> The baud-rate multipliers are per-port settable using _jumpers_, nothing
> can be changed in software.  It would be difficult to probe the jumper
> settings in the driver (I guess it would be possible to transmit a character
> and see how log it takes to do it, but this would be annoying because then
> the driver would transmit "garbage" data at boot time).

Right.  That's why I said the multiplier would make sense to set via
driver flags, i.e. a 'flags' declaration for the driver in the config
file.

Sounds like it makes sense to do it per-port.



> > So, regardless of whether the card has 4 or 8 uarts, it always eats 8
> > UARTs worth (64 bytes) of I/O space?
> 
> I'm not sure.  I haven't got any technical docs for the card.  The status
> register may be address decoded separately, so it might be possible to use
> the addresses between the last UART and the status register.  Dunno.
> 
> > Does it give you a good way of telling how many UARTs it actually has?
> 
> AFAIK, no.  The board has 4 or 8 UARTs and the status register (which
> just contains a bit mask of UART interrupt outputs), that seems to be
> all.

That seems pretty consistent with what the other multi-port boards do.


> > Really, you should config_found() only the UARTs that you know are
> > there (or, at least, the UARTs that are "supposed to be" there, but
> > there may be no easy way to do the right thing (other than a driver
> > flag, and if it always eats 64 bytes of I/O space that's probably a
> > waste of time).
> 
> I don't think the I/O space costs that much - isn't it the interrupts
> that ppl always run out of, anyway?

Well, sure, but that wasn't what I was getting at.  Note that the
following is very much a "how things should be done" comment, and
isn't necessarily reflected by all existing drivers.  8-)

When something calls config_found(), that's supposed to mean "I know
this device is here, really I do."  For some multiport boards, where
there are a known number of chips and changing that requires a chip
puller or soldering iron, config_found() is the right thing to do.

In a case where you _know_ you have a variable number of ports, you
probably shouldn't be config_found()ing things that you're not sure
that you have.  One way to be "sure" that you have something, for
instance, is to write a cheap, sleazy probe routine which just makes
sure that something is in that space, and assumes that if there's
something there that it belongs to the board.  (not an unreasonable
assumption.)




cgd