Meeting Pearls 1

home *** CD-ROM | disk | FTP | other *** search

/ Meeting Pearls 1 / Meeting Pearls Vol 1 (1994).iso / installed_progs / text / faqs / linux.howto.networking.part2 < prev next >

Wrap

Internet Message Format | 1994-04-26 | 27.4 KB

Subject: Linux NET-2 HOWTO (part 2/2) Newsgroups: comp.os.linux.announce,comp.os.linux.help,comp.os.linux.admin,news.answers,comp.answers From: terryd@extro.ucc.su.oz.au (Terry Dawson) Date: Mon, 25 Apr 1994 19:52:41 GMT Archive-name: linux/howto/networking/part2 Last-modified: 15 Apr 94 ---This is part 2/2--- After the user logs in, the login(1) program, if it finds and verifies the user ok, will execute the shell program 'dip-i' which will execute dip command in input mode (-i). Dip now scans the /etc/net/diphosts file for an entry for the given user name. Therefore, each slip user must also have an entry in /etc/net/diphosts. 5.6 /etc/net/diphosts /etc/net/diphosts is used by dip to lookup preset configurations for remote hosts. These remote hosts might be users dialing-into your linux machine, or they might be for machines that you dial into with your linux machine. The general format for /etc/net/diphosts is as follows: Suwalt::145.71.34.1:SLIP uwalt:CSLIP,1006 ^ ^ ^ ^ ^ ^ | | | | | \___ MTU | | | | \_________ protocol (SLIP, CSLIP, | | | | KISS, PPP) | | | \___________________ comment field ("gecos" :-) | | \________________________________ IP address of the other | | side, or host.domain.name | \___________________________________ unused (compat. with passwd) \________________________________________ login name (as returned by getpwuid(getuid()) ) An example /etc/net/diphosts entry for a remote slip user might be: Sfredm::145.71.34.1:SLIP uwalt:SLIP,296 which specifies a SLIP link with MTU==296, or Sfredm::145.71.34.1:SLIP uwalt:CSLIP,1006 which specifies a CSLIP-capable link with MTU of 1006. When a user logs in, they will receive a normal login, and password prompt, at which they should enter their slip-login userid and password. If they check out ok, then the user will see no special messages, they should just change into slip mode at their end, and then they should be able to connect ok, and be configured with the paramters from the diphosts file. 5.7 Configuring PLIP interfaces. PLIP is like SLIP, in that it is used for providing point to point IP links between machines, except that it is designed to use the Parallel ports on your machine instead of the serial ports. Because it is possible to transfer more than one bit at a time with the Parallel port, it is possible to attain higher speeds with the plip interface than with the serial interface. In addition, even the simplest of parallel ports, printer ports, can be used, in lieu of you having to purchase conmparatively expensive 16550AFN UARTs for your serial ports. When compiling the kernel, there is only one file that might need to be looked at. That file is net/drv/plip/global.h, and it contains timers in mS. The defaults are probably going to be fine, unless you have an especially slow computer, in which case you might have to increase them on the machine at the other end of the link. A sample configuration for a plip interface might be: #!/bin/sh # Portion of /etc/rc.d/rc.inet1 for PLIP connection to local machine IPADDR='192.148.64.1' # Replace with YOUR IP Address REMADDR='212.194.167.1' # Replace with the address of YOUR OTHER HOST ifconfig pl0 $IPADDR pointopoint $REMADDR up # Configure PLIP interface route add default gw $REMADDR # Route to other machine. # End The pointopoint parameter has the same meaning as for SLIP, in that it specifies the address of the machine at the other end of the link. In almost all respects, you can treat a plip interface as though it were a slip interface, except that neither 'dip' nor 'slattach' can be, or are used. 5.7.1 PLIP Cabling Diagram. PLIP has been designed to use cables with the same pinout as those commonly used by the better known of the dos based pc-pc file transfer programs. The pinout diagram (taken from net/drv/plip/README) looks as follows: Pin Name Connect pin - pin: -------------- ---------------------- GROUND 25 - 25 D0->ERROR 2 - 15 15 - 2 D1->SLCT 3 - 13 13 - 3 D2->PAPOUT 4 - 12 12 - 4 D3->ACK 5 - 10 10 - 5 D4->BUSY 6 - 11 11 - 6 D5 7* D6 8* D7 9* STROBE output 1* AUTOFD output 14* INIT output 16* SLCTIN output 17* Do not connect the pins marked with an asterisk (`*'). They are D5 (pin 7), D6 (pin 8) and D7 (pin 9). STROBE is pin 1, FEED is pin 14. Extra grounds are on pins 18, 19, 20, 21, 22, 23 and 24. If the cable you are using has a metallic shield it should be connected to the metallic DB-25 shell at one end only. 6. PPP (Under construction). There is now some ALPHA PPP software available. For the latest information relating to it, join the PPP channel on the niksula.hut.fi list server, and keep your eye on comp.os.linux.development. Already there have been some encouringly good reports for it. The PPP software comes in two parts. Some kernel modifications, and the ppp daemon. They are available at the following locations: ftp://ftp.gang.umass.edu/user/michael/linux-ppp-0.1.5.tgz ftp://ftp.gang.umass.edu/user/michael/pppd-0.1.4.tgz Please check that there isn't a later version there, and be sure to read any README files or the like that are there as well, as they will tell you how to install, where to report bugs and the like. 7. AX.25 (Under construction). Alan Cox has some experimental AX.25 code available for testing. 8. Are You Stuck ? Really ? Then you should read the man pages for ifconfig and route, included in the net-032 package, and understand their functions. These commands have a lot of flexibility, and because everyone's network setup is different, you may find a way to use ifconfig and route to get your connection working. If you do, feel free to send me some mail so I can include it in the next update of the NET-2 HOWTO. Because of my limited amount of experimental data, most of the discussion above is about my own setup, and I'd like to generalize it as much as possible. Matt is currently writing a set of scripts to simplify NET-2 configuration. You can pick up the pre-alpha release from tc.cornell.edu, in the file /pub/mdw/netconf-0.3.tar.z. These scripts maintain a small database of network configuration info, and allow you to easily modify and configure your network interface. The scripts are far from complete: Matt has been waiting until the NET-2 interface itself stabilizes a bit more before upgrading it further. Another good place to look for help on setting up NET-2 is the O'Reilly and Associated book 'TCP/IP Network Administration', the one with the crab on the cover. Keep in mind that NET-2 is now a "standard" implementation of TCP/IP, this means that ifconfig and route work the same under Linux as they do on other UNIX systems. Keep also in mind that some of the arguments and options may differ slightly from those in the book. You might also search out the following documents which are an excellent source of tutorial information on tcp/ip: athos.rutgers.edu:/runet -rw-r--r-- 1 0 0 176218 Oct 20 1989 tcp-ip-admin.doc -rw-r--r-- 1 0 0 214199 Oct 20 1989 tcp-ip-admin.ps -rw-r--r-- 1 0 0 92106 Oct 20 1989 tcp-ip-intro.doc -rw-r--r-- 1 0 0 111478 Oct 20 1989 tcp-ip-intro.ps Also keep in mind that NET-2 _is_ developing very rapidly, it's one of the newest additions to the Linux kernel. Thus, all of the bugs haven't been worked out yet, so there may be some problems. However, a good rule of thumb is that if you were able to get TCP/IP working under kernels before 0.99.pl10, you should be able to get it working under NET-2 as well. There are still some issues dealing with performance to be fixed, but overall the system works. And, as with everything in Linux development, time will cure what ails NET-2. If it's absolutely unusable to you, go back to an earlier kernel version, and wait until things develop further. The code is still fairly new. 9. Common Problems and Solutions. Now that the NET-2 HOWTO has been out for a while, I've been able to gather some common problems (and answers!). Here are some things which I have learned from hearing from readers. If you run into a problem which should be included here, please send it along (even if you have the solution!). QUESTION: How do I know what version of NET software I am running? ANSWER: In the kernel messages when you boot your machine, you should see a line that describes your networking code. For example, mine looks like: Linux version 0.99.14l+NET-2EB4 (root@albert.vk2ktj.ampr.org.) This line, not terribly obviously, tells you that I am running NET-2E, Beta 4. QUESTION: When I try to use the network, or use SLIP, I get the error message "Network not reachable". What should I do? ANSWER: This message means that a machine somewhere in the path did not have a route to the destination network. Until you can