RE: [Xen-ia64-devel] RE: [PATCH] fixed serial console support

[Alex Williamson <alex.williamson@xxxxxx>]

On Sun, 2005-12-04 at 19:44 -0800, Magenheimer, Dan (HP Labs Fort
Collins) wrote:
> > 
> > Yep, forcing uart->irq to 0 in ns16550_init_postirq allows the
> > boot to complete.
> 
> and then recompiling xenlinux with 8250 console = n allows
> me to use ^A^A^A to get to the xen keyboard handler!  Cool!
> 
> Of course forcing irq to 0 in ns16550.c isn't going to be
> very popular with the Cambridge folks :-)  I'll try it on
> another machine Monday to see if the problem is unique to
> my test machine.

   Well, it can easily be done in the architecture specific code before
we call ns16550_init().  I don't think the irq setup code currently even
works for legacy interrupts, so it's not a terrible temporary hack.  I'm
still more than a little curious where irq 9 is coming from for you.
The PCDP code doesn't even attempt to set the irq.

   I looked a little bit at how 8250 can be left enabled and not
interfere w/ xen console.  I don't think it's as trivial as I was
hoping.  Even if we were to hide the specific UART xen is using, I think
there would still be a naming conflict for ttyS0 for any of the other
UARTs.  Does anybody know why xen tries to make a fake serial line for
dom0?  Seems like it would be easier if we did something like USB serial
devices and made up something like ttyXEN0.  I suppose it's probably for
compatibility, but unfortunately the xen console is setup after the real
serial ports are enumerated, and even if that is corrected, ttyS0 isn't
properly reserved and the serial layer blows up trying to overwrite it.
Thanks,

        Alex



_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel