Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!spool.mu.edu!uwm.edu!math.ohio-state.edu!howland.erols.net!agate!theos.com!deraadt From: deraadt@theos.com (Theo de Raadt) Newsgroups: comp.unix.bsd.misc,comp.unix.bsd.bsdi.misc,comp.unix.bsd.freebsd.misc,comp.unix.bsd.netbsd.misc Subject: Re: Differences between BSDI, FreeBSD and NetBSD? Date: 05 Jan 1997 20:08:44 GMT Organization: Theo Ports Kernels For Fun And Profit Lines: 50 Message-ID: <DERAADT.97Jan5130844@zeus.theos.com> References: <6OBgx1wrNgB@me-tech.PFM-Mainz.de> <1997Jan4.095835.23223@wavehh.hanse.de> <x7dhgkwius7.fsf@the-light-fantastic.MIT.EDU> NNTP-Posting-Host: zeus.theos.com In-reply-to: ghudson@mit.edu's message of 04 Jan 1997 23:57:12 -0500 Xref: euryale.cc.adfa.oz.au comp.unix.bsd.misc:1880 comp.unix.bsd.bsdi.misc:5525 comp.unix.bsd.freebsd.misc:33663 comp.unix.bsd.netbsd.misc:5082 In article <x7dhgkwius7.fsf@the-light-fantastic.MIT.EDU> ghudson@mit.edu (Greg Hudson) writes: A fix for this problem, in the NetBSD VM, was developed by an OpenBSD developer (Niklas Hallqvist, niklas@appli.se). It's been in OpenBSD for a while, but until quite recently it had a bug which could cause systems to crash under heavy load (a VM object is freed while still in use). Around Christmas, Niklas developed a fix for the crashing bug, and it was committed to NetBSD-current on January 3. In terms of releases, NetBSD 1.2 still has the swap leak problem, while I believe OpenBSD 2.0 has the crashing bug instead. There were two problems. Once resulted in a number of 0xdeadbeef messages showing up on the console (ie. kernel warning that memory was modified in the kernel after a free); this was eventually diagnosed to cause a crash as well. The other problem was different and would also cause a different crash. These two were fixed about a month apart from each other. However I should note that only one person has ever reported a crash from these datastructure races found in OpenBSD 2.0; and he's running a PPro into the ground with thousands of invocations of sendmail. Far more people have suffered from the previous swap leak problem. Over the past few months it has been pleasant to see this guy's uptime go from 18 hours to the current 2+ weeks (now he's getting hit by other problems ;-) I would definately argue that the OpenBSD 2.0 crash was an improvement over the old behaviour. By the time an OpenBSD 2.0 box crashed from the vm leak fix BUG it would long before have run into serious problems from the old vm swap leak problem. > OpenBSD is supposed to have some performance fixes, although still > far away from FreeBSD's memory performance. There was a performance fix bandied about for low-swap conditions. That wasn't really a performance fix. It was a kludge around a stupid design. We've made a number of reliability fixes first, rather than those affecting performance. OpenBSD 2.0 shipped with some fixes for this, like some of the i386 ptdi panics were resolved. In current it is much better, and a 512MB i386 machine should work reliably using the stock kernel. We'd love to have a better vm system, but without someone to spend a lot of time to replace it these fixes are very worthwhile. -- This space not left unintentionally unblank. deraadt@theos.com www.OpenBSD.org -- We're fixing security problems so you can sleep at night.