Return to BSD News archive
Newsgroups: comp.periphs.scsi,comp.os.linux.hardware,comp.unix.bsd.freebsd.misc,comp.os.linux.setup Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msunews!agate!usenet.kornet.nm.kr!news.kreonet.re.kr!europa.chnt.gtegsc.com!howland.reston.ans.net!ix.netcom.com!netcom.com!stephenk From: stephenk@netcom.com (Stephen Knilans) Subject: Re: SCSI PCI host adapter Message-ID: <stephenkDEM2I3.AF4@netcom.com> Organization: Netcom Online Communications Services (408-241-9760 login: guest) References: <danielDEE700.Guv@netcom.com> <42lg2j$pks@news.parc.xerox.com> <42p9v5$hbn@bonnie.tcd-dresden.de> Date: Fri, 8 Sep 1995 23:56:27 GMT Lines: 47 Sender: stephenk@netcom6.netcom.com Xref: euryale.cc.adfa.oz.au comp.periphs.scsi:37266 comp.os.linux.hardware:15580 comp.unix.bsd.freebsd.misc:5883 comp.os.linux.setup:20131 In article <42p9v5$hbn@bonnie.tcd-dresden.de> joerg_wunsch@uriah.heep.sax.de writes: >Peter MacLeod <macleod@adoc.xerox.com> wrote: > >[disks do not work any longer after replacing the SCSI controller] > >>In the case of real operating systems which have SCSI managers and drivers, >>and ignore the whole BIOS mess, I would think that disks swapped between >>cards would be compatible, especially if one didn't try to boot off of the >>SCSI drive. This might be naive, however--many card vendors don't understand >>the purpose of a standard like SCSI, and might "be clever" and break it for >>no good reason. > >Booting is just the only issue. It's done by the firmware (or: BIOS), >so the operating system cannot do anything here. For a non-boot disk, >you are free to label it however you like, and they should be >exchangeable between different controllers. > >For bootable disks, those where the operating system boots off sector >0 (ie., the MBR is the bootstrap of the operating system itself) will >be exchangeable, too, as long as the BIOS idea of the geometry is >sufficient to load the bootstrap. (For *BSD, the bootstrap is 15 >sectors long, so every usual geometry idea will work.) Actually, the problem is that drives USED to be controlled directly. The controllers were named for the method RLL or MFM. Today, I think most drives are STILL RLL, but you can't really control the drive directly. They are controlled through a protocol that is based on a two phase system. The controller, because old O/S' used track/sector/head has to convert that to blocks. The drive then has to convert the blocks into track/sector/head! The method could vary GREATLY. In fact, it is concievable that reading 15 sectors from one drive could actually only read 1 sector from a head! It could then be VERY efficient! Each platter could be offset slightly, or multiple controllers could be used, so that reading 30 sectors off a 15 platter drive would take no more time(with regard to head seek and rotational latency), than reading one sector! Hence, even reading 15 blocks may not be the same from one disk to another. You probably wouldn't have a problem with THAT unless it was a multi surface removable pack though. It IS amazing that nobody ever seemed to think of this. Adaptec even has TWO translation methods! Hence, the SAME controller may be incompatible with ITSELF! As I recall, one is more compatible with old drives, and one is more flexible with larger drives. Steve