KVM: Allow host IRQ sharing for assigned PCI 2.3 devices

PCI 2.3 allows to generically disable IRQ sources at device level. This
enables us to share legacy IRQs of such devices with other host devices
when passing them to a guest.

The new IRQ sharing feature introduced here is optional, user space has
to request it explicitly. Moreover, user space can inform us about its
view of PCI_COMMAND_INTX_DISABLE so that we can avoid unmasking the
interrupt and signaling it if the guest masked it via the virtualized
PCI config space.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>

5 files changed, 230 insertions, 29 deletions

diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
index 59a38264a0ed..6386f8c0482e 100644
--- a/Documentation/virtual/kvm/api.txt
+++ b/Documentation/virtual/kvm/api.txt
@@ -1169,6 +1169,14 @@ following flags are specified:
 
 /* Depends on KVM_CAP_IOMMU */
 #define KVM_DEV_ASSIGN_ENABLE_IOMMU     (1 << 0)
+/* The following two depend on KVM_CAP_PCI_2_3 */
+#define KVM_DEV_ASSIGN_PCI_2_3          (1 << 1)
+#define KVM_DEV_ASSIGN_MASK_INTX        (1 << 2)
+
+If KVM_DEV_ASSIGN_PCI_2_3 is set, the kernel will manage legacy INTx interrupts
+via the PCI-2.3-compliant device-level mask, thus enable IRQ sharing with other
+assigned devices or host devices. KVM_DEV_ASSIGN_MASK_INTX specifies the
+guest's view on the INTx mask, see KVM_ASSIGN_SET_INTX_MASK for details.
 
 The KVM_DEV_ASSIGN_ENABLE_IOMMU flag is a mandatory option to ensure
 isolation of the device.  Usages not specifying this flag are deprecated.
@@ -1441,6 +1449,39 @@ The "num_dirty" field is a performance hint for KVM to determine whether it
 should skip processing the bitmap and just invalidate everything.  It must
 be set to the number of set bits in the bitmap.
 
+4.60 KVM_ASSIGN_SET_INTX_MASK
+
+Capability: KVM_CAP_PCI_2_3
+Architectures: x86
+Type: vm ioctl
+Parameters: struct kvm_assigned_pci_dev (in)
+Returns: 0 on success, -1 on error
+
+Allows userspace to mask PCI INTx interrupts from the assigned device.  The
+kernel will not deliver INTx interrupts to the guest between setting and
+clearing of KVM_ASSIGN_SET_INTX_MASK via this interface.  This enables use of
+and emulation of PCI 2.3 INTx disable command register behavior.
+
+This may be used for both PCI 2.3 devices supporting INTx disable natively and
+older devices lacking this support. Userspace is responsible for emulating the
+read value of the INTx disable bit in the guest visible PCI command register.
+When modifying the INTx disable state, userspace should precede updating the
+physical device command register by calling this ioctl to inform the kernel of
+the new intended INTx mask state.
+
+Note that the kernel uses the device INTx disable bit to internally manage the
+device interrupt state for PCI 2.3 devices.  Reads of this register may
+therefore not match the expected value.  Writes should always use the guest
+intended INTx disable value rather than attempting to read-copy-update the
+current physical device state.  Races between user and kernel updates to the
+INTx disable bit are handled lazily in the kernel.  It's possible the device
+may generate unintended interrupts, but they will not be injected into the
+guest.
+
+See KVM_ASSIGN_DEV_IRQ for the data structure.  The target device is specified
+by assigned_dev_id.  In the flags field, only KVM_DEV_ASSIGN_MASK_INTX is
+evaluated.
+
 4.62 KVM_CREATE_SPAPR_TCE
 
 Capability: KVM_CAP_SPAPR_TCE
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 9477dc6cccae..6866083a48c1 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2143,6 +2143,7 @@ int kvm_dev_ioctl_check_extension(long ext)
         case KVM_CAP_XSAVE:
         case KVM_CAP_ASYNC_PF:
         case KVM_CAP_GET_TSC_KHZ:
+        case KVM_CAP_PCI_2_3:
                 r = 1;
                 break;
         case KVM_CAP_COALESCED_MMIO:
diff --git a/include/linux/kvm.h b/include/linux/kvm.h
index acbe42939089..6c322a90b92f 100644
--- a/include/linux/kvm.h
+++ b/include/linux/kvm.h
@@ -588,6 +588,7 @@ struct kvm_ppc_pvinfo {
 #define KVM_CAP_TSC_DEADLINE_TIMER 72
 #define KVM_CAP_S390_UCONTROL 73
 #define KVM_CAP_SYNC_REGS 74
+#define KVM_CAP_PCI_2_3 75
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
@@ -784,6 +785,9 @@ struct kvm_s390_ucas_mapping {
 /* Available with KVM_CAP_TSC_CONTROL */
 #define KVM_SET_TSC_KHZ           _IO(KVMIO,  0xa2)
 #define KVM_GET_TSC_KHZ           _IO(KVMIO,  0xa3)
+/* Available with KVM_CAP_PCI_2_3 */
+#define KVM_ASSIGN_SET_INTX_MASK  _IOW(KVMIO,  0xa4, \
+                                       struct kvm_assigned_pci_dev)
 
