* [PATCH v2 1/2] KVM: x86: Move nested_run_pending to kvm_vcpu_arch
2026-03-12 23:48 [PATCH v2 0/2] KVM: x86: Syzkaller nested_run_pending defense Sean Christopherson
@ 2026-03-12 23:48 ` Sean Christopherson
2026-03-12 23:48 ` [PATCH v2 2/2] KVM: x86: Suppress WARNs on nested_run_pending after userspace exit Sean Christopherson
2026-04-08 0:14 ` [PATCH v2 0/2] KVM: x86: Syzkaller nested_run_pending defense Sean Christopherson
2 siblings, 0 replies; 4+ messages in thread
From: Sean Christopherson @ 2026-03-12 23:48 UTC (permalink / raw)
To: Sean Christopherson, Paolo Bonzini; +Cc: kvm, linux-kernel, Yosry Ahmed
From: Yosry Ahmed <yosry@kernel.org>
Move nested_run_pending field present in both svm_nested_state and
nested_vmx to the common kvm_vcpu_arch. This allows for common code to
use without plumbing it through per-vendor helpers.
nested_run_pending remains zero-initialized, as the entire kvm_vcpu
struct is, and all further accesses are done through vcpu->arch instead
of svm->nested or vmx->nested.
No functional change intended.
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Yosry Ahmed <yosry@kernel.org>
[sean: expand the commend in the field declaration]
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
arch/x86/include/asm/kvm_host.h | 9 +++++++
arch/x86/kvm/svm/nested.c | 18 ++++++-------
arch/x86/kvm/svm/svm.c | 16 ++++++------
arch/x86/kvm/svm/svm.h | 4 ---
arch/x86/kvm/vmx/nested.c | 46 ++++++++++++++++-----------------
arch/x86/kvm/vmx/vmx.c | 16 ++++++------
arch/x86/kvm/vmx/vmx.h | 3 ---
7 files changed, 57 insertions(+), 55 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index d3bdc9828133..45171b607cf2 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1099,6 +1099,15 @@ struct kvm_vcpu_arch {
*/
bool pdptrs_from_userspace;
+ /*
+ * Set if an emulated nested VM-Enter to L2 is pending completion. KVM
+ * must not synthesize a VM-Exit to L1 before entering L2, as VM-Exits
+ * can only occur at instruction boundaries. The only exception is
+ * VMX's "notify" exits, which exist in large part to break the CPU out
+ * of infinite ucode loops, but can corrupt vCPU state in the process!
+ */
+ bool nested_run_pending;
+
#if IS_ENABLED(CONFIG_HYPERV)
hpa_t hv_root_tdp;
#endif
diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index 5ff01d2ac85e..1b0e0336ef11 100644
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -925,7 +925,7 @@ static void nested_vmcb02_prepare_control(struct vcpu_svm *svm)
* the CPU and/or KVM and should be used regardless of L1's support.
*/
if (guest_cpu_cap_has(vcpu, X86_FEATURE_NRIPS) ||
- !svm->nested.nested_run_pending)
+ !vcpu->arch.nested_run_pending)
vmcb02->control.next_rip = vmcb12_ctrl->next_rip;
svm->nmi_l1_to_l2 = is_evtinj_nmi(vmcb02->control.event_inj);
@@ -937,7 +937,7 @@ static void nested_vmcb02_prepare_control(struct vcpu_svm *svm)
if (is_evtinj_soft(vmcb02->control.event_inj)) {
svm->soft_int_injected = true;
if (guest_cpu_cap_has(vcpu, X86_FEATURE_NRIPS) ||
- !svm->nested.nested_run_pending)
+ !vcpu->arch.nested_run_pending)
svm->soft_int_next_rip = vmcb12_ctrl->next_rip;
}
@@ -1142,11 +1142,11 @@ int nested_svm_vmrun(struct kvm_vcpu *vcpu)
if (!npt_enabled)
vmcb01->save.cr3 = kvm_read_cr3(vcpu);
- svm->nested.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = 1;
if (enter_svm_guest_mode(vcpu, vmcb12_gpa, true) ||
!nested_svm_merge_msrpm(vcpu)) {
- svm->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
svm->nmi_l1_to_l2 = false;
svm->soft_int_injected = false;
@@ -1288,7 +1288,7 @@ void nested_svm_vmexit(struct vcpu_svm *svm)
/* Exit Guest-Mode */
leave_guest_mode(vcpu);
svm->nested.vmcb12_gpa = 0;
- WARN_ON_ONCE(svm->nested.nested_run_pending);
+ WARN_ON_ONCE(vcpu->arch.nested_run_pending);
kvm_clear_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu);
@@ -1498,7 +1498,7 @@ void svm_leave_nested(struct kvm_vcpu *vcpu)
struct vcpu_svm *svm = to_svm(vcpu);
if (is_guest_mode(vcpu)) {
- svm->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
svm->nested.vmcb12_gpa = INVALID_GPA;
leave_guest_mode(vcpu);
@@ -1683,7 +1683,7 @@ static int svm_check_nested_events(struct kvm_vcpu *vcpu)
* previously injected event, the pending exception occurred while said
* event was being delivered and thus needs to be handled.
