KVM: selftests: Verify non-postable IRQ remapping in IRQ test

Extend the eventfd IRQ test with an '-n' flag to route a subset of device
interrupts as NMIs (Non-Maskable Interrupts) into the guest using an
alternating pattern of 4 NMIs followed by 4 regular interrupts.

While this adds coverage for NMI injection, the primary goal is to
validate KVM's handling of non-postable interrupt delivery (AMD and Intel
IOMMUs only support posting fixed IRQs targeting a single vCPU).  KVM
has historically bungled handling transitions between posted and remapped
modes.  Use NMIs to stress the transitions, because they are a reliable,
architectural way to force these code paths.

Signed-off-by: David Matlack <dmatlack@google.com>
Co-developed-by: Josh Hilke <jrhilke@google.com>
Signed-off-by: Josh Hilke <jrhilke@google.com>
[sean: add GUEST_RECEIVED_INTERRUPT(), massage changelog]
Link: https://patch.msgid.link/20260626213534.3866178-15-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
This commit is contained in:
Sean Christopherson 2026-06-26 14:35:27 -07:00
parent 38950af64a
commit 25e268b76e

View File

@ -17,11 +17,17 @@
static u64 timeout_ns = 2ULL * 1000 * 1000 * 1000;
static bool guest_ready_for_irqs[KVM_MAX_VCPUS];
static bool guest_received_irq[KVM_MAX_VCPUS];
static bool guest_received_nmi[KVM_MAX_VCPUS];
static bool irq_affinity;
static bool done;
#define GUEST_RECEIVED_IRQ(__vcpu) \
SYNC_FROM_GUEST_AND_READ((__vcpu)->vm, guest_received_irq[(__vcpu)->id])
#define GUEST_RECEIVED_NMI(__vcpu) \
SYNC_FROM_GUEST_AND_READ((__vcpu)->vm, guest_received_nmi[(__vcpu)->id])
#define GUEST_RECEIVED_INTERRUPT(__vcpu, __nmi) \
((__nmi) ? GUEST_RECEIVED_NMI(__vcpu) : GUEST_RECEIVED_IRQ(__vcpu))
static u32 guest_get_vcpu_id(void)
{
@ -35,6 +41,11 @@ static void guest_irq_handler(struct ex_regs *regs)
x2apic_write_reg(APIC_EOI, 0);
}
static void guest_nmi_handler(struct ex_regs *regs)
{
WRITE_ONCE(guest_received_nmi[guest_get_vcpu_id()], true);
}
static void guest_code(void)
{
x2apic_enable();
@ -103,7 +114,7 @@ static void trigger_interrupt(struct vfio_pci_device *device, int eventfd)
static void kvm_route_msi(struct kvm_vm *vm, u32 gsi, struct kvm_vcpu *vcpu,
u8 vector)
u8 vector, bool use_nmi)
{
struct {
struct kvm_irq_routing header;
@ -114,7 +125,7 @@ static void kvm_route_msi(struct kvm_vm *vm, u32 gsi, struct kvm_vcpu *vcpu,
.gsi = gsi,
.type = KVM_IRQ_ROUTING_MSI,
.u.msi.address_lo = 0xFEE00000 | (vcpu->id << 12),
.u.msi.data = vector,
.u.msi.data = use_nmi ? NMI_VECTOR | (4 << 8) : vector,
},
};
@ -146,13 +157,14 @@ static const char *probe_iommu_type(void)
static void help(const char *name)
{
printf("Usage: %s [-a] [-d <segment:bus:device.function>] [-e] [-h] [-i nr_irqs] [-t iommu_type]\n", name);
printf("Usage: %s [-a] [-d <segment:bus:device.function>] [-e] [-h] [-i nr_irqs] [-n] [-t iommu_type]\n", name);
printf("\n");
printf("Tests KVM interrupt routing and delivery via irqfd.\n");
