virtio: VIRTIO_F_NOTIFY_ON_EMPTY to force callback on empty

diff -r 95a02f0e0e21 drivers/virtio/virtio_ring.c
@@ -253,13 +253,6 @@ irqreturn_t vring_interrupt(int irq, voi
 	if (unlikely(vq->broken))
 		return IRQ_HANDLED;
 
-	/* Other side may have missed us turning off the interrupt,
-	 * but we should preserve disable semantic for virtio users. */
-	if (unlikely(vq->vring.avail->flags & VRING_AVAIL_F_NO_INTERRUPT)) {
-		pr_debug("virtqueue interrupt after disable for %p\n", vq);
-		return IRQ_HANDLED;
-	}
-
 	pr_debug("virtqueue callback for %p (%p)\n", vq, vq->vq.callback);
 	if (vq->vq.callback)
 		vq->vq.callback(&vq->vq);

diff -r 95a02f0e0e21 include/linux/virtio_config.h
@@ -14,6 +14,10 @@
 #define VIRTIO_CONFIG_S_DRIVER_OK	4
 /* We've given up on this device. */
 #define VIRTIO_CONFIG_S_FAILED		0x80
+
+/* Do we get callbacks when the ring is completely used, even if we've
+ * suppressed them? */
+#define VIRTIO_F_NOTIFY_ON_EMPTY	24
 
 #ifdef __KERNEL__
 #include <linux/virtio.h>

diff -r a2e94bbd1589 Documentation/lguest/lguest.c
@@ -157,6 +157,9 @@ struct virtqueue
 
 	/* The routine to call when the Guest pings us. */
 	void (*handle_output)(int fd, struct virtqueue *me);
+
+	/* Outstanding buffers */
+	unsigned int inflight;
 };
 
 /* Remember the arguments to the program so we can "reboot" */

@@ -702,6 +705,7 @@ static unsigned get_vq_desc(struct virtq
 			errx(1, "Looped descriptor");
 	} while ((i = next_desc(vq, i)) != vq->vring.num);
 
+	vq->inflight++;
 	return head;
 }
 

@@ -719,6 +723,7 @@ static void add_used(struct virtqueue *v
 	/* Make sure buffer is written before we update index. */
 	wmb();
 	vq->vring.used->idx++;
+	vq->inflight--;
 }
 
 /* This actually sends the interrupt for this virtqueue */

@@ -726,8 +731,9 @@ static void trigger_irq(int fd, struct v
 {
 	unsigned long buf[] = { LHREQ_IRQ, vq->config.irq };
 
-	/* If they don't want an interrupt, don't send one. */
-	if (vq->vring.avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+	/* If they don't want an interrupt, don't send one, unless empty. */
+	if ((vq->vring.avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+	    && vq->inflight)
 		return;
 
 	/* Send the Guest an interrupt tell them we used something up. */

@@ -1107,6 +1113,7 @@ static void add_virtqueue(struct device 
 	vq->next = NULL;
 	vq->last_avail_idx = 0;
 	vq->dev = dev;
+	vq->inflight = 0;
 
 	/* Initialize the configuration. */
 	vq->config.num = num_descs;

@@ -1368,6 +1375,7 @@ static void setup_tun_net(const char *ar
 
 	/* Tell Guest what MAC address to use. */
 	add_feature(dev, VIRTIO_NET_F_MAC);
+	add_feature(dev, VIRTIO_F_NOTIFY_ON_EMPTY);
 	set_config(dev, sizeof(conf), &conf);
 
 	/* We don't need the socket any more; setup is done. */

diff -r a2e94bbd1589 Documentation/lguest/lguest.c
@@ -157,6 +157,9 @@ struct virtqueue
 
 	/* The routine to call when the Guest pings us. */
 	void (*handle_output)(int fd, struct virtqueue *me);
+
+	/* Outstanding buffers */
+	unsigned int inflight;
 };
 
 /* Remember the arguments to the program so we can "reboot" */

@@ -702,6 +705,7 @@ static unsigned get_vq_desc(struct virtq
 			errx(1, "Looped descriptor");
 	} while ((i = next_desc(vq, i)) != vq->vring.num);
 
+	vq->inflight++;
 	return head;
 }
 

@@ -719,6 +723,7 @@ static void add_used(struct virtqueue *v
 	/* Make sure buffer is written before we update index. */
 	wmb();
 	vq->vring.used->idx++;
+	vq->inflight--;
 }
 
 /* This actually sends the interrupt for this virtqueue */

@@ -726,8 +731,9 @@ static void trigger_irq(int fd, struct v
 {
 	unsigned long buf[] = { LHREQ_IRQ, vq->config.irq };
 
-	/* If they don't want an interrupt, don't send one. */
-	if (vq->vring.avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+	/* If they don't want an interrupt, don't send one, unless empty. */
+	if ((vq->vring.avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
+	    && vq->inflight)
 		return;
 
