[Prev][Next][Index][Thread]

Summary: CAM SCSI and compressing tape drives



A week ago I asked after the magic numbers needed to make all
the densities work on an Exabyte 8505 under Ultrix.  Several people
sent me extracts from their data/cam_data.c files and other wisdom.
Here's a summary of what I learned by merging all the data together.

To add support for the 8505, one must
a) add an entry to cam_devdesc_tab in data/cam_data.c, containing the
proper device ID string and a pointer to a suitable density table.
It suffices to copy the existing TKZ09 entry, changing just those
two items.
b) invent a new density table with the desired mapping between
tape densities and minor device numbers.  The density codes for
the 8505 are:

0x0	default; 8500c on the 8505
0x14	8200 (low density, no compression)
0x90	8200c (low density, compressed)
0x15	8500 (high density, no compression)
0x8c	8500c (high density, compressed)
0x7f	`no change'

0x14 and 0x15 are the same numbers as SEQ_54000_BPI and SEQ_45434_BPI
(constants used in DEC's tables); hence one can just use the TKZ09
density table to get at 8200 (rmt0l and rmt0a) and 8500 (rmt0h and
rmt0m) densities.

Notice that (contrary to my initial worry and guess) the compressed
modes are just treated as densities; there's no need to put anything
new in den_compress_code, and you shouldn't set DENS_COMPRESS_VALID.

Some places have set up 8505-specific mode select tables (though some
contained the same bytes as tkz09_mod).  I kept tkz09_mod; it works
fine so far.

No two responses used the same mapping between minor device numbers and
hardware densities, so I felt free to use my own.  It is also worth
pointing out that the tape drive senses the density (including
compression) automatically when reading, regardless of the tape device
used.

Thanks to Tom Szilagyi (tom@xxxxxxxxxxxxxxx), Jeffery G. Smith
(smithj@xxxxxxxxxxxxxxxx), Andrea Zuccollo (andrea@xxxxxxxxxxxx),
and Simon Burge (simonb@xxxxxxxxxxxxxxxxxxxxxxxxxxx) for their
responses.

Norman Wilson
University of Toronto
norman@xxxxxxxxxxxxxxxxx


Follow-Ups: