Merge remote-tracking branch 'cohuck/virtio-ccw-upstr' into staging

# By Cornelia Huck (2) and Christian Borntraeger (1)
# Via Cornelia Huck
* cohuck/virtio-ccw-upstr:
  virtio-ccw: Wire up guest and host notifies.
  virtio-ccw: Wire up ioeventfd.
  s390/virtio-ccw: Fix virtio reset

Message-id: 1372177538-9812-1-git-send-email-cornelia.huck@de.ibm.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

hw/s390x/css.c

@@ -68,7 +68,7 @@ int css_create_css_image(uint8_t cssid, bool default_image)
     return 0;
 }
 
-static uint16_t css_build_subchannel_id(SubchDev *sch)
+uint16_t css_build_subchannel_id(SubchDev *sch)
 {
     if (channel_subsys->max_cssid > 0) {
         return (sch->cssid << 8) | (1 << 3) | (sch->ssid << 1) | 1;

hw/s390x/css.h

@@ -90,6 +90,7 @@ bool css_devno_used(uint8_t cssid, uint8_t ssid, uint16_t devno);
 void css_subch_assign(uint8_t cssid, uint8_t ssid, uint16_t schid,
                       uint16_t devno, SubchDev *sch);
 void css_sch_build_virtual_schib(SubchDev *sch, uint8_t chpid, uint8_t type);
+uint16_t css_build_subchannel_id(SubchDev *sch);
 void css_reset(void);
 void css_reset_sch(SubchDev *sch);
 void css_queue_crw(uint8_t rsc, uint8_t erc, int chain, uint16_t rsid);

hw/s390x/virtio-ccw.c

@@ -63,6 +63,91 @@ VirtIODevice *virtio_ccw_get_vdev(SubchDev *sch)
     return vdev;
 }
 
+static int virtio_ccw_set_guest2host_notifier(VirtioCcwDevice *dev, int n,
+                                              bool assign, bool set_handler)
+{
+    VirtQueue *vq = virtio_get_queue(dev->vdev, n);
+    EventNotifier *notifier = virtio_queue_get_host_notifier(vq);
+    int r = 0;
+    SubchDev *sch = dev->sch;
+    uint32_t sch_id = (css_build_subchannel_id(sch) << 16) | sch->schid;
+
+    if (assign) {
+        r = event_notifier_init(notifier, 1);
+        if (r < 0) {
+            error_report("%s: unable to init event notifier: %d", __func__, r);
+            return r;
+        }
+        virtio_queue_set_host_notifier_fd_handler(vq, true, set_handler);
+        r = s390_assign_subch_ioeventfd(event_notifier_get_fd(notifier), sch_id,
+                                        n, assign);
+        if (r < 0) {
+            error_report("%s: unable to assign ioeventfd: %d", __func__, r);
+            virtio_queue_set_host_notifier_fd_handler(vq, false, false);
+            event_notifier_cleanup(notifier);
+            return r;
+        }
+    } else {
+        virtio_queue_set_host_notifier_fd_handler(vq, false, false);
+        s390_assign_subch_ioeventfd(event_notifier_get_fd(notifier), sch_id,
+                                    n, assign);
+        event_notifier_cleanup(notifier);
+    }
+    return r;
+}
+
+static void virtio_ccw_start_ioeventfd(VirtioCcwDevice *dev)
+{
+    int n, r;
+
+    if (!(dev->flags & VIRTIO_CCW_FLAG_USE_IOEVENTFD) ||
+        dev->ioeventfd_disabled ||
+        dev->ioeventfd_started) {
+        return;
+    }
+    for (n = 0; n < VIRTIO_PCI_QUEUE_MAX; n++) {
+        if (!virtio_queue_get_num(dev->vdev, n)) {
+            continue;
+        }
+        r = virtio_ccw_set_guest2host_notifier(dev, n, true, true);
+        if (r < 0) {
+            goto assign_error;
+        }
+    }
+    dev->ioeventfd_started = true;
+    return;
+
+  assign_error:
+    while (--n >= 0) {
+        if (!virtio_queue_get_num(dev->vdev, n)) {
+            continue;
+        }
+        r = virtio_ccw_set_guest2host_notifier(dev, n, false, false);
+        assert(r >= 0);
+    }
+    dev->ioeventfd_started = false;
+    /* Disable ioeventfd for this device. */
+    dev->flags &= ~VIRTIO_CCW_FLAG_USE_IOEVENTFD;
+    error_report("%s: failed. Fallback to userspace (slower).", __func__);
+}
+
+static void virtio_ccw_stop_ioeventfd(VirtioCcwDevice *dev)
+{
+    int n, r;
+
+    if (!dev->ioeventfd_started) {
+        return;
+    }
+    for (n = 0; n < VIRTIO_PCI_QUEUE_MAX; n++) {
+        if (!virtio_queue_get_num(dev->vdev, n)) {
+            continue;
+        }
+        r = virtio_ccw_set_guest2host_notifier(dev, n, false, false);
+        assert(r >= 0);
+    }
+    dev->ioeventfd_started = false;
+}
+
 VirtualCssBus *virtual_css_bus_init(void)
 {
     VirtualCssBus *cbus;
@@ -187,6 +272,7 @@ static int virtio_ccw_cb(SubchDev *sch, CCW1 ccw)
         }
         break;
     case CCW_CMD_VDEV_RESET:
+        virtio_ccw_stop_ioeventfd(dev);
         virtio_reset(dev->vdev);
         ret = 0;
         break;
@@ -309,10 +395,16 @@ static int virtio_ccw_cb(SubchDev *sch, CCW1 ccw)
             ret = -EFAULT;
         } else {
             status = ldub_phys(ccw.cda);
+            if (!(status & VIRTIO_CONFIG_S_DRIVER_OK)) {
+                virtio_ccw_stop_ioeventfd(dev);
+            }
             virtio_set_status(dev->vdev, status);
             if (dev->vdev->status == 0) {
                 virtio_reset(dev->vdev);
             }
+            if (status & VIRTIO_CONFIG_S_DRIVER_OK) {
+                virtio_ccw_start_ioeventfd(dev);
+            }
             sch->curr_status.scsw.count = ccw.count - sizeof(status);
             ret = 0;
         }
@@ -540,6 +632,7 @@ static int virtio_ccw_exit(VirtioCcwDevice *dev)
 {
     SubchDev *sch = dev->sch;
 
