Return to BSD News archive
Newsgroups: comp.os.386bsd.questions Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!howland.reston.ans.net!spool.mu.edu!cass.ma02.bull.com!syd.bull.oz.au!melb.bull.oz.au!zen!sjg From: sjg@zen.void.oz.au (Simon J. Gerraty) Subject: [386bsd] NFS problems with dodgy Sun. Organization: Zen programming... Message-ID: <1993Oct12.141637.10409@zen.void.oz.au> Date: Tue, 12 Oct 1993 14:16:37 GMT Lines: 75 Anybody had problems with file corruption via NFS ? My old sun386i has just returned home from a six month visit to the repair shop... (it got hit by lightning) Its diagnostics still say the CPU is faulty and it now loses 15 minutes per hour! The repairer claimed it was fully functional. As you can imagine, I'm expecting it to leave again soon. So I didn't want to spend much time making it an integral part of my network at home... The first thing I wanted to do is update the ksh on it. So I cvs check'd out a copy of pdksh and NFS mounted /users from my 386bsd system (i486 running 386bsd-0.1+0.2.4 - yes I will upgrade to FreeBSD or NetBSD when I have some spare time...) and configured it for SunOS-4.0 The compile went well for quite some time and then died with cc reporting a syntax error in one of the .c files. I more'd the file (on the Sun) and it appeared to contain a few bytes of crap about half way through. Funny, it compiled ok on the BSD system the otherday.... I more'd the same file in the same directory on the 486 and it was ok! But of course NFS on the Sun had it as bad. Rm the file and cvs update to replace it. Now more on the Sun showed it to be ok. So start the compile again. A few minutes later the same problem with another file. Later, trying to run the old sun X binaries via NFS gave exciting though random results (usually not good). I was trying to avoid populating the tiny disk left on the Sun... restoring the X binaries to a local filesystem made everything work ok. So, the files are corrupted via NFS. But is it the 486 or the Sun at fault? rcp,ftp etc work fine. NFS sucks. I tried: i486$ rcp sun:/vmunix /tmp i486$ rcp /tmp/vmunix sun:/tmp sun$ cmp /vmunix /tmp/vmunix # ok Running the rcp's from the Sun works too. But: sun$ cd /users/sjg # NFS mounted from i486 sun$ cp /vmunix tf1 # read local write NFS i486$ cmp /tmp/vmunix ~/tf1 # fails fails at random points in the file (/tmp/vmunix is left over from previous test and should be ok as copying back to the Sun it still compares ok). sun$ cp tf1 tf2 # read and write NFS sun$ cmp tf1 tf2 # fails sun$ cmp /vmunix tf2 # fails sun$ cp tf1 /tmp # read NFS write local sun$ cmp tf1 /tmp/tf1 # fails So it is only NFS in trouble and just reading is enough to corrupt data. I tried cutting rsize and wsize to 1024 and even 512 bytes with no improvement. Now, the hard bit... which is at fault? The Sun I hope, but I've not had a 386bsd system on a network before so I cannot satisfy myself that the i486 is completely blamless. So... has anyone else experienced problems as described above? For the record I'm using ed0 and a WD8003EBT. [At least I now know that my WD8003 is jumpered correctly - I hope:-)] --sjg -- Simon J. Gerraty <sjg@zen.void.oz.au> #include <disclaimer> /* imagine something _very_ witty here */