Re: uhidev(9) improvements

[Nick Hudson <nick.hudson%gmx.co.uk@localhost>]

On 25/01/2022 01:09, Taylor R Campbell wrote:
> The attached patch series partially fixes a bug in the uhidev(9) API
> (uhidev_stop didn't do anything at all; now it is merely racy).
>
> But more importantly, the patch series makes struct uhidev_softc
> opaque, so the uhidev(9) API can have a more stable ABI.  This will
> give us better latitude to fix this race -- and potentially others --
> later on and pull them up to netbsd-10 after it branches.

Looks good to me.

+       /* XXX Can this just use sc->sc_udev, or am I mistaken?  */
+       usbd_interface2device_handle(sc->sc_iface, &dev);

I think yes.


> I have tested uhid(4) and ukbd(4).  I haven't tested ucycom(4), which
> uses uhidev(9) in a way that nothing else does -- to do async xfers on
> the output pipe.  Can anyone review and/or test ucycom(4)?


ucycom(4) works as well as it did before the change.

Nick