From: Digestifier To: Linux-Admin@senator-bedfellow.mit.edu Reply-To: Linux-Admin@senator-bedfellow.mit.edu Date: Thu, 28 Oct 93 14:14:35 EDT Subject: Linux-Admin Digest #128 Linux-Admin Digest #128, Volume #1 Thu, 28 Oct 93 14:14:35 EDT Contents: Re: How to use a secondary IDE interface. (John Ngai) Drive mapping problems... (Arthur Harvey) Re: Fails: mounting Linux filesystem fro (Gareth Bult) Re: lilo, vmlinuz zImage (Craig Sanders) LAN Software for Linux (Dhawal Tyagi) Re: Chatting programs without net (tim) Re: syslogd hangs inetd (SuperGirl) Re: Help with routing under Linux (Dan Warburton) Bug in shadow-packages (patch included) (Hendrik G. Seliger) ---------------------------------------------------------------------------- From: ngai@nova.bellcore.com (John Ngai) Subject: Re: How to use a secondary IDE interface. Reply-To: ngai@apostle.bellcore.com Date: Wed, 27 Oct 1993 16:52:41 GMT I want to thank you for all the tips I received over email about using the 2nd IDE controller. It is wonderful to be in a group of people so willing to help out. Now I have a follow up question. I read all these *great* benefits of using SCSI with multiple drives, that allows concurrent disk access. This makes me wonder if I can derive similar performance benefits if I put the swap and root (mostly system executables) partitions on a separate hard disk under IDE controller 1, and user home directories on another hard disk under IDE controller 2. I can see 3 partitioning scenarios: (1) Swap/root and /home on same IDE hard disk. (2) Swap/root and /home on different hard disks but under same IDE controller. (3) Swap/root and /home on different hard disks and each under a different IDE controller. Case (1) is slow since the disk head has to move back and forth to access swap and user date files. Case (2) seems to solve the problem of case (1), but will there be access overhead for the IDE controller to switch between its two drives? Case (3) may or may not solve the problem in case (2), depending on if LINUX can simultaneously issue read commands to both controllers to get the two drives going. I am ignorant about the IDE interface. Can someone comment on the relative merits of these cases? Thanks -- John ------------------------------ From: harvey@gate.ee.lsu.edu (Arthur Harvey) Subject: Drive mapping problems... Date: 27 Oct 1993 16:35:09 GMT I've got .9.11 running on my secondary drive on my primary drive controller ( ultrastor 12f - ESDI). It runs great - But it will not allow me to access my 740mb Hitachi which is the primary drive on that controller. I've got the sector mapping on the drive enabled so that I can keep a dos file system on that drive as well ,but the linux can't seem to handle it. I've tried rebuilding my kernel with xt support and defining the drive parameters - still no luck. Would anybody out there know where I can find some info on Linux and drive mapping? Didn't I read somewhere that someone was working on a Drive-HOWTO ? __________________________________________________________________________ ARTHUR HARVEY ...To aimlessly meander where no man has aimlessly meandered before... __________________________________________________________________________ ------------------------------ From: gareth@gblinux.demon.co.uk (Gareth Bult) Subject: Re: Fails: mounting Linux filesystem fro Date: Wed, 27 Oct 1993 00:09:50 GMT The 'correct' format for /etc/exports is; (flags) for example; / remote(rw) to allow machine 'remote' to mount my root with read/write. Permission denied is often associates with /etc/hosts.equiv and /.rhosts. (or am I teaching my Granny...) Regards Gareth. ------------------------------ From: cas@muffin.apana.org.au (Craig Sanders) Subject: Re: lilo, vmlinuz zImage Date: 27 Oct 93 05:35:42 GMT pcg@aber.ac.uk (Piercarlo Grandi) writes: [...much deleted...] >One can not only put several alternative kernels ona floppy, have >different boot floppies, and so on. It is also useful to have another >floppy with a miniroot filesystem on it, for disaster recovery; to use >it one simply boots the usual floppy, but with a different root >directory (the floppy), and perhaps with a suitable RAM disk size. yep, this makes sense. I did something like this today in order to create a boot disk which had support for the qic-02 tape drive in the kernel so i could just boot from that floppy and restore my system from tape without having to go through the hassle of inserting 30 disks into the drive again. what i did was: 1. used dd to create an image of the SLS 'a1' install disk onto my hard disk. 2. used dd to copy that image back onto another floppy. (there's probably an easier way to do steps 1. & 2. but it seemed like the fastest way at the time) 3. mounted the floppy as /mnt 4. deleted a few un-necessary files (like do.install, install.info, compress was replaced with a sym-link to gzip etc) so there was enough room for my kernel. 5. copied my working kernel onto the floppy (I'm using a clean 0.99pl13 kernel as the SLS released has broken or non-existant support for my wangtek en5099-24 tape drive) 6. made sure that mt, tar, gzip, were on the disk 7. mknod /mnt/dev/tape c 12 4 8. cd /mnt/etc/lilo 9. export ROOT=/mnt 10. ./install 11. sync; cd / ; umount /mnt 12. used dd to make an image of this new 'tape restore' floppy so I can easily make another one anytime i want. and of course, reboot and test if it worked :-) I now have a boot floppy which I can use to bring linux back up even if the whole hd gets wiped clean. This is good, because I just spent the entire weekend installing linux - i had to reinstall from disk 'a1' three times before I got it right. I **NEVER** want to put those 30 disks into my drive again! All I have to do is make sure I have an up-to-date backup. >Having a boot floppy has at least two advantages, apart from flexibility >in choosing what/how to boot: if you have another OS, you need not >change its booting arrangements at all, and installing Linux can be >totally transparent; there is no problem with limitations of the disk >BIOS, e.g. wrt more than 1204 cylinders, as the boot is entirely from >the floppy, and once the Linux kernel is running, there is no problem. OK. this paragraph interests me. One of the reasons I had to reinstall several times over the weeked appeared to be that i was running into the old 1024 cylinder limit. I've got an old Priam 638 ESDI drive (1225 cyl * 15 heads * 36 sectors, approx 320MB). This drive worked very nicely under OS/2 with Boot Manager when I just had a 100MB partition for linux. however, I wanted to get rid of OS/2 and go to a pure linux system, so i backed up everything to tape, repartitioned, and re-installed linux. After numerous attempts, I finally got it working, but it wont boot off the hard disk. I'm not certain, but I think the problems were due to the fact that my main linux partition wasn't fully contained within the 1024 cylinder limit (there was some cryptic message from fdisk saying something about 'partition /dev/hda?? has different physical and logical endings'). i ended up partitioning it so that my partition table looks like: The number of cylinders for this disk is set to 1225. This is larger than 1024, and may cause problems with some software. Command (m for help): p Disk /dev/hda: 15 heads, 36 sectors, 1225 cylinders Units = cylinders of 540 * 512 bytes Device Boot Begin Start End Blocks Id System /dev/hda1 * 1 1 4 1062 a OPUS /dev/hda2 * 5 5 1023 275130 83 Linux extfs /dev/hda3 1024 1024 1225 54540 5 Extended /dev/hda5 1024 1024 1084 16469+ 82 Linux swap /dev/hda6 * 61 1085 1225 38069+ 6 DOS 16-bit >=32M I've kept the OS/2 Boot Manager (/dev/hda1) because I also want to boot from /dev/hda6 into MS-DOS occasionally. I guess I can do this with just lilo as well, so i'll probably get rid of it next time around. what happens now is that when I try to boot from the hard disk I get an error message saying "Partition table invalid" or something like that. I can, at least, boot from a floppy but I don't really like having to do this. what i'd like to do is to repartition the disk (AGAIN!!) and reformat and install so that it boots from the hard disk. I don't really care that much whether I've got a bootable dos partition or not (it would be nice, though). so, my questions (to you and to anyone else who cares to comment) are: - can i use lilo to do this? - what would you say was a good/optimal partitioning scheme? - what should i be particularly careful of when repartitioning, re-installing, etc? i'm glad i made that 'tape restore' disk now - I can experiment without risking having to put those 30 damned disks in again. -- Craig Sanders ======================THIS SPACE INTENTIONALLY LEFT BLANK====================== The Utility Muffin Research Kitchen muffin!cas@zikzak.apana.org.au ------------------------------ From: tyagi@birch.ee.vt.edu (Dhawal Tyagi) Subject: LAN Software for Linux Date: 27 Oct 1993 23:30:45 GMT Reply-To: tyagi@birch.ee.vt.edu Hi everybody! I'm planning to install Linux on my machine, but before that I'd like to make sure that I can do everything that I want to do. Is there any LAN software available for Linux like Microsoft LAN Manager for SCO Unix is available from SCO. Is it free or can I buy it from some company? Thanks, please e-mail if you can Dhawal. Home Address: Office Address: 424 North Main St., Apt #E 1700, Pratt Dr., #209 Blacksburg, VA 24060. Information Systems Building TEL#: (703)951-3979 Corporate Research Center Blacksburg, VA 24061 TEL#: (703)231-3905, 231-8398. ************************************************************************** Chand tasveer-e-butan, chand haseenon ke khutoot, Baad marne ke mere, Ghar se ye saaman nikla. -Mirza Asadullah Khan "Ghalib". ************************************************************************** ------------------------------ Crossposted-To: comp.os.linux.misc From: tim@stingray.micro.umn.edu (tim) Subject: Re: Chatting programs without net Date: Thu, 28 Oct 1993 02:08:50 GMT a few years ago i wrote a talk program that uses AF_UNIX sockets. it looks quite alot like the BSD talk, but you don't need network sockets. if there is interest, i'll ftp it out somewhere. tim ------------------------------ From: root@calvin.iaf.nl (SuperGirl) Subject: Re: syslogd hangs inetd Date: 27 Oct 1993 12:19:32 +0100 >My net-2 connection seems to hang when syslogd is running. >Apparently, I seem to be getting this error at random times ONLY when >I'm running syslogd. Any suggestions? >recvfrom unix: Bad file number >Currently, syslogd is off (even though inetd still routes through >tcpd)... although, I don't like the idea of leaving syslogd off. >Any suggestions? Hmm, this is no solution, you simple won't see them anymore :-( But i admit, i'm having them too, using an 13e kernel... ------------------------------ From: warb@faatcrl.faa.gov (Dan Warburton) Subject: Re: Help with routing under Linux Date: 28 Oct 93 02:17:52 GMT smithc@kelvin.physics.uq.oz.au (Craig Smith) writes: >: >>Now, the fun part. When I try using "routed". The routes change to: >: >Not surprising. With a bogus routing table, things only get worse if >: >you run 'routed'. Do you really need to run routed? How about just setting up a default route /etc/route add default gw xxx.xxx.xxx.xxx ? to a smart router or single route off your network. static routing works for me. -- Dan Warburton warb@faa.gov warb@tgf.tc.faa.gov warb@pilot.njin.net TGF (Target Generation Facility, an Air Traffic Control Simulator) FAA Technical Center, Atlantic City International Airport, NJ 08405 Mail Stop ACN-400c Phone: 609-485-4480 ------------------------------ From: hank@Blimp.automat.uni-essen.de (Hendrik G. Seliger) Subject: Bug in shadow-packages (patch included) Date: Thu, 28 Oct 1993 07:47:54 GMT Reply-To: hank@Blimp.automat.uni-essen.de Hi everyone! When trying to make the group administration of our linux machines easier (ie have other people administer the groups) I found a bug in the shadow package (actually, more than one): In gpasswd (gpmain.c): - if an group entry is changed, the name of the group has been written as passwd-entry, restricting this group; - the administrators field has not