Re: [PATCH v2 1/5] target/riscv: Add a virtualised MMU Mode

[Alistair Francis]

On Wed, Oct 28, 2020 at 8:13 AM Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 10/28/20 7:42 AM, Alistair Francis wrote:
> > Add a new MMU mode that includes the current virt mode.
> >
> > Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
> > ---
> >  target/riscv/cpu-param.h  | 10 +++++++++-
> >  target/riscv/cpu.h        |  4 +++-
> >  target/riscv/cpu_helper.c |  6 +++++-
> >  3 files changed, 17 insertions(+), 3 deletions(-)
> >
> > diff --git a/target/riscv/cpu-param.h b/target/riscv/cpu-param.h
> > index 664fc1d371..0db6e23140 100644
> > --- a/target/riscv/cpu-param.h
> > +++ b/target/riscv/cpu-param.h
> > @@ -18,6 +18,14 @@
> >  # define TARGET_VIRT_ADDR_SPACE_BITS 32 /* sv32 */
> >  #endif
> >  #define TARGET_PAGE_BITS 12 /* 4 KiB Pages */
> > -#define NB_MMU_MODES 4
> > +/*
> > + * The current MMU Modes are:
> > + *  - U  mode 0b000
> > + *  - S  mode 0b001
> > + *  - M  mode 0b011
> > + *  - HU mode 0b100
> > + *  - HS mode 0b101
> > + */
> > +#define NB_MMU_MODES 6
> >
> >  #endif
> > diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> > index 87b68affa8..5d8e54c426 100644
> > --- a/target/riscv/cpu.h
> > +++ b/target/riscv/cpu.h
> > @@ -363,7 +363,9 @@ void QEMU_NORETURN riscv_raise_exception(CPURISCVState 
> > *env,
> >  target_ulong riscv_cpu_get_fflags(CPURISCVState *env);
> >  void riscv_cpu_set_fflags(CPURISCVState *env, target_ulong);
> >
> > -#define TB_FLAGS_MMU_MASK   3
> > +#define TB_FLAGS_MMU_MASK   7
> > +#define TB_FLAGS_PRIV_MMU_MASK                3
> > +#define TB_FLAGS_PRIV_HYP_ACCESS_MASK   (1 << 2)
> >  #define TB_FLAGS_MSTATUS_FS MSTATUS_FS
> >
> >  typedef CPURISCVState CPUArchState;
> > diff --git a/target/riscv/cpu_helper.c b/target/riscv/cpu_helper.c
> > index 3eb3a034db..453e4c6d8a 100644
> > --- a/target/riscv/cpu_helper.c
> > +++ b/target/riscv/cpu_helper.c
> > @@ -30,6 +30,10 @@ int riscv_cpu_mmu_index(CPURISCVState *env, bool ifetch)
> >  #ifdef CONFIG_USER_ONLY
> >      return 0;
> >  #else
> > +    if (riscv_cpu_virt_enabled(env)) {
> > +        return env->priv | TB_FLAGS_PRIV_HYP_ACCESS_MASK;
> > +    }
>
> This is wrong.  You only want to set this flag in response to one of the
> hypervisor special instructions.  This is setting it any time virt is enabled.

Isn't that ok though? I thought this was the correct approach.

Now we have a MMU context for Virtual guests (VS) which have different
contexts to the hypervisor (S). It also then means that when doing the
hypervisor access load/stores we can re-use the existing MMU context
from when the Hypervisor guest was running.

Alistair

>
>
> r~