Skip to contentSkip to conversation

KVM: x86: Exempt in-kernel PIC from "disappearing" interrupt warning

PATCH syzbot 4 participants 5 messages Jun 25 – Jul 7

Subsystems (3)
lore
Share & tools
lore

syzbot

author syzbot
#0 lore raw

From: Alexander Potapenko <[email protected]>

A warning can be triggered in kvm_check_and_inject_events() when an interrupt disappears between the time it is checked via kvm_cpu_has_injectable_intr() and the time it is fetched via kvm_cpu_get_interrupt(). This occurs because the warning incorrectly assumes that if an interrupt is injectable, fetching it must always return a valid interrupt vector (i.e., not -1).

However, this assumption is broken by level-triggered interrupts in the in-kernel PIC that are deasserted concurrently by another thread. For example, if a misconfigured PIT or a PCI device asserts and then immediately deasserts a level-triggered interrupt, the vCPU thread might see the pending interrupt during the check but find it gone during the fetch, resulting in kvm_cpu_get_interrupt() returning -1.

The warning manifests as follows:

------------[ cut here ]------------
irq == -1
WARNING: arch/x86/kvm/x86.c:10860 at kvm_check_and_inject_events
arch/x86/kvm/x86.c:10860 [inline]
WARNING: arch/x86/kvm/x86.c:10860 at vcpu_enter_guest
arch/x86/kvm/x86.c:11356 [inline]
WARNING: arch/x86/kvm/x86.c:10860 at vcpu_run+0x57ec/0x7950
arch/x86/kvm/x86.c:11770
RIP: 0010:kvm_check_and_inject_events arch/x86/kvm/x86.c:10860 [inline]
RIP: 0010:vcpu_enter_guest arch/x86/kvm/x86.c:11356 [inline]
RIP: 0010:vcpu_run+0x57ec/0x7950 arch/x86/kvm/x86.c:11770
Call Trace:
 <TASK>
 kvm_arch_vcpu_ioctl_run+0x1193/0x2070 arch/x86/kvm/x86.c:12125
 kvm_vcpu_ioctl+0xa61/0xfd0 virt/kvm/kvm_main.c:4470
 vfs_ioctl fs/ioctl.c:51 [inline]
 __do_sys_ioctl fs/ioctl.c:597 [inline]
 __se_sys_ioctl+0xfc/0x170 fs/ioctl.c:583
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0x174/0x580 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
 </TASK>

Since this is a legitimate Time-Of-Check to Time-Of-Use (TOCTOU) race condition for the in-kernel PIC, WARN_ON_ONCE() must not be used for this case. Update the warning to exempt the in-kernel PIC, while preserving it for other interrupt sources (e.g. APIC) as they are not expected to exhibit this behavior.

Fixes bf672720e83c ("KVM: x86: check the kvm_cpu_get_interrupt result before using it")
Assisted-by Gemini:gemini-3.1-pro-preview Gemini:gemini-3-flash-preview syzbot
Reported-by [email protected]

---
v2:
- Restrict the warning exemption to the in-kernel PIC case.
- Remove the pr_err_ratelimited() logging.
- Preserve the WARN_ON_ONCE() for non-PIC interrupt sources.

v1:
https://lore.kernel.org/all/[email protected]/T/
---

arch/x86/kvm/x86.c +3 -1
index 0550359ed..f1681aa9f 100644
@@ -10857,7 +10857,9 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
 		if (r) {
 			int irq = kvm_cpu_get_interrupt(vcpu);
 
-			if (!WARN_ON_ONCE(irq == -1)) {
+			WARN_ON_ONCE(irq == -1 && !pic_in_kernel(vcpu->kvm));
+
+			if (irq != -1) {
 				kvm_queue_interrupt(vcpu, irq, false);
 				kvm_x86_call(inject_irq)(vcpu, false);
 				WARN_ON(kvm_x86_call(interrupt_allowed)(vcpu, true) < 0);

base-commit: 8cd9520d35a6c38db6567e97dd93b1f11f185dc6

Signature · 4 lines
See https://goo.gle/syzbot-ai-patches for information about AI-generated patches.
You can comment on the patch as usual, syzbot will try to address
the comments and send a new version of the patch if necessary.
syzbot engineers can be reached at [email protected].
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

New issues:
- [Medium] The !pic_in_kernel(vcpu->kvm) check inadvertently suppresses the warning for APIC bugs in the default full-irqchip mode, contradicting the commit's intent.

Pre-existing issues:
- [High] A malicious guest can still trigger the WARN_ON_ONCE in kvm_check_and_inject_events when KVM is configured with a split irqchip and Xen emulation is enabled. --

commit a01ce83aaf4bff44e757aa3ac163f97e316b3056
Author: Alexander Potapenko <[email protected]>

KVM: x86: Exempt in-kernel PIC from "disappearing" interrupt warning

This commit addresses a time-of-check to time-of-use race condition in the in-kernel PIC that triggers a kernel warning when an interrupt disappears between check and fetch. It modifies kvm_check_and_inject_events() to exempt the in-kernel PIC from this warning. The intent is to preserve the warning for other interrupt sources like the APIC.

Quoted 3 lines
arch/x86/kvm/x86.c +0 -0

[ ... ]

arch/x86/kvm/x86.c static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu, · line 10860

Quoted 6 lines
arch/x86/kvm/x86.c +1 -1
@@ -10857,7 +10857,9 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
 		if (r) {
 			int irq = kvm_cpu_get_interrupt(vcpu);
 
-			if (!WARN_ON_ONCE(irq == -1)) {
+			WARN_ON_ONCE(irq == -1 && !pic_in_kernel(vcpu->kvm));

[Severity: Medium]
Does this inadvertently suppress the warning for APIC bugs in the default full-irqchip mode?

Since pic_in_kernel() evaluates to true when the in-kernel PIC is present, checking !pic_in_kernel() disables the warning entirely for all interrupt sources in that configuration. This appears to contradict the commit message intent of preserving the warning for other sources like the APIC.

[Severity: High]
This is a pre-existing issue, but can a malicious guest still trigger this warning when KVM is configured with a split irqchip and Xen emulation is enabled?

If a guest concurrently toggles its evtchn_upcall_pending flag in memory while the host executes the event injection loop, it seems this path is directly reachable. Because pic_in_kernel() is false for a split irqchip, the warning could still fire and potentially lead to a host panic if panic_on_warn is enabled.

Quoted 5 lines
arch/x86/kvm/x86.c +2 -0
+
+			if (irq != -1) {
 				kvm_queue_interrupt(vcpu, irq, false);
 				kvm_x86_call(inject_irq)(vcpu, false);
 				WARN_ON(kvm_x86_call(interrupt_allowed)(vcpu, true) < 0);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#2 lore raw
Quoted 20 lines · On Thu, Jun 25, 2026, [email protected] wrote:
Quoted 3 lines
arch/x86/kvm/x86.c +0 -0

[ ... ]

arch/x86/kvm/x86.c static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu, · line 10860

Quoted 6 lines
arch/x86/kvm/x86.c +1 -1
@@ -10857,7 +10857,9 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
 		if (r) {
 			int irq = kvm_cpu_get_interrupt(vcpu);
 
-			if (!WARN_ON_ONCE(irq == -1)) {
+			WARN_ON_ONCE(irq == -1 && !pic_in_kernel(vcpu->kvm));

[Severity: Medium]
Does this inadvertently suppress the warning for APIC bugs in the default full-irqchip mode?

Since pic_in_kernel() evaluates to true when the in-kernel PIC is present, checking !pic_in_kernel() disables the warning entirely for all interrupt sources in that configuration. This appears to contradict the commit message intent of preserving the warning for other sources like the APIC.

Quoted 4 lines

[Severity: High]
This is a pre-existing issue, but can a malicious guest still trigger this warning when KVM is configured with a split irqchip and Xen emulation is enabled?

