Re: [Patch v4] x86/traps: improvements to {rd, wr}msr_hypervisor_regs()

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 08/10/13 09:52, Jan Beulich wrote:
>>>> On 07.10.13 at 15:46, Andrew Cooper <andrew.cooper3@citrix.com> wrote:
>> --- a/xen/arch/x86/traps.c
>> +++ b/xen/arch/x86/traps.c
>> @@ -595,55 +595,50 @@ DO_ERROR_NOCODE(TRAP_copro_error,     coprocessor_error)
>>  DO_ERROR(       TRAP_alignment_check, alignment_check)
>>  DO_ERROR_NOCODE(TRAP_simd_error,      simd_coprocessor_error)
>>  
>> +/*
>> + * Returns 0 if not handled, and non-0 for error. (The calling semantics are
>> + * in need of some work)
>> + */
>>  int rdmsr_hypervisor_regs(uint32_t idx, uint64_t *val)
>>  {
> The comment here still isn't in line with the existing callers. Non-
> zero means success afaict. There simply is no path resulting in
> an error here so far.
>
>> +    switch ( idx - base )
>>      {
>> -    case 0:
>> +    case 0: /* Write hypercall page.  Reads are invalid. Hand a #GP back. */
>>      {
>>          *val = 0;
>> -        break;
>> +        return 1;
> And the above means that there's no #GP being "handed back"
> here either.
>
> Jan
>

/sigh - Quite right.  I misread the where the breaks were going in the
callers.

~Andrew