Re: [PATCH v3] target/riscv/kvm: add satp mode for host cpu

[Andrew Jones <ajones@ventanamicro.com>]

On Tue, May 20, 2025 at 01:58:45PM +0200, Andrew Jones wrote:
> On Tue, May 20, 2025 at 06:41:03PM +0800, Meng Zhuo wrote:
> > This patch adds host satp mode while kvm/host cpu satp mode is not
> > set.
> > 
> > Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2931
> > Signed-off-by: Meng Zhuo <mengzhuo@iscas.ac.cn>
> > ---
> 
> Please provide a changelog here. I had to go reread the comments on v1 to
> guess at what might have changed.
> 
> >  target/riscv/cpu.c         |  3 +--
> >  target/riscv/cpu.h         |  1 +
> >  target/riscv/kvm/kvm-cpu.c | 20 +++++++++++++++++++-
> >  3 files changed, 21 insertions(+), 3 deletions(-)
> > 
> > diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
> > index d92874baa0..a84edd3a3b 100644
> > --- a/target/riscv/cpu.c
> > +++ b/target/riscv/cpu.c
> > @@ -433,8 +433,7 @@ const char *satp_mode_str(uint8_t satp_mode, bool is_32_bit)
> >      g_assert_not_reached();
> >  }
> >  
> > -static void set_satp_mode_max_supported(RISCVCPU *cpu,
> > -                                        uint8_t satp_mode)
> > +void set_satp_mode_max_supported(RISCVCPU *cpu, uint8_t satp_mode)
> 
> In v1 I suggested we add some sanity checking to this function to ensure
> we can never set an invalid mode (one that's higher than the max
> supported).

It appears [1], which I just learned exists (that series doesn't CC
qemu-riscv) adds an assert for the check (among other changes).

[1] https://lore.kernel.org/all/20250406070254.274797-3-pbonzini@redhat.com/

Thanks,
drew

> 
> >  {
> >      bool rv32 = riscv_cpu_mxl(&cpu->env) == MXL_RV32;
> >      const bool *valid_vm = rv32 ? valid_vm_1_10_32 : valid_vm_1_10_64;
> > diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> > index b56d3afa69..d7136f1d72 100644
> > --- a/target/riscv/cpu.h
> > +++ b/target/riscv/cpu.h
> > @@ -915,6 +915,7 @@ char *riscv_cpu_get_name(RISCVCPU *cpu);
> >  
> >  void riscv_cpu_finalize_features(RISCVCPU *cpu, Error **errp);
> >  void riscv_add_satp_mode_properties(Object *obj);
> > +void set_satp_mode_max_supported(RISCVCPU *cpu, uint8_t satp_mode);
> >  bool riscv_cpu_accelerator_compatible(RISCVCPU *cpu);
> >  
> >  /* CSR function table */
> > diff --git a/target/riscv/kvm/kvm-cpu.c b/target/riscv/kvm/kvm-cpu.c
> > index 82f9728636..18fbca1a08 100644
> > --- a/target/riscv/kvm/kvm-cpu.c
> > +++ b/target/riscv/kvm/kvm-cpu.c
> > @@ -999,6 +999,23 @@ static void kvm_riscv_destroy_scratch_vcpu(KVMScratchCPU *scratch)
> >      close(scratch->kvmfd);
> >  }
> >  
> > +static void kvm_riscv_init_satp_mode(RISCVCPU *cpu, KVMScratchCPU *kvmcpu)
> > +{
> > +    CPURISCVState *env = &cpu->env;
> > +    struct kvm_one_reg reg;
> > +    int ret;
> > +    uint64_t val;
> > +
> > +    reg.id = RISCV_CONFIG_REG(env, satp_mode);
> > +    reg.addr = (uint64_t)&val;
> > +    ret = ioctl(kvmcpu->cpufd, KVM_GET_ONE_REG, &reg);
> > +    if (ret != 0) {
> > +        error_report("Unable to retrieve satp from host, error %d", ret);
> 
> This will output "Unable..., error -1" on error. We should output at least
> the errno and preferably the errno string. I see that this is just a copy
> of a pattern used throughout this file though, so we need a patch fixing
> all of the ioctl call error_reports.
> 
> > +    }
> > +
> > +    set_satp_mode_max_supported(cpu, val);
> > +}
> > +
> >  static void kvm_riscv_init_machine_ids(RISCVCPU *cpu, KVMScratchCPU *kvmcpu)
> >  {
> >      struct kvm_one_reg reg;
> > @@ -1302,6 +1319,7 @@ static void riscv_init_kvm_registers(Object *cpu_obj)
> >      kvm_riscv_init_machine_ids(cpu, &kvmcpu);
> >      kvm_riscv_init_misa_ext_mask(cpu, &kvmcpu);
> >      kvm_riscv_init_cfg(cpu, &kvmcpu);
> > +    kvm_riscv_init_satp_mode(cpu, &kvmcpu);
> >  
> >      kvm_riscv_destroy_scratch_vcpu(&kvmcpu);
> >  }
> > @@ -1980,7 +1998,7 @@ static bool kvm_cpu_realize(CPUState *cs, Error **errp)
> >          }
> >      }
> >  
> > -   return true;
> > +    return true;
> >  }
> >  
> >  void riscv_kvm_cpu_finalize_features(RISCVCPU *cpu, Error **errp)
> > -- 
> > 2.39.5
> > 
> >
> 
> Otherwise,
> 
> Reviewed-by: Andrew Jones <ajones@ventanamicro.com>