Re: [PATCH v5 5/5] KVM: PPC: Book3s: mmio: Deliver DSI after emulation failure

[Nicholas Piggin <npiggin@gmail.com>]

Excerpts from Fabiano Rosas's message of January 26, 2022 7:56 am:
> MMIO emulation can fail if the guest uses an instruction that we are
> not prepared to emulate. Since these instructions can be and most
> likely are valid ones, this is (slightly) closer to an access fault
> than to an illegal instruction, so deliver a Data Storage interrupt
> instead of a Program interrupt.
> 
> BookE ignores bad faults, so it will keep using a Program interrupt
> because a DSI would cause a fault loop in the guest.
> 
> Suggested-by: Nicholas Piggin <npiggin@gmail.com>
> Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>

Thanks this looks good to me. (And thanks for updating patch 4/5 with
the kvm debug print helper.)

Reviewed-by: Nicholas Piggin <npiggin@gmail.com>

> ---
>  arch/powerpc/kvm/emulate_loadstore.c | 10 +++-------
>  arch/powerpc/kvm/powerpc.c           | 22 ++++++++++++++++++++++
>  2 files changed, 25 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/powerpc/kvm/emulate_loadstore.c b/arch/powerpc/kvm/emulate_loadstore.c
> index 48272a9b9c30..cfc9114b87d0 100644
> --- a/arch/powerpc/kvm/emulate_loadstore.c
> +++ b/arch/powerpc/kvm/emulate_loadstore.c
> @@ -73,7 +73,6 @@ int kvmppc_emulate_loadstore(struct kvm_vcpu *vcpu)
>  {
>  	u32 inst;
>  	enum emulation_result emulated = EMULATE_FAIL;
> -	int advance = 1;
>  	struct instruction_op op;
>  
>  	/* this default type might be overwritten by subcategories */
> @@ -98,6 +97,8 @@ int kvmppc_emulate_loadstore(struct kvm_vcpu *vcpu)
>  		int type = op.type & INSTR_TYPE_MASK;
>  		int size = GETSIZE(op.type);
>  
> +		vcpu->mmio_is_write = OP_IS_STORE(type);
> +
>  		switch (type) {
>  		case LOAD:  {
>  			int instr_byte_swap = op.type & BYTEREV;
> @@ -355,15 +356,10 @@ int kvmppc_emulate_loadstore(struct kvm_vcpu *vcpu)
>  		}
>  	}
>  
> -	if (emulated == EMULATE_FAIL) {
> -		advance = 0;
> -		kvmppc_core_queue_program(vcpu, 0);
> -	}
> -
>  	trace_kvm_ppc_instr(inst, kvmppc_get_pc(vcpu), emulated);
>  
>  	/* Advance past emulated instruction. */
> -	if (advance)
> +	if (emulated != EMULATE_FAIL)
>  		kvmppc_set_pc(vcpu, kvmppc_get_pc(vcpu) + 4);
>  
>  	return emulated;
> diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
> index acb0d2a4bdb9..82d889db2b6b 100644
> --- a/arch/powerpc/kvm/powerpc.c
> +++ b/arch/powerpc/kvm/powerpc.c
> @@ -309,6 +309,28 @@ int kvmppc_emulate_mmio(struct kvm_vcpu *vcpu)
>  		kvmppc_get_last_inst(vcpu, INST_GENERIC, &last_inst);
>  		kvm_debug_ratelimited("Guest access to device memory using unsupported instruction (opcode: %#08x)\n",
>  				      last_inst);
> +
> +		/*
> +		 * Injecting a Data Storage here is a bit more
> +		 * accurate since the instruction that caused the
> +		 * access could still be a valid one.
> +		 */
> +		if (!IS_ENABLED(CONFIG_BOOKE)) {
> +			ulong dsisr = DSISR_BADACCESS;
> +
> +			if (vcpu->mmio_is_write)
> +				dsisr |= DSISR_ISSTORE;
> +
> +			kvmppc_core_queue_data_storage(vcpu, vcpu->arch.vaddr_accessed, dsisr);
> +		} else {
> +			/*
> +			 * BookE does not send a SIGBUS on a bad
> +			 * fault, so use a Program interrupt instead
> +			 * to avoid a fault loop.
> +			 */
> +			kvmppc_core_queue_program(vcpu, 0);
> +		}
> +
>  		r = RESUME_GUEST;
>  		break;
>  	}
> -- 
> 2.34.1
> 
>