Disable devices?

hansolo77

I'm having a constant reoccurring issue with my setup, in where the system has a hard time deciding what controller it should be using when launching games. Any games that run inside RetroArch using the Libretro system works fine. I've also got a few non-LR ports running like they should. But I've had a go about it in a weird way.

History:
I wanted to build a retro system that would let me play with an Xbox controller 99% of the time. That was easy. The remaining 1% of the time, I wanted to be able to use a legit Nintendo NES controller, or a Sega Genesis controller, or an Atari Joystick. I got all the necessary Controller->USB Adapters (Raphnet for NES, and MayFlash for Sega/Atari). EmulationStation and RetroArch was able to identify and use them just fine. However, I had problems with RetroArch and some ports not accepting input from those other controllers unless I ran a secondary driver (XBOXDRV) to map them to keyboard controls. I then had to leave EmulationStationi and RetroArch "unconfigured" with those controllers, and instead mapped a keyboard. Everything seems to be working.

But it's not. Every now and then, something will get crossed. My Xbox controller doesn't respond in the game, but if I plug in an NES or Sega controller, suddenly it works. The first solution to this was to have @meleu create a script that should be the "be all end all" solution by assigning specific devices to be Player 1, Player 2, etc. Thus the Joystick-Selection tool was made. But it still didn't solve the problem. I've gone back and tried resetting it over and over again, and things still get messed up somewhere.

My most recent adventure was to install OpenBOR to my system. It's a port that doesn't use a controller configuration from EmulationStation, and doesn't run through RetroArch. So it requires it's own controller configuration. By default, it works with a keyboard, and that DOES work. @cyperghost had written a series of scripts to help out with this, by allowing you to download a pre-made configuration. I tried that. It didn't work. I couldn't figure out why, as everybody else seemed to be getting along great. So I figured the old "controller assignment" issue was back again. I took my console apart and started unplugging things. With only the Xbox dongle (and the MayFlash adapter actually) connected, it was working fine. But once I connected my Raphnet adapter, it just died.

So somewhere, some how, the Raphnet is getting assigned as the #1 device to be used for all situations. And only through using EmulationStation, RetroArch, or a series of runcommand-onstart scripts to use XBOXDRV am I able to override it. It's a real pain in the neck. This is what is prompting the creation of this thread. My asking for help again.

Is there a way to disable a device so it's not being detected and identified? Specifically, I'm thinking about doing something like another runcommand-onstartmodification similar to:

if system = openbor
disable raphnet

Of course, re-enable through the runcommand-onend. I think this will solve most, if not all my problems. If I can specifically disable a device, then I won't have that issue. I can even figure out the precise /dev location. I just don't know Linux well enough make this happen.

Can you help me?

mitu

I think the easiest way to 'disable' a controller is to modify the permissions on the corresponding /dev/input entry. By default, I think any /dev/input/js* will be set with owner=root, group=input and the permission owner=rw, group=rw (rw=read/write). The pi user is the input group so it's able to access the controller device. If you can selectively remove the group=rw permissions on the device in the onstart script and re-add them at the end.

hansolo77

Ah.. you think blocking rw access will "disable" the device? I did some searching and came up with a 3rd party tool "hud-ctrl" but it didn't work. I will try modifying the access to the device on my next day off. It would be a chmod thing right?

mitu

@hansolo77 said in Disable devices?:

It would be a chmod thing right?

Yes, something chmod g-rw /dev/input/jsX to disable read/write access and then chmod g+rw /dev/input/jsX to restore it afterwards. Make sure you use sudo chmod ... to run the commands.

cyperghost

@mitu Yes that seems to be the hard way but is a route. But how can be assured which controller is related to which device-tree-number?

hansolo77

(Just turning gears in my head here, can't test the theory right now)....

Can you chmod through /dev/input/by-id or /dev/input/by-path rather than /dev/input/jsX? I seem to recall having difficulties with the Joystick Selection tool because the js# would change on random during boot up.

mitu

@hansolo77 No, symlinks have their own permissions. However, you can use realpath to get the target of the symlink.

# by id
pi@retropie:~ $ ls -l /dev/input/by-id/usb-0583_USB_2-axis_8-button_gamepad-joystick
lrwxrwxrwx 1 root root 6 Jul 19 02:46 /dev/input/by-id/usb-0583_USB_2-axis_8-button_gamepad-joystick -> ../js0
pi@retropie:~ $ realpath /dev/input/by-id/usb-0583_USB_2-axis_8-button_gamepad-joystick
/dev/input/js0

# by path
pi@retropie:~ $ ls -l /dev/input/by-path/platform-3f980000.usb-usb-0\:1.4\:1.0-joystick
lrwxrwxrwx 1 root root 6 Jul 19 02:46 /dev/input/by-path/platform-3f980000.usb-usb-0:1.4:1.0-joystick -> ../js0
pi@retropie:~ $ realpath /dev/input/by-path/platform-3f980000.usb-usb-0\:1.4\:1.0-joystick
/dev/input/js0

hansolo77

Something to experiment with at least..

hansolo77

Ok, got home and did a test..

pi@retropie:~ $ ls -l /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick
lrwxrwxrwx 1 root root 6 Jul 23 00:34 /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick -> ../js3

Then I did a sudo reboot and ran it again:

pi@retropie:~ $ ls -l /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick
lrwxrwxrwx 1 root root 6 Jul 23 22:26 /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick -> ../js0

If it stayed the same, I could try and use that command to temporary disable js3, but after reboot it's now js0. This is my problem I'm always having. The Pi doesn't enumerate the devices according to their plugged in location. I thought I had it figured out at one point, even official Raspberry Pi documentation said it should go "top left, bottom left, top right, bottom right". But this clearly isn't the case. It might be because the Xbox dongle hasn't fully "booted up" yet, as the little light blinks a good while before it's fully up and ready. The Raphnet is probably faster, so it takes over. But this doesn't happen all the time, which is what is so frustrating.

Is there a way to maybe have a script or something that can detect which port it's running in and THEN disable it? Something like

Execute "ls -l /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick"
Export results to "Variable"
Execute chmod to disable "Variable"
-----
Then on return from game - Execute chmod to enable "Variable"

Probably no way to do that huh? Have it intelligently identify the jsX variable by executing that command before hand.

mitu

@hansolo77 Since the ID is the same, you'll get the jsX device by executing

realpath /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick

Your variable can be something like

raphnetJS=$(realpath /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick)
chmod g-rw /dev/input/$raphnetJS

cyperghost

@mitu Wow that is a nice solution. But I think you have to strip the /dev/input because realpath is /dev/input/js# to change filestate sudo command is needed.

I think that will work ... so to disable that device chmod g-rw to enable chmod g+rw

hansolo77

I just tested this combination:

pi@retropie:~ $ ls -l /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick
lrwxrwxrwx 1 root root 6 Jul 23 22:54 /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick -> ../js0

pi@retropie:~ $ sudo chmod g-rw /dev/input/js0

The result is that the controller is still active (I plugged it in to test). Just for "fun" I ran this with that same result:

pi@retropie:~ $ ls -l /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick
lrwxrwxrwx 1 root root 6 Jul 23 22:54 /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick -> ../js0

Did I format my chmod wrong?

EDIT Another quick find:

pi@retropie:~ $ ls -l /dev/input
total 0
drwxr-xr-x 2 root root     280 Jul 24 21:02 by-id
drwxr-xr-x 2 root root     280 Jul 24 21:02 by-path
crw-rw---- 1 root input 13, 64 Jul 23 22:54 event0
crw-rw---- 1 root input 13, 65 Jul 23 22:54 event1
crw-rw---- 1 root input 13, 66 Jul 23 22:54 event2
crw-rw---- 1 root input 13, 67 Jul 23 22:54 event3
crw-rw---- 1 root input 13, 68 Jul 23 22:54 event4
crw-rw---- 1 root input 13, 69 Jul 23 22:54 event5
crw-rw---- 1 root input 13, 70 Jul 23 22:54 event6
crw-rw---- 1 root input 13, 71 Jul 24 21:02 event7
crw------- 1 root input 13,  0 Jul 23 22:54 js0
crw-rw---- 1 root input 13,  1 Jul 23 22:54 js1
crw-rw---- 1 root input 13,  2 Jul 23 22:54 js2
crw-rw---- 1 root input 13,  3 Jul 23 22:54 js3
crw-rw---- 1 root input 13,  4 Jul 24 21:02 js4
crw-rw---- 1 root input 13, 63 Jul 23 22:54 mice
crw-rw---- 1 root input 13, 32 Jul 23 22:54 mouse0

So it DID set the -rw.. but the device is still reading input from the controller... :(

EDIT 2
I give up for tonight. Did a bunch of googling, came up with a potential option from here:
https://github.com/AntiMicro/antimicro/wiki/Methods-for-Disabling-Joystick-Reading-in-Linux#permission-change
Essentially, using chmod 000 rather than chmod g-rw. Unfortunately, it didn't work either. I tried doing it on the jsX, event, and the by-id path. Nothing I tried worked. The controller was still responsive even though I confirmed through ls -l that the device was no longer readable or writable. Bummer there.. Any other ideas?

hansolo77

I think I might have found a solution. Something I remember coming across before.. If the "raphnet" adapter is identified as js0, I can do a sudo mv to change it's ID. SO...