+    virtio_ccw_stop_ioeventfd(dev);
     if (sch) {
         css_subch_assign(sch->cssid, sch->ssid, sch->schid, sch->devno, NULL);
         g_free(sch);
@@ -801,8 +894,106 @@ static void virtio_ccw_reset(DeviceState *d)
 {
     VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
 
+    virtio_ccw_stop_ioeventfd(dev);
     virtio_reset(dev->vdev);
     css_reset_sch(dev->sch);
+    dev->indicators = 0;
+    dev->indicators2 = 0;
+}
+
+static void virtio_ccw_vmstate_change(DeviceState *d, bool running)
+{
+    VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
+
+    if (running) {
+        virtio_ccw_start_ioeventfd(dev);
+    } else {
+        virtio_ccw_stop_ioeventfd(dev);
+    }
+}
+
+static bool virtio_ccw_query_guest_notifiers(DeviceState *d)
+{
+    VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
+
+    return !!(dev->sch->curr_status.pmcw.flags & PMCW_FLAGS_MASK_ENA);
+}
+
+static int virtio_ccw_set_host_notifier(DeviceState *d, int n, bool assign)
+{
+    VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
+
+    /* Stop using the generic ioeventfd, we are doing eventfd handling
+     * ourselves below */
+    dev->ioeventfd_disabled = assign;
+    if (assign) {
+        virtio_ccw_stop_ioeventfd(dev);
+    }
+    return virtio_ccw_set_guest2host_notifier(dev, n, assign, false);
+}
+
+static int virtio_ccw_set_guest_notifier(VirtioCcwDevice *dev, int n,
+                                         bool assign, bool with_irqfd)
+{
+    VirtQueue *vq = virtio_get_queue(dev->vdev, n);
+    EventNotifier *notifier = virtio_queue_get_guest_notifier(vq);
+    VirtioDeviceClass *k = VIRTIO_DEVICE_GET_CLASS(dev->vdev);
+
+    if (assign) {
+        int r = event_notifier_init(notifier, 0);
+
+        if (r < 0) {
+            return r;
+        }
+        virtio_queue_set_guest_notifier_fd_handler(vq, true, with_irqfd);
+        /* We do not support irqfd for classic I/O interrupts, because the
+         * classic interrupts are intermixed with the subchannel status, that
+         * is queried with test subchannel. We want to use vhost, though.
+         * Lets make sure to have vhost running and wire up the irq fd to
+         * land in qemu (and only the irq fd) in this code.
+         */
+        if (k->guest_notifier_mask) {
+            k->guest_notifier_mask(dev->vdev, n, false);
+        }
+        /* get lost events and re-inject */
+        if (k->guest_notifier_pending &&
+            k->guest_notifier_pending(dev->vdev, n)) {
+            event_notifier_set(notifier);
+        }
+    } else {
+        if (k->guest_notifier_mask) {
+            k->guest_notifier_mask(dev->vdev, n, true);
+        }
+        virtio_queue_set_guest_notifier_fd_handler(vq, false, with_irqfd);
+        event_notifier_cleanup(notifier);
+    }
+    return 0;
+}
+
+static int virtio_ccw_set_guest_notifiers(DeviceState *d, int nvqs,
+                                          bool assigned)
+{
+    VirtioCcwDevice *dev = VIRTIO_CCW_DEVICE(d);
+    VirtIODevice *vdev = dev->vdev;
+    int r, n;
+
+    for (n = 0; n < nvqs; n++) {
+        if (!virtio_queue_get_num(vdev, n)) {
+            break;
+        }
+        /* false -> true, as soon as irqfd works */
+        r = virtio_ccw_set_guest_notifier(dev, n, assigned, false);
+        if (r < 0) {
+            goto assign_error;
+        }
+    }
+    return 0;
+
+assign_error:
+    while (--n >= 0) {
+        virtio_ccw_set_guest_notifier(dev, n, !assigned, false);
+    }
+    return r;
 }
 
