Powering off via UI while using a Mausberry Circuit

cyperghost

@caver01 sorry edited this
maybe @meleu gots a solution ... i know only the emulationstation path

EDIT:
Yes... for restart system, shutdown system and restart ES there are temp files named by the usecase

and that's the way to prove ;) within the mausberry... you're right ;)
My first attempt does not work as the loop is only triggered if the MB-button is pressed. But as it checks some file you can use this via [[ -f /tmp/es-shutdown ]] to do some things ;) I guess to use this check into the power = 0, do loop and to set power = 1... this will end the loop and the shutdown is initiated.

But let @meleu look into... ;) He knows the system much better

cyperghost

@caver01 @meleu

I think the extended "mausberry" script with 3 GPIOs will look like this --> https://ghostbin.com/paste/5f8t6
But it may happen that ES deletes the tmp file BEFORE the mausberry script will catch it (During the 1s sleep) - That's nasty ;)

I think @meleu will introduce the inotify thing now as it acts as an "interrupt for filesystems" ;)

meleu

@caver01 said in Powering off via UI while using a Mausberry Circuit:

@meleu Yes, it will. but what can I do?

Try this solution here: https://retropie.org.uk/forum/topic/12895/ensuring-es-gracefully-finish-and-save-metadata-in-every-system-shutdown

Let me know if you find some problem. ;-)

caver01

@meleu This is cool, and I am going to try setting up the service as it greatly simplifies the Muasberry script (or any shutdown button technology for that matter). However, I don't believe it will satisfy the problem identified in the original post here.

The issue is that the Mausberry circuit does not notice that the shutdown has occurred unless it was triggered by the hardwired button. So, selecting QUIT, and SHUTDOWN in ES leaves the circuit still powered ON, and subsequently, the Pi is in low-power sleep mode. Worse, the MB circuit is now stuck thinking it is ON, and won't cut power to the Pi.

If you trigger the shutdown with the wired power button, the circuit starts paying attention to the GPIO IN pin that you have configured in the mausberry script. Once that drops, it cuts power. But, crucially, it seems to ignore that pin until it detects a short on its wired hardware button/switch.

So far, the only way to get around this problem for me has been to hook into the ES shutdown command and send a signal to another GPIO pin which in turn triggers a transistor wired in parallel with the mausberry switch. Instead of actually shutting down, this makes the mausberry circuit think I just pressed the physical button. It then triggers its GPIO OUT pin, which the mausberry script is watching, does the safe shutdown, and the circuit is now 'listening' for GPIO IN to drop --then it cuts power completely.

It's a clever circuit, and your new service will help simplify things for this and similar solutions, but the locked up circuit after software-triggered shutdown remains.

caver01

@meleu So, now I am thinking that I can simply add my echo command to your service which should give the circuit a 'tap on the shoulder' and get it to start watching for the Pi to do down. Is that your thinking too?

meleu

@caver01 yeah, put that command in the beggining of killes.sh. ;-)

caver01

@meleu that didn't work. So, I reverted emulationstation.sh back to default, then followed your instructions to create the service, setup the killes.sh in /home/pi/bin

As far as I know, it saves the metadata, but what it doesn't do is trigger the mausberry circuit.

I switched back to my original pyton script for the circuit. Then I added these lines at the beginning of killes.sh:

#this is the GPIOpin connected to the 1k-ohm resistor, leading to the base of NPN transistor (emitter and collector to power switch poles)
GPIOpin=18

# Export the GPIOpin and and set direction
echo "$GPIOpin" > /sys/class/gpio/export
echo "out" > /sys/class/gpio/gpio$GPIOpin/direction
# Trigger the transistor on GPIOpin to short the mausberry switch
echo "1" > /sys/class/gpio/gpio$GPIOpin/value

I need to export the pin used for the transistor because I am not using the BASH script, and nothing in the Python script uses the transistor. Even if I did define the transistor pin in the python script, I would still need to export it here (or somewhere) because python interacts with GPIO differently than BASH and does not need the pins exported to user space.

