prev next index

Unix Notes
Updated April 10, 2002
Created May 31, 2001


Compaq Linux Web Site: -> ProLiant

Compaq's Open Source Web Site:

Other Good Links: - for great Linux intro / admin stuff - for all kinds of guides, HOWTOs, FAQs, and man pages - Browse / Search Linux source code - for all kinds of Linux info - for some really good links to other Linux sites - for some really good links to good software

Loud Fan Noise in Some Servers:

Linux Health Agents will reduce fan noise in some servers (ML370)

SA5300 and RH70:

The SA5300 controller is supported under RH70 text mode install (not the GUI install).

SCO505 - "Please wait while SCO logs in ..."

ML330/SA221 - installing SCO 505 - "Please wait while SCO logs in ......"
SCO had installed just fine; however, system seems to hang trying to get into graphics screen.
Pressing CTRL+ALT+F1 will produce a login prompt. EU was using an upgrade license during installation. We ran SCOADMIN, reentered the license for SCO 505, system came up normally on the next reboot.

Red Hat 6.2 Linux and NC3122, NC3134, or NC3135 NIC's:

The installation may hang at "Performing Post Install Configuration"
To correct the problem, download the new Red Hat Linux 6.2 boot and
update disks at:
Two updates are required: RHBA-2000:015-01 and RHBA-2000:010-01, which
are available under "Bug Fixes for Version 6.2".

SCO 505 and SA5300:

SS4.90 released w/ EFS 5.38. Currently the Assisted install is not supported with SCO505 and the SA5300. The EFS documentation states you must perform a manual install. I am having the customer do this to see if it will work: Perform an assisted install of SCO505, when it boots back to the boot: prompt, key in the mem string (mem=1m-16m,16m-128m/n [SCO TA #: 109935: I get "Memory excessively fragmented" when attempting to install OpenServer 5.0.5 using a Boot Time Loadable Driver (BTLD).]), then key in the defbootstr link=ciss hd=Sdsk Srom=wd(0,0,0,0). Then continue w/ normal Assisted install, see what happens.
Here's feedback from the customer who tried this:
-----Original Message-----
From: Steven xxxxx
Sent: Thursday, October 19, 2000 12:07 PM
To: Black, Richard
Subject: xxxxxxx
Hello Richard,
SmartStart 4.9 assisted installation of SCO 5.0.5 (on DL380 with 5304 Array
Controller) proceeded without incident after using the string
mem=1m-16m,16m-256m/n followed by defbootstr link="ciss" hd=Sdsk
Steve xxxxx

EFS 5.38 for SCO 5.0.x:

The SCO 505 EFS now consists of the first 3 disks instead of the first 2 as of EFS 5.38.

DL360 and Caldera OpenServer 2.3, embedded dual port NIC:

The DL360 dual NIC's are on the secondary bus and will not be detected by Caldera 2.3. This is covered quite clearly in the How-To and patches are available to address the issue. Plain and simple, get the new boot disk from Caldera and do the install.

The dual integrated NIC's are on the secondary bus, you need caldera's patch to access it.

For Caldera OpenServer 2.3, Caldera now provides an updated installation diskette plus an updated kernel. The software and documentation can be found at Caldera: l.144.peerbus-243

Linux and RA4100

You can use the cpqfc driver, which has been released and is in Red Hat 7.0.

It currently (Jan 2001) only supports one card: the 64-bit/66MHz Fibre Channel Host Adapter.

This information is on the Options Support Matrix, at:

Here's the Compaq opensource page for this driver:

Here's the FreshMeat page for this driver:

When using this driver, your arrays will be accessed through the standard SCSI devices, such as /dev/sda, /dev/sdb, etc. This is different from our other array controllers which use their own device names (like /dev/ida/c0d0, etc.).

Feedback from customer on Linux and RA4100:

Hi Richard
thanks. I applied the cpqfc patch, compiled a new kernel and it
worked fine. Both DL360's can now see the RA4100 as /dev/sda.
thanks again
Eric (Using Compaq Storage Works 64-bit/66-MHz Fibre Channel Host Adapter opt#120186-B21)

With some Compaq/Emulex cards (64bit PCI to fibre - Emulex, DS-KGPSA-CY, sp# 176804-002), booting to SS4.80 is not able to configure the RA4100. 2 workarounds for this exist:

1) install NT to the system on some storage other than the RA4100 (you cannot boot to the RA4100). Install the NTSSD. Then install the ACU. Configure the array, then install Linux, then use the array in Linux. you can either remove NT after the RA4100 is configured or you can leave it around in case you need to do further configuration later on.

