Re: [PATCH 3/4] riscv: KVM: Apply insn-def to hfence encodings

[Andrew Jones]

On Mon, Aug 29, 2022 at 02:07:59PM +0530, Anup Patel wrote:
> On Fri, Aug 19, 2022 at 7:32 PM Andrew Jones <ajones@xxxxxxxxxxxxxxxx> wrote:
> >
> > Introduce hfence instruction encodings and apply them to KVM's use.
> > With the self-documenting nature of the instruction encoding macros,
> > and a spec always within arm's reach, it's safe to remove the
> > comments, so we do that too.
> >
> > Signed-off-by: Andrew Jones <ajones@xxxxxxxxxxxxxxxx>
> > ---
> >  arch/riscv/include/asm/insn-def.h |   8 ++
> >  arch/riscv/kvm/tlb.c              | 117 ++++--------------------------
> >  2 files changed, 21 insertions(+), 104 deletions(-)
> >
> > diff --git a/arch/riscv/include/asm/insn-def.h b/arch/riscv/include/asm/insn-def.h
> > index 4cd0208068dd..cd1c0d365f47 100644
> > --- a/arch/riscv/include/asm/insn-def.h
> > +++ b/arch/riscv/include/asm/insn-def.h
> > @@ -79,4 +79,12 @@
> >  #define RS1(v)         __REG(v)
> >  #define RS2(v)         __REG(v)
> >
> > +#define OPCODE_SYSTEM  OPCODE(115)
> > +
> > +#define HFENCE_VVMA(vaddr, asid)       \
> > +       INSN_R(OPCODE_SYSTEM, FUNC3(0), FUNC7(17), RD(0), vaddr, asid)
> > +
> > +#define HFENCE_GVMA(gaddr, vmid)       \
> > +       INSN_R(OPCODE_SYSTEM, FUNC3(0), FUNC7(49), RD(0), gaddr, vmid)
> > +
> >  #endif /* __ASM_INSN_DEF_H */
> > diff --git a/arch/riscv/kvm/tlb.c b/arch/riscv/kvm/tlb.c
> > index 1a76d0b1907d..f742a0d888e1 100644
> > --- a/arch/riscv/kvm/tlb.c
> > +++ b/arch/riscv/kvm/tlb.c
> > @@ -12,22 +12,7 @@
> >  #include <linux/kvm_host.h>
> >  #include <asm/cacheflush.h>
> >  #include <asm/csr.h>
> > -
> > -/*
> > - * Instruction encoding of hfence.gvma is:
> > - * HFENCE.GVMA rs1, rs2
> > - * HFENCE.GVMA zero, rs2
> > - * HFENCE.GVMA rs1
> > - * HFENCE.GVMA
> > - *
> > - * rs1!=zero and rs2!=zero ==> HFENCE.GVMA rs1, rs2
> > - * rs1==zero and rs2!=zero ==> HFENCE.GVMA zero, rs2
> > - * rs1!=zero and rs2==zero ==> HFENCE.GVMA rs1
> > - * rs1==zero and rs2==zero ==> HFENCE.GVMA
> > - *
> > - * Instruction encoding of HFENCE.GVMA is:
> > - * 0110001 rs2(5) rs1(5) 000 00000 1110011
> > - */
> > +#include <asm/insn-def.h>
> >
> >  void kvm_riscv_local_hfence_gvma_vmid_gpa(unsigned long vmid,
> >                                           gpa_t gpa, gpa_t gpsz,
> > @@ -41,31 +26,14 @@ void kvm_riscv_local_hfence_gvma_vmid_gpa(unsigned long vmid,
> >         }
> >
> >         for (pos = gpa; pos < (gpa + gpsz); pos += BIT(order)) {
> > -               /*
> > -                * rs1 = a0 (GPA >> 2)
> > -                * rs2 = a1 (VMID)
> > -                * HFENCE.GVMA a0, a1
> > -                * 0110001 01011 01010 000 00000 1110011
> > -                */
> > -               asm volatile ("srli a0, %0, 2\n"
> > -                             "add a1, %1, zero\n"
> > -                             ".word 0x62b50073\n"
> > -                             :: "r" (pos), "r" (vmid)
> > -                             : "a0", "a1", "memory");
> > +               asm volatile (HFENCE_GVMA("%0", "%1")

Thank you for the review, Anup! I'd also like to get opinions on whether
the caller should quote the register tokens or the call should be made as,
e.g. HFENCE_GVMA(%0, %1), and then do the quoting inside the macro for
C callers. I could go either way, but I'm starting to lean towards moving
the quoting into the macros.

Thanks,
drew