Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.mira.net.au!news.netspace.net.au!news.mel.connect.com.au!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!insync!news.azstarnet.com!news.sprintlink.net!news-ana-24.sprintlink.net!neonramp.com!cynjut.neonramp.com!cynjut.neonramp.com!not-for-mail From: burgess@cynjut.neonramp.com (Dave Burgess) Newsgroups: comp.unix.bsd.netbsd.announce,comp.unix.bsd.freebsd.announce,comp.answers,news.answers,comp.unix.openbsd.announce Subject: [comp.unix.bsd] NetBSD, FreeBSD, and OpenBSD FAQ (Part 8 of 10) Supersedes: <386bsd-faq-8-859467259@cynjut.neonramp.com> Followup-To: comp.unix.bsd.netbsd.misc Date: 13 Apr 1997 02:00:28 -0500 Organization: Dave's House in Omaha Lines: 855 Approved: news-answers-request@MIT.Edu,cgd@sun-lamp.cs.berkeley.edu Expires: 05/01/97 02:00:04 CDT Message-ID: <386bsd-faq-8-860914804@cynjut.neonramp.com> References: <386bsd-faq-1-860914804@cynjut.neonramp.com> Reply-To: burgess@cynjut.neonramp.com (BSD FAQ Maintainer) NNTP-Posting-Host: cynjut.neonramp.com Keywords: FAQ 386bsd NetBSD FreeBSD !Linux X-Posting-Frequency: Posted on/about the 13th and the 27th of every month. Xref: euryale.cc.adfa.oz.au comp.unix.bsd.netbsd.announce:409 comp.unix.bsd.freebsd.announce:530 comp.answers:23275 news.answers:11057 Posted-By: auto-faq 3.1.1.2 Archive-name: 386bsd-faq/part8 Section 7. (System Communication and Network Information) 7.0 Communications 7.1 SLIP/CSLIP Serial Line I/P is supported in all versions of PC BSDs. Brian <brian@awfulhak.demon.co.uk> provides us with a rather good explanation of some of the hurdles that must be overcome for a working slip interface. The idea is (overview) that you make a serial line connection to the host, set the line discipline, and tell your router to use this interface as your gateway. You also should set the gateway up as a nameserver. You will need the information in 7.4.1 below to make sense to you before you proceed much further. I suggest you read that now. Sounds easy ? - well it is if you've done it before. The _usual_ way of doing this is as follows: Both server and client must know each others inet addresses. Set these up in /etc/hosts with lines saying 11.22.33.44 host.my.domain.name host 11.22.33.55 client.my.domain.name client where 11.22.33.?? is your inet number, and the following name is the full machine name (and is followed by any number of aliases). SERVER: Create a login - usually Sclientname - and run `sliplogin` as its shell. I've looked at the docs for sliplogin, and it seems fairly straightforward. [Ed.Note - I have; it is.] A fairly common problem on the server is an error that is caused by the lack of a 'sliplogin' entry in the /etc/shells file. Be sure to add sliplogin to your shells file. CLIENT: Set up /etc/resolv.conf to say the following (for the nameserver) domain client.my.domain.name nameserver 11.22.33.55 ** traditional method ** - Log on to the server. This is usually done via kermit or some such program. - Exit the program (or background it if your line wants to drop once the device is closed). - Run `slattach /dev/comport` for whatever "comport" is. On most BSD derived systems, this may be either com0, or cua01, or whatever the correct name is for your site. If you run into an error that says 'not configured' in it, your kernel either does not recognize your port (dmesg will verify that) or your kernel does not have the slip interface built in. See below for the latter. The former could be caused by any one of a dozen problems, including conflicting or incorrectly identified IRQs or port addresses. - Run `ifconfig sl0 net clientname servername netmask 0xffffff00` - Run `route add default servername`. "servername" is your server and "clientname" is your client. It should now be possible to `ping host` ** my method ** Configure /etc/remote Configure /etc/host.dial Run `slip host`. /etc/remote contains an extended `tip` entry. /etc/host.dial contains a login script (and is named in /etc/remote). Oh yes, don't forget to have a line in your kernel config saying pseudo-device sl 2 Without this line, you may get a 'device not configured' or 'TIO...' error because the slip driver is not built into the kernel. Someone else mailed me their instructions for using a SLIP service. Here they are, in all their glory. Hi, I thought I'd drop you this email outlining how I have SLIP set up to dial and communicate, as I remember this being an area in the FAQ which needed some expansion/clarification. What I outline works for me dialing up under NetBSD 0.9. Though I don't know the specific nuances of FreeBSD (e.g. the boot-up configuration of the interfaces - /etc/hostname.sl<n> for NetBSD) this'll be easy for someone to add to, hopefully before you release it (I know there's nothing I hate more than having to convert one setup's info into another nearly, but not quite, the same setup :-) In the last quoted script file (slip-off) the unlock line should read: /usr/local/etc/unlock LCK..$DEVICE 1) Configuring the SLIP interface. Ensure that the sl device is present in your kernel (default with the generic kernels). In NetBSD 0.9 they get assigned in turn with each 'slattach'. Put your own hostname and ip number, and that of your dial up gateway, into your /etc/hosts. This is an example:- 127.0.0.1 localhost 158.152.1.65 gate gate.demon.co.uk 158.152.26.67 blodwen blodwen.demon.co.uk Ensure that your /etc/resolv.conf is set up appropriately, to point to the nameserver of your dial-up provider/link. This is what I use:- domain demon.co.uk nameserver 158.152.1.65 nameserver 158.152.1.193 (you can have up to three nameservers, they -must- be listed by number. If you wish to look in several domains, you can use 'search demon.co.uk,foo.other.domain' etc. up to the limit (a finite length specified in resolver(5).) Also, for more flexibility, there is a nameserver switch package that allows you to change the resolver profiles on the fly; see below. Your sl interface needs to be configured using ifconfig(1) as a link from your own hostname to that of your dial-up host. Manually this can be done by:- /sbin/ifconfig sl0 <your-machine> <dial-up-machine> on NetBSD this can be done at boot-time, by having the following in /etc/hostname.sl0:- inet blodwen.demon.co.uk 255.255.255.0 dest gate.demon.co.uk (You can substitute sl0 for sl<n> if this will after another slattach e.g. for a local machine on a fixed cable) The netmask value (255.255.255.0 in this case) is pretty irrelevant to SLIP because you cannot have a subnet broadcast (so I am informed). I use the chat(1) program (currently available in the FreeBSD-current/ports/ directory) to dial up and enter passwords, etc. My modem is setup for hardware handshaking (a necessity really, for performance), and I use the following sh scripts to do the work. Calling 'demon' dials up and connects. Calling 'demon-down' when on-line shuts it all off. Those who use PPP should be able to work out the changes from the original ppp-on and off scripts. [I call it 'demon' for simplicity] #!/bin/sh # # attach slip and route (calls slip-on script) if /usr/local/etc/slip-on then # this adds the default route to 'gate' which is # my dial-up host route add default gate # put anything here to be run when you are connected # slurp news /usr/local/etc/slurp news.demon.co.uk & # send outgoing news /usr/local/news/bin/nntpsend # kick outgoing email sendmail -q0m else # slip-on failed fi [ here's my /usr/local/etc/slip-on ] Note that you may need to adjust the chat command to deal with the prompts you have to answer or that your modem produces, and the final message (my provider sends HELLO to signify that they are ready. The -v to chat makes it send syslog .info messages, which I then send to my /dev/console. You can remove this if you wish. The following is a modified version of the ppp-on script that comes with chat, altered to set the serial line correctly for the modem. I dial-up at 38400bps on a modem on tty03, you might want to change that too (and make sure that the stty line is all kept on one line). # # slip-on # # Set up a SLIP link # LOCKDIR=/var/spool/lock DEVICE=tty03 PHONE=<phone number here> USER=<your login> PASSWORD=<your password> if [ -f $LOCKDIR/LCK..$DEVICE ] then echo "SLIP device is locked" exit 1 fi /usr/local/etc/fix-cua $DEVICE sleep 16000 < /dev/$DEVICE & /bin/stty -f /dev/$DEVICE gfmt1:cflag=4b00:iflag=c00:lflag=3:oflag=6:discard=f:dsusp=19:eof=4: eol=ff:eol2=ff:erase=0:intr=3:kill=0:lnext=16:quit=1c:reprint=12: start=11:status=ff:stop=13:susp=1a:werase=17:ispeed=38400:ospeed=38400 ( if chat -v -l LCK..$DEVICE ABORT "NO DIALTONE" \ ABORT "NO CARRIER" ABORT BUSY "" ATZ OK ATDT$PHONE "CONNECT 38400" "" "ogin:" "$USER" \ "word:" "\\q$PASSWORD" "HELLO" then /sbin/slattach -h -c -s 38400 $DEVICE exit 0 else echo "SLIP call failed" 1>&2 # remove the sleep holding serial line open /bin/kill -KILL `/bin/ps -ax | /usr/bin/egrep " sleep 16000" \ | /usr/bin/egrep -v "egrep" | /usr/bin/sed 's/^\([ 0-9]*\) .*/\1'/` exit 1 fi ) < /dev/$DEVICE > /dev/$DEVICE When it comes to switching off the line, I use the following. Don't forget to adjust the sl0 as appropriate [ I call it demon-down for simplicity] #!/bin/sh # stop script # /usr/local/etc/slip-off /sbin/ifconfig sl0 down [ and /usr/local/etc/slip-off ] and adjust the DEVICE line as well. #!/bin/sh DEVICE=tty03 kill -KILL `ps -ax | egrep "slattach.*${DEVICE}" | egrep -v "egrep" \ | sed 's/^\([ 0- 9]*\) .*/\1'/` kill -KILL `ps -ax | egrep " sleep 16000" | egrep -v "egrep" \ | sed 's/^\([ 0-9]* \) .*/\1'/` # switch line back to normal (closes line) /bin/stty -f /dev/$DEVICE -clocal # unlock the device /usr/local/etc/unlock LCK..$DEVICE exit 0 And that should do it. Happy SLIPping! 7.2 PPP Implementations of Point to Point Protocol are also available. PPP has been available since the 0.9 release of NetBSD and is in the current releases of FreeBSD, OpenBSD, and NetBSD. It should also be noted that there is a newsgroup that covers the PPP protocol exclusively. It is comp.protocols.ppp. Here is some information for people desiring to set up PPP in there systems: A simple way to do this is to use the "chat" and a chat file. I use the following command to initiate a connection : root# pppd tty01 19200 connect 'chat -v -f chat.my-isp' And in the chat.my-isp file: ABORT BUSY ABORT ERROR ABORT 'NO DIALTONE' ABORT 'NO CARRIER' '' ATZ OK ATDT1234567 CONNECT \d TIMEOUT 5 ogin:\s--ogin:\s mylogin ssword: mypasswd prompt:\s /usr/lib/ppp/ppp This dials, connects and negotiates the addresses from just one line entered. To kill the connection: root # kill `cat /var/run/ppp0.pid` which has the added advantage of hanging up the phone if the modem is set up appropriately. The biggest problem that I ever had with this was working out the chat script and that was debugged by adding the following line in /etc/syslog.conf: # Hand chat debug messages to root local2.debug root The PPP.FAQ was helpful, but I ignored quite a bit of it and depended more on the online manuals. For setting up the PPP daemon, here is some more information: For NetBSD, it turned out that I needed to rebuild the kernel with the following line in my config file, pseudo-device ppp 1 This line adds a device driver to the kernel that does the ppp protocol. Once I had that built in, everything worked the first time. This is the kind of sequence I go through to start ppp: 1. Connect with kermit to my ppp account and login. The account tells me when it starts ppp. 2. "Suspend" kermit (i. e. put it in the background). This gets me back to the shell prompt. (You can get kermit back using the "fg" command) 3. Start "pppd". When the shell prompt returns, I then have Internet access! That's it. This procedure will get you access to machines by using their IP address numbers. You still have to configure a name server in "resolv.conf" in order to get DNS functionality. My resolv.conf looks like this: domain umd.edu # Maryland's domain name nameserver 128.8.5.2 # These are the IP addresses of three nameserver 128.8.126.2 # hosts that can act as name servers nameserver 128.8.126.3 The name servers should be as "close" as possible. Whatever machine manages the modem pool your on would be the best but any machine on your local loop will give you good performance. If your Internet Service Provider uses dynamic addressing, You don't even have to worry about this. It's the point of PPP. It's actually a good thing from a security point of view. Your IP address changes w.r.t. to the rest of the net periodically. By the way, don't forget passwords on all your accounts!! When your on PPP, the rest of the net can see you too, you look like a full Internet host. It is important to look into the following to see if any of them apply to you, if you still have questions: Here is a sample PPP config. 1) Make /etc/ppp directory, then populate as follows: 2) Include the following in '/etc/ppp/options': crtscts modem /* This option opens the port with O_NONBLOCK if there is a connect options specified, and resets CLOCAL when the connection is closed */ 19200 defaultroute netmask 255.255.255.0 ipcp-accept-local ipcp-accept-remote noipdefault connect "/usr/sbin/chat -v -f /etc/ppp/sample.chat" 3) Make sure the modem line (in the '/etc/ttys' file) doesn't have the 'local' or 'softcar' options included. With 'local;, CLOCAL will be set for that line and SIGHUP may or may not be sent, apparently based on the age of the software. The "modem" option in the 'options' file (above) should clear that, but it may or may not ever work. If you have "softcar" in /etc/ttys, then the SIGHUP (in fact, almost all of the signals) will never work because the modem lines are effectively ignored, thereby leaving the modem in whatever mode it is in. 4) Include the following in '/etc/ppp/sample.chat': ABORT BUSY ABORT 'NO CARRIER' '' atdt5551212 rname: {sample} sword: crack-me annex: {whatever} PPP. This setup uses LCP and IPCP (parts of PPP) to negotiate the dynamic IP addresses. This setup assumes an ISP which uses an annex terminal server. It prompts for "Annex username:" and "Annex password:". You then get to the command line prompt: "annex:", at which point "PPP" starts the PPP session. You will have to edit this to suit. If your ISP uses a system where you are automatically connected to PPP when you log in, your '/etc/ppp/sample.chat' file might look like this: ABORT BUSY ABORT 'NO CARRIER' '' atdt5551212 rname: {sample} sword: crackme To implement a 'permanent PPP' dial-up connection, the following has been used (by me specifically!) This works in NetBSD 1.1 or higher, OpenBSD, and perhaps recent versions of FreeBSD: The following line in /etc/ttys works wonders for making a permanent link: tty01 "/usr/sbin/pppd" unknown on rtcsts The file '/etc/ppp/options' looks like: /dev/tty01 115200 connect "/usr/sbin/chat '' 'atdt1' 'ogin:' 'x' 'sword:' 'x'" crtscts defaultroute lock netmask 255.255.255.0 n.n.n.n:n.n.n.n -ip modem mtu 1500 -detach You will, of course, have to make some changes if you have multiple ppp connections. The key here is the '-detach' to make the pppd remain connected to the controlling terminal (the modem). The basic idea is if the link drops (i.e. loses carrier), a hangup signal will be sent to pppd, causing it to exit, and init will restart it. You can also try 'demand dialed PPP' by getting the iij-ppp package from the following: ftp://ftp.iij.ad.jp/pub/network/iij-ppp0.94beta2.tar.gz) It supports BSDI's BSD/386 1.1, FreeBSD-2.0, and NetBSD-1.0, but it should be really easy to make it work with NetBSD-1.1 (it comes with a patched NetBSD-1.0 tun driver... to get it working with 1.1 or -current you will need to make the same patches to NetBSD-1.1's tun driver). You can also try 'dp' at: ftp://phoenix.acn.purdue.edu/dp/dp-4.0.tar.gz 7.3 TCP/IP TCP/IP is an integral part of BSD 4.4 Lite. There are at least five different network card drivers. TCP/IP is fully supported and is available to all users of all derived BSD systems. In fact, many people believe that this area is one of the primary advantages that BSD has over Linux. After all, TCP/IP was invented in BSD. 7.4 UUCP There is an excellent document included in the UUCP directory that describes in detail, what needs to be done to get a working UUCP for derived BSD systems. Look in the /usr/src/gnu/libexec/uucp directory for more information. You can also look in the /usr/share/doc/smm/09.uucpimpl and /usr/share/doc/smm/21.uucpnet if yours are populated. 7.4.1 TIP/CU First thing you need to do is... vi /etc/remote Then remove the two lines at the bottom of the file that mention com1, and com2. Now add the following lines: tty00:dv=/dev/tty00:br#9600: tty01:dv=/dev/tty01:br#9600: That tells tip/cu where to find your com ports. Next you need to be logged in as root and do a: chown uucp.dialer /dev/tty00 chown uucp.dialer /dev/tty01 touch /var/log/aculog chown uucp.dialer /var/log/aculog Make sure that, if you are running newsyslog, you change the owner.group entry in the newsyslog.conf file so that the file ownership is maintained correctly. Then you should be all set, remember "DOS Com1" = tty00, and "DOS Com2" = tty01. So, if your modem is at 0x2F8/IRQ=3 and you access it as the COM2: port from DOS, you would do.. tip tty01 To exit, type <RETURN>~.<RETURN> Many people have other problems with cu. The lock open: procedure is one of them. If you receive the error: lock open: no such file or directory all ports busy You need to create a directory: /var/spool/lock, owned by uucp. If this file already exists and is owned correctly, make sure that the lock file in the directory is deleted. If you receive the error "cu: write: Input/output error" You may be able to fix this by creating an /etc/uucp/ports file (see Taylor UUCP docs). In addition, those of you using cu version 1.04 may need to add the following to their susyem: create an /etc/uucp/ports file that looked like this: port mymodem type modem device /dev/tty01 speed 19200 Now cu knows that the line is connected to a modem it does the right thing regarding setting CLOCAL on the line. You don't even have to have either of local or softcar set in /etc/ttys. Since cu's behavior seems to be correct, I'm happy now. All I need to really make my day though is to have John or Martin to tell me that cu 1.04 still works for them even though they don't have an /etc/uucp/ports (or equivalent HDB or V2 uucp config) file ... :-) 7.4.2 What is the magic incantation that allows the modem to dial? Try 'stty -f /dev/tty0? clocal'. Change the '?' for whatever character is appropriate for your tty port. Remember, DOS COM1 = /dev/tty00 and DOS COM2 = /dev/tty01... Some other things that might cause some problems are the entries in the /etc/remotes file. Try 'com?:dv=/dev/tty0?:br#19200:pa=none' and see if that helps. Remember to replace the '?' with '[01234]' as appropriate. NetBSD-current has implemented the 'ttyflags' program. This will set your com ports according to the options specified in the /dev/ttys files. This is an even better solution than the 'stty ... clocal' fix from above! FreeBSD sets this up a little bit differently by having separate dial in and dial out devices available. The /dev/cua?? devices all have clocal set by default to allow the system to dial out without having a carrier present. If you are using FreeBSD and don't have any cua devices in the /dev/ directory, you need to run the ./MAKEDEV script. See the man page for more information. 7.4.3 My modem on DOS COM3 or DOS COM4 works with DOS, but not with *BSD. It is set up using IRQ 4 (or 3) respectively. One of the MAJOR differences between DOS and *BSD is that *BSD actually USES the IRQ lines (*gasp*)... That means that every device on the ISA bus has to have it's own IRQ. Since COM1 and COM2 (/dev/tty00 and /dev/tty01) are already defined using IRQs 4 and 3 respectively, that means that COM3 and COM4 (/dev/tty02 and /dev/tty03) need to be put onto different IRQs. The default GENERICAHA kernel defines the third com port (COM3 or /dev/tty02) to be on IRQ5. You need to reconfigure your com port (for external modems) or modem (for internal modems) to use IRQ5. The GENERICBBT kernel defines the COM4 port to be on IRQ9 (or 2). If you have to put your devices on other ports, you will need to recompile the kernel. 7.5 How do I configure my nameserver? There are several systems that implement /etc/nsswitch.conf instead of the /etc/resolver.conf database. Each has advantages and disadvantages, and both have been implemented for NetBSD. If you want to use nsswitch, you can get it at 'ftp://?/?'. 7.6 Terminals Since the target machine for most BSD machines is a 386 with no more than a couple of serial ports, most people do not bother with serial terminals. For most problems, a quick perusal of the man pages for the ttys file and getty are enough to get them started. Other than that, most terminal problems are limited to peculiarities of particular terminals. One common problem that appears to crop up from time to time is which wires need to be connected at each end of the cable. Most cables do not, in fact, pass through all lines. If your terminal uses XON/XOFF (DC1/DC3) protocol, a cable of the appropriate twist, either straight through or null modem, can have as few as three lines connecting the two devices. Assuming DB-25 connections at each end, the lines need to go from 2 to 3, 3 to 2, and 7 to 7. These lines are Rx, Tx, and gnd. Other lines that may or may not be required include 4 and 5; and 6, 8, and 20. Normally, these lines would be connected within the 'hood' of the cable (4 to 5 and 6 and 8 to 20) to simulate the functionality of the full blown cable. While full support for CTS/RTS is not available (yet), other support for the remainder of these lines is available or is being worked on in all BSD derived systems. Without this handshaking (particularly pins 6, 8, and 20) your ports may appear to be dead. This is because most of the tty driver for *BSD systems require a Data Carrier Detect to be active before the port will work. For those folks that have hardware flow control working, you need to look in the man page for 'stty' and look around for the -clocal and -ctrcrts options. Once the cable is set up, you will need to make sure that your system is ready. The first thing you will need to do is make all of the devices in the /dev/ directory. A program, called MAKEDEV, is available in the /dev directory. Running this program with the argument 'tty' will make all of the physical tty devices. With that done, arranging for a 'getty' on the port is the next order of business. You will need to edit the '/etc/ttys' file and make one of the tty devices available. If you have connected your terminal to DOS COM1, you will be enabling /dev/tty00. Similarly, if you are connected to COM2:, you will be enabling /dev/tty01 (see the pattern?). There are other names for those ports as well, but when you are talking about terminals, be sure to use the '/dev/tty*' names. If not, you will be completely ignored and treated as an outcast because you obviously have not done any of your homework. One of the other common problems with the SIO driver is that people will often disable all handshaking, and then complain that they cannot get a reliable connection above 9600 baud. Handshaking is the solution to most of these problems. 7.7 My network manager (or UUCP feed site admin) just informed me that the way I have installed sendmail through my UUCP connection and has caused a sendmail loop. Can you help me get sendmail installed correctly? (1) Go into sendmail's source directory tree cd /usr/src/usr.sbin/sendmail/cf/cf (2) Make the missing obj directory first, you need it later... mkdir obj (3) Create a sendmail master configuration file (.mc file). Name it yourname.mc vi yourname.mc (4) Contents of the yourname.mc file: #--------------------------------------------------------------- divert(-1) # # This is the prototype for a site with only a uucp connection # to the world, where smarthost and uucp relay are the same ... # Replace "yourname" with your machines nodename without domain # Replace "smarthost" with your uucp neighbours nodename without # domain i.e. here is myname "knobel" and my smarthost is "gomel", # to which I'm connected with uucp via dialup modem. divert(-1) VERSIONID(`yourname.mc 1.0') include(`../m4/cf.m4') OSTYPE(bsd4.4) FEATURE(nodns)dnl MAILER(local)dnl MAILER(smtp)dnl MAILER(uucp)dnl define(`UUCP_MAX_SIZE', 2000000)dnl define(`SMART_HOST', `uucp-dom:smarthost')dnl define(`UUCP_RELAY', `uucp-dom:smarthost')dnl #-------------------------------------------------------------- In the siteconfig directory (.../sendmail/cf/siteconfig) create a file uucp.yourname Put a list of machines into this file to which you have active uucp/mail connection. Generally only the name of your smarthost .... Unknown addresses are routed to your smarthost .... siteconfig/uucp.yourname: #---------------------------------------------------------------------- SITE(nodename_of_your_smarthost) #---------------------------------------------------------------------- (5) create the new sendmail configuration file, which will be stored under obj/yourname.cf, by typing make yourname.cf (6) After that copy obj/yourname.cf to /etc/sendmail.cf (7) It's up to you to browse through the systems global aliases file ((etc/aliases), where important mail aliases are stored. After editing this file you should invoke the command newaliases to update the corresponding database file newaliases (8) Then create/edit the file "/etc/sendmail.cw". This file contains alias names of your system (a list of additional names under that your system might receive e-mail): yourname yourname.uucp yourname.domain (9) Then create a file /etc/mailertable: Here you have to say what else (uucp addresses, too) has to be sent to your smarthost ... .uucp uucp-uudom:name_of_your_uucp_smarthost (10) Create the hash table the following way: makemap hash /etc/mailertable.db < /etc/mailertable Remember, if you make any changes you have to rebuild the aliases database by typing: newaliases (11) BTW: You do not need to create a frozen config file, since sendmail on FreeBSD 1.X and NetBSD aren't compiled with that option turned on. (12) ``Hot files'' with more information (see sendmail src tree): FAQ KNOWNBUGS RELEASE_NOTES cf/README 7.8 Can network attached assets be used by/from NetBSD? FreeBSD? OpenBSD? Yes, they can, assuming the machine at the other end of the connections is reasonably cooperative. The specifics are up to the remote machine, but a couple of things that you can start looking for that will help are provided below: - Ask the system administrator of the machine in question if it is OK for you to use whatever it is you need. This is more a matter of manners than a technical issue. - For NFS mounted disk drives, make sure that you are not prevented from using the assets by the /etc/exports (or equivalent) file. This goes for CD-ROMs as well as regular mounted disks. - There are a completely different set of concerns for tapes and printers. Each system implements these in slightly different ways. Check with your system manager or documentation for more information. - Diskless booting is possible for all three systems, for a little more detail, see below. Note that not all network clients are created equal. There may be semantic differences between what you EXPECT to happen and what actually happens. Your best bet at that point is to get with the local system manager and talk to him or her about what you should be expecting on the system and what is actually happening. An excellent example is the semantics of file group accounts when a new file is created on an NFS machine. The semantics of the create will be based on the OS on the SERVER, so it will be whatever SysV or Sun thinks is correct, not what we expect from the BSD side. There is a package available which can also be used by *BSD which will allow your machine to be visible to LanManager or Windows NT clients. The package is called 'SAMBA' and includes information about how to configure the package to work with NetBSD, OpenBSD, and FreeBSD. Works good for me. 7.8.1 Is it possible to Network boot a NetBSD machine from a network on a diskless Sparc? Yes, it's even more than possible, it actually works! Since OpenBSD tracks NetBSD closely, it is also possible to do this with OpenBSD. Anders Magnusson (ragge@ludd.luth.se) has run it on diskless SLCs, and the only problem they had was when the machine got heavily loaded it ran out of mbufs (also sometimes a problem for regular systems). It is reportedly faster than SunOS4 as long there was lots of free memory in the machine. The fix for the 'running out of mbufs' problem is covered in section .... For the most part, setting up a diskless client is fairly straightforward. The FreeBSD diskless FAQ gives step by step instructions for setting up bootp and the other programs that need to be configured. 7.8.2 I have been working with FreeBSD 1.5.1 with some machines configured as diskless. How can I do the same for 2.0R (i.e., Which are the magic words to put in the Kernel configuration file?) In FreeBSD, there is a file called Diskless.FAQ in the usr/share/FAQ directory. It describes the procedures for diskless booting of an i386 on FreeBSD. Please note that netboot.[cr]om programs from FreeBSD 1.1.5.1 do not work with 2.0 and vice versa. Note that this is just a pointer to the Diskless FAQ. You can get the file from ftp.freebsd.org. -- Dave Burgess (The man of a thousand E-Mail addresses) *bsd FAQ Maintainer / SysAdmin for the NetBSD system in my spare bedroom "Just because something is stupid doesn't mean there isn't someone that doesn't want to do it...."