NetNews Usenet Archive 1992 #20

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

/ NetNews Usenet Archive 1992 #20 / NN_1992_20.iso / spool / comp / unix / xenix / sco / 2835 < prev next >

Wrap

Internet Message Format | 1992-09-08 | 51.8 KB

Path: sparky!uunet!elroy.jpl.nasa.gov!swrinde!news.dell.com!natinst.com!balkan!chinacat!chip From: chip@chinacat.unicom.com (Chip Rosenthal) Newsgroups: comp.unix.xenix.sco Subject: Welcome to comp.unix.xenix.sco [monthly posting] Summary: answers to frequently asked questions Message-ID: <sco-xenix-920905@chinacat.unicom.com> Date: 6 Sep 92 00:20:10 GMT Expires: Thu, 5 Nov 1992 00:20:11 GMT Organization: Unicom Systems Development, Austin, TX Lines: 1248 Supersedes: <sco-xenix-920807@chinacat.unicom.com> X-Faq: sco-xenix Archive-name: sco-xenix Version: 1.33 @(#) sco-xenix 1.33 92/09/05 18:52:02 Welcome to comp.unix.xenix.sco. This newsgroup is about the XENIX operating system from the Santa Cruz Operation (SCO). If you have an SCO XENIX question, then you've come to the right place. This is NOT the newsgroup for SCO UNIX or Open Desktop questions. Please use comp.unix.sysv386. It is also NOT the right place for questions about XENIX implementations from other vendors. Use comp.unix.xenix.misc. Certain questions are asked frequently on the net. Frequently asked questions (FAQs) get old and boring after a while. We can stamp out FAQs in your lifetime! Please review this message before posting your question to comp.unix.xenix.sco. You should also review the FAQ postings in comp.unix.questions and news.announce.newusers. These answer general questions, such as `How do I delete a file with funny characters in the name?' and `How can I change my shell prompt?' C programmers should also check out the FAQ in comp.lang.c. Whenever we talk about XENIX here, we really mean SCO XENIX System V. Our comments often won't apply to the other kinds of XENIX. Also, be particularly careful if you use XENIX/286. It's been ages since we've used it and some of our suggestions might not apply. This message is maintained and posted monthly by Chip Rosenthal. Your corrections, additions, and comments are encouraged. Please mail them to chip@chinacat.unicom.com or uunet!chinacat!chip. These questions and answers were written by Chip Rosenthal, Jeff Liebermann, Ronald Florence, Chip Salzenberg, and Ronald Khoo. ****************************************************************************** **************************** Questions Answered **************************** ****************************************************************************** ===> Product Questions: QA1. What is the difference between UNIX and XENIX? QA2. What is the minimum system required to run XENIX? QA3. Should I buy UNIX or XENIX? QA4. What are the current XENIX versions? QA5. Where can I get XENIX fixes? QA6. I'm using an older XENIX/386 version. What fixes do I need? QA7. Does SCO have a BBS? ===> General XENIX Questions: QB1. Can't boot. My system used to work. QB2. Configure(ADM) fails when I modify my system name. QB3. I upgraded XENIX. Now I get `parity error' panics. QB4. Login fails without prompting for password. QB5. Can I use `gcc' instead of buying the SCO Development System? QB6. Program `foobar' dumps core with a stack overflow. [XENIX 286] QB7. Spell cannot initialize hash table. QB8. Will my laser printer work with the Text Processing Package? QB9. Where is crypt? ===> Disk and Filesystem Questions: QC1. Can I use a disk bigger than 1024 cylinders? QC2. Why won't XENIX recognize my SCSI disk? QC3. Where is the `/u' filesystem I created? QC4. `WARNING!! - mounting: <usr> as <news>' message when booting. QC5. `POSSIBLE FILE SIZE ERROR' when running `fsck'. QC6. Is disk fragmentation curable? ===> Communications Questions: QD1. How do I add more than 2 COM ports? QD2. Where is the `gettydefs' entry for 19200 serial lines? QD3. How can I have more mscreen(C) sessions? QD4. Upgraded to 2.3.4 and flow control broke with mscreen(C). QD5. Terminals act funny, modems drop characters, help!! QD6. Does XENIX support 16550A UARTs? QD7. How do I increase the number of `clists'? QD8. How do I change the 16550A trigger level? QD9. How do I adjust the `ttyhog' value? QD10. Where can I find TCP/IP? NFS? QD11. BSD based uucp can't connect to my machine. QD12. uucp truncates my system name to seven characters! QD13. What is the uucp `windows' patch? QD14. uucp gives `alarm' messages, eventually dies. ===> Display and Graphics Questions: QE1. How can I do graphics under XENIX? What programs are available? QE2. CGI graphics mangles grey scale on a mono VGA. QE3. CGI graphics fail on a LaserJet in landscape mode. QE4. How can I do a printscreen under XENIX? ===> XENIX and MS-DOS Interoperability Questions: QX1. Can XENIX share a hard disk with MS-DOS? QX2. Can XENIX access MS-DOS disks? QX3. Can XENIX execute MS-DOS programs? QX4. Can MS-DOS execute XENIX programs? ===> The (in)Famous XENIX Cheat-Sheet: QZ1. What the $%*&! is happening? ****************************************************************************** **************************** Product Questions ***************************** ****************************************************************************** QA1. What is the difference between UNIX and XENIX? XENIX is UNIX -- or at least one flavor of it. In the late 70's, Microsoft licensed the UNIX sources from AT&T and ported them to a number of platforms. In those days AT&T would license the UNIX sources but not the UNIX name, thus each company had to invent their own name. Microsoft picked XENIX. Microsoft did not sell XENIX to end users. Instead, they licensed the software to OEMs (Intel, Tandy, Altos, SCO, etc.) who provided a finished end-user package. Microsoft no longer supports XENIX, and in fact never even offered a 286 or 386 version. Several UNIX implementations for the PC architecture have been tried with varying levels of success. SCO XENIX for the PC/XT was one. Unfortunately, the PC/XT lacked hardware features required for robust UNIX operations. The PC/AT offered hardware memory protection, and SCO XENIX/286 took advantage of it. SCO XENIX/386 added demand paged virtual memory. These added features made multiuser PCs viable, and SCO XENIX popular. SCO XENIX starts with a UNIX System III base, throws in several Berkeley enhancements, and adds features to obtain conformance to the System V Interface Definition (SVID). Today, the bulk of the code is from System V. XENIX/386 even has capabilities to execute UNIX programs. However, it differs in many of the SVID `optional' areas people tend to expect of a full System V. For example, SCO XENIX lacks a real `inittab'. You need to go to a real System V Release 3.2, such as SCO UNIX, for all these features. QA2. What is the minimum system required to run XENIX? For a minimal single-user XENIX system, we recommend an 80386SX-16 processor, 2MB to 4MB RAM, and 40MB hard disk. For the full XENIX system (Run Time, Development, and Text Processing), consider an 80386DX-20 processor, 4MB RAM, and 80MB hard disk. Additional resources will be required to support additional users, large applications, networking, or X windows. XENIX/286 will run on a PC/AT, but you will quickly become frustrated with its limitations, especially if you are a programmer. QA3. Should I buy UNIX or XENIX? If you are running XENIX now, don't throw it away and buy UNIX. XENIX will be around for some time. However, if you are new to multi-user computing, you should consider UNIX System V Release 3.2 or Release 4.0. XENIX is a mature product. There will be few future enhancements. UNIX has more goodies and better performance than XENIX. UNIX runs XENIX binaries, so it will do most anything XENIX can -- and more. XENIX requires less memory and disk space. However, hardware prices are declining and UNIX is now pretty stable, so these are not great advantages. On the other hand there are a number of XENIX enthusiasts who wouldn't switch if you gave them a free copy of UNIX. They would gladly forgo the new features to avoid the bloat of System V. QA4. What are the current XENIX versions? Package Version `perms' file ---------------------- -------------- --------------------------- XENIX Runtime System 2.3.4 inst, dsmd, tpmd, rts, ext XENIX Runtime (286) 2.3.2 inst, dsmd, tpmd, rts, ext Development System(386) 2.3.1 soft Development System(286) 2.2.1 soft CGI 1.1.0 cgi Text Processing 2.3.0 text Manual Pages 2.3.0 man VP/ix 1.2.0 vpix TCP/IP Runtime 1.0.1 tcprt TCP/IP Development 1.0.1 tcpdev Streams Runtime 1.0.0 streamsrt Streams Toolkit 1.0.0 streamstk Int'l Supplement(386) 2.1.0 sup.os Int'l Supplement(286) 2.0.0 sup.os The perms column lists files found in the /etc/perms directory. The `#rel=' line in these files will tell you what version you have. QA5. Where can I get XENIX fixes? SCO provides Support Level Supplements (SLS) at no charge. They may be downloaded by anonymous uucp from SCO's `sosco' machine. To retrieve a shopping list of available SLSs, add one of the following to your `/usr/lib/uucp/Systems' file: # Telebit (PEP) modem sosco Any ACU 19200 14084291786u gin:-BREAK-gin:-BREAK-gin: uusls # Hayes V-Series modem sosco Any ACU 19200 14084274470 gin:-BREAK-gin:-BREAK-gin: uusls # 1200 or 2400 bps modem sosco Any ACU 300-2400 14084253502 gin:-BREAK-gin:-BREAK-gin: uusls European folks might prefer to use `scolon' instead. From within the UK, try: scolon Any ACU 300-19200 0923210911u \ gin:-BREAK-gin:-BREAK-gin: uusls word: bbsuucp Outside the UK, replace the leading zero with the country code. A 9600bps Trailblazer answers this line. Delete the `u' at the end of the phone number if you aren't using a Telebit modem. In Germany, you can use: sosffm Any ACU 300-9600 069-50910736 \ "" \r ogin:-BREAK-ogin: uusls sword: bbsuucp This modem does support V.32 transfers, but not PEP. Files on the European machines are as on `sosco' but tend to be a week or so behind. Also, replace `sosco' with `scolon' or `sosffm' in the following examples. Add to your `/usr/lib/uucp/Permissions' file: # REQUEST files only - i.e. cannot send to sosco. # rmail required to get bounced mail and error messages. MACHINE=sosco SENDFILES=yes REQUEST=no COMMANDS=rmail:uucp \ WRITE=/usr/spool/uucppublic:/tmp To retrieve an SLS listing to your machine as /tmp/info: uucp sosco\!/usr/spool/uucppublic/SLS/info /tmp/info This file also provides directions on retrieving, unarchiving, and uncompressing the SLSs. There is also a `descriptions' file with additional details on the available SLSs. SLSs are also available on `uunet'. For example, to get the `info' file, run: uucp uunet\!~/vendor/sco/SLS/info /tmp/info Most of these machines also have Enhanced Feature Supplements (EFSs), games and graphics programs, and new termcap and terminfo terminal definitions in addition to SLS. Look for an `info' or `README' file one directory up from the `SLS' directory for more info. QA6. I'm using an older XENIX/386 version. What fixes do I need? Development System - 2.2.X or older An upgrade to 2.3.1 must be purchased from SCO. - 2.3.1 Install `lng244'. Operating System - 2.2.X or older An upgrade to 2.3.4 must be purchased from SCO. - 2.3.0 or 2.3.1 SCO would like you to purchase a 2.3.4 upgrade. The `xnx137' fix would get you to 2.3.2, but SCO does not provide it any longer. - 2.3.2 Install `xnx296a' to get to version 2.3.3. If you have VP/ix and installed update UFJ, then your kernel is already at 2.3.3. None the less, you still want `xnx296a'. The `xnx296a' update is also known as UFN, and replaces UFJ and UFM. UFM was also known as the `xnx155b' update. You must install the development system before `xnx296a'. `xnx296a' provides you with most of the important fixes in 2.3.4. You will lack some of the nice enhancements such as updated man pages and the Korn shell. SCO preconfigures the kernel for a very minimal system. You should tune the kernel for your exact system configuration. See the "Tuning Kernel Parameters" chapter in the System Administrator's Guide. QA7. Does SCO have a BBS? Yes. The SOS Support BBS is $95/year. This is a good deal. You can: - download fixes, suggestions, and hints. - keyword access to the support problems database. - retrieve product info, release status, and hardware compatibility. - get support via email. - subscription to quarterly Discover newsletter. You can dial-in with a terminal program such as `cu' -- uucp is not required. Access is read-only, but most questions can be answered by the support database. SCO currently doesn't provide SOS access to most overseas users. You might also consider joining the ``SCO Developer Alliance Associates Program.'' This costs $100/year, and also provides access to the BBS. Contact devrel@sco.com for additional information. ****************************************************************************** ************************* General XENIX Questions ************************** ****************************************************************************** QB1. Can't boot. My system used to work. Check the A: drive. If there is a floppy in there your system might spew garbage on the monitor or hang during boot. If this problem occurs right after rebuilding a new kernel, go read the FAQ on disks bigger than 1024 cylinder. QB2. Configure(ADM) fails when I modify my system name. If you select a site name of two, three, or four characters, `configure' will fail with an error message: space.asm(XXX): error A2050: Value out of range If the development system is installed, there is a workaround. Do the following: - cd /usr/sys/conf - Edit `xenixconf'. Search for the `nodename' line and enter the desired system name between the double quotes. For example, for machine name = `foo' this line should read: nodename "foo" - Similarly, edit the `master' file and correct the `nodename' line. For machine name = `foo' it should read: nodename NODE "foo" - Run `make'. This will rebuild the kernel data files and then link a new kernel. - Run `./hdinstall' to install the new kernel. Reboot the system, and the new node name should be in effect. To verify this, run `uname -a' after rebooting. This bug has been fixed in 2.3.4. QB3. I upgraded XENIX. Now I get `parity error' panics. Starting with XENIX 2.3.3, memory parity errors are detected by the kernel. In previous versions, parity errors would cause the system to go silently flaky, but continue running. Now, when a parity error occurs, the machine will panic saying: PANIC: memory failure - parity error If you get this error after installing an update, XENIX didn't break your machine -- it's been broken all along. The most likely cause of this error truly is memory problems. Run a good memory diagnostic overnight to try to locate the problem. Keep the cover on the computer while running diagnostics; heat tends to push a marginal memory chip into failure. This problem can also be caused by a marginal power supply. XENIX boxes often have more power hungry peripherals than your run of the mill MS-DOS PC. For example, suspect the power supply if the panics occur as soon as the tape drive starts spinning. QB4. Login fails without prompting for password. If /dev/tty is missing or incorrect, you will get a `Login incorrect' message without ever being prompted for a password. For security reasons, the `login' program goes directly to /dev/tty to ensure the password comes from the keyboard. Your `tty' device should look something like: crw-rw-rw- 2 bin bin 3, 0 Oct 24 04:09 /dev/tty If necessary, use the `mknod' command to recreate it. QB5. Can I use `gcc' instead of buying the SCO Development System? No. It takes more than a compiler to produce executable programs. You might be able to pick up bits and pieces (like `make' and `yacc') from various places, but not the header files and libraries. You can't compile programs without them, and the only way to get them is with the Development System. QB6. Program `foobar' dumps core with a stack overflow. [XENIX 286] XENIX/286 programs request a fixed amount of space for the stack in the executable file header. A program that exceeds this limit will crash with a core dump. To fix, change the header to ask for more stack space. You can recompile your programs with the `-F' flag to increase the stack size. You can change the stack size of existing programs by running `fixhdr' with the `-F' flag. This is not a problem on 80386 binaries that use a variable stack. QB7. Spell cannot initialize hash table. The `spell' script (included in the optional Text Processing package) uses an 8086 binary. It needs to stuff a bunch of data into a single 64K segment. If you exceed this limit, you will get an error saying something like: table = malloc(0xc848) fails sbrk(0) = 0x3868 spell: cannot initialize hash table You can usually get enough space for `spell' to run by clearing out your environment first. To do this, add the following line near the top of the `/usr/bin/spell' script: unset `env | sed -e '/^PATH=/d' -e '/^PS1=/d' -e 's/=.