2)Find another NT system that already has a fibre channel host bus adapter and the NTSSD and the ACU. Then just move the RA4100 to it, configure it, then move it back to the Linux box.

Note: You cannot install to, nor boot from Fibre Storage.

Tapes on ROC:

SCO 5: clad, when running mkdev tape, increase the port# by 1 to get the value for the bus number. (read man clad for more info). If you have only 1 clad controller installed in the system, then the ha # for that clad controller should be 0. Here is an excerpt from man clad:

Configuring devices connected to the Controllers supported by the clad

Please refer to cpqclad(CPQEFS) in order to configure RAID volumes on
controllers supported by this driver.

To configure tape drives on the clad driver, follow the steps given below.

* Invoke "mkdev tape" and select the SCSI tape device.
* When prompted for the ha driver, enter "clad".
* When prompted for the ha number, enter the ha number of the controller.
* When prompted for the bus number, use the following calculation for the
bus number,
bus number = port number to which the tape drive is connected + 1.
ie, if the tape drive is connected to port 1 of the controller, the bus
number will be 2.
* When prompted for the SCSI Target ID and LUN, enter the appropriate values
for the tape drive.

UW7: Status Unknown
Linux: not yet
Sol: 2nd port - support unknown - investigating

Support Numbers:

Solaris: 1-800-872-4786

Setting the Date in Linux:

How to change the hw time from within Linux
Key in the following two commands

set the date with:
make the date permanent with:
hwclock --utc --systohc

SOLARIS Installation Notes:

Installation notes can be found in the README files on the Compaq DU's for Sol. They will soon be updated so look for some changes coming up soon. The changes necessary should be concerning which CD to boot to. So far, it appears that we need to not use the webstart application. I have also put lots of Sol8 install notes out at -> "Solaris" (


DAAD should no longer be used, ADU should be used instead.

SCO 5.0.5 - Memory Excessively Fragmented - iQ-8423:

"Memory excessively fragmented" when attempting to install SCO 5.0.5 using a BTLD.

TA #: 109935: I get "Memory excessively fragmented" when attempting to install OpenServer 5.0.5 using a Boot Time Loadable Driver (BTLD).

There are two solutions:

1) Get the updated SCO 5.0.5 boot disk that has the memory string already keyed in or

2) Use the following boot string which will do the same thing as the updated boot disk. This memory string is to be keyed in alone at the boot: prompt, do not combine it with the defbootstr command. The memory string is to be entered first, then at the second boot: prompt you will key in the normal defbootstr.

Memory String Format:
mem=1m-16m,16m-<total system memory>m/n

Example Memory String Entry (64MB System):

How do I determine what SCSI ID the tape is currently set for in SCO 5:

Make sure a tape is even configured by looking in "mkdev tape" -> configure a SCSI Tape drive -> View current configuration.
Then to determine the SCSI ID, do: cat /etc/conf/cf.d/mscsi
Another way is "sconf -v", this will work all the time in SCO505, in SCO504 beware, read the following article for SCO 504:
TA #: 109675: The command "sconf -v" causes my system to panic with "trap type 0x0E".

PL3000 (P09), SCO 5, Enable SMP, system hangs at ctrl-d prompt - iQ-8923:

HW: PL3000, BIOS: P09 - 12/99 (550MHz, 12/7/99)
SW: SCO 5.0.5
MOF: Enabled SMP, System now hangs at control-d prompt. EU didn't see message that stated: Warning APIC - No Bios Information Found for IRQ1
SOL: Update to the latest ROM and the latest SCU.

UnixWare 7 Agents - not starting - keyword "#" not found:

The following link has all the details on upgrading the EFS in SCO
Sep 01 11:42:33 cmafdtnsmuxd: ../../mailcfg: line 1: keyword # not found (SMUX4026)
Sep 01 11:42:33 cmafdtnsmuxd: ../../mailcfg: line 1: illegal character 0x2a (SMUX4026)
SW: UW7.1.0, Agents 4.70, EFS 7.34, (NOT a NON-Stop cluster)
MOF: Agents won't start.
SOL: removed the pound signs from /opt/compaq/mailcfg. Agents are now able to be started.

HP OpenView:

Check the table at the following link for "Version Support":
If the versions do not match anything on the matrix, then it is not supported. Problems usually do not occur when the correct versions are used.

