Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!foxhound.dsto.gov.au!fang.dsto.gov.au!yoyo.aarnet.edu.au!myall.awadi.com.au!myall!blymn From: blymn@mallee.awadi.com.au (Brett Lymn) Newsgroups: comp.os.386bsd.questions Subject: Re: NETBSD and freezing problems Date: 22 Oct 1993 02:24:15 GMT Organization: AWA Defence Industries Lines: 28 Message-ID: <BLYMN.93Oct21202416@mallee.awadi.com.au> References: <29n0in$446@darkstar.UCSC.EDU> NNTP-Posting-Host: mallee.awadi.com.au In-reply-to: krostrin@cats.ucsc.edu's message of 15 Oct 1993 20:18:31 GMT >>>>> On 15 Oct 1993 20:18:31 GMT, krostrin@cats.ucsc.edu (Ken Ostrin) said: Ken> NNTP-Posting-Host: as215-ws-14.ucsc.edu Ken> I have seen several posts about NetBSD freezing (leaving the HDD light going Ken> but not doing anything) I am having the same problem, and I wondered if there Ken> has been any real solutions yet? In my case it seems to be related to swap. Ken> When I take the "swapon" command out of etc/rc the machine doesn't seem to Ken> crash, however, when I put it back in it almost always crashes. I need There are a few bugs that will cause the symptom you describe. One of them was in the original wd.c drivers (I'm talking 0.1 here), they trusted the hardware to do the right thing all the time, sadly this is not the case on some machines when the disk activity rises. I did make a fix for this under 386BSD 0.1, I do not know if the fix made it into NetBSD. From what I understood a "better" version of the fix (something about using kernel timeouts instead of the poll loop I had). Can anyone using NetBSD confirm/deny this is fixed? What you need to look for is some logic around the loops that check for the controller/disk being busy in wdstart() that will reset and restart the operation if the disk/controller stays busy for too long. I do have a netbsd0.8 version of my patch that someone else ported for me if anyone wants to try jamming it into netbsd0.9 -- Brett Lymn