KVM: Enable MSI for device assignment
Sheng Yang [Mon, 24 Nov 2008 06:32:56 +0000 (14:32 +0800)]
We enable guest MSI and host MSI support in this patch. The userspace want to
enable MSI should set KVM_DEV_IRQ_ASSIGN_ENABLE_MSI in the assigned_irq's flag.
Function would return -ENOTTY if can't enable MSI, userspace shouldn't set MSI
Enable bit when KVM_ASSIGN_IRQ return -ENOTTY with
KVM_DEV_IRQ_ASSIGN_ENABLE_MSI.

Userspace can tell the support of MSI device from #ifdef KVM_CAP_DEVICE_MSI.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>

include/linux/kvm.h
virt/kvm/kvm_main.c

index bb283c3..0997e6f 100644 (file)
@@ -392,6 +392,9 @@ struct kvm_trace_rec {
 #endif
 #define KVM_CAP_IOMMU 18
 #define KVM_CAP_NMI 19
+#if defined(CONFIG_X86)
+#define KVM_CAP_DEVICE_MSI 20
+#endif
 
 /*
  * ioctls for VM fds
index 228c1d1..bf36ae9 100644 (file)
@@ -159,9 +159,15 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
         * finer-grained lock, update this
         */
        mutex_lock(&assigned_dev->kvm->lock);
-       kvm_set_irq(assigned_dev->kvm,
-                   assigned_dev->irq_source_id,
-                   assigned_dev->guest_irq, 1);
+       if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_GUEST_INTX)
+               kvm_set_irq(assigned_dev->kvm,
+                           assigned_dev->irq_source_id,
+                           assigned_dev->guest_irq, 1);
+       else if (assigned_dev->irq_requested_type &
+                               KVM_ASSIGNED_DEV_GUEST_MSI) {
+               assigned_device_msi_dispatch(assigned_dev);
+               enable_irq(assigned_dev->host_irq);
+       }
        mutex_unlock(&assigned_dev->kvm->lock);
        kvm_put_kvm(assigned_dev->kvm);
 }
@@ -197,6 +203,8 @@ static void kvm_free_assigned_device(struct kvm *kvm,
 {
        if (irqchip_in_kernel(kvm) && assigned_dev->irq_requested_type)
                free_irq(assigned_dev->host_irq, (void *)assigned_dev);
+       if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI)
+               pci_disable_msi(assigned_dev->dev);
 
        kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier);
        kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id);
@@ -242,6 +250,11 @@ static int assigned_device_update_intx(struct kvm *kvm,
                return 0;
 
        if (irqchip_in_kernel(kvm)) {
+               if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI) {
+                       free_irq(adev->host_irq, (void *)kvm);
+                       pci_disable_msi(adev->dev);
+               }
+
                if (!capable(CAP_SYS_RAWIO))
                        return -EPERM;
 
@@ -265,6 +278,41 @@ static int assigned_device_update_intx(struct kvm *kvm,
        return 0;
 }
 
+#ifdef CONFIG_X86
+static int assigned_device_update_msi(struct kvm *kvm,
+                       struct kvm_assigned_dev_kernel *adev,
+                       struct kvm_assigned_irq *airq)
+{
+       int r;
+
+       /* x86 don't care upper address of guest msi message addr */
+       adev->guest_msi.address_lo = airq->guest_msi.addr_lo;
+       adev->guest_msi.data = airq->guest_msi.data;
+       adev->ack_notifier.gsi = -1;
+
+       if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI)
+               return 0;
+
+       if (irqchip_in_kernel(kvm)) {
+               if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_INTX)
+                       free_irq(adev->host_irq, (void *)adev);
+
+               r = pci_enable_msi(adev->dev);
+               if (r)
+                       return r;
+
+               adev->host_irq = adev->dev->irq;
+               if (request_irq(adev->host_irq, kvm_assigned_dev_intr, 0,
+                               "kvm_assigned_msi_device", (void *)adev))
+                       return -EIO;
+       }
+
+       adev->irq_requested_type = KVM_ASSIGNED_DEV_GUEST_MSI |
+                                  KVM_ASSIGNED_DEV_HOST_MSI;
+       return 0;
+}
+#endif
+
 static int kvm_vm_ioctl_assign_irq(struct kvm *kvm,
                                   struct kvm_assigned_irq
                                   *assigned_irq)
@@ -301,9 +349,30 @@ static int kvm_vm_ioctl_assign_irq(struct kvm *kvm,
                }
        }
 
-       r = assigned_device_update_intx(kvm, match, assigned_irq);
-       if (r)
-               goto out_release;
+       if (assigned_irq->flags & KVM_DEV_IRQ_ASSIGN_ENABLE_MSI) {
+#ifdef CONFIG_X86
+               r = assigned_device_update_msi(kvm, match, assigned_irq);
+               if (r) {
+                       printk(KERN_WARNING "kvm: failed to enable "
+                                       "MSI device!\n");
+                       goto out_release;
+               }
+#else
+               r = -ENOTTY;
+#endif
+       } else if (assigned_irq->host_irq == 0 && match->dev->irq == 0) {
+               /* Host device IRQ 0 means don't support INTx */
+               printk(KERN_WARNING "kvm: wait device to enable MSI!\n");
+               r = 0;
+       } else {
+               /* Non-sharing INTx mode */
+               r = assigned_device_update_intx(kvm, match, assigned_irq);
+               if (r) {
+                       printk(KERN_WARNING "kvm: failed to enable "
+                                       "INTx device!\n");
+                       goto out_release;
+               }
+       }
 
        mutex_unlock(&kvm->lock);
        return r;