Problems on an SMP system running SCO 5.0.5/5.0.4:

If a SCO 5.0.5/5.0.4 SMP system continually has problems, hangs, whatever, there are a few steps to take to correct it. Usually it is that patches were not applied in the correct order - even though patches may not have been touched in a long time. Here is a short summary of the steps needed to correct the situation:

  • Upgrade to the latest ROM's (Sys ROM, Array ROMs, HDD ROMs).
  • Run the latest version SCU and save changes.
  • Follow the steps to remove and re-add the EFS and SCO patches (, be sure they follow it completely and correctly (get a full backup, get reports from the system [NIC info, inspect, hwconfig, etc.], unconfigure CPQ NIC's, remove in this order: CPQ Agents, CPQ EFS, all SCO patches, SMP, reapply in this order: SMP, SCO patches listed [some patches added, but not listed as a recommended patch may cause problems], EFS, Agents).
  • Then clear NVRAM through advanced mode of SCU ( and follow correct procedures listed there to reconfigure using the latest SCU.

My 2nd processor keeps dying in SCO 5.0.5/5.0.4:

If you have a system where you keep loosing all processors but processor #1, then this should be a case where the OS is corrupting NVRAM and thus taking the 2nd, 3rd, and 4th (if you have them) processors offline (For some reason, you may notice that the failures start with the last processor and works its way toward the first processor -- 4th proc, 3rd proc, 2nd proc.). Upon a reboot you get the 212 processor failed. You can mark the processor as repaired, and it will survive a couple of reboots without entering the OS. But when you go back into the OS and come out again, usually (but not always) the processors are marked failed again. Sometimes it can take a while for the processors to be marked as failed. In this situation, you will notice that PROCESSOR #1 IS NEVER REPORTED AS BEING FAILED. Even if you switch processors around, and even switch PPMs around, you will still notice that proc #1 is never marked as failed.

The issue is that the OS is corrupting NVRAM. On SCO 5.0.4 and 5.0.5, there is a patch that must be applied and then afterwards, you must clear NVRAM through advanced mode of SCU. Remember to apply the patch, you MUST sandwich it in the correct location, this means removing off several layers to apply the patch, apply it, then put those layers back on. You can find all the instructions at

Upgrading the Compaq EFS in SCO 5.0.5/5.0.4:

The following link has all the details on upgrading the EFS in SCO 5.0.5/5.0.4: I usually cover the steps with the customer making sure that they cover the main overview (SCO, SMP, PATCHES, EFS, AGENTS), the upgrade process (currently steps 1-15), and the pitfalls (listed in the notes section immediately following the upgrade steps. The most important information to give the customer is that they MUST BE IN SINGLE USER MODE. The 2nd most important piece of information is DO NOT OVERWRITE AN OLDER EFS WITH A NEWER ONE on SCO OpenServer (You get to keep both pieces when it breaks - and it WILL break - count on it.).

Network Streams in SCO 5:

"My network runs for a while, then stops working, the server is ok, nobody can connect." The answer to this one is that they are probably running out of Network Streams. This is a tunable parameter in the SCO 5 kernel. Type "netstat -m", check the last 3 lines (1] total configured streams memory: xxxx 2] streams memory in use: xxxx 3] maximum streams memory used: xxxx).
TA #: 107566: How do I tune STREAMS resources under OpenServer 5 (TCP/IP 2.0.0)?
To begin configuring the kernel, be in single user mode, do the following: "cd /etc/conf/cf.d" "./configure"
Then adjust NSTRPAGES and NSTREAM. We usually double NSTRPAGES and adjust NSTREAM until everything is stable. We usually leave STRSPLITFRAC alone. The default value for NSTRPAGES is 500, we usually run NSTREAM up to around 4096 or so.

Bootable CD-ROMs - IDE vs. SCSI:

IDE is a simple and easy format. There are not many ways to do IDE so it is easy to build a CD that is bootable on IDE. For SCSI, there are so many different SCSI cards (and drivers) that they won't all fit on a bootable image (usually limited to about 1.44M - even on CD). What this means is this: if you want to make a CD that is bootable on a SCSI CD-ROM drive, you first need to know which SCSI controller will control that. You then grab that driver and then you can make a bootable CD for that SCSI controller. The reason that Compaq SmartStart CD-ROMs are bootable on Compaq SCSI CD-ROM drives used on Compaq SCSI Controllers is that Compaq has only a couple of SCSI drivers and puts them both on the bootable image and they are done. Compaq SmartStart CD is not likely at all to be bootable on a SCSI CD-ROM drive plugged into a NON-Compaq SCSI controller.

