Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!news.kei.com!news.mathworks.com!tank.news.pipex.net!pipex!news.sprintlink.net!news.cloud9.net!cloud9.net!tls From: tls@cloud9.net (Thor Lancelot Simon) Newsgroups: comp.periphs.scsi,comp.os.linux.hardware,comp.os.linux.setup,comp.unix.bsd.freebsd.misc Subject: Re: SCSI diskIO Date: 25 Aug 1995 04:57:37 GMT Organization: Cloud 9 Internet, White Plains, New York, USA Lines: 42 Message-ID: <41jl81$3d8@news.cloud9.net> References: <417ee7$1kp@trauma.rn.com> <DDs7pM.KLA@ritz.mordor.com> <41g8re$11e@news.cloud9.net> <41hqvn$a1p@bonnie.tcd-dresden.de> NNTP-Posting-Host: cloud9.net Xref: euryale.cc.adfa.oz.au comp.periphs.scsi:36349 comp.os.linux.hardware:14376 comp.os.linux.setup:18368 comp.unix.bsd.freebsd.misc:5151 In article <41hqvn$a1p@bonnie.tcd-dresden.de>, J Wunsch <joerg_wunsch@uriah.heep.sax.de> wrote: >Thor Lancelot Simon <tls@cloud9.net> wrote: > >>You also -- IMHO -- *don't* want to have to debug the NetBSD/FreeBSD >>NCR driver if it breaks. It makes extensive use of self-modifying >>code. In fact, both the code running on the host CPU and the code >>running on the NCR modify the code that runs on the NCR. This is >>unpleasant to contemplate. > >Why should i want to? Stefan Esser's answers always returned in time. That's completely true. The driver authors have done a great job of keeping up with bug reports. On my side of the fence, however, the NetBSD core team seems to have been perpetually too busy to keep up with all the bugfixes -- and there have been a fairly large number. The driver seems fairly stable now and with the one or two NetBSD-specific changes that have been made rolled back in, which I hope will happen, NetBSD will probably track it more closely. Nonetheless, the result has been that from time to time I've needed to debug the driver myself. Stefan's answer squares pretty well with the limited understanding I've come to of the NCR code, but I still get very, very nervous _any_ time I touch self-modifying code. Even the "magic number" that's the length of the NCR code and the offsets of all the labels (required to let the host CPU modify them or jump to specific points in the NCR code) puts me off a bit. I do run several NCR controllers in important production systems, and I've had no trouble for some time now. The only problem I still have is that the 825 hardware itself seems to have some problem that makes it very, very slow when talking to certain wide disk drives -- definitely an issue to watch out for, and one that Symbios acknowledged they'd heard of last time I called them, though they didn't say if they knew why it happens. I usually now use NCR810s for narrow disks and BT956 or AHA2940W controllers for wide ones. -- Thor Lancelot Simon tls@cloud9.net Don't let your mouth write no check that your tail can't cash. --Bo Diddley