On Wed, May 21, 2014 at 12:10:45PM +0100, James Hogan wrote:
> On 20/05/14 15:47, Andreas Herrmann wrote:
> > From: David Daney <david.daney@xxxxxxxxxx>
> >
> > This returns the CPUNum from the low order Ebase bits.
> >
> > Signed-off-by: David Daney <david.daney@xxxxxxxxxx>
> > Signed-off-by: Andreas Herrmann <andreas.herrmann@xxxxxxxxxxxxxxxxxx>
> > ---
> > arch/mips/include/asm/mipsregs.h | 5 +++++
> > 1 file changed, 5 insertions(+)
> >
> > diff --git a/arch/mips/include/asm/mipsregs.h b/arch/mips/include/asm/mipsregs.h
> > index 3e025b5..f110d48 100644
> > --- a/arch/mips/include/asm/mipsregs.h
> > +++ b/arch/mips/include/asm/mipsregs.h
> > @@ -1916,6 +1916,11 @@ __BUILD_SET_C0(brcm_cmt_ctrl)
> > __BUILD_SET_C0(brcm_config)
> > __BUILD_SET_C0(brcm_mode)
> >
> > +static inline unsigned int mips_cpunum(void)
> > +{
> > + return read_c0_ebase() & 0x3ff; /* Low 10 bits of ebase. */
> > +}
>
> If this is going to go in mips generic code I think it should be clearly
> defined, especially in the presence of MT, otherwise perhaps it makes
> sense for it to go in a paravirt specific header?
It's just wrapper to read ebase_cpunum. Currently only used in the
paravirt-code (to get CPUnum for a guest CPU -- which eventually is
read from guest cp0 context).
I am not sure whether it needs to be moved to a paravirt specific
header.
> I.e. does it return the core number of the running VPE (if so it should
> probably do something like below as in decode_configs() and go in
> smp.h), or does it simply always return that field in ebase register (in
> which case it should probably have ebase in the name and a comment to
> clarify that it doesn't necessarily map directly to core/vpe number).
Under KVM (MIPSVZ) it effectively returns vcpu_id and thus such a
comment could be added for clarification.
So, should we name it get_ebase_cpunum() but keep it in this header
(and also replace the 1 or 2 occurrences of "read_c0_ebase() & 0x3ff"
in the non-paravirt code with it)?
> unsigned int core = read_c0_ebase() & 0x3ff;
> if (cpu_has_mipsmt)
> core >>= fls(smp_num_siblings) - 1;
>
> Cheers
> James
Andreas
