[lug] Etch AMD kernels not working

[Gary Hodges]

Gary Hodges wrote:
> Gary Hodges wrote:
>> Gary Hodges wrote:
>>> Gary.Hodges at noaa.gov wrote:
>>>>
>>>> ----- Original Message -----
>>>> From: "David L. Anselmi" <anselmi at anselmi.us>
>>>> Date: Saturday, April 5, 2008 9:19 pm
>>>> Subject: Re: [lug] Etch AMD kernels not working
>>>>
>>>>> Gary.Hodges at noaa.gov wrote:
>>>>> [...]
>>>>>> With the SMP k7 machine there are no error messages.  It just 
>>>>> stops.  A
>>>>>> common place to stop was just after a FATAL message about a 
>>>>> thermal(?)> and fan module.  I get these same messages with every 
>>>>> Etch machine I
>>>>>> have.  Otherwise it would stop in seemingly random places.  After 
>>>>> eth0> one time, after drive information another.  Sometimes after a 
>>>>> "File> system is clean" message.  Just different places.
>>>>>
>>>>> I think the file system is checked after the kernel is booted.
>>>>>
>>>>> So maybe there's a kernel module loading that doesn't work anymore. 
>>>>> You can look in /etc/modules.
>>>>
>>>> Nice suggestion.  I just logged into that machine and there are a 
>>>> couple
>>>> there that I believe were associated with hangs on boot.  I'll try it
>>>> again first thing Monday and report back.
>>>
>>> I thought I was getting somewhere when I got to the KDE login screen, 
>>> but the machine locked up when I tried logging in.  During subsequent 
>>> reboots the machine stopped in several different places (some listed 
>>> below):
>>>
>>> FATAL: Error inserting thermal 
>>> (lib/modules/2.6.18-6-k7/kernel/drivers/acpi/thermal.ko): No such device
>>>
>>> Waiting for /dev to be fully populated
>>>
>>> SCSI device sdc: 143550456 512-byte hdwr sectors (73498 MB)
>>>
>>> So I'm back to running the 486 kernel.
>>
>> For those interested...  I did a fresh install on a separate drive. 
>> Same issue.  About half the time I'd get to login prompt, and about 
>> half the time it stops during the boot process.  For kicks I tried a 
>> different SCSI card.  Same results.  When I get a chance I'll compile 
>> a kernel and see if that works.  For the time being I'm back to 
>> running the 486 kernel with one CPU.
> 
> I just played around with booting a custom compiled kernel.  The good 
> news is the machine boots up fine.  The bad news is I have no video.  No 
> more time to play today, but there seems to be something incompatible 
> with the pre-compiled k7 kernels and my k7 hardware.  Is that a bug?

A lot of self replies, but I have some new information.  It seems the 
issue is SMP, not a particular kernel flavor.  I compiled several SMP 
486 and k7 kernels over the past week or so, and all locked the machine 
either during boot or within 5 minutes of logging in.  I just grabbed 
the config file of the single CPU 486 kernel that has been working fine 
since the Etch upgrade a few weeks ago.  This is the default install 
kernel.  The only change I made was adding SMP support and setting 
Enhanced Real Time Clock Support as instructed in the SMP info section. 
  That kernel locked the machine during boot.

To review:
AMD SMP machine ran 24/7 for the past 4-5 years without any trouble.
Tyan S2466 MB
2x Athlon 2800+ CPUs
Will not run as SMP since recent Etch upgrade.

Gary