demonstrate otherwise, it is the courteous thing to do, assume it is your machine. This is usually an indication that either your ifconfig or route commands are in some way wrong. You can look at the status of your ifconfig by using the command "ifconfig" by itself. This should tell you what NET-2 thinks your IP address, netmask, etc. are. You can use the command "route" by itself to get routing information. This will tell you what routes you have set up and what gateways (if any). The best way to test a SLIP or network connection is to use "ping" with IP addresses only. If you use hostnames, as in "ping loomer", if some part of name lookup isn't working you'll have trouble. To test just the network, NOT name lookup, use only IP addresses, as in "ping 128.253.154.32". For SLIP connections the best thing to do is to ping your SLIP server. If nothing comes back, then something is wrong with your slip port configuration, double check all of the steps detailed above. Try using "dip -v" which will print debugging information while DIP is dialing the server. If you get a response from your slip server, but not from anywhere else, then you are probably missing your default route, you may need to use the commands: # route del <your slip server address> # route add default gw <your slip server address> to get SLIP talking to the server. Once you can talk to the server, everything SHOULD work (if your server is set up correctly!). For Ethernet connections, try pinging your gateway. If you can talk to your gateway, you should be able to talk to the outside world. You may need more than one route (that is, more than one gateway). For example, some universities use one gateway for on-campus networks and another for off-campus networks. Either way, try pinging addresses on your local network, and remote addresses. If you can ping all addresses ok remote to your network, and some on your local network, but not others on your local network, then check your netmask setting. If the "network not reachable" message means that you can't talk to your gateway. This can be due to several things: a) Wrong route or ifconfig commands b) Ethernet card problems (see below) c) You didn't compile the kernel correctly (see below). d) There is in fact some sort of network failure elsewhere. QUESTION: I keep getting the error "eth0: transmit timed out". What does this mean? ANSWER: This usually means that your Ethernet cable is unplugged, or that the setup parameters for your card (I/O address, IRQ, etc.) are not set correctly. Check the messages at boot time and make sure that your card is recognized with the correct Ethernet address. If it is, check that there is no conflict with any other hardware in your machine, eg you might have a soundblaster sharing the same IRQ or i/o control port. QUESTION: I get errors "check Ethernet cable" when using the network. ANSWER: You probably have your Ethernet card configured incorrectly. For Etherlink cards, in the file /usr/src/linux/driver/net/CONFIG, change the line EL_OPTS = -UEL2_AUI to EL_OPTS = -DEL2_AUI This tells the card to use the AUI cable interface. Just make sure that all of the options for your card are set correctly in the CONFIG file, and rebuild your kernel. QUESTION: When I use NET-2, I get a "General protection" error or a panic from the kernel. How can I fix this? ANSWER: Remember that the NET-2 code is still on the buggy side, just because it's in mid-development. If you get a kernel panic while using NET-2, write down the EIP address (and the other information given in the panic message). The EIP is the address where the kernel paniced, usually of the form 0008:xxxxxxxx where "0008" is the segment descriptor for the kernel text, and "xxxxxxxx" is the offset into that segment (80386 programmers will know what this means). Use the command nm /usr/src/linux/tools/system | sort -n or nm /usr/src/linux/tools/zSystem | sort -n depending on whether or not you use a compressed kernel (zImage). This will print a listing of all symbols in the kernel text, simply scan down the list and look for the function that contains the EIP address in the kernel dump. There's the culprit. However, in some cases the EIP can be misleading; the kernel may panic at a place which is complete irrelevant to where the actual problem occurred. However, it is a good starting place; first, locate the function which contains the EIP address, and then check out the kernel code to see what might