KVM: Fix guest shared interrupt with in-kernel irqchip

Every call of kvm_set_irq() should offer an irq_source_id, which is
allocated by kvm_request_irq_source_id(). Based on irq_source_id, we
identify the irq source and implement logical OR for shared level
interrupts.

The allocated irq_source_id can be freed by kvm_free_irq_source_id().

Currently, we support at most sizeof(unsigned long) different irq sources.

[Amit: - rebase to kvm.git HEAD
       - move definition of KVM_USERSPACE_IRQ_SOURCE_ID to common file
       - move kvm_request_irq_source_id to the update_irq ioctl]

[Xiantao: - Add kvm/ia64 stuff and make it work for kvm/ia64 guests]
Signed-off-by: NSheng Yang <sheng@linux.intel.com>
Signed-off-by: NAmit Shah <amit.shah@redhat.com>
Signed-off-by: NXiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

arch/ia64/include/asm/kvm_host.h

@@ -417,6 +417,9 @@ struct kvm_arch {
 	struct list_head assigned_dev_head;
 	struct dmar_domain *intel_iommu_domain;
 	struct hlist_head irq_ack_notifier_list;
+
+	unsigned long irq_sources_bitmap;
+	unsigned long irq_states[KVM_IOAPIC_NUM_PINS];
 };
 
 union cpuid3_t {

arch/ia64/kvm/kvm-ia64.c

@@ -778,6 +778,9 @@ static void kvm_init_vm(struct kvm *kvm)
 	kvm_build_io_pmt(kvm);
 
 	INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
+
+	/* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
+	set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap);
 }
 
 struct  kvm *kvm_arch_create_vm(void)
@@ -941,9 +944,8 @@ long kvm_arch_vm_ioctl(struct file *filp,
 			goto out;
 		if (irqchip_in_kernel(kvm)) {
 			mutex_lock(&kvm->lock);
-			kvm_ioapic_set_irq(kvm->arch.vioapic,
-						irq_event.irq,
-						irq_event.level);
+			kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
+				    irq_event.irq, irq_event.level);
 			mutex_unlock(&kvm->lock);
 			r = 0;
 		}

arch/x86/include/asm/kvm_host.h

@@ -364,6 +364,9 @@ struct kvm_arch{
 
 	struct page *ept_identity_pagetable;
 	bool ept_identity_pagetable_done;
+
+	unsigned long irq_sources_bitmap;
+	unsigned long irq_states[KVM_IOAPIC_NUM_PINS];
 };
 
 struct kvm_vm_stat {

arch/x86/kvm/i8254.c

@@ -545,6 +545,12 @@ struct kvm_pit *kvm_create_pit(struct kvm *kvm)
 	if (!pit)
 		return NULL;
 
+	mutex_lock(&kvm->lock);
+	pit->irq_source_id = kvm_request_irq_source_id(kvm);
+	mutex_unlock(&kvm->lock);
+	if (pit->irq_source_id < 0)
+		return NULL;
+
 	mutex_init(&pit->pit_state.lock);
 	mutex_lock(&pit->pit_state.lock);
 	spin_lock_init(&pit->pit_state.inject_lock);
@@ -587,6 +593,7 @@ void kvm_free_pit(struct kvm *kvm)
 		mutex_lock(&kvm->arch.vpit->pit_state.lock);
 		timer = &kvm->arch.vpit->pit_state.pit_timer.timer;
 		hrtimer_cancel(timer);
+		kvm_free_irq_source_id(kvm, kvm->arch.vpit->irq_source_id);
 		mutex_unlock(&kvm->arch.vpit->pit_state.lock);
 		kfree(kvm->arch.vpit);
 	}
@@ -595,8 +602,8 @@ void kvm_free_pit(struct kvm *kvm)
 static void __inject_pit_timer_intr(struct kvm *kvm)
 {
 	mutex_lock(&kvm->lock);
-	kvm_set_irq(kvm, 0, 1);
-	kvm_set_irq(kvm, 0, 0);
+	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);
 }
 

arch/x86/kvm/i8254.h

@@ -44,6 +44,7 @@ struct kvm_pit {
 	struct kvm_io_device speaker_dev;
 	struct kvm *kvm;
 	struct kvm_kpit_state pit_state;
+	int irq_source_id;
 };
 
 #define KVM_PIT_BASE_ADDRESS	    0x40

arch/x86/kvm/x86.c

@@ -1742,7 +1742,8 @@ long kvm_arch_vm_ioctl(struct file *filp,
 			goto out;
 		if (irqchip_in_kernel(kvm)) {
 			mutex_lock(&kvm->lock);
-			kvm_set_irq(kvm, irq_event.irq, irq_event.level);
+			kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
+				    irq_event.irq, irq_event.level);
 			mutex_unlock(&kvm->lock);
 			r = 0;
 		}
@@ -4013,6 +4014,9 @@ struct  kvm *kvm_arch_create_vm(void)
 	INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
 	INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
 
