Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!noc.netcom.net!news.consultix.com!newshost.marcam.com!news.mathworks.com!tank.news.pipex.net!pipex!in2.uu.net!news.erinet.com!inquo!inquo.net!pete From: pete@inquo.net (Pete Kruckenberg) Newsgroups: news.software.nntp,comp.unix.bsd.freebsd.misc Subject: INN sharedactive.patch problems with FreeBSD 2.0.5R Date: 25 Aug 1995 20:11:04 GMT Organization: inQuo Internet (801) 530-7160 Lines: 34 Distribution: world Message-ID: <41laoo$ajq@xymox.dsw.com> NNTP-Posting-Host: inquo.net X-Newsreader: TIN [version 1.2 PL2] Xref: euryale.cc.adfa.oz.au news.software.nntp:15391 comp.unix.bsd.freebsd.misc:5259 Jacques Caron (jcaron@pressimage.net) wrote: : In article <410j0i$lr1@xymox.dsw.com>, pete@inquo.net (Pete Kruckenberg) wrote: : >I put inn1.4unoff-2 on my FreeBSD 2.0.5R machine last night, and : >everything worked but the patch for nnrpd that allows it to use shared : >memory to share the active file. Whenever nnrpd would be activated, it : >would die with a signal 12 (bad argument to system call). : > : >Does anyone have nnrpd running with this patch on FreeBSD (or any other : >BSD4.4-based system)? Anyone have any ideas what might be causing this : >problem? What's the best/easiest way to debug a daemon like this? : Did you activate the SYSVSHM and SYSVSEM options when building your kernel? OK, I missed that one. I did put these into the kernel, and now it runs. However, it seems to be quite unstable. For instance, I had 8 simultaneous NNRP sessions going. I then killed the NNRP sessions (*not* nnrpd, but the session on the remote host). This locked up nnrpd. This was with XSEM_UNDO as both 0 and SEM_UNDO. Has anyone had good experiences running this on any platform? It seems quite unstable to me? I think that the whole semaphore thing might need to be re-done somehow, to allow some kind of recovery without having to reboot the news server. How about using a timer of some sort, or a file, so that the semaphore can be over-ridden if necessary? Or, could it possibly be rewritten to have just one write-access nnrpd session, which updates the shared memory, and the rest are read-only? I'll look into these possibilities, but I'd like to hear the experiences and ideas of other people. Pete Kruckenberg pete@inquo.net