Merge commit 'linus/master' into HEAD

Conflicts:
	MAINTAINERS

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>

4 files changed, 396 insertions, 69 deletions

diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
index 018c070a357f..3a43ebf83a49 100644
--- a/drivers/virtio/virtio.c
+++ b/drivers/virtio/virtio.c
@@ -31,21 +31,37 @@ static ssize_t modalias_show(struct device *_d,
 	return sprintf(buf, "virtio:d%08Xv%08X\n",
 		       dev->id.device, dev->id.vendor);
 }
+static ssize_t features_show(struct device *_d,
+			     struct device_attribute *attr, char *buf)
+{
+	struct virtio_device *dev = container_of(_d, struct virtio_device, dev);
+	unsigned int i;
+	ssize_t len = 0;
+
+	/* We actually represent this as a bitstring, as it could be
+	 * arbitrary length in future. */
+	for (i = 0; i < ARRAY_SIZE(dev->features)*BITS_PER_LONG; i++)
+		len += sprintf(buf+len, "%c",
+			       test_bit(i, dev->features) ? '1' : '0');
+	len += sprintf(buf+len, "\n");
+	return len;
+}
 static struct device_attribute virtio_dev_attrs[] = {
 	__ATTR_RO(device),
 	__ATTR_RO(vendor),
 	__ATTR_RO(status),
 	__ATTR_RO(modalias),
+	__ATTR_RO(features),
 	__ATTR_NULL
 };
 
 static inline int virtio_id_match(const struct virtio_device *dev,
 				  const struct virtio_device_id *id)
 {
-	if (id->device != dev->id.device)
+	if (id->device != dev->id.device && id->device != VIRTIO_DEV_ANY_ID)
 		return 0;
 
-	return id->vendor == VIRTIO_DEV_ANY_ID || id->vendor != dev->id.vendor;
+	return id->vendor == VIRTIO_DEV_ANY_ID || id->vendor == dev->id.vendor;
 }
 
 /* This looks through all the IDs a driver claims to support.  If any of them
@@ -118,13 +134,14 @@ static int virtio_dev_probe(struct device *_d)
 		if (device_features & (1 << i))
 			set_bit(i, dev->features);
 
+	dev->config->finalize_features(dev);
+
 	err = drv->probe(dev);
 	if (err)
 		add_status(dev, VIRTIO_CONFIG_S_FAILED);
-	else {
-		dev->config->finalize_features(dev);
+	else
 		add_status(dev, VIRTIO_CONFIG_S_DRIVER_OK);
-	}
+
 	return err;
 }
 
@@ -185,6 +202,8 @@ int register_virtio_device(struct virtio_device *dev)
 	/* Acknowledge that we've seen the device. */
 	add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE);
 
+	INIT_LIST_HEAD(&dev->vqs);
+
 	/* device_register() causes the bus infrastructure to look for a
 	 * matching driver. */
 	err = device_register(&dev->dev);
diff --git a/drivers/virtio/virtio_balloon.c b/drivers/virtio/virtio_balloon.c
index 9c76a061a04d..26b278264796 100644
--- a/drivers/virtio/virtio_balloon.c
+++ b/drivers/virtio/virtio_balloon.c
@@ -204,6 +204,9 @@ static int balloon(void *_vballoon)
 static int virtballoon_probe(struct virtio_device *vdev)
 {
 	struct virtio_balloon *vb;
+	struct virtqueue *vqs[2];
+	vq_callback_t *callbacks[] = { balloon_ack, balloon_ack };
+	const char *names[] = { "inflate", "deflate" };
 	int err;
 
 	vdev->priv = vb = kmalloc(sizeof(*vb), GFP_KERNEL);
@@ -218,22 +221,17 @@ static int virtballoon_probe(struct virtio_device *vdev)
 	vb->vdev = vdev;
 
