🚀 go-pugleaf

In article <1992Nov2.154711.4667@dmc.com>, munroe@dmc.com (Dick Munroe) writes:
> GKDRIVER question:
> 
> SCSI is bigendian, VAXen are little endian, I assume that the GKDRIVER simply
> takes the commands and arguments given it and transmits them blindly (after
> all, it IS a pass through driver).  Which should mean that when I write SCSI
> commands, I have to worry about swapping bytes in portions of the commands that
> take words, etc.
> 
> Is this assumption correct?

It may or may not.  I never looked at it from that perspective.  As far as
I can remember, every SCSI device manual I've ever examined including the ANSI
specs, always clearly spelled out the contents of each BYTE of the command.
In general non byte quantities are relatively rare or and are used consistently
in each "group" of related commands that have a similar/identical parameter 
(i.e. read/write/seek all specify a LBN and do it exactly the same way for the
Group 0 commands, and do it exactly the same way for the Group 1 commands).

Take for example, a Group 0 WRITE command setup given LBA is the block number
to write and LEN is the number of blocks to write:

	cdb[0] = 0x0A;
	cdb[1] = (LBA>>16)&0x1f;
	cdb[2] = (LBA>>8)&0xff;
	cdb[3] = LBA&0xff;
	cdb[4] = LEN;
	
Although technically this implies big-endian interpretation, it certainly isn't
quite a byte swap operation since it involves parts of various bytes, AND not
an even number of bytes.

Note the above command sets 21 bits of LBA.  Given 512 byte blocks, this is
maximally 30 bits of addressability which is exactly the 1 GB limit the VMS
DKDRIVER had pre VMS V5.4.

--
Mark Pizzolato - INFO COMM Computer Consulting, Redwood City, Ca
PHONE:	(415)369-9366	UUCP:  decwrl!infopiz!mark or uunet!lupine!infopiz!mark
DOMAIN:	mark@infocomm.com