[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-smp
Subject:    Re: Big Fix for 2.2.1
From:       Linux Lists <lists () cyclades ! com>
Date:       1999-01-27 23:18:43
[Download RAW message or body]


On Wed, 27 Jan 1999, Andy Poling wrote:
> 
> On Wed, 27 Jan 1999, Linux Lists wrote:
> > Please don't do that. The Cyclades driver supports both ISA and PCI cards,
> > and it follows correctly the current definition, by ioremap'ing addresses
> > above 1MB and using readX/writeX for _all_ accesses to MMIO resources (in
> > ISA or PCI). Doing this change would mean having to test which card I'm
> > accessing _every time_ I do a readX/writeX, which is completelly
> > unacceptable (for obvious performance reasons).
> 
> Perhaps I'm naive, or missing something really obvious here, but...
> 
> Shouldn't you already know which bus-type of card you're accessing long
> before you get to the performance-critical code in any driver?

The issue here is that I thought Linus mentioned a change in the ioremap
function, which would require ISA cards to be accessed by
read_ISA_X/write_ISA_X and others by readX/writeX. Since my driver uses
readX/writeX to access _both_ (PCI and ISA), that would mean having to add
the following for every call to this functions:

if (ISA)
    read_ISA_X/write_ISA_X;
else
    readX/writeX;


Interestingly enough, Linus _didn't_ mention anything about changing the
ioremap function (and I have no idea where I got this idea from). Thus, my
e-mail is completely unfounded and can be ignored.

Sorry for the unnecessary noise.

Regards and Apologies,
Ivan

-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to majordomo@vger.rutgers.edu

[prev in list] [next in list] [prev in thread] [next in thread]