 	/* We expect two virtqueues. */
-	vb->inflate_vq = vdev->config->find_vq(vdev, 0, balloon_ack);
-	if (IS_ERR(vb->inflate_vq)) {
-		err = PTR_ERR(vb->inflate_vq);
+	err = vdev->config->find_vqs(vdev, 2, vqs, callbacks, names);
+	if (err)
 		goto out_free_vb;
-	}
 
-	vb->deflate_vq = vdev->config->find_vq(vdev, 1, balloon_ack);
-	if (IS_ERR(vb->deflate_vq)) {
-		err = PTR_ERR(vb->deflate_vq);
-		goto out_del_inflate_vq;
-	}
+	vb->inflate_vq = vqs[0];
+	vb->deflate_vq = vqs[1];
 
 	vb->thread = kthread_run(balloon, vb, "vballoon");
 	if (IS_ERR(vb->thread)) {
 		err = PTR_ERR(vb->thread);
-		goto out_del_deflate_vq;
+		goto out_del_vqs;
 	}
 
 	vb->tell_host_first
@@ -241,10 +239,8 @@ static int virtballoon_probe(struct virtio_device *vdev)
 
 	return 0;
 
-out_del_deflate_vq:
-	vdev->config->del_vq(vb->deflate_vq);
-out_del_inflate_vq:
-	vdev->config->del_vq(vb->inflate_vq);
+out_del_vqs:
+	vdev->config->del_vqs(vdev);
 out_free_vb:
 	kfree(vb);
 out:
@@ -264,8 +260,7 @@ static void virtballoon_remove(struct virtio_device *vdev)
 	/* Now we reset the device so we can clean up the queues. */
 	vdev->config->reset(vdev);
 
-	vdev->config->del_vq(vb->deflate_vq);
-	vdev->config->del_vq(vb->inflate_vq);
+	vdev->config->del_vqs(vdev);
 	kfree(vb);
 }
 
diff --git a/drivers/virtio/virtio_pci.c b/drivers/virtio/virtio_pci.c
index 330aacbdec1f..193c8f0e5cc5 100644
--- a/drivers/virtio/virtio_pci.c
+++ b/drivers/virtio/virtio_pci.c
@@ -42,6 +42,26 @@ struct virtio_pci_device
 	/* a list of queues so we can dispatch IRQs */
 	spinlock_t lock;
 	struct list_head virtqueues;
+
+	/* MSI-X support */
+	int msix_enabled;
+	int intx_enabled;
+	struct msix_entry *msix_entries;
+	/* Name strings for interrupts. This size should be enough,
+	 * and I'm too lazy to allocate each name separately. */
+	char (*msix_names)[256];
+	/* Number of available vectors */
+	unsigned msix_vectors;
+	/* Vectors allocated */
+	unsigned msix_used_vectors;
+};
+
+/* Constants for MSI-X */
+/* Use first vector for configuration changes, second and the rest for
+ * virtqueues Thus, we need at least 2 vectors for MSI. */
+enum {
+	VP_MSIX_CONFIG_VECTOR = 0,
+	VP_MSIX_VQ_VECTOR = 1,
 };
 
 struct virtio_pci_vq_info
@@ -60,6 +80,9 @@ struct virtio_pci_vq_info
 
 	/* the list node for the virtqueues list */
 	struct list_head node;
+
+	/* MSI-X vector (or none) */
+	unsigned vector;
 };
 
 /* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
@@ -109,7 +132,8 @@ static void vp_get(struct virtio_device *vdev, unsigned offset,
 		   void *buf, unsigned len)
 {
 	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
-	void __iomem *ioaddr = vp_dev->ioaddr + VIRTIO_PCI_CONFIG + offset;
+	void __iomem *ioaddr = vp_dev->ioaddr +
+				VIRTIO_PCI_CONFIG(vp_dev) + offset;
 	u8 *ptr = buf;
 	int i;
 
