[PATCH] iommu: Make sysfs robust for non-API groups

[PATCH] iommu: Make sysfs robust for non-API groups

[Robin Murphy]

Groups created by VFIO backends outside the core IOMMU API should never
be passed directly into the API itself, however they still expose their
standard sysfs attributes, so we can still stumble across them that way.
Take care to consider those cases before jumping into our normal
assumptions of a fully-initialised core API group.

Fixes: 3f6634d997db ("iommu: Use right way to retrieve iommu_ops")
Reported-by: Jan Stancek <jstancek@redhat.com>
Tested-by: Jan Stancek <jstancek@redhat.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---

/me has a vested interest in not going backwards on dev_iommu_ops() :)

 drivers/iommu/iommu.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index 29906bc16371..41ea2deaee03 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -510,6 +510,13 @@ int iommu_get_group_resv_regions(struct iommu_group *group,
 	list_for_each_entry(device, &group->devices, list) {
 		struct list_head dev_resv_regions;
 
+		/*
+		 * Non-API groups still expose reserved_regions in sysfs,
+		 * so filter out calls that get here that way.
+		 */
+		if (!device->dev->iommu)
+			break;
+
 		INIT_LIST_HEAD(&dev_resv_regions);
 		iommu_get_resv_regions(device->dev, &dev_resv_regions);
 		ret = iommu_insert_device_resv_regions(&dev_resv_regions, head);
@@ -2977,7 +2984,7 @@ static ssize_t iommu_group_store_type(struct iommu_group *group,
 	if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
 		return -EACCES;
 
-	if (WARN_ON(!group))
+	if (WARN_ON(!group) || !group->default_domain)
 		return -EINVAL;
 
 	if (sysfs_streq(buf, "identity"))
-- 
2.35.3.dirty

Re: [PATCH] iommu: Make sysfs robust for non-API groups

[Joerg Roedel]

On Wed, May 04, 2022 at 01:39:58PM +0100, Robin Murphy wrote:
> Groups created by VFIO backends outside the core IOMMU API should never
> be passed directly into the API itself, however they still expose their
> standard sysfs attributes, so we can still stumble across them that way.
> Take care to consider those cases before jumping into our normal
> assumptions of a fully-initialised core API group.
> 
> Fixes: 3f6634d997db ("iommu: Use right way to retrieve iommu_ops")
> Reported-by: Jan Stancek <jstancek@redhat.com>
> Tested-by: Jan Stancek <jstancek@redhat.com>
> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Robin Murphy <robin.murphy@arm.com>
> ---
> 
> /me has a vested interest in not going backwards on dev_iommu_ops() :)
> 
>  drivers/iommu/iommu.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)

Applied, thanks.