*//'` QB8. Will my laser printer work with the Text Processing Package? XENIX `troff' is based on a version written long before the advent of laser printers. The only thing old troff understands is C/A/T typesetters. Chris Lewis's `psroff' allows old troff to work with various devices, including Postscript and LaserJet printers. Version 3.0 is available from the comp.sources.unix archives. For production work, consider purchasing a device independent troff. Some of the suppliers of `ditroff' are Elan, Image Network, Leverage, and SoftQuad. Another alternative is the `groff' package by the Free Software Foundation. You can find it in any of the FSF archives -- but you will need `g++' to compile it. You will still need a backend program to use a LaserJet with `groff'. Some possibilities are `psroff' and `dvi2xx' on ftp.wu-wien.ac.at. QB9. Where is crypt? The DES encryption algorithm, the basis for `crypt', is categorized by the US Government as munitions. (Three cheers for the cold war!) SCO removes the crypt(S) procedure and crypt(C) command from XENIX so it may be exported. US domestic customers may contact SCO to obtain SLS LNG190B. If you simply need a crypt(S) procedure to compile programs that do password checking, you may get the SLS LNG255B `International Crypt SLS' from any of the aforementioned archive sites. This provides a crypt(S) procedure that supports encryption but not decryption, and it does not include the crypt(C) command. ****************************************************************************** ********************** Disk and Filesystem Questions *********************** ****************************************************************************** QC1. Can I use a disk bigger than 1024 cylinders? Yes. However, your root filesystem cannot extend beyond the 1024th cylinder, because that's as far as the BIOS can reach when booting the system. The BIOS is bypassed after XENIX boots, so cylinders beyond 1024 can be accessed once XENIX is running. It is not necessary to use a sector translation mode controller to reduce the number of cylinders. The translation mode is very slow. Early SCO documentation wrongly claimed large disks weren't supported. This is the source of confusion over this issue. QC2. Why won't XENIX recognize my SCSI disk? Not all versions of XENIX support SCSI and those that do won't run all SCSI adapters. First, verify your N1 floppy says `Type: 386GT' and not `Type: 386AT'. The AT version of XENIX (now obsolete) does not include any SCSI support. The GT version supports a limited number of SCSI host adapters. The Adaptec AHA-154x is supported. We've also used compatibles such as the Bustek BT-542B. 2.3.4 also supports WD1009V-SE1/2 and WD-7000-FASST2. If you've got something different, you probably need to get a driver from the board vendor. The `SCO Hardware Compatibility Guide' (contact SCO sales for one) should be consulted whenever you have any hardware compatibility questions. Some typical installation mistakes are: - SCSI ID jumper for the root (boot) hard disk must be ID=0. - The BIOS must be set for drive C:=NONE in the CMOS setup program. - I/O port, memory address, or interrupt misconfiguration/conflict. - Terminating resistors installed only at ends of ribbon cable. - The parity jumper should be enabled on all drives and controllers. A word on terminating resistors: Terminators must be installed on the two devices at the end of the SCSI chain. If you have no external devices then one end of the chain will be the SCSI adapter -- so it gets terminators. The last device on the ribbon cable also needs terminators. The terminators must be removed from everything else on the SCSI chain. Note: If you are using XENIX 386GT 2.3.1-2.3.3 you should install xnx252b for ISA or xnx150 for MCA. This is an improved SCSI driver that yields a significant performance improvement. QC3. Where is the `/u' filesystem I created? When you create filesystems during the install procedure or with `mkdev hd', the filesystem is initialized but it isn't configured into the system. To perform that final step you need to run `mkdev fs'. For example, if you created a `/u' filesystem, run: mkdev fs /dev/u /u QC4. `WARNING!! - mounting: <usr> as <news>' message when booting. Due to a bug in the `mkdev fs' program, filesystems mounted below the root directory are labeled wrong. For example, if you mount a device `/dev/news' onto directory `/usr/spool/news', then you will get a warning message when going into multi-user mode. That's because `mkdev fs' labels the filesystem with the first component in the pathname (`usr' in this example) when it really should be the last component in the pathname (`news' in this example). This message is merely an annoyance and will not hamper system operation. To eliminate the warning, relabel the filesystem with the last component of the mount directory. For the above example, you may change the label from `usr' to `news' by running the command: fsname -s news /dev/news QC5. `POSSIBLE FILE SIZE ERROR' when running `fsck'. If `fsck' says `POSSIBLE FILE SIZE ERROR I=nnnn' don't panic. It almost always means `fsck' found a `sparse file', not a true filesystem error. A sparse file has holes in it. They are created when a program skips over sections of a file it is writing. The XENIX filesystem fills in those sections with zeros. The XENIX filesystem doesn't even both allocating disk blocks for the sections full of zeros, thus creating a sparse file. The complaint occurs because `fsck' calculates the number of disk blocks needed to store the file, counts up the number of disk blocks actually used by the file, and complains that the two don't match. When you get a file size error from `fsck' you should note both the `I=nnnn' inode number logged in the message and the filesystem which was being checked. You can see what the file is by running the following command: ncheck -i <nnnn> <fsname> where <nnnn> is the number logged in the error message and <fsname> is the pathname of the filesystem device, e.g. `/dev/u'. Chances are the file will either be a dbm(S) database file (in which case ignore the complaint, it's expected) or a `core' dump (in which case delete the `core' file). QC6. Is disk fragmentation curable? Not easily. However, the steps to slow down fragmentation are pretty simple. XENIX maintains a list of available disk blocks. As files are