*/
- bool block_nested_exceptions = svm->nested.nested_run_pending;
+ bool block_nested_exceptions = vcpu->arch.nested_run_pending;
/*
* New events (not exceptions) are only recognized at instruction
* boundaries. If an event needs reinjection, then KVM is handling a
@@ -1858,7 +1858,7 @@ static int svm_get_nested_state(struct kvm_vcpu *vcpu,
kvm_state.size += KVM_STATE_NESTED_SVM_VMCB_SIZE;
kvm_state.flags |= KVM_STATE_NESTED_GUEST_MODE;
- if (svm->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
kvm_state.flags |= KVM_STATE_NESTED_RUN_PENDING;
}
@@ -1995,7 +1995,7 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
svm_set_gif(svm, !!(kvm_state->flags & KVM_STATE_NESTED_GIF_SET));
- svm->nested.nested_run_pending =
+ vcpu->arch.nested_run_pending =
!!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
svm->nested.vmcb12_gpa = kvm_state->hdr.svm.vmcb_pa;
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index d98fbc0e58e8..ece115d47044 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -3821,7 +3821,7 @@ static void svm_fixup_nested_rips(struct kvm_vcpu *vcpu)
{
struct vcpu_svm *svm = to_svm(vcpu);
- if (!is_guest_mode(vcpu) || !svm->nested.nested_run_pending)
+ if (!is_guest_mode(vcpu) || !vcpu->arch.nested_run_pending)
return;
/*
@@ -3969,7 +3969,7 @@ bool svm_nmi_blocked(struct kvm_vcpu *vcpu)
static int svm_nmi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
struct vcpu_svm *svm = to_svm(vcpu);
- if (svm->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
if (svm_nmi_blocked(vcpu))
@@ -4011,7 +4011,7 @@ static int svm_interrupt_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
struct vcpu_svm *svm = to_svm(vcpu);
- if (svm->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
if (svm_interrupt_blocked(vcpu))
@@ -4234,7 +4234,7 @@ static void svm_complete_soft_interrupt(struct kvm_vcpu *vcpu, u8 vector,
* the soft int and will reinject it via the standard injection flow,
* and so KVM needs to grab the state from the pending nested VMRUN.
