Return to BSD News archive
Path: sserve!manuel!munnari.oz.au!hp9000.csc.cuhk.hk!uakari.primate.wisc.edu!sdd.hp.com!cs.utexas.edu!uunet!usc!sol.ctr.columbia.edu!destroyer!ubc-cs!uw-beaver!yenbut From: yenbut@cs.washington.edu (Voradesh Yenbut) Newsgroups: comp.unix.bsd Subject: Re: Hang up on heavy disk load (Re: Adaptec SCSI bug?) Message-ID: <1992Aug6.055725.8222@beaver.cs.washington.edu> Date: 6 Aug 92 05:57:25 GMT References: <1992Aug3.150008.188@cotton.nc.u-tokyo.ac.jp>> <TMH.92Aug3185956@hektor.cs.tu-berlin.de> <1552@hcshh.hcs.de> Sender: news@beaver.cs.washington.edu (USENET News System) Organization: Computer Science & Engineering, U. of Washington, Seattle Lines: 41 Let me share some of my experience. My system had hang up problem on SCSI driver since I used MINIX. The problem occurred after a SCSI drive had been accessed for a while. I found that invalid CCB messages (if I recalled correctly, the actual message showed only either code number or none at all) may occur and then the 1542B board did not make an interrupt so the system hanged. After spending some time debugging, I found lowering the DMA transfer speed of 1542B to the lowest one, i.e., 3.3 MB/s, fixes the problem. The same problem occurred when I run BSD386, so my patch to as.c was applied to lower the speed, but I found two other problems. 1) The SCSI disk could not be read reliably (multiple runs of fsck on the same file system did not return the same result) and the system may hang or crash (usually with vm_fault) after heavy disk load -- etcdist had never been completely extracted. The problems did not occur with an ESDI disk on my system. After several desparate attempts, last weekend I found if a (64KB) buffer is provided inside as.c for 1542B to put data in by disk read operations and then the buffer content is copied back to requester's buffer. Disk can be read more reliably and etcdist can be fully extracted. It sounds to me like a buffer may be paged out during a DMA operation, or something is at odd on SCATTER reading with multiple buffers. 2) Occasionally "NMI port 61 a0, port 70 ff" appears or the system gets panic with trap type 19 when a SCSI disk is accessed. When I get the message, the system may have to be rebooted afterward. So far, I haven't figured out what was wrong. I really appreciate if anybody has information on this. My system uses: Micronics 486 80486-25 MHz ISA system CPU board with the following disks and controllers: Ultrastor 12(F) with WREN 3 ESDI as the first drive Adaptec 1542B with CDC94171 and Microp 1598