deleted, the released blocks are appended to this free list. Ideally, one would allocate adjacent blocks for new files. However, blocks are added to and removed from the free list without regard for location. Eventually, new files will be created with their data blocks scattered all around the disk. This fragmentation is undesirable because it slows down disk accesses. To slow down the tendency towards fragmentation, periodically reorder the blocks in the free list. To do this, dismount the filesystem (with `/etc/umount') and run `fsck -S'. You will get a message saying: FILE SYSTEM NOT MODIFIED, STILL DIRTY. Don't worry -- ignore that message. To determine how badly your filesystems are fragmented, run `fsanalyze' by Michael J. Young, from comp.sources.misc. You will need patchlevel 3 to analyze XENIX filesystems. The conventional cure for filesystem fragmentation is: - Make a full backup of the filesystem. Use `cpio' (or `afio' if available). `dump' and `volcopy' do an image backup that will restore in a fragmented manner. `tar' does not backup device nodes, pipes, or directories. - Recreate an empty filesystem with `divvy' or `mkfs'. `divvy' is easier to use -- simply select the `create' option. However, `mkfs' will be required if you want a non-default number of inodes. Invoke `divvy' with: /etc/divvy -c 1 -b 1 # for 1st physical disk /etc/divvy -c 1 -b 1 -p 1 # for 2nd physical disk - Reload the filesystem from the backup. - Regenerate the `lost+found' directory. Run: # Substitute appropriate name for `foo'. /etc/fsck /dev/foo /etc/mount /dev/foo /foo mkdir /foo/lost+found for i in 1 2 3 4 5 6 7 8 ; do for j in 1 2 3 4 5 6 7 8 ; do touch /foo/lost+found/x$i$j done done rm /foo/lost+found/x?? /etc/umount /dev/foo Be very careful with this procedure. We suggest you verify your backup is readable before zapping the filesystem. One mistake (especially when running `mkfs' or `divvy') will trash everything. The filesystem you trash may be your own. ****************************************************************************** ************************* Communications Questions ************************* ****************************************************************************** QD1. How do I add more than 2 COM ports? Dumb serial cards (and internal modems) may be configured for XENIX as COM3 and COM4 -- with one big caveat. Each COM port requires its own interrupt number. If you have an unused IRQ and the Development System, you may build additional COM ports into the kernel. To illustrate the procedure, we will add a COM3 configured as follows: I/O Addr: 0x3E8 IRQ: 5 device: /dev/tty3a Major=5, Minor=16 /dev/tty3A Major=5, Minor=144 1. Edit `/usr/sys/io/sioconf.c'. The two lines in this file that describe COM1 and COM2 are: {0,IBM_BOARD, 1,4,0, (sd)0x3f8,0, 0,MCRBIT3}, /*ibm COM1*/ {1,IBM_BOARD, 1,3,8, (sd)0x2f8,0, 0,MCRBIT3}, /*ibm COM2*/ Add a line that says: {2,IBM_BOARD, 1,5,16, (sd)0x3e8,0, 0,MCRBIT3}, /*ibm COM3*/ See `/usr/sys/io/sioconf.h' for details. 2. Edit `/usr/sys/conf/master'. The serial ports are defined by the line: sio 4 0577 104 sio 0 0 5 1 7 3 4 33 34 The last four fields define the interrupt numbers used. Change the `33' to a `5' since COM3 will be on IRQ5. For reference, here is the mapping between IRQ lines and the master file `magic numbers': IRQ0 0 IRQ4 4 IRQ8 30 IRQ12 34 IRQ1 1 IRQ5 5 IRQ9 31 IRQ13 35 IRQ2 31 IRQ6 6 IRQ10 32 IRQ14 36 IRQ3 3 IRQ7 7 IRQ11 33 IRQ15 37 3. Rebuild the kernel. Run: cd /usr/sys/conf make Do not use `link_xenix' here. You need to run `make' to get `sioconf.o' and the kernel data files rebuilt. 4. Install the new kernel. Run: /usr/sys/conf/hdinstall This will make a backup of your current kernel to `/xenix.old'. 5. Create the device nodes. Run: mknod /dev/tty3a c 5 16 mknod /dev/tty3A c 5 144 Reboot your system, and the new serial port should be there. QD2. Where is the `gettydefs' entry for 19200 serial lines? For historical reasons, terminal speeds of 19200 and 38400 are called `EXTA' and `EXTB', not `B19200' and `B38400'. Entries `n' and `o' in the `/etc/gettydefs' file are for 19200 and 38400, respectively. Once logged in you can change the line speed with the `stty' command: stty 9600 # set to 9600 bps (bits per second) stty exta # set to 19200 bps stty extb # set to 39400 bps A smart serial card is recommended for these two highest speeds. QD3. How can I have more mscreen(C) sessions? The mscreen(C) facility provides multiple sessions from a serial terminal by `multiplexing' your terminal (tty) line onto several pseudo-terminal (pty) lines. XENIX is shipped with eight ptys enabled, that provides a system-wide limit of eight sessions through the mscreen(C) command. First off, you need to determine whether you are allocating mscreen(C) sessions wisely. For example, SCO ships an `/etc/mscreencap' file that runs six sessions on a WY-60 terminal. If you edit `mscreencap' and cut this down to three you immediately double the number of people who can run mscreen(C). You cannot increase this number beyond eight if you are running TCP/IP. (More about this in a moment.) If you are not running TCP/IP, then you can increase the number of possible mscreen(C) sessions by: first creating more ptys in the kernel, second creating the device nodes for the new ptys, and finally placing a `getty' upon each of these new ptys. To create more ptys in the kernel, increase the `nspttys' kernel parameter. With XENIX 2.3.4, run the `configure' command interactively and select the `Multiscreens' option. For older versions of XENIX, run `configure' manually. For example, to provide sixteen ptys, run: cd /usr/sys/conf ./configure nspttys=16 Once `nspttys' is set, run `link_xenix' and `hdinstall' to build and install a new kernel. Next, you need to create the device nodes in the /dev directory for these additional ptys. Unfortunately, mscreen(C) uses