Eric W. Biederman wrote:
> Bari Ari <[EMAIL PROTECTED]> writes:
>
>
>>Eric W. Biederman wrote:
>>
>>
>>>Bari Ari <[EMAIL PROTECTED]> writes:
>>>
>>>
>>>
>>>
>>>>For embedded apps. USB is becoming a preferred method for booting since
>>>>serial ports are no longer included on the many of the latest chipsets.
>>>>
>>>>
>>>Which generally makes using a serial console pretty hard doesn't it.
>>>Or in other words what do people use for a console on a board like
>>>that. I admit I have heard interesting stories about ring buffers, in
>>>ram but that doesn't sound quite approriate either.
>>>
>>>My memory has it that most serial ports are on superio chips anyway so
>>>I don't see their lack of inclusion on a chipset as a problem.
>>>
>>
>>For many embedded apps cost and or board space is very much a concern
>>and to add the $5 for the superI/O just to get a serial console is out
>>of the question.
>>
>>Intel has been trying to push the serial ports into legacy for the past
>>few years, so serial ports have been disappearing from many boxes. A USB
>>console may be the simplest approach to development and debug on the up
>>coming machines.
>>
>
> I think it is microsoft rather than intel.
>
> But at any rate the specification for removing serial ports
> also calls for adding a debug port. Which amounts to a set of pins
> connected to the LPC bus where you can plug in a superio chip.
The majority of LinuxBIOS apps. seem to be for the highly religious
Beowulf community that only use OTS commodity motherboards. These
motherboards will probably have or at least have a slot for a superI/O
with a serial port, so serial console is fine. The all-in-one boxes out
that don't have any slots and only USB for expansion will probably
remain in the desktop world.
I may be the only nut here designing dense nodes that will only have the
bareness for cluster apps that are stripped down to just CPU, RAM, and
network controllers without any superI/O.
Bari
