2021-08-18 03:40:36

by Hou Wenlong

Subject: [PATCH] kvm: fix wrong exception emulation in check_rdtsc

According to Intel's SDM Vol2 and AMD's APM Vol3, when
CR4.TSD is set, use rdtsc/rdtscp instruction above privilege
level 0 should trigger a #GP.

Fixes: d7eb82030699e ("KVM: SVM: Add intercept checks for remaining group7 instructions")
Signed-off-by: Hou Wenlong <[email protected]>
---
arch/x86/kvm/emulate.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index 2837110e66ed..c589ac832265 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -4206,7 +4206,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
u64 cr4 = ctxt->ops->get_cr(ctxt, 4);

if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
- return emulate_ud(ctxt);
+ return emulate_gp(ctxt, 0);

return X86EMUL_CONTINUE;
}
--
2.31.1

2021-08-18 21:10:29

by Jim Mattson

Subject: Re: [PATCH] kvm: fix wrong exception emulation in check_rdtsc

On Tue, Aug 17, 2021 at 8:36 PM Hou Wenlong
<[email protected]> wrote:
>
> According to Intel's SDM Vol2 and AMD's APM Vol3, when
> CR4.TSD is set, use rdtsc/rdtscp instruction above privilege
> level 0 should trigger a #GP.
>
> Fixes: d7eb82030699e ("KVM: SVM: Add intercept checks for remaining group7 instructions")
> Signed-off-by: Hou Wenlong <[email protected]>
Reviewed-by: Jim Mattson <[email protected]>

2021-08-18 21:45:20

by Sean Christopherson

Subject: Re: [PATCH] kvm: fix wrong exception emulation in check_rdtsc

On Wed, Aug 18, 2021, Hou Wenlong wrote:
> According to Intel's SDM Vol2 and AMD's APM Vol3, when
> CR4.TSD is set, use rdtsc/rdtscp instruction above privilege
> level 0 should trigger a #GP.
>
> Fixes: d7eb82030699e ("KVM: SVM: Add intercept checks for remaining group7 instructions")
> Signed-off-by: Hou Wenlong <[email protected]>
> ---
> arch/x86/kvm/emulate.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
> index 2837110e66ed..c589ac832265 100644
> --- a/arch/x86/kvm/emulate.c
> +++ b/arch/x86/kvm/emulate.c
> @@ -4206,7 +4206,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
> u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
>
> if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
> - return emulate_ud(ctxt);
> + return emulate_gp(ctxt, 0);

Heh, I was having some serious deja vu, but the fix I was thinking of was
for em_rdpid, and that was changing #GP -> #UD (commit a9e2e0ae6860).

Reviewed-by: Sean Christopherson <[email protected]>

> return X86EMUL_CONTINUE;
> }
> --
> 2.31.1
>

2021-09-21 19:30:19

by Paolo Bonzini

Subject: Re: [PATCH] kvm: fix wrong exception emulation in check_rdtsc

On 18/08/21 05:36, Hou Wenlong wrote:
> According to Intel's SDM Vol2 and AMD's APM Vol3, when
> CR4.TSD is set, use rdtsc/rdtscp instruction above privilege
> level 0 should trigger a #GP.
>
> Fixes: d7eb82030699e ("KVM: SVM: Add intercept checks for remaining group7 instructions")
> Signed-off-by: Hou Wenlong <[email protected]>
> ---
> arch/x86/kvm/emulate.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
> index 2837110e66ed..c589ac832265 100644
> --- a/arch/x86/kvm/emulate.c
> +++ b/arch/x86/kvm/emulate.c
> @@ -4206,7 +4206,7 @@ static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
> u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
>
> if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
> - return emulate_ud(ctxt);
> + return emulate_gp(ctxt, 0);
>
> return X86EMUL_CONTINUE;
> }
>

Queued, thanks.

Paolo