a very poor naming strategy for ptys beyond the 8th. Once you go beyond the /dev/ttyp7 in the base XENIX distribution, mscreen(C) next looks for not /dev/ttyp8, but rather /dev/ttyp01. Therefore you should create the eighth pty with the commands: cd /dev mknod ttyp01 c 54 8 mknod ptyp01 c 55 8 chmod 666 ptyp01 chmod 644 ttyp01 You might notice that these names are the same ones used by `rlogind' and `telnetd' under SCO TCP/IP. Unfortunately, to use these ptys for mscreen(C) you need to place a `getty' on them -- but if you do then rlogin/telnet will break. Those running TCP/IP (and not weak of heart) can try editing the `/dev/ttyp' string in the `mscreen' binary and totally rename the pty devices. Finally, edit `/etc/ttys' (and `/etc/ttytype') to tell them about the new pty lines, and run `enable' upon each of the lines. QD4. Upgraded to 2.3.4 and flow control broke with mscreen(C). The 2.3.4 version of mscreen(C) was changed to run the tty line totally raw. This is required so that the ^S and ^Q keys are passed onto programs that need them, such as `emacs' and Foxbase. An unfortunate side effect of handling flow control on the pty side rather than the tty side is that response to flow control is a lot more sluggish. Now, when you hit ^S you might get another screenful of data before the display actually freezes. If you never run any applications that require full 8-bit transparency, one way to make flow control more responsive is to rename `mscreen' to `/usr/bin/mscreen.sco' and install the following script in its place: : tty=`tty` ( sleep 5 ; stty ixon <$tty ) & exec mscreen.sco QD5. Terminals act funny, modems drop characters, help!! If you are losing characters on your serial lines, consider the following steps: * Ensure that flow control is working right. * If you are running dumb COM ports, either upgrade the UART chips to 16550As or switch to a smart serial card. * If you have a lot of users or lots of high speed data lines, increase the number of `clist' buffers in the kernel. * If you upgraded to 16550A UARTs and this helped but did not solve the problem, then adjust the FIFO trigger level. * If you are running high speed lines on a loaded system, adjust the `ttyhog' value. When diagnosing serial problems, start at the top of this list and work down. If your flow control is broke, it will do no good to go poking around inside your XENIX kernel. QD6. Does XENIX support 16550A UARTs? The guts of a standard COM port is a UART (universal asynchronous receiver/transmitter) chip. Most serial ports use the 16450 or an equivalent chip. The 16550A is a pin-compatible, drop in replacement which adds 16-characters of on-chip FIFO buffering. The FIFOs reduce system loading and increase performance. Starting with 2.3.4, the `sio' driver recognizes 16550A UARTs and enables the on-chip FIFOs. Older versions of XENIX merely treat this chip as an expensive 16450. You can get 16550A support with any version of XENIX by installing the FAS driver written by Uwe Doering and based upon Jim Murray's `asy' driver. FAS was published in alt.sources, and is available at most reputable archive sites. We recommend sticking with the National Semiconductor NS16550A. Many manufactures make usable 16450 clones, but some of the 16550 compatible chips aren't so compatible. Also, older versions of the 16550 are known to be broken. Ensure yours have an `A' suffix. QD7. How do I increase the number of `clists'? XENIX buffers terminal characters in data structures called `clists'. There is a fixed pool of clists on the system, and when that supply is exhausted incoming characters are dropped. Therefore, the more active users you have on your system, the larger this pool of clists should be. The default configuration is 100 clists. Each terminal line requires an average of 5 to 10 clists. High speed modems need more. A conservative strategy would be to allocate 15 clists per active terminal. For example, if you have 16 serial ports and 4 console screens active at once, allocate (16+4)*15 or 300 clists. To see the number of clists on your system, login as root and run: cd /usr/sys/conf ./configure -y NCLIST To change the number of clists, say to 300, login as root and run: cd /usr/sys/conf ./configure NCLIST=300 then rebuild and install a new kernel and reboot. QD8. How do I change the 16550A trigger level? If installing a 16550A UART helped but didn't totally fix a lost character problem, and your system is heavily loaded, then there is something else you can try. If you installed a 16550A and there was no effect at all, then skip this -- it won't help. The XENIX `sio' driver initializes the 16550A with a trigger level of 14. This means that under continual data streams the UART will wait until 14 characters accumulate before telling XENIX to come get them. This gives the system two character times to get in and service the COM port. If XENIX does not service the port in that time the 16-character buffer will overflow and and characters will be dropped. This might happen on a heavily loaded system. The trigger level is initialized at boot time from the 16-byte `sio_fifoctl[]' array. This array specifies the UART control register initialization values for device minor numbers 0 through 15. Or more exactly, the minor number is ANDed with 0x0F to get the index into this array. For example, the initialization for /dev/tty2a (minor=8) is in `sio_fifoctl[8]'. The possible values are as follows: sio_fifoctl: 0xCF 0x8F 0x4F 0x0F trigger level: 14 8 4 1 The standard value is 0xCF. A smaller value will notify the system earlier that characters are waiting, thus providing the system more time to service the port, but increasing the number of interrupts generated by the port. As an example, here is how to change the trigger level for /dev/tty2a (and /dev/tty2A) to 8 characters (code=0x8F): cp /xenix /xenix.save # make a backup! adb -w /xenix # use "/etc/_fst" if no "adb" * sio_fifoctl+8/x sio_fifoctl+0x8: 0xcfcf | | | `---- This is the 8-bit value for the | port we want (minor=8). | `------ This is the 8-bit value for the next port (minor=9). * sio_fifoctl+0x8/w 0xcf8f sio_fifoctl+0x8: 0xcfcf= 0xcf8f * $q QD9. How do I adjust the `ttyhog' value? Since XENIX has a fixed pool of clists, the kernel tries to prevent one tty from hogging all of them. The limit is defined by `ttyhog', and defaults to 256 (0x100 hex). This limit might be too low if you've got some particularly fast devices on a particularly slow system. To change the `ttyhog' value you must patch the kernel image. The following increases `ttyhog' to 512 (0x200 hex). cp /xenix /xenix.save # make a backup! adb -w /xenix # use "/etc/_fst" if no "adb" * ttyhog/x _ttyhog: 0x100 * ttyhog/w 0x200 _ttyhog: 0x100= 0x200 * $q QD10. Where can I find TCP/IP? NFS? TCP/IP = many places. NFS = forget it. Some TCP/IP vendors are SCO, Novell (the old `Excelan Lan Workplace' package), and Network Research Corp. Our preference is for the SCO package. It is offered in two pieces: the runtime system and the development package. Please note that you need at least the SCO Streams Runtime option before installing TCP/IP. Many of the BSD4.3 networking utilities are provided, and the socket interface follows the BSD4.3 model. NFS for XENIX is not and will not be available. Newer versions of UNIX provide either a `vnode' (BSD UNIX) or `filesystem switch' (System V 3.2 and beyond) mechanism, that supports new filesystem types. With XENIX, the filesystem support must be coded right into the kernel. The task of merging NFS code with VP/ix hacks, XENIX-net hacks, etc. is more than can be justified. QD11. BSD based uucp can't connect to my machine. BSD based uucp sends even parity. XENIX uucp expects no parity. Add: "" P_ZERO to the chat script in their (not your) `Systems' or `L.sys' file. For example, the following works on a Sun 3: xnxbox Any uucp 19200 5553333 "" P_ZERO "" \r in:--in: nuucp QD12. uucp truncates my system name to seven characters! This misfeature was introduced in the xnx155b supplement. The story we hear is that SCO broke their `uucp' to cater to systems that choke on long system names. There is a workaround. For system name = `verylongname', add to every record in the /usr/lib/uucp/Permissions file: MYNAME=verylongname Next, rename /usr/bin/uuname to /usr/bin/uuname.sco. Replace /usr/bin/uuname with: : if [ $# -ne 1 -o "X$1" != "X-l" ] ; then exec /usr/bin/uuname.sco $@ exit 1 fi head -1 /etc/systemid exit $? This allows programs using `uuname' to receive the correct, untruncated name. QD13. What is the uucp `windows' patch? uucp transmits data in packets, usually 64 bytes apiece. An acknowledgement is required for every packet sent -- but not immediately. A window of unacknowledged packets is allowed. The default window size is three, that means `uucico' won't stop sending unless it falls more than three packets behind. A three packet window isn't big enough for very fast media and long delay media. The `uucico' binary can be patched to increase the window size by the following manual procedure: cd /usr/lib/uucp cp uucico uucico.old # make a backup copy!! adb -w uucico * windows/x # display windows value _windows: 0x3 * windows/w 7 # change it to seven _windows: 0x3= 0x7 * $q # done If you have XENIX 2.3.4 but not the Development System, use `/etc/_fst' instead of `adb'. If you can't find adb on your system, look on the N02 disk. This change increases the window size to seven -- the maximum possible. A similar change must be performed on the remote system's uucico because the actual window size is negotiated as part of the protocol startup. QD14. uucp gives `alarm' messages, eventually dies. If you run uucp in debugging mode (`uutry -x9') and see it get stuck like: alarm 1 send 37777777621 alarm 2 send 37777777621 alarm 3 send 37777777621 alarm 4 send 37777777621 and eventually die, this means your system got jammed waiting for a response that never came. The most common cause of this problem running uucp across a modem with XON/XOFF flow control enabled (e.g. a Telebit modem with S58=3). You can't do that; you must run with no flow control or RTS/CTS handshaking (on a Telebit that would be S58=0 or S58=2, respectively). Other possible causes of this problem are serial cards with broken handshaking or running out of clists. ****************************************************************************** ********************** Display and Graphics Questions ********************** ****************************************************************************** QE1. How can I do graphics under XENIX? What programs are available? There are two approaches to graphics programming. At the very lowest level, the screen(HW) manual page describes ioctl() codes to access the video display memory and controller. Device-independent graphics is provided by the Computer Graphics Interface (CGI) package included with the XENIX Development System. CGI drivers are provided for VGA, EGA, and Hercules displays; Epson, LaserJet, and Postscript printers; and HP plotters. Several programs have been posted to the net that run under CGI, such as gnuplot, starchart, and gif. See comp.sources.misc. QE2. CGI graphics mangles grey scale on a mono VGA. SCO fixed their monochrome VGA driver in version 2.3.3. The default color mapping in CGI maps many colors with the identical total intensity. They appear as the same shade grey-scale display. Try using the CGI functions vs_color() and/or vsc_table() to create a color map that works with your display. Try vsf_style() to create hatching or other fill-patterns in place of colors. QE3. CGI graphics fail on a LaserJet in landscape mode. A bug in the CGI 1.1.0 LaserJet driver causes