We are trying to integrate an HP 7980S (SCSI) 9-track tape drive with an IBM
RS/6000 S7A system, AIX 4.3.2. The combination works - almost. We're
experiencing problems with tapes written on other systems (IBM mainframe,
Data General). Specifically, we are trying to read tapes which have a
blocking factor of 6500 bytes (50 records of 130 bytes each) and which
typically consist of at least one full block and a partial final block. The
tapes are closed with a tape mark. AIX returns an I/O error after reading
the last block of the tape (errno = 05).
If I use the 'dd' command (dd if=/dev/rmt2 of=junk bs=6500), dd accurately
recovers the data on the tape but reports an I/O error. Any attempt to treat
the tape as a 'file' (e.g. cat /dev/rmt2 >junk) results in an error with no
data being recovered (the application we are migrating is heavily dependent
upon this method of access).
The device is defined in AIX with a block size of zero (variable block
length). Previous controlled tests, writing and reading from AIX using dd,
tar or cpio exhibit the same behavior if the amount of data written isn't an
exact multiple of 512 (this might actually be 1024... I'm a little fuzzy on
that detail right now and I'm away from the system).
I played with the "Suppress Illegal Length Indication" option and actually
made some progress by enabling it, but the previously described symptoms are
where I am at the moment. Also, I've noticed that the descriptions of the
configuration option value ranges don't always match the documentation. For
example, option 88 is documented as having ON/OFF values, but it displays a
setting of 17 on the front panel and can be cycled through a range of
numerical values, none of which are documented.
If you've shared a similar experience and have a solution or if you know
where I can find more useful documentation, I'd love to hear from you. I'd
also greatly appreciate it if you'd reply by e-mail as well as posting your
p.s. We're using the 7980S because it was already available(in the
original, unopened packaging) and we made the assumption that "It's SCSI -
it ought to work!".
Mark R. Rinfret
CBSI, Providence, RI