[okl4-developer] roottask error. any ideas?

[Boy Dfx]

Hello.

I'm getting back to you for some help or thoughts regarding the following situation.

I have installed OKL4 under a number of different Linux OS-es, by always using arm-linux-3.4.4 crosscompiler from the nicta site, and a proof version of skyeye (a version that worked on a verified image.sim). 

Sometimes, the build just works from the beginning (I can enter busybox and play around with my gumstix simulated system) and sometimes it crashes with a roottask page fault (like in the following lines). The roottask error changes in different instalations. It can be that one PC running a linux distro works and another using  matching hardware does not. 

I might have left out some packages from time to time at instalation time, but nothing really essential (like gcc, python or anything corelated through,etc). Example: Centos 5 - working, Centos 5 - not working, Fedora core 4 -working, Fedora Core 3  -working, openSUSE - working.

I just want to understand what is the source for this particular problem. By doing some cross-checking I can tell you it is not a skyeye problem. If I have 2 systems with the same hardware, one with l4 working and one without, if I load the image.sim from the working build into both skyeye instances then the simulation works. 

Somehow the image.sim gets corrupted either by some x86 code being loaded into it or by an executable taking precedence over another, or something not created during the build.

The PATH variable modification is the same, folder instalations the same  (I install arm-linux-gcc in the same folder and tar -xvzf the same way).  

Can you think of a package or some enviromental variable that when missing can cause this kind of behaviour?

Initially I thought there was a bug in File Roller that was fixable by always using tar. That is because I noticed that in some systems (like FC3, FC4) File Roller refused to create tools/cust and other empty folders in okl4 directory and arm-linux-3.4.4 directory. I thought I got to the bottom of this (and also sent you an email to warn other people about this potential problem and how to stay away), because simply using tar solved my problem. Until when the tar solution did not work on a CentOS5 instalation. It is not python or gcc version related because on a similar machine it does.

What I use for my builds in general:
python 2.4.4 (except on CentOS 5 where I use python 2.4.3)
gcc
arm-linux-3.4.4 (ffrom NICTA)
skyeye-1.2.4 (built with make NO_DBCT=1 NO_BFD=1)
skyeye-1.2.3 (built with make NO_DBCT=1 NO_BFD=1 STATIC=1 NO_LCD=1 PPC_DISABLED=1)


Listing of the error.

skyeye -c /proj/okl4_release_1.5.2/tools/sim_config/gumstix.skyeye -e /proj/okl4_release_1.5.2/build/images/image.sim 
big_endian is false.
arch: arm
cpu info: xscale, pxa25x, 69052100, fffffff0, 2 
mach info: name pxa_lubbock, mach_init addr 0x8060fa0
uart_mod:0, desc_in:, desc_out:, converter:
SKYEYE: use xscale mmu ops
start addr is set to 0xa0000000 by exec file.

OKL4 - (provider: Open Kernel Labs) built on Feb 19 2008 17:52:17 using gcc version 3.4.4.

 *****************************************
 *  Iguana system, Copyright 1996-2006   *
 *  ERTOS - National ICT Australia       *
 *****************************************

vbase is 0xa4000000, vend is 0xcfffffff
addr is 0xa4000000, vbase is 0xa4000000
roottask read pagefault at 0000001c, ip=800038bc - deny
--- KD# roottask pagefault  ---
> 



       
---------------------------------
Never miss a thing.   Make Yahoo your homepage.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.okl4.org/pipermail/developer/attachments/20080219/33af9dbc/attachment.htm