Current ia64 boxes (and potentially ia32 boxes using EFI and HCDP)
have enough information in firmware to make a reasonably educated
guess on where the console lives.  This patch tries to make use of
it and "do the right thing" when it can, without getting in the way
when it shouldn't.

   The code is triggered off whether there's a console port found
in the HCDP table.  At that point, it tries to determine what the
current command line enables for consoles.  The EFI ConOut variable
is also used to try to guess whether the system has VGA.  In short,
if the system only has one device specified for ConOut and you do
not pass a "console=tty..." argument, it will just work.  There's
some other nice parts of it that try to print a courtesy message to
the port that's not being used for the console to indicate where the
console has been started.  These only work for MMIO serial ports in
the HCDP and boxes where the EFI memory map seems to indicate a VGA
memory hole.  Doing the same for I/O port space UARTs should be trivial
to add, but I don't have the hardware/firmware combination to test it.
FWIW, I believe early ia64 boxes set the EFI memory type wrong for
the VGA hole.  I don't try to figure that out so you won't get the
courtesy messages on VGA for them.

   Let me know what you think and if you find any unexpected behavior.
It should not adversely affect any currently working setups.  The
current patch is against 2.6.0-test4.  Portions of this code are
leveraged from a previous implementation by Khalid Aziz.  Thanks,

	Alex

-- 
Alex Williamson                             HP Linux & Open Source Lab