[Comp.Sci.Dept, Utrecht] Note from archiver<at>cs.uu.nl: This page is part of a big collection of Usenet postings, archived here for your convenience. For matters concerning the content of this page, please contact its author(s); use the source, if all else fails. For matters concerning the archive as a whole, please refer to the archive description or contact the archiver.

Subject: [comp.unix.bsd] NetBSD, FreeBSD, and OpenBSD FAQ (Part 5 of 10)

This article was archived around: 13 Oct 1997 02:00:17 -0500

All FAQs in Directory: 386bsd-faq
All FAQs posted in: comp.unix.bsd.netbsd.announce, comp.unix.bsd.freebsd.announce, comp.unix.openbsd.announce
Source: Usenet Version

Posted-By: auto-faq Archive-name: 386bsd-faq/part5
Section 4. (System Additions) 4.0 Introduction If you have written some addition to the kernel or some other part of the system, or know of one that feel should be mentioned, send mail to Dave Burgess (burgess@cynjut.neonramp.com) with all the relevant information, and it will be added for the next release. 4.1 Common (sort of) Kernel-related problems 4.1.1 Sometimes I have trouble with my system resetting the terminal to seven bit mode. Isn't BSD eight bit clean? The answer is "sort of". The problem seems to come from the fact that the <sgtty.h> interface is not guaranteed to be eight bit clean. The <termios.h> interface is better, and should be eight bit clean in all cases. If you find an application that uses the <sgtty.h> interface, you should either contact the author and try and get them to use the termios interface or port the code yourself. See section 5 for more Terminfo/Termlib information, as well as a discussion of the new curses library that is available. 4.1.2 How do you implement quotas on Net/2 derived BSD systems? From: tinguely@plains.NoDak.edu (Mark Tinguely) maybe you did not complete the setup, here is a step-by-step instructions to get them to work: 1) make a kernel with "options QUOTA" installed 2) edit /etc/fstab and include the kinds of quotas you want, below I used "userquota", you could also add "groupquota". /dev/wd0h /usr ufs rw,userquota 1 2 3) for each filesystem that is in /etc/fstab that uses quota, create the file "quota.user" (and "quota.group if appropriate). Above I have user quotas in the /usr filesystem, so I would: # touch /usr/quota.user 4) scan filesystem for files ownership (and/or group ownership). # quotacheck -a 5) now you can add individual quota limits, if you want to add the same quotas to the many people, then make a template and replicate the template. If they change for each user, then edit seperately. # edquota tinguely (an editor is kicked up and says something like: Quotas for user tinguely: /usr: blocks in use: 11876, limits (soft = 0, hard = 0) inodes in use: 891, limits (soft = 0, hard = 0) a limit of 0 means "unlimited". Change these to the appropriate number of blocks. A soft limit generates a warning, and can be exceed for period of time (7 days?), after which time a soft limit is treated like a hard limit. A hard limit denies new writes. to replicate a template (for this example let us assume "tinguely" is the template): # edquota -p tinguely user1 user2 user3 ... userN 6) turn quotas on (usually done in the /etc/rc file, but turn it on manually so you do not have to reboot right now: # quotaon that should take care of setting up quotas. You can look at the status of use of files with repquota, the -a option lists all filesystems with quotas. 4.1.3 What are the correct permissions for the /tmp, /usr/tmp, and /var/tmp directories? All of these directories should be owned by bin, group bin, mode 1777. This turns on the sticky bit, so that the only people who can remove a file from these directories are the owner and root. 4.2 Available kernel add-ons 4.2.1 Loadable Kernel Modules Several strides have been made in the past to reduce the amount of 'cruft' that gets into the default kernel. One way is to make the kernel so hard to use that practically no one but a person with precisely the 'right' hardware would be able to use the system Another way is to implement something called 'LKM's or "Loadable Kernel Modules". These are run-time extensions to the system that allow the distribution kernel to not include things that people might want, but not nxbeed until they get the system up and running. While the security concerns of LKMs are valid, their implementation is such a win that the research to implement them is well worth it. It was really _very_ simple to make these, so this is nothing spectacular. Just something to keep from having to recompile just to add msdosfs support to a machine. ;) To try this: 1) get ftp://ftp.flame.org/pub/netbsd/lkm.tar.gz 2) untar it somewhere. It will create a subdirectory called lkm and all extracted files will go in it. (I use /usr/src, but that may be a bad place) 3) follow the directions in lkm/README Please mail suggestions, and (especially) fixes and more modules to Michael Graff <explorer@flame.org>. Once it is clean enough, I'll send it in as a send-pr and see what happens. :) One question which still needs to be resolved is where should these *.o LKM's be installed? The directory '/usr/lkm' would be a good idea, with the output (modload's -o option) in /var/run/lkm or something like that. 4.3 Other program building type problems. 4.3.1 I am building a program that requires access to the crypt library. Either I have it and it isn't getting copied into the executable, or I don't have it; why? This is actually two separate questions, but they are close enough to the same that I can answer them here. The first problem that anyone building a 'crypt' aware program needs to remember is that the crypt library is a separate library and requires a '-lcrypt' to be added at the end of the link line. The other half of the problem is the 'US Non Export' policy for DES encryption. There are several good sources (about one per country) for non-US crypt libraries. IF you are outside the US and need one, look around on some of the NetBSD/FreeBSD/OpenBSD FTP sites in the 'local area'. By the way. I don't have any good URLs for Mars, so you might be out of luck. OpenBSD doesn't appear to have this problem, since it is a "Canadian" product rather than an American one. Thanks to this, there is no restriction on exporting (or importing) the crypt library, so it is no longer needed. With version 2.1 of OpenBSD, the crypt library doesn't even exist; it is included in the standard library for the system. 4.3.2 I am having trouble with long file names in my libraries. It seems like there is a 16 character limit in the library somewhere. There is a 16 character limit, sort of. The most likely symptom for this is that the header for the file _after_ the long file name will be mangled. It turns out that there is a "T" option that may not be documented very well that provides the correct functionality for long filename support in ar. 4.3.3 I'm getting annoyed with having this "conflicting types for `sys_errlist'" problem show up nearly every time I build a program. What do I need to do? Remove the sys_errlist reference in the source you're compiling. You can either delete it (there are advantages to just deleting it) or you can wrap a "#ifdef __NetBSD__/#endif" (obviously only if you are running NetBSD, FreeBSD and OpenBSD have a similar mechanism) pair around it. There are religious issues regarding the use of sys_errlist that involve either system security (most declaration allow the error list to be written to) or system internals (there's already a well-defined library call that performs the sys_errlist lookup). An anonymous example is included below: Most stupid packages such as GCC expects extern (char*)sys_errlist[] whereas 4.4Lite based systems have more secure extern const (char*) const sys_errlist[] declaration. Just kick that "cccp.c" in the butt and modify the suspicious line. Hard to believe GCC still doesn't do that. You're going to have to do lots of this modification as you encounter more of such programs. 4.4 System Administration Questions 4.4.1 Where can I get good books about NetBSD or FreeBSD? There is a set of books produced by O'Reilly and associates that describe in some detail the 4.4 BSD system. The six volume set includes a book on system administration which directly pertains to the operation and management of NetBSD and FreeBSD. Also see the Section 1 for a good list of the books that folks use for the system. There is also a good list of books (specifically about writing device drivers) in the 'pcvt' distributions in NetBSD, FreeBSD, and OpenBSD. It is in a file called 'Bibliography' and contains the pcvt author's list of device driver books. 4.4.2 I am concerned about system security. What should I do to protect my system from net attacks? With the release of the System Administrators Tool for Analyzing Networks (SATAN), network security has suddenly become a serious issue. There are a few things you can do. -- Get, read, and understand the CERT advisories -- Get SATAN and run it against your own system or network. Fix whatever it finds as holes -- Get courtney, a program that was written to recognize a SATAN attack pattern and notify you whenever someone tries to probe your system -- Log all failed login attempts (see below) 4.4.3 How can I log failed login attempts? Failed logins are logged (without the attempted login name) at LOG_NOTICE priority. Failed logins are logged _with_ the attempted login name at LOG_NOTICE priority, and with the LOG_AUTHPRIV facility. If you set up some lines in syslog.conf like: # The authpriv log file should be restricted access; # these messages shouldn't go to terminals or publically-readable files. authpriv.* /var/log/secure Make absolutely sure, though, that it's really what you want: logging actual supplied logins is often a great way to offer cleartext passwords to an adversary... Which is why you have authpriv.* /var/log/secure ...,authpriv.none,... /var/log/messages So none of the authpriv messages (those that actually display the failed login) goto /var/log/messages, but they do go to /var/log/secure (which you have with 600 perms.) Bear in mind that this still does not prevent someone that has hacked into your system with root privs from reading them. See 4.4.2 for more information. 4.4.4 Can I use a Concatenated Filesystem with NetBSD? The "ccd" device (in -current) provides the capability to span a file system across multiple hard drive partitions. Jason Thorpe <thorpej@nas.nasa.gov> has been working on it; if you try it and have problems, here are the debug instructions: Why, when I type "ccdconfig ccd0 16 none /dev/wd0a > /dev/wd1a", do I get back "ccdconfig: ioctl (CCDIOCSET): /dev/ccd0d: Device not configured"? Considering that the error comes froom the ioctl (rather than the open) I'm tempted to say it comes from either the vn_open() or subsequent VOP_*() operations on the components. If you compile your kernel with `options CCDDEBUG' and set the ccddebug variable (near the top of ccd.c or with the ddb) to 0x03, you should be able to see where it fails. If you could send me that information, that would be most helpful. Might be the same problem I had; it turns out that the partitions that you build your concatenated disk device from must not be marked "unused" in their native disks' labels. This "device not configured" is the way ccdconfig informs you of this condition... :-) Actually, I guess this indicates a need for a special "ccd component" type entry for disklabel? Or should the partition simply be marked as a "raw" partition, sharing this type with database log partitions etc? 'Der Mouse' (mouse@collatz.mcrcim.mcgill.edu) adds: Personally, I think ccd has no business looking at those partition types. But I definitely think a special ccd-component partition type is _not_ the way to go; if nothing else, it makes life hard for people running ports using non-NetBSD disk partitions. For example, under NetBSD/sparc on a disk with a SunOS label, there are no partition types in the label, so it would be impossible to use a ccd that insisted on a special partition type on such a disk. 4.4.5 I am really new to Unix System Administration. I need some real basic help. What is the System Administrator's user name? I can't log in as 'su'. What does that message mean when I log in as root. Both of these indicate a newness to Unix System Administration that many of the core team members don't even remember. The sysadmin user-id is "root", although you typically don't want to log in directly as root. A better solution is to log in as root once, create a user-id and password for yourself. Once you are done with that, you need to modify the /etc/groups file. This identifies the users that are allowed to be part of particular groups. Add your UID to the "wheel" group and log off. With a real UID and password and your UID identified as a "wheel" member. you will be able to use the 'su' command to log in as yourself and then "switch users" to root. That, by the way, is also what that cryptic message "Don't log in as 'root', use 'su' instead." means. Are there any books I can 'bootstrap' myself with? Yes. Here are a couple: (1) Nemeth, Snyder, and Seebass, "Unix System Administration i Handbook" (2) Horspool, "The Berkeley Unix Environment" How about some code examples? ftp://ftp.sterling.com:/usenet/alt.sources/ ftp://ftp.sterling.com:/usenet/comp.sources/misc ftp://ftp.sterling.com:/usenet/comp.sources.unix ftp://wuarchive.wustl.edu:/usenet/alt.sources/articles/ /usenet/comp.sources.misc/ /usenet/comp.sources.unix/ ftp://src.doc.ic.ac.uk:/usenet/alt.sources/articles/ /usenet/comp.sources.misc/ /usenet/comp.sources.unix/ 4.5.6 How do I change the default shell for a user? There are three ways, listed here from most difficult to least difficult: 1) Use 'vi' to edit the /etc/master.passwd. Once you have changed the entry you want modified, run the program "pwd_mkdb -p /etc/master.passwd". This will rebuild the password database and update your /etc/passwd file. Cd to the /var/yp directory and run a "make" to update your 'NIS' database (important only if you are using NIS). 2) Use the 'vipw' program and make your change. This automatically rebuilds the password database and the /etc/passwd file. You still need to update the 'NIS' database on your own. 3) Use 'chpass', 'chfn', or 'chsh' programs to update the appropriate entry in the password database. These have the advantage of update the NIS stuff automatically. 4.5 Daemon questions 4.5.1 I'd like to use amd to mount a file system (/dev/sd0f aka /usr/local) on another machine as "/usr/local". What's the magic? There are several ways to achieve 'amd nirvana'. Each of these elements below is an important consideration for getting amd to work correctly. The "-" means use these as defaults, so you need an entry without a "-". Also, I think one "-..." overrides the previous one completely. As a start, you can use the following in your amd.project file: usr/local opts:=rw;type:=nfs;rhost:=hostname;rfs:=/usr/local Then run "amd /usr/local /your/map/name -type:=direct". One word of warning, however. In NetBSD 1.0, I couldn't get direct mount points to work for some reason. I don't know if this has been fixed or not. If you are using a NetBSD 1.0 (or earlier) system, make /usr/local a real symbolic link into an automount filesystem. Another instance of the amd.project file might look like this: /defaults type:=nfs;opts:=rw,soft,intr,grpid local \ host==hostname;type:=link;fs=/usr/local ||\ host!=hostname;rhost:=hostname;rfs:=/usr/local You amd.master file might look like this: /project amd.project Here's another example which auto-mounts /usr/src from another machine: grizu% ls -lad /usr/src lrwxr-xr-x 1 root wheel 29 Dec 30 15:33 \ (split by ed.) /usr/src -> /tmp_mnt/mounts/src10/usr/src grizu% cat /etc/amd/master /net /etc/amd/net /tmp_mnt/mounts /etc/amd/src grizu% cat /etc/amd/net /defaults type:=host;fs:=${autodir}/${rhost};rhost:=${key} * opts:=ro,soft,intr grizu% cat /etc/amd/src /defaults type:=host;fs:=${autodir}/${rhost}; src10 opts:=rw,soft,intr;rhost:=rfhu1001 grizu% grep ^amd /etc/netstart amd=YES amd_dir=/tmp_mnt # AMD's mount directory amd_master=/etc/amd/master # AMD 'master' map grizu% ls -la /tmp_mnt total 9 drwxr-xr-x 5 root wheel 512 Jan 3 12:09 . drwxr-xr-x 26 root wheel 1024 Jan 3 12:09 .. dr-xr-xr-x 3 root wheel 512 Nov 20 19:29 ftp.uni-regensburg.de dr-xr-xr-x 2 root wheel 512 Jan 4 10:43 mounts dr-xr-xr-x 4 root wheel 512 Dec 11 08:18 rfhu1001 grizu% ls -la /tmp_mnt/mounts total 3 dr-xr-xr-x 2 root wheel 512 Jan 4 10:43 . drwxr-xr-x 5 root wheel 512 Jan 3 12:09 .. I guess that's all. rfhu1001 is the NFS server, grizu the client. 4.5.2 I am having trouble with my nameserver refusing to accept 'nslookup's from my SunOS machine after I installed the resolver fix. The exact error message is "*** Can't find server name for address Query refused". Can you help? From the README file in the BIND distribution: Versions of NSLOOKUP up through BIND 4.8.3's used IQUERY to ask the local server for information about the server's own name. I assume that this was done in a "what the heck, nothing uses these, how can we contrive a need?" sort of spirit. I removed this code as of BIND 4.9's NSLOOKUP and had it use the standard gethostbyaddr() mechanisms (which depend on normal queries of PTR data). Disabling INVQ and putting "options fake-iquery" in the boot file will cause IQUERY to be answered bogusly but in a way that old nslookup programs won't trip on. INVQ is disabled by default in conf/options.h. So, your options are: - Add "options fake-iquery" to named.boot and restart the server - Replace your old, broken nslookup with the one in the 4.9.3 BIND distribution. -- Enable INVQ in conf/options.h, then rebuild and re-install named. This latter option isn't guaranteed to work. If you point an old version of nslookup at a server, and the server either is not authoritative for a zone containing the A RR matching the address you are sending the query to, or if this A RR is not in it's cache, then nslookup will still fail even if the server has the INVQ option turned on. 4.5.3 Are there any alternatives to 'NIS' available for NetBSD, et al.? Yes, there is 'hesiod' which provides (according to Ted Lemon <mellon@fugue.com>i)another way of distributing databases like /etc/passwd, /etc/services, /etc/groups, and so on. It uses DNS, which is (IMHO) slightly more robust and less easily subverted than NIS, and doesn't claim to provide authentication (authentication is Kerberos's job), so as part of a complete system, I think it's a much better solution. It certainly has a smaller installed base than NIS, though. There is also Kerberos IV, which provides similar functionality. It is now fully integrated into the all of the *BSD systems and works well for network wide authentication. 4.6 Adding new and removing old users. 4.6.1 Where can I FTP the 'adduser' program? There is one you can FTP (see the URL below). You will need to be able to use 'vipw' to make it work, but that shouldn't be a big problem for most people. ftp://ftp.quick.com.au/pub/unix/adduser.sh The man page is there too.. ftp://ftp.quick.com.au/pub/unix/adduser.8 4.6.2 Where can I get a 'rmuser' script? There is a Perl script called 'removeuser' which should be available from one of the 'CPAN' sites. As soon as someone has a URL, let me know. There is also a FreeBSD 2.2 'rmuser' program which does everything the remove program does, plus removes crontab and at command entries. -- Dave Burgess Network Engineer - Nebraska On-Ramp, Inc. *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...."