[systemd-devel] [PATCH] console-getty.service: conflict with vconsole-setup

[Ramkumar Ramachandra]

Lennart Poettering wrote:
> /dev/tty0 always points to the foreground VC tty. Are you saying UML
> provides /dev/tty0 but not /dev/tty1, /dev/tty2, ...? I am pretty sure
> that UML should be fixed to provide neither or both, evertyhing else is
> broken.

Thanks for all the information.  With some help from Al, I finally
figured out what's happening:

0. um Linux does not do VT102 emulation.  This is a known limitation.

1. In its default configuration, it connects the guest's /dev/console
to the host's fd:0,fd:1.

2. In its default configuration (I just sent a patch changing this),
it connects the guest's /dev/tty[1-16] to a /dev/pts/N on the the host
(allocated dynamically).  Then, to get a prompt on the guest, the user
can do `screen /dev/pts/N` or `minicom /dev/pts/N` on the host.

3. It provides con#= and ssl#= kernel parameters to override how all
the console and serial devices on the guest should be connected to the
host.

So, I can provide a patch for getty-generator to detect um Linux and
switch to getty on /dev/console, but vconsole-setup will still fail;
what should we do about that?