Re: [RFC/PATCH] printk: Fix preferred console selection with multiple matches

[Petr Mladek]

On Mon 2019-12-16 11:09:33, Benjamin Herrenschmidt wrote:
> On Wed, 2019-12-11 at 10:17 +0100, Petr Mladek wrote:
> > The reverse search of list of console does not work for ttySX
> > consoles because the number is omitted when matching. And the messages
> > will appear only on the first matched serial console. There is
> > a paragraph about this in the commit message of my patch.
> 
> About that specific issue...
> 
> I see indeed that 8250_core.c registers a "generic" console with index
> -1 which will match whetever we hit first in the array.
> 
> This is actually wrong isn't it ? Without any change such as what we've
> been proposing, it means that an arch doing add_preferred_console of
> any ttyS* will override anything on the command line, and it also means
> that a command line with multiple ttyS entries will stop at the first
> one, not the last one.
> 
> IE. In both case the code will select a console that isn't
> preferred_console... or am I missing something subtle ?
> 
> So yes, fixing that will "regress" in the sense that it will change the
> behaviour, but to make it match what's documented... am I wrong ?
> 
> The question then becomes what's the most broken ? Changing the
> behaviour that might have become expected or leaving the (alledgedly)
> broken behaviour in place ?

I though the same. But then I found the following in
Documentation/admin-guide/serial-console.rst

  "If no console device is specified, the first device found capable of
  acting as a system console will be used. At this time, the system
  first looks for a VGA card and then for a serial port. So if you don't
  have a VGA card in your system the first serial port will automatically
  become the console."


In addition, there is the "documentation" how systemd handles serial
console and login prompts, see
http://0pointer.de/blog/projects/serial-console.html


I agree that the current behavior is wrong. I really would like
to change it. But the result would be that some users will not
get login prompt after a kernel update. I am afraid that we are
blocked by the "do not break userspace" rule.

Best Regards,
Petr