From: Digestifier To: Linux-Misc@senator-bedfellow.mit.edu Reply-To: Linux-Misc@senator-bedfellow.mit.edu Date: Tue, 17 Aug 93 07:36:56 EDT Subject: Linux-Misc Digest #11 Linux-Misc Digest #11, Volume #1 Tue, 17 Aug 93 07:36:56 EDT Contents: Re: Linux Satellites (Cheapest Linux Hardware Configuration) (Ralph Doncaster) Re: INN1.4 under Linux - WOW !!!!!! (Marc G Fournier) Re: INN1.4 under Linux - WOW !!!!!! (Marc G Fournier) Linux diskette labels (Chris Johnston) Re: High speed modems & Linux (Jim Graham) Re: From your friends at UNIXWorld (Ken Sorensen) often get "illegal instruction" w. latex and mf (Markus Berndt) Re: Linux comm programs, and mod tracker? (Sean Camanyag) Re: mouse problem on mcc distribution (Allan Dean) Re: Comments on the MCC Interim Release (Tuomas J Lukka) Re: NFS boot? dickless linux? (Mika Liljeberg) Re: Linux Satellites (Cheapest Linux Hardware Configuration) (Tuomas J Lukka) ---------------------------------------------------------------------------- From: rdoncast@bcoc61.on.bell.ca (Ralph Doncaster) Subject: Re: Linux Satellites (Cheapest Linux Hardware Configuration) Reply-To: rdoncast@yo_dud Date: Tue, 17 Aug 1993 01:27:27 GMT In article <53370@sdcc12.ucsd.edu> imcclogh@cs.ucsd.edu (Ian McCloghrie) writes: kenney@stein.u.washington.edu (Michael Kenney) writes: >In article <24ja52$nmc@indigo.imp.ch>, Philippe Steindl wrote: >>Hello, >> >>sounds alright, but don't forget that these cheapo ethernet cards are >>TwistedPair and not Thinwire (am I wrong?). Twistedpair needs a central >>unit collecting all the client wires .. and these units are expensive. >You can get an 8 port twisted pair hub for about $200.00. If you >can indeed get twisted pair cards for $30 (vs. $90 - $100 for thinnet) I've seen AUI+BNC ne2000 clones on sale for ~$60, at area retailers. (I went ahead and bought the SMC though, for $125) I've been told that, if you only want two hosts on your network, can just run a twisted pair cable between the two hubs. I've not tried it (never having had the hardware or inclination lying around) but it sounds reasonable. Of course, it only works if you've only got TWO hosts on your network. -- I've connected my Linux box and DOS box with 2 Boca BEN100 ethernet cards based on the AMD79c960. These have only 10baseT connectors; I made a cable with 2 rj-45's, crossing the tx and rx; works fine. The cards cost me CDN$73 each from Tech Data. (about $55US at the current exchange rate). They do DMA bus-mastering so the performance should be good (haven't done any benchmarks yet, and I just got the LANCE driver working before I went on vacation.) -Ralph -- Ralph Doncaster, computer consultant Bell Sygma Telecomm Solutions home email: ralph@dci.pinetree.org ph:(613)781-6774 fax:781-7677 " my other computer is a Cray-3 " email: rdoncast@on.bell.ca ------------------------------ Crossposted-To: news.software.nntp,news.software.nn From: marcf@nexus.yorku.ca (Marc G Fournier) Subject: Re: INN1.4 under Linux - WOW !!!!!! Date: Tue, 17 Aug 1993 04:44:40 GMT rsalz@rodan.UU.NET (Rich Salz) writes: >In <24gd2v$o5i@nz12.rz.uni-karlsruhe.de> s_titz@ira.uka.de (Olaf Titz) writes: >>running daemon that eats up more and more memory as time goes... >innd does not eat up more and more memory. It reaches a steady-state fairly >quickly. If you set INND_DBZINCORE to 0 then it doesn't use much memory >at all. Hmmm...can someone explain this DBZINCORE a little mnore to me? I tried -DMMAP, changed READ to MMAP the other day, compiled and installed aeverything and it totally killed my system...well...killed is a bad word, but basically, innd kept accepting articles, kept building overview files...but when I tried to connect to the server, it basically told me that no files existed. I went over to the server and looked around, and found that the active file hadn't been updated in over 12 hours *raised eyebrows* Again, this is under Linux pl12 with the net-2 code, which