Justin T. Gibbs wrote:
>
> The 2742AT only has 4 SCBs too, but the command FIFOs are different
> which means you can use SCB paging to get more commands active at once
> (you can't on the prior to Rev E aic7770s that are on the non 'A' cards).

My machine reports: (under NetBSD/i386 -current)
ahc0 at eisa0 slot 11: Adaptec AIC-7770 SCSI (on motherboard)
ahc0: Using Level Sensitive Interrupts
ahc0: aic7770 >= Rev E, Twin Channel, A SCSI Id=7, B SCSI Id=7, 4 SCBs

> There is still some latent bug in the SCB paging code though.  It's 
> very hard to reproduce unless you have many drives (greater than 8)
> since it seems to only happen when you have 30 or more commands pending.
> I'm still hunting for that bug.

If this is not related to the "sleep" issue, I can trigger the SCB
queue error on demand.  Well, ok.. If I hook up my external subsystem
with these drives:

sd2 at scsibus3 targ 0 lun 0: <SEAGATE, ST41650, 7628> SCSI2 0/direct fixed
sd2: 1350MB, 2107 cyl, 15 head, 87 sec, 512 bytes/sec
sd3 at scsibus3 targ 1 lun 0: <MICROP, 1528-15MD1066702, DD24> SCSI1 0/direct fixed
sd3: 1280MB, 2100 cyl, 15 head, 83 sec, 512 bytes/sec
sd4 at scsibus3 targ 2 lun 0: <MICROP, 1528-15MD1066702, DD24> SCSI1 0/direct fixed
sd4: 1280MB, 2100 cyl, 15 head, 83 sec, 512 bytes/sec
sd5 at scsibus3 targ 3 lun 0: <MICROP, 1528-15MD1066702, DD24> SCSI1 0/direct fixed
sd5: 1280MB, 2100 cyl, 15 head, 83 sec, 512 bytes/sec

on the "B" channel, it will blow up fairly consistently.
(but only if I'm mv/cp/whatever'ing between disks)

-Andrew