • week6 • week2