Re: [PARPORT] Tape backup with mt & tar

Art Wildman (
Wed, 25 Aug 1999 01:07:35 -0400

Thanks for the detailed explanation, I had no idea the problem requires a kernel
hack - thought something may have been broke with my version of mt.

1) Thanks for your work! - we depend on you guys to help make all this cheap
hardware we slap together work over an ancient parallel port. I could always go
out and buy an expensive SCSI Tape Drive and PCMCIA contoller, but that would
cost more than my laptop, and wouldn't be any fun.
2) Just a thought from leftfield - Does the "ftape" loadable kernel module use
the same conventions for writing EOFs? - the drive hardware is similar although
the interface is different. Perhaps you could borrow something from the ftape
sources. Hope this helps -good luck.
FTAPE Homepage...
"ftape supports drives that conform to the QIC-117 and one of the QIC-40,
QIC-80, QIC-40, QIC-3010, and QIC-3020 standards as well as the Ditto 2GB and
Ditto Max (Pro) tape drives which partly use proprietary extensions of the
QIC-117 command set. ftape supports also certain kinds of parallel port tape
drives. Check out the Ftape HOWTO for more details.

ftape supports neither QIC-02, IDE (ATAPI), nor SCSI tape drives. SCSI drives
are accessed as /dev/[n]st[0-7] and are supported by the kernel through the SCSI
drivers. ATAPI tape drives are supported by the kernel since 1.3.46 through the
IDE drivers."

Linux Journal Backup Articles

"J. Scott Berg" wrote:
> On Tue, 24 Aug 1999, Art Wildman wrote:
> > # mt -f /dev/npt0 fsf 1
> > pt0: HP COLORADO 8GBe, master, blocksize 512, 4007 MB
> > pt0: Unimplemented mt_op 1
> > /dev/pt0: Invalid argument
> > ... how do I skip forward to next archive???
> >
> > This makes storing multiple tar files on a tape useless, if you
> > can't retrieve them. Please explain what's going on.
> Pretty much what it says: "unimplemented mt_op." The only mt
> operation implemented at this point is rewind, as you discovered.
> I've implemented the filemark-related mt ops, but there are a couple
> issues to get settled before releasing it to prime time:
> 1) This is the first time I've written anything into the kernel
> proper, so there's a substantive chance that it will crash your
> kernel. It doesn't crash mine, but YMMV. Actually, what I've done is
> virtually identical to what is done for a rewind, so I would hope it
> would be harmless, but I'm waiting for feedback to make sure.
> 2) As far as I can tell, the SCSI tape driver and the IDE tape driver
> both have different conventions for writing filemarks. I've adopted a
> third one B~). I think (?) the different conventions are irrelevant if
> you never make a weof mt ioctl.
> Since I sent this to Grant 1-1/2 weeks ago and haven't heard back,
> I'll post the kernel patch (against 2.2.11) with the cavaet that you
> don't complain when it trashes your hard drive and that you promise
> not to try to write filemarks. Please send me any feedback.
> --
> J. Scott Berg
> 3025 E. Amy Ln. (812) 339-8368
> Bloomington, IN 47408-4220

   Art Wildman -    
"Share what you know. Learn what you dont."   - Motto

-- To unsubscribe, send mail to: -- -- with the single word "unsubscribe" in the body of the message. --

This archive was generated by hypermail 2.0b3 on Wed 25 Aug 1999 - 01:07:33 EDT