@@ -123,7 +147,8 @@ static void vp_set(struct virtio_device *vdev, unsigned offset,
 		   const void *buf, unsigned len)
 {
 	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
-	void __iomem *ioaddr = vp_dev->ioaddr + VIRTIO_PCI_CONFIG + offset;
+	void __iomem *ioaddr = vp_dev->ioaddr +
+				VIRTIO_PCI_CONFIG(vp_dev) + offset;
 	const u8 *ptr = buf;
 	int i;
 
@@ -164,6 +189,37 @@ static void vp_notify(struct virtqueue *vq)
 	iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_NOTIFY);
 }
 
+/* Handle a configuration change: Tell driver if it wants to know. */
+static irqreturn_t vp_config_changed(int irq, void *opaque)
+{
+	struct virtio_pci_device *vp_dev = opaque;
+	struct virtio_driver *drv;
+	drv = container_of(vp_dev->vdev.dev.driver,
+			   struct virtio_driver, driver);
+
+	if (drv && drv->config_changed)
+		drv->config_changed(&vp_dev->vdev);
+	return IRQ_HANDLED;
+}
+
+/* Notify all virtqueues on an interrupt. */
+static irqreturn_t vp_vring_interrupt(int irq, void *opaque)
+{
+	struct virtio_pci_device *vp_dev = opaque;
+	struct virtio_pci_vq_info *info;
+	irqreturn_t ret = IRQ_NONE;
+	unsigned long flags;
+
+	spin_lock_irqsave(&vp_dev->lock, flags);
+	list_for_each_entry(info, &vp_dev->virtqueues, node) {
+		if (vring_interrupt(irq, info->vq) == IRQ_HANDLED)
+			ret = IRQ_HANDLED;
+	}
+	spin_unlock_irqrestore(&vp_dev->lock, flags);
+
+	return ret;
+}
+
 /* A small wrapper to also acknowledge the interrupt when it's handled.
  * I really need an EIO hook for the vring so I can ack the interrupt once we
  * know that we'll be handling the IRQ but before we invoke the callback since
@@ -173,9 +229,6 @@ static void vp_notify(struct virtqueue *vq)
 static irqreturn_t vp_interrupt(int irq, void *opaque)
 {
 	struct virtio_pci_device *vp_dev = opaque;
-	struct virtio_pci_vq_info *info;
-	irqreturn_t ret = IRQ_NONE;
-	unsigned long flags;
 	u8 isr;
 
 	/* reading the ISR has the effect of also clearing it so it's very
@@ -187,34 +240,137 @@ static irqreturn_t vp_interrupt(int irq, void *opaque)
 		return IRQ_NONE;
 
 	/* Configuration change?  Tell driver if it wants to know. */
-	if (isr & VIRTIO_PCI_ISR_CONFIG) {
-		struct virtio_driver *drv;
-		drv = container_of(vp_dev->vdev.dev.driver,
-				   struct virtio_driver, driver);
+	if (isr & VIRTIO_PCI_ISR_CONFIG)
+		vp_config_changed(irq, opaque);
 
-		if (drv && drv->config_changed)
-			drv->config_changed(&vp_dev->vdev);
+	return vp_vring_interrupt(irq, opaque);
+}
+
+static void vp_free_vectors(struct virtio_device *vdev)
+{
+	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
+	int i;
+
+	if (vp_dev->intx_enabled) {
+		free_irq(vp_dev->pci_dev->irq, vp_dev);
+		vp_dev->intx_enabled = 0;
 	}
 
-	spin_lock_irqsave(&vp_dev->lock, flags);
-	list_for_each_entry(info, &vp_dev->virtqueues, node) {
-		if (vring_interrupt(irq, info->vq) == IRQ_HANDLED)
-			ret = IRQ_HANDLED;
+	for (i = 0; i < vp_dev->msix_used_vectors; ++i)
+		free_irq(vp_dev->msix_entries[i].vector, vp_dev);
+	vp_dev->msix_used_vectors = 0;
+
+	if (vp_dev->msix_enabled) {
+		/* Disable the vector used for configuration */
+		iowrite16(VIRTIO_MSI_NO_VECTOR,
+			  vp_dev->ioaddr + VIRTIO_MSI_CONFIG_VECTOR);
+		/* Flush the write out to device */
+		ioread16(vp_dev->ioaddr + VIRTIO_MSI_CONFIG_VECTOR);
+
+		vp_dev->msix_enabled = 0;
+		pci_disable_msix(vp_dev->pci_dev);
 	}
-	spin_unlock_irqrestore(&vp_dev->lock, flags);
+}
 
