Subject: Re: How about to config Hayes ESP with glue device rather than magical option?
To: Jason Thorpe <thorpej@nas.nasa.gov>
From: Chris G. Demetriou <cgd@pa.dec.com>
List: current-users
Date: 09/08/1997 09:35:22
> On 08 Sep 1997 08:08:02 +0900 
>  enami tsugutomo <enami@but-b.or.jp> wrote:
> 
>  > isa0 at pcib0
>  > hayesesp0 at isa0 port 0x280-0x287 irq 5
>  > com3 at hayesesp0 port 0x3e8-0x3f0: 1024 byte fifo
> 
> Actually.... I don't understand why we have "hayesesp0 at 0x280", and
> then "come3 at hayesesp0 at 0x3e8".  Why the 0x3e8?  What does that
> _mean_?

I don't know what it means, but it's wrong.  8-)  you should be able
to specify multiple I/O space (and other) addresses for a single ISA
device.  "hayesesp" doesn't provide an ISA bus, so the hierarchy above
is ... not really sensible.

While i'm at it, it really should be just:

hayesesp0 at isa0 port ....

I.e. no 'com' at all.  ("All this and more could be yours, if the
16x50 driver were ever made into a 'real' front-end/back-end -split
driver!")


>  > (1) `com at isa' disables FIFO when system shutdown, but `com at
>  > multi' doesn't.  For Hayes ESP, it did since it was same as `com at
>  > isa', but is it actually needed?  In my case, it works without it but
>  > since the comment says that it depends on BIOS, I can't say it is
>  > unnecessary.  Then, why it doesn't need for multi-port card?  Does it
>  > need for on board one?
> 
> ...this might be a work-around for buggy super-i/o chips?  If it doesn't
> hurt in the general case, why not just always disable the FIFO?

I seem to recall that this was so that losing BIOSes would recognized
the ports on reboot (!!!).



cgd