SCO OpenServer Release 5.0.7 Late News

This document supplements the SCO OpenServer Release 5.0.7 documentation with important notes and known limitations with possible workarounds:


Revision history

Check back here periodically. New information, including additional limitations and workarounds, is published as it becomes available.



Installation, upgrades, and licensing



Hardware installation notes and limitations


Mylex and LSI Logic drivers out-of-date
The Mylex (mdac) and LSI Logic (slha) host adapter drivers provided in this release are out-of-date.

If your hardware is not supported by these versions of the mdac or slha drivers, you may need to acquire the current manufacturer's BTLD:


SMP systems panic during boot
SCO OpenServer systems that are also installed with SMP may experience panics when booting, during kernel initialization. Testing has revealed that this problem typically occurs on SMP systems that also have:

In these cases, the panic dump usually reveals that the following command was the current process on CPU1:

/etc/sysdump -qi /dev/swap -o /dev/swap

There are two options for working around this problem. If you need to use USB devices, enter the following bootstring at the boot prompt:

defbootstr swap=none

(If desired, you can add the swap device back after booting your system by running the swap -a /dev/swap command.)

The other option is to disable the usb_ohci driver by entering the following bootstring at the boot prompt:

defbootstr disable=usb_ohci

A fix for this problem will be available in the near future.


Installing with Adaptec 2916x host adapters
If you want to install SCO OpenServer on a device connected to an Adaptec 2916x SCSI host adapter or compatible devices, you must use the ad160 driver from the CD-ROM Boot Disk or from the SCSI host adapter Boot Time Loadable Driver (BTLD) disk. The ad160 driver is not included on the floppy Boot Disk.

If you boot from the SCO OpenServer System CD-ROM, the driver will be available for the installation process. If you boot from the floppy Boot Disk, you must first install the ad160 driver from the BTLD disk, as described in ``Installing boot-time loadable drivers (BTLDs)''.


Installing on Adaptec dual channel host adapters
If you want to install SCO OpenServer on a device connected to an Adaptec Ultra160 Dual Channel SCSI host adapter, it is recommended that your CD-ROM and hard drives be on the primary bus of the first configured adapter. If the devices are attached to different channels or devices, the order of controllers may be changed when the system is rebooted, in which case the devices may not be recognized and the system may fail to start. This is a known bug that will be addressed in a future driver release from Adaptec.

If your CD-ROM and hard drives are attached to different channels (for example, the CD-ROM drive is on channel 0 and the hard drive is on channel 1), you can try to work around this problem with the following steps:

  1. When booting the installation, use the following bootstring to recognize the hard drive: 
       Sdsk=ad160(1,0,<ID>,0)
    where <ID> is the SCSI ID of the hard drive.

  2. To boot the system after the installation is complete, use the following bootstring: 
       Sdsk=ad160(0,0,<ID>,0)
    where <ID> is the SCSI ID of the hard drive.

  3. When the system is running, do one of the following:

    • Recommended method
      Edit the /etc/conf/cf.d/mscsi file and change the following line: 
         ad160 Sdsk 1 <ID> 0 0

      to: 

         ad160 Sdsk 0 <ID> 0 0

      and then relink the kernel and reboot the system.

    • Alternate method
      Append the following to the DEFBOOTSTR line in the /etc/default/boot file: 
         Sdsk=ad160(0,0,<ID>,0)

Problem installing with IBM ipsraid BTLD
If you use the IBM ipsraid BTLD to install your system, you may see a number of the following error message at some point during the installation: 
   dynamic linker: dlvr_audit: can't open /dev/zero

You can safely ignore these messages. Press <Return> to continue the installation.


Error in the PC Card host adapter documentation
The ``Adding SCSI PC Card host adapters'' section in the SCO OpenServer Handbook contains an error. The commands that you run to relink and reboot the kernel, as specified in Step 3, should be replaced with:

# /etc/conf/cf.d/link_unix -y
# shutdown


Blank or garbled display at boot time
Some systems incorrectly initialize the Primary Display Type bits in the CMOS Equipment Byte (0x14). Because SCO OpenServer relies on reading the Primary Display Type bits to determine which video driver to use, this initialization problem may result in SCO OpenServer incorrectly identifying the video as CGA instead of VGA.

This problem can manifest itself in a variety of ways, including:

This problem has been observed on some Intel 845 and 850 series motherboards, as well as IBM, Gateway, and Compaq systems and systems using the AMI BIOS issued between early 2002 and mid-2003. In most cases, this problem made it impossible to install SCO OpenServer on the above systems.

If you experience this problem, using the following bootstring to install osr;:

defbootstr cn.primary=vga


Video adapter might require accelerated video driver
Most video adapters can be operated with the generic VESA video driver. A more limited range of video adapters can be operated with accelerated, hardware-specific drivers. When a hardware-specific driver exists, the SCOadmin Video Configuration Manager will identify it. Accelerated drivers usually perform faster than the VESA driver.

If you experience problems with the operation of a video adapter, try switching between its accelerated and VESA drivers. An adapter not behaving correctly with one driver might operate correctly with the other. Specifically, if you experience system hangs using the VESA driver with either the Number Nine SR9 AGP video adapter or the Diamond Multimedia Stealth III Savage 4 video adapter, you should configure and use the corresponding accelerated video driver to avoid this problem.


Laptop display resolution
On certain laptop systems, setting the resolution too high can cause the display to lock up. For example, if 1024x768 does not work on your display, try 800x600.

IBM Thinkpad 600E system hangs during installation
During installation on an IBM Thinkpad 600E system, the installation might hang after loading the ramfs image.

To avoid this problem, start the installation by typing the following bootstring at the boot prompt:

defbootstr mem=1m-65280k


Toshiba laptops cannot install from CD-ROM
Toshiba laptop systems use a standard for booting from CD-ROM drives that is not supported by SCO OpenServer.

To install SCO OpenServer on a Toshiba laptop, you will first need to create a floppy Boot Disk. For information on how to do this, see ``SCO OpenServer installation media''.


Toshiba notebook might require bootstring
If your notebook has trouble reading from the floppy drive, verify that the bootstring contains fd.fifo=0. If not, add it to your installation bootstring. See the bootstring(HW) manual page for more information.

After the installation is complete, you should also add this bootstring to the /etc/default/boot file. For more information, see the boot(HW) manual page.


Text mode stretching on Toshiba laptops
Switching from text mode to graphics mode and then back to text mode might disable text mode stretching (that is, the text screen does not fill the entire LCD display). This problem has been observed on some Toshiba laptops using the Chips & Technology 65550 and 65554 video adapters. To restore stretched text mode, run your system setup and re-enable ``Text Mode Stretch'', or reboot your system.

Mouse on Toshiba notebook requires bootstring
By default, SCO OpenServer does not correctly detect or handle the built-in mouse (buttons and movement) on a Toshiba notebook.

To configure the mouse correctly for a Toshiba notebook, add kbm.id2chk to your bootstring when you start the installation. At the boot prompt, enter:

defbootstr kbm.id2chk


Configuring mice on laptops
When configuring internal mice on most laptops, select Keyboard mouse when prompted for the mouse type.

Problems with Toshiba laptops and PC Card adapters
If you are using an older Toshiba laptop with a PC Card adapter, you must set the PC Card Controller Mode in the laptop's BIOS to either ``Auto-Selected'' or ``CardBus/16-bit''.

The PC Card driver (pcic) will fail to start during system boot if the Controller Mode is set to ``PCIC Compatible''.


PCMCIA network adapters not autodetected
PCMCIA network adapters are not autodetected during the SCO OpenServer installation.

During the installation process, you can manually select and configure the PCMCIA network adapter, or you can select to defer networking configuration and then configure the adapter using the Network Configuration Manager after the installation is complete. For help configuring your network adapter after the system is installed, see Configuring Network Connections in the SCO OpenServer online documentation.


PC Card plug-and-play conflicts cause system hangs
If you are using plug-and-play PC Cards, conflicts in IRQs and I/O addresses can result in a hang during system startup.

If you experience this problem, you should reconfigure your PC Cards to avoid these conflicts. See ``Adding PC Card devices'' in the SCO OpenServer Handbook for information on determining the IRQs and I/O addresses that are available on your system.


