Re: Ata & atapi bug

[vanoven <vanoven%free.fr@localhost>]

On 10/07/2012 08:44, David Laight wrote:
> On Tue, Jul 10, 2012 at 01:47:43AM +0200, vanoven wrote:
> > Hi,
> > I got trouble booting with the last kernel on my old nc8000 laptop : my
> > ATA drive was not seen anymore while cdrom drive was.
> >
> > after looking in source i found a problem in
> > /cvsroot/src/sys/dev/scsipi/atapi_wdc.c
> >
> > with regard to this  diff
> >
> > @@ -321,12 +321,12 @@ wdc_atapi_probe_device(struct atapibus_s
> >                          ata_probe_caps(drvp);
> >                  else {
> >                          s = splbio();
> > -                       drvp->drive_flags &= ~DRIVE_ATAPI;
> > +                       drvp->drive_type = DRIVET_NONE;
> >                          splx(s);
> >                  }
> >          } else {
> >                  s = splbio();
> > -               drvp->drive_flags &= ~DRIVE_ATAPI;
> > +               drvp->drive_type = DRIVET_NONE;
> >                  splx(s);
> >          }
> >   }
> >
> > my harddrive is changed from DRIVET_ATA to DRIVET_NONE in the new version.
> > I add
> >      if ( drvp->drive_type == DRIVET_ATAPI )
> >              drvp->drive_type = DRIVET_NONE;
> > to fix that
> >
> > Hope it can help.
> Those two parts of the diff look like error paths - so shouldn't be relevant.
> Exactly where did you add the code to set DRIVET_NONE?
> Your CDROM will probably need DRIVET_ATAPI.
>
> I applied a small change to the driver yesterday (patch from Manuel)
> which fixes issues on a lot of other systems.
> Does that one fix yours?
>
>         David


I replace

drvp->drive_type = DRIVET_NONE;

in line 324 and 329 of atapi_wdc.c by

if ( drvp->drive_type == DRIVET_ATAPI )
            drvp->drive_type = DRIVET_NONE;


my ata drive and atapi cdrom go through wdc_atapi_probe_device() function
so my ide hard drive is changed to DRIVET_NONE by that error paths.


atapi devices are probed first in sys/dev/ata/ata.c

in function atabusconfig_thread(void *arg)

line 308: (*atac->atac_atapibus_attach)(atabus_sc);
 after that line my hdd get DRIVET_NONE

from
ch_drive : 0x0 , type = 0x1   (DRIVET_ATA)
ch_drive : 0x1 , type = 0x2   (DRIVET_ATAPI)
to
ch_drive : 0x0 , type = 0x0   (DRIVET_NONE)
ch_drive : 0x1 , type = 0x2   (DRIVET_ATAPI)


I guess that all devices in channel are probed by code in atapi_wdc.c ( 
config_found_ia() line 140 )

I tried the patch from Manuel to but it didn't work, i only have old ide drives 
in that laptop so i guess ahci stuff is not used here.