John McSwain wrote:
> This machine has a AMD Sempron 1.6G on a ECS755-A2 with 1024 GB RAM, an 
> 80 GB SATA Drive (main operating system and swap space) and a 160 GB ide 
> drive (home directories and swap sapce).  Has a floopy and CD.  Uses MB 
> network connection and has a Enermax EG495P-UE PS which should be 
> sufficient.  It has an old simple vidoe card of some type as I rarely do 
> anything with it from the monitor.  This machine is a server for my 
> network that sits in the DMZ, providing mail, web server, SSH, NTP, etc.
>
>   
>> as an expensive fix as MBs were *not* cheap.
>>     
>
> I'm thinking it could easily be an MB problem.  I've used AMD processors 
> in the past and I had another AMD machine that I always had problem 
> building LFS on because in each compile it would stop with an error but 
> eventually finish it.  My own impression is that the problem with AMD is 
> not the CPU but the chipsets that others build for them.  I've never had 
> any problem on an intel chipset MB.
>
> <snip>
>   
I had this exact problem about 2 years ago building my LFS based MythTV 
box.  I chased odd problems with Myth and compile errors in the Kernel, 
GCC and Glib.  I had the hardware to test everything on a different 
system - swap hardware to a different system and see where the problem 
followed.

It ended up being the AMD64 3000 that was bad.  Fortunately it was still 
under warranty and AMD exchanged it with out question when I referred to 
the SIG11 problem.  Follow the trouble shooting procedure at 
http://www.bitwizard.nl/sig11/

NOTE: I never saw "fatal signal 11", the errors I saw were random.  But 
the sites troubleshooting procedure and explanations were dead on.

I was able to consistently reproduce the error by writing a script to 
repeatedly compile a full blown kernel.  If it compiles correctly the 
first time then it should compile correctly 30 times. And I would 
recommend letting it run overnight once you think you have the problem 
fixed, mine sometimes wouldn't fail until the 7th pass then not again 
until th 15th". 

Memtest86+ wasn't very helpful, while it did show errors after an 
extended run it didn't mean the memory was bad.  AMD places the memory 
controller on the CPU, not the chipset like Intel.  (That's a big part 
of why the BTX form factor didn't catch on)

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page

Reply via email to