Bug 7032

System: Dell PowerEdge 2300 server, dual CPUs, with a
serial console configured on /dev/ttyS0 at 9600 baud in
LILO and the kernel.

 After upgrading from RH 6.0 to 6.1, our system (running
the exact same kernel as before the upgrade, as we use a
custom kernel) developed an inability to consistently do
unattended reboots. If we left LILO boot the system via
its timeout, it would print:
	Loading <kernel version>...........
and then there would be a long string of ASCII NUL's dumped
out the serial port and nothing much would happen. (In some
experiments, the monitor would display the kernel hanging
right after the 'Console: colour VGA+ 80x25' line.) It
appears that if we explicitly hit return on the serial
console at the LILO promtp, the system would come up.

(It's hard to be precise on the diagnostics, because this
 is our production server and only reboots in situations
 where it needs to be up as soon as possible afterwards.)

 We reverted LILO back to lilo-0.21-6, the version from
RH 6.0, and in one test since then the problem has gone
away. (And the system was trouble-free until our upgrade
to 6.1.)