Argh, yes.

Quoted 3 lines

If a guest concurrently toggles its evtchn_upcall_pending flag in memory while the host executes the event injection loop, it seems this path is directly reachable.

Even more coarsely, userspace could disable Xen event stuff entirely via KVM_XEN_HVM_CONFIG. The easiest fix would be to condition the WARN on !IS_ENABLED(CONFIG_KVM_XEN). We could track if the VM has ever used Xen emulation, but I don't think that's worth doing given that many deployments can and should disable Xen emulation.

The other annoying issue is that vmx_check_nested_events() has the same flaw. And I definitely don't want to toss that WARN either, because it recently found a real bug that actually matters.

I'll send a v2 in the near-ish future to weed out CONFIG_KVM_XEN, and fix the nVMX case as well, hopefully without having to duplicate too much code.

#3 lore raw
In reply to syzbot
Quoted 69 lines · syzbot' via syzkaller-bugs

From: Alexander Potapenko <[email protected]>

A warning can be triggered in kvm_check_and_inject_events() when an interrupt disappears between the time it is checked via kvm_cpu_has_injectable_intr() and the time it is fetched via kvm_cpu_get_interrupt(). This occurs because the warning incorrectly assumes that if an interrupt is injectable, fetching it must always return a valid interrupt vector (i.e., not -1).

However, this assumption is broken by level-triggered interrupts in the in-kernel PIC that are deasserted concurrently by another thread. For example, if a misconfigured PIT or a PCI device asserts and then immediately deasserts a level-triggered interrupt, the vCPU thread might see the pending interrupt during the check but find it gone during the fetch, resulting in kvm_cpu_get_interrupt() returning -1.

The warning manifests as follows:

------------[ cut here ]------------
irq == -1
WARNING: arch/x86/kvm/x86.c:10860 at kvm_check_and_inject_events
arch/x86/kvm/x86.c:10860 [inline]
WARNING: arch/x86/kvm/x86.c:10860 at vcpu_enter_guest
arch/x86/kvm/x86.c:11356 [inline]
WARNING: arch/x86/kvm/x86.c:10860 at vcpu_run+0x57ec/0x7950
arch/x86/kvm/x86.c:11770
RIP: 0010:kvm_check_and_inject_events arch/x86/kvm/x86.c:10860 [inline]
RIP: 0010:vcpu_enter_guest arch/x86/kvm/x86.c:11356 [inline]
RIP: 0010:vcpu_run+0x57ec/0x7950 arch/x86/kvm/x86.c:11770
Call Trace:
 <TASK>
 kvm_arch_vcpu_ioctl_run+0x1193/0x2070 arch/x86/kvm/x86.c:12125
 kvm_vcpu_ioctl+0xa61/0xfd0 virt/kvm/kvm_main.c:4470
 vfs_ioctl fs/ioctl.c:51 [inline]
 __do_sys_ioctl fs/ioctl.c:597 [inline]
 __se_sys_ioctl+0xfc/0x170 fs/ioctl.c:583
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0x174/0x580 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
 </TASK>

Since this is a legitimate Time-Of-Check to Time-Of-Use (TOCTOU) race condition for the in-kernel PIC, WARN_ON_ONCE() must not be used for this case. Update the warning to exempt the in-kernel PIC, while preserving it for other interrupt sources (e.g. APIC) as they are not expected to exhibit this behavior.

Fixes bf672720e83c ("KVM: x86: check the kvm_cpu_get_interrupt result before using it")
Assisted-by Gemini:gemini-3.1-pro-preview Gemini:gemini-3-flash-preview syzbot
Reported-by [email protected]

---
v2:
- Restrict the warning exemption to the in-kernel PIC case.
- Remove the pr_err_ratelimited() logging.
- Preserve the WARN_ON_ONCE() for non-PIC interrupt sources.

v1:
https://lore.kernel.org/all/[email protected]/T/
---

arch/x86/kvm/x86.c +0 -0
index 0550359ed..f1681aa9f 100644

arch/x86/kvm/x86.c static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu, · line 10862

Quoted 7 lines
arch/x86/kvm/x86.c +3 -1
@@ -10857,7 +10857,9 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
                if (r) {
                        int irq = kvm_cpu_get_interrupt(vcpu);
 
-                       if (!WARN_ON_ONCE(irq == -1)) {
+                       WARN_ON_ONCE(irq == -1 && !pic_in_kernel(vcpu->kvm));
+
+                       if (irq != -1) {

Hmm, no, that looks weird.
Sorry for the noise, please ignore.

Quoted 12 lines
arch/x86/kvm/x86.c +0 -0
                                kvm_queue_interrupt(vcpu, irq, false);
                                kvm_x86_call(inject_irq)(vcpu, false);
                                WARN_ON(kvm_x86_call(interrupt_allowed)(vcpu, true) < 0);

base-commit: 8cd9520d35a6c38db6567e97dd93b1f11f185dc6
--
See https://goo.gle/syzbot-ai-patches for information about AI-generated patches. You can comment on the patch as usual, syzbot will try to address the comments and send a new version of the patch if necessary. syzbot engineers can be reached at [email protected].

#4 lore raw
Quoted 11 lines · Aleksandr Nogikh

On Thu, Jun 25, 2026 at 11:10 PM 'syzbot' via syzkaller-bugs

Quoted 10 lines

https://lore.kernel.org/all/[email protected]/T/
---

arch/x86/kvm/x86.c +0 -0
index 0550359ed..f1681aa9f 100644
@@ -10857,7 +10857,9 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
                if (r) {
                        int irq = kvm_cpu_get_interrupt(vcpu);
 
Quoted 7 lines

arch/x86/kvm/x86.c static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu, · line 10862

Quoted 4 lines
arch/x86/kvm/x86.c +3 -1
-                       if (!WARN_ON_ONCE(irq == -1)) {
+                       WARN_ON_ONCE(irq == -1 && !pic_in_kernel(vcpu->kvm));
+
+                       if (irq != -1) {

Hmm, no, that looks weird.
Sorry for the noise, please ignore.

Looks right to me? FWIW, this is what I had thrown together locally:

---

Author:     Sean Christopherson <[email protected]>
AuthorDate: Thu Jun 25 08:46:48 2026 -0700
Commit:     Sean Christopherson <[email protected]>
CommitDate: Thu Jun 25 09:00:49 2026 -0700

KVM: x86: Don't WARN if IRQ disappears because it was cleared from the PIC

When getting a to-be-injected IRQ, don't WARN if the IRQ disappeared and the VM has an in-kernel PIC, as the ExtINT handling that's routed through KVM's virtual PIC is tracked per-VM, not per-vCPU. If another vCPU grabs the IRQ, or deasserts the interrupt (which is level-triggered), then it's both expected and "fine" for a

Keep the assert for split IRQCHIP VMs to help detect KVM bugs, as userspace is responsible for routing ExtINT to the intended vCPU, i.e. once an ExtINT is pending, it can't be cleared without holding the vCPU's mutex, and thus false positives are impossible.

x86/kvm/x86.c +3 -1
index 0626e835e9eb..7feddeeb819d 100644
@@ -7686,10 +7686,12 @@ static int kvm_check_and_inject_events(struct kvm_vcpu *vcpu,
                if (r) {
                        int irq = kvm_cpu_get_interrupt(vcpu);
 
-                       if (!WARN_ON_ONCE(irq == -1)) {
+                       if (likely(irq != -1)) {
                                kvm_queue_interrupt(vcpu, irq, false);
                                kvm_x86_call(inject_irq)(vcpu, false);
                                WARN_ON(kvm_x86_call(interrupt_allowed)(vcpu, true) < 0);
+                       } else {
+                               WARN_ON_ONCE(!pic_in_kernel(vcpu->kvm));
                        }
                }
                if (kvm_cpu_has_injectable_intr(vcpu))