 /*
  * ioctls for vcpu fds
@@ -857,6 +861,8 @@ struct kvm_s390_ucas_mapping {
 #define KVM_SET_ONE_REG           _IOW(KVMIO,  0xac, struct kvm_one_reg)
 
 #define KVM_DEV_ASSIGN_ENABLE_IOMMU     (1 << 0)
+#define KVM_DEV_ASSIGN_PCI_2_3          (1 << 1)
+#define KVM_DEV_ASSIGN_MASK_INTX        (1 << 2)
 
 struct kvm_assigned_pci_dev {
         __u32 assigned_dev_id;
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index e42d85ae8541..ec171c1d0878 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -546,6 +546,7 @@ struct kvm_assigned_dev_kernel {
         unsigned int entries_nr;
         int host_irq;
         bool host_irq_disabled;
+        bool pci_2_3;
         struct msix_entry *host_msix_entries;
         int guest_irq;
         struct msix_entry *guest_msix_entries;
@@ -555,6 +556,7 @@ struct kvm_assigned_dev_kernel {
         struct pci_dev *dev;
         struct kvm *kvm;
         spinlock_t intx_lock;
+        struct mutex intx_mask_lock;
         char irq_name[32];
         struct pci_saved_state *pci_saved_state;
 };
diff --git a/virt/kvm/assigned-dev.c b/virt/kvm/assigned-dev.c
index ece80612b594..08e05715df72 100644
--- a/virt/kvm/assigned-dev.c
+++ b/virt/kvm/assigned-dev.c
@@ -55,22 +55,66 @@ static int find_index_from_host_irq(struct kvm_assigned_dev_kernel
         return index;
 }
 
-static irqreturn_t kvm_assigned_dev_thread(int irq, void *dev_id)
+static irqreturn_t kvm_assigned_dev_intx(int irq, void *dev_id)
 {
         struct kvm_assigned_dev_kernel *assigned_dev = dev_id;
+        int ret;
 
-        if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_INTX) {
-                spin_lock(&assigned_dev->intx_lock);
+        spin_lock(&assigned_dev->intx_lock);
+        if (pci_check_and_mask_intx(assigned_dev->dev)) {
+                assigned_dev->host_irq_disabled = true;
+                ret = IRQ_WAKE_THREAD;
+        } else
+                ret = IRQ_NONE;
+        spin_unlock(&assigned_dev->intx_lock);
+
+        return ret;
+}
+
+static void
+kvm_assigned_dev_raise_guest_irq(struct kvm_assigned_dev_kernel *assigned_dev,
+                                 int vector)
+{
+        if (unlikely(assigned_dev->irq_requested_type &
+                     KVM_DEV_IRQ_GUEST_INTX)) {
+                mutex_lock(&assigned_dev->intx_mask_lock);
+                if (!(assigned_dev->flags & KVM_DEV_ASSIGN_MASK_INTX))
+                        kvm_set_irq(assigned_dev->kvm,
+                                    assigned_dev->irq_source_id, vector, 1);
+                mutex_unlock(&assigned_dev->intx_mask_lock);
+        } else
+                kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id,
+                            vector, 1);
+}
+
+static irqreturn_t kvm_assigned_dev_thread_intx(int irq, void *dev_id)
+{
+        struct kvm_assigned_dev_kernel *assigned_dev = dev_id;
+
+        if (!(assigned_dev->flags & KVM_DEV_ASSIGN_PCI_2_3)) {
+                spin_lock_irq(&assigned_dev->intx_lock);
                 disable_irq_nosync(irq);
                 assigned_dev->host_irq_disabled = true;
-                spin_unlock(&assigned_dev->intx_lock);
+                spin_unlock_irq(&assigned_dev->intx_lock);
         }
 
-        kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id,
-                    assigned_dev->guest_irq, 1);
+        kvm_assigned_dev_raise_guest_irq(assigned_dev,
+                                         assigned_dev->guest_irq);
+
+        return IRQ_HANDLED;
+}
+