Re: [PATCH v2 2/2] KVM: x86: On emulation failure, convey the exit reason to userspace

[David Edmondson <david.edmondson@oracle.com>]

On Friday, 2021-07-09 at 15:47:08 UTC, Sean Christopherson wrote:

> On Tue, Jul 06, 2021, David Edmondson wrote:
>> Should instruction emulation fail, include the VM exit reason in the
>> emulation_failure data passed to userspace, in order that the VMM can
>> report it as a debugging aid when describing the failure.
>
> ...
>
>> @@ -7473,7 +7474,14 @@ static void prepare_emulation_failure_exit(struct kvm_vcpu *vcpu)
>>  		memcpy(run->emulation_failure.insn_bytes,
>>  		       ctxt->fetch.data, insn_size);
>>  	}
>> +
>> +	run->emulation_failure.ndata = 4;
>> +	run->emulation_failure.flags |=
>> +		KVM_INTERNAL_ERROR_EMULATION_FLAG_EXIT_REASON;
>> +	run->emulation_failure.exit_reason =
>> +		static_call(kvm_x86_get_exit_reason)(vcpu);
>>  }
>
> ...
>
>> diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
>> index d9e4aabcb31a..863195371272 100644
>> --- a/include/uapi/linux/kvm.h
>> +++ b/include/uapi/linux/kvm.h
>> @@ -282,6 +282,7 @@ struct kvm_xen_exit {
>>  
>>  /* Flags that describe what fields in emulation_failure hold valid data. */
>>  #define KVM_INTERNAL_ERROR_EMULATION_FLAG_INSTRUCTION_BYTES (1ULL << 0)
>> +#define KVM_INTERNAL_ERROR_EMULATION_FLAG_EXIT_REASON       (1ULL << 1)
>>  
>>  /* for KVM_RUN, returned by mmap(vcpu_fd, offset=0) */
>>  struct kvm_run {
>> @@ -404,6 +405,12 @@ struct kvm_run {
>>  			__u64 flags;
>>  			__u8  insn_size;
>>  			__u8  insn_bytes[15];
>> +			/*
>> +			 * The "exit reason" extracted from the
>> +			 * VMCS/VMCB that was the cause of attempted
>> +			 * emulation.
>> +			 */
>> +			__u64 exit_reason;
>
> Rather than providing just the exit reason and adding another kvm_x86_ops hook,
> I would prefer to extend kvm_x86_get_exit_info() to also provide the exit reason
> and use that.  E.g. on VMX, all exceptions funnel through a single exit reason.
> Dumping exit_info_{1,2} and error_code in addition to intr_info might not be all
> that useful, but I can't see in harm either, and more info is generally a good
> thing.
>
> The only other user of kvm_x86_get_exit_info() is for tracepoints, those could
> be modified to not pass in the exit reason.

Okay.

>>  		} emulation_failure;
>>  		/* KVM_EXIT_OSI */
>>  		struct {
>> -- 
>> 2.30.2
>> 

dme.
-- 
What did you learn today? I learnt nothing.