Linux and the Intel Network Controller (Compaq NIC with Intel chipset, RH6.2, DL360):

By default Red Hat installs the eepro100 driver rather than the e100 driver. Compaq has an e100 driver that they have checked out and certified. The e100 driver performs much better than the e100 driver. The driver from Compaq is located at:

It appears that there is a problem with the eepro100 driver that gets loaded by default in Red Hat 6.2 on the DL360 and other servers which contain the Intel based Compaq NIC. The solution on this is to get the e100 driver from Intel and install that. The driver supplied by Intel in the download is configured for Red Hat 6.2 non-SMP. If the system is or will be configured for SMP, then the driver will need to be rebuilt with SMP enabled (edit the e100 "Makefile" and turn on SMP support); otherwise, the driver will not function. Let us know if this does not resolve the issue. Here is a shortcut to the same driver, but this one is on Intel's site:

Clearing NVRAM Through Advanced Mode of SCU:

Please do not use the Maintenance Switch unless you cannot successfully boot to the System Configuration Utility.
Here's a complete document on how to clear NVRAM:

2/8GB DAT, HP C1536-00480, AS# 142019-001 / SP# 142074-001 - iQ-10572:

Default switch settings = all on and switch #3 off, as seen here (
It appears that some of the C1536 spares are shipping with all switches to the on position. This is causing the tape drive to have a blinking amber light and it will not accept a tape (motor doesn't engage). After locating the default settings on HP's site, tape drive worked. The Compaq documentation does not show the default settings.

SCO 5.0.6 is Currently Shipping:

SCO 5.0.6 is currently shipping; however, the latest SmartStart 4.80 does not have an entry for 5.0.6. We don't have any media yet to test this, but it can be certain that one cannot do an assisted install of SCO 5.0.6 by using SS 4.80 and choosing assisted install of SCO 5.0.5. One must choose to do a manual install at this time. This also means that we do not currently fully support SCO 5.0.6, so results may vary. EFS 5.36 was observed to work on SCO 5.0.6. No more info available at this time.

SS4.90 and SCO 5.0.6:

SS 4.90 supports SCO 5.06 assisted install. It did however drop 5.04 assisted install support. Also EFS 5.38 which is on SS 4.90 has the new cnet driver and is supposed to fix the "adapter check interrupt's" on cnet NIC's.

Solaris installs - Panic:

We had a Sol 8 install on an ML370 that required a modification to the 06/00 DCA Disk. The modification turns off ACPI. The exact configuration for this one method was ML370, SA3200, Sol8. We added the following line to the end of the solaris/bootenv.rc file on the Sol8 DCA disk.
setprop acpi-user-options '0x2'

Solaris DCA Disk Location:
Don't use the FCS version, I don't know what it is. Use the updated DCA from Solaris' web site because it may contain fixes for the install.

When does Solaris Prompt for the Compaq DU's:

Solaris prompts for the DU's at the VERY FIRST screen you can interact with. On this very first screen, at the very first keystroke you can press, press "F4 - Add Driver". -- If they've hit any other keys or went to any other screens, then it's probably best to have them restart.

Working with a DOS floppy:

SCO 5: doscp, dosdir, dosformat, dosrm, ... ex: dosdir a:
UW7 - method UW7.0.0 only: doscp, dosdir, dosrm, ... These were part of the dosmerge package and are no longer included in UW7.
UW7 - method UW7.0.1 and later: mcopy, mdir, mdel, (mattrib, mbadblocks, mcd, mcopy, mdel, mdeltree, mdir, mformat, minfo, mlabel, mmd, mmount, mpartition, mrd, mread, mmove, mren mtoolstest, mtype, mwrite and mzip) These are part of the mtools package on the SCO Skunkworks CD, install using pkgadd. The mtools commands use a UNIX style pattern to match with rather than DOS style e.g.: * to match all files not *.*
Linux - method #1: mcopy, mdir, mdel, mformat, mren, ... ex: mdir a:
Linux - method #2: mount -t msdos /dev/fd0 /mnt/floppy

Online Inspect Report:

SCO5: /usr/local/compaq/svr/bin/cmastdeqd > /tmp/inspect.out
Note: must have agents installed.
Linux: install and use the survey utility ( - unknown if agents are required - I've always put survey in after the agents.


