On Thu, 19 Dec 2013, David Herrmann wrote:

[off-list ref] wrote:

quoted

we used to set the parent of the input device as the parent of
the hid bus. This was introduced when we created hid as a real bus, and
to keep backward compatibility. Now, it's time to proper set the parent
so that sysfs has an idea of which input device is attached to
which hid device.

Signed-off-by: Benjamin Tissoires <redacted>
---

Hi Jiri,

well, the regression test did not showed anything bad, and a look at the hid
drivers also showed that no drivers should be armed by that.

I think this is valuable because it will give a better understanding of the
actual mapping hardware/inputs. Like a touchscreen with pen + touch will have
the same hid parent, and I expect X/Wayland to be able to detect this at some
point to keep the mapping correctly between the two input devices and the screen.

I also need that for hid-replay, so that I can be sure which input is attached
to which uhid device, and give up the heuristics I currently use.

I was just wondering where we have multiple HID devices on a single
parent, but yeah, uhid is a good example. I actually have no
objections to this patch and it looks fine. But I cannot tell whether
anyone relies on this.

I'd say we should give it a try.

Agreed, let's take it for 3.14, and if anyone complains about breakage 
caused by this (which I don't expect), we'll revert to the previous state.

-- 
Jiri Kosina
SUSE Labs