As a test, I can still drop to command prompt and enter the echo commands above manually and it will trigger the circuit and power down, so it should be possible.

I am not certain that shutdown is even invoking the killes.sh.

caver01

I though of another problem incorporating the Mausberry trigger—reboots will become shutdowns. That’s not the end of the world, but worth mentioning.

caver01

@meleu Ok, so I think I know why your solution did not work for my case. I hadn't flagged my killes.sh as executable. I am hoping that resolves the problem (I was inadvertently NOT watching the other thread when you added that step.) I am not at home to try it, but I did the CHMOD remotely and I am anxious to test it. If my addition to the killes.sh triggers my mausberry circuit to turn off, we have a great solution. This would simplify mine and many other 3rd party shutdown needs as handling the metadata issues safely. It's a smart idea.

One bit of fallout I mentioned above with my change is that every restart would trigger my mausberry to cut power. I would lose the reboot ability, but this is a worth sacrifice for simplifying everything else in my opinion.

Also, I plan to drop the use of the mausberry script completely--I am shifting to my GPIO inputs solution--GPIOnext--to handle the shutdown trigger. I will connect the mauseberry's input wire to a 3.3v GPIO rail.

meleu

@caver01 said in Powering off via UI while using a Mausberry Circuit:

every restart would trigger my mausberry to cut power. I would lose the reboot ability

I can't see why my systemd trick would cause that (edit: oh, I've just remembered that you added some lines to killes.sh). Can you please make a test to check?

Do all the steps I showed on that topic and try to reboot your system. Describe here what happens.

And then perform this command: sudo systemctl disable killes; sudo systemctl stop killes and reboot your system. Describe here what happens.

I'm anxious for your feedback. :-)

caver01

@meleu Yes, that's the plan. It's going to be several hours before I can test. You remembered my addition. . . I am adding a line that is going to twiddle GPIO 18 to my transistor, which should tell the MB circuit it is time to start watching its GPIO connection for power-down (of course it will try to initiate a shutdown too at that point).

meleu

@caver01 Oh... Maybe adding those lines to the killes.sh isn't the perfect solution. Let's solve the saving metadata issue first and later we can try to solve your shutdown/reboot issue. ;-)

caver01

@meleu Agreed. However, is there a way, during a soft reboot to detect whether the shutting down is a result of a reboot request as opposed to a shutdown/poweroff? If a reboot command leaves some trace that we could try to detect with a conditional, I could hold back that command if found.

meleu

@caver01 I know that there are "systemd ways" to launch a script for reboot or poweroff exclusively. I'm not a systemd expert but I know it can be done.

caver01

@meleu LOL and here I thought you did all of that exciting sysetmd research!

meleu

@caver01 I just researched enough to find that solution. And then went to do something more interesting (usually play Mega Man :-) ).

caver01

@meleu said in Powering off via UI while using a Mausberry Circuit:

Do all the steps I showed on that topic and try to reboot your system. Describe here what happens.

And then perform this command: sudo systemctl disable killes; sudo systemctl stop killes and reboot your system. Describe here what happens.

SO, for starters, I have already done what you prescribed in the other thread to build the service and the killes.sh. Mine is till in the ~/bin folder, which is fine by me for now. After setting the killes.sh as executable, it works great! That is to say, I can do a software shutdown from command line for instance, and it definitely saves my metadata. Favorites and custom collections are awesome. Moreover, my additional line which triggers the NPN transistor does indeed let the Mausberry circuit know that a shutdown is imminent and it properly notices the drop on the GPIO it is monitoring and it cuts power. I really like this clever solution.

It also cuts power when I initiate a reboot. That makes sense to me, and for now I am willing to live with it. The point is that software shutdown requests properly exit ES and no longer lock up my Mausberry circuit.

Seems to me, the improvement to this would be to set a conditional statement before my GPIO command to trigger the transistor such that it checks to see if this is a shutdown/poweroff or if this is a reboot. I just don't know what to check to verify one way or the other--yet.

As for your command above, I initiated the command as you suggested and the system immediately did a shutdown, mausberry circuit powered off.