Installing systems with multiple CD-ROM drives
The CD-ROM drive that you use to install SCO OpenServer is configured into the kernel as your default drive. If your system contains additional CD-ROM drives, you will need to configure them manually by running mkdev cdrom. See ``Adding SCSI/EIDE CD-ROM drives'' for more information.

ATAPI and USB CD-ROM drives are automatically recognized by the kernel, even if they are not configured. For these drive types, it is necessary to run mkdev cdrom to create device nodes and to permanently assign the drive unit numbers. You can collect hardware information about recognized CD-ROM drives by running hwconfig -h.


Problems booting SCSI systems with ATAPI media devices
If your system uses a SCSI primary hard disk and additionally contains an ATAPI media device (such as a CD-ROM or LS-120 drive) that is attached to an IDE controller, a problem with the IDE/ATAPI (wd) driver may result in your system being unable to boot from the hard disk.

To avoid this problem, enter the following bootstring at the boot prompt:

defbootstr hd=Sdsk

This bootstring specifies the SCSI hard disk as your boot device.



Installation notes and limitations


Do not use boot floppies that do not match your SCO OpenServer release
If you need to use a boot floppy to begin the SCO OpenServer Release 5.0.7 installation, you must use either the floppy Boot Disk provided in the SCO OpenServer Release 5.0.7 media kit or one of the floppy Boot Disks that you created from the floppy images on the SCO OpenServer Release 5.0.7 System CD-ROM.

Boot floppies from earlier releases of SCO OpenServer cannot be used to install SCO OpenServer Release 5.0.7 and Release 5.0.7 boot floppies cannot be used to install previous versions of SCO OpenServer. If you use the wrong release of boot floppy to install SCO OpenServer, it may appear that the installation was successful but system behavior will be unpredictable -- possibly unusable -- because of licensing conflicts.


Non-primary hard disk divisions not available
Some divisions configured during installation on a second SCSI hard disk might not be available after installation and are displayed by divvy(ADM) as not named.

To make those divisions available, use divvy after installation to name the divisions, but do not make any other changes with divvy. Then, install the updated division table to make all divisions available as configured.


Configuring non-root filesystems during ISL produces an error message
If you select to configure non-root filesystems during the SCO OpenServer installation, there are some circumstances where the following error message may display after the Setting file permissions phase of the installation: 
   expr: syntax error /opt/K/SCO/Unix/5.0.7Hw/cntl/ccsisl.post: test: argument expected

If you see this message, the installation will complete normally but some of the system configuration files may not be set up correctly. After rebooting the system at the end of the installation, check for the following possible problems:


Installing from a serial console
To install SCO OpenServer from a serial console onto a machine that does not have a video adapter, the required settings are 9600 baud, 8 data bits, one stop bit, and direct connect to COM1.

Installing Compaq IDA devices on non-Compaq platforms
The Compaq IDA driver checks the system BIOS to verify that it is on a Compaq system. If you use Compaq IDA hardware on non-Compaq platforms, you must add ida.nocheck to your installation bootstring or the driver will not function. See the bootstring(HW) manual page for more information.

After the installation is complete, you must also add this bootstring to the /etc/default/boot file. See the boot(HW) manual page for more information.


Installing Internet Services
If you want to install Internet Services, be sure the component is selected during initial system load. You cannot use the Software Manager to add the Internet Services after initial system load (whether Fresh or Upgrade installations).

UDK compatibility libraries and JRE generate installation warnings
When the UDK compatibility libraries and the Java Runtime Environment (JRE) are installed during the SCO OpenServer installation process, a number of installation error messages are logged in files in the /var/adm/install/logs directory. These log files include:

j2jre131.log
j2plg131.log
libC.log
libc.log
libm.log
OSRcmppkg.log
OSRcompat.log
urwfonts.log

In all cases, the following error messages can be ignored: 

   UX:pkginstall: ERROR: unable to create symbolic link from </usr/adm/pkg/<pkgname>/
   predepend> to </usr/options/<pkgname>.name>
   UX:pkginstall: WARNING: e-mail notification may have failed
   UX:pkgadd: WARNING: e-mail notification may have failed



Upgrade notes and limitations


Back up the /usr/local directory
When you perform an Upgrade installation, the system attempts to preserve and restore the contents of the /usr/local directory. If you have added files to the /usr/local directory or its subdirectories, you should manually back up the directory before performing an upgrade.

Cannot access tape drive on /dev/rct0 after upgrading
After doing an Upgrade installation, an existing tape drive will no longer be accessible on /dev/rct0. To fix this, run the following commands:

# rm /dev/rct0 /dev/nrct0 /dev/xct0
# ln -s /dev/rStp0 /dev/rct0
# ln -s /dev/nrStp0 /dev/nrct0
# ln -s /dev/xStp0 /dev/xct0


Upgrading USB devices
If a USB Zip drive or any other USB mass storage device is attached to your SCO OpenServer system, you must unconfigure these devices before beginning the Upgrade installation.

To unconfigure USB mass storage devices, edit the /etc/conf/cf.d/mscsi file and delete any lines that begin with the usbha string.

After you have completed the Upgrade installation, you can reconfigure your USB mass storage devices for use on the SCO OpenServer Release 5.0.7 system. See the usb(HW) manual page for more information.


SCO Internet Security Package does not work after upgrading
The SCO Internet Security Package does not work after you complete an Upgrade installation. To fix this:

  1. As root, run the Software Manager.

  2. Highlight SCO Internet Security Package then select Remove Software from the Software menu.

  3. Click on More Options in the dialog box.

  4. Select Leave Loaded then click on Remove.

  5. When the product is removed, select Install New from the Software menu.

  6. Select Loaded Software as the Media Device and click on Continue.

  7. Highlight SCO Internet Security Package then click on Install.

  8. Shut down and reboot the system.

Remove SCODB before upgrading
If SCODB, the kernel debugger, is linked into the kernel prior to an in-place upgrade, the upgrade fails with an error message similar to: 
   undefined symbol - scodbinit
To avoid this problem, before trying to upgrade to SCO OpenServer Release 5.0.7, deactivate SCODB in the link kit by changing Y to N in the file /etc/conf/sdevice.d/scodb.

If the system has already been upgraded, you can correct the problem by editing the file /etc/conf/cf.d/mdevice. However, be careful not to damage the format of this file because you cannot reconstruct it without reinstalling the system. On the line for ``scodb'', change the second field from ``P'' to ``PI''. (Because of the risk involved in editing /etc/conf/cf.d/mdevice, it is safer to deactivate SCODB, as described above, before upgrading.)



Licensing and registration notes and limitations

Some changes in the licensing policy have been introduced in SCO OpenServer Release 5.0.7. Important differences to note include:



Please note the following licensing and registration issues:


Upgrade licenses
If you want to upgrade your SCO OpenServer Release 5.0.5 or 5.0.6 system by performing one of the Fresh installation options, you must have both an SCO OpenServer Upgrade License and the original operating system license that accompanied your Release 5.0.5 or 5.0.6 system.

To properly license your system:

  1. Complete the Fresh installation to upgrade your system. When prompted, enter the data from your Upgrade license.

    When the installation is complete, however, the system will be in an unlicensed state.

  2. Boot the system into single-user mode. You are unable to complete the licensing of your system in multiuser mode.

  3. As root, run the License Manager and add the data from your original operating system license. When you are finished, exit the License Manager.

  4. Run /etc/sco_pmd. You can ignore the warning message that the policy manager was restarted.

  5. Once the policy manager is running, reboot the system so that the core kernel services are licensed.

See ``Licensing and registering SCO OpenServer products'' for more information on using the License Manager.


SCO Update Service licenses
If you purchased an SCO Update Service license or an SCO OpenServer Upgrade license with the SCO Update Service bundle, verify that the parameters on your new and original operating system licenses match. Specifically, if your original operating system license is a non-callback license, then your new operating system license must also be a non-callback license or the SCO Update Service will not function.

Registering the SCO Update Service
The SCO Update Service will not function unless you have registered the SCO Update Service license or you have a non-callback SCO Update Service license. Specifically, you will not be able install future Maintenance and Upgrade packs until both your operating system and the SCO Update Service license are registered. (Note that registering the SCO Update Service license automatically registers your operating system license as well.) See ``Registering products'' for details on registering products.