DL380: Use CH1 for tape (int or ext), mkdev tape, clad, HA+1

Backing Up the Kernel:

SCO 5: TA #: 105196: How can I copy a kernel to the read-only /stand filesystem?
Linux: ...(for now read the info at

Major and Minor Numbers:

SCO 5: TA #: 104383: Explanation of major and minor numbers.

Determining the running OS and Version:

SCO 5: uname -X, third line shows the release, for ex.: "Release = 3.2v5.0.4"
UW7: uname -f, last line shows version, for ex.: "version=7.1.1"
Linux - Red Hat: 1) cat /etc/redhat-release 2) uname -a

OS Support Matrix:


Compaq Hardware: INSPECT, ADU (Use ADU instead of DAAD)
SCO 5: hwconfig, swconfig, uname -X, /var/adm/messages, /var/adm/syslog, /var/spool/compaq/agenterrs.log, health log (cpqutil | tee -a /tmp/healthlog.out) (cpqutil -> 4. server health log)
UW7: /var/spool/compaq/agenterrs.log, /usr/adm/log/osmlog
Linux: /etc/fstab, fdisk -l, /etc/lilo.conf, /etc/conf.modules, /var/log/messages, /bin/dmesg >/tmp/dmesg.out (not /var/log/dmesg), /var/log/boot.log, /etc/mtab, /proc/cpuinfo, /proc/meminfo, /etc/sysconfig/hwconf (kudzu's hw db file of currently installed hw), /proc/scsi/scsi (hdd's / tapes attached to embedded SCSI), /proc/array/ida0 (Compaq Array Controller), uname -a, cat /etc/issue.

Variable vs. Fixed Block Sizes during Backup / Restore:

For many reasons, it is better to go with a fixed block size rather than a variable one when performing tape backups and restores. One should test various block sizes to see which one gives the best results (speed vs. tape use). Use your OS's tape setblock command to specify a fixed block size. Keep increasing the block size until the optimum value is found. The block size should match the average size of the files being backed up. The same block size is required for both the backup and restore. Most tape drives will default to variable block size. A variable block size during a tar or cpio backup will cause the backup/restore to take a very long time or even fail.
TA #: 102733: A DAT tape written on one DAT drive cannot be read on another.
TA #: 105433: I cannot restore my Open Server 3.0 tape backups on OpenServer Release 5.
The man page for tapecntl (UW7) states "Using variable-length block mode when writing to magnetic tapes is discouraged because it may not work correctly ... Magnetic tape should always be written in fixed-length block mode, even though you are free to change the default fixed-block length ...".

Setting and Getting Tape Block Sizes:

SCO 5 - set: tape -a 512 setblk
SCO 5 - get: tape getblk (a return of 0 means variable block size)
UW7 - set: tapecntl -f 512 /dev/rmt/ntape1
UW7 - get: n/a
Linux - set: mt -f /dev/tape defblksize
Linux - get: mt -f /dev/tape status
Linux - /dev/tape should point to /dev/st0

Tape Status:

SCO 5: tape -s status /dev/rct0
UW7: n/a
Linux: mt -f /dev/tape status

Resetting tape drives:

SCO 5 - tape reset: tape -s reset /dev/rct0
SCO 5 - SCSI bus reset: cpqsmu c -tcha -n0 -r (this resets the first SCSI channel, ha=0, because "n" is the host bus adapter number)
UW7: tapecntl -r /dev/rmt/ntape1
Linux: n/a

SCO 5 - Adding Tape Drives:

sconf -v may be useful in gathering parameters to use in mkdev tape. The ha (Host Adapter) number can also be seen (and is configurable) in SCU.
If you have mkdev tape problems, then clean out mkdev tape as much as possible (remove all the tape entries you can) and then check with he following 5 SCO TA's at
TA #: 109266: Unable to remove the last tape drive in mkdev tape.
TA #: 105776: Unable to rebuild the kernel after all tape drives are removed.
TA #: 104978: idmkreg reports "error updating space.c for Stp" when the kernel is relinking.
TA #: 109447: Kernel relink fails with "too many array initializers" after adding a SCSI tape unit.
TA #: 110973: What files get out of sync after trying to unsuccessfully add a tape drive?
There is a chance under SCO 5.0.4 w/ rs504c installed for the system to Panic with a trap type 0x0E when running "sconf -v" and you have two SCSI host adapters. Here is the SCO TA as well as the fix:
TA #: 109675: The command "sconf -v" causes my system to panic with "trap type 0x0E".
When adding a tape drive in SCO 5, you use mkdev tape. You need to have 4 values to add a tape: Host Bus Adapter #, SCSI Bus (always 0), Target ID (SCSI ID), and LUN (All Compaq tape drives will have a LUN of 0 - LUN is for multiple devices on the same SCSI ID). So with SCSI Bus and LUN both being 0, the Target ID is the SCSI ID you assigned the drive, the only question left should be what is the Host Bus Adapter #. This value is determined by using sconf -v or by looking in SCU. This value is also configurable in SCU.