*/
- if (is_guest_mode(vcpu) && svm->nested.nested_run_pending)
+ if (is_guest_mode(vcpu) && vcpu->arch.nested_run_pending)
svm_set_nested_run_soft_int_state(vcpu);
/*
@@ -4537,11 +4537,11 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu, u64 run_flags)
nested_sync_control_from_vmcb02(svm);
/* Track VMRUNs that have made past consistency checking */
- if (svm->nested.nested_run_pending &&
+ if (vcpu->arch.nested_run_pending &&
!svm_is_vmrun_failure(svm->vmcb->control.exit_code))
++vcpu->stat.nested_run;
- svm->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
}
svm->vmcb->control.tlb_ctl = TLB_CONTROL_DO_NOTHING;
@@ -4910,7 +4910,7 @@ bool svm_smi_blocked(struct kvm_vcpu *vcpu)
static int svm_smi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
struct vcpu_svm *svm = to_svm(vcpu);
- if (svm->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
if (svm_smi_blocked(vcpu))
@@ -5028,7 +5028,7 @@ static int svm_leave_smm(struct kvm_vcpu *vcpu, const union kvm_smram *smram)
goto unmap_save;
ret = 0;
- svm->nested.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = 1;
unmap_save:
kvm_vcpu_unmap(vcpu, &map_save);
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index ff1e4b4dc998..d3186956ec4b 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -215,10 +215,6 @@ struct svm_nested_state {
*/
void *msrpm;
- /* A VMRUN has started but has not yet been performed, so
- * we cannot inject a nested vmexit yet. */
- bool nested_run_pending;
-
/* cache for control fields of the guest */
struct vmcb_ctrl_area_cached ctl;
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index 937aeb474af7..f1543a6ad524 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -2273,7 +2273,7 @@ static void vmx_start_preemption_timer(struct kvm_vcpu *vcpu,
static u64 nested_vmx_calc_efer(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
{
- if (vmx->nested.nested_run_pending &&
+ if (vmx->vcpu.arch.nested_run_pending &&
(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_EFER))
return vmcs12->guest_ia32_efer;
else if (vmcs12->vm_entry_controls & VM_ENTRY_IA32E_MODE)
@@ -2513,7 +2513,7 @@ static void prepare_vmcs02_early(struct vcpu_vmx *vmx, struct loaded_vmcs *vmcs0
/*
* Interrupt/Exception Fields
*/
- if (vmx->nested.nested_run_pending) {
+ if (vmx->vcpu.arch.nested_run_pending) {
vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
vmcs12->vm_entry_intr_info_field);
vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE,
@@ -2621,7 +2621,7 @@ static void prepare_vmcs02_rare(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
vmcs_write64(GUEST_PDPTR3, vmcs12->guest_pdptr3);
}
- if (kvm_mpx_supported() && vmx->nested.nested_run_pending &&
+ if (kvm_mpx_supported() && vmx->vcpu.arch.nested_run_pending &&
(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS))
vmcs_write64(GUEST_BNDCFGS, vmcs12->guest_bndcfgs);
}
@@ -2718,7 +2718,7 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
!(evmcs->hv_clean_fields & HV_VMX_ENLIGHTENED_CLEAN_FIELD_GUEST_GRP1);
}
- if (vmx->nested.nested_run_pending &&
+ if (vcpu->arch.nested_run_pending &&
(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS)) {
kvm_set_dr(vcpu, 7, vmcs12->guest_dr7);
vmx_guest_debugctl_write(vcpu, vmcs12->guest_ia32_debugctl &
@@ -2728,13 +2728,13 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
vmx_guest_debugctl_write(vcpu, vmx->nested.pre_vmenter_debugctl);
}
- if (!vmx->nested.nested_run_pending ||
+ if (!vcpu->arch.nested_run_pending ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_CET_STATE))
vmcs_write_cet_state(vcpu, vmx->nested.pre_vmenter_s_cet,
vmx->nested.pre_vmenter_ssp,
vmx->nested.pre_vmenter_ssp_tbl);
- if (kvm_mpx_supported() && (!vmx->nested.nested_run_pending ||
+ if (kvm_mpx_supported() && (!vcpu->arch.nested_run_pending ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS)))
vmcs_write64(GUEST_BNDCFGS, vmx->nested.pre_vmenter_bndcfgs);
vmx_set_rflags(vcpu, vmcs12->guest_rflags);
@@ -2747,7 +2747,7 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
vcpu->arch.cr0_guest_owned_bits &= ~vmcs12->cr0_guest_host_mask;
vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits);
- if (vmx->nested.nested_run_pending &&
+ if (vcpu->arch.nested_run_pending &&
(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT)) {
vmcs_write64(GUEST_IA32_PAT, vmcs12->guest_ia32_pat);
vcpu->arch.pat = vmcs12->guest_ia32_pat;
@@ -3349,7 +3349,7 @@ static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu,
* to bit 8 (LME) if bit 31 in the CR0 field (corresponding to
* CR0.PG) is 1.