Runtime system



Hardware configuration notes and limitations


Floppy drive I/O errors
If you receive I/O errors when accessing the floppy drive on high-activity machines and you know the floppy media is sound, your system might be sending data faster than the floppy drive can process it. You can fix the problem by enabling FIFO buffering for the floppy device. To do so, edit /etc/conf/pack.d/fd/space.c and reset these variables from the initial value of 0:

fd_enable_FIFO
Set to 1 to enable FIFO buffering.

fd_FIFOthresh
Set the buffer threshold (in bytes) to a sufficiently high value for your system; the range is 1-15. We recommend setting the value to 13 for most Pentium processors, although you may have to experiment to find the appropriate value for your system.

System panics when IDE tape drive is configured but not present
If your system is configured for an IDE tape drive but the physical drive is not present, the system may panic when a tape status command is run. If you experience this problem, you should remove the tape drive from your system configuration by running mkdev tape. Select Configure a SCSI or Enhanced IDE tape drive, then choose Remove an Enhanced IDE tape drive.

PCI adapters requiring interrupts are recognized but do not work
Certain computers (in particular, some DEC models) have the ability to disable interrupts in the PCI BIOS. If an adapter is recognized but does not seem to be working properly, you can determine whether this is a problem by running hwconfig(C) and examining the output for the adapter. If the output contains the string vec=255, re-enable interrupts using your computer's hardware setup program. This does not apply to PCI adapters that do not require interrupts, such as graphics adapters.

Differences in the new USB implementation
The USB implementation in SCO OpenServer Release 5.0.7 is substantially different from the one provided in SCO OpenServer Release 5.0.6a. In particular:

For more details on the differences between the SCO OpenServer Release 5.0.6a and SCO OpenServer Release 5.0.7 USB implementations, compare the usb(HW) manual pages provided with both releases.


USB tape drives are not supported in SCO OpenServer Release 5.0.7
USB tape drives are not currently supported in SCO OpenServer Release 5.0.7. References to USB tape drive support in the SCO OpenServer documentation, including the usb(HW) manual page, are incorrect. USB tape drive support will be available in the near future.

USB host controllers cause system hangs or error messages while initializing
Due to variations in hardware and BIOS implementations, the operating system may hang or issue diagnostic messages while trying to initialize USB host controllers (HCIs). If you experience problems of this nature, try the following:

  1. You can completely disable the kernel's USB drivers with the following bootstring: 
       defbootstr disable=usb_uhci,usb_ohci,usb_ehci

    Even if you need USB enabled to complete the SCO OpenServer installation (for example, your CD-ROM drive is a USB device), this step is a quick way to determine if your problems are associated with USB. If you disable all of the HCIs on your system and your system problems are eliminated, you can be certain that USB is the source of trouble.

    Many machines have more than one type of HCI. You can experiment by disabling different combinations of HCI drivers to see if you can pinpoint the exact source of your USB problem.

    If you determine that USB is causing problems on your system, but you need USB enabled to complete your installation, proceed with the following steps.

  2. Check to see if USB controllers are enabled in BIOS setup. Some systems have separate BIOS setup options for enabling USB controllers and their interrupts. To use USB, make sure both options are enabled. If you do not want to use USB and it seems to be causing system problems, make sure that both options are disabled.

  3. Check your system's BIOS setup for USB keyboard configuration. This may appear as Legacy keyboard port emulation, Legacy keyboard emulation, USB keyboard support, or some other similar option. Depending on your system, using BIOS keyboard emulation may help or cause system problems. If the kernel is hanging while attempting to enable USB HCIs, change the current keyboard emulation setting to the alternate value and see if that helps.

    If the only USB device that you need is a keyboard and the BIOS supports keyboard emulation, an option is to disable the SCO OpenServer USB drivers (see Step 1) and allow the BIOS to fully control the keyboard. If you do this, SCO OpenServer recognizes the keyboard as a PS/2 (legacy) keyboard.


LED states on a hot-plugged USB keyboard
If you hot-plug a USB keyboard, the LEDs do not reflect the current shift state of the keyboard until you press a key that affects the state of the LEDs -- for example, the <CapsLock> or <NumLock> keys.

USB device enumeration errors
If you are booting a system with USB devices attached, you may see one of the following messages: 
   Descriptor load failed during enumeration
or 
   Device address set failed during enumeration

These messages are an indicator that one or more of your USB devices were not detected and are probably not functional. To work around this problem, unplug any USB devices that are not working then plug them back in.

A fix for this device enumeration problem will be made available as soon as possible.


Addendum to supported PCI serial cards
In addition to the PCI serial cards listed in ``PCI serial cards'' in the SCO OpenServer Handbook, the following cards are also supported in SCO OpenServer Release 5.0.7:

PCI serial card ports and baud rates

Manufacturer Board type Number of ports Maximum baud rate
    RS-232 RS-485  
    interface interface  
Digiboard ClassicBoard PCI 4 4 - 921600
Digiboard ClassicBoard PCI 8 8 - 921600


New PC Card modems supported in SCO OpenServer
The following PC Card modems (PCMCIA and CardBus) are now supported in SCO OpenServer Release 5.0.7:

Modem/Serial Cards
3Com 56K Global Modem (3CCM156B)
3Com Megahertz 56K Cellular Modem (3CCM556)
ActionTec 56K DataLink PC Card (FM560LK)
Socket Communications I/O
Xircom 56K Modem PC Card (XM5600)
Xircom RealPort Modem 56K Global Access (RM56G)


Configuring PC Card serial modems
To configure a PC Card serial modem, you must select the correct serial port to use. To configure the modem on the COM1 port:

  1. Run the Serial Manager or the mkdev serial command.

  2. If currently configured, delete IBM-COM1.

  3. Click on Add and select PC-Card-COM1 from the point-and-pick list.

  4. Click on Close.

  5. Relink the kernel. You must shut down and reboot the system for your changes to take effect.

You can also configure your PC Card modem to use the COM2 and COM3 ports by selecting PC-Card-COM2 or PC-Card-COM3.

NOTE: SCO OpenServer only supports one configured PC Card fax, modem, or serial card at any given time.

NOTE: Before configuring a PC Card serial modem, make sure that you have disabled the corresponding internal COM port in your laptop's BIOS. If you fail to do this, the PC Card modem will not work and your system may hang when it is restarted.

If you experience this problem, reboot the system and use the unix.old kernel at the boot prompt until you have corrected the situation.

There have been some reports of PC Card modems hanging after periods of use. This problem is under investigation. If you experience this problem, you will need to use an external modem instead.


Error when configuring PC Card modems for dial-in
When configuring some PC Card modems for dial-in, you may see an error message similar to: 
   WARNING: serial: Garbage or loose cable on dev 0, port shut down

You can ignore this message. Your modem is configured correctly.


LinkSys EtherFast 10/100 PC Card not supported
Although the LinkSys EtherFast 10/100 (PCMPC100) PC Card shares the same PC Card ID with the supported LinkSys EC2T Combo PCMCIA EthernetCard -- and it is autodetected by the Network Configuration Manager as an EC2T card -- the PCMPC100 card is not supported in SCO OpenServer Release 5.0.7. Even if you configure this card using the Network Configuration Manager, it will not work on SCO OpenServer.



Commands and System Software notes and limitations



Documentation and help


DocView search index not created automatically
The DocView search system, based on the htdig search engine, requires that a search index be generated. Because indexing the SCO OpenServer online Documentation Library can take awhile, the index is not generated during the installation of DocView.

To generate the search index, run the following command:

#  /usr/lib/docview/conf/rundig

For more information, see ``Generating the search index'', which is accessible from the SCO OpenServer online Documentation Library.


Serving manual pages using the man(C) facility
The man(C) facility for serving manual pages using the MANSERVER and MANSOCKETFILE variables does not work as documented in SCO OpenServer Release 5.0.7. Manual pages can be served on a network using the DocView help system; for more information, see ``Using DocView''.

IPX/SPX context-sensitive help
When you install the documentation for IPX/SPX, the context-sensitive help is not automatically installed. To access context-sensitive help, you must also install the TCP/IP documentation.

Help on Index broken in several SCOadmin managers
Help on Index in several SCOadmin managers is resolving incorrectly to the ``Controlling processes with the Process Manager'' topic in the Handbook instead of to the appropriate Index.

Affected managers include:

Event Logs
Hardware/Kernel
International Settings
License
Modem
Process
Software
System Defaults
System Shutdown
System Time

A fix for this problem will be available in a future release. In the meantime, the indexes for the SCO OpenServer Handbook and the System Administration Guide are available from the SCO OpenServer online Documentation Library.


Viewing AFPS documentation in DocView
To fully integrate the AFPS documentation with the new DocView documentation server, you must install the Advanced File and Print Server 4.0.2 Supplement (SLS APP619B). This supplement is available on the Optional Services CD-ROM.

If you do not install this supplement, you can still access the AFPS documentation from the SCOhelp Home Page link on the DocView home page.


Optional Services upgrade documentation
Some documentation for products on the Optional Services CD-ROM might refer to upgrading Optional Services from SCO OpenServer Release 5.0.0 or 5.0.2. Note that, while the information in the documentation is still applicable, upgrades of Optional Services products are only supported from SCO OpenServer Release 5.0.5 and later.



Kernel issues


Hyper-Threading technology not supported
The Hyper-Threading technology that is provided in Intel Xeon and some Pentium 4 processors is not supported in SCO OpenServer.

If your system BIOS provides an option for enabling and disabling Hyper-Threading, you should disable it. The BIOS may refer to this functionality as Logical Processors or Virtual CPU.


k_trap warning
On any Pentium Pro system, you might see the message: 
   WARNING: k_trap - Intel Reserved Fault (#15) trap in kernel mode
According to Intel Errata for the Pentium P6, this message reports about an interrupt that can be safely ignored.

Kernel link failure with large kernels
The SCO OpenServer kernel may fail to link if the total load size of the kernel is greater than 8MB. If you encounter this problem, you see messages similar to: 
   i386ld locore.o 40: can't allocate section .bss into owner KV_loads
   i386ld locore.o 40: illegal operator in expression
   ERROR: Can not link-edit unix

Total load size includes the size of .text (executable code), .data (initialized data), and .bss (uninitialized data) kernel segments. It does not include data that is dynamically allocated by the kernel after it is loaded (for example, the buffer cache). The kernel load size is heavily influenced by the size of the device drivers that are linked in and the values of some kernel parameters that allocate memory.

If you experience this problem, you must omit drivers from the kernel or decrease tunable parameters. The memory costs of many tunable parameters are documented in the Performance Guide. Driver sizes can be checked by running size(CP) on the driver's Driver.o and space.o files. (Link the kernel with link_unix -p to preserve space.o files generated during the link.)



Files and filesystems


Filesystems
Note the following:

File size
If you need to use files larger than 1GB on an SCO OpenServer system, change the definition of ULIMIT from ``2097151'' to a value of up to ``4194303''. Use the configure(ADM) command to modify the ULIMIT parameter. You can also change the limit imposed by login(M) in the /etc/default/login file. (See ``User and group configuration'' for more information.) The maximum size supported is 2GB.

File versioning and directories
When file versioning is in effect (see ``Retrieving deleted files''), one or more saved copies of a file may exist in a directory, in addition to the current version of the file. (The files are distinguished by a version number suffix, separated from the filename by a semicolon.) The kernel parameter MAXVDEPTH is used to control the number of saved copies that exist at any one time. If the file is changed and there are already MAXVDEPTH previous versions present in the directory, the oldest previous version will be deleted and a new version created. In the following example, MAXVDEPTH is set to 3. 
   $ ls -l
   total 8
   -rw-r--r--    1 root     sys          126 Feb 16 16:54 file
   -rw-r--r--    1 root     sys          126 Feb 16 16:51 file;1
   -rw-r--r--    1 root     sys          126 Feb 16 16:52 file;2
   -rw-r--r--    1 root     sys          126 Feb 16 16:53 file;3
   $ rm file
   $ ls -l
   total 6
   -rw-r--r--    1 root     sys          126 Feb 16 16:52 file;2
   -rw-r--r--    1 root     sys          126 Feb 16 16:53 file;3
   -rw-r--r--    1 root     sys          126 Feb 16 16:54 file;4
   $
Directories can be versioned as well, but the behavior of directory versioning is slightly different. A directory can normally only be removed by rmdir when it contains no files at all. This still applies when versioning is enabled, and a directory containing no files will be removed completely. If, however, a directory contains file versions but no current files, the directory will be versioned. Directory versions follow the same naming convention as file versions.

Because MAXVDEPTH does not apply to directories, it is possible for any number of directory versions of the same name to be created. If a directory version exists and a file is then created with the same name, the directory version will always be preserved. If the MAXVDEPTH limit for the file is exceeded, the oldest non-directory version of that name will be removed. If possible, avoid mixing file and directory versions in this way; administration of such mixed versions requires some care.



Commands


ELF, COFF, and OMF/x.out binary support
The SCO OpenServer kernel supports native use of ELF and COFF binaries that use the SCO OpenServer ABI (Application Binary Interface) and/or the iBCS2 (Intel Binary Compatibility Standard, 2nd edition) ABI. OMF/x.out binaries that use the XENIX and/or iBCS2 ABI are also supported. Dynamic ELF binaries compiled on SCO UnixWare SVR4 or SVR5 platforms that use the SVR4 ABI are supported on SCO OpenServer through the OSRcompat module.

The hw(ADM) command not updated for the PC Card subsystem
The hw(ADM) command has not yet been updated to use the PC Card subsystem. However, hw does currently have some PC Card detection abilities.

To see additional PC Card information, run the pccardslot command.


OSRcompat package and JRE not supported on earlier SCO OpenServer releases
The OSRcompat package and the Java Runtime Environment (JRE) that are provided in SCO OpenServer Release 5.0.7 are not supported for use on systems prior to Release 5.0.6a. You may install these packages on earlier SCO OpenServer releases, but you do so at your own risk. The software may not function correctly.

System hangs after running reboot(ADM) command
On some systems, running reboot(ADM) may cause the machine to lock up after shutting down, such that you need to do a hardware reset to restore the system.

When a system is running in multiuser mode, it is recommended that you always use the init(M) process instead of reboot or haltsys. Use init 0 to power off the system. Use init 6 to reboot the system.



SCOadmin managers


Account Manager
Note the following:

Backup Manager
When restoring a file, no error message is displayed if the file is not extracted. Be sure the desired file is actually on the backup and that you specify the correct pathname.

Filesystem Manager
Note the following:

Modem Manager
Configuring a modem with the Modem Manager in the character (CHARM) interface may hang the manager when exiting. If you then kill the Modem Manager process, the error Error:/etc/mkdev modem did not complete successfully appears on the screen. If this problem occurs, you will need to configure your modem using the graphical interface instead of the character environment.

Printer Manager
Hewlett-Packard JetDirect® Ethernet printers display with serial attributes when queried.

If you cannot transfer print jobs between two local printers as a non-root user, it may be because you have remote printing installed on your system.

To work around this problem, you need to access the non-remote command, lpmove(ADM), from the command line. For example, you could enter:

/usr/lpd/local/lpmove request-id localprinter


Print Job Manager
Note the following:

System Defaults Manager
Note the following:

User Equivalence Manager
Depending on the type of name service you are using, you might find the User Equivalence Manager does not create .rhosts files that allow remote access. The User Equivalence Manager uses the output of hostname(ADMN) in the .rhosts file. Some network configurations require only the host name itself, not the fully-qualified name (which includes the domain) that is returned by hostname.

Virtual Disk Manager
The Virtual Disk Manager, as a software-only implementation, is designed primarily for data integrity rather than high availability. Availability requires additional hardware support: hot-swap-capable drives, chassis, host adapters, and possibly host adapter drivers.

Without additional hardware support, the Virtual Disk Manager can only guarantee data recoverability. It cannot guarantee that the system will continue running if a partial hardware failure occurs. The effects of a hardware failure depend on the hardware itself and its drivers, which must be deliberately designed to present a ``clean'' failure mode to the vdisk(HW) software.

Note the following issues with the Virtual Disk Manager:



System administration and configuration

Note the following system administration and configuration issues:


Applications in scoansi dump core
If you have an existing application that began dumping core with SCO OpenServer Release 5.0.6a and your TERM variable is set to scoansi, try setting $TERM to scoansi -1k and restarting the application. For example,

If this corrects the problem, this means that the application has a termcap buffer that is too small. Contact your application provider to correct the problem. Application providers should review the corrected termcap(F) manual page.