 	/* Send the Guest an interrupt tell them we used something up. */

@@ -1107,6 +1113,7 @@ static void add_virtqueue(struct device 
 	vq->next = NULL;
 	vq->last_avail_idx = 0;
 	vq->dev = dev;
+	vq->inflight = 0;
 
 	/* Initialize the configuration. */
 	vq->config.num = num_descs;

@@ -1368,6 +1375,7 @@ static void setup_tun_net(const char *ar
 
 	/* Tell Guest what MAC address to use. */
 	add_feature(dev, VIRTIO_NET_F_MAC);
+	add_feature(dev, VIRTIO_F_NOTIFY_ON_EMPTY);
 	set_config(dev, sizeof(conf), &conf);
 
 	/* We don't need the socket any more; setup is done. */

diff -r 95a02f0e0e21 drivers/net/virtio_net.c
@@ -44,6 +44,7 @@ struct virtnet_info
 	/* The skb we couldn't send because buffers were full. */
 	struct sk_buff *last_xmit_skb;
 
+	/* If we need to free in a timer, this is it. */
 	struct timer_list xmit_free_timer;
 
 	/* Number of input buffers, and max we've ever had. */

@@ -51,6 +52,7 @@ struct virtnet_info
 
 	/* For cleaning up after transmission. */
 	struct tasklet_struct tasklet;
+	bool free_in_tasklet;
 
 	/* Receive & send queues. */
 	struct sk_buff_head recv;

@@ -74,7 +76,7 @@ static void skb_xmit_done(struct virtque
 	/* Suppress further interrupts. */
 	svq->vq_ops->disable_cb(svq);
 
-	/* We were waiting for more output buffers. */
+	/* We were probably waiting for more output buffers. */
 	netif_wake_queue(vi->dev);
 
 	/* Make sure we re-xmit last_xmit_skb: if there are no more packets

@@ -240,6 +242,8 @@ static void free_old_xmit_skbs(struct vi
 	}
 }
 
+/* If the virtio transport doesn't always notify us when all in-flight packets
+ * are consumed, we fall back to using this function on a timer to free them. */
 static void xmit_free(unsigned long data)
 {
 	struct virtnet_info *vi = (void *)data;

@@ -300,7 +304,7 @@ static int xmit_skb(struct virtnet_info 
 	num = skb_to_sgvec(skb, sg+1, 0, skb->len) + 1;
 
 	err = vi->svq->vq_ops->add_buf(vi->svq, sg, num, 0, skb);
-	if (!err)
+	if (!err && !vi->free_in_tasklet)
 		mod_timer(&vi->xmit_free_timer, jiffies + (HZ/10));
 
 	return err;

@@ -315,6 +319,8 @@ static void xmit_tasklet(unsigned long d
 		vi->svq->vq_ops->kick(vi->svq);
 		vi->last_xmit_skb = NULL;
 	}
+	if (vi->free_in_tasklet)
+		free_old_xmit_skbs(vi);
 	netif_tx_unlock_bh(vi->dev);
 }
 

@@ -455,6 +461,10 @@ static int virtnet_probe(struct virtio_d
 	vi->vdev = vdev;
 	vdev->priv = vi;
 
+	/* If they give us a callback when all buffers are done, we don't need
+	 * the timer. */
+	vi->free_in_tasklet = virtio_has_feature(vdev,VIRTIO_F_NOTIFY_ON_EMPTY);
+
 	/* We expect two virtqueues, receive then send. */
 	vi->rvq = vdev->config->find_vq(vdev, 0, skb_recv_done);
 	if (IS_ERR(vi->rvq)) {

@@ -474,7 +487,8 @@ static int virtnet_probe(struct virtio_d
 
 	tasklet_init(&vi->tasklet, xmit_tasklet, (unsigned long)vi);
 
-	setup_timer(&vi->xmit_free_timer, xmit_free, (unsigned long)vi);
+	if (!vi->free_in_tasklet)
+		setup_timer(&vi->xmit_free_timer, xmit_free, (unsigned long)vi);
 
 	err = register_netdev(dev);
 	if (err) {

@@ -513,7 +527,8 @@ static void virtnet_remove(struct virtio
 	/* Stop all the virtqueues. */
 	vdev->config->reset(vdev);
 
-	del_timer_sync(&vi->xmit_free_timer);
+	if (!vi->free_in_tasklet)
+		del_timer_sync(&vi->xmit_free_timer);
 
 	/* Free our skbs in send and recv queues, if any. */
 	while ((skb = __skb_dequeue(&vi->recv)) != NULL) {

