From a58d9166a756a0f4a6618e4f593232593d6df134 Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Wed, 31 Mar 2021 06:24:43 -0400 Subject: [PATCH 1/2] KVM: SVM: load control fields from VMCB12 before checking them Avoid races between check and use of the nested VMCB controls. This for example ensures that the VMRUN intercept is always reflected to the nested hypervisor, instead of being processed by the host. Without this patch, it is possible to end up with svm->nested.hsave pointing to the MSR permission bitmap for nested guests. This bug is CVE-2021-29657. Reported-by: Felix Wilhelm Cc: stable@vger.kernel.org Fixes: 2fcf4876ada ("KVM: nSVM: implement on demand allocation of the nested state") Signed-off-by: Paolo Bonzini --- arch/x86/kvm/svm/nested.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c index 9e4c226dbf7d..061a00f91af5 100644 --- a/arch/x86/kvm/svm/nested.c +++ b/arch/x86/kvm/svm/nested.c @@ -225,7 +225,7 @@ static bool nested_vmcb_check_controls(struct vmcb_control_area *control) return true; } -static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12) +static bool nested_vmcb_check_save(struct vcpu_svm *svm, struct vmcb *vmcb12) { bool vmcb12_lma; @@ -257,7 +257,7 @@ static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12) if (kvm_valid_cr4(&svm->vcpu, vmcb12->save.cr4)) return false; - return nested_vmcb_check_controls(&vmcb12->control); + return true; } static void load_nested_vmcb_control(struct vcpu_svm *svm, @@ -440,7 +440,6 @@ int enter_svm_guest_mode(struct vcpu_svm *svm, u64 vmcb12_gpa, int ret; svm->nested.vmcb12_gpa = vmcb12_gpa; - load_nested_vmcb_control(svm, &vmcb12->control); nested_prepare_vmcb_save(svm, vmcb12); nested_prepare_vmcb_control(svm); @@ -484,7 +483,10 @@ int nested_svm_vmrun(struct vcpu_svm *svm) if (WARN_ON_ONCE(!svm->nested.initialized)) return -EINVAL; - if (!nested_vmcb_checks(svm, vmcb12)) { + load_nested_vmcb_control(svm, &vmcb12->control); + + if (!nested_vmcb_check_save(svm, vmcb12) || + !nested_vmcb_check_controls(&svm->nested.ctl)) { vmcb12->control.exit_code = SVM_EXIT_ERR; vmcb12->control.exit_code_hi = 0; vmcb12->control.exit_info_1 = 0; From 3c346c0c60ab06a021d1c0884a0ef494bc4ee3a7 Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Wed, 31 Mar 2021 06:28:01 -0400 Subject: [PATCH 2/2] KVM: SVM: ensure that EFER.SVME is set when running nested guest or on nested vmexit Fixing nested_vmcb_check_save to avoid all TOC/TOU races is a bit harder in released kernels, so do the bare minimum by avoiding that EFER.SVME is cleared. This is problematic because svm_set_efer frees the data structures for nested virtualization if EFER.SVME is cleared. Also check that EFER.SVME remains set after a nested vmexit; clearing it could happen if the bit is zero in the save area that is passed to KVM_SET_NESTED_STATE (the save area of the nested state corresponds to the nested hypervisor's state and is restored on the next nested vmexit). Cc: stable@vger.kernel.org Fixes: 2fcf4876ada ("KVM: nSVM: implement on demand allocation of the nested state") Signed-off-by: Paolo Bonzini --- arch/x86/kvm/svm/nested.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c index 061a00f91af5..c9e7b86350d6 100644 --- a/arch/x86/kvm/svm/nested.c +++ b/arch/x86/kvm/svm/nested.c @@ -229,6 +229,13 @@ static bool nested_vmcb_check_save(struct vcpu_svm *svm, struct vmcb *vmcb12) { bool vmcb12_lma; + /* + * FIXME: these should be done after copying the fields, + * to avoid TOC/TOU races. For these save area checks + * the possible damage is limited since kvm_set_cr0 and + * kvm_set_cr4 handle failure; EFER_SVME is an exception + * so it is force-set later in nested_prepare_vmcb_save. + */ if ((vmcb12->save.efer & EFER_SVME) == 0) return false; @@ -382,7 +389,14 @@ static void nested_prepare_vmcb_save(struct vcpu_svm *svm, struct vmcb *vmcb12) svm->vmcb->save.gdtr = vmcb12->save.gdtr; svm->vmcb->save.idtr = vmcb12->save.idtr; kvm_set_rflags(&svm->vcpu, vmcb12->save.rflags); - svm_set_efer(&svm->vcpu, vmcb12->save.efer); + + /* + * Force-set EFER_SVME even though it is checked earlier on the + * VMCB12, because the guest can flip the bit between the check + * and now. Clearing EFER_SVME would call svm_free_nested. + */ + svm_set_efer(&svm->vcpu, vmcb12->save.efer | EFER_SVME); + svm_set_cr0(&svm->vcpu, vmcb12->save.cr0); svm_set_cr4(&svm->vcpu, vmcb12->save.cr4); svm->vmcb->save.cr2 = svm->vcpu.arch.cr2 = vmcb12->save.cr2; @@ -1188,6 +1202,8 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu, */ if (!(save->cr0 & X86_CR0_PG)) goto out_free; + if (!(save->efer & EFER_SVME)) + goto out_free; /* * All checks done, we can enter guest mode. L1 control fields