Questions and Answers Common to Unix, Xenix and ODT *************************************************** How do I stop banners from printing? ==================================== You need to edit the file /etc/default/lpd. You need one of the following lines: For Xenix: BANNERS=0 For Unix: BANNERS=nobanner Note that there are some Unix printer interface scripts which do not use /etc/default/lpd, and you must use an option to these to disable banners. Also, some Unix printer interface scripts expect the Xenix syntax above. Aren't standards wonderful? Should you encounter one of these, if you're reasonably adept at shell scripts, you might want to cut and paste the section that reads /etc/default/lpd from a script that works properly. Of course, be sure you make note of your changes so that you can redo them the next time an upgrade replaces your printer drivers. ____________________________________________________________________________ Are there any screen savers? ============================ Unix (and Xenix 2.3.4) have a built-in screen saver for VGA only. You have to reconfigure the kernel for this to work. It doesn't work with all hardware, but try it first. Also, it has been reported that VP/ix may not be compatible with this screen saver. To enable the screensaver, set the kernel variable TBLNK to the number of seconds of inactivity which should trigger the screensaver, relink, and reboot. ____________________________________________________________________________ Is tar/cpio a good backup program? ================================== tar is not; cpio is, to some degree. tar will not back up things like device nodes (and, prior to OpenServer Release 5, it will also not back up empty directories), so a tar backup will not catch anything in /dev, for example, and you will find your device nodes missing when you do your restore. cpio will catch these things. Neither one is very good at verification. You can dd the tape to make sure you can read the whole thing, and run it through tar or cpio ... but they'll just check the file headers to make sure they make some sense. If you need better verification, try one of the products listed below. Most third-party backup programs do many things better than the standard utilities included with the OS, including things like making much better emergency recovery diskettes, byte-for-byte verification (if you want), compression, more options for things like nondestructive restore, etc. Many of us swear by them. gnu tar is a significantly better backup utility, and is available on many archive sites listed in the Administrative FAQ. There is also a shareware tar/cpio archive checker called tapechk, written by Nigel Horne < njh@smsltd.demon.co.uk>. A demonstration version is available at ftp://garbo.uwasa.fi/unix/util/tapechk.sco.tar.Z ____________________________________________________________________________ How do I compress my backups? ============================= Well, you could just run the output of tar, cpio, or whatever through compress, but if even one bit of your tape or diskette goes bad, you'll lose the rest of the backup. Not recommended at all, unless of course you don't actually care if your backups work - but if you didn't care, you wouldn't be doing any, right? A better solution would be a third-party product. The next answer lists a few ... if you produce, market, or use one that's not listed below but which you feel should be, please send me the information. ____________________________________________________________________________ What are some third-party backup/recovery products? =================================================== There are a couple of categories here - products which are mostly aimed at one or a small number of Unix machines, and those which are aimed at enterprise-wide, multiplatform backup. The following two lists are NOT meant to be all-inclusive, but merely a sample of some of the better-known products. First, the ones aimed at one or a few Unix machines: o BackupEDGE (Microlite Corp., 2315 Mill Street, Aliquippa, PA 15001-2228; info@microlite.com; (888) BKP-EDGE or (724) 375-6711; http://www.microlite.com/) o BRU (Enhanced Software Technologies Inc., 5016 S. Ash Avenue Suite 109, Tempe, AZ 85282; swinfo@estinc.com; (800) 998-8649 or (602) 820-0042; http://www.estinc.com/) o Lone-Tar (Lone Star Software Inc., 13987 W. Annapolis Court, Mt. Airy, MD 21771; sales@cactus.com; (301) 829-1622 or (800) LONE-TAR; http://www.cactus.com/) o Ctar (Unitrends Software Corp., 1601 Oak Street, Suite 201, Myrtle Beach, SC 29677; sales@unitrends.com; (800) 648-2827 or (803) 626-2878; http://www.unitrends.com/) These products tend to be fast and robust, generally offer data compression, and tend to be able to handle errors on the backup media. Many also include, or can optionally be purchased with, utilities to create automated emergency recovery diskettes which are much friendlier and easier to use than the ones you can produce with standard SCO utilities. Now, a few for those with more ambitious backup plans ... this section is under construction and hopefully I'll have some more contact info shortly. o The Backup Professional (Lone Star Software Inc., 13987 W. Annapolis Court, Mt. Airy, MD 21771; sales@cactus.com; (301) 829-1622 or (800) LONE-TAR; http://www.cactus.com/bp.html) o ARCserve/Open (The Santa Cruz Operation Inc., 400 Encinal Street, Santa Cruz, CA 95061; info@sco.com; (800) SCO-UNIX or (408) 425-7222; http://www.sco.com/) o Legato (415) 812-6000 A variety of backup products was reviewed in the September 1997 issue of SCO World Magazine. ____________________________________________________________________________ I don't like being restricted to 14 character filenames ======================================================= If you're running Xenix, or a version of Unix prior to 3.2v4, I'm afraid you're stuck. Unix 3.2v4, however, includes long filename support on all EAFS filesystems. OSR5 adds two new filesystems, HTFS and DTFS, which also support long filenames. More information on long filenames can be found in the section dealing with Unix. ____________________________________________________________________________ How do I get a copy of adb? =========================== If you have the Development System, you already have /bin/adb. If not, you may need to grab a copy from your distribution, or it may already have been installed, depending on your OS and version. It could be called /bin/adb (older Xenix) or /etc/_fst (Unix and recent versions of Xenix). If you don't have either of these, look through the files in /etc/perms for them; in Xenix 2.3.4, you will find one of each, which will be in fact the exact same file but on two different diskettes. If the volume on which the file you want is mountable (you can check this in the manual, or use the dtype command), then mount it and copy the file off. Otherwise, use tar to extract the file, keeping in mind that the filenames on your diskettes are all written with relative paths (i.e. ./bin/adb, not /bin/adb). Note that if you look in the Unix documentation, it may well tell you that you need /bin/adb, when in fact it's called /etc/_fst. ____________________________________________________________________________ I can't find crypt ================== Most (all?) of SCO's release notes state that due to American government restrictions aimed at trying to prevent unfriendly nations from having access to data encryption technology, SCO does not ship crypt with their products. If you live in the States and would like crypt(C) and the crypt(S) libraries, contact SCO support. This is also worth trying in Canada, as the particular regulation in question permits export of such technology to Canada; however, I don't know if SCO will honour such requests. There is also an international version of crypt available from the usual places as lng225b. ____________________________________________________________________________ What do I need to compile programs? =================================== If you have free OpenServer, you already have a license to install the development system; the Web page on which you license free OpenServer gave you several keys and codes, including one to license the development system. Xenix, Unix and ODT do not ship with program development tools. These are unbundled into packages known as Development Systems. The rationale behind this is that many users of SCO systems are using off-the-shelf software and never need to write a line of C code. If everyone was forced to buy the development system whether they needed it or not, some of the customers might get upset. There is a periodic flame war about this; this is not the place to discuss it. You can buy the Development System for any of the three environments listed above as a separate package including the compiler, header files, libraries, lex, yacc, linker, and other tools. Additionally, development systems are available for other packages such as TCP/IP; these development systems add the include files, libraries, etc. required to program for the package in question. The ODT Dev Sys includes the development tools for all of the other goodies (e.g. TCP/IP, X) that are bundled into ODT. Since OSR5 generally bundles the various runtime packages (e.g. TCP/IP) with the OS, it also bundles the same development packages, so there are not the same development system add-ons in OSR5 that there were in previous versions. There are versions of gcc (the Gnu C Compiler) freely available for SCO systems. On older SCO operating systems, however, you will probably need the development system, as the header and library files you need are part of it and not part of the operating system itself. This problem has been alleviated in OpenServer Release 5, as the headers and libraries are now shipped as part of the base operating system and are available even if you have not purchased the development system. gcc sources and binaries for OpenServer Release 5 only are on the free Skunkware family of CD-ROMs; for more info, see http://www.sco.com/skunkware/faq.html or read the section below entitled "What is Skunkware?" gcc sources and binaries are also available on Robert Lipe's home page: http://www.dgii.com/people/robertl/ or ftp://ftp.dgii.com/users/robertl/scods/ These are mirrored by SCO at http://www.sco.com/skunkware/gds/ and ftp://ftp.sco.com/skunkware/gds/ You can also look at a different version at ftp://ftp.sco.com/skunkware/osr5/devtools/gcc/ and http://www.sco.com/skunkware/osr5/devtools/gcc/ For those who want to find this based on a keyword search: programming programmer library libraries developer source ____________________________________________________________________________ What does the NCALL kernel parameter affect? ============================================ NCALL controls the size of the kernel callout table. The kernel has the ability to schedule some action at a given real time; this is often used by device drivers and by the nap(S) system call. The size of this table is set by NCALL. If the system message "timeout table overflow" appears on your console, NCALL should be increased. Increasing NCALL is not expensive in terms of memory or CPU overhead, as the structure is small (16 bytes per entry) and stored sorted, so it is best to be generous with these entries. ____________________________________________________________________________ How do I reset the root password if I forget it? (part 1) ========================================================= This procedure will work for Xenix, and for Unix as well if you are using a very relaxed security level (one which stores encrypted passwords directly in /etc/passwd). If you're using a higher security level on Unix, look for part 2 below. Boot the system from your emergency boot diskettes (if you didn't make these and keep them up to date, shame on you, but you should be able to use N1/N2 instead, and see the entry on crashing out of these diskettes below). Next, mount /dev/hd0root /mnt; this will mount your hard drive's root filesystem on /mnt. Edit /mnt/etc/passwd. The first line will be your root line, such as root:encryptedpasswordgoeshere:0:0:God,Everywhere:/:/bin/sh Edit out the encrypted password (don't touch anything else!) so that the line reads something like root::0:0:God,Everywhere:/:/bin/sh Save the file and shut down. Reboot from the hard drive. Your root password has now been removed, and you can reset it normally. ____________________________________________________________________________ How do I reset the root password if I forget it? (part 2) ========================================================= This is another procedure involving manually editing files, and is specific to SCO Unix 3.2v4.0 through 3.2v4.2. The location of the encrypted passwords depends on the security settings. Look in /etc/passwd, /etc/shadow, and /tcb/files/auth/r/root; one or more of these will be used depending on how you have security configured. Follow the procedure in part 1 above; instead of editing /etc/passwd, edit the appropriate file(s) from the above list, and delete the encrypted password field. Note that formatting is critical; while you can delete the contents of the field, you must not remove separators, and making seemingly minor errors such as leaving blank lines can cause problems. Save, shut down, and reboot. C2 security will complain about what you've done; to make it happy, run /etc/fixmog. You may also want to run /tcb/bin/integrity and /etc/tcbck. ____________________________________________________________________________ How do I reset the root password if I forget it? (part 3) ========================================================= This procedure will work for any variant of SCO Xenix or Unix. As above, boot from your emergency boot diskettes and mount /dev/hd0root /mnt to gain access to your hard drive's root filesystem. Now, run /mnt/bin/chroot /mnt "/mnt/bin/passwd root" (check the chroot man page for more info on how it works). As before, shut down and reboot. It has been reported that on 3.2v4.2 (and possibly others), this must be done in two steps: /mnt/bin/chroot /mnt "/bin/su root", followed by passwd. If it doesn't work with the quotes, try it without. ____________________________________________________________________________ How do I crash out of the install script? ========================================= On OpenServer Release 5, boot from the boot diskette, and at the Boot: prompt, type tools. This is not an undocumented option to the boot command, but rather a special line in /etc/default/boot on the installation diskette - so you can't use it from anywhere but your installation boot diskette. For older SCO Unix/Xenix/ODT releases, wait until the question early in the process that asks you what