*/
- if (to_vmx(vcpu)->nested.nested_run_pending &&
+ if (vcpu->arch.nested_run_pending &&
(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_EFER)) {
if (CC(!kvm_valid_efer(vcpu, vmcs12->guest_ia32_efer)) ||
CC(ia32e != !!(vmcs12->guest_ia32_efer & EFER_LMA)) ||
@@ -3627,15 +3627,15 @@ enum nvmx_vmentry_status nested_vmx_enter_non_root_mode(struct kvm_vcpu *vcpu,
kvm_service_local_tlb_flush_requests(vcpu);
- if (!vmx->nested.nested_run_pending ||
+ if (!vcpu->arch.nested_run_pending ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS))
vmx->nested.pre_vmenter_debugctl = vmx_guest_debugctl_read();
if (kvm_mpx_supported() &&
- (!vmx->nested.nested_run_pending ||
+ (!vcpu->arch.nested_run_pending ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_BNDCFGS)))
vmx->nested.pre_vmenter_bndcfgs = vmcs_read64(GUEST_BNDCFGS);
- if (!vmx->nested.nested_run_pending ||
+ if (!vcpu->arch.nested_run_pending ||
!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_CET_STATE))
vmcs_read_cet_state(vcpu, &vmx->nested.pre_vmenter_s_cet,
&vmx->nested.pre_vmenter_ssp,
@@ -3844,7 +3844,7 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
* We're finally done with prerequisite checking, and can start with
* the nested entry.
*/
- vmx->nested.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = 1;
vmx->nested.has_preemption_timer_deadline = false;
status = nested_vmx_enter_non_root_mode(vcpu, true);
if (unlikely(status != NVMX_VMENTRY_SUCCESS))
@@ -3876,12 +3876,12 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
!nested_cpu_has(vmcs12, CPU_BASED_NMI_WINDOW_EXITING) &&
!(nested_cpu_has(vmcs12, CPU_BASED_INTR_WINDOW_EXITING) &&
(vmcs12->guest_rflags & X86_EFLAGS_IF))) {
- vmx->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
return kvm_emulate_halt_noskip(vcpu);
}
break;
case GUEST_ACTIVITY_WAIT_SIPI:
- vmx->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
kvm_set_mp_state(vcpu, KVM_MP_STATE_INIT_RECEIVED);
break;
default:
@@ -3891,7 +3891,7 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
return 1;
vmentry_failed:
- vmx->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
if (status == NVMX_VMENTRY_KVM_INTERNAL_ERROR)
return 0;
if (status == NVMX_VMENTRY_VMEXIT)
@@ -4288,7 +4288,7 @@ static int vmx_check_nested_events(struct kvm_vcpu *vcpu)
* previously injected event, the pending exception occurred while said
* event was being delivered and thus needs to be handled.
*/
- bool block_nested_exceptions = vmx->nested.nested_run_pending;
+ bool block_nested_exceptions = vcpu->arch.nested_run_pending;
/*
* Events that don't require injection, i.e. that are virtualized by
* hardware, aren't blocked by a pending VM-Enter as KVM doesn't need
@@ -4657,7 +4657,7 @@ static void sync_vmcs02_to_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
if (nested_cpu_has_preemption_timer(vmcs12) &&
vmcs12->vm_exit_controls & VM_EXIT_SAVE_VMX_PREEMPTION_TIMER &&
- !vmx->nested.nested_run_pending)
+ !vcpu->arch.nested_run_pending)
vmcs12->vmx_preemption_timer_value =
vmx_get_preemption_timer_value(vcpu);
@@ -5056,7 +5056,7 @@ void __nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 vm_exit_reason,
vmx->nested.mtf_pending = false;
/* trying to cancel vmlaunch/vmresume is a bug */
- WARN_ON_ONCE(vmx->nested.nested_run_pending);
+ WARN_ON_ONCE(vcpu->arch.nested_run_pending);
#ifdef CONFIG_KVM_HYPERV
if (kvm_check_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu)) {
@@ -6679,7 +6679,7 @@ bool nested_vmx_reflect_vmexit(struct kvm_vcpu *vcpu)
unsigned long exit_qual;
u32 exit_intr_info;
- WARN_ON_ONCE(vmx->nested.nested_run_pending);
+ WARN_ON_ONCE(vcpu->arch.nested_run_pending);
/*
* Late nested VM-Fail shares the same flow as nested VM-Exit since KVM
@@ -6775,7 +6775,7 @@ static int vmx_get_nested_state(struct kvm_vcpu *vcpu,
if (is_guest_mode(vcpu)) {
kvm_state.flags |= KVM_STATE_NESTED_GUEST_MODE;
- if (vmx->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
kvm_state.flags |= KVM_STATE_NESTED_RUN_PENDING;
if (vmx->nested.mtf_pending)
@@ -6850,7 +6850,7 @@ static int vmx_get_nested_state(struct kvm_vcpu *vcpu,
void vmx_leave_nested(struct kvm_vcpu *vcpu)
{
if (is_guest_mode(vcpu)) {
- to_vmx(vcpu)->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
nested_vmx_vmexit(vcpu, -1, 0, 0);
