[PATCH v3 3/3] kvm: x86: Enable Intel XSAVES for guest

[PATCH v3 3/3] kvm: x86: Enable Intel XSAVES for guest

[Wanpeng Li]

Exporse intel xsaves feature to guest.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
---
 arch/x86/kvm/cpuid.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index a4f5ac4..0d919bc 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -267,6 +267,7 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
 	unsigned f_rdtscp = kvm_x86_ops->rdtscp_supported() ? F(RDTSCP) : 0;
 	unsigned f_invpcid = kvm_x86_ops->invpcid_supported() ? F(INVPCID) : 0;
 	unsigned f_mpx = kvm_x86_ops->mpx_supported() ? F(MPX) : 0;
+	unsigned f_xsaves = kvm_x86_ops->xsaves_supported() ? F(XSAVES) : 0;
 
 	/* cpuid 1.edx */
 	const u32 kvm_supported_word0_x86_features =
@@ -322,7 +323,7 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
 
 	/* cpuid 0xD.1.eax */
 	const u32 kvm_supported_word10_x86_features =
-		F(XSAVEOPT) | F(XSAVEC) | F(XGETBV1);
+		F(XSAVEOPT) | F(XSAVEC) | F(XGETBV1) | f_xsaves;
 
 	/* all calls to cpuid_count() should be made on the same cpu */
 	get_cpu();
-- 
1.9.1

Re: [PATCH v3 3/3] kvm: x86: Enable Intel XSAVES for guest

[Radim Krčmář]

2014-12-02 19:21+0800, Wanpeng Li:
> Exporse intel xsaves feature to guest.

0xD.1:ebx ought to be non-zero with XSAVES, even if IA32_XSS is known to
be 0, so we'll need to set it after Paolo's patch.

> Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
> ---
>  arch/x86/kvm/cpuid.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index a4f5ac4..0d919bc 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -267,6 +267,7 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
>  	unsigned f_rdtscp = kvm_x86_ops->rdtscp_supported() ? F(RDTSCP) : 0;
>  	unsigned f_invpcid = kvm_x86_ops->invpcid_supported() ? F(INVPCID) : 0;
>  	unsigned f_mpx = kvm_x86_ops->mpx_supported() ? F(MPX) : 0;
> +	unsigned f_xsaves = kvm_x86_ops->xsaves_supported() ? F(XSAVES) : 0;
>  
>  	/* cpuid 1.edx */
>  	const u32 kvm_supported_word0_x86_features =
> @@ -322,7 +323,7 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
>  
>  	/* cpuid 0xD.1.eax */
>  	const u32 kvm_supported_word10_x86_features =
> -		F(XSAVEOPT) | F(XSAVEC) | F(XGETBV1);
> +		F(XSAVEOPT) | F(XSAVEC) | F(XGETBV1) | f_xsaves;
>  
>  	/* all calls to cpuid_count() should be made on the same cpu */
>  	get_cpu();
> -- 
> 1.9.1

--- rant ---
The documentation isn't clearly backward-compatible, it took me a while
to understand why

  EBX  Bits 31-00: The size in bytes of the XSAVE area containing all
       states enabled by XCRO | IA32_XSS.

Is 0 on my old machine.

Different section gives a better hint by mentioning XSAVES directly
 * EBX enumerates the size (in bytes) required by the XSAVES
   instruction for an XSAVE area containing all the state components
   corresponding to bits currently set in XCR0 | IA32_XSS.