Re: [Xen-devel] [PATCH] x86/cpuid: AVX-512 Feature Detection

["Kang, Luwei" <luwei.kang@xxxxxxxxx>]

OK, no problem.


-----Original Message-----
From: Andrew Cooper [mailto:andrew.cooper3@xxxxxxxxxx] 
Sent: Tuesday, June 28, 2016 4:47 PM
To: Kang, Luwei <luwei.kang@xxxxxxxxx>; xen-devel@xxxxxxxxxxxxx
Cc: jbeulich@xxxxxxxx; Wang, Yong Y <yong.y.wang@xxxxxxxxx>; Peng, Chao P 
<chao.p.peng@xxxxxxxxx>
Subject: Re: [PATCH] x86/cpuid: AVX-512 Feature Detection

On 28/06/16 06:51, Luwei Kang wrote:
> @@ -1136,9 +1136,16 @@ void pv_cpuid(struct cpu_user_regs *regs)
>      case XSTATE_CPUID:
>  
>          if ( !is_control_domain(currd) && !is_hardware_domain(currd) 
> )
> +        {
>              domain_cpuid(currd, 1, 0, &tmp, &tmp, &_ecx, &tmp);
> +            domain_cpuid(currd, 0x07, 0, &tmp, &_ebx, &tmp, &tmp);
> +        }
>          else
> +        {
>              _ecx = cpuid_ecx(1);
> +            cpuid_count(0x07, 0, &tmp, &_ebx, &tmp, &tmp);
> +        }
> +

In addition to Jan's comments, having _ecx from one leaf and _ebx from a 
different leaf collected at the same time is liable to cause confusion.

Please split the cpuid call for leaf 7 out from here, and put it in the next 
hunk, just like the way the hvm_cpuid() side works.

~Andrew

>          _ecx &= pv_featureset[FEATURESET_1c];
>  
>          if ( !(_ecx & cpufeat_mask(X86_FEATURE_XSAVE)) || subleaf >= 
> 63 ) @@ -1157,6 +1164,14 @@ void pv_cpuid(struct cpu_user_regs *regs)
>                                 xstate_sizes[_XSTATE_YMM]);
>              }
>  
> +            if ( _ebx & cpufeat_mask(X86_FEATURE_AVX512F) )
> +            {
> +                xfeature_mask |= XSTATE_OPMASK | XSTATE_ZMM | XSTATE_HI_ZMM;
> +                xstate_size = max(xstate_size,
> +                                  xstate_offsets[_XSTATE_HI_ZMM] +
> +                                  xstate_sizes[_XSTATE_HI_ZMM]);
> +            }
> +
>              a = (uint32_t)xfeature_mask;
>              d = (uint32_t)(xfeature_mask >> 32);
>              c = xstate_size;


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel