Re: [PATCH 1/4] x86: base support for the TS-5500 platform

[Vivien Didelot <vivien.didelot@savoirfairelinux.com>]

On Wed, 24 Aug 2011 14:03:11 -0700,
"H. Peter Anvin" <hpa@zytor.com> wrote:

> On 08/24/2011 01:23 PM, Vivien Didelot wrote:
> > +
> > +/**
> > + * ts5500_pre_detect() - check for TS-5500 specific features
> > + *
> > + * It is not safe to read ID register if we are not sure that it's
> > a TS SBC.
> > + * Since TS SBCs don't have DMI support, it is safer to check for
> > a TS-5500
> > + * specific feature such as the processor.
> > + */
> > +static int ts5500_pre_detect(void)
> > +{
> > +	/* Check for AMD ElanSC520 Microcontroller */
> > +	if (!(cpu_info.x86_vendor == X86_VENDOR_AMD &&
> > +	      cpu_info.x86 == AMD_ELAN_FAMILY &&
> > +	      cpu_info.x86_model == AMD_ELAN_SC520))
> > +		return -ENODEV;
> > +
> > +	return 0;
> > +}
> > +
> 
> Doesn't this mean you will read this unsafe register on any Elan SC520
> platform?  Is that okay?
> 
> 	-hpa
> 
Indeed this will be run on any Elan SC520 based platforms if the user
has asked for TS-5500 support in the menuconfig. I could add checks for
other hardware such as the Ethernet controller, but it would require
including pci headers and delaying the initcall (using late_initcall
instead of postcore_initcall).
That's why I think it is not really worth adding more tests. They
could be added later in this function if there is really a need.

Regards,
Vivien.