Bug#385179: debian-installer: [installation report : powerpc] IBM pseries p505 single CPU, no virtualization

[stappers@stappers.nl (Geert Stappers)]

On Tue, Aug 29, 2006 at 11:49:39PM +0200, Sven Luther wrote:
> On Tue, Aug 29, 2006 at 11:15:49PM +0200, Geert Stappers wrote:
> 
> > How knows the kernel which one to use?
> 
> the kernel will output to all of those existing, while the interactive one
> being the last one. Since i have not yet seen a case where hvsi0, hvc0 and
> ttyS0 are present at the same time, this works just fine.

Mmm. I would reduce the four consoles to the one that is actually used.

> Obviously, the p505 lacking a graphical output, tty0 is also not existent.
> 
> I can't say this is the most smart thing to do, the kernel guys must have been
> smoking something serious when they decided to name the serial devices in a
> thousand different ways.
> 
> I had to resort to that since the p505 firmware is not able to add bootargs on
> the commandline when using a zimage, and the above is what the 2.6.17
> powerpc64 kernels has as default commandline.

Mmm, that makes the reduction of consoles some what harder.


    <snip/> 
> > 
> > I wonder why serial output dies there, as I see no reason "to switch console"
> 
> Indeed. Do you have an idea of what comes next ? And who is doing the above
> output ? I thought it was coldplug, but since enabling set -x and various
> debug doesn't help, these outputs probably come from the kernel messages
> directly.
> 
> I guess the console is directed to somewhere random, and thus not shown. What
> is strange is that when doing the same on a JS21 blade in a blade center, it
> works just fine, so something funny is going on here.


I guess a console buffer is full. It waits for a handshake to change
level. ( e.g. 'CTS, Clear To Send' is seen as not active )
On the JS21 is CTS as being active, which allows to continue the output
to _all_ console, even the non existing consoles.


Cheers
Geert Stappers