RE: hid-sensor-hub and autoloading of HID sensor modules

I mean if your patch is merged, when new kernel is synced in the product line, we can't depend on auto loading of hid-sensor-hub. So we need to do something similar as you suggested.

Thanks,
Srinivas



-----Original Message-----
From: Alexander Holler [mailto:holler@ahsoftware.de] 
Sent: Monday, December 10, 2012 1:33 PM
To: Pandruvada, Srinivas
Cc: linux-iio@vger.kernel.org; linux-input@vger.kernel.org; matteo.dameno@st.com
Subject: Re: hid-sensor-hub and autoloading of HID sensor modules

Am 10.12.2012 22:22, schrieb Pandruvada, Srinivas:

This will be a problem for current implementations, which are about to get shipped.
Once I get chance, I will start looking at this idea. Meanwhile, you need to add hook to vendor and product id in the table.

Sorry, but I don't understand what that "shipped" has to do with the implementation in the kernel. And I don't seen any problems as only the implementation inside the kernel will change, nothing else.

Regards,

Alexander

Thanks,
Srinivas


-----Original Message-----
From: Alexander Holler [mailto:holler@ahsoftware.de]
Sent: Monday, December 10, 2012 1:08 PM
To: linux-iio@vger.kernel.org; linux-input@vger.kernel.org; 
Pandruvada, Srinivas; matteo.dameno@st.com
Subject: hid-sensor-hub and autoloading of HID sensor modules

Hello,

as I've just posted a patch to get rid of the vendor and device IDs for USB HID sensor hubs, there is still one problem left: hid-sensor-hub doesn't automatically load the necessary drivers (if they are modules).

Here is a suggestion about how to solve that:

Create HID groups HID_GROUP_SENSOR_ACCEL, HID_GROUP_SENSOR_TIME, ...) for every type of (supported) sensor (instead of the one HID_GROUP_SENSOR_HUB), scan for their usage pages (e.g.
HID_USAGE_SENSOR_ACCEL, HID_USAGE_SENSOR_TIME, ...) set the group and so load the final module and not hid-sensor-hub. hid-sensor-hub will then get loaded automatically.

As I don't have any real sensor hubs (and no business with them), I think others should implement such, if that idea got approved to make sense. ;) That just would be some few lines, but I couldn't test them with real-world stuff.

Regards,

Alexander