Re: [PATCH] fix issues with loading PCI ide drivers as modules (linux 2.6.0)

[Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>]

On Monday 05 of January 2004 03:09, Davin McCall wrote:
> Sure, the current code doesn't cause a crash - but it's very, very ugly. It
> generates some confusing error messages, and it makes it look like the
> module has taken control of the IDE interfaces but really the drives
> haven't been re-probed etc.
>
> Is this not worth fixing?

You are right.  Thanks for very good explanation.

> > Ehh, more hwif->chipset crap.
>
> Alright, this newer patch below mostly avoids the "hwif->chipset crap" (it
> doesn't introduce any new chipset types). But it has to export the
> "initializing" variable from ide.c (I changed its name to
> "ide_initializing").

You don't need to export "initializing" variable from ide.c,
just use "pre_init" variable from setup-pci.c :-).

> Plus, everything works as before - including "idex=..." parameters.

Except when using them for IDE PCI modules with non default ports:
- hwif->chipset is set to ide_generic during boot
- main IDE driver initialization
- module load fails (because hwif->chipset == ide_generic && !initializing)

You can fix it by replacing all current occurrences of ide_generic by some
new type (ide_forced).  It will also clear confusion about ide_generic name.

> @@ -1343,6 +1343,7 @@
>  			int unit;
>  			if (!hwif->present)
>  				continue;
> +			if (hwif->chipset == ide_unknown) hwif->chipset = ide_generic;

very minor nitpick:

if (hwif->chipset == ide_unknown)
	hwif->chipset = ide_generic;

Please correct patch and I will merge it.

cheers,
--bart