}
free_nested(vcpu);
@@ -7008,7 +7008,7 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
if (!(kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE))
return 0;
- vmx->nested.nested_run_pending =
+ vcpu->arch.nested_run_pending =
!!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
vmx->nested.mtf_pending =
@@ -7054,7 +7054,7 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
return 0;
error_guest_mode:
- vmx->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
return ret;
}
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index b15b4662b653..6e4b12a5849c 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -5170,7 +5170,7 @@ bool vmx_nmi_blocked(struct kvm_vcpu *vcpu)
int vmx_nmi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
- if (to_vmx(vcpu)->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
/* An NMI must not be injected into L2 if it's supposed to VM-Exit. */
@@ -5197,7 +5197,7 @@ bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu)
int vmx_interrupt_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
- if (to_vmx(vcpu)->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
/*
@@ -6009,7 +6009,7 @@ static bool vmx_unhandleable_emulation_required(struct kvm_vcpu *vcpu)
* only reachable if userspace modifies L2 guest state after KVM has
* performed the nested VM-Enter consistency checks.
*/
- if (vmx->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return true;
/*
@@ -6693,7 +6693,7 @@ static int __vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
* invalid guest state should never happen as that means KVM knowingly
* allowed a nested VM-Enter with an invalid vmcs12. More below.
*/
- if (KVM_BUG_ON(vmx->nested.nested_run_pending, vcpu->kvm))
+ if (KVM_BUG_ON(vcpu->arch.nested_run_pending, vcpu->kvm))
return -EIO;
if (is_guest_mode(vcpu)) {
@@ -7621,11 +7621,11 @@ fastpath_t vmx_vcpu_run(struct kvm_vcpu *vcpu, u64 run_flags)
* Track VMLAUNCH/VMRESUME that have made past guest state
* checking.
*/
- if (vmx->nested.nested_run_pending &&
+ if (vcpu->arch.nested_run_pending &&
!vmx_get_exit_reason(vcpu).failed_vmentry)
++vcpu->stat.nested_run;
- vmx->nested.nested_run_pending = 0;
+ vcpu->arch.nested_run_pending = 0;
}
if (unlikely(vmx->fail))
@@ -8382,7 +8382,7 @@ void vmx_setup_mce(struct kvm_vcpu *vcpu)
int vmx_smi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
{
/* we need a nested vmexit to enter SMM, postpone if run is pending */
- if (to_vmx(vcpu)->nested.nested_run_pending)
+ if (vcpu->arch.nested_run_pending)
return -EBUSY;
return !is_smm(vcpu);
}
@@ -8427,7 +8427,7 @@ int vmx_leave_smm(struct kvm_vcpu *vcpu, const union kvm_smram *smram)
if (ret != NVMX_VMENTRY_SUCCESS)
return 1;
- vmx->nested.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = 1;
vmx->nested.smm.guest_mode = false;
}
return 0;
diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
index 70bfe81dea54..db84e8001da5 100644
--- a/arch/x86/kvm/vmx/vmx.h
+++ b/arch/x86/kvm/vmx/vmx.h
@@ -138,9 +138,6 @@ struct nested_vmx {
*/
bool enlightened_vmcs_enabled;
- /* L2 must run next, and mustn't decide to exit to L1. */
- bool nested_run_pending;
-
/* Pending MTF VM-exit into L1. */
bool mtf_pending;
--
2.53.0.851.ga537e3e6e9-goog
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH v2 2/2] KVM: x86: Suppress WARNs on nested_run_pending after userspace exit
2026-03-12 23:48 [PATCH v2 0/2] KVM: x86: Syzkaller nested_run_pending defense Sean Christopherson
2026-03-12 23:48 ` [PATCH v2 1/2] KVM: x86: Move nested_run_pending to kvm_vcpu_arch Sean Christopherson
@ 2026-03-12 23:48 ` Sean Christopherson
2026-04-08 0:14 ` [PATCH v2 0/2] KVM: x86: Syzkaller nested_run_pending defense Sean Christopherson
2 siblings, 0 replies; 4+ messages in thread
From: Sean Christopherson @ 2026-03-12 23:48 UTC (permalink / raw)
To: Sean Christopherson, Paolo Bonzini; +Cc: kvm, linux-kernel, Yosry Ahmed
To end an ongoing game of whack-a-mole between KVM and syzkaller, WARN on
illegally cancelling a pending nested VM-Enter if and only if userspace
has NOT gained control of the vCPU since the nested run was initiated. As
proven time and time again by syzkaller, userspace can clobber vCPU state
so as to force a VM-Exit that violates KVM's architectural modelling of
VMRUN/VMLAUNCH/VMRESUME.
