I am experimenting with some additional caps that use bits in capabilities2/capenable2 and that's how I ran into this.

In D39865#917264, @melifaro wrote:

Ping :-)
I'd love to get rid of if_clone_advanced() before 14 and cxgbe is the only remaining user :-)

Almost there. Now the new ifnet is created successfully but there is an extra 0 in its name. The ifnet should have been t6nex0 and not t6nex00.

The if_name2unit call in if_clone_createif_nl -> ifc_handle_unit seems to be the problem. It is returning EINVAL because it doesn't like the 0 in "t6nex0".

I tried this change but the cloned interface was not created.

Sorry I missed this previously. This man page has a dependency on groff and I'd like to get rid of it eventually (the dependency, not the man page). But my manpage-fu is weak and I'm not sure how to figure out exactly which parts need groff. Will these proposed changes make any difference as far as reliance on groff goes? Okay to check this in either way.

I'm done looking at the cxgb part, will look at cxgbe next.

Was this done with some coccinelle type script or something similar? I have a bunch of code that is not upstream yet and it is not ifAPI style. If there's a script I could run pre-commit to make sure any new code is ifAPI clean it would be quite useful.

@brd what kind of workload do you see the improvements with?

Incorporate feedback from kib.

In D37270#846696, @kib wrote:

Should the current version number for ACS also get a symbolic name?

Here are a couple of examples from the motherboard that i tested it on, with the available caps disabled and enabled.

Please add a "Fixes: e398922eaf6" to make it easier to determine the commits to MFC together.

Please go ahead. I'll follow up with commits that make this work with suspend/reset, when hw access is not allowed.

Please commit with boolean &&.