Introduce nhop_ref_locked_object().

[markj]

Summary

Use it when we know that the object must have a positive reference
count. This provides two benefits:

• We can assert that the refcount is positive.
• We can use a cheaper operation to acquire the reference count.

I am open to alternate name suggestions.

[melifaro]

Thank you for raising this diff! Was going to do the same, but you managed to do it faster :-)
LGTM, please see the naming comment :-)

sys/net/route/nhop_ctl.c

609 ↗

(On Diff #73969)

Naming is hard :-)
Maybe we can do it slightly differently, like leave this function as nhop_ref_object and the original one as nhop_try_ref_object ?

I'd prefer not to mention locked in the name, as nhop objects have their locks, but we don't use them in refcounting.

What do you think?

[markj]

sys/net/route/nhop_ctl.c

609 ↗

(On Diff #73969)

I think that's much better than my name, will fix.

[markj]

Introduce nhop_try_ref_object() instead.

I note that we also have nhop_ref_any() and nhop_free_any() as well,
but they are unused.

[melifaro]

In D25535#565494, @markj wrote:

Introduce nhop_try_ref_object() instead.

I note that we also have nhop_ref_any() and nhop_free_any() as well,
but they are unused.

That's the route multipath part, which will be introduced soon.