KVM: switch irq injection/acking data structures to irq_lock

Protect irq injection/acking data structures with a separate irq_lock
mutex. This fixes the following deadlock:

CPU A                               CPU B
kvm_vm_ioctl_deassign_dev_irq()
  mutex_lock(&kvm->lock);            worker_thread()
  -> kvm_deassign_irq()                -> kvm_assigned_dev_interrupt_work_handler()
    -> deassign_host_irq()               mutex_lock(&kvm->lock);
      -> cancel_work_sync() [blocked]

[gleb: fix ia64 path]
Reported-by: NAlex Williamson <alex.williamson@hp.com>
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: NGleb Natapov <gleb@redhat.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

arch/ia64/kvm/kvm-ia64.c

@@ -1000,10 +1000,10 @@ long kvm_arch_vm_ioctl(struct file *filp,
 			goto out;
 		if (irqchip_in_kernel(kvm)) {
 			__s32 status;
-			mutex_lock(&kvm->lock);
+			mutex_lock(&kvm->irq_lock);
 			status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
 				    irq_event.irq, irq_event.level);
-			mutex_unlock(&kvm->lock);
+			mutex_unlock(&kvm->irq_lock);
 			if (ioctl == KVM_IRQ_LINE_STATUS) {
 				irq_event.status = status;
 				if (copy_to_user(argp, &irq_event,

arch/x86/kvm/i8254.c

@@ -654,10 +654,10 @@ static void __inject_pit_timer_intr(struct kvm *kvm)
 	struct kvm_vcpu *vcpu;
 	int i;
 
-	mutex_lock(&kvm->lock);
+	mutex_lock(&kvm->irq_lock);
 	kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 1);
 	kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 0);
-	mutex_unlock(&kvm->lock);
+	mutex_unlock(&kvm->irq_lock);
 
 	/*
 	 * Provides NMI watchdog support via Virtual Wire mode.

arch/x86/kvm/lapic.c

@@ -424,7 +424,9 @@ static void apic_set_eoi(struct kvm_lapic *apic)
 		trigger_mode = IOAPIC_LEVEL_TRIG;
 	else
 		trigger_mode = IOAPIC_EDGE_TRIG;
+	mutex_lock(&apic->vcpu->kvm->irq_lock);
 	kvm_ioapic_update_eoi(apic->vcpu->kvm, vector, trigger_mode);
+	mutex_unlock(&apic->vcpu->kvm->irq_lock);
 }
 
 static void apic_send_ipi(struct kvm_lapic *apic)
@@ -448,7 +450,9 @@ static void apic_send_ipi(struct kvm_lapic *apic)
 		   irq.trig_mode, irq.level, irq.dest_mode, irq.delivery_mode,
 		   irq.vector);
 
+	mutex_lock(&apic->vcpu->kvm->irq_lock);
 	kvm_irq_delivery_to_apic(apic->vcpu->kvm, apic, &irq);
+	mutex_unlock(&apic->vcpu->kvm->irq_lock);
 }
 
 static u32 apic_get_tmcct(struct kvm_lapic *apic)

arch/x86/kvm/x86.c

@@ -2136,10 +2136,10 @@ long kvm_arch_vm_ioctl(struct file *filp,
 			goto out;
 		if (irqchip_in_kernel(kvm)) {
 			__s32 status;
-			mutex_lock(&kvm->lock);
+			mutex_lock(&kvm->irq_lock);
 			status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
 					irq_event.irq, irq_event.level);
-			mutex_unlock(&kvm->lock);
+			mutex_unlock(&kvm->irq_lock);
 			if (ioctl == KVM_IRQ_LINE_STATUS) {
 				irq_event.status = status;
 				if (copy_to_user(argp, &irq_event,
@@ -2385,12 +2385,11 @@ static int emulator_read_emulated(unsigned long addr,
 	 */
 	mutex_lock(&vcpu->kvm->lock);
 	mmio_dev = vcpu_find_mmio_dev(vcpu, gpa, bytes, 0);
+	mutex_unlock(&vcpu->kvm->lock);
 	if (mmio_dev) {
 		kvm_iodevice_read(mmio_dev, gpa, bytes, val);
-		mutex_unlock(&vcpu->kvm->lock);
 		return X86EMUL_CONTINUE;
 	}
-	mutex_unlock(&vcpu->kvm->lock);
 
