This way, user-space can figure out which profile currently is
configured. Take note that this may also involve a different button
mapping, and only profile 0 has paddles forcibly reported as their
distinct events (otherwise mapped to A, B, X, and Y by default).

Link: #428
Signed-off-by: Kai Krakow <[email protected]>

Link: #198
Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

`dkms remove` may fail because a previous installation has been broken
or tampered with. This will cause the remaining uninstallation to fail.
Fix this by logging that we proceed anyways.

Suggested-by: flodavid <[email protected]>
Link: #563
Signed-off-by: Kai Krakow <[email protected]>

This change is backwards compatible with older versions.

Link: https://github.com/systemd/systemd/blob/781d9d0789379d1ea1f2ecefb804d41e9c8b6c38/NEWS#L22-L32
Closes: #559
Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

The system should trigger loading of the `uhid` module when `bluetooth`
is loaded to properly support BLE devices which provide the HID bus
from user-space through bluez.

This may still leave a gap for a race-condition if both `bluetooth` and
the BLE device appear at the same time. `uhid` should be present before
any device even connects from user-space, otherwise `hid-xpadneo` may
not see the HID bus early enough.

I'm unsure if we also should provide `weakdep` to trigger inclusion of
`uhid` into initrd - but since Xbox controllers are not critical input
devices during early boot, this is probably not our business. OTOH,
`softdep` may already trigger inclusion on its own (untested).

Hint: This feels wrong because xpadneo is the HID consumer, not the HID
provider. We should not need to do that. But it prevents bug reports
from users in the future.

Tested-by: flodavid <[email protected]>
Link: #563 (comment)
Signed-off-by: Kai Krakow <[email protected]>

AI checks identified the check being inverted.

Fixes: 25d0ea1 ("xpadneo, installer: Get rid of ERTM work-around by requiring kernel 5.12")
Signed-off-by: Kai Krakow <[email protected]>

Do not leak the error pointer into the battery struct if psy
registration failed.

Signed-off-by: Kai Krakow <[email protected]>

If `ida_simple_get()` fails, we didn't properly bubble up the error to
the device manager.

Signed-off-by: Kai Krakow <[email protected]>

battery.desc.name was never freed during device removal because this
previously caused a kernel crash when xpadneo tried to free the name
itself. Fix this by using the devm memory manager so we don't need to
clean up memory by ourselves.

Also fix input_dev->name which has been an oversight previously.

Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

This confuses code analysis.

Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>

Signed-off-by: Kai Krakow <[email protected]>