Hi Jiri, Henrik,

On 04/23/2012 12:27 AM, Jiri Kosina wrote:
> On Sun, 22 Apr 2012, Henrik Rydberg wrote:
>
>> This patchset contains a resolution to the problem with
>> driver-specific modification of the report descriptor. The core of the
>> problem lies with the dual semantics of hid_parse_report(), which is
>> therefore split into two functions. As a consequence, the hid core can
>> handle the rebind case internally, with no changes to the
>> drivers. Out-of-tree drivers will work fine as long as they operate in
>> the same way as the in-kernel drivers.
>>
>> The first two patches are bug fixes found in the process. I am hoping
>> for some test feedback on those. The third patch is the main one.
>
> Hi Henrik,
>
> thanks a lot for looking into this. I will look into the patches in more
> detail tomorrow. I am now just adding Nikolai to CC, as he was the first
> one to bring this up and actually has a driver that exposed the fact that
> just freeing/reinstatiating rdesc is not enough.

I've managed to throw together a script which rebinds a device to a specific
out-of-tree HID driver automatically. Please find it attached. I'll be
bundling it with the modules, probably.

I'm planning to use it with udev rules looking like this one:

SUBSYSTEM=="hid", ACTION=="add", ENV{HID_ID}=="0003:0000172F:*", \
         RUN+="/usr/local/bin/digimend-rebind"

So, the out-of-tree HID module problem is finally solved, at least for my
current uses. Thanks a lot!

Sincerely,
Nick