On Fri, May 23, 2025 at 07:49:40AM +0000, Tian, Kevin wrote:
From: Nicolin Chen nicolinc@nvidia.com Sent: Sunday, May 18, 2025 11:21 AM
Now the new ucmd-based object allocator eases the finalize/abort routine, apply this to all existing allocators that aren't protected by any lock.
Upgrade the for-driver vIOMMU alloctor too, and pass down to all existing viommu_alloc op accordingly.
Note that __iommufd_object_alloc_ucmd() builds in some static tests that cover both static_asserts in the iommufd_viommu_alloc(). Thus drop them.
I may overlook something, but at a quick glance the following check Is not covered?
static_assert(__same_type(struct iommufd_viommu, \((drv_struct *)NULL)->member)); \
The container_of() inside __iommufd_object_alloc_ucmd() covers that:
#define container_of(ptr, type, member) ({ \ void *__mptr = (void *)(ptr); \ static_assert(__same_type(*(ptr), ((type *)0)->member) || \ <<== here __same_type(*(ptr), void), \ "pointer type mismatch in container_of()"); \ ((type *)(__mptr - offsetof(type, member))); })
Btw this patch doesn't convert all object allocations to the new helper, e.g. ioas. What is the criteria for when to use the old helpers vs. the new helpers?
Criteria is "that aren't protected by any lock" mentioned in the line 2. The ioas has an ioas_creation_lock around finalize() :(
Thanks Nic