Core dump tunables
An undocumented feature in SCO OpenServer is the ability to control the processes that dump core using tunable parameters in the /etc/conf/pack.d/kernel/space.c file. For more information, see the comments in the file.

WARNING: Use extreme caution when modifying this file. Accidental changes may make your system unusable.


Perl and the PHP Scripting Language for Apache
By default, the Apache package has PHP enabled. The PHP that ships with SCO OpenServer Release 5.0.7 has an extensive list of optional modules that are also enabled. Additionally, mod_perl is enabled by default, linking a full Perl interpreter into the Apache web server.

The PHP and mod_perl modules can consume a considerable amount of memory. If you are not using them, you should either stop Apache from loading them by editing /usr/lib/apache/conf/cf.d/httpd.conf, or edit /usr/lib/php/php.ini to fine-tune PHP to load only the extra modules required by your PHP scripts. Search for extension= in the php.ini file to see the list of extensions that are loaded by default.


PHP phpinfo() function reports several library versions incorrectly
The PHP phpinfo() function reports several library versions incorrectly. This is because PHP was compiled with earlier versions of these libraries present on the system. The table below lists the library, the version that phpinfo() reports, and the actual library version installed:

Library Version reported by phpinfo() Correct version
OpenSSL 0.9.6g 0.9.6h
OpenLDAP 2.1.9 2.1.12
libxml 2.4.28 2.5.2
libxslt 1.0.23 1.0.25
libexslt 0.7.14 0.7.16


Reducing the number of console multiscreens with configure(ADM)
If you use configure(ADM) to reduce the number of console multiscreens, you should also edit /etc/default/mapchan and /etc/default/mapkey and remove entries for console ttys that no longer exist. If this is not done, mapchan will produce errors similar to the following when it is run during system startup: 
   could not open /dev/tty08
   could not open /dev/tty09

SCODB
The DBKEY tunable parameter now applies only to serial consoles. The debugger key on video console keyboards is set using the ``debug'' keyword in the keyboard map (see the mapkey(M) manual page) and defaults to <Ctrl>-<Alt>-D.

SCO Visual Tcl
If the documentation for an option argument does not list the G attribute and you attempt to use VtGetValues to obtain the value of that option argument, the interpreter will hang.

Volution Manager client not started by default
The Volution Manager client is automatically installed as part of the SCO OpenServer system. However, you must manually start the client before you can make use of Volution Manager functionality. See ``Managing your system with Caldera Volution Manager'' for information on how to do this.



Mail


MMDF Alias Administration Manager
Note the following:

MMDF Host Administration Manager
The MMDF Host Administration Manager ignores attempts to remove the last remaining host in a given table. If you must remove a single host from a table, edit the corresponding table file in the /usr/mmdf/table directory (this is the default location for table files, but this can be changed in /usr/mmdf/mmdftailor).

SCOmail
Note the following:

SCOmail online help
The SCOmail online help has not been updated to reflect these additions to SCOmail functionality:

automatic saves
SCOmail automatically saves any messages under composition if scomail is shut down while composition is in progress.

automatic prompts
SCOmail automatically prompts you to resume any saved messages the next time scomail is started. After all messages have been ``resumed'' and disposed of by sending the mail or abandoning the compose session, the resume dialog will no longer appear when scomail is started.

changing types or character sets
Incoming mail attachments can have their type changed to whatever the user prefers. For example, if someone sends a GIF image (picture) and accidentally marks the type as generic-data, scomail allows the type to be changed to GIF (or any other type), and the attachment will be displayed properly.

Incoming mail messages or attachments can have their character set changed to whatever the user prefers. For example, some systems use multiple character sets (European users can use ISO-8859-1 through ISO-8859-10); if someone sends a plain ASCII text file, but marks it as ISO-8869-7 by mistake, it may not display properly.

To change either the type or character set of a message in your mailbox, highlight the message or attachment, select Change Attributes from the Message menu, and make the necessary modification.



Localization


Setting the system locale
SCO OpenServer binaries that call setlocale(S) will only work using those locales supported on SCO OpenServer systems. You get setlocale syntax errors on kernel relink when the locale has a ``-'' in it. If a non-localized or non-SCO product, such as ghostview, is run, a large number of error messages are displayed, saying that an error in the locale occurred when setting language-dependent environment variables, such as those which start with the prefix LC_*. Avoid this by setting the LANG variable to german_germany.8859 or french_france.8859 (choose a locale that does not include a ``-'' in the locale name).

To allow applications to run correctly, set LANG (and any other LC_* variables as required) to a value from /etc/default/lang that does not contain the ``-'' character in the environment in which the application is to be run. For example, if an application binary's name is foo, you could enter the following in the Korn shell to run the application in the ``french_france.8859'' locale:

env LANG=french_france.8859 foo

If LANG is not set, the locale will default to the first entry in /etc/default/lang, ``C_C.C'' (the POSIX locale). This will not produce any warnings or errors.

If LANG is set to an entry that is not present in /etc/default/lang, errors or warnings might be produced because setlocale will be forced to read entries in the file containing the ``-'' character.


Building DocView search index in the German locale produces error message
If you are using the German locale, running /usr/lib/docview/conf/rundig to generate the DocView search index may generate an error message similar to: 
   Warning: unknown locale!

You can safely ignore this message -- it has no effect on the indexing process.


Restarting the calendar server with IQM_LANGUAGE
The steps given in ``Calendars'' to restart the calendar server with the initial LANG setting are not correct.

Below are the correct steps:

  1. Change to the /usr/lib/sco/oadb directory and remove all of the data files in the caldata directory with:

    cd /usr/lib/sco/oadb
    rm caldata/*

  2. Enter:

    DBKEY=6373
    IQMFILE=/usr/adm/ISL/iqm_file
    . $IQMFILE
    LANG=$IQM_LANGUAGE
    export DBKEY IQMFILE LANG

  3. Change to the /usr/lib/scosh/utilbin directory and rebuild the database with:

    cd /usr/lib/scosh/utilbin
    ./calbuild

You should now be able to administer the calendar database.

To make this change permanent on your system, you must also:

  1. Edit the file /etc/rc2.d/P95calserver and find the line: 
    DBKEY=6373;     export DBKEY

  2. Before that line, add: 
    IQMFILE=/usr/adm/ISL/iqm_file
. $IQMFILE
LANG=$IQM_LANGUAGE
export IQMFILE LANG
This will enable the calendar server to start using the locale that was set at installation time, regardless of the system locale setting.



Graphical Environment notes and limitations



X server


Maximum number of local X servers
SCO OpenServer supports up to 8 local X servers.

VGA and Super-VGA drivers do not run if you switch multiscreens
If you are using a Super VGA or VGA-only adapter that does not contain a graphics acceleration chip, the driver cannot render graphics in the background when you switch to a different multiscreen. X applications will block (stop running) until you return to the multiscreen on which the X server is running. This occurs with the mw driver and the following Chips & Technology drivers: ct8, ct16, ct65548_8, ct65548_16, ct65550_8, ct65550_16.

Drawing problems when moving Desktop icons
Occasionally, if Desktop icons are moved across the top of the Desktop under the menu bar, incorrect pixels are drawn. Using Refresh from the window manager Root menu clears the incorrectly drawn pixels.

Color problems when using TrueColor and DirectColor with SCO X clients
The SCOcolor and SCOpaint clients do not work with TrueColor and DirectColor visual types. DirectColor and TrueColor adapters typically support 32K, 64K, and 16 Million colors, but SCOcolor and SCOpaint clients do not work in these modes. If you want to use these clients, configure your graphics adapter for 256 colors.

This problem also results if you start Xsco with the -static option.

The SCOcolor and SCOpaint clients also do not work with fixed colormap graphics adapters.


Problems switching screens
You may experience problems switching screens on newer and faster machines. When you switch from the X Server (graphics screen) to one of the text screens, the text screen may become unusable.

You can correct this problem by adding a wait(1000000) call to the end of the SetText procedure in the grafinfo file for the configured adapter: 

   PROCEDURE SetText
   {
        /* restore text mode */
        r0 = 0x0003;
        int10(r0, 1);
        wait(1000000);
   }

To determine the location of your adapter's grafinfo file, consult the configuration information for your adapter in ``Supported video adapters''.



