home *** CD-ROM | disk | FTP | other *** search
- Path: dialin35.zdv.uni-mainz.de!user
- From: bauec002@goofy.zdv.uni-mainz.de (Christian Bauer)
- Newsgroups: comp.sys.amiga.emulations
- Subject: Re: SS3.4 ejects my CD's
- Date: Wed, 20 Mar 1996 21:16:14 +0100
- Organization: .uni-mainz.de - "None of the known nameservers are responding."
- Distribution: world
- Message-ID: <bauec002-2003962116150001@dialin35.zdv.uni-mainz.de>
- References: <xRKJtMD3E3aHzB@thain.ods.de> <4ibja1$kpr@news.uni-paderborn.de> <xS+lYMD3E3aiz8@thain.ods.de>
- NNTP-Posting-Host: dialin35.zdv.uni-mainz.de
-
- In article <xS+lYMD3E3aiz8@thain.ods.de>, thorin@ods.de (Michael Boehmer) wrote:
-
- >RS> And you get your answer....the fastlane may have bazillions of bugs
- >RS> but THIS..i repeat...THIS..is NOT a problem of the Fastlane device.
- >RS> The reason the requester pops up is that Shapeshifter allows the MAC
- >RS> to pass a SCSIDirekt Request to the device where it asks the CDRom
- >RS> for 1 Block=2048 Bytes but tells the dma engine only to transfer 512
- >
- >Who tells it ? The MacOS or ShapeShifter ?
-
- The Mac program/driver that initiated the transfer.
-
- >Why do so many other host adapters work happily (even with DMA turned on)
- >with the ShapeShifter, but only the Fastlane Z3 not ?!?!?!?
-
- There are two (completely unrelated!) problems with the Fastlane (and
- derivates (CyberSCSI, Blizzard)):
-
- 1. XXL-sized "Phase Error" requesters. These used to pop up at boot time
- when a CD was in the drive under ShSh <=3.3 because MacOS checks all
- connected SCSI devices for a valid boot block on startup without checking
- the device class first (it even tries to boot from scanners and media
- changers) and reads only 512 bytes of the first sector. This is no
- problem for ShapeShifter and it doesn't expect anything from the SCSI
- controller than to report an error condition, but it becomes annoying,
- so starting with 3.4, only (removable) hard disks are checked for
- boot blocks.
-
- 2. The z3scsi.device (and derivates) reports wrong values in scsi_Actual
- of HD_SCSICMD requests when DMA is turned on. This is the major problem
- with the Fastlane which also caused the Oktagon to fail under ShapeShifter
- and which (for the Oktagon) is fixed by SSOktaPatch. ShapeShifter relies
- on correct scsi_Actual values as there is no general way to determine
- the length of SCSI data transfers from within the Mac SCSI Manager.
- The "summer" mode tries to work around this by assuming fixed block
- sizes for hard disks and CD-ROMs (as ShapeShifter 3.0 did).
-
- Bye,
- Christian
-
- --
- / Christian Bauer, Max-Planck-Str.60, 55124 Mainz, Germany
- \/ WWW homepage: http://www.uni-mainz.de/~bauec002/
-