been scanned, not allowing more than one administrator. In libshadow.a (sgroupio.c): - the number of administrators has not been calculated (typical block-copy-problem), so that the number of members has been used to allocate memory for the administrators, giving a segmentation error if there're more admins. than members. I've sent the following patch to the author/editor of the shadow package, hoping he'll incorporate it in a new release. By the way, when making the library libshadow.a, the file sppwd.o is not included by the Makefile, which results in some programs not compiling. My Makefile has been so much tampered with that it's better not to distribute any diffs, but it should be no problem: make sppwd.o; ar rv libshadow.a sppwd.o; ranlib libshadow.a; rm sppwd.o So here's the diffs: ==================snippy============== *** gpmain.c.orig Fri Dec 4 13:13:33 1992 --- gpmain.c Wed Oct 27 18:12:18 1993 *************** *** 10,19 **** --- 10,23 ---- * * This software is provided on an AS-IS basis and the author makes * no warrantee of any kind. */ + /* + * Multi-admin repaired by HGS, Oct. 26, 1993 + */ + #include #include #include "pwd.h" #include "shadow.h" #include *************** *** 290,300 **** grent.gr_mem[i] = (char *) 0; #ifdef SHADOWGRP if (sg = getsgnam (group)) { sgent = *sg; sgent.sg_name = strdup (sg->sg_name); ! sgent.sg_passwd = strdup (sg->sg_name); for (i = 0;sg->sg_mem[i];i++) ; sgent.sg_mem = (char **) malloc (sizeof (char *) * (i + 1)); for (i = 0;sg->sg_mem[i];i++) --- 294,304 ---- grent.gr_mem[i] = (char *) 0; #ifdef SHADOWGRP if (sg = getsgnam (group)) { sgent = *sg; sgent.sg_name = strdup (sg->sg_name); ! sgent.sg_passwd = strdup (sg->sg_passwd); for (i = 0;sg->sg_mem[i];i++) ; sgent.sg_mem = (char **) malloc (sizeof (char *) * (i + 1)); for (i = 0;sg->sg_mem[i];i++) *************** *** 328,353 **** sg = &sgent; } #endif ! /* ! * The policy for changing a group is that 1) you must be root ! * or 2) you must be the first listed member of the group. The ! * first listed member of a group can do anything to that group ! * that the root user can. ! */ ! ! if (! amroot) { ! if (grent.gr_mem[0] == (char *) 0) ! goto failure; ! ! if (strcmp (grent.gr_mem[0], name) != 0) ! goto failure; ! } ! ! /* * Removing a password is straight forward. Just set the * password field to a "". */ if (rflg) { --- 332,356 ---- sg = &sgent; } #endif ! /* Changed by HGS ! * The policy for changing a group is that 1) you must be root ! * or 2) you must be in the list of administrators ! * who can do anything to that group ! * that the root user can. ! */ ! ! if (! amroot) { /*HGS*/ ! for(i=0; (sgent.sg_adm[i] != (char *) 0) && ! (strcmp(sgent.sg_adm[i], name) != 0); i++); ! if (sgent.sg_adm[i] == (char *) 0) ! goto failure; ! } ! ! /* * Removing a password is straight forward. Just set the * password field to a "". */ if (rflg) { *** sgroupio.c.orig Fri Dec 4 14:10:41 1992 --- sgroupio.c Wed Oct 27 18:10:49 1993 *************** *** 99,108 **** --- 99,111 ---- if (! (sgr->sg_mem[i] = strdup (sgrent->sg_mem[i]))) return 0; sgr->sg_mem[i] = 0; + for (i = 0;sgrent->sg_adm[i];i++) /*HGS*/ + ; + sgr->sg_adm = (char **) malloc (sizeof (char *) * (i + 1)); for (i = 0;sgrent->sg_adm[i];i++) if (! (sgr->sg_adm[i] = strdup (sgrent->sg_adm[i]))) return 0; =========snappy============================ Hank -- ====================================================================== Hendrik G. Seliger Universitaet Essen hank@Blimp.automat.uni-essen.de Schuetzenbahn 70 Tel.: +49-201-183-2898 45117 Essen, Germany ====================================================================== "Handling interrupts is simple." (G. Pajari) "Interrupts are an unpleasant fact of life." (A. Tanenbaum) -- Martin Preusser, Universitaet Essen GH, Tel.: 0201/183-2176 ------------------------------ ** 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-Admin-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.admin) via: Internet: Linux-Admin@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-Admin Digest ******************************