Desktop and Graphical Environment utilities


Desktop
Note the following:

SCO Panner window manager
Note the following:

Bell
You cannot use the scobell client to change the system bell volume on the console. This does work on an X terminal, however.

Calendar
The View Free window always displays the full width of your screen. If you try to resize the window, it automatically reverts to its original dimensions.

Color
On a high-security system, any non-default color palette used during the previous session is not preserved. However, that palette will still appear in the palette list when scocolor is run. The user can select this palette, and it will be preserved for all future sessions.

Lock
In some instances, scolock does not release its lock on the display when a key is pressed. In these cases, either the password prompt does not appear when running scolock in its default mode or you do not return to your Graphical Environment session if running scolock in screen-saver mode.

If either situation occurs, press a mouse button to bring up the password prompt or remove the screen saver.


Paint
Note the following:

Print Setup
When used for networked printers, the Print Setup program that is accessed from the Printer icon includes buttons in the dialog box that cannot be used to make your selections. Instead, enter the options you want in the ``Print Command'' field.

SCOterm
Note the following:



Networking notes and limitations



Address Allocation Server (AAS)



Non-SCO LLI driver compatibility

The STREAMS implementation in SCO OpenServer Release 5 has changed from that of earlier releases so that the memory within a single data block is not guaranteed to reside in a single, physically-contiguous block. This creates problems for STREAMS drivers that control devices with DMA transfer functions. All SCO-provided network drivers have been modified for compatibility with the new STREAMS architecture; non-SCO LLI drivers for DMA-based network adapters (such as ZYNX) may not work correctly with the default SCO OpenServer STRMAXBLK setting.

The new STRMAXBLK kernel parameter defines the maximum size of STREAMS buffers. The default value of the STRMAXBLK parameter is 524288, but it can be set to any value that is a power of two between 4096 and 524288 (4KB to 512KB). Setting STRMAXBLK to 4096 has the effect of ensuring that all buffers allocated from STREAMS will be located in a physically contiguous block, which enables older LLI drivers to work properly with SCO protocol stacks and other networking products. Note that changing STRMAXBLK to 4096 does not ensure that LLI drivers can work with non-SCO protocols that map in private data blocks.

If you are using a DMA-based network adapter with a non-SCO provided LLI driver, use the Hardware/Kernel Manager to set the value of STRMAXBLK to 4096 bytes, then relink the kernel and reboot the system.



TCP/IP


Configuration and startup changes for TCP/IP
Under certain conditions, it might appear that the network on your SCO OpenServer system is not functioning properly. Upon closer inspection, you might discover that a default route has not been set or that a routing daemon did not start as expected. This might be due to changes in how configuration information is stored and changes in the TCP/IP startup script.

You might encounter these problems if you:

To get your network functioning properly:

  1. Determining the network status of your system.

    You can check if a default route exists by displaying the routing tables with:

    netstat -r

    This command produces output similar to: 

      Routing tables
  Destination      Gateway            Flags    Refs      Use  Interface
  default          gateway1           UGS        38   355378  net0
  localhost        localhost          UH         37    52026  lo0
  198.0.0          system1            UC          1        0  net0
  system1          localhost          UGHS        2       30  lo0
  BASE-ADDRESS.MCA system1            UGS         0        0  net0
    (If you have trouble with name resolution, you can display network addresses instead of system and network names by adding the -n option to the netstat(TC) command.)

    If the destination column does not contain an entry for ``default'', then a default route has not been configured on your system.

    You can check if a router daemon is running by entering:

    ps -e | egrep "gated|routed"

    If the command returns with no output, no router daemon is running, which is normal for many network configurations because it is usually sufficient to have a default route that redirects outbound network traffic to a specific network interface. A router daemon is only needed on certain complex network configurations. If a router daemon is needed, your network administrator should be able to identify the correct daemon for your network. In no case should more than one routing daemon be running.

    To see if the resolver configuration file has been set up as you expect, take a quick look at it with:

    cat /etc/resolv.conf

    If the file does not exist or does not have the values you expect, you can edit this file manually -- see the resolv.conf(SFF) manual page .

  2. Understand how the installation has changed and where these network configuration parameters are stored.

    The network configuration screen of the initial installation has some new fields for TCP/IP configuration, which affects how network configuration files are created and used.


    Gateway address
    You can configure a default route by supplying the IP address of the system to which network traffic should be directed if no other routes can resolve the destination. The value you place in this field is stored in the file /etc/default/tcp as the GATEWAY variable, which is used by the /etc/tcp startup script to determine if a default route should be established. If the variable is set, the startup script executes the following command:

    route add default $GATEWAY $DFLT_METRIC

    If you leave this field blank during initial installation, the ``GATEWAY='' line in /etc/default/tcp exists without a value assigned. Because the variable is not set, /etc/tcp does not establish a default route.

    NOTE: DFLT_METRIC is also in /etc/default/tcp to specify a default metric for the route command, but is not set. You can edit the /etc/default/tcp file to place the desired value on the ``DFLT_METRIC='' line.


    Nameserver addresses
    You can configure a primary DNS nameserver and a secondary DNS nameserver by supplying the IP addresses, which are then stored in /etc/resolv.conf. If you leave both of these fields blank during initial installation, the file /etc/resolv.conf is not created.

    NOTE: To supply a third nameserver address, you should edit the /etc/resolv.conf file manually after installation.

  3. Understand how TCP/IP startup behavior has changed.

    Two routing daemons (routed and gated) are available, but they are mutually exclusive. One (but not both) of these daemons can be configured to start by editing the configuration file /etc/default/tcp. The STARTDAEMON[0-9] variables allow you to start up to ten TCP/IP-related daemons. You need to supply the absolute path to the daemon to be used and the arguments with which to start the daemon, respectively. If the command contains spaces or characters special to the shell, they must be quoted.

    These variables are not set during installation, so routing daemons will not be started automatically for you. This is fine if you have specified a gateway address, or do not require routing tables, but if your system participates in a larger network where routing tables are updated through the routing daemon propagation protocols, you must specify which daemon to use and its arguments by editing the /etc/default/tcp file and setting the appropriate STARTDAEMONx variable. For example: 

       STARTDAEMON0='/etc/routed -q'
   STOPDAEMON0='tcp:routed'
    or 
       STARTDAEMON4='/etc/gated -f /etc/gated.conf'
   STOPDAEMON4='tcp:gated'

    These variables might not exist in the /etc/default/tcp file if you have performed an upgrade. If automatic start-up of one of these daemons is required, simply add the desired variable and appropriate definitions to the /etc/default/tcp file.

    Be sure to unset the GATEWAY variable, if its existence is inappropriate with the running of one of these daemons.

    The SECURESHELL variable, also set in the /etc/default/tcp file, controls startup of the Secure Shell (SSH). By default, this variable is set to ``YES'', which means that the sshd(8) daemon is automatically started when TCP/IP starts. Even if sshd isn't started automatically, the pseudo-random number generator daemon (prngd(1)) is always run because the OpenSSH library is dependent on it.

    See the comments in the /etc/default/tcp file for more details on the syntax of the STARTDAEMON[0-9], STOPDAEMON[0-9], and SECURESHELL variables. Read the manual pages for routed(ADMN), gated(ADMN), and gated.conf(SFF) for further information on setting up the routed and gated daemons.

  4. Restart the network or the system (if necessary) to fix the problems you have encountered due to these changes.

    A change made to /etc/resolv.conf is used by new processes requiring name resolution services. Some daemons, or long-running programs, might need to be restarted to reread /etc/resolv.conf. If you detect inconsistent name resolution behavior, you might need to either restart the services that are not working properly or reboot the system.

    Changes to /etc/default/tcp require a restart of TCP/IP using one of these methods:

    • In most situations, you can restart the TCP/IP network by simply running these commands:

      /etc/tcp stop
      /usr/bin/sd tcp start

      The new network environment should reflect the changes you made.

      or

    • In some situations, you might need to reboot the system (for example, if network does not start up properly or if some service that requires network connectivity became confused when the network was restarted).

rcmd and rlogin appear to not work after setting up .rhosts file
If the root .rhosts file has read or write permissions greater than 0600 (that is, any combination of group or other permissions readable or writable), the file will be ignored, which will cause rcmd and rlogin to not work as it did previously.

