The fake-hid-input API (bluez/input/ plugin) calls the ->connect callback quite randomly. The ->connect callback (inside struct fake_hid, NOT struct fake_input) is called inside input_device_connect() and fake_hid_connadd(). However, input_device_connect() calls it only if the "fake" member is set, which actually is NEVER set at this point, because the "fake" member is set on "hsp" connections only and never on "hid" connections. input_device_connect() works only on "hid" connections (which is actually written in uppercase "HID" and the device creation uses lowercase "hid" which should be fixed for consistency (although it might be case insensitive, I don't know)). So input_device_connect() never calls this ->connect callback. fake_hid_connadd() calls the callback only if a second connection is created on the same device (which never happens because fake_hid_connadd is only called on PSM 0x11/0x13 HID connections and there can only be ONE connection on this PSM). So the whole ->connect callback in struct fake_hid seems useless, same as the code testing whether fake_hid_connadd() already has an open connection. All in all it is quite confusing reading input/device.c because the fake_input structure is used for two purposes, once rfcomm connections of "hsp" connections and second for HID-fake connections which is l2cap. It is not clear to me, why devices registered with fake_input_register use the same structure (struct fake_input) as fake-HID devices registered with input_device_register(and then fake_hid_connadd). The latter one only need the "uinput" member of this structure while all other members are redundant. I try to implement a Wii-Remote driver as fake-hid input driver which actually works and allows the remote to establish connections to my host and vice-versa, however, trying to understand input/device.c code is quite impossible and getting this ->connect callback to work is also not possible. I would be glad to get some help here because my wiimote-uinput driver needs some initialization before the ->event callback can be called and doing this inside the uinput_setup callback seems misplaced as long as there is a ->connect callback. David -- To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
