Path: chonsp.franklin.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Unix near-death experience (was Re:Operating System Definition) Date: 09 Oct 1998 22:21:42 +0200 Organization: My own Private Self Lines: 70 Message-ID: References: <36134cd5.0@122.122.122.1> <3613B0CC.2AA4494F@danet.com> <§5NlT1.1442$Cf.1854726@news6.ispnews.com> <361e5bf0.0@news3.uswest.net> X-Newsreader: Gnus v5.3/Emacs 19.34 cmikk@tig.oss.uswest.net (Chris Mikkelson) writes: > > The problem is that X essentially "takes over" the keyboard and > screen. When it crashes, it doesn't release them. The "console > switching" keys ALT-F? require cooperation on the part of X or > the console driver, by the way, so they won't work either. One of the weaknesses of Linux (and possibley also *BSD and SCO). User processes touching the hardware directly. Nice for performance, bad for stability. Sort of an Microsoftism :-). I hope GGI will close this hole. Very interesting variant: beta version X server dies, xdm simply starts an new one, I log in and continue 5 secods after crash. Meanwhile squid (web cache) had nicely downloaded the page I had just started to get when the crash happened. No time lost. > What does work is to login to your box from another network box, > or a serial dumb terminal, and kill and restart X. I have done that one a few times (from notenook via ether to desktop). > Granted, I *have* had X crash my box, for real, a few times. That > was on FreeBSD 2.1.5 or something similarly aged. In those cases, > the machine simply rebooted (cold), no freezing. I once had an incompatible S3 video card (Actix), fast under DOS/Win with their special drivers. Deadly under Linux, full frozen system. Reset switch time. The card was free. Got rid of it for an ATI. > ObFolklore-or-new-thread?: Unix near-death experiences. New thread, one with folklore, not advocacy. > What I believe happened was that I triggered a bug in my disk, SCSI > controller, or the FreeBSD driver for my SCSI controller that > effectively disabled the disks. The networking stack was at least > sane enough to deal with ICMP, but I couldn't load telnetd or sshd > from the disk to log in :-(. Reluctantly, I rebooted the box. Just like what I did this afternoon on my SGI Indy at work. It still runs IRIX 5.3 which unlike 6.* uses automounter instead of the newer autofs. I am used to ummounting autofsed NFS mounts when they don't disappear on their own on the 6.* boxes. On 5.3 this is deadly, it kills the NFS client entirely, as I discovered today. Result: both Netscapes (loaded from server) are dead. I can not open new shells (.cshrc tests for server mount and if it is there "source"s cshrc from the server). The 2 root shells I had open had blocked on NFS before I noticed it. "su"ing from an user shell tried to get cshrc. The only stuff still running were the 3 root xterms to the Sun servers I was working on (the unmount was part of testing an change there) but remote login also failed for cshrc... HOw to convert an SGI into an 3 session X terminal in one easy lesson. After 5 minutes I conseded that it was reboot time. 6 weeks of of uptime (since having to unplug for moving the box) down the drain :-(. -- *** New home Addresses Mail and Web *** home: neil@franklin.ch.remove http://neil.franklin.ch/ work: franklin@arch.ethz.ch.remove http://caad.arch.ethz.ch/~franklin/ Microsoft is Software Communism, Fight for GNU Freedom!