This change tightens the security handling for use of root .rhosts files (disallowing use if readable by group or other) and puts the software behavior in agreement with what has been documented in previous releases.

If you see this problem, remove group and other read and write permissions from the .rhosts file with the following command:

chmod 600 .rhosts


Incompatibility between bootpd(ADMN) and tftpd(ADMN) in secure mode
An incompatibility exists between bootpd(ADMN) and tftpd(ADMN) when tftpd is operated in secure mode.

In a typical scenario, a host needing system boot information (a BOOTP client) requests of a BOOTP server a file containing that information. The BOOTP server responds to the client confirming the existence of the file. The client host then executes the tftp(TC) command to request the transfer of that file from the server to itself. The server host's tftp server, tftpd(ADMN), fields the tftp request and forwards the file.

When tftpd is operating in secure mode, it prepends a secure directory path to any file pathname it receives in a tftp request. It then looks for the file by that complete pathname. If the BOOTP client host already includes the secure directory path in its tftp request, tftpd in secure mode creates a pathname for a file that does not exist when it prepends the secure directory path to the path sent in the tftp request.

For example, the BOOTP client requests the file /tftpboot/bootfile. The BOOTP server confirms the existence of this file. The BOOTP client host requests this file with tftp. The secure directory path of the tftpd server is /tftpboot. The tftpd server prepends this secure directory path to /tftpboot/bootfile and looks for the file /tftpboot/tftpboot/bootfile which it cannot find.

There are two ways to work around this problem:



Domain Name Service (DNS)


Changes to the contents of /etc/resolv.conf
In this release of SCO OpenServer, the /etc/resolv.conf file might be created for you during a fresh installation; however, if you examine this file, you might notice that the domain line is missing, which is intentional because the domain line is redundant with the search line. While the domain keyword is still legal, if the domain line appears after the search line, the search line is ignored. To prevent confusion or misunderstanding, the /etc/resolv.conf file is now created without the domain line.

Other configuration information placed in the /etc/resolv.conf file during installation includes:

The /etc/resolv.conf file is created only if nameserver addresses and the system domain name were entered during initial installation.

For more detailed instructions on resolver configuration, see the resolv.conf(SFF) manual page.



SCO Gateway for NetWare

Note the following issues with SCO Gateway for NetWare:


Printing NetWare files on SCO Gateway for NetWare printers
When the data file is located on a NetWare server, a non-root user cannot submit a SCO Gateway for NetWare print request if root is not logged in to the printer's NetWare server.

To avoid this problem, copy the data file to a non-NetWare filesystem and submit the print request from there. Or execute:

cat filename | lp

Another solution is to log in as root on the appropriate NetWare server.


Printing to Altos NetWare for UNIX servers
There are two problems caused by differences between NetWare for Altos/SCO UNIX v3.01B and Novell® NetWare® 3.11.

Undefined symbols appear when trying to load NLMs on NetWare 3.11 server
You must install a NetWare server patch. The patch is available from ftp.novell.com, as /pub/updates/nwos/nw312/libup5.exe.

NLMs may allow unauthorized access to files
If you chmod(C) a directory to mode 0777, NetWare read and write rights are granted to all trustees. If you then change the directory to mode 0, the read and write rights are not removed. DOS users still have access to files in that directory. This also applies with modes 0070, 0007, and others.

NLMs may prevent access to directory
A directory created with mode 0111 will only add the NetWare read and write rights for the owner; the group and others will have no rights. In other words, creating a directory with mode 0111 is equivalent to creating it with mode 0100.

nwlogin/nwlogout fail if login connection is lost
If a NetWare server goes down or is rebooted while a user is logged in through the SCO Gateway for NetWare to the server and is using the server's filesystem, some utilities will not work properly. Both nwlogin(NWG) and nwlogout(NWG) will fail. nwwho(NWG) will say that the user is logged in, whereas nwwhoami(NWG) will say that the user is not logged in. slist(NWG) will hang if the server is the primary server.

CAUTION: Do not try to restart the SCO Gateway for NetWare by using nuc start, nuc stop, or nuc restart. These commands will fail and you must then reboot the system.

To work around this problem:

  1. Stop all processes that are using the NetWare filesystems. For example, users currently in NetWare directories should change directory to a local UNIX filesystem.

  2. Wait for about two minutes to allow all automounted filesystems to be unmounted. Manually unmount all mounted NUCFS filesystems.
The utilities should now work correctly.

Locale for NetWare auto-login window
The NetWare auto-login window is displayed when a user accesses a NetWare filesystem while not logged in to the NetWare server. The auto-login window is displayed in the locale that was in effect when the Gateway for NetWare was started, not in the user's locale.

If a user's locale is different from that used when the Gateway for NetWare was started, which is typically the system's default, do not enable auto-login for that user. Instead, use nwlogin(NWG) or the NetWare Access or Login managers to manually log in to the NetWare server before attempting to access any server files.



NFS


Unmounting an inoperative remote filesystem
Unmounting an inoperative remote filesystem during system shutdown can sometimes hang the local node. For example, if you import a filesystem and the exporting machine halts operation, your client machine tries to hold the mount to reattach the filesystem when the host machine restarts. If you halt your local system while an imported filesystem is down, your system can hang while trying to unmount the inoperative filesystem.

Invalid NFS client credential errors
An SCO NFS client mount request to a non-SCO NFS server might return the following error message from mountd(NADM): 
   Invalid client credentials error
This occurs because the version of NFS running on the server system has a secure RPC (Remote Procedure Call) mechanism. This mechanism denies other machines permission to mount filesystems if the client machine is not properly configured.

If you encounter this problem, you must remove any security options from the mountd command on the server; these options are specified in the NFS startup scripts. See the mountd(NADM) manual page for more information.

CAUTION: Changing the security options for mountd might not be a supported modification for some NFS implementations. Consult the server system's NFS documentation before making changes.


Printing with pcnfs
pcnfsd(NADM) supports printing requests from NFS clients on PC systems by returning a pathname to the server's spool directory that the client may use; NFS exports the server's spool directory. However, NFS does not export the spool directory automatically. You must manually add an entry to the server's /etc/exports file similar to the following: 
   /usr/spool/pcnfs    -"with appropriate options"    # comment
For further information, see the pcnfsd(NADM) manual page. However, note that pcnfsd does not check /etc/exports to see if the spool directory is exported when a PCNFSD_PR_INIT request is received. See the exportfs(NADM) manual page for more information on NFS exporting procedures.

Delayed errors from write() operations
Applications which write data to a file using NFS may not receive some errors relating to the filesystem or media at the time of the write() operation. Examples of such an error are out of space and permission denied. These errors are reported when a second write is done or when the file is closed. This occurs because NFS does write-behind of blocks from the client's buffer cache, as discussed in the manual page for biod(NADM). Therefore, although the write() system call has returned, the data may still be in transit.

Swapping to a file in an NFS mounted filesystem
If you configure your system to swap to a file in an NFS mounted filesystem, your system might hang if it begins to swap.

Mounting local filesystems using NFS
It is not advisable to use the loopback interface to NFS mount a local filesystem. Mounting local filesystems using NFS may cause deadlock conditions and so hang your system.

lp -- remote printing failing
The lp command in SCO OpenServer Release 5 does not copy a file to the spool area unless the -c option is specified. Remote printing from a PCNFS client will fail if the lp command specified in /etc/pcnfsd.conf for the remote printer does not include the -c option. If the printer is not defined in /etc/pcnfsd.conf, psnfsd uses a default invocation of lp that includes the -c option. Omitting the -c option can also cause remote printing to fail if it is implemented using the /usr/spool/lp/remote file and the file to be printed is not accessible to the print service on the remote machine.

NFS -- defect when flushing memory-mapped files
Applications that write to memory-mapped files over NFS will fail with the error Permission denied even if the user has write permission on the files. This happens because the buffer flushing daemon (bdflush) cannot flush the contents of a memory-mapped file to a remote filesystem mounted over NFS unless an entry giving root access to the local host exists in /etc/exports on the NFS server. For example, the following entry in /etc/exports would allow host1 and host2 to have root access to the exported filesystem /usr/mercury: 
   /usr/mercury    -root=host1:host2



NIS


Using rmuser with NIS
In the NIS environment, rmuser(ADM) does not access the file /etc/passwd.local, which holds the local system account information. The command rmuser only accesses the file /etc/passwd, which holds distributed account information in the NIS environment. Therefore, to change local-only accounts, you must edit the /etc/passwd.local file manually.

