[Qemu-devel] [PATCH 8/8] vhost: backend masking support

[Michael S. Tsirkin]

Support backend guest notifier masking in vhost-net:
create eventfd at device init, when masked,
make vhost use that as eventfd instead of
sending an interrupt.

Signed-off-by: Michael S. Tsirkin <address@hidden>
---
 hw/vhost.c      | 95 +++++++++++++++++++++++++++++++++++++++++++++++++--------
 hw/vhost.h      | 10 ++++++
 hw/vhost_net.c  | 27 ++++++++++++++--
 hw/vhost_net.h  |  3 ++
 hw/virtio-net.c | 18 +++++++++++
 5 files changed, 137 insertions(+), 16 deletions(-)

diff --git a/hw/vhost.c b/hw/vhost.c
index 4fa5007..cee8aad 100644
--- a/hw/vhost.c
+++ b/hw/vhost.c
@@ -612,7 +612,7 @@ static void vhost_log_stop(MemoryListener *listener,
     /* FIXME: implement */
 }
 
-static int vhost_virtqueue_init(struct vhost_dev *dev,
+static int vhost_virtqueue_start(struct vhost_dev *dev,
                                 struct VirtIODevice *vdev,
                                 struct vhost_virtqueue *vq,
                                 unsigned idx)
@@ -681,16 +681,11 @@ static int vhost_virtqueue_init(struct vhost_dev *dev,
         goto fail_kick;
     }
 
-    file.fd = event_notifier_get_fd(virtio_queue_get_guest_notifier(vvq));
-    r = ioctl(dev->control, VHOST_SET_VRING_CALL, &file);
-    if (r) {
-        r = -errno;
-        goto fail_call;
-    }
+    /* Clear and discard previous events if any. */
+    event_notifier_test_and_clear(&vq->masked_notifier);
 
     return 0;
 
-fail_call:
 fail_kick:
 fail_alloc:
     cpu_physical_memory_unmap(vq->ring, virtio_queue_get_ring_size(vdev, idx),
@@ -708,7 +703,7 @@ fail_alloc_desc:
     return r;
 }
 
-static void vhost_virtqueue_cleanup(struct vhost_dev *dev,
+static void vhost_virtqueue_stop(struct vhost_dev *dev,
                                     struct VirtIODevice *vdev,
                                     struct vhost_virtqueue *vq,
                                     unsigned idx)
@@ -746,11 +741,39 @@ static void vhost_eventfd_del(MemoryListener *listener,
 {
 }
 
+static int vhost_virtqueue_init(struct vhost_dev *dev,
+                                struct vhost_virtqueue *vq, int n)
+{
+    struct vhost_vring_file file = {
+        .index = n,
+    };
+    int r = event_notifier_init(&vq->masked_notifier, 0);
+    if (r < 0) {
+        return r;
+    }
+
+    file.fd = event_notifier_get_fd(&vq->masked_notifier);
+    r = ioctl(dev->control, VHOST_SET_VRING_CALL, &file);
+    if (r) {
+        r = -errno;
+        goto fail_call;
+    }
+    return 0;
+fail_call:
+    event_notifier_cleanup(&vq->masked_notifier);
+    return r;
+}
+
+static void vhost_virtqueue_cleanup(struct vhost_virtqueue *vq)
+{
+    event_notifier_cleanup(&vq->masked_notifier);
+}
+
 int vhost_dev_init(struct vhost_dev *hdev, int devfd, const char *devpath,
                    bool force)
 {
     uint64_t features;
-    int r;
+    int i, r;
     if (devfd >= 0) {
         hdev->control = devfd;
     } else {
@@ -768,6 +791,13 @@ int vhost_dev_init(struct vhost_dev *hdev, int devfd, 
const char *devpath,
     if (r < 0) {
         goto fail;
     }
+
+    for (i = 0; i < hdev->nvqs; ++i) {
+        r = vhost_virtqueue_init(hdev, hdev->vqs + i, i);
+        if (r < 0) {
+            goto fail_vq;
+        }
+    }
     hdev->features = features;
 
