1 files changed, 87 insertions, 8 deletions

diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index d06cc04079..f24d7da783 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -160,9 +160,59 @@ static const KVMCapabilityInfo kvm_required_capabilites[] = {
 static NotifierList kvm_irqchip_change_notifiers =
     NOTIFIER_LIST_INITIALIZER(kvm_irqchip_change_notifiers);
 
+struct KVMResampleFd {
+    int gsi;
+    EventNotifier *resample_event;
+    QLIST_ENTRY(KVMResampleFd) node;
+};
+typedef struct KVMResampleFd KVMResampleFd;
+
+/*
+ * Only used with split irqchip where we need to do the resample fd
+ * kick for the kernel from userspace.
+ */
+static QLIST_HEAD(, KVMResampleFd) kvm_resample_fd_list =
+    QLIST_HEAD_INITIALIZER(kvm_resample_fd_list);
+
 #define kvm_slots_lock(kml)      qemu_mutex_lock(&(kml)->slots_lock)
 #define kvm_slots_unlock(kml)    qemu_mutex_unlock(&(kml)->slots_lock)
 
+static inline void kvm_resample_fd_remove(int gsi)
+{
+    KVMResampleFd *rfd;
+
+    QLIST_FOREACH(rfd, &kvm_resample_fd_list, node) {
+        if (rfd->gsi == gsi) {
+            QLIST_REMOVE(rfd, node);
+            g_free(rfd);
+            break;
+        }
+    }
+}
+
+static inline void kvm_resample_fd_insert(int gsi, EventNotifier *event)
+{
+    KVMResampleFd *rfd = g_new0(KVMResampleFd, 1);
+
+    rfd->gsi = gsi;
+    rfd->resample_event = event;
+
+    QLIST_INSERT_HEAD(&kvm_resample_fd_list, rfd, node);
+}
+
+void kvm_resample_fd_notify(int gsi)
+{
+    KVMResampleFd *rfd;
+
+    QLIST_FOREACH(rfd, &kvm_resample_fd_list, node) {
+        if (rfd->gsi == gsi) {
+            event_notifier_set(rfd->resample_event);
+            trace_kvm_resample_fd_notify(gsi);
+            return;
+        }
+    }
+}
+
 int kvm_get_max_memslots(void)
 {
     KVMState *s = KVM_STATE(current_accel());
@@ -1662,9 +1712,13 @@ int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
     return kvm_update_routing_entry(s, &kroute);
 }
 
-static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int rfd, int virq,
+static int kvm_irqchip_assign_irqfd(KVMState *s, EventNotifier *event,
+                                    EventNotifier *resample, int virq,
                                     bool assign)
 {
+    int fd = event_notifier_get_fd(event);
+    int rfd = resample ? event_notifier_get_fd(resample) : -1;
+
     struct kvm_irqfd irqfd = {
         .fd = fd,
         .gsi = virq,
@@ -1672,8 +1726,33 @@ static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int rfd, int virq,
     };
 
     if (rfd != -1) {
-        irqfd.flags |= KVM_IRQFD_FLAG_RESAMPLE;
-        irqfd.resamplefd = rfd;
+        assert(assign);
+        if (kvm_irqchip_is_split()) {
+            /*
+             * When the slow irqchip (e.g. IOAPIC) is in the
+             * userspace, KVM kernel resamplefd will not work because
+             * the EOI of the interrupt will be delivered to userspace
+             * instead, so the KVM kernel resamplefd kick will be
+             * skipped.  The userspace here mimics what the kernel
+             * provides with resamplefd, remember the resamplefd and
+             * kick it when we receive EOI of this IRQ.
+             *
+             * This is hackery because IOAPIC is mostly bypassed
+             * (except EOI broadcasts) when irqfd is used.  However
+             * this can bring much performance back for split irqchip
+             * with INTx IRQs (for VFIO, this gives 93% perf of the
+             * full fast path, which is 46% perf boost comparing to
+             * the INTx slow path).
+             */
+            kvm_resample_fd_insert(virq, resample);
+        } else {
+            irqfd.flags |= KVM_IRQFD_FLAG_RESAMPLE;
+            irqfd.resamplefd = rfd;
+        }
+    } else if (!assign) {
+        if (kvm_irqchip_is_split()) {
+            kvm_resample_fd_remove(virq);
+        }
     }
 
     if (!kvm_irqfds_enabled()) {
@@ -1769,7 +1848,9 @@ int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint)
     return -ENOSYS;
 }
 
-static int kvm_irqchip_assign_irqfd(KVMState *s, int fd, int virq, bool assign)
+static int kvm_irqchip_assign_irqfd(KVMState *s, EventNotifier *event,
+                                    EventNotifier *resample, int virq,
+                                    bool assign)
 {
     abort();
 }
@@ -1783,15 +1864,13 @@ int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg)
 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
                                        EventNotifier *rn, int virq)
 {
-    return kvm_irqchip_assign_irqfd(s, event_notifier_get_fd(n),
-           rn ? event_notifier_get_fd(rn) : -1, virq, true);
+    return kvm_irqchip_assign_irqfd(s, n, rn, virq, true);
 }
 
 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
                                           int virq)
 {
-    return kvm_irqchip_assign_irqfd(s, event_notifier_get_fd(n), -1, virq,
-           false);
+    return kvm_irqchip_assign_irqfd(s, n, NULL, virq, false);
 }
 
 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,