speedtch Makefile, 1.12, 1.13 cxacru2.c, 1.3, 1.4 usbatm.h, 1.5, 1.6 usbatm2.c, 1.15, 1.16

[Roman Kagan]

  Hi Duncan,

On Fri, Jan 28, 2005 at 02:58:20PM +0000, Duncan Sands wrote:
> New registration model, basically Roman's patch plus control over USB probing
> given to the minidrivers.  Hopefully Matthieu will like it :)  It works like
> this (actually, since I haven't tested it, maybe it doesn't work at all):
> 
> minidrivers register a struct usbatm_driver with usbatm, and register a
> struct usb_driver with usb.  Most minidrivers should be happy with letting
> usbatm do all the work: they can just use the usbatm probe and disconnect
> methods:
> 
> ...
>         .probe          = usbatm_usb_probe,
>         .disconnect     = usbatm_usb_disconnect,
> ...
> 
> However, if a driver needs extra flexibility it can use its own methods:
> 
> ...
> 	.probe		= matthieu_usb_probe,
> 	.disconnect	= matthieu_usb_disconnect
> ...
> 
> It is OK to register minidrivers in the probe method, so matthieu_usb_probe
> etc could be:
> 
> int matthieu_usb_probe(...) {
> 	...query the device, find the best driver to use...
> 	if ((ret = usbatm_register(best_driver)) < 0)
> 		return ret;
> 	if ((ret = usbatm_usb_probe(intf, id)) < 0) {
> 		usbatm_deregister(best_driver);
> 		return ret;
> 	}
> 	return 0;
> }

Now I'm really confused.  I thought the whole purpose of these
usbatm_(de)register was to eliminate the need for the subdrivers to
write their own probe and disconnect adhering to a certain calling
order.

If you decide to make the subdrivers provide their probe methods, why do
you need all this complexity with registration at all?

Instead you can make usbatm_usb_probe take a pointer to the subdriver as
a parameter:

int usbatm_usb_probe (struct usb_interface *intf,
			const struct usb_device_id *id,
			struct usbatm_driver *driver)
{
	...
	instance->driver = driver;
	...
}


In subdriver's probe you pass your struct usbatm_driver, and that's all
you need.  No registration, no locks, no lists, no risk of multiple
subdrivers registering the same usb_device_id (breaking the list stuff
BTW).

Or am I overlooking something?

Cheers,
  Roman.