On Fri, 11 Oct 2002 bob@cemetary.homeunix.org wrote:

# I have an SS2 happily running 1.5.4_ALPHA

{I would guess that's the "cemetary" (sic).}

# It has a HP C3725S 2 gig scsi disk, which is know to have issues with
# tagged command queueing (ie. it says it can but it lies). OpenBSD 3.0
# caused major filesystem corruption when installed on this disk but
# NetBSD 1.5 has been rock solid for months.
#
# probe(esp0:1:0): max sync rate 3.84MB/s
# sd0 at scsibus0 target 1 lun 0: <HP, C3725S, 6039> SCSI2 0/direct fixed
# sd0: 2047 MB, 3703 cyl, 9 head, 125 sec, 512 bytes/sect x 4194058
# sectors
#
# Now I am thinking of upgrading to 1.6 - are there any gotchas wrt
# sun4c/esp/tagged command queueing that are likely to cause me headaches
# should I upgrade?

I'd say go in and set the appropriate flag in 0xff0000 to disable
tagged queueing in your kernel (someone please correct me if those
are not the right flag bits for this).

I have a further question (two, actually):

Have we already broached the subject of having the driver attempt tagged
queueing commands and automatically disabling TQ if they fail?

How about completely disabling TQ in the GENERIC kernel, with an easy-
to-find comment pointing out where TQ is disabled (and how to re-enable
it on a subsequent build)?  It might seem a PITA to those with smart
disks, but on those with not-so-smart or "lying" disks, bootstrapping
with a GENERIC kernel becomes next to impossible short of putting in a
TQ-capable disk.

Merely a thought; apologies if this has already been covered.
[I don't see comments in the GENERIC config which say ANYTHING
about tagged queueing.  I think this needs to be fixed.]

# --Andrew

				--*greywolf;
--
Microsoft:	"Where do you want to go today?"
Linux:		"Where do you want to be tomorrow?"
NetBSD:		"Are you guys coming, or what?"