Re: [PATCH 06/10] HWMON: Convert coretemp to x86 cpuid autoprobing

[Jean Delvare <khali@linux-fr.org>]

Hi Andi,

Thanks a lot for working on this. I wanted to give it a try long ago but
could never find the time.

On Wed, 7 Dec 2011 18:40:01 -0800, Guenter Roeck wrote:
> Andi,
> 
> On Wed, Dec 07, 2011 at 07:41:19PM -0500, Andi Kleen wrote:
> > From: Andi Kleen <ak@linux.intel.com>
> > 
> > Use the new x86 cpuid autoprobe interface for the Intel coretemp
> > driver.
> > 
> > Cc: fenghua.yu@intel.com
> > Cc: khali@linux-fr.org
> > Cc: guenter.roeck@ericsson.com
> > Signed-off-by: Andi Kleen <ak@linux.intel.com>
> > ---
> >  drivers/hwmon/coretemp.c |   17 ++++++++++++++---
> >  1 files changed, 14 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/hwmon/coretemp.c b/drivers/hwmon/coretemp.c
> > index 104b376..5de1579 100644
> > --- a/drivers/hwmon/coretemp.c
> > +++ b/drivers/hwmon/coretemp.c
> > @@ -39,6 +39,7 @@
> >  #include <linux/moduleparam.h>
> >  #include <asm/msr.h>
> >  #include <asm/processor.h>
> > +#include <asm/cpu_device_id.h>
> >  
> >  #define DRVNAME	"coretemp"
> >  
> > @@ -756,13 +757,23 @@ static struct notifier_block coretemp_cpu_notifier __refdata = {
> >  	.notifier_call = coretemp_cpu_callback,
> >  };
> >  
> > +static struct x86_cpu_id coretemp_ids[] = {

Can't this be made const?

> > +	{ X86_VENDOR_INTEL, X86_FAMILY_ANY, X86_MODEL_ANY, X86_FEATURE_DTS },
> > +	{}
> > +};
> > +MODULE_DEVICE_TABLE(x86cpu, coretemp_ids);
> > +
> >  static int __init coretemp_init(void)
> >  {
> >  	int i, err = -ENODEV;
> >  
> > -	/* quick check if we run Intel */
> > -	if (cpu_data(0).x86_vendor != X86_VENDOR_INTEL)
> > -		goto exit;
> > +	/*
> > +	 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
> > +	 * sensors. We check this bit only, all the early CPUs
> > +	 * without thermal sensors will be filtered out.
> > +	 */
> > +	if (!x86_match_cpu(coretemp_ids))
> > +		return -ENODEV;
>
> X86_FEATURE_DTS is checked elsewhere in the driver, in function get_core_online().
> Can that check be removed ?

Is it not possible in theory to mix a CPU with DTS and one without DTS
on the same system? There may be a reason why the DTS flag was checked
in get_core_online() (i.e. on a per-core basis) rather than in
coretemp_init(). I seem to recall that someone explicitly asked for it.

To avoid any risk of regression, I wouldn't touch this part of the code.

-- 
Jean Delvare