+	/* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
+	set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap);
+
 	return kvm;
 }
 

include/linux/kvm_host.h

@@ -37,6 +37,8 @@
 #define KVM_REQ_UNHALT             6
 #define KVM_REQ_MMU_SYNC           7
 
+#define KVM_USERSPACE_IRQ_SOURCE_ID	0
+
 struct kvm_vcpu;
 extern struct kmem_cache *kvm_vcpu_cache;
 
@@ -306,15 +308,18 @@ struct kvm_assigned_dev_kernel {
 	int host_irq;
 	int guest_irq;
 	int irq_requested;
+	int irq_source_id;
 	struct pci_dev *dev;
 	struct kvm *kvm;
 };
-void kvm_set_irq(struct kvm *kvm, int irq, int level);
+void kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level);
 void kvm_notify_acked_irq(struct kvm *kvm, unsigned gsi);
 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 				   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);
 
 #ifdef CONFIG_DMAR
 int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,

virt/kvm/irq_comm.c

@@ -25,15 +25,23 @@
 #include "ioapic.h"
 
 /* This should be called with the kvm->lock mutex held */
-void kvm_set_irq(struct kvm *kvm, int irq, int level)
+void kvm_set_irq(struct kvm *kvm, int irq_source_id, int irq, int level)
 {
+	unsigned long *irq_state = (unsigned long *)&kvm->arch.irq_states[irq];
+
+	/* Logical OR for level trig interrupt */
+	if (level)
+		set_bit(irq_source_id, irq_state);
+	else
+		clear_bit(irq_source_id, irq_state);
+
 	/* Not possible to detect if the guest uses the PIC or the
 	 * IOAPIC.  So set the bit in both. The guest will ignore
 	 * writes to the unused one.
 	 */
-	kvm_ioapic_set_irq(kvm->arch.vioapic, irq, level);
+	kvm_ioapic_set_irq(kvm->arch.vioapic, irq, !!(*irq_state));
 #ifdef CONFIG_X86
-	kvm_pic_set_irq(pic_irqchip(kvm), irq, level);
+	kvm_pic_set_irq(pic_irqchip(kvm), irq, !!(*irq_state));
 #endif
 }
 
@@ -58,3 +66,31 @@ void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 {
 	hlist_del(&kian->link);
 }
+
+/* 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,
+				sizeof(kvm->arch.irq_sources_bitmap));
+	if (irq_source_id >= sizeof(kvm->arch.irq_sources_bitmap)) {
+		printk(KERN_WARNING "kvm: exhaust allocatable IRQ sources!\n");
+		irq_source_id = -EFAULT;
+	} else
+		set_bit(irq_source_id, bitmap);
+	return irq_source_id;
+}
+
+void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id)
+{
+	int i;
+
+	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");
+		return;
+	}
+	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);
+}

virt/kvm/kvm_main.c

@@ -105,14 +105,12 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
 	 */
 	mutex_lock(&assigned_dev->kvm->lock);
 	kvm_set_irq(assigned_dev->kvm,
+		    assigned_dev->irq_source_id,
 		    assigned_dev->guest_irq, 1);
 	mutex_unlock(&assigned_dev->kvm->lock);
 	kvm_put_kvm(assigned_dev->kvm);
 }
 
-/* FIXME: Implement the OR logic needed to make shared interrupts on
- * this line behave properly
- */
 static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id)
 {
 	struct kvm_assigned_dev_kernel *assigned_dev =
@@ -134,7 +132,7 @@ static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian)
 
 	dev = container_of(kian, struct kvm_assigned_dev_kernel,
 			   ack_notifier);
-	kvm_set_irq(dev->kvm, dev->guest_irq, 0);
+	kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0);
 	enable_irq(dev->host_irq);
 }
 
@@ -146,6 +144,7 @@ static void kvm_free_assigned_device(struct kvm *kvm,
 		free_irq(assigned_dev->host_irq, (void *)assigned_dev);
 
 	kvm_unregister_irq_ack_notifier(kvm, &assigned_dev->ack_notifier);
+	kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id);
 
 	if (cancel_work_sync(&assigned_dev->interrupt_work))
 		/* We had pending work. That means we will have to take
@@ -215,6 +214,11 @@ static int kvm_vm_ioctl_assign_irq(struct kvm *kvm,
 		match->ack_notifier.gsi = assigned_irq->guest_irq;
 		match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq;
 		kvm_register_irq_ack_notifier(kvm, &match->ack_notifier);
+		r = kvm_request_irq_source_id(kvm);
+		if (r < 0)
+			goto out_release;
+		else
+			match->irq_source_id = r;
 
 		/* Even though this is PCI, we don't want to use shared
 		 * interrupts. Sharing host devices with guest-assigned devices