Simple Backups / Tests:

SCO 5 - Backup: tar -cvf /dev/rStp0 etc
SCO 5 - Test: tar -tvf /dev/rStp0
UW7 - Backup: tar -cvf /dev/rmt/ntape1 ????
UW7 - Test: tar -tvf /dev/rmt/ntape1 ????
Linux - Backup: tar -cvf /dev/tape etc
Linux - Test: tar -tvf /dev/tape

Note: Never use a leading slash when backing up. The leading slash will cause you to use absolute pathnames, rather than relative pathnames. The problem happens when trying to restore a backup made with absolute pathnames - you will not be able to check out the backup first, when restoring, it will try to go directly to the absolute directory. With relative pathnames you may extract the archive to a temporary directory, make any compares as necessary, and copy which files you want to where you want to - you get full controll when using relative pathnames when backing up.
Just in case you need to inspect a backup that was created with absolute pathnames, the best way I can think of is to use a chroot environment. See chroot.html.

Mounting the CD-ROM:

TA #: 108273: Mount commands for CD-ROM drives.
SCO 5: mount -r -f HS,lower /dev/cd0 /mnt
UW7: mount -F cdfs -r -o fperm=555 /dev/cdrom/c?b?t?l? /mnt
c=ctrlr, b=bus, t=target_id(scsi id), l=lun
Linux: mount /mnt/cdrom

Network Card Info:

SCO 5: ifconfig -a
UW7: ifconfig -a
Linux: ifconfig

Network Tools:

SCO 5: ifconfig -a, netstat -rm, netstat -r, netconfig, ping, telnet, route, /etc/hosts
UW7: ifconfig -a, netstat -rm, netstat -r, netconfig, ping, telnet, route, /etc/hosts
Linux: ifconfig, netstat -rm, netstat -r, netconf, ping, telnet, route, /etc/hosts, /etc/resolv.conf, hostname, hostname -f

Admin Tools:

SCO 5: scoadmin, custom, netconfig
UW7: scoadmin
Linux: linuxconf, netconf, xf86config, userconf

Text Pagers, Editors, Output Helpers, and Filters, Timers, Help Files, File Utils, ... :

Pagers: less, more, tail, head
Editors: vi, pico
Output Helpers: tee, script
Filters: Grep, sed, awk
Timers / Schedulers: at, cron
Help Utilities: man, info
File Utilities: find, slocate, updatedb
Process Management: ps, top, cpqmon
Other Utils: watch, fuser, time
SCO 5: more, cat, tail, head, vi, tee, script, grep, cron, man, find, ps, cpqmon
UW7: more, cat, tail, head, vi, tee, script, grep, cron, man, find, ps
Linux: less, more, cat, tail, head, vi, pico, tee, script, grep, at, cron, man, info, find, slocate, updatedb, ps, top, watch
man -k searchstring
man -K searchstring
find searchdir -type f -exec grep searchstring \{\} --with-filename \;
find /etc -type f -exec grep -i redhat \{\} --with-filename \;
touch /tmp/timestamp.txt -t 200007301215; find / -newer /tmp/timestamp.txt


Runlevels control what services run by default and which ones don't.
Runlevel 0 is halt. In this mode the system halts (all services halt).
Runlevel 6 is reboot. In this mode the system halts then reboots.
Runlevel 1 is Single User Mode. In this mode networking is not loaded and filesystems are not mounted automatically.
Runlevel 2-5 is Multiuser Mode. In this mode the system fully comes up (It brings up the networking and filesystems and allows users to log in.)
Each OS has their own default multiuser runlevel: SCO 5 is "2", UW7 is "3", Linux is "3" or "5" depending if you chose to load X-Windows on startup or not.
The other available multiuser runlevels (runlevels 2 - 5 in the respective OS's) can be customized to create a specialized runlevel (specialized runtime environment) controlling which services start and which ones don't when in that particular runlevel.

