Subject: Re: Multiple values for locators
To: John Kohl <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: tech-kern
Date: 03/25/1997 16:35:57
CH> So, does anyone have a reasonable objection to the above?

JTK> Not an objection, but a couple questions:
JTK> (a) what would you do with FOOCF_LOCNAME_DEFAULT?  set it to 0,0,0,0?
JTK>     Or define several of them (one for each position)?

CH> wdc0 at isa? port 0x1f0,0x3f6 irq 14

JTK>Presumably this should also "size 8,1" for the two port ranges?  [This
JTK>is a peeve of mine; most of the ISA drivers could/should have a size
JTK>locator, and the 0x666 crock in isa.c can go away.]

Not an objection but a question: how are the sizes supposed to be

There are drivers (e.g., the isa if_ed) in the tree that work on NICs
from several different vendors that happen to use the same chipset;
the ioport and memory size used by these drivers varies widly,
depending on the NIC actually found.

If the sizes are set in a config file for the `expected card at a
given IOport, that would break another vendor card at that address, if
the actual card needed a bigger IOport size than the locator allowed.

If the sizes are not-to-be-exceeded maxima, I can't see what the point
of adding them is: the drivers presumably know the max size anyway.
OK, I'm stupid.  Would someone mind explaining, please?
<question mode off>

<rant mode on>
The muliple-card support in a single driver is broken, IMHO, and has
been since we split the cfdriver and cfattach structures.  Again IMHO,
the right fix for the size (and other) problems this causes is to
split those drivers into separate front-ends for each card. That would
also kill the BROKEN_INDIRECT_CONFIG use of locators in that driver at
the same time.
<rant mode off>

I have yet to get Official Blessing to split the if_ed driver.
Please consider this a formal (and final) request...