Re: [atomic-devel] EC2 PV/HVM AMIs

[Dusty Mabe <dustymabe gmail com>]

On Mon, Jul 21, 2014 at 04:25:32PM -0400, Matthew Miller wrote:
> On Mon, Jul 21, 2014 at 02:33:55PM -0400, Andy Grimm wrote:
> > Fedora isn't going to stay this way for long, though  They are
> > intending to register HVM images.  Actually, I do not believe that
> > hvc0 is the correct console device for HVM images (I will double-check
> > this), and there is no issue specifying more than one console device.
> > The most recent RHEL 6 HVM image I registered specified "console=tty0
> > console=ttyS0 console=hvc0" just to cover all the bases.  I don't know
> > which of them was right, but I got console output.  :)
> 
> There is, or was -- I haven't rechecked -- a bug where specifying
> non-existent console devices caused systemd to spawn a getty for them
> anyway, and possibly out of control. 
> 
> Also, when multiple consoles are listed, output goes to them all, but input
> comes from the last one. And, only last one is "/dev/console", which means
> that programs which open and write to /dev/console (cloud-init, I'm lookin'
> at you) are sensitive to the ordering here.

Yes. That is my understanding as well. I beleive that it is the
designed behavior (see [1]).

> 
> All of this might eventually mean that we need to eventually suck it up and
> generate tailored images for each environment -- we'd been wanting to avoid
> that, because it saves QA for us _and_ for users.

Agreed. A unified image is extremely desireable. 

> 
> In any case, consider this a big, waving "help wanted" flag. :)

I have some capabilities in this area. I don't mind helping get this
project off the ground. 

- Dusty 

[1] https://www.kernel.org/doc/Documentation/serial-console.txt