I believe if you move the block from here until MODULE_VERSION to the end of the file all the prototypes can go away as the functions will just be ordered correctly and frankly having the MODULE lines in the middle of the file feels weird (at least from the [PCI] drivers I have cached in my mind.

The put the bus probe/attach/detach functions before that. and then on top the auxiliary ones, usually the read/write wrappers ones comes first.

Maybe I al too wpaul-ified with the voices in my head ;-) but it's a structure I am used to and I assume this does not diff to NetBSD anymore in any reasonable way, does it? If it does, then disregard.

label at beginning of line.

dead code below?

Normally I would expect SRCS= udbc.c SRCS+= all the (bus etc) header files or with a \ but the implementation first.

I find this highly confusing to have += and \ mixed in a file; also the re-wrapping of long lines feels annoying; Can we not just expand them into one-a-line (maybe upfront and then add the single new line).

I assume the current format is for grouping by type but that could easily be a comment for a block then if we don't want to sort them alphanum; kind-of make it more like the level up modules/Makefile?

This hyphen was for a long abandoned parser.

I'm trying to increase the consistency and predictability of the manual. Predictability is one of the greatest strengths of manuals, but it requires consistency, like we have in every other section's synopses. I gave a 5 min lightning talk about section 4 synopses at BSDCan:

https://people.freebsd.org/~ziaee/tmp/bsdcan25lightning-ziaee.webm

This should go earlyish in DESCRIPTION, HARDWARE should say something like:

The
.Nm
driver supports the debugging channel provided by XHCI compliant controllers.

If I'm understanding this patch correctly.

I think we should move these to STANDARDS