On Thu, Mar 31, 2005 at 05:05:12PM +0200, Hauke Fath wrote:
> Am 31.03.2005 um 15:46 Uhr +0200 schrieb Manuel Bouyer:
> >On Wed, Mar 30, 2005 at 10:04:56PM +0200, Hauke Fath wrote:
> >> At 8:48 Uhr +0200 30.3.2005, Manuel Bouyer wrote:
> >> >> There are two more machines, same controllers with a later firmware
> >> >> revision, boards same family, only 64Bit PCI. When I put the NetBSD 
> >> disks
> > > >> into one of those machines, I found that they wouldn't even boot
> >> >> multi-user before locking up. The boxes run debian with a
> >>
> >> [...]
> >>
> >> >It looks like an issue with the MPT driver, not scsipi or softdeps.
> >> >Maybe the newer firmware have a sightly different interface with the
> >> >driver ?
> >>
> >> Swapping the controller cards between the machines made no difference, so
> >> it looks like a 32bit vs. 64bit PCI issue (bandwidth?).
> >
> >I don't follow you, do you mean you have a box with 64bit PCI slot,
> >one with 32bit PCI slot, and only the 32bit one has issues with the mpt
> >driver ?
> 
> The two machines running netbsd-2 ATM are 32bit pci boards. They 
> sort-of-work under light load with the occasional panic during the 
> daily cron jobs. Moving the disks to the two newer machines (64bit 
> pci) makes NetBSD jam while booting to multi-user. Using the older 
> controllers (all are 64bit pci cards) makes no difference.
> 
> So both machine generations have the mpt problem, only it hits much 
> worse/earlier with the 64bit pci boards.

It's just a supposition, but it's possible that the firmware of the
adapter is running out of ressources. It then gives back a status message
which isn't handled properly by the driver. It's possible that this
status code appeared in newer version of the firmware, and the driver wasn't
updated for it.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--