I run a test.

pi@retropie:~ $ ls -l /dev/input/by-id
total 0
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-event-joystick -> ../event6
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-if01-event-joystick -> ../event7
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-if01-joystick -> ../js4
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-joystick -> ../js3
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-event-kbd -> ../event3
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if01-event-mouse -> ../event4
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if01-mouse -> ../mouse0
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if02-event-kbd -> ../event5
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-©Microsoft_Xbox_360_Wireless_Receiver_for_Windows_E1594E70-event-joystick -> ../event2
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-©Microsoft_Xbox_360_Wireless_Receiver_for_Windows_E1594E70-joystick -> ../js2
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-raphnet.net_nes2usb_1228-event-joystick -> ../event1
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-raphnet.net_nes2usb_1228-joystick -> ../js1

The "raphnet" is located on js1. Everything should work like it should. Test running OpenBOR, and YUP! The Xbox controller works just fine. Reboot and test again:

pi@retropie:~ $ ls -l /dev/input/by-id
total 0
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-event-joystick -> ../event6
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-if01-event-joystick -> ../event7
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-if01-joystick -> ../js4
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-Mayflash_Ltd_Mayflash_MD_USB_Adapter-joystick -> ../js3
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-event-kbd -> ../event3
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if01-event-mouse -> ../event4
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if01-mouse -> ../mouse0
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-Microsoft_Microsoft®_2.4GHz_Transceiver_v8.0-if02-event-kbd -> ../event5
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-©Microsoft_Xbox_360_Wireless_Receiver_for_Windows_E1594E70-event-joystick -> ../event2
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-©Microsoft_Xbox_360_Wireless_Receiver_for_Windows_E1594E70-joystick -> ../js2
lrwxrwxrwx 1 root root 9 Jul 26 23:55 usb-raphnet.net_nes2usb_1228-event-joystick -> ../event1
lrwxrwxrwx 1 root root 6 Jul 26 23:55 usb-raphnet.net_nes2usb_1228-joystick -> ../js0

(ok so I just copied the same block and changed it but...)
Now the "Raphnet" is located on js0. Test it in OpenBOR and the result is the Xbox controller doesn't work. So let's try this... Reboot, confirm "raphnet" is located on js0. It is, so I execute this:

sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick /dev/input/js99
sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick /dev/input/event99

Then I try and run OpenBOR. The game is working with the Xbox controller like it should again! The trick here is to move the controller's ID to something beyond all the other controller ID's. The game appears to then be loading up the next "first" controller Linux has set up (in this case, the Xbox). The hard part it took me a while to figure out is that I have to do this BEFORE launching an OpenBOR game. If not, the configs are out of sync or something and don't work even if I remember to move them and re-load the game. It only works if I move the ID before the first time I load an OpenBOR game. But it works!

So now I just need a way to add this to my runcommand-onstart script. I tried creating an IF/THEN for the $1 system of "openbor" but it didn't work. I also tried to flat out add the commands to very top of the script without any IF/THEN, but that didn't work either.

Any thoughts?

hansolo77

Ok, got a working runcommand-onstartscript going, but it's not pretty. Had some complications but it's all figured out now.

if [ "$1" = "openbor" ]
then
sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick /dev/input/js99
sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick /dev/input/event99
fi

The "not pretty" part is that this is a suicide modification. I don't know how to store the original ID's as variables to be recalled in the runcommand-onend script to put things back to normal. Going this route solves the problem for OpenBOR, but kills any other games that would use that device. What I need is a script that will do this:

if [ "$1" = "openbor" ] && [ "/dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick" = "js0" ]
then
# store variable for /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick as this changes randomly too and ties with the jsX
sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick /dev/input/js99
sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick /dev/input/event99
fi

and runcommand-onend:

if [ "$1" = "openbor" ] && [ "/dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick" = "js99" ]
then
# recall stored variable for eventX
sudo mv /dev/input/js99 /dev/input/js0
sudo mv /dev/input/event99 /dev/input/eventX #variable
fi

Does this sound possible? I know my &&'s and ='s are probably wrong, this is coming from a google search lol. Does the runcommand-onend remember the $1 variable when a game is exiting?

hansolo77

Anybody have any experience with this care to offer some helpful advice? :) Sorry I gotta ask. Once I get this figured out and working, I promise to leave you guys alone for a month or so before I try and do something else lol.

mediamogul

@hansolo77

I'm not entirely sure I understand the workflow you're aiming for, but if not exactly correct, this may at least serve as inspiration for a better solution.

In theory, the following should store the two event names as variables, export them and then check to see if they're present to allow the if/then statement to continue.

runcommand-onstart:

RNJS="$(find /dev/input/by-id/ -name '*raphnet.net_nes2usb*USB-joystick*')"
RNEJS="$(find /dev/input/by-id/ -name '*raphnet.net_nes2usb*USB-event-joystick*')"

export RNJS RNEJS

if [ "$1" = "openbor" ] && [ -n "$RNJS" ] && [ -n "$RNEJS" ]; then
  sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick /dev/input/js99
  sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick /dev/input/event99
fi

With the event name variables exported, the following should put everything back as it was.

runcommand-onend:

if [ "$1" = "openbor" ] && [ -n "$RNJS" ] && [ -n "$RNEJS" ]; then
  sudo mv /dev/input/js99 "/dev/input/by-id/""$RNJS"
  sudo mv /dev/input/event99 "/dev/input/by-id/""$RNEJS"
fi

Does the runcommand-onend remember the $1 variable when a game is exiting?

It is aware of that variable, yes.

hansolo77

@mediamogul Thanks a lot for this! I will experiment and report back. I wasn't sure how a bash script would work as far as storing variables. My experience in writing a "program" from scratch is limited to my days in high school with TI-BASIC on the TI-85 graphing calculator. LOL! It looks like we name the variable, then the command to locate it. The parts I'm hung up on are the programming commands, like find, -name, and -n. find is pretty self explanatory, as is -name. Reading it across though, are you sure this is correct? I'm not defining $RNJS, just RNJS (missing the preceeding $)... or is that what the export does? Also, I'm still not seeing where the test against the variable is detecting it as js0... It looks like all we're doing is testing for the existence of the $RNJS variable...

What if we do:

if [ "$1" = "openbor" ] && [ "/dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick" = "js0" ]; then
  RNJS="$(find /dev/input/by-id/ -name '*raphnet.net_nes2usb*USB-joystick*')"
  RNEJS="$(find /dev/input/by-id/ -name '*raphnet.net_nes2usb*USB-event-joystick*')"  
  export RNJS RNEJS
  sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick /dev/input/js99
  sudo mv /dev/input/by-id/usb-raphnet.net_nes2usb_1228-event-joystick /dev/input/event99
fi

Or is that just overly complicating things?

mediamogul

@hansolo77 said in Disable devices?:

It looks like all we're doing is testing for the existence of the $RNJS variable...

That is all that it's doing. I'm really not sure how you'd correlate an event name to a jsX assignment for the if/then statement. Perhaps you could get there by extracting and manipulating information from:

cat /proc/bus/input/devices

in some way, but to my knowledge you can't automatically correlate them by just using:

if  [ "/dev/input/by-id/usb-raphnet.net_nes2usb_1228-joystick" = "js0" ]; then

hansolo77

Well I'm getting no where fast. I thought I had the solution, by moving the js assignment. It worked, but now it's not. Even when doing it manually. Now I'm having a problem where the Xbox controller isn't being enumerated as js1. Instead, it's coming up randomly too as js4 sometimes (so OpenBOR's controller mappings need changed and it says "P3 BUTTON 1" etc. Then other times it's coming up as js1 (Player 2), or even a js6 (Player 5) one time. This whole things is ridiculously complicated now. I think I'm going to skip all of it and add another XBOXDRV setting. I really didn't want to do that, because it seems like each new entry takes a little bit longer for the runcommand-onstart to process. But it looks like it's still the best way to go.

mediamogul

@hansolo77

You never can tell, inspiration usually strikes me just as I've given up. However, xboxdrv will definitely work. One way to keep the script size down is to offload the lengthy xboxdrv launch commands to a separate script, leaving just the if/then statements that call the script in one line. I run a little over 300 xboxdrv maps and my runcommand-onstart file is only 32k.