Re: Getting device 0fe6:9700 Kontron to work with NM

[Larry Finger <Larry Finger lwfinger net>]

On 12/02/2011 12:51 PM, Dan Williams wrote:
> On Fri, 2011-12-02 at 13:27 +0200, Perazim wrote:
> > Update: Turns out that this device was repacked by manufacturer without
> > documenting NIC chip change. The correct driver is qf9700 which is not
> > in the standard kernel and I found the source for on the internet and
> > built against my kernel. Once done and installed in modules, I can
> > modprobe this and device is brought up and NM recognizes it.
> >
> > Now I need to figure out how to get this to happen automatically when
> > the device is plugged.
>
> It's a kernel problem.  If it's the same USB ID as the old device with
> the old chip, then you'll need to talk to various kernel mailing lists
> (linux-usb probably, or netdev) to see how to figure this out.  The
> problem is that drivers mostly use USB IDs for autoloading, and if this
> device has the same USB IDs then it'll conflict with the old driver.
> You can't just add the same USB ID to two different drivers, there's
> more coordination that needs to happen.  Unfortunately I don't know what
> the coordination is off the top of my head :(

The general fix for this kind of thing is to find some value in the device 
configuration registers that is different for the two devices, and return driver 
load failure when the wrong one is found. An example is found for PCI ID 
0x10ec:0x8192, which occurs for both RTL8192SE and RTL8192E devices. In this 
case, drivers rtl8192se and r8192e read the PCI revision. If the value is 1, 
then rtl8192xe returns the invalid device code, while r8192e returns the invalid 
code when the revision is 0x10.

For your device, you would need to find some value in the USB configuration that 
differs, and do the same kind of logic. Both drivers will be loaded, but only 
the "right" one will be active.

Larry