doesn't mean much to Rich (sorry) If someone can give me an idea of why I can't do this DBZINCORE...and can give me an idea of why I would want to use it...my meager understanding is that it keeps everything in 'memory' and thereby makes things run faster(?) On a different note, does anyone know of support for, or upcoming support for, overview files in nn? Something that would allow me to make use of what inn is already building, like trn 3.2 does? Thanks... Marc ------------------------------ From: marcf@nexus.yorku.ca (Marc G Fournier) Subject: Re: INN1.4 under Linux - WOW !!!!!! Date: Tue, 17 Aug 1993 04:51:28 GMT devet@adv.win.tue.nl (Arjan de Vet) writes: >(as is used for example in the INN newgroup script). These problems will >be fixed in version 1.13 which is more /bin/sh compatible than version >1.12. approx date for release? :) Marc ------------------------------ From: chris@dudes.surfcty.com (Chris Johnston) Subject: Linux diskette labels Date: Mon, 16 Aug 93 20:47:01 PDT I have put together Linux diskette labels for those who want them. They are done in Macintosh freehand, but i think I have successfully exported them into postscript files. I think they look cool. They are based upon the Linux logo that you can find posted on tsx-11.mit.edu in /pub/linux/postscript. These files are not small. They print onto Avery 5139 3.5" laser printer labels. If anyone wants the files (4 of 'em) I'll email them to you. If I get enough requests, I'll see if someone will want to post them onto some linux site (tsx-11 maybe?) or a mirror or whatever. If you want the source files, fine... Gladly give them to you. If you want the labels pre-printed for you after you look at them, we can work on price. All I want is my cost of the labels plus postage - that's all. Let me know what anyone thinks... Chris ****************************************************************** ** Chris D. Johnston 714/289-8543 ** ** Surf City Software 714/289-1002 (fax) ** ** chris@surfcty.com Surf.City (AppleLink) ** ** 1095 N. Main Street, M-1, Orange, California 92667 ** ** The Cart Before the Horse? Preposterous! ** ****************************************************************** ------------------------------ From: jim@n5ial.mythical.com (Jim Graham) Subject: Re: High speed modems & Linux Date: Tue, 17 Aug 1993 01:53:17 GMT Note followups. This question really should have been in cdm. I'm re-directing followups in that direction. Also, this is a reply to *TWO* posts---the first is the original post, from kfisher3@mach1.wlu.ca (Kevin G. Fisher), and the second is a followup post from entropy@world.std.com (Lawrence Foard). The first bit deals mostly with the 16550, and the next bit deals with serial port speeds, throughput, etc. >In article kfisher3@mach1.wlu.ca (Kevin G. Fisher) >writes: >>I just purchased a 14.4 modem (yay!). [....] >>Whenever I'm doing a file transfer at >>14.4, if I do anything (ie start up XV or something) while the transfer is >>happening, it looses packets until whatever I started is loaded. I realize >>this problem is probably caused by the cpu, uart or whatever being unable to >>keep up... >>Am I correct in this assumption? Most likely, yes. When data arrives at your computer, the UART triggers an interrupt telling the CPU that it needs to be serviced. If the CPU is busy servicing another interrupt that happens to have a higher priority, the UART has to wait. The older UART chips, the 8250 and 16450 (and chips that emulate them), only had a 1 byte FIFO (first-in first-out) buffer, meaning that if the data in the buffer wasn't taken care of before another character arrived, the buffer is overrun, and the first character is lost. The 16550A, and more recently, the 16550AF, have a 16 byte FIFO buffer. This means that instead of having only one character time for the CPU to service the interrupt, you now have up to 16 (read on...this varies), which gives you some breathing