home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.aux
- Path: sparky!uunet!destroyer!ncar!noao!noao.edu!tody
- From: tody@noao.edu (Doug Tody)
- Subject: Re: Yet more about 'tc' (blocking issues)
- Message-ID: <1992Nov17.213516.27266@noao.edu>
- Sender: news@noao.edu
- Nntp-Posting-Host: lepus.tuc.noao.edu
- Reply-To: tody@noao.edu
- Organization: National Optical Astronomy Observatories
- References: <1228@jagubox.gsfc.nasa.gov> <1992Nov11.032712.26391@noao.edu> <1241@jagubox.gsfc.nasa.gov> <1243@jagubox.gsfc.nasa.gov> <1244@jagubox.gsfc.nasa.gov> <1992Nov14.060233.3593@noao.edu> <1245@jagubox.gsfc.nasa.gov> <1246@jagubox.gsfc.nasa.gov>
- Date: Tue, 17 Nov 1992 21:35:16 GMT
- Lines: 46
-
- In article <1246@jagubox.gsfc.nasa.gov>, jim@jagubox.gsfc.nasa.gov (Jim Jagielski) writes:
- > jim@jagubox.gsfc.nasa.gov (Jim Jagielski) writes:
- >
- > >tody@noao.edu (Doug Tody) writes:
- >
- > >>I have no information on the SCSI manager and am only grasping for straws,
- > >>but a remote possibility exists if the SCSI manager allows transfers which
- > >>are a multiple of N times the device block size. In that case, setting the
- > >>block size to 1 byte would allow odd size transfers to the device.
- >
- > >This may be possible... Of course, 'tc' would 1st have to issue a
- > >Read Block Limits to see if the device can do down as low as 1byte and
- > >then do the Mode Select... This also would mean that things be set up
- > >so that the device isn't _really_ in variable mode but in a small fixed
- > >blocksize mode...
- >
- > Well, it looks like the above idea works! (Thanks Doug).
- >
- > Buttttt... performance under this mode is a super dog. After all, the
- > tape drive must handshake for each and every _byte_ transfered.
-
- Oh dear... I was suggesting this only in the hopes that it might serve to
- fake out the SCSI manager. It don't think it will work to actually set this
- block size at the device level, because this changes the tape format and we
- are back to the problem of having a tape format which can't be read on a
- different system. What I was hoping, lacking information on the 3.0 SCSI
- manager, was that the SCSI manager could be told by the driver that the
- device block size was 1, that the SCSI manager would hopefully permit device
- transfers N times this value, and that the actual device block size would be
- set unbeknownst to the SCSI manager, using the SCSI set mode command to the
- device (I'm still not completely convinced this won't work, though it would
- be a horrible kludge).
-
- It sounds to me like this SCSI manager was written to do fixed block mode
- i/o to SCSI disk devices. The stock tc driver wants the tape device to look
- like a disk in order to use this SCSI manager code. It might still be
- possible to work around this, but I am kind of stuck until the 3.0 device
- driver kit comes out, as I have no source for any of this stuff. Even so,
- if the limitation lies further into the A/UX kernel we may be stuck.
-
- In any case, thanks for the work on the tc driver Jim, we appreciate your
- efforts. Just being able to have 1K fixed block mode i/o helps a lot.
-
- --
- Doug Tody National Optical Astronomy Observatories
- tody@noao.edu P.O. Box 26732, Tucson, Arizona, 85726
-