Return to BSD News archive
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.mira.net.au!news.netspace.net.au!serval.net.wsu.edu!netnews.nwnet.net!oracle.pnl.gov!osi-east2.es.net!cronkite.nersc.gov!dancer.ca.sandia.gov!overload.lbl.gov!agate!howland.reston.ans.net!usc!bloom-beacon.mit.edu!satisfied.elf.com!usenet From: *Hobbit* <hobbit@asylum.sf.ca.us> Newsgroups: comp.os.386bsd.bugs Subject: more on the "routed" thing, and a lesson... Date: 28 Jan 1995 01:59:39 EST Organization: large Lines: 29 Sender: root Approved: God Expires: 1 Apr 95 12:34 Message-ID: <3gcqrp$q90@satisfied.elf.com> NNTP-Posting-Host: asylum.sf.ca.us I may have spoken too soon, and realized today that I might *still* have skewed versions of important .h files like pmap.h, sysctl.h, ioctl.h ... This is my fault, and the next step is to pick through various makefiles to see how /usr/include is generated. Is /usr/src/include the right place to start?? There are .h files *all* over the place. At any rate, I've temporarily fixed my "routed" problems by sleazing together a little thing that just squirts the One Correct RIP Packet at the provider's router every 30 seconds. Takes up much less memory than gated would... WRT "mysterious crashes" -- a lot of new clone machines are being shipped with questionable factory BIOS setups, especially with regard to wait states. A good example: My extant memory is 70ns, but I just stuck 4 more meg of 80ns in. Because I'm paranoid I ran a fairly thorough memory test under DOS, which liked it fine, but then FBSD crashed after about 30 seconds of intense disk-reading. I went back to DOS to poke some more, and noticed that if I diddled the "turbo" switch between fast and slow during the memory testing, I'd get failures, but *only* in the region of 80ns memory. Aha, a clue. Setting the DRAM wait state from 0 to 1 fixed it, and a lesson was learned. I suspect that the "eventual" crashes above were caused by misreads in the buffer cache, because it wouldn't happen until I'd done something to use a lot of memory [such as copying a whole pile of files], and when the pages started landing in the slower memory from 8M to 12M, boom... _H*