mc%miranda.uucp@moc.jpl.nasa.gov (Mike Caplinger) (03/30/89)
I have a driver for a SCSI CDROM device that worked correctly under 3.4. When I tried to recompile it under 4.0, it failed to compile because the SCSI device unit block, struct scsi_unit in scsi.h, no longer has an element struct buf un_utab, which is described in a comment as "queue of requests". There have been a number of changes to the SCSI stuff, and I was able to figure most of them out by reading include files, but nowhere can I find a reference to the SCSI request queue. Messing with the SCSI stuff has always been difficult, because Sun's "open systems" policy apparently doesn't extend to the addition of new SCSI devices -- there's no documentation I can find about the structure of how SCSI devices work. Note that the CDROM is not being treated as a block device (only raw access is supported), but I still don't know enough about the SCSI driver system internals to do the job. Any help would be appreciated. The only thing I need to know is where to enqueue a SCSI I/O request for the generic SCSI driver code to find. thanks, Mike Caplinger, Ground Data Systems Scientist ASU/Caltech Mars Observer Camera Project mc@moc.jpl.nasa.gov
henry@uunet.uu.net (04/24/89)
>Messing with the SCSI stuff has always been difficult, because Sun's "open >systems" policy apparently doesn't extend to the addition of new SCSI >devices -- there's no documentation I can find about the structure of how >SCSI devices work... The fact is, Sun's "open systems" policy is a sham, a scam, and a swindle. When you compare Sun to Mips -- which doesn't make 1/10th as much fuss about "open systems" and is one hell of a lot more open about things like the kernel and the hardware -- you quickly realize this. I'm indebted to my frequent partner in crime and fellow Sun victim, Geoff Collyer, for summing it up nicely: "`Open systems' means they want your wallet to open." Henry Spencer at U of Toronto Zoology uunet!attcan!utzoo!henry henry@zoo.toronto.edu
ian@sq.sq.com (Ian Darwin) (05/10/89)
>Messing with the SCSI stuff has always been difficult, because Sun's "open >systems" policy apparently doesn't extend to the addition of new SCSI >devices -- there's no documentation I can find about the structure of how >SCSI devices work... [Henry's and Geoff's comments about "open systems for open wallets" deleted] Funny you should raise this: "As part of Sun's continuing efforts to enhance SunOS, SPARCstation 1 incorporates a new SCSI driver. This driver supports a more consistent set of SCSI operations, thereby simplifying the addition of new SCSI devices to the system. Equally important is the support for synchronous SCSI for high-speed data transfers." -- from an article on the design of the 4/110 (SPARCstation 1, a name that puts me in mind of a spark-station), by Bechtolsheim et al in Sun Technology, Spring 1989, page 53. They talk about revised documentation too, but don't specifically mention documenting the SCSI driver.