-	return ret;
+static int vp_enable_msix(struct pci_dev *dev, struct msix_entry *entries,
+			  int *options, int noptions)
+{
+	int i;
+	for (i = 0; i < noptions; ++i)
+		if (!pci_enable_msix(dev, entries, options[i]))
+			return options[i];
+	return -EBUSY;
+}
+
+static int vp_request_vectors(struct virtio_device *vdev, unsigned max_vqs)
+{
+	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
+	const char *name = dev_name(&vp_dev->vdev.dev);
+	unsigned i, v;
+	int err = -ENOMEM;
+	/* We want at most one vector per queue and one for config changes.
+	 * Fallback to separate vectors for config and a shared for queues.
+	 * Finally fall back to regular interrupts. */
+	int options[] = { max_vqs + 1, 2 };
+	int nvectors = max(options[0], options[1]);
+
+	vp_dev->msix_entries = kmalloc(nvectors * sizeof *vp_dev->msix_entries,
+				       GFP_KERNEL);
+	if (!vp_dev->msix_entries)
+		goto error_entries;
+	vp_dev->msix_names = kmalloc(nvectors * sizeof *vp_dev->msix_names,
+				     GFP_KERNEL);
+	if (!vp_dev->msix_names)
+		goto error_names;
+
+	for (i = 0; i < nvectors; ++i)
+		vp_dev->msix_entries[i].entry = i;
+
+	err = vp_enable_msix(vp_dev->pci_dev, vp_dev->msix_entries,
+			     options, ARRAY_SIZE(options));
+	if (err < 0) {
+		/* Can't allocate enough MSI-X vectors, use regular interrupt */
+		vp_dev->msix_vectors = 0;
+		err = request_irq(vp_dev->pci_dev->irq, vp_interrupt,
+				  IRQF_SHARED, name, vp_dev);
+		if (err)
+			goto error_irq;
+		vp_dev->intx_enabled = 1;
+	} else {
+		vp_dev->msix_vectors = err;
+		vp_dev->msix_enabled = 1;
+
+		/* Set the vector used for configuration */
+		v = vp_dev->msix_used_vectors;
+		snprintf(vp_dev->msix_names[v], sizeof *vp_dev->msix_names,
+			 "%s-config", name);
+		err = request_irq(vp_dev->msix_entries[v].vector,
+				  vp_config_changed, 0, vp_dev->msix_names[v],
+				  vp_dev);
+		if (err)
+			goto error_irq;
+		++vp_dev->msix_used_vectors;
+
+		iowrite16(v, vp_dev->ioaddr + VIRTIO_MSI_CONFIG_VECTOR);
+		/* Verify we had enough resources to assign the vector */
+		v = ioread16(vp_dev->ioaddr + VIRTIO_MSI_CONFIG_VECTOR);
+		if (v == VIRTIO_MSI_NO_VECTOR) {
+			err = -EBUSY;
+			goto error_irq;
+		}
+	}
+
+	if (vp_dev->msix_vectors && vp_dev->msix_vectors != max_vqs + 1) {
+		/* Shared vector for all VQs */
+		v = vp_dev->msix_used_vectors;
+		snprintf(vp_dev->msix_names[v], sizeof *vp_dev->msix_names,
+			 "%s-virtqueues", name);
+		err = request_irq(vp_dev->msix_entries[v].vector,
+				  vp_vring_interrupt, 0, vp_dev->msix_names[v],
+				  vp_dev);
+		if (err)
+			goto error_irq;
+		++vp_dev->msix_used_vectors;
+	}
+	return 0;
+error_irq:
+	vp_free_vectors(vdev);
+	kfree(vp_dev->msix_names);
+error_names:
+	kfree(vp_dev->msix_entries);
+error_entries:
+	return err;
 }
 
