On Sat, 2022-01-22 at 23:13 +0300, Denis Valeev wrote:
> The bug occurs on #GP triggered by VMware backdoor when eax value is
> unaligned. eax alignment check should not be applied to non-SVM
> instructions because it leads to incorrect omission of the instructions
> emulation.
> Apply the alignment check only to SVM instructions to fix.
>
> Fixes: d1cba6c92237 ("KVM: x86: nSVM: test eax for 4K alignment for GP errata workaround")
>
> Signed-off-by: Denis Valeev <lemniscattaden@xxxxxxxxx>
> ---
> This bug breaks nyx-fuzz (https://nyx-fuzz.com) that uses VMware backdoor
> as an alternative way for hypercall from guest user-mode. With this bug
> a hypercall interpreted as a GP and leads to process termination.
>
> arch/x86/kvm/svm/svm.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> index e64f16237b60..b5e4731080ef 100644
> --- a/arch/x86/kvm/svm/svm.c
> +++ b/arch/x86/kvm/svm/svm.c
> @@ -2233,10 +2233,6 @@ static int gp_interception(struct kvm_vcpu *vcpu)
> if (error_code)
> goto reinject;
>
> - /* All SVM instructions expect page aligned RAX */
> - if (svm->vmcb->save.rax & ~PAGE_MASK)
> - goto reinject;
> -
> /* Decode the instruction for usage later */
> if (x86_decode_emulated_instruction(vcpu, 0, NULL, 0) != EMULATION_OK)
> goto reinject;
> @@ -2254,8 +2250,13 @@ static int gp_interception(struct kvm_vcpu *vcpu)
> if (!is_guest_mode(vcpu))
> return kvm_emulate_instruction(vcpu,
> EMULTYPE_VMWARE_GP | EMULTYPE_NO_DECODE);
> - } else
> + } else {
> + /* All SVM instructions expect page aligned RAX */
> + if (svm->vmcb->save.rax & ~PAGE_MASK)
> + goto reinject;
> +
> return emulate_svm_instr(vcpu, opcode);
> + }
>
> reinject:
> kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
Oops.
Reviewed-by: Maxim Levitsky <mlevitsk@xxxxxxxxxx>
Thanks,
Best regards,
Maxim Levitsky