room. Now, that's only the first improvement...there's more. Every time the CPU has to stop what it's doing and service an interrupt, there is a certain amount of system overhead associated with this change (this is called ``context-switcing overhead''). In order to work to reduce the total amount of context-switching overhead associated with handling the incoming data stream, the 16550A is able to wait, and only issue an interrupt after either a certain number of characters (can be set to 1, 4, 8, or 14 --- 4 is probably your best setting --- see a recent discussion on cdm for details), or a certain amount of time since the last character (4 character times) was placed into the FIFO (w/o being serviced). Along with this, it provides the ability for the interrupt service routine (the software on the computer that services the interrupt) to read more than one character at a time. In other words, instead of issuing an interrupt for *EVERY* character, the UART now issues an interrupt for every n characters, and the system can retrieve all of the characters in the FIFO while servicing that one interrupt. What does this all boil down to? Better performance and the ability to support much higher speeds, even in a multi-tasking environment. >>I seem to remember in the days of OS/2 >>2.0's intro people running into the same problem, and the solution being to >>get a faster UART in their pc....... Yep---see above. You probably need to look into getting a 16550AF. Before you do, however, you need to make sure your serial board has either an 8250 or 16450 installed. Some (many) serial boards have chips that emulate these instead of having the actual chips. A *VISUAL* inspection is required to determine what you have. If you have either an 8250 or 16450, and if it's in a socket, your task is extremely simple. Otherwise, e-mail me. Btw, the 16550AF normally runs for $9 to $12 (US) plus shipping. I wouldn't advise paying any more than that (there are those who try to sell the chip for as high as $50 !!!). In article entropy@world.std.com (Lawrence Foard) writes: >Inorder to run at 14K baud on an 8250 your >machine has to keep up with atleast 1400 interrupts/second. Yes, but with a 14,400 bps modem, you'd never set the serial port to only 14.4 kb (since you say baud above, and not bps, it is assumed that you're referring to the serial port speed, as there is no such thing as a 14,400 baud modem for normal POTS lines). You'd want to set the serial port to *AT LEAST* 19.2 kb, preferably 38.4 or 57.6 kb. And as far as the 1440 cps / 1440 interrupts/second figure, you're forgetting about the async-to-sync conversion from V.42 (or MNP3). The actual data rate between modems (before any compression at all) is going to be more along the lines of 1724 cps (V.32bis w/ V.42 and a clean phone line). Assuming you have things setup on both ends such that there isn't a bottleneck between the modem and the computer, you're looking at more in the range of 1650 to 1724 interrupts/second. If you add data compression to this (and assuming you're not transferring pre-compressed data), the number gets even higher. How high depends on the data. If you want to get an idea of the ranges there, see the current discussion in cdm on this. Oh well, this post is long enough. :-) Virtually, --jim -- #include 73 DE N5IAL (/4) ==========================< Running Linux 0.99 PL9 >========================== INTERNET: jim@n5ial.mythical.com | j.graham@ieee.org ICBM: 30.23N 86.32W AMATEUR RADIO: (packet station temporarily offline) AMTOR SELCAL: NIAL ============================================================================== E-mail me for information about KAMterm (host mode for Kantronics TNCs). ------------------------------ From: ksore@atr-14 (Ken Sorensen) Subject: Re: From your friends at UNIXWorld Date: 16 Aug 93 16:13:10 GMT kevin@frobozz.sccsi.com (Kevin Brown) writes: : In article johnsonm@calypso.oit.unc.edu (Michael K. Johnson) writes: : >In article kevin@frobozz.sccsi.com (Kevin Brown) writes: : > : > Yup. Some things will just bring any Unix box down, e.g. : > "while ( 1 ) fork();", but that's not something particular to Linux. : > : >Yes it is -- It won't crash linux. : I always love doing this and watching my system grind to a near halt... main() { void *foo; for (;;) foo = (void *)malloc( ); } All it takes is a ctrl-c to kill, but if your in X, and not in that xterm... FORGET IT! You might as well shut the machine down. I did try this: #define MYSTRING "Let's have a party!" #define MEM 100 main() { char *foo; for (;;) { if ((foo = (char *)malloc( MEM )) == NULL) { printf("Oops! No more memory.\n"); exit(1); } else { strcpy(foo, MYSTRING); } } } I watched the memory drop down to nothing, and I never got the "Opps!..." And I also didn't get a SEGV either. Is this a problem with the libc malloc? The kernel? Or what??? BTW, I have 16M of memory + 16M swap. -- Ken Sorensen -= ksore@atr-2s.hac.com =- ------------------------------ From: berndt@numerik.uni-duesseldorf.de (Markus Berndt) Crossposted-To: comp.os.linux.help,de.comp.os.linux Subject: often get "illegal instruction" w. latex and mf Date: 17 Aug 1993 08:31:13 GMT Hi Linuxers, when I use latex or mf the program very often dies with "illegel instruction". Normally one would say that a program runs or it does not, but on my linux box mf works sometimes, allthough not very often (in 10% it works). Latex has the same behaviour. So I suspected a hardware problem and it seems really to be one: I've got a i486 ISA with an old OPTI chipset, 256kb cache 2 ide drives (one Maxtor 7120A and one Conner 30204) Et4000 I've installed SLS 1.03 with 0.99p11 (the pre Aug 11 rel of SLS) There is a Turbo Switch Jumper on the Motherboard and when I close it (slower mode) mf and latex work fine in most cases. When I run the board in fast mode an disable the external cache latex and mf work fine! I played with the BIOS Setup (waitstates, non caheable block, ...) but only disabeling the external cache helps. It is very strange, that all the other binaries like gcc XF86_SVGA and also my numerical mathematics codes (compiled with gcc 2.4.5 and liked with libc.so.4.4.1) work fine with the external cache enabled. The TeX binaries in SLS 1.03 are from feb 93 and they are linked with DLL Jump 4.3. Should I use newer binaries, should I buy a new motherboard (@#$%#$) or should I disable my external cache RAM for ever? I would be very happy about suggestions, explanations or comments because parts of my diploma thesis want to be texed ;-) Markus =============================================================================== Markus Berndt =============================================================================== Angewandte Mathematik | private adress: Heinrich Heine Universitaet Duesseldorf | Universitaetsstrasse 1 | Dachsbau 11 40225 Duesseldorf, Germany | 40789 Monheim, Germany Tel.: ++49 211 311 3460 | Tel.: ++49 2173 54419 e-mail: berndt@numerik.uni-duesseldorf.de | =============================================================================== ------------------------------ From: bgc@netcom.com (Sean Camanyag) Subject: Re: Linux comm programs, and mod tracker? Date: Tue, 17 Aug 1993 09:09:11 GMT byron@cc.gatech.edu (Byron A Jeff) writes: >Off the top of my head there kermit, minicom, seyon, and ecu. What exactly >are you looking for in a comm program? I guess what I'm looking for is something similar to the dos comm programs Qmodem or Telix. I suppose I could use the dos emulator but I'd rather keep looking for another linux comm program, or wait until the emulator is well beyond the alpha stage. I'll be trying out minicom and pcomm, which I overlooked the first time I FTP'd to sunsite. I'm hoping one of these two will be adequate for me; I'll find out after I finish reading newsgroups. -=======================================================================- | Sean