Merge tag 'vfio-v3.20-rc1' of git://github.com/awilliam/linux-vfio

Pull VFIO updates from Alex Williamson:

 - IOMMU updates based on trace analysis
 - VFIO device request interface

* tag 'vfio-v3.20-rc1' of git://github.com/awilliam/linux-vfio:
  vfio-pci: Add device request interface
  vfio-pci: Generalize setup of simple eventfds
  vfio: Add and use device request op for vfio bus drivers
  vfio: Tie IOMMU group reference to vfio group
  vfio: Add device tracking during unbind
  vfio/type1: Add conditional rescheduling
  vfio/type1: Chunk contiguous reserved/invalid page mappings
  vfio/type1: DMA unmap chunking

drivers/vfio/pci/vfio_pci.c

@@ -239,9 +239,12 @@ static int vfio_pci_get_irq_count(struct vfio_pci_device *vdev, int irq_type)
 
 			return (flags & PCI_MSIX_FLAGS_QSIZE) + 1;
 		}
-	} else if (irq_type == VFIO_PCI_ERR_IRQ_INDEX)
+	} else if (irq_type == VFIO_PCI_ERR_IRQ_INDEX) {
 		if (pci_is_pcie(vdev->pdev))
 			return 1;
+	} else if (irq_type == VFIO_PCI_REQ_IRQ_INDEX) {
+		return 1;
+	}
 
 	return 0;
 }
@@ -464,6 +467,7 @@ static long vfio_pci_ioctl(void *device_data,
 
 		switch (info.index) {
 		case VFIO_PCI_INTX_IRQ_INDEX ... VFIO_PCI_MSIX_IRQ_INDEX:
+		case VFIO_PCI_REQ_IRQ_INDEX:
 			break;
 		case VFIO_PCI_ERR_IRQ_INDEX:
 			if (pci_is_pcie(vdev->pdev))
@@ -828,6 +832,20 @@ static int vfio_pci_mmap(void *device_data, struct vm_area_struct *vma)
 			       req_len, vma->vm_page_prot);
 }
 
+static void vfio_pci_request(void *device_data, unsigned int count)
+{
+	struct vfio_pci_device *vdev = device_data;
+
+	mutex_lock(&vdev->igate);
+
+	if (vdev->req_trigger) {
+		dev_dbg(&vdev->pdev->dev, "Requesting device from user\n");
+		eventfd_signal(vdev->req_trigger, 1);
+	}
+
+	mutex_unlock(&vdev->igate);
+}
+
 static const struct vfio_device_ops vfio_pci_ops = {
 	.name		= "vfio-pci",
 	.open		= vfio_pci_open,
@@ -836,6 +854,7 @@ static const struct vfio_device_ops vfio_pci_ops = {
 	.read		= vfio_pci_read,
 	.write		= vfio_pci_write,
 	.mmap		= vfio_pci_mmap,
+	.request	= vfio_pci_request,
 };
 
 static int vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)

drivers/vfio/pci/vfio_pci_intrs.c

@@ -763,46 +763,70 @@ static int vfio_pci_set_msi_trigger(struct vfio_pci_device *vdev,
 	return 0;
 }
 
-static int vfio_pci_set_err_trigger(struct vfio_pci_device *vdev,
-				    unsigned index, unsigned start,
-				    unsigned count, uint32_t flags, void *data)
+static int vfio_pci_set_ctx_trigger_single(struct eventfd_ctx **ctx,
+					   uint32_t flags, void *data)
 {
 	int32_t fd = *(int32_t *)data;
 
-	if ((index != VFIO_PCI_ERR_IRQ_INDEX) ||
-	    !(flags & VFIO_IRQ_SET_DATA_TYPE_MASK))
+	if (!(flags & VFIO_IRQ_SET_DATA_TYPE_MASK))
 		return -EINVAL;
 
