quoted

In my drawer, I have a patchset that aims to remove all usbhid
dependence, from all the drivers. Perhaps the attached patch is
something to consider here?

yep, removing usbhid dependencies is a good thing.
See my review below :)

quoted

From 5d9a791e0a9e41bcea0fcb286e2849b403675f37 Mon Sep 17 00:00:00 2001
From: Henrik Rydberg <redacted>
Date: Mon, 2 Jul 2012 20:38:21 +0200
Subject: [PATCH 3/7] hid: extend interface with report requests

---
 drivers/hid/usbhid/hid-core.c | 14 ++++++++++++++
 include/linux/hid.h           | 33 +++++++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+)

diff --git a/drivers/hid/usbhid/hid-core.c b/drivers/hid/usbhid/hid-core.c
index de1f9ac..3c618da 100644
--- a/drivers/hid/usbhid/hid-core.c
+++ b/drivers/hid/usbhid/hid-core.c

@@ -1251,6 +1251,18 @@ static int usbhid_power(struct hid_device *hid, int lvl)
        return r;
 }

+static void usbhid_send(struct hid_device *hid, struct hid_report *rep, int req)
+{
+       switch (req) {
+       case HID_REQ_GET_REPORT:
+               usbhid_submit_report(hid, rep, USB_DIR_IN);
+               break;
+       case HID_REQ_SET_REPORT:
+               usbhid_submit_report(hid, rep, USB_DIR_OUT);
+               break;
+       }
+}
+
 static struct hid_ll_driver usb_hid_driver = {
        .parse = usbhid_parse,
        .start = usbhid_start,

@@ -1259,6 +1271,8 @@ static struct hid_ll_driver usb_hid_driver = {
        .close = usbhid_close,
        .power = usbhid_power,
        .hidinput_input_event = usb_hidinput_input_event,
+       .send = usbhid_send,

the name here is a little bit misleading. You are using "send" to also
"get" reports...
Maybe "hid_request" is a better name. This will allows us to add the
missing function:
Get_Descriptor, Set_Descriptor, Get_Report Request, Set_Report
Request, Get_Idle, Set_Idle, Get_Protocol, Set_Protocol and maybe
others - even WAIT for instance :)

Sounds good, I'll ponder this a bit.

quoted

+       .wait = usbhid_wait_io,

is it really necessary to wait for IO? (I think it will not be one
line for hid over i2c...).

We can certainly skip it for the scope of your patchset, but last time
I checked, it was needed in quite a few places.

Thanks,
Henrik