-/* the config->find_vq() implementation */
 static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
-				    void (*callback)(struct virtqueue *vq))
+				    void (*callback)(struct virtqueue *vq),
+				    const char *name)
 {
 	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
 	struct virtio_pci_vq_info *info;
 	struct virtqueue *vq;
 	unsigned long flags, size;
-	u16 num;
+	u16 num, vector;
 	int err;
 
 	/* Select the queue we're interested in */
@@ -233,6 +389,7 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
 
 	info->queue_index = index;
 	info->num = num;
+	info->vector = VIRTIO_MSI_NO_VECTOR;
 
 	size = PAGE_ALIGN(vring_size(num, VIRTIO_PCI_VRING_ALIGN));
 	info->queue = alloc_pages_exact(size, GFP_KERNEL|__GFP_ZERO);
@@ -247,7 +404,7 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
 
 	/* create the vring */
 	vq = vring_new_virtqueue(info->num, VIRTIO_PCI_VRING_ALIGN,
-				 vdev, info->queue, vp_notify, callback);
+				 vdev, info->queue, vp_notify, callback, name);
 	if (!vq) {
 		err = -ENOMEM;
 		goto out_activate_queue;
@@ -256,12 +413,43 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
 	vq->priv = info;
 	info->vq = vq;
 
+	/* allocate per-vq vector if available and necessary */
+	if (callback && vp_dev->msix_used_vectors < vp_dev->msix_vectors) {
+		vector = vp_dev->msix_used_vectors;
+		snprintf(vp_dev->msix_names[vector], sizeof *vp_dev->msix_names,
+			 "%s-%s", dev_name(&vp_dev->vdev.dev), name);
+		err = request_irq(vp_dev->msix_entries[vector].vector,
+				  vring_interrupt, 0,
+				  vp_dev->msix_names[vector], vq);
+		if (err)
+			goto out_request_irq;
+		info->vector = vector;
+		++vp_dev->msix_used_vectors;
+	} else
+		vector = VP_MSIX_VQ_VECTOR;
+
+	 if (callback && vp_dev->msix_enabled) {
+		iowrite16(vector, vp_dev->ioaddr + VIRTIO_MSI_QUEUE_VECTOR);
+		vector = ioread16(vp_dev->ioaddr + VIRTIO_MSI_QUEUE_VECTOR);
+		if (vector == VIRTIO_MSI_NO_VECTOR) {
+			err = -EBUSY;
+			goto out_assign;
+		}
+	}
+
 	spin_lock_irqsave(&vp_dev->lock, flags);
 	list_add(&info->node, &vp_dev->virtqueues);
 	spin_unlock_irqrestore(&vp_dev->lock, flags);
 
 	return vq;
 
+out_assign:
+	if (info->vector != VIRTIO_MSI_NO_VECTOR) {
+		free_irq(vp_dev->msix_entries[info->vector].vector, vq);
+		--vp_dev->msix_used_vectors;
+	}
+out_request_irq:
+	vring_del_virtqueue(vq);
 out_activate_queue:
 	iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
 	free_pages_exact(info->queue, size);
@@ -270,21 +458,27 @@ out_info:
 	return ERR_PTR(err);
 }
 
-/* the config->del_vq() implementation */
 static void vp_del_vq(struct virtqueue *vq)
 {
 	struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev);
 	struct virtio_pci_vq_info *info = vq->priv;
-	unsigned long flags, size;
+	unsigned long size;
 