@@ -538,7 +553,7 @@ static unsigned int features[] = {
 static unsigned int features[] = {
 	VIRTIO_NET_F_CSUM, VIRTIO_NET_F_GSO, VIRTIO_NET_F_MAC,
 	VIRTIO_NET_F_HOST_TSO4, VIRTIO_NET_F_HOST_UFO, VIRTIO_NET_F_HOST_TSO6,
-	VIRTIO_NET_F_HOST_ECN,
+	VIRTIO_NET_F_HOST_ECN, VIRTIO_F_NOTIFY_ON_EMPTY,
 };
 
 static struct virtio_driver virtio_net = {

diff -r 95a02f0e0e21 drivers/net/virtio_net.c
@@ -44,6 +44,7 @@ struct virtnet_info
 	/* The skb we couldn't send because buffers were full. */
 	struct sk_buff *last_xmit_skb;
 
+	/* If we need to free in a timer, this is it. */
 	struct timer_list xmit_free_timer;
 
 	/* Number of input buffers, and max we've ever had. */

@@ -51,6 +52,7 @@ struct virtnet_info
 
 	/* For cleaning up after transmission. */
 	struct tasklet_struct tasklet;
+	bool free_in_tasklet;
 
 	/* Receive & send queues. */
 	struct sk_buff_head recv;

@@ -74,7 +76,7 @@ static void skb_xmit_done(struct virtque
 	/* Suppress further interrupts. */
 	svq->vq_ops->disable_cb(svq);
 
-	/* We were waiting for more output buffers. */
+	/* We were probably waiting for more output buffers. */
 	netif_wake_queue(vi->dev);
 
 	/* Make sure we re-xmit last_xmit_skb: if there are no more packets

@@ -240,6 +242,8 @@ static void free_old_xmit_skbs(struct vi
 	}
 }
 
+/* If the virtio transport doesn't always notify us when all in-flight packets
+ * are consumed, we fall back to using this function on a timer to free them. */
 static void xmit_free(unsigned long data)
 {
 	struct virtnet_info *vi = (void *)data;

@@ -300,7 +304,7 @@ static int xmit_skb(struct virtnet_info 
 	num = skb_to_sgvec(skb, sg+1, 0, skb->len) + 1;
 
 	err = vi->svq->vq_ops->add_buf(vi->svq, sg, num, 0, skb);
-	if (!err)
+	if (!err && !vi->free_in_tasklet)
 		mod_timer(&vi->xmit_free_timer, jiffies + (HZ/10));
 
 	return err;

@@ -315,6 +319,8 @@ static void xmit_tasklet(unsigned long d
 		vi->svq->vq_ops->kick(vi->svq);
 		vi->last_xmit_skb = NULL;
 	}
+	if (vi->free_in_tasklet)
+		free_old_xmit_skbs(vi);
 	netif_tx_unlock_bh(vi->dev);
 }
 

@@ -455,6 +461,10 @@ static int virtnet_probe(struct virtio_d
 	vi->vdev = vdev;
 	vdev->priv = vi;
 
+	/* If they give us a callback when all buffers are done, we don't need
+	 * the timer. */
+	vi->free_in_tasklet = virtio_has_feature(vdev,VIRTIO_F_NOTIFY_ON_EMPTY);
+
 	/* We expect two virtqueues, receive then send. */
 	vi->rvq = vdev->config->find_vq(vdev, 0, skb_recv_done);
 	if (IS_ERR(vi->rvq)) {

@@ -474,7 +487,8 @@ static int virtnet_probe(struct virtio_d
 
 	tasklet_init(&vi->tasklet, xmit_tasklet, (unsigned long)vi);
 
-	setup_timer(&vi->xmit_free_timer, xmit_free, (unsigned long)vi);
+	if (!vi->free_in_tasklet)
+		setup_timer(&vi->xmit_free_timer, xmit_free, (unsigned long)vi);
 
 	err = register_netdev(dev);
 	if (err) {

@@ -513,7 +527,8 @@ static void virtnet_remove(struct virtio
 	/* Stop all the virtqueues. */
 	vdev->config->reset(vdev);
 
-	del_timer_sync(&vi->xmit_free_timer);
+	if (!vi->free_in_tasklet)
+		del_timer_sync(&vi->xmit_free_timer);
 
 	/* Free our skbs in send and recv queues, if any. */
 	while ((skb = __skb_dequeue(&vi->recv)) != NULL) {

@@ -538,7 +553,7 @@ static unsigned int features[] = {
 static unsigned int features[] = {
 	VIRTIO_NET_F_CSUM, VIRTIO_NET_F_GSO, VIRTIO_NET_F_MAC,
 	VIRTIO_NET_F_HOST_TSO4, VIRTIO_NET_F_HOST_UFO, VIRTIO_NET_F_HOST_TSO6,
-	VIRTIO_NET_F_HOST_ECN,
+	VIRTIO_NET_F_HOST_ECN, VIRTIO_F_NOTIFY_ON_EMPTY,
 };
 
 static struct virtio_driver virtio_net = {