On Mon, Sep 09, 2013 at 04:09:47PM +0200, Paolo Bonzini wrote:
> Il 08/09/2013 11:54, Gleb Natapov ha scritto:
> > On Thu, Sep 05, 2013 at 02:21:54PM +0200, Paolo Bonzini wrote:
> >> A guest can still attempt to save and restore XSAVE states even if they
> >> have been masked in CPUID leaf 0Dh. This usually is not visible to
> >> the guest, but is still wrong: "Any attempt to set a reserved bit (as
> >> determined by the contents of EAX and EDX after executing CPUID with
> >> EAX=0DH, ECX= 0H) in XCR0 for a given processor will result in a #GP
> >> exception".
> >>
> >> The patch also performs the same checks as __kvm_set_xcr in KVM_SET_XSAVE.
> >> This catches migration from newer to older kernel/processor before the
> >> guest starts running.
> >>
> >> Cc: kvm@xxxxxxxxxxxxxxx
> >> Cc: Gleb Natapov <gleb@xxxxxxxxxx>
> >> Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> >> ---
> >> arch/x86/kvm/cpuid.c | 2 +-
> >> arch/x86/kvm/x86.c | 10 ++++++++--
> >> arch/x86/kvm/x86.h | 1 +
> >> 3 files changed, 10 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> >> index a20ecb5..d7c465d 100644
> >> --- a/arch/x86/kvm/cpuid.c
> >> +++ b/arch/x86/kvm/cpuid.c
> >> @@ -182,7 +182,7 @@ static bool supported_xcr0_bit(unsigned bit)
> >> {
> >> u64 mask = ((u64)1 << bit);
> >>
> >> - return mask & (XSTATE_FP | XSTATE_SSE | XSTATE_YMM) & host_xcr0;
> >> + return mask & KVM_SUPPORTED_XCR0 & host_xcr0;
> >> }
> >>
> >> #define F(x) bit(X86_FEATURE_##x)
> >> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> >> index 3625798..801a882 100644
> >> --- a/arch/x86/kvm/x86.c
> >> +++ b/arch/x86/kvm/x86.c
> >> @@ -586,6 +586,8 @@ int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
> >> return 1;
> >> if ((xcr0 & XSTATE_YMM) && !(xcr0 & XSTATE_SSE))
> >> return 1;
> >> + if (xcr0 & ~KVM_SUPPORTED_XCR0)
> >> + return 1;
> > Shouldn't we check guest's cpuid here?
>
> Yes.
>
> >> if (xcr0 & ~host_xcr0)
> >> return 1;
> >> kvm_put_guest_xcr0(vcpu);
> >> @@ -2980,10 +2982,14 @@ static int kvm_vcpu_ioctl_x86_set_xsave(struct kvm_vcpu *vcpu,
> >> u64 xstate_bv =
> >> *(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)];
> >>
> >> - if (cpu_has_xsave)
> >> + if (cpu_has_xsave) {
> >> + if (xstate_bv & ~KVM_SUPPORTED_XCR0)
> >> + return -EINVAL;
> >> + if (xstate_bv & ~host_xcr0)
> >> + return -EINVAL;
> > And here?
>
> Here it'd be nice, but we cannot due to backwards compatibility. We
> agreed to change KVM_GET_XSAVE and make it only report states that are
> included in CPUID, but we still need to do migration from old kernels
> that report all states.
>
> If we change KVM_SET_XSAVE to look at CPUID, and the CPUID does not
> include AVX, migration will fail from old kernel (KVM_GET_XSAVE reports
> all states) to new kernel (KVM_SET_XSAVES checks against CPUID).
>
Yeah :(
--
Gleb.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