-	spin_lock_irqsave(&vp_dev->lock, flags);
-	list_del(&info->node);
-	spin_unlock_irqrestore(&vp_dev->lock, flags);
+	iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL);
+
+	if (info->vector != VIRTIO_MSI_NO_VECTOR)
+		free_irq(vp_dev->msix_entries[info->vector].vector, vq);
+
+	if (vp_dev->msix_enabled) {
+		iowrite16(VIRTIO_MSI_NO_VECTOR,
+			  vp_dev->ioaddr + VIRTIO_MSI_QUEUE_VECTOR);
+		/* Flush the write out to device */
+		ioread8(vp_dev->ioaddr + VIRTIO_PCI_ISR);
+	}
 
 	vring_del_virtqueue(vq);
 
 	/* Select and deactivate the queue */
-	iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL);
 	iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
 
 	size = PAGE_ALIGN(vring_size(info->num, VIRTIO_PCI_VRING_ALIGN));
@@ -292,14 +486,57 @@ static void vp_del_vq(struct virtqueue *vq)
 	kfree(info);
 }
 
+/* the config->del_vqs() implementation */
+static void vp_del_vqs(struct virtio_device *vdev)
+{
+	struct virtqueue *vq, *n;
+
+	list_for_each_entry_safe(vq, n, &vdev->vqs, list)
+		vp_del_vq(vq);
+
+	vp_free_vectors(vdev);
+}
+
+/* the config->find_vqs() implementation */
+static int vp_find_vqs(struct virtio_device *vdev, unsigned nvqs,
+		       struct virtqueue *vqs[],
+		       vq_callback_t *callbacks[],
+		       const char *names[])
+{
+	int vectors = 0;
+	int i, err;
+
+	/* How many vectors would we like? */
+	for (i = 0; i < nvqs; ++i)
+		if (callbacks[i])
+			++vectors;
+
+	err = vp_request_vectors(vdev, vectors);
+	if (err)
+		goto error_request;
+
+	for (i = 0; i < nvqs; ++i) {
+		vqs[i] = vp_find_vq(vdev, i, callbacks[i], names[i]);
+		if (IS_ERR(vqs[i]))
+			goto error_find;
+	}
+	return 0;
+
+error_find:
+	vp_del_vqs(vdev);
+
+error_request:
+	return PTR_ERR(vqs[i]);
+}
+
 static struct virtio_config_ops virtio_pci_config_ops = {
 	.get		= vp_get,
 	.set		= vp_set,
 	.get_status	= vp_get_status,
 	.set_status	= vp_set_status,
 	.reset		= vp_reset,
-	.find_vq	= vp_find_vq,
-	.del_vq		= vp_del_vq,
+	.find_vqs	= vp_find_vqs,
+	.del_vqs	= vp_del_vqs,
 	.get_features	= vp_get_features,
 	.finalize_features = vp_finalize_features,
 };
@@ -310,7 +547,7 @@ static void virtio_pci_release_dev(struct device *_d)
 	struct virtio_pci_device *vp_dev = to_vp_device(dev);
 	struct pci_dev *pci_dev = vp_dev->pci_dev;
 
-	free_irq(pci_dev->irq, vp_dev);
+	vp_del_vqs(dev);
 	pci_set_drvdata(pci_dev, NULL);
 	pci_iounmap(pci_dev, vp_dev->ioaddr);
 	pci_release_regions(pci_dev);
@@ -369,21 +606,13 @@ static int __devinit virtio_pci_probe(struct pci_dev *pci_dev,
 	vp_dev->vdev.id.vendor = pci_dev->subsystem_vendor;
 	vp_dev->vdev.id.device = pci_dev->subsystem_device;
 
-	/* register a handler for the queue with the PCI device's interrupt */
-	err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED,
-			  dev_name(&vp_dev->vdev.dev), vp_dev);
-	if (err)
-		goto out_set_drvdata;
-
 	/* finally register the virtio device */
 	err = register_virtio_device(&vp_dev->vdev);
 	if (err)
-		goto out_req_irq;
+		goto out_set_drvdata;
 
 	return 0;
 
