Jason Thorpe <thorpej@nas.nasa.gov> écrit :
> On Fri, 28 Nov 1997 07:28:28 +0100 (MET) 
>  Martin Husemann <martin@rumolt.teuto.de> wrote:
> 
>  > Why isn't this mask calculated at runtime? Or the mapping retried with another
>  > irq if one fails?
> 
> Unfortunately, I know of no way to do this.  When you get an IRQ that
> doesn't work, one of two things will happen:
> 
> 	(a) Your machine will hang when the first interrupt comes in.
> 
> 	(b) Your device will never get interrupts.
> 
> (a) is hard to recover from.  And (b) is hard to distinguish from e.g.
> "nothing is happening, no, really".
> 
> There are some BIOS calls that one can make to "solve" some of the
> problems people have mentioned here, i.e. getting a decent IRQ, getting
> a decent I/O base, etc.  Unfortunately, it doens't appear as if there
> is a protected-mode interface to these calls (like there is for APM),
> which makes using them problematic.

Anyway, how can I solve my problem?  The obvious way is not to
use IRQ 9 for the PCMCIA but I don't know if it is OK from the
NetBSD core team point of view.

I've been using NetBSD on my laptop for two and a half years now
with various PCMCIA tweaks and I'd really like to install NetBSD
1.3 without having anything more to do (yes, I'm lazy :-)  I may
not be the only one but there have been very few messages on this
thread and I don't know if it's because nobody has this problem
or because nobody cares.

A (maybe) better approach would be to implement something like
FreeBSD's -c boot loader option, which I found very convenient to
dynamically adapt the running kernel to the machine's
configuration.

-- 
Marc Baudoin   -=-   <Marc.Baudoin@solsoft.com>
Security On-Line Software (Solsoft) SA