 	vcpu->mmio_needed = 1;
 	vcpu->mmio_phys_addr = gpa;
@@ -2440,12 +2439,11 @@ static int emulator_write_emulated_onepage(unsigned long addr,
 	 */
 	mutex_lock(&vcpu->kvm->lock);
 	mmio_dev = vcpu_find_mmio_dev(vcpu, gpa, bytes, 1);
+	mutex_unlock(&vcpu->kvm->lock);
 	if (mmio_dev) {
 		kvm_iodevice_write(mmio_dev, gpa, bytes, val);
-		mutex_unlock(&vcpu->kvm->lock);
 		return X86EMUL_CONTINUE;
 	}
-	mutex_unlock(&vcpu->kvm->lock);
 
 	vcpu->mmio_needed = 1;
 	vcpu->mmio_phys_addr = gpa;
@@ -2768,7 +2766,6 @@ static void kernel_pio(struct kvm_io_device *pio_dev,
 {
 	/* TODO: String I/O for in kernel device */
 
-	mutex_lock(&vcpu->kvm->lock);
 	if (vcpu->arch.pio.in)
 		kvm_iodevice_read(pio_dev, vcpu->arch.pio.port,
 				  vcpu->arch.pio.size,
@@ -2777,7 +2774,6 @@ static void kernel_pio(struct kvm_io_device *pio_dev,
 		kvm_iodevice_write(pio_dev, vcpu->arch.pio.port,
 				   vcpu->arch.pio.size,
 				   pd);
-	mutex_unlock(&vcpu->kvm->lock);
 }
 
 static void pio_string_write(struct kvm_io_device *pio_dev,
@@ -2787,14 +2783,12 @@ static void pio_string_write(struct kvm_io_device *pio_dev,
 	void *pd = vcpu->arch.pio_data;
 	int i;
 
-	mutex_lock(&vcpu->kvm->lock);
 	for (i = 0; i < io->cur_count; i++) {
 		kvm_iodevice_write(pio_dev, io->port,
 				   io->size,
 				   pd);
 		pd += io->size;
 	}
-	mutex_unlock(&vcpu->kvm->lock);
 }
 
 static struct kvm_io_device *vcpu_find_pio_dev(struct kvm_vcpu *vcpu,
@@ -2831,7 +2825,9 @@ int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
 	val = kvm_register_read(vcpu, VCPU_REGS_RAX);
 	memcpy(vcpu->arch.pio_data, &val, 4);
 
+	mutex_lock(&vcpu->kvm->lock);
 	pio_dev = vcpu_find_pio_dev(vcpu, port, size, !in);
+	mutex_unlock(&vcpu->kvm->lock);
 	if (pio_dev) {
 		kernel_pio(pio_dev, vcpu, vcpu->arch.pio_data);
 		complete_pio(vcpu);
@@ -2895,9 +2891,12 @@ int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
 
 	vcpu->arch.pio.guest_gva = address;
 
+	mutex_lock(&vcpu->kvm->lock);
 	pio_dev = vcpu_find_pio_dev(vcpu, port,
 				    vcpu->arch.pio.cur_count,
 				    !vcpu->arch.pio.in);
+	mutex_unlock(&vcpu->kvm->lock);
+
 	if (!vcpu->arch.pio.in) {
 		/* string PIO write */
 		ret = pio_copy_data(vcpu);

include/linux/kvm_host.h

@@ -371,7 +371,8 @@ int kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level);
 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 				   struct kvm_irq_ack_notifier *kian);
-void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier *kian);
+void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
+				   struct kvm_irq_ack_notifier *kian);
 int kvm_request_irq_source_id(struct kvm *kvm);
 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 

virt/kvm/eventfd.c

@@ -57,10 +57,10 @@ irqfd_inject(struct work_struct *work)
 	struct _irqfd *irqfd = container_of(work, struct _irqfd, inject);
 	struct kvm *kvm = irqfd->kvm;
 
