Read a tape past EOT
2007-12-25 9:03:00
Well, this isn't so much a summary as it is a documentation of failure!
My original question asked how can I read a tape past the logical-end-of-tape
marker? I received several good replies, from all over the world! The ones
having to do with removing power to the tape drive while writing seem logic-
ally to stand the best chance of succeeding, yet, they haven't. I even went
so far as to write a short program to write records one-at-a-time to the
drive, waiting in-between so that I could remove power - also no luck. I also
tried reading (in a program, using dd, etc.) past LEOT, but the read always
returns a "zero" length; read (or perhaps the device driver) definitely knows
the difference between the first & second EOF marks on the tape.
When examining the tape after a write/power-off procedure, it always shows the
number of records I wrote, then an EOF then an EOT mark (using tcopy to
examine). At the same time, I always get an I/O or media error on the console.
Two things disturb me. One person mentioned that certain version of SunOS
won't read past an I/O error (hopefully not true in 4.1.3_U1), and another
person mentioned ejecting the tape between powering off the drive, and turning
it back on. Can this be done with an Exabyte 8500 drive? You don't suppose
the sneaky device writes an EOT upon power-up, having sensed the tape is in
an unusual position?
Anyway, further clues will be appreciated. Here are some details:
8mm tape written in 8500 (5GB) format dump/restore under SunOS4.1.3.
The tape constains /, /usr, and /home partitions. I want to get some data
from the /home partition, if possible.
The tapes were "erased" by writing a very short tar file to the beginning of
each one.
Powering off the drive while writing (via dd, program control, etc.) seems
not to work, upon trying to re-read the tape. Trying to read the tape on a
system now running SunOS 4.1.3_U1. There are Solaris 2.3 machines around...
Using a Sun "lunchbox" desktop Exabyte 8500 8mm drive to try to read tapes.
One person mentioned a 20% success rate. Am I being impatient? :->
Here's an ominous little bit from the "st" man page:
Reading past logical EOT is transparent to the user. Read-
ing is only stopped by reading EOF's. For 1/2" reel dev-
ices, it is possible to read off the end of the reel if you
read past the two file marks which mark the end of recorded
media. All other devices have safeguards to eliminate this
problem.
Many thanks to:
Bruce L. Harrell (bharrell@digit.com)
Nico Garcia (raoul@mit.edu)
Jimi Xenidis (jimix@resunix.ri.sickkids.on.ca)
John Cheshire (floating@float.demon.co.uk)
Steve Harris (etnibsd!vsh@uunet.uu.net)
Yuan Lu (lu@pgc.emr.ca)
Maxim Kozover (kozover@bimacs.cs.biu.ac.il)
Ian Camm (i.camm@fml.co.uk)
Steve Lucius (srl@eas069.msd.ray.com)
Regards,
Dave Edwards
MITRE Corp.
J-Lab Asst. Mgr.
Comments
Got something to say?
You must be logged in to post a comment.
