Re: [PATCH v4 1/3] powerpc/book3s64/radix/tlb: tlbie primitives for process-scoped invalidations from guests

[Bharata B Rao <bharata@linux.ibm.com>]

On Wed, Feb 17, 2021 at 11:24:48AM +1100, David Gibson wrote:
> On Mon, Feb 15, 2021 at 12:05:40PM +0530, Bharata B Rao wrote:
> > H_RPT_INVALIDATE hcall needs to perform process scoped tlbie
> > invalidations of L1 and nested guests from L0. This needs RS register
> > for TLBIE instruction to contain both PID and LPID. Introduce
> > primitives that execute tlbie instruction with both PID
> > and LPID set in prepartion for H_RPT_INVALIDATE hcall.
> > 
> > While we are here, move RIC_FLUSH definitions to header file
> > and introduce helper rpti_pgsize_to_psize() that will be needed
> > by the upcoming hcall.
> > 
> > Signed-off-by: Bharata B Rao <bharata@linux.ibm.com>
> > ---
> >  .../include/asm/book3s/64/tlbflush-radix.h    |  18 +++
> >  arch/powerpc/mm/book3s64/radix_tlb.c          | 122 +++++++++++++++++-
> >  2 files changed, 136 insertions(+), 4 deletions(-)
> > 
> > diff --git a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> > index 94439e0cefc9..aace7e9b2397 100644
> > --- a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> > +++ b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> > @@ -4,6 +4,10 @@
> >  
> >  #include <asm/hvcall.h>
> >  
> > +#define RIC_FLUSH_TLB 0
> > +#define RIC_FLUSH_PWC 1
> > +#define RIC_FLUSH_ALL 2
> > +
> >  struct vm_area_struct;
> >  struct mm_struct;
> >  struct mmu_gather;
> > @@ -21,6 +25,20 @@ static inline u64 psize_to_rpti_pgsize(unsigned long psize)
> >  	return H_RPTI_PAGE_ALL;
> >  }
> >  
> > +static inline int rpti_pgsize_to_psize(unsigned long page_size)
> > +{
> > +	if (page_size == H_RPTI_PAGE_4K)
> > +		return MMU_PAGE_4K;
> > +	if (page_size == H_RPTI_PAGE_64K)
> > +		return MMU_PAGE_64K;
> > +	if (page_size == H_RPTI_PAGE_2M)
> > +		return MMU_PAGE_2M;
> > +	if (page_size == H_RPTI_PAGE_1G)
> > +		return MMU_PAGE_1G;
> > +	else
> > +		return MMU_PAGE_64K; /* Default */
> > +}
> 
> Would it make sense to put the H_RPT_PAGE_ tags into the
> mmu_psize_defs table and scan that here, rather than open coding the
> conversion?

I will give this a try and see how it looks.

Otherwise the changes in the patch which are mainly about
introducing primitives that require to set both PID and LPID
for tlbie instruction - do they look right?

Regards,
Bharata.