libusual interface [was Re: libusual and ub]

[Pete Zaitcev]

On Thu, 15 Dec 2005 00:37:41 -0500, Bill Nottingham <notting at redhat.com> wrote:

> If you aren't using libusual to load the modules (installer), you have to
> hardcode which module to load somewhere... Jeremy was sending me a (bad :) )
> patch for this just the other day... with the ids in the module, you just
> use the modalias.

Suddenly, the problem came into focus. I wish one of you cc:-ed me on it.
Like I mentioned previously, when libusual is in effect, modprobing
backend drivers makes no sense. With that in mind, I do not see a
fundamental problem. We did have an explicit call to modprobe usb-storage.
So now we would have something like 'modprobe libusual bias="ub"'.
Is there a problem with passing of arguments?

> You still get a USB hotplug event for the USB device, with the modalias,
> etc. Obviously, running modprobe with that won't load anything, and
> libusual has already claimed the device.

Have you actually tried it? Loads like a charm here, with 2.6.14-1.1740_FC5
RPM, as it should. This is the scenario:
 - Since we build with libusual static, there is no hotplug event.
   module.alias is not involved.
 - libusual gets ->probe
 - libusual loads the driver (if not loaded)
 - The driver attaches to the device

> > Plug TEAC CD-210PU into a Rawhide system biased to ub and you'll see.
> 
> Heh, all I have is a craptastic usb-floppy.

I must have mixed you with someone. When the story with toggles just
broke, people were running with these things, fangs out, fur on fire,
as if every engineer had them. Tim bought me one such 210PU at CDW.

-- Pete