Determining the Current Runlevel:

SCO 5: who -r
SingleUser = S, Default Multiuser Runlevel = 2
UW7: who -r
SingleUser = S, Default Multiuser Runlevel = 3
Linux: runlevel
SingleUser = S, Default Multiuser Runlevel = 3 (no GUI) / 5 (GUI) - configurable in /etc/inittab

Changing Runlevels:

SCO 5: init # (where # is the runlevel to go to)
UW7: init # (note, init 1 is not recommended for UW 7, see "getting to runlevel 1" to do this right)
Linux: /sbin/telinit # (where # is the runlevel to go to)

Getting to Runlevel 1 / Single User Mode:

Single User Mode means no networking and no mounted file systems (that is, they are not loaded automatically)
SCO 5: During boot, you will be prompted with:

Type CONTROL-d to proceed with normal startup,
(or give root password for system maintenance):

Key in the root password and you will enter Single User Mode. You may also, from multiuser mode, type init S; however, the keystrokes in scoadmin, custom, and netconfig get messed up and don't seem to function well, so I don't recommend performing the command "init S".
UW7: press spacebar during the UW7 banner, then type S, then type go or boot
Linux: On boot at the LILO: prompt, type the kernel name and S, for ex: linux S

Startup Scripts:

SCO 5:
Linux: /etc/inittab, /etc/rc.d/rc.sysinit, /etc/rc.d/rc, /etc/rc.d/rc.local, /etc/rc.d/init.d, /etc/rc.d/rc3.d, /etc/rc.d/rc5.d

Relinking the Kernel / Building a Kernel:

SCO 5 - kernel tuning: 1) cd /etc/conf/cf.d/ 2) ./configure (that's "dot slash, configure")
SCO 5 - kernel relinking: 1) cd /etc/conf/cf.d/ 2) ./link_unix (that's "dot slash, link underscore unix")
Linux: See
extra info unknown: man idbuild

Shutting Down:

SCO 5: shutdown -g0 -y -i6
g=gracetime (1=1minute), y=answer_yes, i6=reboot
UW7: shutdown -g0 -y
Linux: shutdown -g0 -y, shutdown -g0 -y -i6, shutdown -h now, shutdown -r now
h=halt, r=reboot

Setting Environment Variables:

SCO 5: VARIABLE=setting; export VARIABLE
Linux: export VARIABLE=setting

Default Term Types:

SCO 5: TERM=scoansi; export TERM
Linux: export TERM=linux

Compaq and Linux Archive:

linux-kernel Archive:

Lots of Notes:
search the archive:
Alt archive location:
Alt archive location:
Alt archive location:
Alt archive location:

The Linux Kernel:

Linux Training Course:

Common Locations of Executables:

SCO 5: /etc, ...
UW7: /etc, ...
Linux: /bin, /sbin, /usr/bin, /usr/sbin

F10 System Partition:

DL360, DL380, ML370, ML530, ML570
PL1000, PL1200, PL1500, PL1600, PL1850R, PL2000, PL2500, PL3000, PL5000, PL5500, PL6000, PL6400, PL6500, PL7000

ML370 DOES have the F10 System Partition.

F10 Bios Setup:

ML330, ML350, PS720?, PL400?


Standard - DL360, DL380, DL580
Optional - ML370, ML570


ML330, ML350, ML530








Boot Strings:

SCO 5.0.5 - embedded SCSI: defbootstr link="cha" hd=Sdsk Srom=wd(0,0,0,0)
SCO 5.0.5 - Array Controller: defbootstr disable=ida link="cha ida" hd=ida0 Srom=wd(0,0,0,0)
SCO 5.0.5 - Array Controller: defbootstr link="clad" hd=Sdsk Srom=wd(0,0,0,0)
Linux: linux text expert
Linux: linux text expert mem=24M


SCO 5: cpqefsver

Linux - ROC, and Tape Drives:

Currently (June 2001) the array driver in Linux (cpqarray) does not support tape drives. This means choosing between one of two choices: 1) get a separate controller for the tape drive or 2) get an array controller for the drives, remove the ROC cache chip, and leave the tape drive on the embedded SCSI (will use the updated ncr53c8xx driver)

Linux - and Original Smart:

Currently, under Red Hat 6.2, this driver can't be loaded as a module. The install will work with the driver as a module; however, booting to the system with cpqarray.o loaded as a module will fail. The solution to this is: at some point (during install, after install, before install, whatever) compile a kernel with cpqarray built in statically. After you do this, get this into /boot either during install or after install. Edit /etc/lilo.conf to use this new kernel, run /sbin/lilo and then reboot onto the array controller. This allows you to boot straight to the Original Smart using the cpqarray.o driver built in statically. See for complete instructions.

