Technical article 109445

After installing SCO UnixWare on a HP/Compaq server it relinks every time I reboot.

OpenServer 6.0.0UnixWare 7.0.0UnixWare 7.0.1UnixWare 7.1.0UnixWare 7.1.1UnixWare 7.1.2UnixWare 7.1.3UnixWare 7.1.4Open UNIX 8

Written · Last revised

Problem

I have an HP/Compaq server system with a dual-channel integrated "cpqsc" (Compaq SCSI Ultra Wide) controller for my hard drives and an integrated IDE controller for my CD-ROM drive which I used for the installation. After installing SCO UnixWare and getting all the drive controllers recognized, the system goes into a relink and reboot loop indefinitely if I don't press <Esc> when prompted during the boot process. I get the following message just before the system reboots itself:

The Unix Operating System will now be rebuilt to reflect recent configuration changes:

Strike ENTER when ready
or ESC to stop
If I press <Enter>, the kernel is rebuilt, then the system reboots
and the process repeats.  If I press <Esc>, the system continues to
boot without rebuilding the kernel. In most cases, the system may
function normally at this point. In other cases, the boot process
may fail with:
unable to mount /stand

Cause

This is due to a conflict in the resmgr(1M) database.  The most
likely cause of this conflict is multiple items using the same
UNIT number. In this case the conflict is between the IDE
controller and the second channel of the cpqsc SCSI controller.
The IDE controller requires a UNIT number of 1, but the second
channel of the cpqsc SCSI controller is using UNIT number 1. The
following is a sample output of the command "resmgr" specific to
this case:

KEY MODNAME UNIT IPL ITYPE IRQ IOADDR MEMADDR DMAC BINDCPU BRDBUSTYPE BRDID SLOTENTRYTYPE BUSNUM 10 cram 1 - - - 70 7f - - - - - - - 1 - 11 atup 0 - - - - - - - - - - - - 1 - 12 atup 0 - - - 40 5f - - - - - - - 1 - 13 atup 0 - - - 20 3f - - - - - - - 1 - 14 atup 0 - - - a0 bf - - - - - - - 1 - 15 asyc 1 9 1 4 3f8 3ff - - - - - - - 1 - 16 asyc 1 9 1 3 2f8 2ff - - - - - - - 1 - 17 pnt - 6 4 3 7000 701f - - -1 - 4 0x10222000 0 - 0 18 cpqsc 0 5 4 10 7100 71ff 2000000 20000ff - - 4 0x10000001 0 - 0 19 m320 1 6 1 12 - - - - - - 1 - - - - 20 cpqsc 1 5 4 15 7400 74ff 2200000 22000ff - - 4 0x1000000F 5 - 0 21 ide 2 5 4 5 1f0 1ff - - - - 4 - 4 - 0

The third field (or column) reflects the current UNIT number for each line item.

An alternate cause would be that the controller order within the server's BIOS is set incorrectly so the Resource Manager and the server's BIOS never agree.

Solution

Check that the controller order is correct within the server's BIOS at boot time.

Use the DCU (Device Configuration Utility) utility from the character console to change the UNIT number for IDE from 2 to 1 and change the UNIT number for the second channel of cpqsc from 1 to 2. To do this, take these steps:

(1) When the system boots and gets to the kernel rebuild point, press the <Esc> key to avoid rebooting. When the system has booted completely, change to the character console screen with one of the following key sequences:

UnixWare 2.1.x: <Ctrl>Alt>SysRq> and then the "h" key. UnixWare 7.x.x: <Ctrl>Alt><F2>

(2) Log in as root and enter the Device Configuration Utility (DCU) by typing the following command:

# /sbin/dcu

(3) When the menu for the DCU comes up, highlight the "Hardware Device Configuration" and press <Enter>.

(4) Using the <PageUp> and <PageDown> keys or the arrow keys, locate the second entry for "Compaq Wide-Ultra SCSI" and place the blinking cursor on the beginning of that line. Press the <F7> key. You will see the following example screen:

Advanced Parameter Selection
Device Name             Compaq Wide-Ultra SC
BindCPU                 -
Unit                    1
IPL                     5
ITYPE                   4
Use the down arrow key to place the blinking cursor under the
"Unit" number and change the number from 1 to 2 by pressing the
number 2 key.  Then press the <F10> key to be returned to the
"Hardware Device Configuration" screen.

(6) Now locate the entry for "IDE" and place the blinking cursor at the beginning of that line. Press <F7>. You will see the following example screen:

Advanced Parameter Selection
Device Name             IDE
BindCPU                 -
Unit                    2
IPL                     5
ITYPE                   3

Use the down arrow key to place the blinking cursor under the "Unit" number and change the number from 2 to 1 by pressing the number 1 key. Then press <F10> to be returned to the "Hardware Device Configuration" screen.

(7) Press <F10> again to be returned to the main menu and highlight the line "Apply Changes & Exit DCU". Press <Enter> at that point to apply changes and exit the program.

