[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#541831: installation-reports: Sunfire T2000 success (one minor issue)



On Monday 17 August 2009, Stephen Gran wrote:
> > It sounds like the "rsc console" is something different, which could
> > explain what you saw. Maybe we need special handling for this case.
>
> The rsc is much like an iLo or alom.  It's just another management
> interface that offers a 'console' interface to manage the OS.  I did
> the install using this interface, obtained by running 'break -c'.

OK. So the problem is that there is no direct link between the rsc and the 
(serial) console to be used for the installed system.

> > Questions:
> > - What were the contents of /etc/inittab after the installation? Were
> >   the entries for tty[1-6] commented out or not?
> [contents of /etc/inittab]

So regular consoles were enabled. That's expected. I mainly asked to 
confirm.

> #T0:23:respawn:/sbin/getty -L ttyS0 9600 vt100
> #T1:23:respawn:/sbin/getty -L ttyS1 9600 vt100
>
> I had to uncomment T0 to get a console with 'console -f' from the rsc.

So the rsc is somehow linked to ttyS0.

> ~ # console-type
> serial

This is at least one clue: D-I *does* detect it's a serial connection.

> ~ # readlink /proc/self/fd/0
> /dev/console

That's a logical value if it isn't /dev/ttyS0.

> > - If the first is not "serial" and/or the last is not /dev/ttyS0,
> > then how could the need to set up serial console be detected?
>
> That's a very good question.  Interestingly, I see this in dmesg:
> [   32.032662] console handover: boot [earlyprom0] -> real [tty0]
>
> So at least at that point, the kernel thinks it _is_ tty0.

That's fairly normal. In the past we've also seen on sparc:
[    0.000000] console [earlyprom0] enabled
[   53.918290] Console: colour dummy device 80x25
[   53.971361] console handover: boot [earlyprom0] -> real [tty0]
[   59.676537] Console: switching to mono PROM 80x34
[   64.238061] Console: ttyS0 (SAB82532)
[   64.308951] console [ttyS0] enabled

Is there anything after that first message in your case? The full boot log 
would be useful to have for this issue.

> Unfortunately, I don't know enough about whether there is a difference
> between what you attach to with 'console -f' in the rsc and what you
> attach to with 'break -c'.  I naively think that they should be the
> same thing, but it is possible that they are not.

/me is completely lost here.

> > - Does the installer work correctly and does a listener for serial
> > console get set up for the installed system if you boot D-I with
> > console=ttyS0?
>
> No:
> steal-ctty: No such file or directory
> steal-ctty: No such file or directory
> steal-ctty: No such file or directory
> Over and over.

That might be a bug in itself that's worth fixing. The source for 
steal-ctty is at [1]. It is called by [2], which does the actual 
detection. The loop is probably because init keeps failing and resets.

Can you find out what /var/run/console-device contains at that point?
I'd expect /dev/ttyS0, but it would be good to know for sure.
You can probably find out by booting with BOOT_DEBUG=3 and then in
/sbin/reopen-console add a 'set -x' plus near the end a 'sleep 30' so you 
have time to read the output.

> So, this doesn't look like a problem the installer can
> fix.  It's just odd that the OS doesn't see the serial console on the
> same device as the installer does:
>
> sgran@zee:~$ readlink /proc/self/fd/0
> /dev/ttyS0
>
> Well, I'm a bit mystified about how this could be fixed, so I don't
> mind if you want to close it as unresolvable or punt it over to the
> sparc porters to see if they can shed any light on it.

We have solved similar problems in that past. The only thing that's needed 
is *some* way of identifying this case. Possibly there's something 
in /proc from openprom? We could then do something similar as was done 
for powerpc (see the code after "# Set up virtualized..." in [3]).

The final option would be to add a dialog that's displayed if we are 
uncertain and that just asks what to use. Might be useful for other cases 
too.

Main question is if you're willing and able to do a bit more digging :-)

Cheers,
FJP

[1]http://svn.debian.org/wsvn/d-i/trunk/packages/rootskel/src/sbin/steal-ctty.c
[2]http://svn.debian.org/wsvn/d-i/trunk/packages/rootskel/src/sbin/reopen-console-linux
[3]http://svn.debian.org/wsvn/d-i/trunk/packages/finish-install/finish-install.d/90console



Reply to: