Reproducing "Bluetooth On" Behaviour

[lumping-sugar66]

Hello folks,

I'm trying to turn on Bluetooth programmatically without needing Nickel for alternative readers on my Kobo Clara 2E. I wrote the patch to turn Bluetooth on and off using NickelMenu, and it works perfectly fine, but when I try to run the services that get triggered when Nickel is used, I get unreliable behavior.

I need some help in making sense of all the programs that are running.

Here's what I've found out:

ps dump changes after turning Bluetooth on via Nickel:

[root@kobo ~]# ps
PID   USER     TIME  COMMAND
......
 5173 root      0:00 /bin/bluealsa -S -i hci0
 5170 root      0:00 /sbin/getty -L ttymxc0 (or ttymxc1) 115200 vt100
...... # this is where I turn on bluetooth
 3431 root      0:00 /sbin/hciattach -n ttymxc1 any 1500000 flow -t 20
 3434 root      0:00 [hci0]
 3435 root      0:00 [hci0]
 3440 root      0:00 /libexec/bluetooth/bluetoothd

Running hciattach gets the hci0 links so that bluetoothd can hook the Bluetooth agent up. And then I run headless bluetoothd.

Here's my script:

# cat on.sh 
#!/bin/bash
cd "$(dirname "$0")"

# Check if bluetoothd is running
if ! pgrep bluetoothd > /dev/null
then
    echo "bluetoothd is not running."

    echo "Killing getty and bluealsa" # An experiment
    pkill bluealsa &
    pkill getty &
    sleep 1

    echo "bluealsastuff" # Content: # /bin/bluealsa -S -i hci0 &
                                                  # pkill bluetoothd

    sh bluealsastuff.sh  > /dev/null 2>&1 &

    echo "Adding uhid.ko.." # load the kernel patch for the device recognition
    sh ./adduhid.sh

    echo "hci services.."
    sh ./runhci.sh # nohup /sbin/hciattach -n ttymxc1 any 1500000 flow -t 20 &

    echo "Starting bluetoothd.."
    /libexec/bluetooth/bluetoothd  > /dev/null 2>&1 &

    echo "Sleep 1"
    sleep 1
    bluetoothctl power on

    echo "HCI0 up..."
    hciconfig hci0 up

    echo "Success.."
else
    echo "bluetoothd is already running. No action taken."
fi

exit 0

This script is unreliable in turning on Bluetooth correctly. Maybe 40% of the time it works. The biggest problem I'm facing is when hci0 interface doesn't get up:

# hciconfig hci0
hci0:	Type: Primary  Bus: UART
	BD Address: A4:3C:D7:10:62:44  ACL MTU: 1016:5  SCO MTU: 60:12
	DOWN 
	RX bytes:1711 acl:29 sco:0 events:95 errors:0
	TX bytes:1597 acl:29 sco:0 commands:56 errors:0

This is probably because I'm unable to run hciattach process in a healthy way so, the hci0 interfaces aren't running healthily because even I do

hciconfig hci0 up

I get either device busy or timeout.

If I run Bluetooth once from Nickel and then go run the script, I see that my chances are better. It's also a must to have WiFi running for Bluetooth to be turned on for Nickel btpt to run. After doing this once, it's not necessary for the script to have wifi on.

Any help in making this process a bit more reliable is appreciated.

For what it's worth, this is how I turn off Bluetooth:

cat off.sh 
#!/bin/bash
cd "$(dirname "$0")"

# Check if bluetoothd is running
if pgrep bluetoothd > /dev/null
then
    echo "bluetoothd is running."

    echo "Turning off Bluetooth and killing bluetoothd."
#    bluetoothctl power off
    pkill bluetoothd

    echo "Remove uhid.ko.."
    sh ./removeuhid.sh

    echo "Killing hciattach.."
    pkill hciattach

    echo "Success.."
else
    echo "bluetoothd is not running. No action taken."
fi

The biggest reason I have to kill all these processes is to prevent excessive battery drain. I am getting close to 3% per hour drain if bluetoothd is running, regardless of whether I use bluetoothctl to power off the device or not.

The killing of bluealsa and getty (I have no idea what they do) was an experiment as they seem to be using the ttymxc1 sometimes and I thought maybe isn't making the device busy or something. I haven't been able to verify this 100%.

Thank you