 	/* DATA_NONE/DATA_BOOL enables loopback testing */
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
-		if (vdev->err_trigger)
-			eventfd_signal(vdev->err_trigger, 1);
+		if (*ctx)
+			eventfd_signal(*ctx, 1);
 		return 0;
 	} else if (flags & VFIO_IRQ_SET_DATA_BOOL) {
 		uint8_t trigger = *(uint8_t *)data;
-		if (trigger && vdev->err_trigger)
-			eventfd_signal(vdev->err_trigger, 1);
+		if (trigger && *ctx)
+			eventfd_signal(*ctx, 1);
 		return 0;
 	}
 
 	/* Handle SET_DATA_EVENTFD */
 	if (fd == -1) {
-		if (vdev->err_trigger)
-			eventfd_ctx_put(vdev->err_trigger);
-		vdev->err_trigger = NULL;
+		if (*ctx)
+			eventfd_ctx_put(*ctx);
+		*ctx = NULL;
 		return 0;
 	} else if (fd >= 0) {
 		struct eventfd_ctx *efdctx;
 		efdctx = eventfd_ctx_fdget(fd);
 		if (IS_ERR(efdctx))
 			return PTR_ERR(efdctx);
-		if (vdev->err_trigger)
-			eventfd_ctx_put(vdev->err_trigger);
-		vdev->err_trigger = efdctx;
+		if (*ctx)
+			eventfd_ctx_put(*ctx);
+		*ctx = efdctx;
 		return 0;
 	} else
 		return -EINVAL;
 }
+
+static int vfio_pci_set_err_trigger(struct vfio_pci_device *vdev,
+				    unsigned index, unsigned start,
+				    unsigned count, uint32_t flags, void *data)
+{
+	if (index != VFIO_PCI_ERR_IRQ_INDEX)
+		return -EINVAL;
+
+	/*
+	 * We should sanitize start & count, but that wasn't caught
+	 * originally, so this IRQ index must forever ignore them :-(
+	 */
+
+	return vfio_pci_set_ctx_trigger_single(&vdev->err_trigger, flags, data);
+}
+
+static int vfio_pci_set_req_trigger(struct vfio_pci_device *vdev,
+				    unsigned index, unsigned start,
+				    unsigned count, uint32_t flags, void *data)
+{
+	if (index != VFIO_PCI_REQ_IRQ_INDEX || start != 0 || count != 1)
+		return -EINVAL;
+
+	return vfio_pci_set_ctx_trigger_single(&vdev->req_trigger, flags, data);
+}
+
 int vfio_pci_set_irqs_ioctl(struct vfio_pci_device *vdev, uint32_t flags,
 			    unsigned index, unsigned start, unsigned count,
 			    void *data)
@@ -844,6 +868,12 @@ int vfio_pci_set_irqs_ioctl(struct vfio_pci_device *vdev, uint32_t flags,
 				func = vfio_pci_set_err_trigger;
 			break;
 		}
+	case VFIO_PCI_REQ_IRQ_INDEX:
+		switch (flags & VFIO_IRQ_SET_ACTION_TYPE_MASK) {
+		case VFIO_IRQ_SET_ACTION_TRIGGER:
+			func = vfio_pci_set_req_trigger;
+			break;
+		}
 	}
 
 	if (!func)

drivers/vfio/pci/vfio_pci_private.h

@@ -58,6 +58,7 @@ struct vfio_pci_device {
 	struct pci_saved_state	*pci_saved_state;
 	int			refcnt;
 	struct eventfd_ctx	*err_trigger;
+	struct eventfd_ctx	*req_trigger;
 };
 
 #define is_intx(vdev) (vdev->irq_type == VFIO_PCI_INTX_IRQ_INDEX)

drivers/vfio/vfio.c

@@ -63,6 +63,11 @@ struct vfio_container {
 	void				*iommu_data;
 };
 
+struct vfio_unbound_dev {
+	struct device			*dev;
+	struct list_head		unbound_next;
+};
+
 struct vfio_group {
 	struct kref			kref;
 	int				minor;
@@ -75,6 +80,8 @@ struct vfio_group {
 	struct notifier_block		nb;
 	struct list_head		vfio_next;
 	struct list_head		container_next;
+	struct list_head		unbound_list;
+	struct mutex			unbound_lock;
 	atomic_t			opened;
 };
 
@@ -204,6 +211,8 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
 	kref_init(&group->kref);
 	INIT_LIST_HEAD(&group->device_list);
 	mutex_init(&group->device_lock);
+	INIT_LIST_HEAD(&group->unbound_list);
+	mutex_init(&group->unbound_lock);
 	atomic_set(&group->container_users, 0);
 	atomic_set(&group->opened, 0);
 	group->iommu_group = iommu_group;
@@ -264,13 +273,22 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group)
 static void vfio_group_release(struct kref *kref)
 {
 	struct vfio_group *group = container_of(kref, struct vfio_group, kref);
+	struct vfio_unbound_dev *unbound, *tmp;
+	struct iommu_group *iommu_group = group->iommu_group;
 
 	WARN_ON(!list_empty(&group->device_list));
 
+	list_for_each_entry_safe(unbound, tmp,
+				 &group->unbound_list, unbound_next) {
+		list_del(&unbound->unbound_next);
+		kfree(unbound);
+	}
+
 	device_destroy(vfio.class, MKDEV(MAJOR(vfio.group_devt), group->minor));
 	list_del(&group->vfio_next);
 	vfio_free_group_minor(group->minor);
 	vfio_group_unlock_and_free(group);
+	iommu_group_put(iommu_group);
 }
 
 static void vfio_group_put(struct vfio_group *group)
@@ -440,17 +458,36 @@ static bool vfio_whitelisted_driver(struct device_driver *drv)
 }
 
 /*
- * A vfio group is viable for use by userspace if all devices are either
- * driver-less or bound to a vfio or whitelisted driver.  We test the
- * latter by the existence of a struct vfio_device matching the dev.
+ * A vfio group is viable for use by userspace if all devices are in
+ * one of the following states:
+ *  - driver-less
+ *  - bound to a vfio driver
+ *  - bound to a whitelisted driver
+ *
+ * We use two methods to determine whether a device is bound to a vfio
+ * driver.  The first is to test whether the device exists in the vfio
+ * group.  The second is to test if the device exists on the group
+ * unbound_list, indicating it's in the middle of transitioning from
+ * a vfio driver to driver-less.
  */
 static int vfio_dev_viable(struct device *dev, void *data)
 {
 	struct vfio_group *group = data;
 	struct vfio_device *device;
 	struct device_driver *drv = ACCESS_ONCE(dev->driver);
+	struct vfio_unbound_dev *unbound;
+	int ret = -EINVAL;
 
-	if (!drv || vfio_whitelisted_driver(drv))
+	mutex_lock(&group->unbound_lock);
+	list_for_each_entry(unbound, &group->unbound_list, unbound_next) {
+		if (dev == unbound->dev) {
+			ret = 0;
+			break;
+		}
+	}
+	mutex_unlock(&group->unbound_lock);
+
+	if (!ret || !drv || vfio_whitelisted_driver(drv))
 		return 0;
 
 	device = vfio_group_get_device(group, dev);
@@ -459,7 +496,7 @@ static int vfio_dev_viable(struct device *dev, void *data)
 		return 0;
 	}
 
-	return -EINVAL;
+	return ret;
 }
 
 /**
@@ -501,6 +538,7 @@ static int vfio_iommu_group_notifier(struct notifier_block *nb,
 {
 	struct vfio_group *group = container_of(nb, struct vfio_group, nb);
 	struct device *dev = data;
+	struct vfio_unbound_dev *unbound;
 
 	/*
 	 * Need to go through a group_lock lookup to get a reference or we
@@ -550,6 +588,17 @@ static int vfio_iommu_group_notifier(struct notifier_block *nb,
 		 * stop the system to maintain isolation.  At a minimum, we'd
 		 * want a toggle to disable driver auto probe for this device.
 		 */
+
+		mutex_lock(&group->unbound_lock);
+		list_for_each_entry(unbound,
+				    &group->unbound_list, unbound_next) {
+			if (dev == unbound->dev) {
+				list_del(&unbound->unbound_next);
+				kfree(unbound);
+				break;
+			}
+		}
+		mutex_unlock(&group->unbound_lock);
 		break;
 	}
 
@@ -578,6 +627,12 @@ int vfio_add_group_dev(struct device *dev,
 			iommu_group_put(iommu_group);
 			return PTR_ERR(group);
 		}
+	} else {
+		/*
+		 * A found vfio_group already holds a reference to the
+		 * iommu_group.  A created vfio_group keeps the reference.
+		 */
+		iommu_group_put(iommu_group);
 	}
 
 	device = vfio_group_get_device(group, dev);
@@ -586,21 +641,19 @@ int vfio_add_group_dev(struct device *dev,
 		     dev_name(dev), iommu_group_id(iommu_group));
 		vfio_device_put(device);
 		vfio_group_put(group);
-		iommu_group_put(iommu_group);
 		return -EBUSY;
 	}
 
 	device = vfio_group_create_device(group, dev, ops, device_data);
 	if (IS_ERR(device)) {
 		vfio_group_put(group);
-		iommu_group_put(iommu_group);
 		return PTR_ERR(device);
 	}
 
 	/*
-	 * Added device holds reference to iommu_group and vfio_device
-	 * (which in turn holds reference to vfio_group).  Drop extra
-	 * group reference used while acquiring device.
+	 * Drop all but the vfio_device reference.  The vfio_device holds
+	 * a reference to the vfio_group, which holds a reference to the
+	 * iommu_group.
 	 */
 	vfio_group_put(group);
 
@@ -655,8 +708,9 @@ void *vfio_del_group_dev(struct device *dev)
 {
 	struct vfio_device *device = dev_get_drvdata(dev);
 	struct vfio_group *group = device->group;
-	struct iommu_group *iommu_group = group->iommu_group;
 	void *device_data = device->device_data;
+	struct vfio_unbound_dev *unbound;
+	unsigned int i = 0;
 
 	/*
 	 * The group exists so long as we have a device reference.  Get
@@ -664,14 +718,49 @@ void *vfio_del_group_dev(struct device *dev)
 	 */
 	vfio_group_get(group);
 
+	/*
+	 * When the device is removed from the group, the group suddenly
+	 * becomes non-viable; the device has a driver (until the unbind
+	 * completes), but it's not present in the group.  This is bad news
+	 * for any external users that need to re-acquire a group reference
+	 * in order to match and release their existing reference.  To
+	 * solve this, we track such devices on the unbound_list to bridge
+	 * the gap until they're fully unbound.
+	 */
+	unbound = kzalloc(sizeof(*unbound), GFP_KERNEL);
+	if (unbound) {
+		unbound->dev = dev;
+		mutex_lock(&group->unbound_lock);
+		list_add(&unbound->unbound_next, &group->unbound_list);
+		mutex_unlock(&group->unbound_lock);
+	}
+	WARN_ON(!unbound);
+
 	vfio_device_put(device);
 
-	/* TODO send a signal to encourage this to be released */
-	wait_event(vfio.release_q, !vfio_dev_present(group, dev));
+	/*
+	 * If the device is still present in the group after the above
+	 * 'put', then it is in use and we need to request it from the
+	 * bus driver.  The driver may in turn need to request the
+	 * device from the user.  We send the request on an arbitrary
+	 * interval with counter to allow the driver to take escalating
+	 * measures to release the device if it has the ability to do so.
+	 */
+	do {
+		device = vfio_group_get_device(group, dev);
+		if (!device)
+			break;
 
-	vfio_group_put(group);
+		if (device->ops->request)
+			device->ops->request(device_data, i++);
 
-	iommu_group_put(iommu_group);
+		vfio_device_put(device);
+
+	} while (wait_event_interruptible_timeout(vfio.release_q,
+						  !vfio_dev_present(group, dev),
+						  HZ * 10) <= 0);
+
+	vfio_group_put(group);
 
 	return device_data;
 }

drivers/vfio/vfio_iommu_type1.c

@@ -66,6 +66,7 @@ struct vfio_domain {
 	struct list_head	next;
 	struct list_head	group_list;
 	int			prot;		/* IOMMU_CACHE */
+	bool			fgsp;		/* Fine-grained super pages */
 };
 
 struct vfio_dma {
@@ -264,6 +265,7 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 	unsigned long limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
 	bool lock_cap = capable(CAP_IPC_LOCK);
 	long ret, i;
+	bool rsvd;
 
 	if (!current->mm)
 		return -ENODEV;
@@ -272,10 +274,9 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 	if (ret)
 		return ret;
 
-	if (is_invalid_reserved_pfn(*pfn_base))
-		return 1;
+	rsvd = is_invalid_reserved_pfn(*pfn_base);
 
-	if (!lock_cap && current->mm->locked_vm + 1 > limit) {
+	if (!rsvd && !lock_cap && current->mm->locked_vm + 1 > limit) {
 		put_pfn(*pfn_base, prot);
 		pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n", __func__,
 			limit << PAGE_SHIFT);
@@ -283,7 +284,8 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 	}
 
 	if (unlikely(disable_hugepages)) {
-		vfio_lock_acct(1);
+		if (!rsvd)
+			vfio_lock_acct(1);
 		return 1;
 	}
 
@@ -295,12 +297,14 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 		if (ret)
 			break;
 
-		if (pfn != *pfn_base + i || is_invalid_reserved_pfn(pfn)) {
+		if (pfn != *pfn_base + i ||
+		    rsvd != is_invalid_reserved_pfn(pfn)) {
 			put_pfn(pfn, prot);
 			break;
 		}
 
-		if (!lock_cap && current->mm->locked_vm + i + 1 > limit) {
+		if (!rsvd && !lock_cap &&
+		    current->mm->locked_vm + i + 1 > limit) {
 			put_pfn(pfn, prot);
 			pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n",
 				__func__, limit << PAGE_SHIFT);
@@ -308,7 +312,8 @@ static long vfio_pin_pages(unsigned long vaddr, long npage,
 		}
 	}
 
-	vfio_lock_acct(i);
+	if (!rsvd)
+		vfio_lock_acct(i);
 
 	return i;
 }
@@ -346,12 +351,14 @@ static void vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma)
 	domain = d = list_first_entry(&iommu->domain_list,
 				      struct vfio_domain, next);
 
-	list_for_each_entry_continue(d, &iommu->domain_list, next)
+	list_for_each_entry_continue(d, &iommu->domain_list, next) {
 		iommu_unmap(d->domain, dma->iova, dma->size);
+		cond_resched();
+	}
 
 	while (iova < end) {
-		size_t unmapped;
-		phys_addr_t phys;
+		size_t unmapped, len;
+		phys_addr_t phys, next;
 
 		phys = iommu_iova_to_phys(domain->domain, iova);
 		if (WARN_ON(!phys)) {
@@ -359,7 +366,19 @@ static void vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma)
 			continue;
 		}
 
-		unmapped = iommu_unmap(domain->domain, iova, PAGE_SIZE);
+		/*
+		 * To optimize for fewer iommu_unmap() calls, each of which
+		 * may require hardware cache flushing, try to find the
+		 * largest contiguous physical memory chunk to unmap.
+		 */
+		for (len = PAGE_SIZE;
+		     !domain->fgsp && iova + len < end; len += PAGE_SIZE) {
+			next = iommu_iova_to_phys(domain->domain, iova + len);
+			if (next != phys + len)
+				break;
+		}
+
+		unmapped = iommu_unmap(domain->domain, iova, len);
 		if (WARN_ON(!unmapped))
 			break;
 