     hdev->memory_listener = (MemoryListener) {
@@ -795,6 +825,10 @@ int vhost_dev_init(struct vhost_dev *hdev, int devfd, 
const char *devpath,
     memory_listener_register(&hdev->memory_listener, &address_space_memory);
     hdev->force = force;
     return 0;
+fail_vq:
+    while (--i >= 0) {
+        vhost_virtqueue_cleanup(hdev->vqs + i);
+    }
 fail:
     r = -errno;
     close(hdev->control);
@@ -803,6 +837,10 @@ fail:
 
 void vhost_dev_cleanup(struct vhost_dev *hdev)
 {
+    int i;
+    for (i = 0; i < hdev->nvqs; ++i) {
+        vhost_virtqueue_cleanup(hdev->vqs + i);
+    }
     memory_listener_unregister(&hdev->memory_listener);
     g_free(hdev->mem);
     g_free(hdev->mem_sections);
@@ -869,6 +907,37 @@ void vhost_dev_disable_notifiers(struct vhost_dev *hdev, 
VirtIODevice *vdev)
     }
 }
 
+/* Test and clear event pending status.
+ * Should be called after unmask to avoid losing events.
+ */
+bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n)
+{
+    struct vhost_virtqueue *vq = hdev->vqs + n;
+    assert(hdev->started);
+    return event_notifier_test_and_clear(&vq->masked_notifier);
+}
+
+/* Mask/unmask events from this vq. */
+void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n,
+                         bool mask)
+{
+    struct VirtQueue *vvq = virtio_get_queue(vdev, n);
+    int r;
+
+    assert(hdev->started);
+
+    struct vhost_vring_file file = {
+        .index = n,
+    };
+    if (mask) {
+        file.fd = event_notifier_get_fd(&hdev->vqs[n].masked_notifier);
+    } else {
+        file.fd = event_notifier_get_fd(virtio_queue_get_guest_notifier(vvq));
+    }
+    r = ioctl(hdev->control, VHOST_SET_VRING_CALL, &file);
+    assert(r >= 0);
+}
+
 /* Host notifiers must be enabled at this point. */
 int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev)
 {
@@ -900,7 +969,7 @@ int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice 
*vdev)
         goto fail_mem;
     }
     for (i = 0; i < hdev->nvqs; ++i) {
-        r = vhost_virtqueue_init(hdev,
+        r = vhost_virtqueue_start(hdev,
                                  vdev,
                                  hdev->vqs + i,
                                  i);
@@ -925,7 +994,7 @@ int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice 
*vdev)
 fail_log:
 fail_vq:
     while (--i >= 0) {
-        vhost_virtqueue_cleanup(hdev,
+        vhost_virtqueue_stop(hdev,
                                 vdev,
                                 hdev->vqs + i,
                                 i);
@@ -946,7 +1015,7 @@ void vhost_dev_stop(struct vhost_dev *hdev, VirtIODevice 
*vdev)
     int i, r;
 
     for (i = 0; i < hdev->nvqs; ++i) {
-        vhost_virtqueue_cleanup(hdev,
+        vhost_virtqueue_stop(hdev,
                                 vdev,
                                 hdev->vqs + i,
                                 i);
diff --git a/hw/vhost.h b/hw/vhost.h
index 6f6a906..44c61a5 100644
--- a/hw/vhost.h
+++ b/hw/vhost.h
@@ -18,6 +18,7 @@ struct vhost_virtqueue {
     void *ring;
     unsigned long long ring_phys;
     unsigned ring_size;
+    EventNotifier masked_notifier;
 };
 
 typedef unsigned long vhost_log_chunk_t;
@@ -53,4 +54,13 @@ void vhost_dev_stop(struct vhost_dev *hdev, VirtIODevice 
*vdev);
 int vhost_dev_enable_notifiers(struct vhost_dev *hdev, VirtIODevice *vdev);
 void vhost_dev_disable_notifiers(struct vhost_dev *hdev, VirtIODevice *vdev);
 
+/* Test and clear masked event pending status.
+ * Should be called after unmask to avoid losing events.
+ */
+bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n);
+
+/* Mask/unmask events from this vq.
+ */
+void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n,
+                          bool mask);
 #endif
diff --git a/hw/vhost_net.c b/hw/vhost_net.c
index ae2785d..d3a04ca 100644
--- a/hw/vhost_net.c
+++ b/hw/vhost_net.c
@@ -109,6 +109,9 @@ struct vhost_net *vhost_net_init(NetClientState *backend, 
int devfd,
         (1 << VHOST_NET_F_VIRTIO_NET_HDR);
     net->backend = r;
 
