Serial console removal in production should be done with systemd masking about meta-balena HOT 8 CLOSED

This is something I wanted to do for a long time now. I will take care of it.

from meta-balena.

@telphan , @floion agreed to fix this.

from meta-balena.

@agherzan , @petrosagg , @telphan I believe we should better off remove the kernel cmdline "console" argument for production altogether instead of masking. Masking this serial getty service is not possible, as systemd instantiates this service based on the console for each particular device

from meta-balena.

In that case we need a unified way of manipulating command line arguments. As well something that would be able to switch to a debug mode.

from meta-balena.

@floion are you sure systemd starts getty on all console arguments? I don't find any documentation about this and don't seem to be able to replicate on device.

from meta-balena.

@agherzan I am pretty sure it does. You can check the current rpi-basic-image in meta-rpi, that misses the console=tty1 argument and no getty gets spawned before is is added to the kernel command line.

from meta-balena.

Here is the information that I was looking for @floion http://0pointer.net/blog/projects/serial-console.html . We need to sync on this @telphan @floion . I agreed that on production there should be no console kernel argument to avoid having getty spawned on those. But this is somehow a different issue than the one where systemd-serialgetty.bb enables the ones in SERIAL_CONSOLE which we currently handle my removing the files.

from meta-balena.

This was fixed in the meanwhile by removing the getty systemd generator.

from meta-balena.