-	mutex_lock(&kvm->lock);
+	mutex_lock(&kvm->irq_lock);
 	kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, irqfd->gsi, 1);
 	kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, irqfd->gsi, 0);
-	mutex_unlock(&kvm->lock);
+	mutex_unlock(&kvm->irq_lock);
 }
 
 /*

virt/kvm/irq_comm.c

@@ -62,6 +62,8 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
 	int i, r = -1;
 	struct kvm_vcpu *vcpu, *lowest = NULL;
 
+	WARN_ON(!mutex_is_locked(&kvm->irq_lock));
+
 	if (irq->dest_mode == 0 && irq->dest_id == 0xff &&
 			kvm_is_dm_lowest_prio(irq))
 		printk(KERN_INFO "kvm: apic: phys broadcast and lowest prio\n");
@@ -113,7 +115,7 @@ static int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
 	return kvm_irq_delivery_to_apic(kvm, NULL, &irq);
 }
 
-/* This should be called with the kvm->lock mutex held
+/* This should be called with the kvm->irq_lock mutex held
  * Return value:
  *  < 0   Interrupt was ignored (masked or not delivered for other reasons)
  *  = 0   Interrupt was coalesced (previous irq is still pending)
@@ -125,6 +127,8 @@ int kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level)
 	unsigned long *irq_state, sig_level;
 	int ret = -1;
 
+	WARN_ON(!mutex_is_locked(&kvm->irq_lock));
+
 	if (irq < KVM_IOAPIC_NUM_PINS) {
 		irq_state = (unsigned long *)&kvm->arch.irq_states[irq];
 
@@ -175,19 +179,26 @@ void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin)
 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 				   struct kvm_irq_ack_notifier *kian)
 {
+	mutex_lock(&kvm->irq_lock);
 	hlist_add_head(&kian->link, &kvm->arch.irq_ack_notifier_list);
+	mutex_unlock(&kvm->irq_lock);
 }
 
-void kvm_unregister_irq_ack_notifier(struct kvm_irq_ack_notifier *kian)
+void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
+				    struct kvm_irq_ack_notifier *kian)
 {
+	mutex_lock(&kvm->irq_lock);
 	hlist_del_init(&kian->link);
+	mutex_unlock(&kvm->irq_lock);
 }
 
-/* The caller must hold kvm->lock mutex */
 int kvm_request_irq_source_id(struct kvm *kvm)
 {
 	unsigned long *bitmap = &kvm->arch.irq_sources_bitmap;
-	int irq_source_id = find_first_zero_bit(bitmap,
+	int irq_source_id;
+
+	mutex_lock(&kvm->irq_lock);
+	irq_source_id = find_first_zero_bit(bitmap,
 				sizeof(kvm->arch.irq_sources_bitmap));
 
 	if (irq_source_id >= sizeof(kvm->arch.irq_sources_bitmap)) {
@@ -197,6 +208,7 @@ int kvm_request_irq_source_id(struct kvm *kvm)
 
 	ASSERT(irq_source_id != KVM_USERSPACE_IRQ_SOURCE_ID);
 	set_bit(irq_source_id, bitmap);
+	mutex_unlock(&kvm->irq_lock);
 
 	return irq_source_id;
 }
@@ -207,6 +219,7 @@ void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id)
 
 	ASSERT(irq_source_id != KVM_USERSPACE_IRQ_SOURCE_ID);
 
+	mutex_lock(&kvm->irq_lock);
 	if (irq_source_id < 0 ||
 	    irq_source_id >= sizeof(kvm->arch.irq_sources_bitmap)) {
 		printk(KERN_ERR "kvm: IRQ source ID out of range!\n");
@@ -215,19 +228,24 @@ void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id)
 	for (i = 0; i < KVM_IOAPIC_NUM_PINS; i++)
 		clear_bit(irq_source_id, &kvm->arch.irq_states[i]);
 	clear_bit(irq_source_id, &kvm->arch.irq_sources_bitmap);
+	mutex_unlock(&kvm->irq_lock);
 }
 
