• Add short comment about unmap_mapping_range() parameters
• min -> ulmin

These symbols are required to load on 13-STABLE kernel module built on 13.0-RELEASE. 13-STABLE alone does not require them exposed

In D32169#726062, @hselasky wrote:

Does "buildkernel WITH_OFED=YES" still build?

Use PMAP_HAS_DMAP instead of LP64
Optimize sg_pcopy_to_buffer() for !PMAP_HAS_DMAP case
Move kmap/kunmap implementation in to header file
Drop debugging printf

Thank you for the time spent on it

Replace KASSERT with printf.

Or you are talking to be more error-prone to bugs in userspace drivers?

IMO, Although printf is better in general case, KASSERT is still good choice for i915kms.
It has some cleanup code which removes old mappings[1] and it works
So when i915kms mmaps same pages again (it happens sometimes), nothing bad happens.
I think we should not try to hide problem if the cleanup stopped to work as it happened e.g. on 5.5->5.6 update

Revision updated according to last @kib comment

Add check for owning of the lock for page->object

Do full cycle retry if vm_page_busy_acquire() has failed

Pass VM_ALLOC_WAITFAIL flag to vm_page_busy_acquire()

Allow to steal pages grom any object

Move "page still belongs to shm_obj" check to right place

Add unconditional execution of vm_pager_page_unswapped()

Eliminate lock ordering through use of vm_page_remove()

Clarify _Static_assert reason

It is unmodified file from OpenBSD: https://github.com/openbsd/src/blob/master/sys/dev/pci/drm/include/linux/poison.h

Yes, it is partial. kfree_rcu in linux_file_free() is new as compared with old version.

In D31672#723880, @hselasky wrote:

The problem is that the refcount mechanism access f->_file->f_count . You need to free both "f" and "_file" using kfree_rcu() for this to work.
"f" is the Linux file structure and "_file" is the FreeBSD kernel's file structure.

Revert close() handler to previous state

In D32068#723767, @hselasky wrote:

Looks good. Was this code tested?

In D31672#723772, @hselasky wrote:

Are you sure this works, that other structure pointers inside the Linux file structure won't be accessed and will also need to be freed by RCU?

Wait for a grace period when anonymous file is closed with fput().