| Message ID | 1417691470-5221-2-git-send-email-wanpeng.li@linux.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
2014-12-04 19:11+0800, Wanpeng Li: > The section of CPUID(EAX=0xd, ECX=1) in the spec which commit > f5c2290cd01e (KVM: cpuid: mask more bits in leaf 0xd and subleaves) > mentioned is older than SDM. > > EBX: Bits 31-00: The size in bytes of the XSAVE area containing all > states enabled by XCR0|IA32_XSS. Well, CPUs without XSAVES return 0 there, so we would emulate them incorrectly ... (I don't mind much, it is reserved.) > The the value of EBX should represent the size of XCR0 related XSAVE > area since IA32_XSS is not used currently. True, but 'supported' is not the state of XCR0, just its supremum. EBX should be set in kvm_update_cpuid(), like [3/4] does. (We can safely drop [2/4].) > Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com> > --- > v1 -> v2: > * add F(XSAVEC) check > > arch/x86/kvm/cpuid.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c > index 646e6e8..5b78e9b 100644 > --- a/arch/x86/kvm/cpuid.c > +++ b/arch/x86/kvm/cpuid.c > @@ -477,7 +477,10 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function, > do_cpuid_1_ent(&entry[i], function, idx); > if (idx == 1) { > entry[i].eax &= kvm_supported_word10_x86_features; > - entry[i].ebx = 0; > + if (entry[i].eax & (F(XSAVES) | F(XSAVEC))) > + entry[i].ebx = xstate_required_size(supported, true); > + else > + entry[i].ebx = xstate_required_size(supported, false); > } else { > if (entry[i].eax == 0 || !(supported & mask)) > continue; > -- > 1.9.1 > -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 04/12/2014 14:14, Radim Kr?má? wrote: > 2014-12-04 19:11+0800, Wanpeng Li: >> The section of CPUID(EAX=0xd, ECX=1) in the spec which commit >> f5c2290cd01e (KVM: cpuid: mask more bits in leaf 0xd and subleaves) >> mentioned is older than SDM. >> >> EBX: Bits 31-00: The size in bytes of the XSAVE area containing all >> states enabled by XCR0|IA32_XSS. > > Well, CPUs without XSAVES return 0 there, so we would emulate them > incorrectly ... (I don't mind much, it is reserved.) I agree it should stay to 0 if !XSAVES && !XSAVEC. For !XSAVEC && XSAVES there's no silicon, so we have some leeway. >> The the value of EBX should represent the size of XCR0 related XSAVE >> area since IA32_XSS is not used currently. > > True, but 'supported' is not the state of XCR0, just its supremum. > EBX should be set in kvm_update_cpuid(), like [3/4] does. > (We can safely drop [2/4].) Still, it's nice to be consistent and return a plausible value to userspace for KVM_GET_SUPPORTED_CPUID. Paolo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c index 646e6e8..5b78e9b 100644 --- a/arch/x86/kvm/cpuid.c +++ b/arch/x86/kvm/cpuid.c @@ -477,7 +477,10 @@ static inline int __do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function, do_cpuid_1_ent(&entry[i], function, idx); if (idx == 1) { entry[i].eax &= kvm_supported_word10_x86_features; - entry[i].ebx = 0; + if (entry[i].eax & (F(XSAVES) | F(XSAVEC))) + entry[i].ebx = xstate_required_size(supported, true); + else + entry[i].ebx = xstate_required_size(supported, false); } else { if (entry[i].eax == 0 || !(supported & mask)) continue;
The section of CPUID(EAX=0xd, ECX=1) in the spec which commit f5c2290cd01e (KVM: cpuid: mask more bits in leaf 0xd and subleaves) mentioned is older than SDM. EBX: Bits 31-00: The size in bytes of the XSAVE area containing all states enabled by XCR0|IA32_XSS. The the value of EBX should represent the size of XCR0 related XSAVE area since IA32_XSS is not used currently. Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com> --- v1 -> v2: * add F(XSAVEC) check arch/x86/kvm/cpuid.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)