(8) Once the changes have been made, it is necessary to rebuild the kernel for the changes to take effect. Do this by shutting the system down normally. When it comes back up, there will be no conflict.

Please note that the latest EFS version available from HP (7.72), 'cpqunits' is removed.

NOTE:     TROUBLESHOOTING:

- The same issue has been reported with HP/Compaq's "ida" and "ciss" Hard Disk controller drivers.

- Alternatively, use the "resmgr" commands to resolve the issue by setting the Unit Number to the correct value, followed by:

# /etc/conf/bin/idconfupdate -f
# /etc/conf/bin/idbuild -F
# init 6
- Check to see if "cpqunits" can be disabled and this can cause the
problem.  To do this change the following in /etc/inittab from:
cpq1::sysinit:/etc/scsi/cpqunits -a >/dev/console 2>&1
cpq2::sysinit:/etc/scsi/cpqunits -o >/dev/console 2>&1
to:
#cpq1::sysinit:/etc/scsi/cpqunits -a >/dev/console 2>&1
#cpq2::sysinit:/etc/scsi/cpqunits -o >/dev/console 2>&1

Reboot the server having first ensured the Unit Number in the resmgr are set correctly, see above.

To identify a fault with "cpqunits" run:

# /etc/scsi/cpqunits -f -o -v -d

a fault can be seen here with -1 as the Unit Number: /etc/scsi/cpqunits:debug: RMgetvals 'MODNAME UNIT BRDBUSTYPE BRDID SLOT DEVNUM FUNCNUM IOADDR' returned <ciss -1 4 0xE11B178 0 1 0 2000 20ff> /etc/scsi/cpqunits:debug: rm_modname ciss /etc/scsi/cpqunits:debug: resource manager unit value -1 /etc/scsi/cpqunits:debug: decimal unit -1 /etc/scsi/cpqunits:debug: using PCI brdid e11b178 /etc/scsi/cpqunits:debug:

               CQHORD    resmgr
               ----------------
matching:       id 0x0e114080 ?= 0x0e11b178
            slot 0x00000000 ?= 0x00000000
          device 0x00000001 ?= 0x00000001
            func 0x00000000 ?= 0x00000000
/etc/scsi/cpqunits:debug:
                 CQHORD    resmgr
                 ----------------
matching:       id 0x0e11ffff ?= 0x0e11b178
            slot 0x00000000 ?= 0x00000000
          device 0x0000000f ?= 0x00000001
            func 0x00000000 ?= 0x00000000
/etc/scsi/cpqunits:debug: No match
/etc/scsi/cpqunits:debug: system unit number -1 (none)
/etc/scsi/cpqunits:verbose: Unit for ciss (rm_key=13) unit -1  matches -1.
A working result would look like:
  ./cpqunits:debug: using PCI brdid 10000010
  ./cpqunits:debug:
              CQHORD    resmgr
              ----------------
matching:       id 0x10000010 ?= 0x10000010
              slot 0x00000000 ?= 0x00000000
       device 0x00000001 ?= 0x00000001
         func 0x00000000 ?= 0x00000000

./cpqunits:debug: Match returning 0 ./cpqunits:debug: system unit number 0 (first) ./cpqunits:verbose: Unit for ida (rm_key=13) unit 0 matches 0. ./cpqunits:debug: rm_key 14

This issue has been reported to SCO Engineering.

- If you are not able to gain access to the server then boot from the Splash Screen, by hitting space bar when the Splash Screen appears during the boot process; when you get a command prompt, issue the commands

b unix.old
go

If the unix.old, or alternate, is not available then use your Emergency Recovery media or installation media to gain access to the server to resolve the issue.

NOTE:     The latest version of HP's EFS (v7.72) now no longer uses "cpqunits".

It is recommended to upgrade to this release if your current hardware platform supports it.

See: http://www.hp.com/support
NOTE:     Under SCO OpenServer Release 6.0.0 it may be necessary to remove the
          "ide" driver with "pkgrm", having first installed a copy of it in
          /local/drivers/ide, should your Secondary IDE controller conflict
          with the Onboard Network Card on IRQ 15.
With SCO OpenServer 6.0.0, removing the secondary IDE controller from
the DCU and whether you or do not run "idconfupdate -f", the IDE
controller will always return after a reboot.  It is, therefore,
necessary to remove it for networking to function.

This has been seen with the Compaq Proliant Generation 1 (G1) DL380 servers.

If you hardware is not supported by a later EFS revision then it may be advisable to "re-install" the EFS HBA.

From the Xinuos technical article database, record 109445. Reproduced with permission. The original is at www.sco.com/ta/109445, which is reachable but blocked from search engines.

Next step

Need this fixed rather than explained?

Tell us the release, what the system runs and what it is doing. We will reply by email with what we would do next.

  • You deal with the people who do the work
  • No obligation, and no upgrade required before we will help
  • We tell you plainly if something is not worth doing

Send an enquiry

We reply by email. If a server is down, say so in the first line. You can also write to help@scoexpert.com.