On Fri, Oct 06, 2006 at 03:06:37PM +0200, Michael Schmitz wrote:
> > i interpreted this exactly that way. the message comes from the scsci
> > controller. the controller is recieving commanf code 54 which besides
> > being an optional command (not part of the standard list) is not even
> > appropriate -- - 54 is for a cd/dvd-r drive not a hd.
> >
> > my impression thence is that maybe the kernel is not even using the
> > correct driver, at all. since it does not follow the current standard of
> > using modules, there is not easy way to check, load, unload drivers
> > as there would be with a sane initrd. that is to me one of the goals
> > of etch, why there are two replacement candidates for mkinitrd and
> > a new format besides.
>
> I rather think the kernel does get the SCSI command queues mixed up or
> corrupted in some way. Maybe some other driver scribbles over the SCSI
> command space. Locking broken, or some DMA process running wild (the only
> thing I can imagine would be Sonic, or IIfx DMA).
Feel free to correct me if I'm wrong, but I don't think this was a IIfx
or a machine with Sonic ethernet. It's possible the 5380 core code has
suffered from bitrot. It's mostly just used by 68k. The only x86 uses
are some old cards that I really doubt get much testing on 2.6.
> > with command codes at controller level scrambled , anyway no
> > wonder that directories get corrupted.
>
> Yup. The question is, how does the corruption happen?
I'd first try to put together a list of models where the corruption
happens. Also, if it's specific to using an older hard drive. My IIfx
has a drive ripped out of a PowerMac, so it's much more modern. I'll try
to get it setup sometime soon. I had to pack away all my 68k hardware a
little while ago and don't have it setup yet.
Brad Boyer
[EMAIL PROTECTED]
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
