lphillips@lpami.wimsey.bc.ca (Larry Phillips) (04/06/90)
In <10667@cbmvax.commodore.com>, mks@cbmvax.commodore.com (Michael Sinz - CATS) writes: >In article <PASSARET.90Apr5112459@brahe.crd.ge.com> <passaretti@crd.ge.com> writes: >> >>Ok, I'm about to get a 500/590 for hacking on (Cheapest >>alternative 'till the 3000 comes out, then a good X >>terminal). My housemate has a 1000/590 and we'd like to >>share SCSI devices. Is this possible?? Is it feasible?? > >Is it possible? Well... You need a new device driver, modify one of the >590s to be at a different SCSI ID, and write some from of arbitration. >(The best bet would be to have one of the devices act like a drive to >the other and then have that one control the disk. This makes it like a >server/client. Sharing the drive otherwise comes into the problem of who >has the right to read/write/cache any part of it.) I am currently awaiting delivery of a 2091, so this and a couple of other comments about the 2091 are of interest to me. I agree that the best bet would be, as you say, to write some form of arbitration making one host adapter look like a disk, but I am wondering if the 2091 supports SCSI bus arbitration (lower level than the arbitration you are speaking of), so that, for example, an IO device could act as a master on the bus. Disconnect/reconnect... I noticed a posting from Dave Haynie that mentioned the 2091 disconnecting between the command and data phases in case something else needs to use the bus. Is this 'generic' disconnect/reconnect capability? It would be real nice if it is, so I could keep the tape streaming without having to go to a separate tape controller. If it has this capability, can it be turned on/off on a physical device basis for those devices that do not support the capability? -larry -- Entomology bugs me. +-----------------------------------------------------------------------+ | // Larry Phillips | | \X/ lphillips@lpami.wimsey.bc.ca -or- uunet!van-bc!lpami!lphillips | | COMPUSERVE: 76703,4322 -or- 76703.4322@compuserve.com | +-----------------------------------------------------------------------+
mwandel@tiger.waterloo.edu (Markus Wandel) (04/08/90)
In article <1356@lpami.wimsey.bc.ca> lphillips@lpami.wimsey.bc.ca (Larry Phillips) writes: > > I agree that the best bet would be, as you say, to write some form of > arbitration making one host adapter look like a disk, but I am wondering if > the 2091 supports SCSI bus arbitration (lower level than the arbitration you > are speaking of), so that, for example, an IO device could act as a master > on the bus. > > Disconnect/reconnect... I noticed a posting from Dave Haynie that mentioned > the 2091 disconnecting between the command and data phases in case something > else needs to use the bus. Just thought I'd mention that one implies the other. You can't have a SCSI system which supports disconnection/reselection without having arbitration implemented, since a SCSI device processing a command while disconnected must be able to reselect the host at any time, and thus arbitrate for use of the bus. This is the same arbitration as is required when you want to have multiple initiators on the bus. In other words, the only issue would be to change the 2091's SCSI device ID, and how easy that is depends on its design. It's probably just a 1-byte patch to the device driver, if you can get it to run out of RAM. Further clarification: The 2091 does not "disconnect between command and data phases". The target device (disk) controls that, and it can disconnect and reconnect as many times as it wants, even between command bytes and after the data phase. The initiator and the target device negotiate the disconnect capability at the beginning of each command, after that everything is up to the target to decide. Markus Wandel mwandel@tiger.waterloo.edu (519) 884-9547
mks@cbmvax.commodore.com (Michael Sinz - CATS) (04/09/90)
In article <1356@lpami.wimsey.bc.ca> lphillips@lpami.wimsey.bc.ca (Larry Phillips) writes: > [ lots-o-stuff deleted ] >Disconnect/reconnect... I noticed a posting from Dave Haynie that mentioned the >2091 disconnecting between the command and data phases in case something else >needs to use the bus. Is this 'generic' disconnect/reconnect capability? It >would be real nice if it is, so I could keep the tape streaming without having >to go to a separate tape controller. If it has this capability, can it be >turned on/off on a physical device basis for those devices that do not support >the capability? Yes, the Deselect/reselect is available on a device (SCSI-ID) by device level. It is stored in the 2091 HDToolBox under the drive type definitions. Yes, this would make streaming possible. /----------------------------------------------------------------------\ | /// Michael Sinz -- CATS/Amiga Software Engineer | | /// PHONE 215-431-9422 UUCP ( uunet | rutgers ) !cbmvax!mks | | /// | |\\\/// "I don't think so," said Ren'e Descartes. | | \XX/ Just then, he vanished. | \----------------------------------------------------------------------/
harald@boink.UUCP (Harald Milne) (04/11/90)
In article <10710@cbmvax.commodore.com>, mks@cbmvax.commodore.com (Michael Sinz - CATS) writes: > Yes, the Deselect/reselect is available on a device (SCSI-ID) by device > level. It is stored in the 2091 HDToolBox under the drive type definitions. > > Yes, this would make streaming possible. Question: Is this feature only available in the 2091 because of change in hardware, or is this just a software change in roms? If this is a software change, do current owners of 2090 and 2090A have an upgrade path from CBM? If 2090 or 2090A owners can't upgrade, then it won't be possible to make use of this feature without alienating them. I can stream without this feature, but it is very difficult to maintain. If everybody can upgrade, then deselect/reselect is a viable option, and everybody can get the same performance. It's an ugly sight watching a tape drive cough and wheeze. It also takes a lot more time. A lot more. > > /----------------------------------------------------------------------\ > | /// Michael Sinz -- CATS/Amiga Software Engineer | > | /// PHONE 215-431-9422 UUCP ( uunet | rutgers ) !cbmvax!mks | > | /// | > |\\\/// "I don't think so," said Ren'e Descartes. | > | \XX/ Just then, he vanished. | > \----------------------------------------------------------------------/ -- Harald Milne RISCy business uunet!ccicpg!boink!harald