your keyboard type is. For character-mode installations, this is a regular textual prompt; for ODT, it's a box in a curses-style installation program. How to break out at this point depends on the OS. Under Xenix, press Del. Under Unix, type shell and press enter. Under ODT, press Control-A. ____________________________________________________________________________ Why can't fsck find my lost+found directory? ============================================ Because you don't have one. It's possible someone deleted it, but the more likely cause is that you didn't use mkdev fs to create it. One of the things that fsck looks for is inodes which are marked as used (i.e. not in the free list) but do not have a directory entry pointing to them. fsck will ask if you wish to reconnect these; if you say yes, it will try to create a file entry in the /lost+found directory on that filesystem. If there is no free space in /lost+found, it is not safe to expand it because the rest of the filesystem may still be corrupt; for information on this one, see below. If there is not /lost+found directory, fsck will tell you that it can't reconnect the file and the data in that file will be lost. ____________________________________________________________________________ I want more space in my lost+found directory ============================================ By default, the mkdev fs script creates 62 empty entries in lost+found. If you'd like to make more, use a variant of the following script: for a in 0 1 2 3 4 5 6 7 8 9 do for b in 0 1 2 3 4 5 6 7 8 9 do > /lost+found/dummy$a$b done done rm -f /lost+found/* The above will create 100 entries. Season to taste. ____________________________________________________________________________ How do I find out serial numbers of my various components? ========================================================== For the OS itself, you can use uname. For Unix, use uname -X; it will print (among other items) the serial number. For Xenix (at least 2.3.3, and probably other releases), uname -u will print the numerical portion (e.g. if your serial number is sco012345, it will print 12345). Starting in Unix 3.2v4.2/ODT 3.0, SCO added /etc/getserno. To find out the serial number of a package, first find out what files are serialized in that package using grep ser= /etc/perms/* (or /etc/perms/packagename if you know it). Then, run /etc/getserno filename, where filename is the name of one of the files that is serialized during installation. Note that not all files listed may actually contain a textual representation of the serial number (for example, none of the binaries in the Unix dev sys do). As a special case, the serial number of the OS itself can be found simply by watching the kernel ID it displays at boot time (or look through /usr/adm/messages for it). ____________________________________________________________________________ How do I solve an "arglist too long"? ===================================== Wildcard expansion (globbing) is performed by the shell. There is a limit of 5120 bytes (5k) for the environment and command line arguments put together, in all versions of SCO Xenix and SCO Unix versions prior to OpenServer 5; more on OSR5 later. See also TA 480563. This is particularly likely to be a problem under X, as it has a habit of using a lot of environment space. It is also a problem when running a command such as ls *.c in a directory with a large number of files which match the filespec. The general solution is to construct your command in such a way that it does not have to include all of the filenames on the command line. You can use the echo command, which is built into the shells and therefore is not subject to the 5k limit. For example, rather than rm V*, you might try echo V* | xargs rm. A similar, but somewhat more complex solution, might involve using the ls command to generate a list of filenames, and then using a command such as grep to filter them; ls | grep '^V' | xargs rm will perform the same task as the above example. You may also find the find command to be useful in this; however, it works recursively so it may not be appropriate in a directory with subdirectories. Please consult the man pages for each of these commands to identify any unexpected side effects they may cause. Another alternative, in cases where the environment is unnecessarily large, is to reduce its size. If you have some environment variables that you never use (be careful with this, as the system or some commands may use things you don't realize), you can permanently remove them in your .profile (or .login for C Shell users). You could also temporarily remove some manually. To run a subshell without any of the environment being passed to it, try running env - sh -c 'command' OpenServer Release 5 makes two changes to cure this problem. The default limit has been increased substantially (to 100k), which should by itself fix almost all instances of "arglist too long". As well, it is now a tunable kernel parameter, so if the default isn't adequate, you can adjust it. One exception: /bin/csh still has a hard-coded limit to the length of a line. If you are using csh, you may wish to replace it with tcsh (discussed below). ____________________________________________________________________________ What versions/configurations am I using? ======================================== WARNING: Many of these commands have different options under different versions of different operating systems, and not all of them are available under all versions of Unix, Xenix, and ODT. I've tried to note such differences but I'm sure many have escaped my attention. Take the following with a grain of salt. Unless noted otherwise, these entries should be applicable to most/all systems. o Kernel Configuration: configure -x | more (for Xenix, run this from /usr/sys/conf; for Unix, run it from /etc/conf/cf.d). This lists the current and default values for tunable kernel parameters. Under Unix, /etc/sysdef prints information including BTLDs (Boot Time Loadable Drivers). o Software Installed: /usr/bin/swconfig -p and /usr/bin/swconfig -a (both for Unix) print various information on installed software. You can look at the permissions lists in /etc/perms/* but you cannot tell from here which parts are installed; use custom for that. Use /usr/bin/displaypkg to display software installed using installpkg. Note that swconfig is not a terribly accurate guide. o Hardware configuration: /etc/hwconfig -h shows most of the installed hardware but not all of it; generally, things like multiport cards don't show up here. Use /etc/hwconfig -hc on Unix 3.2v4.x or later and on Xenix 2.3.4 o System name, version, etc.: uname -X (Unix and Xenix 2.3.4) or uname -a (Xenix 2.3.3 and earlier) o Printer configuration: lpstat -t ____________________________________________________________________________ I have a bad block on my hard drive =================================== You will see error messages going by giving you the sector, cylinder, head, and other nifty information regarding the error. If you can jot this down, it makes it much easier to find the bad block without having to scan the entire drive for it. Shut the system down cleanly (using shutdown). If the error is on the root filesystem, boot from emergency floppies; otherwise, you can boot from the hard drive and enter single-user mode. The rule here is that the filesystem on which the error is located must not be mounted while you try to fix it. If you have a SCSI hard drive, use scsibadblk. It ships with Unix 3.2v4.1 and 3.2v4.2, and ODT 2.0 and 3.0. For Unix 3.2v4.0, install the 4.1 maintenance supplement or upgrade to 4.2 (not a bad idea anyway). For Unix 3.2.2 or ODT 1.1, install unx347a (no longer available). For Xenix 2.3.4, install xnx348a. For OSR5, scsibadblk was rolled into badtrk, so just use badtrk. For older versions of Xenix or Unix, you're out of luck. One other note about SCSI drives; many of them will automatically remap bad blocks, so when you go to run scsibadblk you will not actually find any bad blocks - even if you run a thorough scan of the area where the bad block was reported. This capability is called AWR/ARR. If you see a menu option called something like "Modify target parameters", you can enable and disable AWR and ARR. If you're using a standard drive type (MFM, RLL, ATA, ESDI), use /etc/badtrk. I'd recommend doing a thorough, nondestructive scan of the area where the error message said there was a bad block. Before doing this stuff, have a look at the manual for your specific operating system to see any notes or recommendations made by SCO. If you're not careful here, you might make things worse than they already are (such as by doing a destructive scan, which will wipe out all data on the area you scan). ____________________________________________________________________________ My system is slow ================= First things first - make sure that somebody didn't accidentally turn the Turbo switch off. Don't laugh - I have a client who regularly manages this one. At some sites, it may be wise to disconnect this switch entirely. It might also be wise to run the system's CMOS setup program and ensure that primary and secondary cache is turned on, unless you know for a fact that there's something in your system that won't work properly that way. Turning on BIOS shadowing will generally only speed things up at boot time; with the exception of vbiosd (used to call real-mode video BIOS routines for video mode switching on some video cards in SCO's X11R5 implementation), the BIOS is not used after this point. If you gain the use of extra RAM by disabling BIOS shadowing, you should certainly do so; even if you don't, there may be cases where BIOS shadowing may lead to weird problems (I've even seen a host adapter which wouldn't work at all if its BIOS was shadowed or cached, for example). Under both Unix and Xenix, you can use vmstat to give you an overview of system performance. One problem is that it won't show you what percent of the system's time was spent waiting on I/O devices, and what percent was spent idle; these are both lumped together as idle time. vmstat can be helpful in diagnosing excessive swapping, and in finding if your system is CPU-bound. Unix also offers sar, which is far more advanced than vmstat. It reports on a wide range of system statistics including CPU utilization (system, user, idle, waiting for I/O), memory use, disk cache effectiveness, swapping/paging, and things you've never even thought of. Note that under MPX, it may not be reliable; check your MPX release notes for info (and for information on the mpstat and mpsar programs). One third-party program which may be useful in conjunction with sar is sarcheck (Aurora Software Inc., P. O. Box 1033, Plaistow NH 03865, (603) 382-4200, http://www.sarcheck.com/, 74013.1625@compuserve.com), which translates sar's results into English to identify system performance bottlenecks and suggest possible resolutions for these problems. sarcheck also works on multi-processor systems. There are some other utilities you may wish to use. Some freely- available ones include u386mon, bcw, and cpuhog/iohog/memhog, all of which are available in various TLSes (tls518 for OSR5, tls018d for older versions). u386mon is a general performance monitoring utility which watches about as many different things as sar (but presents the information in a full-screen display format); bcw is the Buffer Cache Watch, which can help you see how well your cache buffers are tuned for your system's actual needs; the hog programs show you processes which are hogging those respective resources. Another commercial product which may be of use is Olympus Tuneup (Olympus Software, (408) 426-7582, olympus@olysoft.com), which will monitor how your system is making use of tunable kernel resources and can perform tuning for you. Multiuser/multitasking/etc. operating systems love extra memory. Xenix will use up to 16 MB; Unix will use much more (how much depends on what version; check your release notes). There are several ways that extra memory is used; here are three of the most important. First, disk buffers; the system uses these for disk cache, and in general, the more, the better. Second, to avoid swapping; while a virtual memory system allows you to access more memory than you actually have, doing so involves the hard drive, which is several orders of magnitude slower than memory. Third, the system keeps recently-used programs in memory; if you access one again, it doesn't have to be reloaded from disk. There are tradeoffs between #1 and #2+#3; the more memory you have, the more generously each can be configured. Note that adding more memory will not cure CPU-bound processes, and will only cure I/O-bound processes if it can be used effectively as a disk cache (often it can, but not always). ____________________________________________________________________________ Why did my region table overflow? ================================= Each process generally consists of several (usually, but not always, three) regions - typically code, data, and stack. Two copies of the same program running at the same time will often share code, reducing the number of regions required; however, there's nothing to stop a program from using more than three regions, either. There is a tunable kernel parameter, NREGION, which specifies the maximum number of regions available. This should always be set to at least three times the number of processes (NPROC), and if you want to be on the safe side, use four times NPROC. Note that in OSR5, by default, both NREGION and NPROC are allocated dynamically. ____________________________________________________________________________ How do I solve "fork failed: no more processes"? ================================================ This is usually one of two things. There is a tunable kernel parameter, NPROC, which determines the maximum number of processes that may be running at any time. You may have exceeded this limit. The usual method of solving this is to increase it a fair bit and see if the problem goes away. If you are running on OSR5, this is unlikely to be the reason, as NPROC is allocated dynamically. There is another tunable kernel parameter, MAXUPRC, which determines the maximum number of processes any one user may have running at one time. Under Unix, for example, a large number of mail messages being processed at once may cause this to be exceeded by MMDF, usually resulting in "uux failure - pipe broke" or similar messages. Once again, increase it and see if the problem goes away. Also, have a look at the console and/or /usr/adm/messages for any system messages which appeared at the same time the user got this message. They may point to another potential reason, such as being out of swap space or exceeding NREGION (see the previous topic). ____________________________________________________________________________ How are minor device numbers assigned by mkdev hd? ================================================== Basically, they start at 64 (the major device number is 1) and go up by 64 each time you run mkdev hd. Don't expect them to be in the same order as your SCSI IDs for the drives unless that's the order you added them in. Also, if you being running mkdev hd but do not complete the process, it will generally already have assigned the next number; the next time you run mkdev hd, it will add another 64 even though you aren't actually using the last drive you started to create. This isn't a problem; it just looks weird. ____________________________________________________________________________ I need fax software. Who makes it? ================================== There are numerous vendors in the Unix fax software market. Many of these make software that runs on Xenix as well as on Unix. Listed below, in no particular order, are company names, product names, and contact information for most of them. As always, I hope this is a reasonably complete list; inclusion or exclusion is not to be construed as a comment on the product or company. Also, there is a fax FAQ posted in comp.dcom.fax (from which this list is derived - note that it is out of date); you may wish to look there. Also, the standard "look through magazines" applies. Fax products are often advertised, and sometimes reviewed. The September 1995 edition of SCO World Magazine reviewed some fax products, for example. o Arnet - ArnetFAX; (615) 834-8000, clarence@arnet.com o Black and White Software - NXFax; (802) 496-8500, nxfax@bandw.com o comFax - Com-M-Tex; +49 89 546130-0 o COS - TruFax; (609) 771-6705, trufax@cosi.com o Faximum Software - Faximum ELS, Faximum PLUS; (604) 925-3600, info@Faximum.com o ICSW - [product name unknown]; (800) 486-7274, (602) 998-8623 o i link GmbH - mix fax; +49 30 216 20 48 o Intuitive Technology - FaxLink; (409) 762-8456 o netCS GmbH - netFAX; +49 30 787999-0 o QUEST systems GmbH - FaxX; +49 231 914028-0, faxx@quest.sub.org o Signify Software Products - i(F)x Faxsoftware for UNIX; +31-(0)3480-30131, gerard@integrity.nl o smoFax - SMO GmbH; +49 721 551971 o UniSal System - FaxTrax; (201) 729-9221 o V Systems - VSI-Fax; (714) 489-8778, info@vsi.com, http://www.vsi.com/ o Company Unknown - FaxFX; (708) 574-3600 o Company Unknown - FAXSMART o Company Unknown - Fax*Starx; (800) 327 9859 You might also want to look at a couple of publicly-available programs. Check out Hylafax at http://www.hylafax.org/ and ftp://ftp.hylafax.org/. ftp://sgi.com/sgi/fax/. You can find more information on mgetty+sendfax at http://wais.leo.org/~doering/mgetty/mgetty_1.html. ____________________________________________________________________________ How much swap space do I need? ============================== There are two factors to consider - how much you actually need for correct operation of your system, and how much you might need in case of a kernel panic. Unix and Xenix are virtual memory operating systems. If you have, say, 16 MB of RAM and a 20 MB swap device, you have 36 MB of virtual memory available, of which the operating system will keep 16 MB in memory (whatever space it uses, plus the most recently used user memory). Therefore, the total of swap space plus physical RAM must equal or exceed the greatest amount of physical memory your system will need. The kernel, however, will check how much swap space is available whenever a program executes a system call which may require more swap space, such as fork() or malloc(). There must be enough free swap space to hold the memory which the system call will allocate, or else the system call will fail. This is a safety precaution which applies even if no swapping is required! So you will need as much swap space as you will have swappable memory (generally, stack and data regions). Under Unix, you can use the crash command to check how much swap space has been allocated. Once in crash, type od -d availsmem to see the value of availsmem. This is a kernel variable which is measured in 4k units (i.e. pages in the i386 memory architecture) and which says how much more memory can be allocated. Any request for a number of pages greater than the current value of availsmem will fail. See TA 482712 for some more information on availsmem. Of course, in order to know how much swap space you need, you need to have an idea of how much total virtual memory your programs will require. Some need more than others. For example, if you're running any Java components, you may need a lot of swap space; check the release notes for details. X Windows generally eats a lot of memory, so you'll want plenty of swap space if you do much work in X. Your applications' documentation may give you information on how much swap/memory they require. The other consideration relates to system panics, and does not apply to Xenix. Should a system panic occur, the kernel will dump the contents of physical memory into the swap device. There is a net.rumour that should you have more physical memory than swap space, it will overwrite whatever's next to the swap device with the dump. It is possible to force the kernel to overwrite something with a poor choice of parameters on the Boot: line (e.g. explicitly giving an incorrect size for the swap device), but without this form of prompting, the kernel will not dump any more memory than it believes will fit into the device specified for dumps. One final comment - swapping is a wonderful thing, in that it allows you to use more memory than you actually have. However, disk is several orders of magnitude slower than memory, and so the more swapping you have, the slower your system will run. If you find your system is swapping and that is having a noticeable effect on performance, you should consider adding memory if your hardware and OS support more memory than you have. ____________________________________________________________________________ Can I add more swap space? ========================== Yes, with caveats. The first one does not apply to Xenix. If you have space in your Unix partition which is not allocated to any device (i.e. is not being used by a filesystem, your swap device, etc.), use /etc/swap to add this to your system's available swap space. Note that free space within a filesystem cannot be used in this manner. Also, this setting only works until the system is shutdown, so if you want it to be done permanently, put it in a file in /etc/rc2.d so it gets run whenever the system goes multiuser. If you have two hard drives, you can split swap space between them, which may improve swapping performance. On OpenServer Release 5, you can also add swap space in the form of a file on one of your filesystems. As with the previous section, you use the swap command, and the added swap space does not become permanent unless you add it to a startup file. I have no benchmarks on this, but I'd expect that swapping to a file is at least a bit slower than swapping to a dedicated swap division. The second approach will work on any SCO operating system, but will require downtime and probably a backup/restore. You can bring the system up from emergency boot diskettes (or from the distribution media; instructions are elsewhere in the FAQ) and adjust your drive's division table. However, in order to adjust the size of a filesystem or swap device, you must delete it and recreate it, so if you need to take space from a filesystem to add it to swap, you will need to backup that filesystem and restore it later. ____________________________________________________________________________ Do haltsys and reboot do a sync()? ================================== Yes. haltsys and reboot are both the same file. In some versions, they are a binary, but rest assured that they do sync(). In other versions, they are a shell script and you can look at them to determine that they do call /bin/sync. If it's at all possible to use shutdown to shut the system down, rather than using haltsys or reboot, do so. shutdown is the proper way to do it; it goes out and kills processes and attempts to shut the system down as cleanly as possible. haltsys and reboot, on the other hand, try to shut the system down as quickly as possible, and any programs which are running will be rudely interrupted. Also, if you're using a caching hard drive controller, be aware that it may not realize the system has been shut down, so even though Xenix or Unix tells you it's *** Safe to power down ***, there may still be data left in the hardware cache that isn't flushed to disk yet. A good, but not foolproof, precaution is to press a key to allow the system to reboot, and not power down until the Boot: prompt comes up. The added time and disk activity may allow the controller to flush its cache. If your caching controller has a specific driver for SCO and you're using that driver, then it can communicate with the operating system to ensure that its buffers are all flushed, and this problem does not arise. Note that using a caching hard drive controller on a caching operating system is generally of little or no use, though there are certainly some cases in which it makes a significant performance difference (but only by defeating the order in which write requests were made by the kernel or the application, possibly decreasing data integrity somewhat if the system crashes). ____________________________________________________________________________ How can I get more than 64k inodes? =================================== SCO Xenix and all versions of SCO Unix up to and including 3.2v4.2 use 16-bit unsigned integers for inode numbers, so there is a limit of slightly under 65 536 inodes available per filesystem. If you are running one of these versions, you'll have to make multiple filesystems, each with 64k or less inodes. OpenServer 5 includes two new filesystems, with 27- and 30-bit inode numbers. These provide approximately 130 million and approximately one billion inodes per filesystem, respectively. ____________________________________________________________________________ Where do I get zmodem? ====================== Some of the ftp sites listed in the Administrative FAQ (posted at the same time as this FAQ) should have versions that are compiled and known to work on SCO systems. See also tls025. If you want source, try ftp://ftp.cs.pdx.edu/pub/zmodem/. This program also handles X and Y modem transmission and reception. Zmodem was designed by Chuck Forsberg of Omen Technology ( http://www.omen.com/). Omen's Web site includes both freely available and commercial communications software. ____________________________________________________________________________ Where do I get kermit? ====================== The Kermit Project at Columbia University can be found at http://www.columbia.edu/kermit/ or kermit@columbia.edu The latest version (as of 25 September 1997) can be found, precompiled for a number of SCO platforms, at http://www.columbia.edu/kermit/ck60.html The companion products for Windows 95 and for DOS/Win3.x can be found at http://www.columbia.edu/kermit/k95.html and http://www.columbia.edu/kermit/mskermit.html ____________________________________________________________________________ I get messages saying "stat() failed: /tmp/croutPPGa00288: no such file" ======================================================================== Basically, this is normal. When cron runs an at, batch, or cron job, it creates a temp file named /tmp/crout* to hold the stdout and stderr of that job. When the job is finished, it mails the results (if any) to the job's owner and then removes the file. In the meantime, some other program (probably a filesystem cleaning daemon) has been scanning for whatever purpose. It did this in two passes; first, it got a list of all files it had to consider (probably by asking the shell to expand "/tmp/*" into a list of files); second, it uses stat() to find out information about each one. Between these two steps, the cron job finished and cron deleted the file, so by the time the second job went to get information about the file, it had vanished. This message is harmless so long as it refers to a cron output temp file. If it refers to some other file, you may want to find out what generated that other file; chances are it's a harmless message, too. ____________________________________________________________________________ Does SCO support my hardware? ============================= If you already have your copy of Unix/Xenix/ODT, the first place to look is in the Release Notes. If you see your hardware listed there, it's supported. However, the Release Notes are not quite up-to-date; this industry changes so quickly that a manual written this month will be out of date by the time it's come back from the printers. The on-line solution is at http://wdb1.sco.com/sdir_web/owa/scodir_display_third_levels?f_category1_id=22 SCO used to make the Hardware Compatibility Handbook available in the form of postscript files, but stopped in late 1997. To be perfectly accurate, the HCH also is not always up-to-date. The best advice is to scout around in the EFS directory on ftp.sco.com and read the doc files for the latest and greatest Advanced Hardware Supplement files. See also the TLS and VCD directories, plus the appropriate AHS directory for your version. ____________________________________________________________________________ How do I get a file off my distribution diskettes? ================================================== The simplest way is to use custom to reinstall that file. If you can't do this for some reason, try the following. Note that with the extensive use of symlinks in OSR5, this method may not be very easy to use on OSR5 because you may not know the actual file you want (which is the file within the SSO tree, not the file in the usual place you're accustomed to finding it). NOTE: This applies ONLY to diskette media. I don't have tape or CD to play with here so I can't be sure on those. Please feel free to send deltas for these media. First, find out which diskette it's on. Use your favourite search tool to look through /etc/perms/* for the file you want. The last field in each line of this file is the diskette label (e.g. N1, X2). Find that diskette. Look on the label to see if it's a mountable filesystem (N1 is; N2 is on some distributions but not on others). If it's a filesystem, mount it and just copy the file. If it's not, it's in tar format, with relative pathnames. If the file you want is, say, /bin/foo, you'd extract it with a command like tar xv2 ./bin/foo ____________________________________________________________________________ Will I have problems upgrading my hardware? =========================================== There are a lot of different situations possible here - far too many to cover. I'll handle a few; any common additions to this list are certainly welcome. Upgrading memory - on the hardware side, that's easy; add the memory and run whatever your hardware uses for a setup program. One gotcha here is that many motherboards with L2 cache can only cache up to a certain amount of memory. Check with your motherboard manufacturer. In many cases, you need at least 64 kB of cache per 16 MB of system memory. From the software side, Xenix handles only 16 MB of RAM; different versions of Unix handle different amounts, so check the release notes for your version. The kernel will automatically see the additional memory, but may not put it to optimal use. There are a few kernel parameters which can auto-tune themselves within certain ranges (the best-known is NBUF), but most are fixed at link time. So the short answer is that the system will _use_ the memory, but perhaps not in the way you want it to unless you configure and link a new kernel. Upgrading the CPU - it should work fine, with one notable exception. Older Unix (and Xenix GT) releases have a timing loop in the ad (Adaptec 154x) driver detection code which will time out on many high-end 486 or higher machines, resulting in the host adapter not being detected. Recent Unix releases have this cured; there is a patch for at least some older Unix versions. See the question on 154x detection in section 3. Using an EIDE drive - first, some background. Traditional hard drives appear to consist of cylinders, heads, and sectors, and the standard hard drive controller driver in SCO products has traditionally expected standard (MFM, RLL, many ESDI, and ATA) hard drives to appear this way. For ATA drives, this is fine up to about 500 MB, at which point the interface details and the BIOS conspire to cause problems. EIDE gets around this by defining a new addressing mode - LBA (Logical Block Addressing). LBA must be supported by your operating system and/or your BIOS, however, and no SCO product prior to OpenServer Release 5 supported LBA mode out of the box. There is a note in the second section of this FAQ on how you may get an EIDE drive beyond 500 MB to work on 3.2v4.x. Also, LBA support is one of the features of uod429a; check the documentation for this patch. If you're running OpenServer Release 5, your EIDE hardware should work just fine. If you're running an earlier release of SCO Unix, or any release of SCO Xenix, your EIDE hardware should work as long as it's in CHS mode and NOT in LBA mode. The other caveat is that since /boot is real-mode code which does its disk access through the BIOS, and since the BIOS can only access up to 1024 cylinders, there are several files and directories which must lie entirely within the first 1024 cylinders in order for you to boot. The easiest way to do this is to keep your entire boot filesystem within the first 1024 cylinders. For fresh OSR5 installations, you have the option of creating separate boot and root filesystems; this is generally a good idea, and means that your root filesystem can be as large as you like (since it's only the boot filesystem which needs to be in the first 1024 cylinders). If you do not select a separate boot filesystem, or if you're running an older version of Unix or Xenix, your entire root filesystem should live within the first 1024 cylinders. Changing SCSI host adapters - there's a big gotcha here. Different host adapters (even the same one, often, if configured differently) use different logical mappings of the drive, and a drive set up under one host adapter may not be readable under another. Even two versions of the same adapter may have this problem; I've seen problems moving from an Adaptec 154xB to a 154xC, and Adaptec recommends reformatting (!). So it may work ... or it may not. Never ever try this without at least one backup which verifies 100% perfectly ... and see the note earlier in this FAQ about using tar or cpio for your backups before considering either to be a good backup program. There is a note in the Unix-specific part of this FAQ on a possible procedure for such a change. See also TA number 483121. Using an ATAPI CD-ROM - SCO Xenix doesn't support CD-ROMs. SCO Unix (at least since version 3.2.2) does, but prior to OpenServer Release 5, they had to be SCSI CD-ROMs. While SCSI is generally a wiser choice, support for ATAPI CD-ROMs has been added to OpenServer Release 5. While many IDE CD-ROMs are ATAPI CD-ROMs, not all IDE CD-ROMs follow the ATAPI spec. Only those which do are supported. Note that uod429a adds support for ATAPI CD-ROMs; check the notes for this patch to determine whether it's applicable to your system. ____________________________________________________________________________ I typed in the wrong serial number! =================================== On some products, there is a command,/etc/serialize, which will do the dirty work for you. Check for this file before trying the second method below. /etc/serialize takes one argument, which is the name of a permissions file, and will ask you for keys. Try the following: cd / ls /etc/perms | while read file do /etc/serialize /etc/perms/$file done It may complain about some files with nothing to serialize; this is normal. Also, it will rewrite binaries and should only be run in single-user mode so that it doesn't clash with files which are currently busy. It will also leave some files named /tmp/*.ser with your serial numbers and activation keys - so you definitely want to clean those up. If you don't have /etc/serialize, there's another way to do it. In your /etc/perms directory, find all of the files which belong to the product in question. Scan each one for a line near the top which begins #ser=; this line lists all files which must be serialized in this package. Many of the files in /etc/perms will have no such line, or will have an empty line; this is normal and these files can be ignored. The exact list of files will vary from release to release. You can now use /etc/brand to reserialize them. Change to the root directory and run /etc/brand serno actkey file [file ...]. For example, if the files are ./etc/getty and ./unix, you'd run cd / /etc/brand sco012345 abafjdlg ./etc/getty ./unix ____________________________________________________________________________ Why does fsck want a scratch file? ================================== There is an archaic limit to how large a filesystem fsck can check using available memory (archaic because it hasn't kept up with the growth in system memory). The exact limit is not something that appears to be documented anywhere, and may also vary between versions and different filesystem types. When this size is exceeded, fsck will want to use a scratch file to hold information while it's running. Before I continue, please read the man page for the -t option to fsck, and pay particular attention to the warning about following it with a space. Failure to do so may destroy data. You may have been prompted at the time you installed the OS to create a scratch division if your root filesystem was too large for fsck to check. If so, you might wish to edit /etc/default/filesys to specify that this should be used if the system has to check the root filesystem after a crash. Add -t /dev/scratch (or whatever you called the scratch filesystem) in the fsckflags= entry for /dev/root. For any filesystem other than root, you can generally use a temporary file on your root filesystem as a scratch file. fsck will create it and delete it automatically, once you've told it what file to use. I usually use /tmp/scratch. If, however, you find you need to fsck /dev/root, which is too big to check without a scratch file and you don't have a scratch filesystem, you still have some choices. A blank (but formatted) floppy diskette will often do the trick. If you're running fsck in single-user mode and you can guarantee that no swapping has taken place and no swapping will take place while you're running fsck, you could use /dev/swap. For Unix 3.2v4.2 and ODT 3.0, see uod418a, which provides a new fsck which may eliminate the need for a scratch file. ____________________________________________________________________________ How can I boot multiple operating systems? ========================================== Your SCO system includes the ability to boot Unix/Xenix or other operating systems; see the man page for boot. This is enough in most, though certainly not all, cases. As always, make and verify at least one backup and at least one set of emergency diskettes before performing any major systems work. This qualifies as major systems work. Sometimes, you may have to deactivate your Unix/Xenix partition, install a new operating system, then reboot from your emergency diskettes and reactivate the Unix/Xenix partition. There is at least one highly-regarded third-party utility which may help. Contact V Communications ((800) 648-8266 or http://www.v-com.com/) for information on System Commander. ____________________________________________________________________________ How do I set disk space quotas? =============================== No filesystem in OSR5, SCO Unix, ODT, or Xenix supports hard quotas. For another approach, look around ftp://ftp.armory.com/pub/admin/ where you will find, among other things, shell scripts to check users' disk space usage and notify users and administrators when users are using more space than you'd like them to. ____________________________________________________________________________ How do I find out what IP address a user logged in from? ======================================================== In OSR5, there are options to who, w, last, and finger which provide this information. In a program, you can fetch this information from /etc/utmpx; #include for the appropriate definitions. For some earlier versions, see nwho (in tls059b). The farther back you go through older versions, the less likely you are to find this sort of information. ____________________________________________________________________________ My ANSI terminal emulator doesn't work correctly ================================================ Unfortunately, everyone has a different definition of the behaviour of an ANSI terminal. The exact definition may also vary between versions of products (SCO has had more than one version of SCO ANSI, and even the rudimentary ANSI support in the ANSI.SYS driver in DOS varies between versions). If your terminal emulation program doesn't specifically mention that it emulates a SCO ANSI terminal, chances are that it's designed to work like ANSI.SYS, and that's not sufficient for SCO ANSI. Many terminal emulation programs have a specific SCO ANSI setting; check with your documentation or contact the vendor. In some cases (particularly via telnet or rlogin), the terminal type you're using is transmitted as part of the connection sequence. Make sure that the terminal type your communications software is reporting is the same as what SCO expects. For example, many programs call their SCO ANSI emulation "SCOANSI", but SCO calls it "ansi", and if your software sends "SCOANSI" as its terminal type, your SCO system will not understand. Many terminal emulation packages allow you to define what terminal type it will say it's using; set this to "ansi". ____________________________________________________________________________ What is Skunkware? ================== For the full story, see http://www.sco.com/skunkware/faq.html The brief summary: Skunkware is a collection of programs which people have compiled for SCO operating systems. These may have been ported by SCO staff on their own time, by their authors, or by anyone else. None of them are supported by SCO. The contents of all Skunkware discs are available from ftp.sco.com. The current version may also be ordered on-line from http://www.sco.com/orders/. ____________________________________________________________________________ Can I replace csh with tcsh? ============================ Yes. SCO's csh is both ancient and broken, and many people don't use it. Note that for many systems there's a specific warning in your manual that you should not use csh as root's shell. Consider replacing it with tcsh, available from a number of places including Skunkware (ftp://ftp.sco.com/skunkware/). There are no known problems in simply replacing /bin/csh with tcsh, with the possible exception SCOAdmin Software complaining about an incorrect checksum when verifying your software. ____________________________________________________________________________ Is my system Year 2000 compliant? ================================= You'll need to check out everything about your system - your hardware (including the BIOS and the motherboard's real-time clock), your operating system, any add-on utilities you may have, and your applications. A brief summary of information gathered from SCO follows; for more detail, and possibly more up-to-date information, please visit SCO's site. Please note that the information below is not guaranteed to be accurate. For official Year 2000 information relating to SCO products, contact SCO. o SCO's Y2k Web site is http://www.sco.com/year2000/ o UnixWare 7 requires the application of the 7.0.1 release supplement. o UnixWare 2.1 requires the application of PTF3015. o UnixWare 2.03 requires the application of PTF2243. o OpenServer 5.0.5 requires the application of patch oss600a. o OpenServer 5.0.2 and 5.0.4 require the application of patch oss601a. o OpenServer 5.0.0 is not compliant but there is an unsupported patch, oss603a, available from SCO. Visit http://www.sco.com/support/osr5_query.html for downloading information. o SCO Unix 3.2v4.x require the application of SLS uod426d. 3.2v4.2, with this patch, qualifies for SCO's Date Processing Limited Warranty; earlier releases of 4.x will take this patch but are not covered by the warranty. o SCO Unix releases 3.2v2.0 and 3.2.0 are not officially listed; however, the documentation for uod426d states that while this supplement has not been extensively tested on these releases, it "should still work as expected" on them. Obviously, there is no warranty on this. o SCO Open Server 3.0 and OpenDesktop 3.0 require the application of SLS uod426d, and are covered by the Date Processing Limited Warranty. Earlier Open Server and OpenDesktop releases are in the same category as the Unix releases on which they are based - uod426d is available and should work, but it has not been extensively tested on them and this combination carries no warranty of any kind. o Xenix 2.3.2 through 2.3.4 require the application of SLS xnx427d. o There are no patches for Xenix versions prior to 2.3.2.