-out_req_irq:
-	free_irq(pci_dev->irq, vp_dev);
 out_set_drvdata:
 	pci_set_drvdata(pci_dev, NULL);
 	pci_iounmap(pci_dev, vp_dev->ioaddr);
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 5c52369ab9bb..a882f2606515 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -23,21 +23,30 @@
 
 #ifdef DEBUG
 /* For development, we want to crash whenever the ring is screwed. */
-#define BAD_RING(_vq, fmt...)			\
-	do { dev_err(&(_vq)->vq.vdev->dev, fmt); BUG(); } while(0)
+#define BAD_RING(_vq, fmt, args...)				\
+	do {							\
+		dev_err(&(_vq)->vq.vdev->dev,			\
+			"%s:"fmt, (_vq)->vq.name, ##args);	\
+		BUG();						\
+	} while (0)
 /* Caller is supposed to guarantee no reentry. */
 #define START_USE(_vq)						\
 	do {							\
 		if ((_vq)->in_use)				\
-			panic("in_use = %i\n", (_vq)->in_use);	\
+			panic("%s:in_use = %i\n",		\
+			      (_vq)->vq.name, (_vq)->in_use);	\
 		(_vq)->in_use = __LINE__;			\
 		mb();						\
-	} while(0)
+	} while (0)
 #define END_USE(_vq) \
 	do { BUG_ON(!(_vq)->in_use); (_vq)->in_use = 0; mb(); } while(0)
 #else
-#define BAD_RING(_vq, fmt...)			\
-	do { dev_err(&_vq->vq.vdev->dev, fmt); (_vq)->broken = true; } while(0)
+#define BAD_RING(_vq, fmt, args...)				\
+	do {							\
+		dev_err(&_vq->vq.vdev->dev,			\
+			"%s:"fmt, (_vq)->vq.name, ##args);	\
+		(_vq)->broken = true;				\
+	} while (0)
 #define START_USE(vq)
 #define END_USE(vq)
 #endif
@@ -52,6 +61,9 @@ struct vring_virtqueue
 	/* Other side has made a mess, don't try any more. */
 	bool broken;
 
+	/* Host supports indirect buffers */
+	bool indirect;
+
 	/* Number of free buffers */
 	unsigned int num_free;
 	/* Head of free buffer list. */
@@ -76,6 +88,55 @@ struct vring_virtqueue
 
 #define to_vvq(_vq) container_of(_vq, struct vring_virtqueue, vq)
 
+/* Set up an indirect table of descriptors and add it to the queue. */
+static int vring_add_indirect(struct vring_virtqueue *vq,
+			      struct scatterlist sg[],
+			      unsigned int out,
+			      unsigned int in)
+{
+	struct vring_desc *desc;
+	unsigned head;
+	int i;
+
+	desc = kmalloc((out + in) * sizeof(struct vring_desc), GFP_ATOMIC);
+	if (!desc)
+		return vq->vring.num;
+
+	/* Transfer entries from the sg list into the indirect page */
+	for (i = 0; i < out; i++) {
+		desc[i].flags = VRING_DESC_F_NEXT;
+		desc[i].addr = sg_phys(sg);
+		desc[i].len = sg->length;
+		desc[i].next = i+1;
+		sg++;
+	}
+	for (; i < (out + in); i++) {
+		desc[i].flags = VRING_DESC_F_NEXT|VRING_DESC_F_WRITE;
+		desc[i].addr = sg_phys(sg);
+		desc[i].len = sg->length;
+		desc[i].next = i+1;
+		sg++;
+	}
+
+	/* Last one doesn't continue. */
+	desc[i-1].flags &= ~VRING_DESC_F_NEXT;
+	desc[i-1].next = 0;
+
+	/* We're about to use a buffer */
+	vq->num_free--;
+
+	/* Use a single buffer which doesn't continue */
+	head = vq->free_head;
+	vq->vring.desc[head].flags = VRING_DESC_F_INDIRECT;
+	vq->vring.desc[head].addr = virt_to_phys(desc);
+	vq->vring.desc[head].len = i * sizeof(struct vring_desc);
+
+	/* Update free pointer */
+	vq->free_head = vq->vring.desc[head].next;
+
+	return head;
+}
+
 static int vring_add_buf(struct virtqueue *_vq,
 			 struct scatterlist sg[],
 			 unsigned int out,
@@ -85,12 +146,21 @@ static int vring_add_buf(struct virtqueue *_vq,
 	struct vring_virtqueue *vq = to_vvq(_vq);
 	unsigned int i, avail, head, uninitialized_var(prev);
 
