Re: IDE problem source ?

[<cah...@po...>]

On Tue, 24 Sep 2002, Roman Zippel wrote:
> On Tue, 24 Sep 2002, Geert Uytterhoeven wrote:
> 
> > I took a closer look at ide-probe.c and compared it to plain 2.4.18 from
> > Marcelo.  Apparently the enable_irq() (and some other code) is present in the
> > APUS version only! Although there are m68k comments in the changes, these
> > changes are not present in the Linux/m68k tree.
> >
> > Anyone who knows why APUS changed ide-probe.c?
> 
> I changed it while adapting the APUS irq code to the generic ppc irq code,
> but I don't really remember what caused that change.
> Anyway, that code works with an A4000, so there must be another problem,
> that just never triggered before, my guess is a timing problem.
> 
> > Perhaps you can try with the original ide-probe.c from Marcelo's tree?
> 
> It's better to find the real problem, that code is bascially already in
> 2.5. Two things which could be tried:
> 1. Add this call to the acknowledge routine before enabling interrupts:
> 	hwif->hw.ack_intr(hwif);
> 2. Move reenabling of interrupts after the request_irq() call.
> 
> The first solution hopefullly works, since the second one would be an ugly
> hack.

I don't know how, but my kernel's started booting. I modified kernel as
you said:

        __restore_flags(flags); /* local CPU only */
        if (hwif->hw.ack_intr && hwif->irq)
        {
                hwif->hw.ack_intr(hwif);
                enable_irq(hwif->irq);
        }
        for (unit = 0; unit < MAX_DRIVES; ++unit) {
                ide_drive_t *drive = &hwif->drives[unit];
                if (drive->present) {
                        ide_tuneproc_t *tuneproc = HWIF(drive)->tuneproc;
                        if (tuneproc != NULL && drive->autotune == 1)
                                tuneproc(drive, 255);
                }
        }

Regards
Krystian Baclawski