printf("-a Affine the device's host IRQ to a random physical CPU\n");
printf("-d Use a VFIO device to send MSI-X interrupts instead of manually signaling the eventfd\n");
printf("-e Set empty GSI routing in-between some interrupts\n");
printf("-i The number of IRQs to generate during the test\n");
printf("-n Deliver 50 percent of IRQs as non-maskable interrupts\n");
printf("-t Override the IOMMU type to use (vfio_type1_iommu or iommufd)\n");
printf("\n");
exit(KSFT_FAIL);
@ -183,11 +195,12 @@ int main(int argc, char **argv)
const char *device_bdf = NULL;
const char *iommu_type = NULL;
int i, j, c, msix, eventfd;
bool use_nmi = false;
struct iommu *iommu;
struct kvm_vm *vm;
int irq, irq_cpu;
while ((c = getopt(argc, argv, "ad:ehi:t:")) != -1) {
while ((c = getopt(argc, argv, "ad:ehi:nt:")) != -1) {
switch (c) {
case 'a':
irq_affinity = true;
@ -201,6 +214,9 @@ int main(int argc, char **argv)
case 'i':
nr_irqs = atoi_positive("Number of IRQs", optarg);
break;
case 'n':
use_nmi = true;
break;
case 't':
iommu_type = optarg;
break;
@ -214,6 +230,7 @@ int main(int argc, char **argv)
vm = vm_create_with_vcpus(nr_vcpus, guest_code, vcpus);
vm_install_exception_handler(vm, vector, guest_irq_handler);
vm_install_exception_handler(vm, NMI_VECTOR, guest_nmi_handler);
if (device_bdf) {
if (!iommu_type)
@ -251,36 +268,45 @@ int main(int argc, char **argv)
for (i = 0; i < nr_irqs; i++) {
const bool do_set_empty_routing = set_empty_routing && (i & BIT(3));
const bool do_use_nmi = use_nmi && (i & BIT(2));
struct kvm_vcpu *vcpu = vcpus[i % nr_vcpus];
struct timespec start;
if (do_set_empty_routing)
kvm_set_empty_gsi_routing(vm);
kvm_route_msi(vm, gsi, vcpu, vector);
kvm_route_msi(vm, gsi, vcpu, vector, do_use_nmi);
if (irq_affinity) {
irq_cpu = kvm_random_u64(&kvm_rng) % get_nprocs();
proc_irq_set_smp_affinity(irq, irq_cpu);
}
for (j = 0; j < nr_vcpus; j++)
for (j = 0; j < nr_vcpus; j++) {
TEST_ASSERT(!GUEST_RECEIVED_IRQ(vcpus[j]),
"IRQ flag for vCPU %d not clear prior to test",
vcpus[j]->id);
TEST_ASSERT(!GUEST_RECEIVED_NMI(vcpus[j]),
"NMI flag for vCPU %d not clear prior to test",
vcpus[j]->id);
}
trigger_interrupt(device, eventfd);
clock_gettime(CLOCK_MONOTONIC, &start);
while (!GUEST_RECEIVED_IRQ(vcpu) &&
while (!GUEST_RECEIVED_INTERRUPT(vcpu, do_use_nmi) &&
timespec_to_ns(timespec_elapsed(start)) <= timeout_ns)
cpu_relax();
TEST_ASSERT(GUEST_RECEIVED_IRQ(vcpu),
"vCPU %d timed out waiting for IRQ (vector 0x%x) from GSI %d (via CPU %d)\n",
vcpu->id, vector, gsi, irq_cpu);
TEST_ASSERT(GUEST_RECEIVED_INTERRUPT(vcpu, do_use_nmi),
"vCPU %d timed out waiting for %s (vector 0x%x) from GSI %d (via CPU %d)\n",
vcpu->id, do_use_nmi ? "NMI" : "IRQ",
do_use_nmi ? NMI_VECTOR : vector, gsi, irq_cpu);
WRITE_AND_SYNC_TO_GUEST(vm, guest_received_irq[vcpu->id], false);
if (do_use_nmi)
WRITE_AND_SYNC_TO_GUEST(vm, guest_received_nmi[vcpu->id], false);
else
WRITE_AND_SYNC_TO_GUEST(vm, guest_received_irq[vcpu->id], false);
}
WRITE_AND_SYNC_TO_GUEST(vm, done, true);