Re: Reproducible FreeBSD 6.0-BETA2 panic - probably ATA-ng related

On Sun, 2005-08-28 at 09:52 +0200, Fabian Keil wrote:
> Gavin Atkinson <gavin.atkinson_at_ury.york.ac.uk> wrote:
> 
> > On Sun, 21 Aug 2005, Fabian Keil wrote:
> > > I own a Plextor PlexWriter Premium, the drive has a buggy firmware
> > > which crashes if you try to burn multi session in SAO mode.
> > > On FreeBSD 6.0-BETA2 a panic is caused:
> > 
> > [snip the CD ROM drive detaching...]
>  
> > It's a known issue.  I see exactly this same panic 80% of the time on my 
> > laptop on resume from ACPI suspend.  I believe it was introduced during 
> > the newbus-ification of ATA-mk3.  On the call to acd_geom_detach, 
> > acd_softc is already null.
> > 
> > (kgdb) f 23
> > #23 0xc04dd936 in acd_geom_detach (arg=0xc16dd680, flag=0)
> >      at /usr/src/sys/dev/ata/atapi-cd.c:199
> > 199         g_wither_geom(cdp->gp, ENXIO);
> > (kgdb) list
> > 194     acd_geom_detach(void *arg, int flag)
> > 195     {
> > 196         struct acd_softc *cdp = device_get_ivars(arg);
> > 197
> > 198         /* signal geom so we dont get any further requests */
> > 199         g_wither_geom(cdp->gp, ENXIO);
> > 200
> > 201         /* fail requests on the queue and any thats "in flight" for this device */
> > 202         ata_fail_requests(arg);
> > 203
> > (kgdb) p cdp
> > $5 = (struct acd_softc *) 0x0
> 
> Thanks for the information. Is this the problem you described in
> <http://lists.freebsd.org/pipermail/cvs-src/2005-May/045876.html>
> and did you already find the causing commit?

Well, I thought so, but now I'm slightly confused.  I had been running
with the ATA-III changes for several weeks before they were committed
without problem, but updated a couple of weeks after they hit the tree
and had this issue.  In my mind I was sure the newbus changes happened a
week or so after the ATA-III commit, but looking at the CVS logs I don't
know any more.  However, I can't find any evidence at the moment that I
actually reported which commit was to blame, and I have no idea now.

I can probably take my laptop back to (say) April 15th and see if I have
any problems, however this is made significantly harder due to the
version library bump...

Gavin

Gavin