I think that the formal correctness is achieved by rendezvous_mtx.
That wmb() was for the benefit of code like:

if (rendezvous_func != NULL) {
  ...
  mtx_lock(rendezvous_mtx);
  // do real things with rendezvous_func
  mtx_unlock(rendezvous_mtx);
}

That is, rendezvous_func is assigned and used always under rendezvous_mtx.
Access outside of the lock are only to check it for being set.

oops, typo

@yuripv , I noted that you ended your suggestions with question marks...
I am not a native speaker, so I am not sure if definite or indefinite article is more appropriate in those sentences.

man4 makefile too

add makefile changes

Hmm, I see that superio.9 could become a first conditionally installed manual page in man9.
Is that okay?
Or would it be better to merge superio.9 into superio.4?
I see a couple of examples in man4, e.g., cpufreq.4.

Thank you very much for the review!
Problems should be fixed now.

cleanups based on the review

In D21886#479253, @emaste wrote:

My use case for something like this is a USB-serial console server where I might have 16 different USB-serial adapters attached to various USB hubs; I intended to address it by using the serial number in each device.

I see.
I think that the problem can be solved quite easily with devd, but maybe it's not good for embedded solutions.
Also, I am not sure how stable the unit numbers are going to be with respect to add-in USB controllers and external USB hubs.

Why is this needed at all? What's the motivation?

rebase

• drop redundant name checks
• move name checks to the check function

Any feedback at all?

The change itself looks good to me.
Thank you!

A general comment: I hope the current review summary will not be a commit message :)

In D21660#472250, @sef wrote:

That's why you can request different numbers. If you want.

So, maybe you'd better debug and fix that problem.
Or add a sleep instead of additional sync-s. And how do you know that 3 will always be enough?
Or maybe the problem was with concurrent writes somewhere else and in some cases even a million sync-s would not be enough...

In D21660#472237, @sef wrote:

Because I was making changes that happened in zfs on sync 8-). I was using multiple syncs to ensure everything did get synced out, and then examining status and structures. (And depending on timing, not everything did get synced out on the first one.)

I think that a need for sync ; sync ; sync is greatly exaggerated.
If you think that it's useful and sync 3 would be an improvement, then could you please explain why?
And without referring to "old wives' tales" :-)

Looks good to me.

A couple more nits.

If this driver just provides a single sysctl, then I think that it would be simpler to add that sysctl to acpi_wmi.
Do you plan to have any other functionality in acpi_wmi_bmof?