[Qemu-devel] Re: [Qemu-commits] [COMMIT 883ecbf] Remove noisy printf when KVM masks CPU features

[Qemu-devel] Re: [Qemu-commits] [COMMIT 883ecbf] Remove noisy printf when KVM masks CPU features

[Avi Kivity]

Anthony Liguori wrote:
> From: Anthony Liguori <aliguori@us.ibm.com>
>
> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
>
> diff --git a/target-i386/helper.c b/target-i386/helper.c
> index bdf242b..5c03e3a 100644
> --- a/target-i386/helper.c
> +++ b/target-i386/helper.c
> @@ -102,7 +102,6 @@ static void kvm_trim_features(uint32_t *features, uint32_t supported,
>      for (i = 0; i < 32; ++i) {
>          mask = 1U << i;
>          if ((*features & mask) && !(supported & mask)) {
> -            printf("Processor feature %s not supported by kvm\n", names[i]);
>              *features &= ~mask;
>          }
>      }
>   

That was intentional, I didn't want to silently drop features the user 
requested as those may impact live migration or load/save (or 
application performance).

Maybe it's better to only do that if -cpu was specified on the command 
line; and exit() as well.  The user can easily pick up the missing bits 
and remove them from the requested cpu features.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

[Qemu-devel] Re: [Qemu-commits] [COMMIT 883ecbf] Remove noisy printf when KVMmasks CPU features

[Anthony Liguori]

Avi Kivity wrote:
> Anthony Liguori wrote:
>> From: Anthony Liguori <aliguori@us.ibm.com>
>>
>> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
>>
>> diff --git a/target-i386/helper.c b/target-i386/helper.c
>> index bdf242b..5c03e3a 100644
>> --- a/target-i386/helper.c
>> +++ b/target-i386/helper.c
>> @@ -102,7 +102,6 @@ static void kvm_trim_features(uint32_t *features, 
>> uint32_t supported,
>>      for (i = 0; i < 32; ++i) {
>>          mask = 1U << i;
>>          if ((*features & mask) && !(supported & mask)) {
>> -            printf("Processor feature %s not supported by kvm\n", 
>> names[i]);
>>              *features &= ~mask;
>>          }
>>      }
>>   
>
> That was intentional, I didn't want to silently drop features the user 
> requested as those may impact live migration or load/save (or 
> application performance).

Why not provide a monitor option to view the list of CPUID features that 
are enabled?

printing to stdio is a bad idea.  We print the names of pty devices to 
stdio and tools have to parse that today.   I hope they're sufficiently 
robust but this isn't something we want to introduce if we don't have to.

> Maybe it's better to only do that if -cpu was specified on the command 
> line; and exit() as well.  The user can easily pick up the missing 
> bits and remove them from the requested cpu features.

Yes, that's also a good idea.

Regards,

Anthony Liguori


-- 
Regards,

Anthony Liguori