Merge linux upstream HID input patch

[BetaXOi]

https://git.kernel.org/pub/scm/linux/kernel/git/dtor/input.git/commit/?id=6ce901eb61aa30ba8565c62049ee80c90728ef14

current status in master branch:
amlogic-3.10) use LibreELEC/linux-amlogic 95ba9d626c0fce672caa296f5911ab9190881642 Fixed
amlogic-3.14) use CoreELEC/linux-amlogic f1bb00821950b0f5cab1b8a820c697ba9fbd3183 NOT Fixed
amlogic-4.9) use CoreELEC/linux-amlogic 6f607ec78af4fc755f41a7dd7e73ad6542f9143b Fixed

In my case, i wrote a shell to inject input_event to event device of my gamepad when catch a special input_event triggered by a button pressed and hold, the kernel without the patch will auto report a input_event of button release, but it works fine if kernel be patched.

Maybe we should merge this patch to packages/linux/patches/amlogic-3.14

[BetaXOi]

I have a LeWGP-201 gamepad, it will create two event device, one will be used for control on Android, and another be used for other, looks like below:

Sx05RE:~ # ll /dev/input/by-id
total 0
lrwxrwxrwx 1 root root 9 Jan 1 08:12 usb-MY-POWER_LeWGP-201-event-if01 -> ../event5
lrwxrwxrwx 1 root root 9 Jan 1 08:12 usb-MY-POWER_LeWGP-201-event-joystick -> ../event4
lrwxrwxrwx 1 root root 6 Jan 1 08:12 usb-MY-POWER_LeWGP-201-joystick -> ../js1
Sx05RE:~ # ll /dev/input/by-path/
total 0
lrwxrwxrwx 1 root root 9 Jan 1 08:00 platform-c8100580.meson-ir-event -> ../event6
lrwxrwxrwx 1 root root 9 Jun 22 2018 platform-dwc2_a-usb-0:1:1.0-event-joystick -> ../event4
lrwxrwxrwx 1 root root 6 Jun 22 2018 platform-dwc2_a-usb-0:1:1.0-joystick -> ../js1
lrwxrwxrwx 1 root root 9 Jun 22 2018 platform-dwc2_a-usb-0:1:1.1-event -> ../event5
lrwxrwxrwx 1 root root 9 Jun 22 2018 platform-gpio_keypad.47-event -> ../event0

The gamepad has no select button but back button for Android, so i'm trying to remap back to select, and remap power to select+start.
I did some test, injecting input_event of BTN_A pressed and holded to event-joystick , and input_event of release after 3 seconds, see: BetaXOi@ce9a38c
But i got a invalid result, input_event of release will auto report by kernel:

Event: time 1560437407.276170, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1560437407.276170, type 1 (Key), code 304 (BtnA), value 1
Event: time 1560437407.276170, -------------- Report Sync ------------
Event: time 1560437407.277813, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1560437407.277813, type 1 (Key), code 304 (BtnA), value 0
Event: time 1560437407.277813, -------------- Report Sync ------------

Event: time 1560437410.276373, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1560437410.276373, -------------- Report Sync ------------

It works fine if kernel patched.

Event: time 1560439280.104830, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1560439280.104830, type 1 (Key), code 304 (BtnA), value 1
Event: time 1560439280.104830, -------------- Report Sync ------------
Event: time 1560439283.105041, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1560439283.105041, type 1 (Key), code 304 (BtnA), value 0
Event: time 1560439283.105041, -------------- Report Sync ------------

[BetaXOi]

My two LeWGP-201 work fine with my custom kernel patched and remap binary program:
BetaXOi@7380726

I add two line in autostart.sh

/storage/remap /dev/input/by-path/platform-dwc2_a-usb-0:1:1.1-event /dev/input/by-path/platform-dwc2_a-usb-0:1:1.0-event-joystick &
/storage/remap /dev/input/by-path/platform-dwc2_b-usb-0:1:1.1-event /dev/input/by-path/platform-dwc2_b-usb-0:1:1.0-event-joystick &

[shantigilbert]

Thanks!
could this remap program work to send a command on certain key combinations? like to kill a running application? maybe we an use to quit all emulators with a key-combo

[BetaXOi]

Yes, we can monitor input_event of event device, and do something we want.
But i don't know how to be suitable for all devices, my source code of remap is only suitable for my LeWGP-201

[shantigilbert]

shouldn't all evdev devices work?

[BetaXOi]

I have no other gamepad to test, i don't know whether the ScanCode is same for all gamepad.

Event: time 1420070488.321441, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1420070488.321441, type 1 (Key), code 304 (BtnA), value 1
Event: time 1420070488.321441, -------------- Report Sync ------------
Event: time 1420070488.433432, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1420070488.433432, type 1 (Key), code 304 (BtnA), value 0
Event: time 1420070488.433432, -------------- Report Sync ------------
Event: time 1420070489.553487, type 4 (Misc), code 4 (ScanCode), value 90002
Event: time 1420070489.553487, type 1 (Key), code 305 (BtnB), value 1
Event: time 1420070489.553487, -------------- Report Sync ------------
Event: time 1420070489.649473, type 4 (Misc), code 4 (ScanCode), value 90002
Event: time 1420070489.649473, type 1 (Key), code 305 (BtnB), value 0
Event: time 1420070489.649473, -------------- Report Sync ------------

[shantigilbert]

how can I test?

[shantigilbert]

I guess they are not, but the key code seems to be the same

Event: time 1561182439.110425, type 1 (Key), code 304 (BtnA), value 0
Event: time 1561182439.110425, -------------- Report Sync ------------
Event: time 1561182439.626431, type 1 (Key), code 305 (BtnB), value 1
Event: time 1561182439.626431, -------------- Report Sync ------------

[BetaXOi]

compile evtest.c and run ./evtest <evdev> catch button info you want.
Then implement the logic code in remap

[shantigilbert]

my other controller does give scancode

Event: time 1561182667.940376, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1561182667.940376, type 1 (Key), code 304 (BtnA), value 1
Event: time 1561182667.940376, -------------- Report Sync ------------
Event: time 1561182668.086607, type 4 (Misc), code 4 (ScanCode), value 90001
Event: time 1561182668.086607, type 1 (Key), code 304 (BtnA), value 0
Event: time 1561182668.086607, -------------- Report Sync ------------
Event: time 1561182668.367918, type 4 (Misc), code 4 (ScanCode), value 90002
Event: time 1561182668.367918, type 1 (Key), code 305 (BtnB), value 1
Event: time 1561182668.367918, -------------- Report Sync ------------
Event: time 1561182668.514142, type 4 (Misc), code 4 (ScanCode), value 90002
Event: time 1561182668.514142, type 1 (Key), code 305 (BtnB), value 0
Event: time 1561182668.514142, -------------- Report Sync ------------

so it seems they are maybe the same scancodes and keycodes?

[BetaXOi]

I have no idea, but i think maybe it's feasible just use Key and Value without ScanCode even

[shantigilbert]

it would be cool to send SIGTERM to a running program with a key combination

maybe define it at runtime
./evkill -K304+305+306 retroarch

[BetaXOi]

Yes, i will write a evkill demo later, maybe tonight.

[BetaXOi]

BetaXOi@ec1b19d

[shantigilbert]

Thanks! I made a package with all the tools, you see anything you would like to change ?

0bf6245

[shantigilbert]

I tried evkill and it works perfectly! now all the emulators that don't have insta-quit can be closed with the gamepad!

I do have a small request, not sure how possible it is (I know close to nothing on this) could it be possible to instead of /dev/input/eventx we could use /dev/input/jsx ? This will make it easier to set-up

thanks again!

[BetaXOi]

Yes,it's possible, i will improve the code

[shantigilbert]

Thank you! I made a few changes, it does not print anything and it exits after the command is called.

[BetaXOi]

please review PR #30

[zang74]

Does this patch mean that volume can be controlled outside of RetroArch?

[shantigilbert]

@zang74 no, but I am curious as to why you would think that, maybe I missed something?

This patch (the original HID input) is a patch that fixes some particular issues with certain keyboards.
The second one is a small program that @BetaXOi made for quitting emulators with a game pad key-combo, because some emulators needed keyboard to quit, not anymore :)

[zang74]

It's in regards to the second one. If it's possible to map a key combo to quitting an emulator, is it not possible to map another to system audio volume?

[shantigilbert]

Not with this current program, it is made solely to quit a running program (in this case an emulator).