 /**************** Virtio-ccw Bus Device Descriptions *******************/
@@ -812,6 +1003,8 @@ static Property virtio_ccw_net_properties[] = {
     DEFINE_VIRTIO_NET_FEATURES(VirtioCcwDevice, host_features[0]),
     DEFINE_VIRTIO_NET_PROPERTIES(VirtIONetCcw, vdev.net_conf),
     DEFINE_NIC_PROPERTIES(VirtIONetCcw, vdev.nic_conf),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -838,6 +1031,8 @@ static Property virtio_ccw_blk_properties[] = {
     DEFINE_PROP_STRING("devno", VirtioCcwDevice, bus_id),
     DEFINE_VIRTIO_BLK_FEATURES(VirtioCcwDevice, host_features[0]),
     DEFINE_VIRTIO_BLK_PROPERTIES(VirtIOBlkCcw, blk),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -864,6 +1059,8 @@ static Property virtio_ccw_serial_properties[] = {
     DEFINE_PROP_STRING("devno", VirtioCcwDevice, bus_id),
     DEFINE_VIRTIO_SERIAL_PROPERTIES(VirtioSerialCcw, vdev.serial),
     DEFINE_VIRTIO_COMMON_FEATURES(VirtioCcwDevice, host_features[0]),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -889,6 +1086,8 @@ static const TypeInfo virtio_ccw_serial = {
 static Property virtio_ccw_balloon_properties[] = {
     DEFINE_PROP_STRING("devno", VirtioCcwDevice, bus_id),
     DEFINE_VIRTIO_COMMON_FEATURES(VirtioCcwDevice, host_features[0]),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -915,6 +1114,8 @@ static Property virtio_ccw_scsi_properties[] = {
     DEFINE_PROP_STRING("devno", VirtioCcwDevice, bus_id),
     DEFINE_VIRTIO_SCSI_PROPERTIES(VirtIOSCSICcw, vdev.parent_obj.conf),
     DEFINE_VIRTIO_SCSI_FEATURES(VirtioCcwDevice, host_features[0]),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -978,6 +1179,8 @@ static Property virtio_ccw_rng_properties[] = {
     DEFINE_PROP_STRING("devno", VirtioCcwDevice, bus_id),
     DEFINE_VIRTIO_COMMON_FEATURES(VirtioCcwDevice, host_features[0]),
     DEFINE_VIRTIO_RNG_PROPERTIES(VirtIORNGCcw, vdev.conf),
+    DEFINE_PROP_BIT("ioeventfd", VirtioCcwDevice, flags,
+                    VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT, true),
     DEFINE_PROP_END_OF_LIST(),
 };
 
