Re: [PATCH v8 15/19] x86/VPMU: Merge vpmu_rdmsr and vpmu_wrmsr

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 07/29/2014 03:46 AM, Jan Beulich wrote:
>>>> On 01.07.14 at 16:37, <boris.ostrovsky@oracle.com> wrote:
>>>>
>> @@ -99,13 +99,21 @@ int vpmu_do_wrmsr(unsigned int msr, uint64_t msr_content)
>>       if ( !(vpmu_mode & XENPMU_MODE_ON) )
>>           return 0;
>>   
>> -    if ( vpmu->arch_vpmu_ops && vpmu->arch_vpmu_ops->do_wrmsr )
>> +    ASSERT((rw == VPMU_MSR_READ) || (rw == VPMU_MSR_WRITE));
>> +
>> +    if ( vpmu->arch_vpmu_ops )
>>       {
>> -        int ret = vpmu->arch_vpmu_ops->do_wrmsr(msr, msr_content);
>> +        int ret;
>>   
>> +        if ( (rw == VPMU_MSR_READ) && vpmu->arch_vpmu_ops->do_rdmsr )
>> +            ret = vpmu->arch_vpmu_ops->do_rdmsr(msr, msr_content);
>> +        else if ( vpmu->arch_vpmu_ops->do_wrmsr )
>> +            ret = vpmu->arch_vpmu_ops->do_wrmsr(msr, *msr_content);
> This won't do what you intend if rw == VPMU_MSR_READ but there's
> no ->do_rdmsr handler.

Right.

I was actually thinking about dropping checks for ops altogether 
(everywhere) and adding nop ops when needed (I think cpuid is the only one).

But if I do this it will be sometime later.


-boris