@@ -367,6 +386,8 @@ static void vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma)
 					     unmapped >> PAGE_SHIFT,
 					     dma->prot, false);
 		iova += unmapped;
+
+		cond_resched();
 	}
 
 	vfio_lock_acct(-unlocked);
@@ -511,6 +532,8 @@ static int vfio_iommu_map(struct vfio_iommu *iommu, dma_addr_t iova,
 			    map_try_harder(d, iova, pfn, npage, prot))
 				goto unwind;
 		}
+
+		cond_resched();
 	}
 
 	return 0;
@@ -665,6 +688,39 @@ static int vfio_iommu_replay(struct vfio_iommu *iommu,
 	return 0;
 }
 
+/*
+ * We change our unmap behavior slightly depending on whether the IOMMU
+ * supports fine-grained superpages.  IOMMUs like AMD-Vi will use a superpage
+ * for practically any contiguous power-of-two mapping we give it.  This means
+ * we don't need to look for contiguous chunks ourselves to make unmapping
+ * more efficient.  On IOMMUs with coarse-grained super pages, like Intel VT-d
+ * with discrete 2M/1G/512G/1T superpages, identifying contiguous chunks
+ * significantly boosts non-hugetlbfs mappings and doesn't seem to hurt when
+ * hugetlbfs is in use.
+ */
+static void vfio_test_domain_fgsp(struct vfio_domain *domain)
+{
+	struct page *pages;
+	int ret, order = get_order(PAGE_SIZE * 2);
+
+	pages = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
+	if (!pages)
+		return;
+
+	ret = iommu_map(domain->domain, 0, page_to_phys(pages), PAGE_SIZE * 2,
+			IOMMU_READ | IOMMU_WRITE | domain->prot);
+	if (!ret) {
+		size_t unmapped = iommu_unmap(domain->domain, 0, PAGE_SIZE);
+
+		if (unmapped == PAGE_SIZE)
+			iommu_unmap(domain->domain, PAGE_SIZE, PAGE_SIZE);
+		else
+			domain->fgsp = true;
+	}
+
+	__free_pages(pages, order);
+}
+
 static int vfio_iommu_type1_attach_group(void *iommu_data,
 					 struct iommu_group *iommu_group)
 {
@@ -758,6 +814,8 @@ static int vfio_iommu_type1_attach_group(void *iommu_data,
 		}
 	}
 
+	vfio_test_domain_fgsp(domain);
+
 	/* replay mappings on new domains */
 	ret = vfio_iommu_replay(iommu, domain);
 	if (ret)

include/linux/vfio.h

@@ -26,6 +26,7 @@
  * @ioctl: Perform ioctl(2) on device file descriptor, supporting VFIO_DEVICE_*
  *         operations documented below
  * @mmap: Perform mmap(2) on a region of the device file descriptor
+ * @request: Request for the bus driver to release the device
  */
 struct vfio_device_ops {
 	char	*name;
@@ -38,6 +39,7 @@ struct vfio_device_ops {
 	long	(*ioctl)(void *device_data, unsigned int cmd,
 			 unsigned long arg);
 	int	(*mmap)(void *device_data, struct vm_area_struct *vma);
+	void	(*request)(void *device_data, unsigned int count);
 };
 
 extern int vfio_add_group_dev(struct device *dev,

include/uapi/linux/vfio.h

@@ -333,6 +333,7 @@ enum {
 	VFIO_PCI_MSI_IRQ_INDEX,
 	VFIO_PCI_MSIX_IRQ_INDEX,
 	VFIO_PCI_ERR_IRQ_INDEX,
+	VFIO_PCI_REQ_IRQ_INDEX,
 	VFIO_PCI_NUM_IRQS
 };