+	START_USE(vq);
+
 	BUG_ON(data == NULL);
+
+	/* If the host supports indirect descriptor tables, and we have multiple
+	 * buffers, then go indirect. FIXME: tune this threshold */
+	if (vq->indirect && (out + in) > 1 && vq->num_free) {
+		head = vring_add_indirect(vq, sg, out, in);
+		if (head != vq->vring.num)
+			goto add_head;
+	}
+
 	BUG_ON(out + in > vq->vring.num);
 	BUG_ON(out + in == 0);
 
-	START_USE(vq);
-
 	if (vq->num_free < out + in) {
 		pr_debug("Can't add buf len %i - avail = %i\n",
 			 out + in, vq->num_free);
@@ -127,6 +197,7 @@ static int vring_add_buf(struct virtqueue *_vq,
 	/* Update free pointer */
 	vq->free_head = i;
 
+add_head:
 	/* Set token. */
 	vq->data[head] = data;
 
@@ -170,6 +241,11 @@ static void detach_buf(struct vring_virtqueue *vq, unsigned int head)
 
 	/* Put back on free list: find end */
 	i = head;
+
+	/* Free the indirect table */
+	if (vq->vring.desc[i].flags & VRING_DESC_F_INDIRECT)
+		kfree(phys_to_virt(vq->vring.desc[i].addr));
+
 	while (vq->vring.desc[i].flags & VRING_DESC_F_NEXT) {
 		i = vq->vring.desc[i].next;
 		vq->num_free++;
@@ -284,7 +360,8 @@ struct virtqueue *vring_new_virtqueue(unsigned int num,
 				      struct virtio_device *vdev,
 				      void *pages,
 				      void (*notify)(struct virtqueue *),
-				      void (*callback)(struct virtqueue *))
+				      void (*callback)(struct virtqueue *),
+				      const char *name)
 {
 	struct vring_virtqueue *vq;
 	unsigned int i;
@@ -303,14 +380,18 @@ struct virtqueue *vring_new_virtqueue(unsigned int num,
 	vq->vq.callback = callback;
 	vq->vq.vdev = vdev;
 	vq->vq.vq_ops = &vring_vq_ops;
+	vq->vq.name = name;
 	vq->notify = notify;
 	vq->broken = false;
 	vq->last_used_idx = 0;
 	vq->num_added = 0;
+	list_add_tail(&vq->vq.list, &vdev->vqs);
 #ifdef DEBUG
 	vq->in_use = false;
 #endif
 
+	vq->indirect = virtio_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC);
+
 	/* No callback?  Tell other side not to bother us. */
 	if (!callback)
 		vq->vring.avail->flags |= VRING_AVAIL_F_NO_INTERRUPT;
@@ -327,6 +408,7 @@ EXPORT_SYMBOL_GPL(vring_new_virtqueue);
 
 void vring_del_virtqueue(struct virtqueue *vq)
 {
+	list_del(&vq->list);
 	kfree(to_vvq(vq));
 }
 EXPORT_SYMBOL_GPL(vring_del_virtqueue);
@@ -338,6 +420,8 @@ void vring_transport_features(struct virtio_device *vdev)
 
 	for (i = VIRTIO_TRANSPORT_F_START; i < VIRTIO_TRANSPORT_F_END; i++) {
 		switch (i) {
+		case VIRTIO_RING_F_INDIRECT_DESC:
+			break;
 		default:
 			/* We don't understand this bit. */
 			clear_bit(i, vdev->features);