Linux - PL4500 and Linux:

The PL4500 seems to be the only server entirely EISA based with the ncr53c8xx chipset. After much searching, it does not appear that the ncr53c8xx supports EISA. The solution is to not use the embedded SCSI, but to install a supported adapter card in a slot. One of the Compaq array controllers can be used (original Smart and Smart-2/E) or any Linux supported (and PL4500 supported) ISA or EISA adapter card (the sim710 driver (ncr53c710 chipset) will support EISA). See for complete instructions.

Linux - Agents and Wellness Driver:

Health Driver, Agents, Survey, e100, and others are all available from:

Linux - RIB/LOE (Lights Out Edition):

Yes, the RIB does work in Linux to control the console. The only parts that doesn't work is the parts that require drivers, this means you can't pull information from the OS. All other parts that doesn't use drivers works ok.

SCO 5 - Adapter Check Interrupt:

We are currently getting "Adapter Check Interrupt" on the network cards under SCO 5. The current solution to this is to downgrade to EFS 5.32, SP10689.EXE. Both EFS 5.34 and 5.36 are affected. If this does not resolve the issue, then contact tech support.

SCO 5 - Invalid Request:

We are currently getting "invalid request" on the array controller. This is not the only format we are seeing:
NOTICE: ida0: slot 6: invalid request on dev 146/0 blk 0
Jul 14 09:18:25 max logical unit=0, block=0, size=1, cmd=0x82, status=0x10
Jul 14 09:18:25 max CMD_BLK: addr=C0149734h csize=00000094h
Jul 14 09:18:25 max CMD HDR: unit=0 prior=0x00000000 ctrl=0x00000025
Jul 14 09:18:25 max REQ HDR: next=0 cmd=0x00000082 ret_code=0x00000010
Jul 14 09:18:25 max PARAM: blknum=0x00000000 blkcnt=0x00000001 cnt=0x00000001 phys addr=0x007DA800 virt addr = 0xC07DA800 bytes=0x00000200
Basically for the invalid request on the ida controller, the result is to either ignore the invalid request for now or remove the Compaq Storage Agent. It looks like the Compaq Storage Agent is giving incorrect information.

Preparing to call Technical Support:

Customarily the following is done:
Make sure you have the latest System ROM, Array Controller ROM, and latest SCU (Be sure to run SCU and save changes).
Here is a sample list of some of the reports needed, more reports may be needed based on the issue:
HWCONFIG > /tmp/hwconfig.out
SWCONFIG > /tmp/swconfig.out
UNAME -X > /tmp/uname.out
tail -1000 /VAR/ADM/MESSAGES > /tmp/messages.out
tail -1000 /VAR/ADM/SYSLOG > /tmp/syslog.out

As always when sending in reports:
CC: CC the Technical Support Engineer if he/she gives you his/her email
Subject: case number (no dashes or spaces in the case number) and put ATTN to the TSE who is helping you
Body: Put the reports as attachments in the body of the email. Please do not have all the reports listed as one long page. You may place any extra error messages or notes into the body of the email as well.

Compaq Hardware - Lpt Ports:

  LPT Port    Standard I/O Address     Compaq I/O Address
    LPT1              378                      3BC
    LPT2              3BC                      378
    LPT3              278                      278
Note, the newer Compaq servers have changed the addresses around to match the Standard I/O Addresses; however, they still haven't change the value around. Eventually Compaq will have the value matching the address that is displayed in SCU. Summary: When the HW is currently set for LPT1, then in the OS you should use the address of 3BC (that is until we change the value on the newer servers). LPT3 has always matched, so there may be less confusion about what the address is set to there.

Here are some current support reference sites:

External SCO/Unix Web Site (hosted by Chris Knapp):
Compaq's Linux Home Page:
SCO's Technical Articles Page:
SCO's Minimum Patch Listings:
OS Support Matrix:
Nice search engine for just about everything else:

Updates to this page:

If anyone has any updates, changes, or additions to this information, please let me know and I will be happy to put the information in.
prev next index
NEW! Search this Site!: