[PATCH V3 1/3] iommu: Add support to change default domain of an iommu_group

Sai Praneeth Prakhya sai.praneeth.prakhya at intel.com
Thu May 28 19:23:33 UTC 2020


Presently, the default domain of an iommu_group is allocated during boot
time (i.e. when a device is being added to a group) and it cannot be
changed later. So, the device would typically be either in identity (also
known as pass_through) mode (controlled by "iommu=pt" kernel command line
argument) or the device would be in DMA mode as long as the machine is up
and running. There is no way to change the default domain type dynamically
i.e. after booting, a device cannot switch between identity mode and DMA
mode.

But, assume a use case wherein the user trusts the device and believes that
the OS is secure enough and hence wants *only* this device to bypass IOMMU
(so that it could be high performing) whereas all the other devices to go
through IOMMU (so that the system is protected). Presently, this use case
is not supported. It will be helpful if there is some way to change the
default domain of a B:D.F dynamically. Hence, add such support.

A privileged user could request the kernel to change the default domain
type of a iommu_group by writing to
"/sys/kernel/iommu_groups/<grp_id>/type" file. Presently, only three values
are supported
1. identity: all the DMA transactions from the device in this group are
             *not* translated by the iommu
2. DMA: all the DMA transactions from the device in this group are
        translated by the iommu
3. auto: change to the type the device was booted with

Note:
1. Default domain of an iommu group with two or more devices cannot be
   changed.
2. The device in the iommu group shouldn't be bound to any driver.
3. The device shouldn't be assigned to user for direct access.

Please see "Documentation/ABI/testing/sysfs-kernel-iommu_groups" for more
information.

Cc: Christoph Hellwig <hch at lst.de>
Cc: Joerg Roedel <joro at 8bytes.org>
Cc: Ashok Raj <ashok.raj at intel.com>
Cc: Will Deacon <will.deacon at arm.com>
Cc: Lu Baolu <baolu.lu at linux.intel.com>
Cc: Sohil Mehta <sohil.mehta at intel.com>
Cc: Robin Murphy <robin.murphy at arm.com>
Cc: Jacob Pan <jacob.jun.pan at linux.intel.com>
Signed-off-by: Sai Praneeth Prakhya <sai.praneeth.prakhya at intel.com>
---
 drivers/iommu/iommu.c | 211 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 210 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index a4c2f122eb8b..2b6cca799055 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -92,6 +92,8 @@ static void __iommu_detach_group(struct iommu_domain *domain,
 static int iommu_create_device_direct_mappings(struct iommu_group *group,
 					       struct device *dev);
 static struct iommu_group *iommu_group_get_for_dev(struct device *dev);
+static ssize_t iommu_group_store_type(struct iommu_group *group,
+				      const char *buf, size_t count);
 
 #define IOMMU_GROUP_ATTR(_name, _mode, _show, _store)		\
 struct iommu_group_attribute iommu_group_attr_##_name =		\
@@ -524,7 +526,8 @@ static IOMMU_GROUP_ATTR(name, S_IRUGO, iommu_group_show_name, NULL);
 static IOMMU_GROUP_ATTR(reserved_regions, 0444,
 			iommu_group_show_resv_regions, NULL);
 
-static IOMMU_GROUP_ATTR(type, 0444, iommu_group_show_type, NULL);
+static IOMMU_GROUP_ATTR(type, 0644, iommu_group_show_type,
+			iommu_group_store_type);
 
 static void iommu_group_release(struct kobject *kobj)
 {
@@ -2847,3 +2850,209 @@ int iommu_sva_get_pasid(struct iommu_sva *handle)
 	return ops->sva_get_pasid(handle);
 }
 EXPORT_SYMBOL_GPL(iommu_sva_get_pasid);
+
+/*
+ * Changes the default domain of a device
+ *
+ * @dev: The *only* device in the group
+ * @group: The group for which the default domain should be changed
+ * @prev_domain: The previous domain that is being switched from
+ * @type: The type of the new default domain that gets associated with the group
+ *
+ * Returns 0 on success and error code on failure
+ *
+ * Note:
+ * 1. Presently, this function is called only when user requests to change the
+ *    group's default domain type through /sys/kernel/iommu_groups/<grp_id>/type
+ *    Please take a closer look if intended to use for other purposes.
+ * 2. Assumes that group->mutex is already taken and releases before returning
+ */
+static int iommu_change_dev_def_domain(struct device *dev,
+				       struct iommu_group *group,
+				       struct iommu_domain *prev_dom, int type)
+{
+	int ret = 0;
+
+	/* Sets group->default_domain to the newly allocated domain */
+	ret = iommu_group_alloc_default_domain(dev->bus, group, type);
+	if (ret)
+		goto out;
+
+	ret = __iommu_attach_device(group->default_domain, dev);
+	if (ret)
+		goto free_new_domain;
+
+	group->domain = group->default_domain;
+
+	ret = iommu_create_device_direct_mappings(group, dev);
+	if (ret)
+		goto free_new_domain;
+
+	/*
+	 * Release the mutex here because ops->probe_finalize() call-back of
+	 * some vendor IOMMU drivers calls arm_iommu_attach_device() which
+	 * in-turn might call back into IOMMU core code, where it tries to take
+	 * group->mutex, resulting in a deadlock.
+	 */
+	mutex_unlock(&group->mutex);
+
+	/* Make sure dma_ops is appropriatley set */
+	iommu_group_do_probe_finalize(dev, group->default_domain);
+	iommu_domain_free(prev_dom);
+	return 0;
+
+free_new_domain:
+	iommu_domain_free(group->default_domain);
+	group->default_domain = prev_dom;
+	group->domain = prev_dom;
+
+out:
+	mutex_unlock(&group->mutex);
+
+	return ret;
+}
+
+static ssize_t iommu_group_store_type(struct iommu_group *group,
+				      const char *buf, size_t count)
+{
+	int ret, req_type = 0, req_auto = 0, dev_def_dom;
+	struct iommu_domain *prev_dom;
+	struct group_device *grp_dev;
+	const struct iommu_ops *ops;
+	struct device *dev;
+
+	if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
+		return -EACCES;
+
+	if (WARN_ON(!group))
+		return -EINVAL;
+
+	if (sysfs_streq(buf, "identity"))
+		req_type = IOMMU_DOMAIN_IDENTITY;
+	else if (sysfs_streq(buf, "DMA"))
+		req_type = IOMMU_DOMAIN_DMA;
+	else if (sysfs_streq(buf, "auto"))
+		req_auto = 1;
+	else
+		return -EINVAL;
+
+	/* Lock the group to make sure that the device count doesn't change */
+	mutex_lock(&group->mutex);
+	if (iommu_group_device_count(group) != 1) {
+		mutex_unlock(&group->mutex);
+		pr_err("Cannot change default domain of a group with two or more devices\n");
+		return -EINVAL;
+	}
+
+	/* Since group has only one device */
+	list_for_each_entry(grp_dev, &group->devices, list)
+		dev = grp_dev->dev;
+
+	/*
+	 * Don't hold onto the group mutex because taking group mutex first and
+	 * then the device lock could potentially cause a deadlock as below.
+	 * Assume two threads T1 and T2. T1 is trying to change default domain
+	 * of an iommu group and T2 is trying to hot unplug a device or release
+	 * [1] VF of a PCIe device which is in the same iommu group. T1 takes
+	 * group mutex and before it could take device lock assume T2 has taken
+	 * device lock and is yet to take group mutex. Now, both the threads
+	 * will be waiting for the other thread to release lock.
+	 *
+	 * [1] Typical device release path
+	 * device_lock() from device/driver core code
+	 *  -> bus_notifier()
+	 *   -> iommu_bus_notifier()
+	 *    -> iommu_release_device()
+	 *     -> ops->release_device() vendor driver calls back iommu core code
+	 *      -> mutex_lock() from iommu core code
+	 */
+	mutex_unlock(&group->mutex);
+
+	/* Check if the device in the group still has a driver bound to it */
+	device_lock(dev);
+	if (device_is_bound(dev)) {
+		pr_err("Device is still bound to driver\n");
+		ret = -EBUSY;
+		goto dev_unlock;
+	}
+
+	/*
+	 * iommu group wasn't locked while acquiring device lock. So, make sure
+	 * that the device count hasn't changed while acquiring device lock.
+	 *
+	 * Changing default domain of an iommu group with two or more devices
+	 * isn't supported because there could be a potential deadlock. Consider
+	 * the following scenario. T1 is trying to acquire device locks of all
+	 * the devices in the group and before it could acquire all of them,
+	 * there could be another thread T2 (from different sub-system and use
+	 * case) that has already acquired some of the device locks and might be
+	 * waiting for T1 to release other device locks.
+	 */
+	mutex_lock(&group->mutex);
+	if (iommu_group_device_count(group) != 1) {
+		pr_err("Cannot change default domain of a group with two or more devices\n");
+		ret = -EINVAL;
+		goto group_unlock;
+	}
+
+	/*
+	 * Check if any user level driver (that doesn't use kernel driver like
+	 * VFIO) is directly using the group. VFIO use case is detected by
+	 * device_is_bound() check above
+	 */
+	if (group->default_domain != group->domain) {
+		pr_err("Group assigned to user level for direct access\n");
+		ret = -EINVAL;
+		goto group_unlock;
+	}
+
+	prev_dom = group->default_domain;
+	if (!prev_dom || !prev_dom->ops || !prev_dom->ops->def_domain_type) {
+		pr_err("'def_domain_type' call back isn't registered\n");
+		ret = -EINVAL;
+		goto group_unlock;
+	}
+
+	ops = prev_dom->ops;
+	dev_def_dom = ops->def_domain_type(dev);
+
+	/* Check if user requested domain is supported by the device or not */
+	if (!req_auto && dev_def_dom && req_type != dev_def_dom) {
+		pr_err("Device cannot be in %s domain\n", buf);
+		ret = -EINVAL;
+		goto group_unlock;
+	}
+
+	if (req_auto) {
+		/*
+		 * If user requested "auto" and the device supports both the
+		 * domains, then default to the domain the device was booted
+		 * with
+		 */
+		if (!dev_def_dom)
+			req_type = iommu_def_domain_type;
+		else
+			req_type = dev_def_dom;
+	}
+
+	/*
+	 * Switch to a new domain only if the requested domain type is different
+	 * from the existing default domain type
+	 */
+	if (prev_dom->type == req_type) {
+		ret = count;
+		goto group_unlock;
+	}
+
+	ret = iommu_change_dev_def_domain(dev, group, prev_dom, req_type);
+	device_unlock(dev);
+	return ret ?: count;
+
+group_unlock:
+	mutex_unlock(&group->mutex);
+
+dev_unlock:
+	device_unlock(dev);
+
+	return ret;
+}
-- 
2.19.1



More information about the iommu mailing list