On Wed, May 10, 2006 at 08:42:17PM -0700, Jeff Rizzo wrote:
> So, this has been bugging me for about six weeks, but I've only had time
> to do semi-exhaustive testing in the last few days.
> 
> _If_ I remember When It Started correctly, I hadn't done anything
> special other than reboot.  But, as of late, I can't use SCSI disks. 
> I've tried different boards (I have three different 53c895 boards, plus
> there's two ahc onboard), different disks, different cables, and an
> external enclosure.   SRM sees the disks fine, and can load the kernel
> still, but I get this when disks are probed:
> 
> scsibus4: waiting 2 seconds for devices to settle...
> probe(esiop2:0:0:0): request sense for a request sense ?
> probe(esiop2:0:0:0): request sense failed with error 22
> probe(esiop2:0:0:0): generic HBA error
> probe(esiop2:0:1:0): request sense for a request sense ?
> probe(esiop2:0:1:0): request sense failed with error 22
> probe(esiop2:0:1:0): generic HBA error
> probe(esiop2:0:2:0): request sense for a request sense ?
> probe(esiop2:0:2:0): request sense failed with error 22
> probe(esiop2:0:2:0): generic HBA error
> probe(esiop2:0:3:0): request sense for a request sense ?
> probe(esiop2:0:3:0): request sense failed with error 22
> probe(esiop2:0:3:0): generic HBA error
> WARNING: can't figure what device matches "SCSI 1 8 0 0 0 0 0"
> 
> ... the same with a 2.1_STABLE or a very -current kernel (netbooted).
> 
> (When I attach the disks to the ahc, I don't get the same errors, but
> the disks show up sans any info, and don't actually work)
> 
> I'm perfectly willing to believe this is hardware, but I don't
> understand why it follows the different scsi parts the way it does.  Is
> it possible I've done something goofy in SRM?  (I have a vague memory of
> playing with settings in an attempt to get it to autoboot around the
> time things stopped working, but I don't recall details)
> 
> Any suggestions?  It's a real shame not to be overheating my office with
> such a nice machine...

Most times when I saw this, it was a hardware issue, and more specifically
2 disks using the same ID, or a pin shortcut on a connector. With HD68
connectors, I've got this several times: pins are not perfectly aligned
and when plugin, one pin slip in the wrong hole, creating a shorcut.
The M connector can be recovered by gently unbending the pin,
but the F one may be dead. You can see it: if the thin plastic between
the 2 holes is brocken, it's likely that the 2 relevant contacts in
the connector are shortcut.

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