@@ -1105,6 +1308,10 @@ static void virtio_ccw_bus_class_init(ObjectClass *klass, void *data)
     bus_class->max_dev = 1;
     k->notify = virtio_ccw_notify;
     k->get_features = virtio_ccw_get_features;
+    k->vmstate_change = virtio_ccw_vmstate_change;
+    k->query_guest_notifiers = virtio_ccw_query_guest_notifiers;
+    k->set_host_notifier = virtio_ccw_set_host_notifier;
+    k->set_guest_notifiers = virtio_ccw_set_guest_notifiers;
 }
 
 static const TypeInfo virtio_ccw_bus_info = {

hw/s390x/virtio-ccw.h

@@ -69,6 +69,11 @@ typedef struct VirtIOCCWDeviceClass {
 /* Change here if we want to support more feature bits. */
 #define VIRTIO_CCW_FEATURE_SIZE 1
 
+/* Performance improves when virtqueue kick processing is decoupled from the
+ * vcpu thread using ioeventfd for some devices. */
+#define VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT 1
+#define VIRTIO_CCW_FLAG_USE_IOEVENTFD   (1 << VIRTIO_CCW_FLAG_USE_IOEVENTFD_BIT)
+
 struct VirtioCcwDevice {
     DeviceState parent_obj;
     SubchDev *sch;
@@ -76,6 +81,9 @@ struct VirtioCcwDevice {
     char *bus_id;
     uint32_t host_features[VIRTIO_CCW_FEATURE_SIZE];
     VirtioBusState bus;
+    bool ioeventfd_started;
+    bool ioeventfd_disabled;
+    uint32_t flags;
     /* Guest provided values: */
     hwaddr indicators;
     hwaddr indicators2;

target-s390x/cpu.h

@@ -1081,6 +1081,7 @@ void kvm_s390_io_interrupt(S390CPU *cpu, uint16_t subchannel_id,
 void kvm_s390_crw_mchk(S390CPU *cpu);
 void kvm_s390_enable_css_support(S390CPU *cpu);
 int kvm_s390_get_registers_partial(CPUState *cpu);
+int kvm_s390_assign_subch_ioeventfd(int fd, uint32_t sch, int vq, bool assign);
 #else
 static inline void kvm_s390_io_interrupt(S390CPU *cpu,
                                         uint16_t subchannel_id,
@@ -1099,6 +1100,11 @@ static inline int kvm_s390_get_registers_partial(CPUState *cpu)
 {
     return -ENOSYS;
 }
+static inline int kvm_s390_assign_subch_ioeventfd(int fd, uint32_t sch, int vq,
+                                                  bool assign)
+{
+    return -ENOSYS;
+}
 #endif
 
 static inline void s390_io_interrupt(S390CPU *cpu,
@@ -1125,4 +1131,14 @@ static inline void s390_crw_mchk(S390CPU *cpu)
     }
 }
 
+static inline int s390_assign_subch_ioeventfd(int fd, uint32_t sch_id, int vq,
+                                              bool assign)
+{
+    if (kvm_enabled()) {
+        return kvm_s390_assign_subch_ioeventfd(fd, sch_id, vq, assign);
+    } else {
+        return -ENOSYS;
+    }
+}
+
 #endif

target-s390x/kvm.c

@@ -932,3 +932,22 @@ void kvm_s390_enable_css_support(S390CPU *cpu)
 void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
+
+int kvm_s390_assign_subch_ioeventfd(int fd, uint32_t sch, int vq, bool assign)
+{
+    struct kvm_ioeventfd kick = {
+        .flags = KVM_IOEVENTFD_FLAG_VIRTIO_CCW_NOTIFY |
+        KVM_IOEVENTFD_FLAG_DATAMATCH,
+        .fd = fd,
+        .datamatch = vq,
+        .addr = sch,
+        .len = 8,
+    };
+    if (!kvm_check_extension(kvm_state, KVM_CAP_IOEVENTFD)) {
+        return -ENOSYS;
+    }
+    if (!assign) {
+        kick.flags |= KVM_IOEVENTFD_FLAG_DEASSIGN;
+    }
+    return kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
+}