From: Digestifier To: Linux-Admin@senator-bedfellow.mit.edu Reply-To: Linux-Admin@senator-bedfellow.mit.edu Date: Fri, 22 Oct 93 21:13:50 EDT Subject: Linux-Admin Digest #122 Linux-Admin Digest #122, Volume #1 Fri, 22 Oct 93 21:13:50 EDT Contents: high speed modem problems with getty2.07b (cliff@trifid.astro.ucla.edu) Re: [Q] libc-4.4.4 incompatible with X 1.2 static libs? (Hal N. Brooks) Re: Has any one ported spice3e2 to linux? (Mike Engelhardt) Has anyone done Gopher on Linux? (Chris Thompson) Re: Linux with 2 Ethernet Cards (Alex Liu) Re: a lost+found magic! (Remy CARD) Re: [Q] libc-4.4.4 incompatible with X 1.2 static libs? (Hal N. Brooks) Re: Has anyone done Gopher on Linux? Re: UNIX sysadmin FAQ- proposal and volunt (Paul Vojta) HELP! ext2 filesystem partly corrupted - e2fsck useless (Mark Kassab) Re: UNIX sysadmin FAQ- proposal and volunteer (Karl Nicholas) Re: UNIX sysadmin FAQ- proposal and volunt (Karl Nicholas) Re: UNIX sysadmin FAQ- proposal and volunt (Karl Nicholas) ---------------------------------------------------------------------------- From: cliff@trifid.astro.ucla.edu Crossposted-To: comp.os.linux Subject: high speed modem problems with getty2.07b Date: Mon, 18 Oct 1993 11:55:35 I've been using 0.99pl10 with getty 2.07b now for a couple of months and I've noticed problems with my v.32bis (14,400 baud) modems. Sometimes an incoming connection will be negotiated successfully and I get a login prompt from my linux box, but anything I try to type seems to be completely ignored. Other times, I can log in but the connection suddenly slows to a crawl. This only seems to happen with v.32bis connections, v.32 usually works just fine. I have two 14,400 baud modems on Com's 1 and 2. The inittab entry sets them up for 38400 baud connections using getty. The gettytab file has entries for the modems starting at 38400baud with hardware flow control enabled. The modems have the following settings &c1&d2&k3&q5&s0&r0s0=2. Tried re-compiling the source for getty/uugetty 2.07b but the executables look exactly like the one I already have. Has anyone else experienced these problems? Please e-mail your replies, I don't get to read this news group very often... cliff cliff@trifid.astro.ucla.edu ------------------------------ From: hal@pollux.cs.uga.edu (Hal N. Brooks) Subject: Re: [Q] libc-4.4.4 incompatible with X 1.2 static libs? Date: 22 Oct 93 19:08:55 GMT Reply-To: hal@pollux.cs.uga.edu (Hal N. Brooks) This is a followup to my own post. This is looking less like an administration related problem, so it's probably the wrong group now, but look at this: {~/ss/qbe}: gdb qbe (gdb) list 97 98 int 99 main(argc, argv) 100 int argc; 101 char *argv[]; 102 { 103 Panel panel; 104 105 /* mtrace();*/ 106 server = xv_init(XV_INIT_ARGC_PTR_ARGV, &argc, argv, NULL); (gdb) run Starting program: /usr/home/hal/ss/qbe/qbe Program received signal 11, Segmentation fault 0xf74b9 in strncpy (s1=0x0, s2=0xbfffe47e "iablm", n=6) at strncpy.c:38 strncpy.c:38: No such file or directory. (gdb) bt #0 0xf74b9 in strncpy (s1=0x0, s2=0xbfffe47e "iablm", n=6) at strncpy.c:38 #1 0xdcd31 in _XGetHostname () I can't believe that XGetHostname() could *ever* have gotten away with calling strncpy() with a NULL pointer for the destination argument. Furthermore, the host name is "hiablm", not "iablm". And why doesn't it dump core when linked with shared libraries? On another note, many people have mentioned Seyon breaking with libc-4.4.4. I never had any problems with it dumping core, but I've found that if a dialing script has a "waitfor something" command, it's fine if the "something" shows up, but it *never* times out if the "something" doesn't arrive. I've tried getting rid of all blank line in all seyon configuration files and turning funMessages off, but it doesn't seem to get around this problem. ====================================================================== Hal N. Brooks Voice: (706) 546-7792 Internet: hal@cs.uga.edu ====================================================================== ------------------------------ From: engel@netcom.com (Mike Engelhardt) Subject: Re: Has any one ported spice3e2 to linux? Date: Fri, 22 Oct 1993 17:18:00 GMT I wrote: > ...But it turns out that the source itself of spice3e2 targeted by > (my) patch was never supposed to be ftp'able. The attention received > by the port(...) got the file spice3e2.tar.Z pulled from *all* anonymous > ftp sites. > (...) > If anyone wants me to port spice3f4 for them let me know. mirons@icarus.ci.net (Michael A. Irons) wirtes: > If spice3f4 is fair game but spice3e2 isn't then this sounds > odd. Is f4 newer then e2?? No, spice3f4 is not fair game for uncontrolled distribution/export. But If you have the source, which at least US citizens can obtain for signing some agreements and a distribution fee, then I would be happy to port it for them. I just couldn't make it freely distributable. And yes, if you have the source, you could legally distribute to me under some simple conditions we would have to arrange. I simply haven't obtained 3f4 because over the years I've purchased three legal copies of Spice for personal use and don't feel I need the very latest because I'm not simulating circuits but charged-particle optics today. Hope this helps to clairify, --Mike Engelhardt {engel@netcom.com,engel@[163.185.50.62]} ------------------------------ From: thompsch@ccmail.orst.edu (Chris Thompson) Subject: Has anyone done Gopher on Linux? Date: 22 Oct 93 19:32:07 GMT Just interested in hearing if anyone has done or is working on a Linux port of Gopher. If this is a dumb question please point me in the right direction. If you are working on it or have it going please drop me a note. Thanks Chris ------------------------------ From: aliu@phakt.usc.edu (Alex Liu) Subject: Re: Linux with 2 Ethernet Cards Date: 22 Oct 93 19:48:27 GMT botelle@jrb3.xrt.upenn.edu (Justin Botelle) writes: >Will Linux support 2 Ethernet Cards in the same machine? Yes, this is possible (I have done it) You must first make sure that the network cards don't conflict in any IRQ, I/O or memory address. Once you have done that, you have to look into the kernel, and modify the file: linux/net/inet/Space.c There are several entries for the network devices. Usually have names like eth0, eth1, eth2, eth3, ... etc this are in a device structure. Usually defaults are eth0 autoprobes and eth1-3 are disabled. What I did was let eth0 still autoprobe, but also added the I/O addr, IRQ, shmem of my network card on eth1. Recompile and presto, you get the 2nd ethercard as eth1. Note, that the comments say that you can add the network cards using "boot-time" setup, but I found no docs on this respect. Also, if you have kernel ver0.99p13 or better, it says that you can also set to autoprobe on the other eth devices. I am just listing what worked for me... > Will it route >between the two nets the cards are attached to? I am trying to do this and unfortunately with no luck. >Will it accomplish this without hacking and recompiling the kernel? You must recompile the kernel to get Linux to recognize the 2nd network card. About geting routing to work, who knows.... >Justin Botelle >botelle@jrb3.xrt.upenn.edu >standard disclaimers apply >__ -- _____________________________________________________________________________ Alejandro Liu |EMail: aliu@usc.edu |All mispellings are intentional 424N Electric Ave. #C |Voice: 818-293-8696 |Anything mentioned here is not Alhambra,CA 91801 |Data: 818-293-XXXX |necessarily true. ------------------------------ From: card@masi.ibp.fr (Remy CARD) Subject: Re: a lost+found magic! Date: 22 Oct 1993 20:16:31 GMT In article , Jim Dodd wrote: ] ] I would recommend you create lost+found directories on _ALL_ of your ] partitions. this can be done by the following script: ] ] mkdir /lost+found ] i=100 ] while [ i -gt 0 ] ] do ] touch tmp$i ] i=`expr $i + 1 ` ] done ] rm tmp* ] ] You do need to do the touch part to insure that there are enough ] slots already in the directory, so e2fsck doesn't have to try ] and allocate additional blocks for these directory entries when ] the free block list might be messed up. Even better, use the program mklost+found contained in the ext2 fs programs to recreate the lost+found directory. Please, note that e2fsck is the only file system checker which uses lost+found in Linux. Checkers for other file systems (e.g. minix fs, ext fs, xia fs) do not use it. So, you don't need to create the directory lost+found if you use other file systems than ext2. ] ] Hope this helps. ] ] ] -- ] Jim Dodd email: jimd@netcom.com -- Remy Card card@masi.ibp.fr ------------------------------ From: hal@pollux.cs.uga.edu (Hal N. Brooks) Subject: Re: [Q] libc-4.4.4 incompatible with X 1.2 static libs? Date: 22 Oct 93 21:46:54 GMT Reply-To: hal@pollux.cs.uga.edu (Hal N. Brooks) I said earlier: >This is a followup to my own post. This is looking less like >an administration related problem, so it's probably the wrong >group now, but look at this: > >{~/ss/qbe}: gdb qbe >(gdb) list >97 >98 int >99 main(argc, argv) >100 int argc; >101 char *argv[]; >102 { >103 Panel panel; >104 >105 /* mtrace();*/ >106 server = xv_init(XV_INIT_ARGC_PTR_ARGV, &argc, argv, NULL); >(gdb) run >Starting program: /usr/home/hal/ss/qbe/qbe > >Program received signal 11, Segmentation fault >0xf74b9 in strncpy (s1=0x0, s2=0xbfffe47e "iablm", n=6) at strncpy.c:38 >strncpy.c:38: No such file or directory. >(gdb) bt >#0 0xf74b9 in strncpy (s1=0x0, s2=0xbfffe47e "iablm", n=6) at strncpy.c:38 >#1 0xdcd31 in _XGetHostname () > >I can't believe that XGetHostname() could *ever* have gotten away >with calling strncpy() with a NULL pointer for the destination >argument. Furthermore, the host name is "hiablm", not "iablm". >And why doesn't it dump core when linked with shared libraries? I just came across a couple of followup articles posted by Warner Losh to c.o.l.help (which I guess is where this should have been) which answers my question. Don't use XFree86 1.2 with libc-4.4.4. I reprint below (and without permission) his posts which coincidentally relate to my problem: >Newsgroups: comp.os.linux.help >From: imp@boulder.parcplace.com (Warner Losh) >Subject: Re: Statically linked ObjectBuilder bins >Date: Thu, 21 Oct 1993 16:07:50 GMT > >In article <2a2gqs$c6f@hermes.acs.ryerson.ca> ylam@acs.ryerson.ca >(York Lam - ACPS/F93) writes: >> I'm trying to make statically linked binaries from source generated by >>Object Builder. Upon execution of this binary, it ended up dumping core >>at a seg. fault. I don't know it's a problem with one of the libraries or >>with g++. >> gcc 2.4.5 >> libX11.so.3.0 >> libc.so.4.4.1 > >Hmmm. I think I know what the problem is. The binary releases of lbc >haven't been completely compatible when statically linking. The >format of stat changed after 4.3.3 so that it had extra info in the >stat structure. In _XGetHostname, there is a call to stat which >should copy the information into the parameter passed to it. However, >since the utsname structure is on the stack and the library is now >writing more data to it, you lose because the pointer to the buffer is >now NULL and the _XGetHostname tries to copy to that NULL pointer. >The shared libraries get around this by having the jumpslot call >__stat rather than stat. This is a known problem with XFree 1.2 >combined wit libc 4.4.1. I've heard reports that you can use XFree >1.3 and all is well. I know that I have at home. > >Warner >-- >Warner Losh imp@boulder.parcplace.COM ParcPlace Boulder >I've almost finished my brute force solution to subtlety. > and a few minutes later, Warner made a minor clarification: >Newsgroups: comp.os.linux.help >Subject: Re: Statically linked ObjectBuilder bins >From: imp@boulder.parcplace.com (Warner Losh) > >In article >imp@boulder.parcplace.com (Warner Losh) writes: >>format of stat changed after 4.3.3 so that it had extra info in the >>stat structure. > >What I ment to say here was > >"The format of utsname change after 4.3.3 so that there was extra info > in the utsname structure." > >Warner >-- >Warner Losh imp@boulder.parcplace.COM ParcPlace Boulder >I've almost finished my brute force solution to subtlety. ------------------------------ From: blake@real.nevada.edu () Subject: Re: Has anyone done Gopher on Linux? Date: Fri, 22 Oct 1993 22:10:25 GMT Chris Thompson (thompsch@ccmail.orst.edu) wrote: : Just interested in hearing if anyone has done or is working on a Linux port : of Gopher. If this is a dumb question please point me in the right : direction. If you are working on it or have it going please drop me a note. : Thanks Chris -- ftp to boombox.micro.umn.edu:pub/gopher/Unix/ and get the latest source (2.09?). untar it, edit conf.h and Makefile.config, then type "make all;make install" Thats all. ========= This is my .sig. If you don't like it, I have others. blake@nevada.edu ------------------------------ From: vojta@powdermilk.berkeley.edu (Paul Vojta) Crossposted-To: comp.admin.policy Subject: Re: UNIX sysadmin FAQ- proposal and volunt Date: 22 Oct 1993 23:10:13 GMT In article , Bert Medley wrote: >In article dtm@Ebay.Sun.COM (Duane T. Mun) writes: [...] >>The one big problem I see is if the users that are nfs mounting >>/var/spool/mail have root access to their machine. They can add a new >>user with a uid of another user (who is also nfs mounting the same >>/var/spool/mail), and do what they like to their mail. >But if they have root access, why bother with the rigamarole? Just touch the >mail. Don't you need root access to add a user? Yes, but the default is that exportfs denies root access to the client machines. (By "touch" I assume you mean read/copy/edit/delete, as opposed to touch(1).) --Paul Vojta, vojta@math.berkeley.edu ------------------------------ Crossposted-To: comp.os.linux.help From: mark@macs.ee.mcgill.ca (Mark Kassab) Subject: HELP! ext2 filesystem partly corrupted - e2fsck useless Reply-To: mark@macs.ee.mcgill.ca Date: Fri, 22 Oct 1993 23:49:17 GMT I don't know where to start. My system had been up but idle for several hours when the power went out. That was during the night, so the power came back and the system rebooted. I decided to run e2fsck (which admittedly I hadn't run for over a month). I have just one Linux partition, so since I obviously couldn't umount it before fsck'ing (I didn't know yet about the mounting read-only and remounting trick), I ran fsck on a non-quiescent filesystem. I don't know if that caused the start of my problems. First, I trusted fsck and ran it with the "-a" option. Next thing I knew, my /tmp and everything in it were gone, my /etc/fstab contained garbage, and who knows what other files were corrupted. After a while of fiddling and doing e2fsck right, I managed to get rid of some problems. However, weird things were happening. Running fsck would sometimes transform my /etc/fstab to a directory! Everything under /lost+found is troublesome: Directories (including /lost+found itself) would suddenly end up with one link instead of 2 or more, and I can't go into them anymore. I tried unlinking /lost+found, creating another, and running fsck to clean up. No chance - the new /lost+found would soon afterwards be corrupted. I was getting errors like: /lost+found: bad directory: size < 24 I was also getting errors like: Block 15527 has been used before. Now in file '/etc/mount'. Clear? I haven't a clue of what's going on. I can understand some inodes having incorrect information (which fsck should correct I presume), but I can't understand why a newly-created lost+found would suddenly end up with 1 or 0 links (at least that's what I could see from 'ls'). Any ideas what's going on? How to get rid of this problem? Are there any tools besides fsck that could help? I would hate to think I have to reformat and reinstall everything! I don't know for sure what files are corrupted. (I presume it's only what was in /tmp + some files in /etc). Note that I'm running SLS 1.03 with 0.99pl12. Thanks in advance for any help. Mark mark@macs.ee.mcgill.ca ------------------------------ Crossposted-To: comp.unix.admin,comp.sys.sun.admin,comp.sys.sgi.admin,comp.admin.policy From: karl@bofa.com (Karl Nicholas) Subject: Re: UNIX sysadmin FAQ- proposal and volunteer Date: Fri, 22 Oct 1993 23:00:00 GMT In article dhesi@rahul.net (Rahul Dhesi) writes: >In <1993Oct18.064938.11806@leland.Stanford.EDU> >mosedale@aeffle.Stanford.EDU (Dan Mosedale) writes: > >>- Mounting /var/spool/mail