Most recent kernel where this bug did *NOT* occur: unknown Distribution: vanilla kernel copiled with FC6 gcc-4.1.1-51.fc6 Hardware Environment: AMD Athlon64 3800+ AM2 on Gigabyte GA-M61PM-S2, latest BIOS Software Environment: Fedora Core 6 Problem Description: Powernow-k8 driver does not load - it shows a message powernow-k8: Processor cpuid 70ff1 not supported on my new AMD Athlon64 3800+ on AM2 connector (Family: 15 Model: 15 Stepping: 1), which does support Cool'n'Quiet. Because of this I cannot use cpufreq in my system. I think it triggers a check (eax & CPUID_XMOD) > CPUID_XMOD_REV_G) from row 524 in arch/i386/kernel/cpu/cpufreq/powernow-k8.c. Macros CPUID_XMOD and CPUID_XMOD_REV_G are defined in ./arch/i386/kernel/cpu/cpufreq/powernow-k8.h: #define CPUID_XMOD 0x000f0000 /* extended model */ #define CPUID_XMOD_REV_G 0x00060000 Steps to reproduce: 1. compile kernel with powernow-k8 support 2. boot on Athlon64 3800+ AM2
Created attachment 11381 [details] x86info output
Known bug; correcting patch was submitted on February 23, 2007. Correcting patch attached.
Created attachment 11385 [details] Change RevF numbers in powernow-k8
not sure what happened to your earlier patch, but I don't seem to be able to find it. Someone else mailed me a patch yesterday which fixes the problem for him, so I merged it and pushed it on to Linus. That one replaced #define CPUID_XMOD_REV_G 0x00060000 with #define CPUID_XMOD_REV_H 0x00070000 Whilst the patch you attached here bumps it to 80000. Does RevH span two revisions?, or were you thinking ahead for RevI :-)
now fixed in Linus tree, and pushed to stable@