Re: [Xen-devel] [PATCH v3 20/62] arm/acpi: Add ACPI support for SMP initialization

[Mark Rutland <mark.rutland@xxxxxxx>]

On Mon, Jan 04, 2016 at 02:51:51PM +0000, Stefano Stabellini wrote:
> On Wed, 30 Dec 2015, Shannon Zhao wrote:
> > On 2015/11/30 22:57, Julien Grall wrote:
> > > Hi Shannon,
> > > 
> > > On 17/11/15 09:40, shannon.zhao@xxxxxxxxxx wrote:
> > >> > diff --git a/xen/arch/arm/psci.c b/xen/arch/arm/psci.c
> > >> > index d800cb6..dede0e1 100644
> > >> > --- a/xen/arch/arm/psci.c
> > >> > +++ b/xen/arch/arm/psci.c
> > >> > @@ -22,6 +22,7 @@
> > >> >  #include <xen/mm.h>
> > >> >  #include <xen/smp.h>
> > >> >  #include <asm/psci.h>
> > >> > +#include <asm/acpi.h>
> > >> >  
> > >> >  /*
> > >> >   * While a 64-bit OS can make calls with SMC32 calling conventions, 
> > >> > for
> > >> > @@ -86,6 +87,9 @@ int __init psci_init_0_1(void)
> > >> >      int ret;
> > >> >      const struct dt_device_node *psci;
> > >> >  
> > >> > +    if ( !acpi_disabled )
> > >> > +        return -EINVAL;
> > > Please explain in the commit message why PSCI 0.1 is not supported on 
> > > ACPI.
> > 
> > Hi,
> > 
> > I check this again. There are not limitations of supporting PSCI version
> > in ACPI SPEC. It should support PSCI 0.1 as well. But look at the code
> > of linux kernel, it says it only supports PSCI 0.2+.
> > 
> > #define ACPI_FADT_PSCI_COMPLIANT    (1)     /* 00: [V5+] PSCI 0.2+ is
> > implemented */
> > 
> > So does it need to be consistent with Linux or support PSCI 0.1 in Xen
> > as well?
> 
> I don't think it needs to be consistent with Linux. I would support PSCI
> 0.1 too.

That's not possible, so I don't follow. Prior to 0.2 the function IDs
are not defined.

The FADT has a single bit which describes PSCI 0.2+ being implemented,
and does not describe function IDs.

You must assume PSCI 0.2+ in order to have a set of function IDs to use.

You should also assume the presence of the rest of the mandatory
portions of PSCI 0.2, as these are also required per the combination of
the PSCI spec and the ACPI spec.

Mark.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel