2.6.x kernels and AMD K6 processor?

[hseaver]

I've been running debian on my i-opener for quite awhile. I'm upgraded to the latest version, etch, and I've got a K6-lll+ cpu running with badflashes kit. I've also been using the 2.6.x kernels ever since they came out. I have no problem with compiling kernels, do it all the time on many machines, but for some reason I can't figure out, when I compile a kernel with the K6 chosen as the cpu, and the cpufreq stuff enabled, the kernel seems to compile fine, builds the initrd fine, and then always crashes halfway thru the boot. I can run kernels compiled for 486 and 586 cpus, but not the K6.
Anybody have a clue what the problem is? I don't have any problem with compiling these very same kernels (optimized for their own cpu) on other machines, just the IOP. I'm wondering if it's the cpufreq that does it, like does it try during boot to crank the cpu way up or something?

[Dark_Stang]

I had a similar problem with compiling kernels on a K7. Everything compiled fine. It just wouldn't boot. I've been using pre-packaged stuff since.

[antidrugue]

I've been running debian on my i-opener for quite awhile. I'm upgraded to the latest version, etch, and I've got a K6-lll+ cpu running with badflashes kit. I've also been using the 2.6.x kernels ever since they came out. I have no problem with compiling kernels, do it all the time on many machines, but for some reason I can't figure out, when I compile a kernel with the K6 chosen as the cpu, and the cpufreq stuff enabled, the kernel seems to compile fine, builds the initrd fine, and then always crashes halfway thru the boot. I can run kernels compiled for 486 and 586 cpus, but not the K6.

You seem to know your stuff, and it appears you are doing it right.

But which exact version did you try ? Was it from vanilla sources, or the Debian ones ?

Before calling it quit, perhaps you can try the latest 2.6.20 vanilla ?

Anybody have a clue what the problem is? I don't have any problem with compiling these very same kernels (optimized for their own cpu) on other machines, just the IOP. I'm wondering if it's the cpufreq that does it, like does it try during boot to crank the cpu way up or something?

Did you build powernow_k6 as a module ? Or statically ? Remember that it is not supported by all K6 CPUs, only the mobile ones.

I've written a small tutorial on how to compile a kernel on Debian. It is mostly basic stuff you probably already know of, but still you can take a look :
http://technowizah.com/2005/12/debia...l-compile.html

Keep us updated on how it goes.

[hseaver]

But which exact version did you try ? Was it from vanilla sources, or the Debian ones ?

Before calling it quit, perhaps you can try the latest 2.6.20 vanilla ?

Did you build powernow_k6 as a module ? Or statically ? Remember that it is not supported by all K6 CPUs, only the mobile ones.
.

I'm using vanilla src from kernel.org. Tried 2.6.17.x and 2.6.19.1 and 2.6.20. I'm currently running 2.19.1 on my other machines, and now running 2.6.18.3-486 on the i-opener, but that was just a straight debian apt-get install, no compiling. But that is also a PITA with dma timer errors when booting.
I compiled Powernow into the kernel, and this is a mobile cpu -- K6-lll+
but just now I reconfiged that into a module, and as much of the cpufreq stuff as I could as modules and am recompiling.

[hseaver]

Well, that seemed to do it -- made powernow a module and now that kernel boots all the way.
Thanks for the help, I should have thought of that.

[antidrugue]

Well, that seemed to do it -- made powernow a module and now that kernel boots all the way.

Nice to hear that. I understand the philosophy being having an entirely static kernel, but on the other hand modules give a certain flexibility (you can load them or not, with whichever parameters you want, etc.). Not to mention that certain elements of the kernel just work better when built as a module : alsa, drm (graphic), etc.

[hseaver]

Actually I've got most stuff as modules, but that was a new thing and I hit y instead of m for some reason.