[PATCH v5 0/5] iommu aux-domain APIs extensions

[PATCH v5 0/5] iommu aux-domain APIs extensions

[Lu Baolu]

Hi Jorge and Alex,

A description of this patch series could be found here.

https://lore.kernel.org/linux-iommu/20200901033422.22249-1-baolu.lu@linux.intel.com/

This version adds some changes according to Alex's review comments.

- Add comments and naming rule for subdevices.
https://lore.kernel.org/linux-iommu/20200910160549.2b176ac5@w520.home/

- Continue detaching even no subdevice parent found.
https://lore.kernel.org/linux-iommu/20200910160547.0a8b9891@w520.home/

- Make subdev_link_device() and subdev_unlink_device() symmetrical.

Please help to review and merge.

Best regards,
baolu

Lu Baolu (5):
  iommu: Add optional subdev in aux_at(de)tach ops
  iommu: Add iommu_at(de)tach_subdev_group()
  iommu: Add iommu_aux_get_domain_for_dev()
  vfio/type1: Use iommu_aux_at(de)tach_group() APIs
  iommu/vt-d: Add is_aux_domain support

 drivers/iommu/intel/iommu.c     | 139 +++++++++++++++++-------
 drivers/iommu/iommu.c           | 184 ++++++++++++++++++++++++++++++--
 drivers/vfio/vfio_iommu_type1.c |  43 ++------
 include/linux/intel-iommu.h     |  17 +--
 include/linux/iommu.h           |  46 +++++++-
 5 files changed, 336 insertions(+), 93 deletions(-)

-- 
2.17.1

[PATCH v5 4/5] vfio/type1: Use iommu_aux_at(de)tach_group() APIs

[Lu Baolu]

Replace iommu_aux_at(de)tach_device() with iommu_at(de)tach_subdev_group().

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/vfio/vfio_iommu_type1.c | 43 +++++----------------------------
 1 file changed, 6 insertions(+), 37 deletions(-)

diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index c255a6683f31..869a71c258d1 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -1690,45 +1690,13 @@ static struct device *vfio_mdev_get_iommu_device(struct device *dev)
 	return NULL;
 }
 
-static int vfio_mdev_attach_domain(struct device *dev, void *data)
-{
-	struct iommu_domain *domain = data;
-	struct device *iommu_device;
-
-	iommu_device = vfio_mdev_get_iommu_device(dev);
-	if (iommu_device) {
-		if (iommu_dev_feature_enabled(iommu_device, IOMMU_DEV_FEAT_AUX))
-			return iommu_aux_attach_device(domain, iommu_device);
-		else
-			return iommu_attach_device(domain, iommu_device);
-	}
-
-	return -EINVAL;
-}
-
-static int vfio_mdev_detach_domain(struct device *dev, void *data)
-{
-	struct iommu_domain *domain = data;
-	struct device *iommu_device;
-
-	iommu_device = vfio_mdev_get_iommu_device(dev);
-	if (iommu_device) {
-		if (iommu_dev_feature_enabled(iommu_device, IOMMU_DEV_FEAT_AUX))
-			iommu_aux_detach_device(domain, iommu_device);
-		else
-			iommu_detach_device(domain, iommu_device);
-	}
-
-	return 0;
-}
-
 static int vfio_iommu_attach_group(struct vfio_domain *domain,
 				   struct vfio_group *group)
 {
 	if (group->mdev_group)
-		return iommu_group_for_each_dev(group->iommu_group,
-						domain->domain,
-						vfio_mdev_attach_domain);
+		return iommu_attach_subdev_group(domain->domain,
+						 group->iommu_group,
+						 vfio_mdev_get_iommu_device);
 	else
 		return iommu_attach_group(domain->domain, group->iommu_group);
 }
@@ -1737,8 +1705,9 @@ static void vfio_iommu_detach_group(struct vfio_domain *domain,
 				    struct vfio_group *group)
 {
 	if (group->mdev_group)
-		iommu_group_for_each_dev(group->iommu_group, domain->domain,
-					 vfio_mdev_detach_domain);
+		iommu_detach_subdev_group(domain->domain,
+					  group->iommu_group,
+					  vfio_mdev_get_iommu_device);
 	else
 		iommu_detach_group(domain->domain, group->iommu_group);
 }
-- 
2.17.1

[PATCH v5 5/5] iommu/vt-d: Add is_aux_domain support

[Lu Baolu]

With subdevice information opt-in through iommu_ops.aux_at(de)tach_dev()
interfaces, the vendor iommu driver is able to learn the knowledge about
the relationships between the subdevices and the aux-domains. Implement
is_aux_domain() support based on the relationship knowledges.

Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 drivers/iommu/intel/iommu.c | 129 ++++++++++++++++++++++++++----------
 include/linux/intel-iommu.h |  17 +++--
 2 files changed, 105 insertions(+), 41 deletions(-)

diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
index 86142ce32f21..9ba314e2872f 100644
--- a/drivers/iommu/intel/iommu.c
+++ b/drivers/iommu/intel/iommu.c
@@ -334,6 +334,8 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,
 				     struct device *dev);
 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
 					    dma_addr_t iova);
+static bool intel_iommu_dev_feat_enabled(struct device *dev,
+					 enum iommu_dev_features feat);
 
 #ifdef CONFIG_INTEL_IOMMU_DEFAULT_ON
 int dmar_disabled = 0;
@@ -1881,6 +1883,7 @@ static struct dmar_domain *alloc_domain(int flags)
 		domain->flags |= DOMAIN_FLAG_USE_FIRST_LEVEL;
 	domain->has_iotlb_device = false;
 	INIT_LIST_HEAD(&domain->devices);
+	INIT_LIST_HEAD(&domain->subdevices);
 
 	return domain;
 }
@@ -2629,7 +2632,7 @@ static struct dmar_domain *dmar_insert_one_dev_info(struct intel_iommu *iommu,
 	info->iommu = iommu;
 	info->pasid_table = NULL;
 	info->auxd_enabled = 0;
-	INIT_LIST_HEAD(&info->auxiliary_domains);
+	INIT_LIST_HEAD(&info->subdevices);
 
 	if (dev && dev_is_pci(dev)) {
 		struct pci_dev *pdev = to_pci_dev(info->dev);
@@ -5152,21 +5155,28 @@ static void intel_iommu_domain_free(struct iommu_domain *domain)
 		domain_exit(to_dmar_domain(domain));
 }
 
-/*
- * Check whether a @domain could be attached to the @dev through the
- * aux-domain attach/detach APIs.
- */
-static inline bool
-is_aux_domain(struct device *dev, struct iommu_domain *domain)
+/* Lookup subdev_info in the domain's subdevice siblings. */
+static struct subdev_info *
+subdev_lookup_domain(struct dmar_domain *domain, struct device *dev,
+		     struct device *subdev)
 {
-	struct device_domain_info *info = get_domain_info(dev);
+	struct subdev_info *sinfo = NULL, *tmp;
+
+	assert_spin_locked(&device_domain_lock);
+
+	list_for_each_entry(tmp, &domain->subdevices, link_domain) {
+		if ((!dev || tmp->pdev == dev) && tmp->dev == subdev) {
+			sinfo = tmp;
+			break;
+		}
+	}
 
-	return info && info->auxd_enabled &&
-			domain->type == IOMMU_DOMAIN_UNMANAGED;
+	return sinfo;
 }
 
-static void auxiliary_link_device(struct dmar_domain *domain,
-				  struct device *dev)
+static void
+subdev_link_device(struct dmar_domain *domain, struct device *dev,
+		   struct subdev_info *sinfo)
 {
 	struct device_domain_info *info = get_domain_info(dev);
 
@@ -5174,12 +5184,13 @@ static void auxiliary_link_device(struct dmar_domain *domain,
 	if (WARN_ON(!info))
 		return;
 
-	domain->auxd_refcnt++;
-	list_add(&domain->auxd, &info->auxiliary_domains);
+	list_add(&sinfo->link_phys, &info->subdevices);
+	list_add(&sinfo->link_domain, &domain->subdevices);
 }
 
-static void auxiliary_unlink_device(struct dmar_domain *domain,
-				    struct device *dev)
+static void
+subdev_unlink_device(struct dmar_domain *domain, struct device *dev,
+		     struct subdev_info *sinfo)
 {
 	struct device_domain_info *info = get_domain_info(dev);
 
@@ -5187,24 +5198,26 @@ static void auxiliary_unlink_device(struct dmar_domain *domain,
 	if (WARN_ON(!info))
 		return;
 
-	list_del(&domain->auxd);
-	domain->auxd_refcnt--;
-
-	if (!domain->auxd_refcnt && domain->default_pasid > 0)
-		ioasid_free(domain->default_pasid);
+	list_del(&sinfo->link_phys);
+	list_del(&sinfo->link_domain);
 }
 
-static int aux_domain_add_dev(struct dmar_domain *domain,
-			      struct device *dev)
+static int aux_domain_add_dev(struct dmar_domain *domain, struct device *dev,
+			      struct device *subdev)
 {
 	int ret;
 	unsigned long flags;
 	struct intel_iommu *iommu;
+	struct subdev_info *sinfo;
 
 	iommu = device_to_iommu(dev, NULL, NULL);
 	if (!iommu)
 		return -ENODEV;
 
+	sinfo = kzalloc(sizeof(*sinfo), GFP_KERNEL);
+	if (!sinfo)
+		return -ENOMEM;
+
 	if (domain->default_pasid <= 0) {
 		int pasid;
 
@@ -5214,7 +5227,8 @@ static int aux_domain_add_dev(struct dmar_domain *domain,
 				     NULL);
 		if (pasid == INVALID_IOASID) {
 			pr_err("Can't allocate default pasid\n");
-			return -ENODEV;
+			ret = -ENODEV;
+			goto pasid_failed;
 		}
 		domain->default_pasid = pasid;
 	}
@@ -5240,7 +5254,10 @@ static int aux_domain_add_dev(struct dmar_domain *domain,
 		goto table_failed;
 	spin_unlock(&iommu->lock);
 
-	auxiliary_link_device(domain, dev);
+	sinfo->dev = subdev;
+	sinfo->domain = domain;
+	sinfo->pdev = dev;
+	subdev_link_device(domain, dev, sinfo);
 
 	spin_unlock_irqrestore(&device_domain_lock, flags);
 
@@ -5251,27 +5268,36 @@ static int aux_domain_add_dev(struct dmar_domain *domain,
 attach_failed:
 	spin_unlock(&iommu->lock);
 	spin_unlock_irqrestore(&device_domain_lock, flags);
-	if (!domain->auxd_refcnt && domain->default_pasid > 0)
+	if (list_empty(&domain->subdevices) && domain->default_pasid > 0)
 		ioasid_free(domain->default_pasid);
+pasid_failed:
+	kfree(sinfo);
 
 	return ret;
 }
 
-static void aux_domain_remove_dev(struct dmar_domain *domain,
-				  struct device *dev)
+static void
+aux_domain_remove_dev(struct dmar_domain *domain, struct device *dev,
+		      struct device *subdev)
 {
 	struct device_domain_info *info;
 	struct intel_iommu *iommu;
+	struct subdev_info *sinfo;
 	unsigned long flags;
 
-	if (!is_aux_domain(dev, &domain->domain))
+	if (!intel_iommu_dev_feat_enabled(dev, IOMMU_DEV_FEAT_AUX) ||
+	    domain->domain.type != IOMMU_DOMAIN_UNMANAGED)
 		return;
 
 	spin_lock_irqsave(&device_domain_lock, flags);
 	info = get_domain_info(dev);
 	iommu = info->iommu;
-
-	auxiliary_unlink_device(domain, dev);
+	sinfo = subdev_lookup_domain(domain, dev, subdev);
+	if (!sinfo) {
+		spin_unlock_irqrestore(&device_domain_lock, flags);
+		return;
+	}
+	subdev_unlink_device(domain, dev, sinfo);
 
 	spin_lock(&iommu->lock);
 	intel_pasid_tear_down_entry(iommu, dev, domain->default_pasid, false);
@@ -5279,6 +5305,10 @@ static void aux_domain_remove_dev(struct dmar_domain *domain,
 	spin_unlock(&iommu->lock);
 
 	spin_unlock_irqrestore(&device_domain_lock, flags);
+
+	if (list_empty(&domain->subdevices) && domain->default_pasid > 0)
+		ioasid_free(domain->default_pasid);
+	kfree(sinfo);
 }
 
 static int prepare_domain_attach_device(struct iommu_domain *domain,
@@ -5334,7 +5364,8 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,
 		return -EPERM;
 	}
 
-	if (is_aux_domain(dev, domain))
+	if (intel_iommu_dev_feat_enabled(dev, IOMMU_DEV_FEAT_AUX) &&
+	    domain->type == IOMMU_DOMAIN_UNMANAGED)
 		return -EPERM;
 
 	/* normally dev is not mapped */
@@ -5359,14 +5390,15 @@ intel_iommu_aux_attach_device(struct iommu_domain *domain,
 {
 	int ret;
 
-	if (!is_aux_domain(dev, domain))
+	if (!intel_iommu_dev_feat_enabled(dev, IOMMU_DEV_FEAT_AUX) ||
+	    domain->type != IOMMU_DOMAIN_UNMANAGED)
 		return -EPERM;
 
 	ret = prepare_domain_attach_device(domain, dev);
 	if (ret)
 		return ret;
 
-	return aux_domain_add_dev(to_dmar_domain(domain), dev);
+	return aux_domain_add_dev(to_dmar_domain(domain), dev, subdev);
 }
 
 static void intel_iommu_detach_device(struct iommu_domain *domain,
@@ -5379,7 +5411,7 @@ static void
 intel_iommu_aux_detach_device(struct iommu_domain *domain, struct device *dev,
 			      struct device *subdev)
 {
-	aux_domain_remove_dev(to_dmar_domain(domain), dev);
+	aux_domain_remove_dev(to_dmar_domain(domain), dev, subdev);
 }
 
 /*
@@ -6035,6 +6067,32 @@ static bool intel_iommu_is_attach_deferred(struct iommu_domain *domain,
 	return attach_deferred(dev);
 }
 
+static int
+intel_iommu_domain_get_attr(struct iommu_domain *domain,
+			    enum iommu_attr attr, void *data)
+{
+	struct dmar_domain *dmar_domain = to_dmar_domain(domain);
+	unsigned long flags;
+	int ret;
+
+	if (domain->type != IOMMU_DOMAIN_UNMANAGED)
+		return -EINVAL;
+
+	switch (attr) {
+	case DOMAIN_ATTR_IS_AUX:
+		spin_lock_irqsave(&device_domain_lock, flags);
+		ret = !IS_ERR_OR_NULL(subdev_lookup_domain(dmar_domain,
+							   NULL, data));
+		spin_unlock_irqrestore(&device_domain_lock, flags);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
 static int
 intel_iommu_domain_set_attr(struct iommu_domain *domain,
 			    enum iommu_attr attr, void *data)
@@ -6088,6 +6146,7 @@ const struct iommu_ops intel_iommu_ops = {
 	.domain_alloc		= intel_iommu_domain_alloc,
 	.domain_free		= intel_iommu_domain_free,
 	.domain_set_attr	= intel_iommu_domain_set_attr,
+	.domain_get_attr	= intel_iommu_domain_get_attr,
 	.attach_dev		= intel_iommu_attach_device,
 	.detach_dev		= intel_iommu_detach_device,
 	.aux_attach_dev		= intel_iommu_aux_attach_device,
diff --git a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h
index 6a3ddaabf3f5..3b112356c6ea 100644
--- a/include/linux/intel-iommu.h
+++ b/include/linux/intel-iommu.h
@@ -526,11 +526,9 @@ struct dmar_domain {
 					/* Domain ids per IOMMU. Use u16 since
 					 * domain ids are 16 bit wide according
 					 * to VT-d spec, section 9.3 */
-	unsigned int	auxd_refcnt;	/* Refcount of auxiliary attaching */
-
 	bool has_iotlb_device;
 	struct list_head devices;	/* all devices' list */
-	struct list_head auxd;		/* link to device's auxiliary list */
+	struct list_head subdevices;	/* all subdevices' list */
 	struct iova_domain iovad;	/* iova's that belong to this domain */
 
 	struct dma_pte	*pgd;		/* virtual address */
@@ -603,14 +601,21 @@ struct intel_iommu {
 	struct dmar_drhd_unit *drhd;
 };
 
+/* Per subdevice private data */
+struct subdev_info {
+	struct list_head link_phys;	/* link to phys device siblings */
+	struct list_head link_domain;	/* link to domain siblings */
+	struct device *pdev;		/* physical device derived from */
+	struct device *dev;		/* subdevice node */
+	struct dmar_domain *domain;	/* aux-domain */
+};
+
 /* PCI domain-device relationship */
 struct device_domain_info {
 	struct list_head link;	/* link to domain siblings */
 	struct list_head global; /* link to global list */
 	struct list_head table;	/* link to pasid table */
-	struct list_head auxiliary_domains; /* auxiliary domains
-					     * attached to this device
-					     */
+	struct list_head subdevices; /* subdevices sibling */
 	u32 segment;		/* PCI segment number */
 	u8 bus;			/* PCI bus number */
 	u8 devfn;		/* PCI devfn number */
-- 
2.17.1

Re: [PATCH v5 0/5] iommu aux-domain APIs extensions

[Joerg Roedel]

On Tue, Sep 22, 2020 at 02:10:37PM +0800, Lu Baolu wrote:
> Hi Jorge and Alex,
> 
> A description of this patch series could be found here.
> 
> https://lore.kernel.org/linux-iommu/20200901033422.22249-1-baolu.lu@linux.intel.com/

Hmm, I am wondering if we can avoid all this hassle and special APIs by
making the mdev framework more visible outside of the vfio code. There
is an underlying bus implementation for mdevs, so is there a reason
those can't use the standard iommu-core code to setup IOMMU mappings?

What speaks against doing:

	- IOMMU drivers capable of handling mdevs register iommu-ops
	  for the mdev_bus.

	- iommu_domain_alloc() takes bus_type as parameter, so there can
	  be special domains be allocated for mdevs.

	- Group creation and domain allocation will happen
	  automatically in the iommu-core when a new mdev is registered
	  through device-driver core code.

	- There should be no need for special iommu_aux_* APIs, as one
	  can attach a domain directly to &mdev->dev with
	  iommu_attach_device(domain, &mdev->dev).

Doing it this way will probably also keep the mdev-special code in VFIO
small.

Regards,

	Joerg

Re: [PATCH v5 0/5] iommu aux-domain APIs extensions

[Lu Baolu]

Hi Joerg,

On 9/24/20 5:55 PM, Joerg Roedel wrote:
> On Tue, Sep 22, 2020 at 02:10:37PM +0800, Lu Baolu wrote:
>> Hi Jorge and Alex,
>>
>> A description of this patch series could be found here.
>>
>> https://lore.kernel.org/linux-iommu/20200901033422.22249-1-baolu.lu@linux.intel.com/
> 
> Hmm, I am wondering if we can avoid all this hassle and special APIs by
> making the mdev framework more visible outside of the vfio code. There
> is an underlying bus implementation for mdevs, so is there a reason
> those can't use the standard iommu-core code to setup IOMMU mappings?

The original purpose of this series is to enable the device driver to
retrieve the aux-domain through iommu core after iommu
ops.aux_attach_dev().

The domain was allocated in vfio/mdev, but it's also needed by the
device driver in mediated callbacks. The idea of this patch series is to
extend the aux-API so that the domain could be saved in group->domain
and get by the mediated driver through the existing
iommu_get_domain_for_dev().

Back when we were developing the aux-domain, I proposed to keep the
domain in vfio/mdev.

https://lore.kernel.org/linux-iommu/20181105073408.21815-7-baolu.lu@linux.intel.com/

It wasn't discussed at that time due to the lack of real consumer. Intel
is now adding aux-domain support in idxd (DMA streaming accelerator)
driver which becomes the first real consumer. So this problem is brought
back to the table.

> 
> What speaks against doing:
> 
> 	- IOMMU drivers capable of handling mdevs register iommu-ops
> 	  for the mdev_bus.
> 
> 	- iommu_domain_alloc() takes bus_type as parameter, so there can
> 	  be special domains be allocated for mdevs.
> 
> 	- Group creation and domain allocation will happen
> 	  automatically in the iommu-core when a new mdev is registered
> 	  through device-driver core code.
> 
> 	- There should be no need for special iommu_aux_* APIs, as one
> 	  can attach a domain directly to &mdev->dev with
> 	  iommu_attach_device(domain, &mdev->dev).
> 
> Doing it this way will probably also keep the mdev-special code in VFIO
> small.

Fully understand now. Thanks for guide.

> 
> Regards,
> 
> 	Joerg
> 

Best regards,
baolu