Dothan/Centrino cpufreq

Dothan/Centrino cpufreq

[athul acharya]

Hey,

I originally sent this to linux-kernel and was advised
that I'd be better off sending it here. 

I noticed that 2.6.7-rc1 was supposed to have Dothan
support for cpufreq, but since neither it nor -rc2
worked  on my machine, I decided to take a look at the
code and see if I couldn't figure something out.  I
saw this:

static const struct cpu_id cpu_id_dothan_a1 = {
        .x86_vendor = X86_VENDOR_INTEL,
        .x86 = 6,
        .x86_model = 13,
        .x86_mask = 1,
};

Attempting to match this against /proc/cpuinfo, I
changed it to

static const struct cpu_id cpu_id_dothan_a1 = {
        .x86_vendor = X86_VENDOR_INTEL,
        .x86 = 6,
        .x86_model = 13,
        .x86_mask = 6,
};

This worked, creating
/sys/devices/system/cpu/cpu0/cpufreq/ and the whole
shebang.  So, the above change needs to be integrated
in order for (some?) Dothans to work.

Thanks,

Athul Acharya


	
		
__________________________________
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/ 

Re: Dothan/Centrino cpufreq

[Jeremy Fitzhardinge]

On Wed, 2004-06-09 at 18:30 -0700, athul acharya wrote:
> Hey,
> 
> I originally sent this to linux-kernel and was advised
> that I'd be better off sending it here. 
> 
> I noticed that 2.6.7-rc1 was supposed to have Dothan
> support for cpufreq, but since neither it nor -rc2
> worked  on my machine, I decided to take a look at the
> code and see if I couldn't figure something out.  I
> saw this:
> 
> static const struct cpu_id cpu_id_dothan_a1 = {
>         .x86_vendor = X86_VENDOR_INTEL,
>         .x86 = 6,
>         .x86_model = 13,
>         .x86_mask = 1,
> };
> 
> Attempting to match this against /proc/cpuinfo, I
> changed it to
> 
> static const struct cpu_id cpu_id_dothan_a1 = {
>         .x86_vendor = X86_VENDOR_INTEL,
>         .x86 = 6,
>         .x86_model = 13,
>         .x86_mask = 6,
> };
> 
> This worked, creating
> /sys/devices/system/cpu/cpu0/cpufreq/ and the whole
> shebang.  So, the above change needs to be integrated
> in order for (some?) Dothans to work.

There seems to be a few things which need fixing in the driver, which
will require a re-org of the various tables:

1. Intel seems to have changed the branding string on Banias (but
without apparently changing the stepping)
2. There seems to be two steppings of Dothan

I'm a bit hesitant to just assume that both Dothan steppings are
identical (but since we're currently just extracting the operating
points from the BIOS, it shouldn't be a big risk).

	J