To detect that userspace has gained control, while minimizing the risk of
operating on stale data, convert nested_run_pending from a pure boolean to
a tri-state of sorts, where '0' is still "not pending", '1' is "pending",
and '2' is "pending but untrusted". Then on KVM_RUN, if the flag is in
the "trusted pending" state, move it to "untrusted pending".
Note, moving the state to "untrusted" even if KVM_RUN is ultimately
rejected is a-ok, because for the "untrusted" state to matter, KVM must
get past kvm_x86_vcpu_pre_run() at some point for the vCPU.
Reviewed-by: Yosry Ahmed <yosry@kernel.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
arch/x86/include/asm/kvm_host.h | 8 +++++++-
arch/x86/kvm/svm/nested.c | 11 +++++++----
arch/x86/kvm/svm/svm.c | 2 +-
arch/x86/kvm/vmx/nested.c | 12 +++++++-----
arch/x86/kvm/vmx/vmx.c | 2 +-
arch/x86/kvm/x86.c | 7 +++++++
arch/x86/kvm/x86.h | 10 ++++++++++
7 files changed, 40 insertions(+), 12 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 45171b607cf2..c470e40a00aa 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1105,8 +1105,14 @@ struct kvm_vcpu_arch {
* can only occur at instruction boundaries. The only exception is
* VMX's "notify" exits, which exist in large part to break the CPU out
* of infinite ucode loops, but can corrupt vCPU state in the process!
+ *
+ * For all intents and purposes, this is a boolean, but it's tracked as
+ * a u8 so that KVM can detect when userspace may have stuffed vCPU
+ * state and generated an architecturally-impossible VM-Exit.
*/
- bool nested_run_pending;
+#define KVM_NESTED_RUN_PENDING 1
+#define KVM_NESTED_RUN_PENDING_UNTRUSTED 2
+ u8 nested_run_pending;
#if IS_ENABLED(CONFIG_HYPERV)
hpa_t hv_root_tdp;
diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index 1b0e0336ef11..3b6d75787f9f 100644
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -1142,7 +1142,7 @@ int nested_svm_vmrun(struct kvm_vcpu *vcpu)
if (!npt_enabled)
vmcb01->save.cr3 = kvm_read_cr3(vcpu);
- vcpu->arch.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING;
if (enter_svm_guest_mode(vcpu, vmcb12_gpa, true) ||
!nested_svm_merge_msrpm(vcpu)) {
@@ -1288,7 +1288,8 @@ void nested_svm_vmexit(struct vcpu_svm *svm)
/* Exit Guest-Mode */
leave_guest_mode(vcpu);
svm->nested.vmcb12_gpa = 0;
- WARN_ON_ONCE(vcpu->arch.nested_run_pending);
+
+ kvm_warn_on_nested_run_pending(vcpu);
kvm_clear_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu);
@@ -1995,8 +1996,10 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
svm_set_gif(svm, !!(kvm_state->flags & KVM_STATE_NESTED_GIF_SET));
- vcpu->arch.nested_run_pending =
- !!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
+ if (kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING)
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING_UNTRUSTED;
+ else
+ vcpu->arch.nested_run_pending = 0;
svm->nested.vmcb12_gpa = kvm_state->hdr.svm.vmcb_pa;
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index ece115d47044..b4c27465bd53 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -5028,7 +5028,7 @@ static int svm_leave_smm(struct kvm_vcpu *vcpu, const union kvm_smram *smram)
goto unmap_save;
ret = 0;
- vcpu->arch.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING;
unmap_save:
kvm_vcpu_unmap(vcpu, &map_save);
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index f1543a6ad524..3fe88f29be7a 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -3844,7 +3844,7 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
* We're finally done with prerequisite checking, and can start with
* the nested entry.