be wrong. Keep in mind that this will only work if you compile your own kernel and have the "system" file associated with it. QUESTION: How can I hang up the phone line when I'm done using SLIP? ANSWER: If you use dip to dial out on the SLIP line, just "kill -9" the dip process itself (dip won't die unless you kill it with SIGKILL or some other signal). When dip dies, the line should hang up. DIPs behaviour is being modified so that it will be more sociable and die when it is supposed to. If you are using the new dip, then 'dip -k' will kill any copy of dip that you have running, and hang up the line as well. If you don't use dip to dial out, either instruct your dialing program to hang up the line, or kill the dialing process. QUESTION: dip doesn't work. How do I make it work ? ANSWER: Check that the file permissions of dip are 6750, that is 'chmod 6750 dip'. Check also that dip is owned by root: 'chown root:dip dip' QUESTION: With SLIP, I get a connection open, but no data flows. ANSWER: This could be a number of things. First, check your routes and be sure that the gateway is set correctly. Attempt to ping your gateway; if you can't, then something is wrong with the routes. Another problem could be that your system and the SLIP server disagree about header compression. With 0.99.pl11 and above, SLIP automatically compresses packet headers. To turn off header compression, check the SL_COMPRESS option in the CONFIG file. In pl14 there will be supplied a 'setencap' command to allow you to configure compression. QUESTION: With SLIP, I get a connection, but after sending a small amount of data, the connection hangs. ANSWER: Probably an MTU problem. The MTU is the maximum packet size available for the network. For SLIP, your MTU is set in your dip dialing script with the "MTU" command. The default value is 1500, which means that the system can send packets of up to 1500 bytes in size. However, some SLIP servers (Berkeley SLIP, for example), use a smaller MTU (around 1006). Another thing to check if you are having erratic SLIP problems is flow control. You need to use hardware (RTS/CTS) flow control on your modem, and your modem and your computer must agree. XON/XOFF flow control is not practical for SLIP. 9.1 Not so common problems and solutions (Mostly NFS). QUESTION: How do I use my existing Novell fileserver with my Linux machine ? ANSWER: If you have the Novell NFS Daemon code then it is easy, just NFS mount the Novell volume that you wish to use. If you don't, and you are really desperate to be able to do this, and you have a spare pc machine laying about, you are in luck. Here is what you do: You configure the spare machine as a normal Novell workstation, mapping the appropriate Fileserver directories to virtual drives as you so desire. You then grab a copy of SOSS (Son Of Stans own Server) from your nearest ftp site, and run it on the same workstation. SOSS is an NFS server that will happily run on just about any pc. This will allow you to NFS export the Novell network drives. It has caveats in that it will not perform as well as directly mounting the Novell fileserver, that it requires another machine, and that it will generate roughly twice as much network traffic, but it will work. Stan's Own Server (NFS server). spdcc.com:pub/sos/soss.zoo spdcc.com:pub/sos/sossexe.zoo A version "couple of bugs fixed: IP numbers and subdirectories with extensions)" is available from: hilbert.wharton.upenn.edu:/pub/tcpip/soss.zip QUESTION: Files get corrupted when using NFS over wider area networks or SLIP, why ? How do I stop it ? ANSWER: Certain vendors (Sun primarily) shipped many machines running NFS without UDP checksums. Great on ethernet, suicide otherwise. UDP checksums can be enabled on most file servers. Linux has it enabled by default from pl13 onwards - but both ends need to have it enabled... QUESTION: Why are my NFS files all read only ? ANSWER: The Linux NFS server defaults to read only. RTFM the 'exports' and nfsd manual pages. With non Linux servers you may also need to alter /etc/exports QUESTION: I mount from a linux nfs server and while ls works I can't read or write files. How do I fix this ? ANSWER: You must mount a Linux filestore with rsize=1024,wsize=1024 (or 2048 if you really want - 1024 is a better choice). QUESTION: I mount from a linux nfs server with a blocksize of between 3500-4000 and it crashes the Linux box regularly, why ? ANSWER: This is a known problem that is being worked on, refer to previous question. Don't you hate answers like that ? :) QUESTION: Can Linux do NFS over TCP ? ANSWER: No. To do this would require someone to spend the time to update the rpc code to add rpc stream record marking. It should work then. QUESTION: Why do I get loads of strange errors trying to mount a machine from a Linux box. ANSWER: This is possibly related to a restriction imposed by older NFS servers. Make sure your users are in 8 groups or less. QUESTION: Why are my Linux NFS clients very slow when writing to Sun & BSD systems ? ANSWER: NFS writes are normally synchronous, meaning that all file- -system changes occur in the order they transmitted, this means that if before NFS will allow you to write any more data, any previous write must have already completed, (you can disable this if you don't mind risking losing data). Worse still, BSD derived kernels, this includes Sun systems, tend to be unable to work in small blocks. Thus when you write 4K of data from a Linux box in the 1K packets it uses, BSD does this: read 4K page alter 1K write 4K back to physical disk read 4K page alter 1K write 4K page back to physical disk etc.. Better systems don't have this problem. The Linux client is however quite slow anyway. QUESTION: I've heard NFS is not secure is this true ? ANSWER: Yes, totally. Running NFS in an uncontrolled environment is rather like leaving your front door open, painting 'On holiday' on your house and posting maps to every known criminal... In a fairly secure environment or when you can recover data from stupid misuse its pretty much OK. The worst someone can easily do is alter all the files on an NFS mounted disk, and/or crash the machine. So long as you don't mount your system files writable you should be mostly safe. QUESTION: I occasionally mount from lots of different places, do I have to mount them all each time I boot ? ANSWER: No you can use the automounter to mount disks as you access them. QUESTION: How do I stop things hanging when a server goes down ? ANSWER: There are three main NFS behaviours: soft: Your NFS client will report an error to the process concerned if an NFS server doesn't answer after a few retries. Most software handles this well - but not all. hard: Your NFS client will try forever unless killed off. Operations will be restarted when the NFS server recovers or reboots. hard,intr: As hard but ^C will also stop the NFS retrying. In a few cases, notably nfs mounted /usr/spool/mail disks, this doesn't help as the shell will be ignoring ^C when it checks you have mail. If you intend to leave your machine unattended, then choosing the 'soft' option is probably best, because while it might cause some problems to an application running, it won't halt your whole machine if a server that it is attached to goes down. If your machine will always have a human operator available, then the 'hard,intr' option might be best. The hard option would be best suited to you if you can afford to wait, and don't want the process writing to the server interrupted at all. QUESTION: Can I use two slip interfaces ? ANSWER: Yes. If you have, for example, three machines which you would like to interconnect, then you most certainly could use two slip interfaces on one machine and connect each of the other machines to it. Simply configure the second interface as you did the first. NOTE that the second interface will require a different IP address to the first. You may need to play with the routing a bit to get it to do what you want, but it should work. 10. Known bugs. There are several known bugs with the NET-2 software. Note that these may or may not be fixed with a newer version of the NET-2 code; therefore, I leave them here. The bugs here are for NET-2d, found in kernels 0.99.pl10, pl11, and pl12, and pl13, and pl14. NET-2e (currently in Beta), when released, may or may not have fixed these bugs. * Bug with route guessing code. If you ifconfig the "lo" interface before the "eth0" interface in rc.inet1, whenever you add a route, it will be added to "lo" instead of "eth0". (Simply use the "route" command by itself; it will display all of your routes. If your "default" route, which should be out on the ethernet, is for device "lo" instead of "eth0", then you're seeing this bug.) This is just a problem with the route guessing code. Several things can fix it: 1) ifconfig/route on "eth0" before "lo" in rc.inet1; or, 2) Set your netmask to 255.0.0.0 (which is reported to work, but I can't guarantee it). This should be fixed in NET-2e. * Missing IP packet fragmentation. Packet fragmentation allows the various protocol layers to "chop up" packets into smaller packets if the MTU (maximum tranfer unit) of one network differs from another. NET-2e should contain packet fragmentation/defragmentation code, but NET-2d currently does not. This now only applies to kernel earlier than pl14+, as it is now supported. * Weak NFS support. There have been a number of success stories with NFS under Linux, however, not all of the support is there. For one thing, the current NFS buffer size is much smaller, and therefore much slower, than other implementations of NFS. 11. Copyright Message. (We're not ogres, nor are we silly). The NET-2-HOWTO is copyright by Terry Dawson and Matt Welsh. A verbatim copy of this document may be reproduced and distributed in any medium, physical or electronic without permission of the authors. Translations are similarly permitted without express permission if such translations include a notice stating who performed the translation, and that it is a translation. Commercial redistribution is allowed and encouraged, however, the authors would like to be notified of any such distributions. Short quotes may be used without prior consent by the authors. Derivative works and partial distributions of the NET-2-HOWTO must include either a verbatim copy of this file, or make a verbatim copy of this file available. If the latter is the case, a pointer to the verbatim copy must be stated at a clearly visible place. In short, we wish to promote dissemination of this information through as many channels as possible. However, we wish to retain copyright on this HOWTO document, and would like to be notified of any plans to redistribute it. Further we desire that ALL information provided in this HOWTO be disseminated. If you have any questions relating to the conditions of this copyright, please contact Matt Welsh, the Linux HOWTO coordinator, at: mdw@sunsite.unc.edu, or +1 607 256 7372. 12. Miscellaneous. I'm sure that I've missed something. This NET-2 HOWTO was thrown together with the help of Matt Welsh, and Jeff Uphoff. Other major contributors have been Alan Cox, Fred van Kempen, and others just like yourself. Hopefully it will help you, and others out there, get networking under Linux. Future plans for the NET-2 HOWTO include a section on setting up your own Linux LAN (with SLIP and/or Ethernet), adventures in routing, and the use of netstat and other network administration under Linux. For now, the information here should be more than enough. :) If you have questions about setting up NET-2, feel free to mail me, or if you have any corrections, additions, or errata for this NET-2 HOWTO, send me any and all changes (cdiffs are nice, but I'm flexible). Of course, thanks to Fred, Linus, Ross, Phil, Paul, Don, Alan, Matt, and everyone else who helped to develop the NET-2 code and work on previous versions of TCP/IP for Linux and the NET-FAQ. Finally, Linux has a complete implementation of TCP/IP. It may not be for everyone yet. But for those who have an itch they want to scratch, happy hacking, here it is. Cheers, Terry Dawson, (terryd@extro.ucc.su.oz.au) 13. Change History. Changes from 1.8: correction to broadcast address calculation, thanks Andr'as Salamon tcp/ip tutorials added thanks to Gilbert Callaghan These annotations at the suggestion of Andy Burgess Shadow password section updated - thanks Rick Sladkey added Slip Server section - thanks Fred added /etc/net/diphosts section - thanks Fred enhanced the netmask description a little Revamped for 0.99.14 Added Index Changes from 1.9: Added change history. Corrected Archive header now that I understand what it is there for Thanks to _everyone_ who helped me understand :) Ammended loopback route details - thanks Jeffrey A. Kintscher. First attempt at enlarging the configuration section to cope with different networks and different distributions thanks Eric Christensen. Reinstated /dev/arp as a required device. Oops. Finally added resolv+(8) man page reference. Tried to clean the slip section a bit. Added leased line/cable slip link config using slattach. Corrected a minor PLIP stoopidity I inflicted that fortunately noone appears to have noticed. Ammended Slip Server config to run a script in lieu of 'dip -i' Fixed numerous tyops and mizpellinks (When will I not ?)