 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 				    struct kvm_irq_mask_notifier *kimn)
 {
+	mutex_lock(&kvm->irq_lock);
 	kimn->irq = irq;
 	hlist_add_head(&kimn->link, &kvm->mask_notifier_list);
+	mutex_unlock(&kvm->irq_lock);
 }
 
 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 				      struct kvm_irq_mask_notifier *kimn)
 {
+	mutex_lock(&kvm->irq_lock);
 	hlist_del(&kimn->link);
+	mutex_unlock(&kvm->irq_lock);
 }
 
 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask)
@@ -235,6 +253,8 @@ void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask)
 	struct kvm_irq_mask_notifier *kimn;
 	struct hlist_node *n;
 
+	WARN_ON(!mutex_is_locked(&kvm->irq_lock));
+
 	hlist_for_each_entry(kimn, n, &kvm->mask_notifier_list, link)
 		if (kimn->irq == irq)
 			kimn->func(kimn, mask);
@@ -250,7 +270,9 @@ static void __kvm_free_irq_routing(struct list_head *irq_routing)
 
 void kvm_free_irq_routing(struct kvm *kvm)
 {
+	mutex_lock(&kvm->irq_lock);
 	__kvm_free_irq_routing(&kvm->irq_routing);
+	mutex_unlock(&kvm->irq_lock);
 }
 
 static int setup_routing_entry(struct kvm_kernel_irq_routing_entry *e,
@@ -325,13 +347,13 @@ int kvm_set_irq_routing(struct kvm *kvm,
 		e = NULL;
 	}
 
-	mutex_lock(&kvm->lock);
+	mutex_lock(&kvm->irq_lock);
 	list_splice(&kvm->irq_routing, &tmp);
 	INIT_LIST_HEAD(&kvm->irq_routing);
 	list_splice(&irq_list, &kvm->irq_routing);
 	INIT_LIST_HEAD(&irq_list);
 	list_splice(&tmp, &irq_list);
-	mutex_unlock(&kvm->lock);
+	mutex_unlock(&kvm->irq_lock);
 
 	r = 0;
 

virt/kvm/kvm_main.c

@@ -62,6 +62,12 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+/*
+ * Ordering of locks:
+ *
+ * 		kvm->lock --> kvm->irq_lock
+ */
+
 DEFINE_SPINLOCK(kvm_lock);
 LIST_HEAD(vm_list);
 
@@ -126,11 +132,7 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
 				    interrupt_work);
 	kvm = assigned_dev->kvm;
 
-	/* This is taken to safely inject irq inside the guest. When
-	 * the interrupt injection (or the ioapic code) uses a
-	 * finer-grained lock, update this
-	 */
-	mutex_lock(&kvm->lock);
+	mutex_lock(&kvm->irq_lock);
 	spin_lock_irq(&assigned_dev->assigned_dev_lock);
 	if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) {
 		struct kvm_guest_msix_entry *guest_entries =
@@ -149,7 +151,7 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
 			    assigned_dev->guest_irq, 1);
 
 	spin_unlock_irq(&assigned_dev->assigned_dev_lock);
-	mutex_unlock(&assigned_dev->kvm->lock);
+	mutex_unlock(&assigned_dev->kvm->irq_lock);
 }
 
 static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id)
@@ -207,7 +209,7 @@ static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian)
 static void deassign_guest_irq(struct kvm *kvm,
 			       struct kvm_assigned_dev_kernel *assigned_dev)
 {
-	kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier);
+	kvm_unregister_irq_ack_notifier(kvm, &assigned_dev->ack_notifier);
 	assigned_dev->ack_notifier.gsi = -1;
 
 	if (assigned_dev->irq_source_id != -1)