Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.mel.connect.com.au!munnari.OZ.AU!news.ecn.uoknor.edu!news.uoknor.edu!news.nodak.edu!netnews1.nwnet.net!news.u.washington.edu!news.alt.net!news.serv.net!solaris.cc.vt.edu!news.mathworks.com!newsxfer.itd.umich.edu!news.eecs.umich.edu!panix!not-for-mail From: jbrann@panix.com (John Brann) Newsgroups: comp.unix.bsd.freebsd.misc,comp.periphs.scsi Subject: Re: SCSI-2 Bus Hangs Date: 28 Dec 1995 15:37:38 -0500 Organization: PANIX Public Access Internet and Unix, NYC Lines: 65 Message-ID: <4buv6i$j2o@panix3.panix.com> References: <4bju5p$ihj@wilma.widomaker.com> NNTP-Posting-Host: panix3.panix.com Keywords: SCSI-2 Lockup Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:11661 comp.periphs.scsi:44018 In <4bju5p$ihj@wilma.widomaker.com> bloyall@widomaker.com (Ben Loyall) writes: >I have a Usenet server that keeps locking up. When it locks, the SCSI-2 >disk in question will have its access light light solidly. The rest of >the machine is still running, but nothing can access the disk. >Eventually, all process block waiting on i/o and the machine is dead. >A complete power down is the only cure. This used to happen on average >every 7 days when the drive was controlled by an AHA-1742/I486dx2 combo. >But since going to NCR-8510/IP100, It is usually once every 24 hours. >The only thing I can attribute that to is the speed increase. >As the disk in question is a newspool on a rather busy newsbox, it gets a >good workout. >The problem has plauged me for at least 9 months. I had always thought >it was a bum hdd. However, the drive in question (Seagate Baracudda 2GB) >was removed this week and a Conner 4.2GB was put in its place. This >morning I found the machine in the same locked state. >I have seen this problem under 4 OSes (Unixware 1.1.3, RH Linux, FreeBSD >2.0.5 & FreeBSD 2.1). It has happened on 2 different motherboard/cpu sets >(Micronics 486-EISA w/486-DX2 & lately, SiS MTI-R526 w/P-100). Also >observed on the following controllers: AHA-1742 (Enhanced only) & and >NCR 8510 based PCI card. >The bus has 4 devices on it. id0=Seagate ST1480N (406MB SCSI-2), >id1=Maxtor 7213S (202MB SCSI-1), id2=Maxtor 7345S (329MB SCSI-1), >id3=Conner CFP4207S (4.2GB Fast SCSI-2). >id0 is at the far end of the bus and is terminated & the controller is >properly terminated. The other devices have their terminators removed. >Any help on this would be greatly appreciated. I suffered similar problems, up until I installed XFree86 v3.1.2. My experience may not be relevant, but the range of places you have seen the problem may be indicative. My machine is just a personal workstation, and doesn't see the kind of workload that a News server gets, so the number of crashes I saw was small. In my case the problem was connected with memory mapping on the PCI bus. I have an Adaptec 2940 and a PCI video card (CL-GD5434). Neither of these came with very extensive docco, so I slapped them onto the motherboard, and everything seemed to work OK. I had 2 problems which were causing system hangs of the type you describe - the first was doing backups to tape using 'dump' (I have a Conner tape drive on the SCSI bus) and the other was connected with doing fancy graphics in Windog. In Xfree86 3.1.2 is a handy program called 'scanpci' which gives a readout of the territorial demands of the various devices on the PCI bus. Sure enough, the graphics and SCSI card were claiming the same address. The address claimed is _not_ configurable but is dependent on the slot occupied by the card. So I swapped the two cards around and hey-presto, problem solved. Since doing this move I have not had a single crash. Good luck John >-- >Ben Loyall | (804) 221-8070 28.8 & down >Widomaker Public Access Internet | Unix shell for 60 hours/month, $20 >Williamsburg, Virginia | PPP/SLIP for 60 hours/month, $30 > Shells, PPP/SLIP, Dedicated IP, Feeds, DNS and MX available