+    net->dev.nvqs = 2;
+    net->dev.vqs = net->vqs;
+
     r = vhost_dev_init(&net->dev, devfd, "/dev/vhost-net", force);
     if (r < 0) {
         goto fail;
@@ -143,9 +146,6 @@ int vhost_net_start(struct vhost_net *net,
     struct vhost_vring_file file = { };
     int r;
 
-    net->dev.nvqs = 2;
-    net->dev.vqs = net->vqs;
-
     r = vhost_dev_enable_notifiers(&net->dev, dev);
     if (r < 0) {
         goto fail_notifiers;
@@ -200,6 +200,17 @@ void vhost_net_cleanup(struct vhost_net *net)
     vhost_dev_cleanup(&net->dev);
     g_free(net);
 }
+
+bool vhost_net_virtqueue_pending(VHostNetState *net, int idx)
+{
+    return vhost_virtqueue_pending(&net->dev, idx);
+}
+
+void vhost_net_virtqueue_mask(VHostNetState *net, VirtIODevice *dev,
+                              int idx, bool mask)
+{
+    vhost_virtqueue_mask(&net->dev, dev, idx, mask);
+}
 #else
 struct vhost_net *vhost_net_init(NetClientState *backend, int devfd,
                                  bool force)
@@ -234,4 +245,14 @@ unsigned vhost_net_get_features(struct vhost_net *net, 
unsigned features)
 void vhost_net_ack_features(struct vhost_net *net, unsigned features)
 {
 }
+
+bool vhost_net_virtqueue_pending(VHostNetState *net, int idx)
+{
+    return -ENOSYS;
+}
+
+void vhost_net_virtqueue_mask(VHostNetState *net, VirtIODevice *dev,
+                              int idx, bool mask)
+{
+}
 #endif
diff --git a/hw/vhost_net.h b/hw/vhost_net.h
index 012aba4..88912b8 100644
--- a/hw/vhost_net.h
+++ b/hw/vhost_net.h
@@ -17,4 +17,7 @@ void vhost_net_cleanup(VHostNetState *net);
 unsigned vhost_net_get_features(VHostNetState *net, unsigned features);
 void vhost_net_ack_features(VHostNetState *net, unsigned features);
 
+bool vhost_net_virtqueue_pending(VHostNetState *net, int n);
+void vhost_net_virtqueue_mask(VHostNetState *net, VirtIODevice *dev,
+                              int idx, bool mask);
 #endif
diff --git a/hw/virtio-net.c b/hw/virtio-net.c
index b756d57..3bb01b1 100644
--- a/hw/virtio-net.c
+++ b/hw/virtio-net.c
@@ -1010,6 +1010,22 @@ static NetClientInfo net_virtio_info = {
     .link_status_changed = virtio_net_set_link_status,
 };
 
+static bool virtio_net_guest_notifier_pending(VirtIODevice *vdev, int idx)
+{
+    VirtIONet *n = to_virtio_net(vdev);
+    assert(n->vhost_started);
+    return vhost_net_virtqueue_pending(tap_get_vhost_net(n->nic->nc.peer), 
idx);
+}
+
+static void virtio_net_guest_notifier_mask(VirtIODevice *vdev, int idx,
+                                           bool mask)
+{
+    VirtIONet *n = to_virtio_net(vdev);
+    assert(n->vhost_started);
+    vhost_net_virtqueue_mask(tap_get_vhost_net(n->nic->nc.peer),
+                             vdev, idx, mask);
+}
+
 VirtIODevice *virtio_net_init(DeviceState *dev, NICConf *conf,
                               virtio_net_conf *net)
 {
@@ -1026,6 +1042,8 @@ VirtIODevice *virtio_net_init(DeviceState *dev, NICConf 
*conf,
     n->vdev.bad_features = virtio_net_bad_features;
     n->vdev.reset = virtio_net_reset;
     n->vdev.set_status = virtio_net_set_status;
+    n->vdev.guest_notifier_mask = virtio_net_guest_notifier_mask;
+    n->vdev.guest_notifier_pending = virtio_net_guest_notifier_pending;
     n->rx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_rx);
 
     if (net->tx && strcmp(net->tx, "timer") && strcmp(net->tx, "bh")) {
-- 
MST