Powernow-k8: Update to support the latest Turion processors

The latest series of Turion X2 processors have a new XFAM model. Add support for them to powernow-k8.h.

@@ -47,7 +47,7 @@ struct powernow_k8_data {
 #define CPUID_XFAM			0x0ff00000	/* extended family */
 #define CPUID_XFAM_K8			0
 #define CPUID_XMOD			0x000f0000	/* extended model */
-#define CPUID_XMOD_REV_MASK		0x00080000
+#define CPUID_XMOD_REV_MASK		0x000c0000
 #define CPUID_XFAM_10H			0x00100000	/* family 0x10 */
 #define CPUID_USE_XFAM_XMOD		0x00000f00
 #define CPUID_GET_MAX_CAPABILITIES	0x80000000

Hi,

The problem occurs when CPU is running on low freq. (800 MHz). The hangs occur (not neccesarily immediately) with Machine Check Exception, however the message is not always displayed (or I am too slow to see it), there were two messages I did not manage to write down (fully) both of them, one of them was CPU. 0: Machine Check Exception: 0....4 Bank 4: b20...70f0f
In place of dots were zeroes I did not manage to count before reboot. Sometimes it does not reboot but just hangs hard with no message but blinking capslock (kernel panic).

System attempted to instal is linux Gentoo, kernel 2.6.23, it does not matter whether the frequency is switched after startup or low from the beginning (switched by bios on startup with battery power. Chipset is Nvidia MCP51.

Also, the system does exactlu the same booted from x86 as well as amd64 livecd gentoo, SystemRescueCD (0.4.3 and 1.0.0) - mostly during some compiling or un-tarring, but sometimes it just does that "out of the box", during boot or right after that).

On 2.4GHz, system is running without any problems (govenor - performance or userspace). It doess not seem to be a memory problem - I have tested it on 2 sets of modules - (original ones 2x1GB 667MHz CL5 as well as elpida 2x2GB 800MHz CL5). Moreover, frequency scaling (or keeping low frequency, unexpectedly) does not affect system stability in bundled vista - which also suggests no memory problem. memtest86 also passes.

Best regards,
Micha³ Pytasz

Additional information:

The full MCE message (on which system hangs) is:

CPU 0: Machine Check Exception: 0000000000000004
Bank 4: b200000000070f0f

Kernel panic - not syncing: CPU context corrupt

However for the first time it occured with 2.4GHz... This time kernel was 2.6.24.

Best regards,
Micha³ Pytasz

I have another problem on mine (Ferrari F5000) and I wonder whether you also have seen something similar. I have a similar Turion (T-60?). Since the clockevents patches already some kernel iterations ago, I need to pass noapictimer as boot parameter.
Otherwise the kernel will freeze at early boot time, but only every second time.
Not every second, but statistically every second time. It could be that I can boot 3 times, then it freezes 2 times ...
It looks like whether it depends that it's random which CPU/apic core is used (setup up) first and if the broken one is taken the APIC timer does not tick and the machine hangs.

I wonder whether you also see this (then I could submit a patch to also use noapictimer for Turions) or whether I really have broken HW here and have to live with the workaround boot parameter.
I wasn't able to see/google my backtrace anywhere:
A division through zero:
setup_boot_APIC_clock
kernel_init
child_rip
vgacon_cursor
kernel_init
child_rip

RIP: clockevent_delta2ns+0x11/0x3b

therefore I could imagine I should go for the boot param?

For your memory problem... Just a guess, but it could be that your machine has voltage regulating problems. I'd rather expect that if run on full power, but I also could imagine this happens on low cpufreq values (if HW is/got broken). I'd run a memtest with lower cpufreq, maybe something like this:
http://www.krugle.org/examples/p-Gk7AKdq1GmR9tNcy/memeater.c
Does this one help to let your machine freeze quicker (still only if cpufreq is lowered -> I'd say it's a HW problem then)?

Thomas

_______________________________________________
Cpufreq mailing list
[email protected]
http://lists.linux.org.uk/mailman/listinfo/cpufreq