I noticed that in a recent version of the GENERIC kernel config, the bt0
device specifies a drq as well as an irq.  I've been using my BT445S with
the bt742a driver for months now with no problems without specifying a drq.

Now, is it necessary to specify a drq to use the BT742/445 in busmastering mode?
I realize that these boards don't use the ISA DMA controller (does the 742 use
the EISA DMA controller?) to do their busmastering, but I just read an 
interesting article on one of the comp.sys.386bsd.* newsgroups
written by a guy who has written device drivers for SCSI adapters on PCs.  In
that message he explained that busmastering boards use the DRQ request channel
to grab the bus, then do their own xfer (while the DMA controller sits idly
by, thinking that it's waiting for instructions), then release the DRQ.
What  I'm not clear on is how this relates to EISA/VLB boards (EISA has its
own DMA controller, which I don't think the BT742 uses anyways, but VLB 
definitely uses the ISA DMA controller and doesn't add any functionality to 
the standard ISA DMA capability).

I also seem to remember someone mentioning that the BT742 will report
"eisa dma" during boot-up if it's been initialized to use "enhanced mode"
busmastering.  I don't see this, of course, since I have a VLB/ISA system,
but does that mean that I'm not really using my BT445S in busmastering mode
under NetBSD?

I specified drq 5 for the bt0 driver in my current kernel, just for kicks,
and though no DRQ is reported at boot-time by the bt0 driver, I'd *swear*
that my system performance has improved during big disk operations (moving
windows around in opaque-move mode under fvwm/XFree2.0/XF86_Mach32 is a lot
less jerky now while compiling the system).  Haven't taken a look at overall
disk performance yet....

Anyways, maybe someone with enough knowledge of the SCSI subsystem can answer
this question for me (does Julian read this list?).

tnx
-dwh-
dhess@cs.stanford.edu


------------------------------------------------------------------------------