via NFS >> - why & why not > >This is easy to write. > > - Mounting /var/spool/mail via NFS > - don't And why not. Please do not make broad statements like this without some argument. As far as I'm concerned, you MUST mount it NFS. This is the only way I could figure out that users will actually get their mail and still be logged onto any given machine within the domain. I've NFS mounted my /var/spool/mail dir from the very first day I ever got a sun, never had any troubles. Thats what the sun sysadmin manuals say to do!!!! -- +---------------------------------------------------------------------------+ | Karl Nicholas (510) 675-3686 | Beer is good food, food is not beer. | | karl@BofA.COM uunet!odinba!karl | *** BofA is not me, I am not BofA *** | +---------------------------------------------------------------------------+ ------------------------------ Crossposted-To: comp.unix.admin,comp.sys.sun.admin,comp.sys.sgi.admin,comp.admin.policy From: karl@bofa.com (Karl Nicholas) Subject: Re: UNIX sysadmin FAQ- proposal and volunt Date: Fri, 22 Oct 1993 23:09:02 GMT In article dtm@Ebay.Sun.COM (Duane T. Mun) writes: >"Ronald" == Ronald Hello writes: > >Ronald> In article 161b@thebes.cc.waikato.ac.nz, hamish@thebes.cc.waikato.ac.nz (Hamish Marson) writes: >rh>Rahul Dhesi (dhesi@rahul.net) wrote: >rh>> In <1993Oct18.064938.11806@leland.Stanford.EDU> >rh>> mosedale@aeffle.Stanford.EDU (Dan Mosedale) writes: >rh> >rh>> >- Mounting /var/spool/mail via NFS >rh>> > - why & why not >rh> >rh>> This is easy to write. >rh> >rh>> - Mounting /var/spool/mail via NFS >rh>> - don't >rh> >Ronald> about? > >The one big problem I see is if the users that are nfs mounting >/var/spool/mail have root access to their machine. They can add a new >user with a uid of another user (who is also nfs mounting the same >/var/spool/mail), and do what they like to their mail. > This is not true. root access for NFS mounted partitions must be explicitly allowed by the NFS server. -root=machine ... -- +---------------------------------------------------------------------------+ | Karl Nicholas (510) 675-3686 | Beer is good food, food is not beer. | | karl@BofA.COM uunet!odinba!karl | *** BofA is not me, I am not BofA *** | +---------------------------------------------------------------------------+ ------------------------------ Crossposted-To: comp.unix.admin,comp.sys.sun.admin,comp.sys.sgi.admin,comp.admin.policy From: karl@bofa.com (Karl Nicholas) Subject: Re: UNIX sysadmin FAQ- proposal and volunt Date: Fri, 22 Oct 1993 23:15:40 GMT In article <1993Oct20.025007.11821@ntg.com> dplatt@ntg.com (Dave Platt) writes: >>rh>> - Mounting /var/spool/mail via NFS > >[1] All of the NFS-client machines send _all_ "local" mail to the NFS > server, for delivery by the server's sendmail/smail daemon. [This > avoids the need to give the client machines trusted-root access to > the spool directory, and also avoids sendmail/sendmail collisions > during delivery.] > > Current versions of SunOS sendmail (and Solaris too, I assume) have > this feature built in... regardless of what the local-delivery > ruleset may say, SunOS 4.1.x sendmail does SMTP delivery to the NFS > server if it realizes that it is an NFS-client of /var/spool/mail. > Quite true. It can however be disabled for those of you who want the mailhost to mount the /var/spool/mail dir NFS for whatever strange reason you may have (I won't go into our strange reason :). There is a line # Remote mode - send through server if mailbox directory is mounted OR Thats does what is described above. For subsidiary machines then add the mailhost (or even mailhost maybe) after the OR. Comment the line out on the mailhost. Should work. -- +---------------------------------------------------------------------------+ | Karl Nicholas (510) 675-3686 | Beer is good food, food is not beer. | | karl@BofA.COM uunet!odinba!karl | *** BofA is not me, I am not BofA *** | +---------------------------------------------------------------------------+ ------------------------------ ** 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 ******************************