[PATCH v2 04/12] KVM: Add irqfd to KVM's list via the vfs_poll() callback

[Sean Christopherson <seanjc@google.com>]

Add the irqfd structure to KVM's list of irqfds in kvm_irqfd_register(),
i.e. via the vfs_poll() callback.  This will allow taking irqfds.lock
across the entire registration sequence (add to waitqueue, add to list),
and more importantly will allow inserting into KVM's list if and only if
adding to the waitqueue succeeds (spoiler alert), without needing to
juggle return codes in weird ways.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 virt/kvm/eventfd.c | 102 +++++++++++++++++++++++++--------------------
 1 file changed, 57 insertions(+), 45 deletions(-)

diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
index 8b9a87daa2bb..99274d60335d 100644
--- a/virt/kvm/eventfd.c
+++ b/virt/kvm/eventfd.c
@@ -245,34 +245,14 @@ irqfd_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync, void *key)
 	return ret;
 }
 
-struct kvm_irqfd_pt {
-	struct kvm_kernel_irqfd *irqfd;
-	poll_table pt;
-};
-
-static void kvm_irqfd_register(struct file *file, wait_queue_head_t *wqh,
-			       poll_table *pt)
-{
-	struct kvm_irqfd_pt *p = container_of(pt, struct kvm_irqfd_pt, pt);
-	struct kvm_kernel_irqfd *irqfd = p->irqfd;
-
-	/*
-	 * Add the irqfd as a priority waiter on the eventfd, with a custom
-	 * wake-up handler, so that KVM *and only KVM* is notified whenever the
-	 * underlying eventfd is signaled.
-	 */
-	init_waitqueue_func_entry(&irqfd->wait, irqfd_wakeup);
-
-	add_wait_queue_priority(wqh, &irqfd->wait);
-}
-
-/* Must be called under irqfds.lock */
 static void irqfd_update(struct kvm *kvm, struct kvm_kernel_irqfd *irqfd)
 {
 	struct kvm_kernel_irq_routing_entry *e;
 	struct kvm_kernel_irq_routing_entry entries[KVM_NR_IRQCHIPS];
 	int n_entries;
 
+	lockdep_assert_held(&kvm->irqfds.lock);
+
 	n_entries = kvm_irq_map_gsi(kvm, entries, irqfd->gsi);
 
 	write_seqcount_begin(&irqfd->irq_entry_sc);
@@ -286,6 +266,49 @@ static void irqfd_update(struct kvm *kvm, struct kvm_kernel_irqfd *irqfd)
 	write_seqcount_end(&irqfd->irq_entry_sc);
 }
 
+struct kvm_irqfd_pt {
+	struct kvm_kernel_irqfd *irqfd;
+	struct kvm *kvm;
+	poll_table pt;
+	int ret;
+};
+
+static void kvm_irqfd_register(struct file *file, wait_queue_head_t *wqh,
+			       poll_table *pt)
+{
+	struct kvm_irqfd_pt *p = container_of(pt, struct kvm_irqfd_pt, pt);
+	struct kvm_kernel_irqfd *irqfd = p->irqfd;
+	struct kvm_kernel_irqfd *tmp;
+	struct kvm *kvm = p->kvm;
+
+	spin_lock_irq(&kvm->irqfds.lock);
+
+	list_for_each_entry(tmp, &kvm->irqfds.items, list) {
+		if (irqfd->eventfd != tmp->eventfd)
+			continue;
+		/* This fd is used for another irq already. */
+		p->ret = -EBUSY;
+		spin_unlock_irq(&kvm->irqfds.lock);
+		return;
+	}
+
+	irqfd_update(kvm, irqfd);
+
+	list_add_tail(&irqfd->list, &kvm->irqfds.items);
+
+	spin_unlock_irq(&kvm->irqfds.lock);
+
+	/*
+	 * Add the irqfd as a priority waiter on the eventfd, with a custom
+	 * wake-up handler, so that KVM *and only KVM* is notified whenever the
+	 * underlying eventfd is signaled.
+	 */
+	init_waitqueue_func_entry(&irqfd->wait, irqfd_wakeup);
+
+	add_wait_queue_priority(wqh, &irqfd->wait);
+	p->ret = 0;
+}
+
 #if IS_ENABLED(CONFIG_HAVE_KVM_IRQ_BYPASS)
 void __attribute__((weak)) kvm_arch_irq_bypass_stop(
 				struct irq_bypass_consumer *cons)
@@ -315,7 +338,7 @@ bool __attribute__((weak)) kvm_arch_irqfd_route_changed(
 static int
 kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 {
-	struct kvm_kernel_irqfd *irqfd, *tmp;
+	struct kvm_kernel_irqfd *irqfd;
 	struct eventfd_ctx *eventfd = NULL, *resamplefd = NULL;
 	struct kvm_irqfd_pt irqfd_pt;
 	int ret;
@@ -414,32 +437,22 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 	 */
 	idx = srcu_read_lock(&kvm->irq_srcu);
 
-	spin_lock_irq(&kvm->irqfds.lock);
-
-	ret = 0;
-	list_for_each_entry(tmp, &kvm->irqfds.items, list) {
-		if (irqfd->eventfd != tmp->eventfd)
-			continue;
-		/* This fd is used for another irq already. */
-		ret = -EBUSY;
-		goto fail_duplicate;
-	}
-
-	irqfd_update(kvm, irqfd);
-
-	list_add_tail(&irqfd->list, &kvm->irqfds.items);
-
-	spin_unlock_irq(&kvm->irqfds.lock);
-
 	/*
-	 * Register the irqfd with the eventfd by polling on the eventfd.  If
-	 * there was en event pending on the eventfd prior to registering,
-	 * manually trigger IRQ injection.
+	 * Register the irqfd with the eventfd by polling on the eventfd, and
+	 * simultaneously and the irqfd to KVM's list.  If there was en event
+	 * pending on the eventfd prior to registering, manually trigger IRQ
+	 * injection.
 	 */
 	irqfd_pt.irqfd = irqfd;
+	irqfd_pt.kvm = kvm;
 	init_poll_funcptr(&irqfd_pt.pt, kvm_irqfd_register);
 
 	events = vfs_poll(fd_file(f), &irqfd_pt.pt);
+
+	ret = irqfd_pt.ret;
+	if (ret)
+		goto fail_poll;
+
 	if (events & EPOLLIN)
 		schedule_work(&irqfd->inject);
 
@@ -460,8 +473,7 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 	srcu_read_unlock(&kvm->irq_srcu, idx);
 	return 0;
 
-fail_duplicate:
-	spin_unlock_irq(&kvm->irqfds.lock);
+fail_poll:
 	srcu_read_unlock(&kvm->irq_srcu, idx);
 fail:
 	if (irqfd->resampler)
-- 
2.49.0.1101.gccaa498523-goog