Camanyag | "I am Homer of Borg. You will be...ooohhh Donuts! | | bgc@netcom.com | | -=======================================================================- ------------------------------ Crossposted-To: comp.windows.x.i386unix,comp.os.linux.help From: amd@cam-orl.co.uk (Allan Dean) Subject: Re: mouse problem on mcc distribution Date: Tue, 17 Aug 1993 09:40:33 GMT If you have a round, mini-din-plug mouse socket, this is probably a PS/2 mouse port. You will have to ... - Change your Xdefaults file so that the mouse is PS/2-style and uses /dev/psaux - Recompile your kernel to include PS/2 mouse support. The MCC standard release doesn't have this compiled in. Allan ------------------------------ From: lukka@klaava.Helsinki.FI (Tuomas J Lukka) Subject: Re: Comments on the MCC Interim Release Date: 13 Aug 1993 11:08:05 +0300 >age of old. Emacs-18.59 new. Emacs-19.15 >software gzip 1.0.x gzip 1.2.3 > old tar 1.1 recent on just about everything Tar 'manpage' is info. >Please don't flame me. I've installed Linux a (small) number of times >on several systems and I doubt that I'll touch SLS again. Same here. >I truly wish that the Linux community would give the MCC-interim >package the spotlight that it deserves (and perhaps a piece of the FAQ >the size of SLS's chunk!). I've come to think of exactly the same. I've gotten the impression that SLS is what it is: a collection of free software, whereas MCC is a lot more like a 'real', polished unixoid operating system. And what about the other installations, let's have comments on those! Who's installed TAMU, Slackware etc? Can you compare? The FAQ is maintained by the users, mostly, and MCC hasn't got the area apparently because less questions about it are asked. I doubt mr MacCauley would have anything against a good installation comparation in there? >The only drawback to using MCC is that it lacks a X-windows >installation. If you doubt that you can handle installing X on your >own, maybe SLS is for you. Then again, maybe not. And installing X is trivial on ones own. Tjl Tjl ------------------------------ From: liljeber@plootu.Helsinki.FI (Mika Liljeberg) Crossposted-To: comp.os.linux.help Subject: Re: NFS boot? dickless linux? Date: 15 Aug 1993 11:03:01 +0300 In article <24kncv$l4s@urmel.informatik.rwth-aachen.de> dak@messua.informatik.rwth-aachen.de (David Kastrup) wrote: > An interesting headline. Computers listening to others are to > be considered female? A harddisk is to be considered a penis > symbol? Wow, what *HAVE* I missed. No wonder my disk has > become that loud and vibrates. I thought it was the bearings, > but... Well, at least he didn't mistake Linus for Linux, as some people do. That would have been quite embarrassing.... > David Kastrup dak@pool.informatik.rwth-aachen.de > Tel: +49-241-72419 Fax: +49-241-79502 > Goethestr. 20, D-52064 Aachen Mika -- Mika Liljeberg Email: liljeber@hydra.Helsinki.FI Helsinki University Mika.Liljeberg@cs.Helsinki.FI Dept. of Computer Science ------------------------------ From: lukka@klaava.Helsinki.FI (Tuomas J Lukka) Subject: Re: Linux Satellites (Cheapest Linux Hardware Configuration) Date: 15 Aug 1993 23:29:03 +0300 >Here is my first crack at it (I almost forgot the video! ;-): >$100.00 - 33 Mhz 386sx motherboard >$120.00 - 4 MB of memory - Net bought of course ;-) [...] Rather, get more memory... 8 is a minimum for serious things, 16 is heavenly. Tjl ------------------------------ ** FOR YOUR REFERENCE ** The service address, to which questions about the list itself and requests to be added to or deleted from it should be directed, is: Internet: Linux-Misc-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.misc) via: Internet: Linux-Misc@NEWS-DIGESTS.MIT.EDU Linux may be obtained via one of these FTP sites: nic.funet.fi pub/OS/Linux tsx-11.mit.edu pub/linux sunsite.unc.edu pub/Linux End of Linux-Misc Digest ******************************