Re: CVS Update: 2.3 (branch: trunk)

[Roman Z. <zi...@li...>]

Hi,

On Mon, 5 Nov 2001, Geert Uytterhoeven wrote:

> >   Revision      Changes    Path
> >   1.4           +89 -55    2.3/drivers/ide/ide-probe.c
> >   1.4           +283 -171  2.3/drivers/ide/ide.c
>
> Since I was interested in seeing this fix, I diffed both files (against 1.3),
> but it looks like lots of other things were changed, too. Is that correct?

No, the number of changes are against the wrong file (1.3 is from
2.4.0-test2, also see the log), you want these diffs:

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/linux-apus/2.3/drivers/ide/ide.c.diff?r1=text&tr1=1.1.1.16&r2=text&tr2=1.4&diff_format=h
http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/linux-apus/2.3/drivers/ide/ide-probe.c.diff?r1=text&tr1=1.1.1.10&r2=text&tr2=1.4&diff_format=h

The problem is that request_irq is between disable_irq/enable_irq and the
ppc version of enable_irq doesn't like this, because request_irq resets
the counter.
The other point is that SA_INTERRUPT is IMO only needed by broken intel
hardware, so there should be rather a ifdef __i386__...
Something else: ide_get_lock/ide_release_lock is only for atari anyway,
so does it need to be conditional during probe? I already removed the
ifdef's but I wasn't sure about this.

bye, Roman