[RFC 7/9] RISC-V: KVM: Implement trap & emulate for hpmcounters

[Andrew Jones]

On Mon, Jul 18, 2022 at 10:02:03AM -0700, Atish Patra wrote:
> As the KVM guests only see the virtual PMU counters, all hpmcounter
> access should trap and KVM emulates the read access on behalf of guests.
> 
> Signed-off-by: Atish Patra <atishp at rivosinc.com>
> ---
>  arch/riscv/include/asm/kvm_vcpu_pmu.h | 16 +++++++++
>  arch/riscv/kvm/vcpu_insn.c            |  1 +
>  arch/riscv/kvm/vcpu_pmu.c             | 47 +++++++++++++++++++++++----
>  3 files changed, 57 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/riscv/include/asm/kvm_vcpu_pmu.h b/arch/riscv/include/asm/kvm_vcpu_pmu.h
> index bffee052f2ae..5410236b62a8 100644
> --- a/arch/riscv/include/asm/kvm_vcpu_pmu.h
> +++ b/arch/riscv/include/asm/kvm_vcpu_pmu.h
> @@ -39,6 +39,19 @@ struct kvm_pmu {
>  #define pmu_to_vcpu(pmu)  (container_of((pmu), struct kvm_vcpu, arch.pmu))
>  #define pmc_to_pmu(pmc)   (&(pmc)->vcpu->arch.pmu)
>  
> +#if defined(CONFIG_32BIT)
> +#define KVM_RISCV_VCPU_HPMCOUNTER_CSR_FUNCS \
> +{ .base = CSR_CYCLEH,      .count = 31, .func = kvm_riscv_vcpu_pmu_read_hpm }, \
> +{ .base = CSR_CYCLE,      .count = 31, .func = kvm_riscv_vcpu_pmu_read_hpm },
> +#else
> +#define KVM_RISCV_VCPU_HPMCOUNTER_CSR_FUNCS \
> +{ .base = CSR_CYCLE,      .count = 31, .func = kvm_riscv_vcpu_pmu_read_hpm },
> +#endif
> +
> +int kvm_riscv_vcpu_pmu_read_hpm(struct kvm_vcpu *vcpu, unsigned int csr_num,
> +				unsigned long *val, unsigned long new_val,
> +				unsigned long wr_mask);
> +
>  int kvm_riscv_vcpu_pmu_num_ctrs(struct kvm_vcpu *vcpu, unsigned long *out_val);
>  int kvm_riscv_vcpu_pmu_ctr_info(struct kvm_vcpu *vcpu, unsigned long cidx,
>  				unsigned long *ctr_info);
> @@ -59,6 +72,9 @@ void kvm_riscv_vcpu_pmu_reset(struct kvm_vcpu *vcpu);
>  #else
>  struct kvm_pmu {
>  };
> +#define KVM_RISCV_VCPU_HPMCOUNTER_CSR_FUNCS \
> +{ .base = 0,      .count = 0, .func = NULL },
> +
>  
>  static inline int kvm_riscv_vcpu_pmu_init(struct kvm_vcpu *vcpu)
>  {
> diff --git a/arch/riscv/kvm/vcpu_insn.c b/arch/riscv/kvm/vcpu_insn.c
> index 0aa334f853c8..7c2a4b1a69f7 100644
> --- a/arch/riscv/kvm/vcpu_insn.c
> +++ b/arch/riscv/kvm/vcpu_insn.c
> @@ -215,6 +215,7 @@ struct csr_func {
>  };
>  
>  static const struct csr_func csr_funcs[] = {
> +	KVM_RISCV_VCPU_HPMCOUNTER_CSR_FUNCS
>  };
>  
>  /**
> diff --git a/arch/riscv/kvm/vcpu_pmu.c b/arch/riscv/kvm/vcpu_pmu.c
> index 3168ed740bdd..5434051f495d 100644
> --- a/arch/riscv/kvm/vcpu_pmu.c
> +++ b/arch/riscv/kvm/vcpu_pmu.c
> @@ -14,6 +14,46 @@
>  #include <asm/kvm_vcpu_pmu.h>
>  #include <linux/kvm_host.h>
>  
> +int kvm_riscv_vcpu_pmu_ctr_read(struct kvm_vcpu *vcpu, unsigned long cidx,
> +				unsigned long *out_val)
> +{
> +	struct kvm_pmu *kvpmu = vcpu_to_pmu(vcpu);
> +	struct kvm_pmc *pmc;
> +	u64 enabled, running;
> +
> +	if (!kvpmu)
> +		return -EINVAL;
> +
> +	pmc = &kvpmu->pmc[cidx];
> +	if (!pmc->perf_event)
> +		return -EINVAL;
> +
> +	pmc->counter_val += perf_event_read_value(pmc->perf_event, &enabled, &running);
> +	*out_val = pmc->counter_val;
> +
> +	return 0;
> +}
> +
> +int kvm_riscv_vcpu_pmu_read_hpm(struct kvm_vcpu *vcpu, unsigned int csr_num,
> +				unsigned long *val, unsigned long new_val,
> +				unsigned long wr_mask)
> +{
> +	struct kvm_pmu *kvpmu = vcpu_to_pmu(vcpu);
> +	int cidx, ret = KVM_INSN_CONTINUE_NEXT_SEPC;
> +
> +	if (!kvpmu)
> +		return KVM_INSN_EXIT_TO_USER_SPACE;
> +	//TODO: Should we check if vcpu pmu is initialized or not!

I guess it depends on the path to this call. It'd be best to keep the
checks to the minimum, so if this isn't a top level call then I'd say
no, but we need to check in the top level.

> +	if (wr_mask)
> +		return KVM_INSN_ILLEGAL_TRAP;
> +	cidx = csr_num - CSR_CYCLE;
> +
> +	if (kvm_riscv_vcpu_pmu_ctr_read(vcpu, cidx, val) < 0)
> +		return KVM_INSN_EXIT_TO_USER_SPACE;
> +
> +	return ret;
> +}
> +
>  int kvm_riscv_vcpu_pmu_num_ctrs(struct kvm_vcpu *vcpu, unsigned long *out_val)
>  {
>  	struct kvm_pmu *kvpmu = vcpu_to_pmu(vcpu);
> @@ -60,13 +100,6 @@ int kvm_riscv_vcpu_pmu_ctr_cfg_match(struct kvm_vcpu *vcpu, unsigned long ctr_ba
>  	return 0;
>  }
>  
> -int kvm_riscv_vcpu_pmu_ctr_read(struct kvm_vcpu *vcpu, unsigned long cidx,
> -				unsigned long *out_val)
> -{
> -	/* TODO */
> -	return 0;
> -}
> -
>  int kvm_riscv_vcpu_pmu_init(struct kvm_vcpu *vcpu)
>  {
>  	int i = 0, num_hw_ctrs, num_fw_ctrs, hpm_width;
> -- 
> 2.25.1
>

Thanks,
drew