No /dev/tty when script is invoked as UDEV trigger

Question

I have raspberry pi with stock raspbian installed. I have also installed RetroPI which works just fine.

What I wanted to do, is to write small autostart script that would start emulationstation (retropi main run script) when game pad is connected via bluetooth. I mave made simple udev rule

pi@raspberrypi:~ $ cat /etc/udev/rules.d/99-zlocal.rules
SUBSYSTEM=="bluetooth",SUBSYSTEMS=="amba", ATTRS{id}=="00241011", ACTION=="add", RUN+="/usr/local/bin/emulator-controll.sh start"

which works as expected, script is ran when gamepad is paired.

The problem is, that if i run my script via eg SSH like

sudo emulator-controll.sh start        #sudo is optional here, but I want the same conditions as in udev trigger context

emulationstation as expected. However, if the same script is used as action trigger for udev rule above, it fails with errors like

Wed Aug 21 00:02:33 CEST 2019
Invoking user: root
Invoked start command
Invoking emulation station start command
Done
'unknown': I need something more specific.
tput: unknown terminal "unknown"
Segmentation fault
/usr/bin/emulationstation: line 23: /dev/tty: No such device or address
tput: unknown terminal "unknown"

I think that for some reasons (im linux newb) tty is not accessible from udev trigger context. Can I fix it somehow?

PS. I have tried to run background deamon (to use named pipes) that is launching on boot time, but effect is exactly the same - no tty in script execution context which makes me think that is is general behavior of background tasks

Script itself is irrelevant IMHO, but I will provide it as well for clarity:

#!/bin/bash
#/bin/date >> /tmp/udev.log
#env >> /tmp/udev.log
#echo "Connected" > /dev/pts/2

log=/var/log/emulator-controll.log

{
readarray -t PIDS <<< $(ps aux | grep [e]mulationstation | awk '{print $2}')
echo ">${PIDS[@]}<"
echo >> $log
/bin/date >> $log
echo "Invoking user: $(whoami)"
case "$1" in
        start )
                echo "Invoked start command" >> $log
                if [[ "${#PIDS[@]}" -ge 3 ]]; then
                        echo "Emulator is already running. No action is taken" >> $log
                else
                        echo "Invoking emulation station start command" >> $log
                        export DISPLAY=:0
                        sudo -u pi nohup emulationstation  &
                fi
                echo "Done" >> $log
                ;;
        stop )
                echo "Invoked stop command" >> $log
                if [[ -z "${PIDS[0]}" ]]; then
                        echo "Emulator is not running, no action is taken" >> $log
                else
                        echo "Killing processes with PIDs ${PIDS[@]}" >> $log
                        for pid in "${PIDS[@]}"; do
                                kill $pid
                        done
                        echo "Done" >> $log
                fi
                ;;
        *)
                echo "Usage: $0 {start|stop}"
esac

} >> $log 2>>$log

Answer 1

When the kernel triggers actions and udev runs scripts as part of the rules, the environment of those scripts is very restricted.

There can be many users logged in simultaneously into a single Linux computer, each user with multiple tty's, so of course a script run from the kernel doesn't have a tty: No user has started this script, so even if one could assign a tty to it, which of the many tty's in use should it be? And the same goes for an X connection (which is also frequently asked).

There are more restrictions, as you can see from man 7 udev:

RUN

Add a program to the list of programs to be executed for a specific device. This can only be used for very short running tasks. Running an event process for a long period of time may block all further events for this or a dependent device. Long running tasks need to be immediately detached from the event process itself. If the option RUN{fail_event_on_error} is specified, and the executed program returns non-zero, the event will be marked as failed for a possible later handling.

If I read your script correctly, you want to start or stop something on a given X connection that will be running for a very long time. This can only work if you split your task in two tasks: Have some kind of demon that is started when the corresponding X server starts (the X server needs to be started so you can get the authorization cookie), and then tell the demon in the udev run script that it should do certain things, like starting or stopping a service.

In this way, the udev run script can exit immediately.

If you are using systemd, you can also start a unit from an udev rule (see e.g. here).

Edit

If you want to use tty1, consider changing the login service (getty) that runs on tty1 to your demon, which you can then trigger from the script.

So e.g. here for details of a possible way to do it, vary as needed.

Answer 2

Instead of trying to attach script to TTY or assign tty to the script, I have managed to achieve my goal by injecting command directly to destination tty's input buffer with this:

sudo writevt /dev/tty1 "emulationstation $(echo -ne '\r')"

this types in command and hits "enter" on the main terminal.

PS. writevt is part of console-tools and that is available on Raspbian (at least today)