all output to appear on a single line. In landscape mode the driver puts out a spurious escape sequence `\033&a864V' that forces all output to the same position on the page. You can either remove the offending escape sequence from the driver with a binary editor (keep a backup copy), or use a filter to clobber that sequence from the output. QE4. How can I do a printscreen under XENIX? MS-DOS users can produce a screen printout with one keystroke. XENIX does not offer this builtin capability. SCO's MultiView product provides this. Also, it can be simulated with the following script: : # @(#) prtscrn Print Screen # execute from cu as ~!prtscrn oldstty=`stty -g` # save stty stty -echo ixon ixoff -ixany # no echo /bin/echo '\033[2i\c' # send screen to host # grab 24 lines, clean up trailing blanks, print it sed -e 's/ *$//' -e 24q | lp stty $oldstty # put stty back when done. Another approach is the `prtscrn' program by Chip Rosenthal. It grabs the contents of any console screen and sends it to standard output. See comp.sources.misc. ****************************************************************************** *************** XENIX and MS-DOS Interoperability Questions **************** ****************************************************************************** QX1. Can XENIX share a hard disk with MS-DOS? Yes. You must: - Install MS-DOS 3.2 bootable partition as the FIRST partition. - Install XENIX bootable as the next partition. - Select XENIX as the `active' partition. At the Boot : prompt either type `dos' for MS-DOS boot or press ENTER for XENIX. If you create any other type of MS-DOS partition (e.g. one of the new larger sized MS-DOS 5.0 partitions) or forget to create the MS-DOS partition before loading XENIX, then you can still access the MS-DOS partition by booting from a floppy disk. QX2. Can XENIX access MS-DOS disks? Yes, with limitations. The commands on the `dos(C)' manual page provide access to MS-DOS disks (floppy and fixed) under XENIX. The limitation is that only conventional MS-DOS 3.2 hard disk partitions can be recognized. Extended MS-DOS partitions cannot. If you run into problems with the dos(C) commands, check that `/etc/default/msdos' provides proper aliases for the A:, B:, etc. devices. Also, if root can access the MS-DOS fixed disk but users can't, check the permissions on /dev/hd0d (the XENIX name of the MS-DOS partition on the first fixed disk). The `mtools' package by Emmet Gray in comp.sources.misc provides better handling of MS-DOS devices under XENIX. Modify the `devices.c' file to understand XENIX device naming. #ifdef M_XENIX struct device devices[] = { {'A', "/dev/install", 0L, 12, 0, (int (*) ()) 0, 0, 0, 0}, {'B', "/dev/install1", 0L, 12, 0, (int (*) ()) 0, 0, 0, 0}, {'C', "/dev/hd0d", 0L, 16, 0, (int (*) ()) 0, 0, 0, 0}, {'D', "/dev/hd1d", 0L, 16, 0, (int (*) ()) 0, 0, 0, 0}, {'\0', (char *) NULL, 0L, 0, 0, (int (*) ()) 0, 0, 0, 0} }; #endif /* M_XENIX */ If you assign `/dev/install' to the `A' drive, you don't need to worry about disk density when reading and writing. However, you need to specify the full device name when formatting. For example, to format a high-density disk in drive A: you must specify the full device name: dosformat /dev/fd096ds15 QX3. Can XENIX execute MS-DOS programs? Yes. You must purchase SCO VP/ix. VP/ix emulates an 8086 PC/XT running MS-DOS. The compatibility provided by VP/ix is fair with a significant performance penalty. VP/ix will not run programs that use 80286 or 80386 instruction codes. VP/ix is not available for XENIX/286. QX4. Can MS-DOS execute XENIX programs? No. With the XENIX Development System, you can compile programs to execute under MS-DOS with the `-dos' command line switch. ****************************************************************************** ********************* The (in)Famous XENIX Cheat-Sheet ********************* ****************************************************************************** QZ1. What the $%*&! is happening? Who am I? /usr/bin/who am i Where am I? /bin/pwd What am I doing? /bin/ps -f Am I having fun yet? /usr/games/worms -R What's my user/group id? /usr/bin/id Who did I login as? /usr/bin/logname What's my terminal? /bin/tty What are my terminal parameters? /bin/stty -a Is terminal flow control working? /usr/games/stars What is my current crontab file? /usr/bin/crontab -l Do I have uucp jobs queued? /usr/bin/uustat What is the system doing? /usr/bin/w How is the system running? /usr/bin/vmstat 10 Has the system been swapping? /usr/bin/vmstat -s Open files, inodes, processes? /bin/pstat | grep '^[0-9]' What are the kernel parameters? /usr/sys/conf/configure -x What versions are installed? egrep "#prd|#set|#rel" /etc/perms/* What are the hardware settings? /etc/hwconfig -h Is /etc/passwd correct? /etc/pwcheck Is /etc/group correct? /etc/grpcheck Is /etc/gettydefs correct? /etc/getty -c /etc/gettydefs Do all files have owners? /usr/bin/quot Who is using all the disk space? /usr/bin/quot When did joe last login? /usr/bin/last joe What files are owned by UID=944? find / -user 944 -exec l -d {} \; Are file perms/owners set OK? cd / ; /etc/fixperm -n /etc/perms/rts How about perms on uucp files? cd / ; /etc/fixperm -nd UUCP /etc/perms/rts Check uucp file consistency. /usr/lib/uucp/uucheck What are uucp access permissions? /usr/lib/uucp/uucheck -v What is the print spooler doing? /usr/bin/lpstat -t Stop the scheduler. /usr/lib/lpshut Restart a printer. /usr/lib/accept printer_name ; /usr/bin/enable printer_name [Following apply only if TCP/IP is installed.] Where am I? /usr/bin/hostname Is system foo alive? /usr/bin/ping foo What machines are connected? /usr/bin/netstat How is the network doing? /usr/bin/netstat 10 Where are the packets going? /usr/bin/netstat -r Network not work? /usr/bin/netstat -s [Thanks to Jeff Liebermann.] [ end of sco-xenix 1.33 ]