*/
- vcpu->arch.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING;
vmx->nested.has_preemption_timer_deadline = false;
status = nested_vmx_enter_non_root_mode(vcpu, true);
if (unlikely(status != NVMX_VMENTRY_SUCCESS))
@@ -5056,7 +5056,7 @@ void __nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 vm_exit_reason,
vmx->nested.mtf_pending = false;
/* trying to cancel vmlaunch/vmresume is a bug */
- WARN_ON_ONCE(vcpu->arch.nested_run_pending);
+ kvm_warn_on_nested_run_pending(vcpu);
#ifdef CONFIG_KVM_HYPERV
if (kvm_check_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu)) {
@@ -6679,7 +6679,7 @@ bool nested_vmx_reflect_vmexit(struct kvm_vcpu *vcpu)
unsigned long exit_qual;
u32 exit_intr_info;
- WARN_ON_ONCE(vcpu->arch.nested_run_pending);
+ kvm_warn_on_nested_run_pending(vcpu);
/*
* Late nested VM-Fail shares the same flow as nested VM-Exit since KVM
@@ -7008,8 +7008,10 @@ static int vmx_set_nested_state(struct kvm_vcpu *vcpu,
if (!(kvm_state->flags & KVM_STATE_NESTED_GUEST_MODE))
return 0;
- vcpu->arch.nested_run_pending =
- !!(kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING);
+ if (kvm_state->flags & KVM_STATE_NESTED_RUN_PENDING)
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING_UNTRUSTED;
+ else
+ vcpu->arch.nested_run_pending = 0;
vmx->nested.mtf_pending =
!!(kvm_state->flags & KVM_STATE_NESTED_MTF_PENDING);
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 6e4b12a5849c..a29896a9ef14 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -8427,7 +8427,7 @@ int vmx_leave_smm(struct kvm_vcpu *vcpu, const union kvm_smram *smram)
if (ret != NVMX_VMENTRY_SUCCESS)
return 1;
- vcpu->arch.nested_run_pending = 1;
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING;
vmx->nested.smm.guest_mode = false;
}
return 0;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 0b5d48e75b65..eec0bde5d3e5 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -11978,6 +11978,13 @@ static void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
static int kvm_x86_vcpu_pre_run(struct kvm_vcpu *vcpu)
{
+ /*
+ * Userspace may have modified vCPU state, mark nested_run_pending as
+ * "untrusted" to avoid triggering false-positive WARNs.
+ */
+ if (vcpu->arch.nested_run_pending == KVM_NESTED_RUN_PENDING)
+ vcpu->arch.nested_run_pending = KVM_NESTED_RUN_PENDING_UNTRUSTED;
+
/*
* SIPI_RECEIVED is obsolete; KVM leaves the vCPU in Wait-For-SIPI and
* tracks the pending SIPI separately. SIPI_RECEIVED is still accepted
diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
index 44a28d343d40..38a905fa86de 100644
--- a/arch/x86/kvm/x86.h
+++ b/arch/x86/kvm/x86.h
@@ -188,6 +188,16 @@ static inline bool kvm_can_set_cpuid_and_feature_msrs(struct kvm_vcpu *vcpu)
return vcpu->arch.last_vmentry_cpu == -1 && !is_guest_mode(vcpu);
}
+/*
+ * WARN if a nested VM-Enter is pending completion, and userspace hasn't gained
+ * control since the nested VM-Enter was initiated (in which case, userspace
+ * may have modified vCPU state to induce an architecturally invalid VM-Exit).
+ */
+static inline void kvm_warn_on_nested_run_pending(struct kvm_vcpu *vcpu)
+{
+ WARN_ON_ONCE(vcpu->arch.nested_run_pending == KVM_NESTED_RUN_PENDING);
+}
+
static inline void kvm_set_mp_state(struct kvm_vcpu *vcpu, int mp_state)
{
vcpu->arch.mp_state = mp_state;
--
2.53.0.851.ga537e3e6e9-goog
^ permalink raw reply [flat|nested] 4+ messages in thread