Re: [Xen-devel] Regression in serial console on ia64

[Markus Armbruster <armbru@xxxxxxxxxx>]

Keir Fraser <Keir.Fraser@xxxxxxxxxxxx> writes:

> On 11/9/07 12:12, "Keir Fraser" <Keir.Fraser@xxxxxxxxxxxx> wrote:
>
>> On 11/9/07 12:06, "Horms" <horms@xxxxxxxxxxxx> wrote:
>> 
>>> I havnen't investigated why, but 52f308b17bae
>>> "xencons: Remove xencons_early_setup() and do not clear use_vt_console."
>>> seems to cause a regression on ia64.
>>> 
>>> The problem is that with this change applied the console
>>> seems to stop just before the end of boot - no more output,
>>> no input.
>> 
>> Is this when booting dom0? And its serial output (via Xen) that is affected?
>> Do you have console=xvc0 on your dom0 cmdline (and no other console= lines)?
>
> My guess is that you have console=tty0 on your cmdline and this is taking
> precedence over your serial console for init output. I've seen this kind of
> issue before on native when trying to have both console=tty0 and
> console=ttyS0 on the cmdline -- init output always seems to go to only one
> or the other. I think it's supposed to be the case that the *last* console=
> on your cmdline is the one that gets /dev/console and thus gets to swallow
> init's output. I'm not sure what the fix is (if any) if you really want your
> init output to go to multiple console sinks.

Kernel messages (printk) go to all registered consoles with flag
CON_ENABLED.

User space talks to /dev/console.  The kernel routes that to the
preferred console, which is the last one on the command line, if any,
or else the first one to register (typically the virtual terminal).

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel