Re: [PATCH 1/2] tpm, tpm_tis: use acpi_driver instead of pnp_driver

[Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>]

On Tue, Sep 29, 2015 at 11:26:53AM -0600, Jason Gunthorpe wrote:
> On Tue, Sep 29, 2015 at 08:07:10PM +0300, Jarkko Sakkinen wrote:
> 
> > -static struct pnp_device_id tpm_pnp_tbl[] = {
> > +static struct acpi_device_id tpm_acpi_tbl[] = {
> >  	{"PNP0C31", 0},		/* TPM */
> >  	{"ATM1200", 0},		/* Atmel */
> >  	{"IFX0102", 0},		/* Infineon */
> > @@ -925,28 +941,34 @@ static struct pnp_device_id tpm_pnp_tbl[] = {
> >  	{"", 0},		/* User Specified */
> >  	{"", 0}			/* Terminator */
> >  };
> 
> Is this OK? I don't know alot about x86 PNP, but I thought the
> pnp_device_id scheme would work with ACPI and legacy PNPBIOS stuff,
> and changing to ACPI means ACPI only?
> 
> If so, should we care? Is there a spec for non-ACPI TPM discovery we
> need to be following here?

I found at least all the IDs listed from drivers/acpi/acpi_pnp.c but you
might be right that they might be (don't know) with pnpbios.

Maybe a better solution would to have two tables and have only MSFT0101
in tpm_acpi_tbl in order to make sure that old functionality is not
broken up because we want this also to the stable kernels.

/Jarkko