]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
iommu: Lock group->mutex in iommu_deferred_attach()
authorNicolin Chen <nicolinc@nvidia.com>
Mon, 15 Dec 2025 21:42:16 +0000 (13:42 -0800)
committerJoerg Roedel <joerg.roedel@amd.com>
Sat, 10 Jan 2026 09:26:42 +0000 (10:26 +0100)
The iommu_deferred_attach() function invokes __iommu_attach_device(), but
doesn't hold the group->mutex like other __iommu_attach_device() callers.

Though there is no pratical bug being triggered so far, it would be better
to apply the same locking to this __iommu_attach_device(), since the IOMMU
drivers nowaday are more aware of the group->mutex -- some of them use the
iommu_group_mutex_assert() function that could be potentially in the path
of an attach_dev callback function invoked by the __iommu_attach_device().

Worth mentioning that the iommu_deferred_attach() will soon need to check
group->resetting_domain that must be locked also.

Thus, grab the mutex to guard __iommu_attach_device() like other callers.

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Tested-by: Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Samiullah Khawaja <skhawaja@google.com>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
drivers/iommu/iommu.c

index 2ca990dfbb884fa9ddf2510113b6cc7a063bf6c3..170e522b5bda4504df8eb6e65961ee140b0158bc 100644 (file)
@@ -2185,10 +2185,17 @@ EXPORT_SYMBOL_GPL(iommu_attach_device);
 
 int iommu_deferred_attach(struct device *dev, struct iommu_domain *domain)
 {
-       if (dev->iommu && dev->iommu->attach_deferred)
-               return __iommu_attach_device(domain, dev, NULL);
+       /*
+        * This is called on the dma mapping fast path so avoid locking. This is
+        * racy, but we have an expectation that the driver will setup its DMAs
+        * inside probe while being single threaded to avoid racing.
+        */
+       if (!dev->iommu || !dev->iommu->attach_deferred)
+               return 0;
 
-       return 0;
+       guard(mutex)(&dev->iommu_group->mutex);
+
+       return __iommu_attach_device(domain, dev, NULL);
 }
 
 void iommu_detach_device(struct iommu_domain *domain, struct device *dev)