NIS passwords
On NIS clients, /bin/yppasswd is not currently linked to /bin/passwd. You must use yppasswd, not passwd, to change your password if your password is exported by an NIS server.

NIS security
NIS requires that the UNIX security system use /etc/passwd, rather than the Trusted Computing Base (TCB), as the master password file. Therefore, NIS can only be used with the ``Traditional'' or ``Low'' security default packages, because the ``High'' and ``Improved'' security default packages use the TCB as the master password file.

Inability to change password
When a non-root user on a copy-only server runs the passwd(C) command without arguments, the message Setting password for user might be followed by an account name other than the user's. The user is then unable to change their password, because that individual user cannot match the password for the other user account.

This problem is caused when the /etc/passwd file contains non-unique UID numbers. To correct this, assign a different UID to one of the accounts.


Simultaneous password changes to distributed accounts
The passwd map can be accidentally erased from the NIS master server if two non-root users change passwords on their distributed accounts within one second of each other, using the passwd command, and at least one of them is logged into the NIS master server. The passwd map is erased, and other distributed users are unable to log into the master server.

To fix the problem, enter the following commands as root to create and propagate a new passwd map:

cd /etc/yp
rm passwd.time
./ypmake passwd



PPP

Note the following issues about the PPP Manager and the PPP Connection Wizard:


Do not change default settings in the PPP Manager
Changing default settings in the PPP Manager causes problems with adding or modifying endpoints. If you do change the default settings, you can undo the changes by editing the file /.scoadmin.pref to remove the entry for pppGUI. For example, remove the lines: 
   pppGUI:
       DYNAMIC= ...
       MANUAL= ...
       INCOMING = ...
       DEDICATED = ...

PPP Manager does not add PPP driver to kernel
The PPP Manager does not add the PPP driver to the link kit. To work around this problem, configure your WAN device and protocol stack with the Network Configuration Manager, then use either the Network Configuration Manager or the PPP Manager to configure PPP. For more information, see ``Configuring network connections'' and ``Configuring PPP''.

PPP Connection Wizard does not add PPP driver to kernel
The PPP Connection Wizard does not add the PPP driver to the link kit. Follow the instructions in the previous paragraph to add the PPP driver. It is recommended that you do this step before you run the PPP Connection Wizard. If you add the PPP driver after you run the PPP Connection Wizard, you will see the following message: 
   ln: cannot create /etc/ppphosts: File already exists

You can safely ignore this message.



Network and remote printing



Internet Services notes and limitations


Strong Encryption enabled by default
In SCO OpenServer Release 5.0.7, the SCO Strong Encryption Supplement is licensed and enabled by default when SCO OpenServer is installed. This supplement turns on 128-bit encryption for Netscape Communicator.

Note that the SCO Strong Encryption Supplement is dependent on the presence of Netscape Communicator. If Netscape Communicator is not installed on your system, the SCO Strong Encryption Supplement is not licensed. Installing Netscape Communicator at a later time also automatically adds the SCO Strong Encryption Supplement license.

Do not remove the SCO Strong Encryption Supplement without also removing Netscape Communicator, otherwise the Netscape browser will cease to function.


Fully qualified domain name required for Squid Caching Proxy Server
The Squid Caching Proxy Server requires that your hostname resolves to a fully qualified domain name (FQDN). This requires that you do one of the following:

Squid also generates a core file on any fatal errors, even fatal configuration errors. This is not indicative of a bug in Squid. The code explicitly calls the abort() function on any fatal error, which generates a core file. Do not send Squid core files that were generated by fatal errors to SCO Support personnel unless explicitly asked to do so.



Optional Services notes and limitations


Layered products and TMPDIR environment variable
The default location for application temporary files is in the /tmp directory of the root filesystem. The TMPDIR environment variable is typically used in shell programming to override this setting. If TMPDIR is set to a location outside the root filesystem, it might interfere with the installation of layered products (such as the Virtual Disk Manager). The most common symptom of this problem is deletion of the mdevice(F) file.

When installing any layered products, ensure that the TMPDIR environment variable is not set or is set to a directory on the root filesystem partition.


NeTraverse Merge sound support
Sound support for NeTraverse Merge and many of the Windows applications that you can run with Merge is available through OSS 3.9.7g, a product provided by 4Front Technologies. To download OSS 3.9.7g for SCO OpenServer, see the 4Front Technologies web site:

www.4front-tech.com

Refer to the OSS documentation for instructions on installing and administering the product. OSS is not compatible for use with the Software Manager or the SCOadmin administration managers.

The latest release of the OSS drivers includes fixes to enhance performance and to reduce the impact of installing OSS on the size of the UNIX kernel. Note that support for the digital equalizer is now discontinued.



Development systems and environments


Intel extended instruction sets
The following extended instruction sets for Intel Pentium III and Pentium 4 chipsets are not supported by the native SCO OpenServer Development System in COFF mode:

These instruction sets are recognized by the SCO OpenServer Development System in ELF mode and by the GNU Compiler Collection (gcc) and the UDK.

Note that there is no kernel support for SSE and SSE2. We recommend that you do not use applications that take advantage of the features of the above technologies, as the risk of data corruption within these applications is high.


SCO OpenServer Development System manual pages
Many of the SCO OpenServer Development System manual pages in Sections S and M have been revised to reflect the updated library functions and header file modifications in SCO OpenServer Release 5.0.7. However, this work has not been completed and some of the affected manual pages are not yet current. The complete set of updated manual pages will be provided in the next SCO OpenServer Upgrade Pack.

Graphics, Web, and X11 Libraries Supplement (SLS OSS631B) and Execution Environment Supplement (SLS OSS646A)
The Optional Services CD-ROM contains two important supplements: the Graphics, Web, and X11 Libraries Supplement (SLS OSS631B), and the Execution Environment Supplement (SLS OSS646A). These two packages are supported for installation on SCO OpenServer Releases 5.0.5 and 5.0.6. (Although not supported by SCO, you can also install these packages on SCO OpenServer Release 5.0.4 at your own risk.) These supplements update systems to match the runtime execution environment of SCO OpenServer Release 5.0.7, and also update core components of the SCO OpenServer Development System, if it is installed.

If you develop any ELF applications on SCO OpenServer Release 5.0.7, then all consumers of that application on releases prior to SCO OpenServer Release 5.0.7 must install OSS646A (and OSS631B if the application uses any of the libraries provided in that package). Otherwise, the application will abort with a run-time link editor error message indicating that the symbol __requires_updated_system_library_set__ could not be found.

COFF binaries have no dependencies on the runtime execution environment. However, none of the libraries provided by OSS631B have COFF libraries, so using them requires you use ELF.


Configuring the kernel to enable SUDS
The SUDS (Software Update for Database Systems) driver that is needed by the SUDS library is configured into the kernel by default when you select the Database services initial system profile during installation (see ``Installation and upgrade checklist'').

If you want to use the SUDS library or use an application that has been linked with the library and you did not configure Database services at installation, follow this procedure. If you are unsure whether an application requires this, refer to the documentation provided with that application.

  1. Log in as root.

  2. Enable the SUDS driver by editing the /etc/conf/sdevice.d/suds file and changing the N in the second field to Y.

  3. Use mkdev aio or the Hardware/Kernel Manager to enable support for asynchronous I/O and relink the kernel. To allow users to lock down memory for use with this type of I/O, update the aiomemlock(F) file to specify the names of users that can use AIO and the maximum amounts of memory they can lock down. Many applications use setuid on their binaries so that you only need to specify a single user. Refer to the documentation provided with the application to see if this is the case for your application. Once you complete this process, reboot the system to take account of the new kernel and the memory locking privileges that have been assigned to particular users.

  4. The process and thread synchronization primitives supported by SUDS can require users of these features to be authorized by the system administrator. This is achieved by assigning all such users to a particular group(s) and giving that group the necessary privileges with the suds_ctrl(ADM) command. This process can be made automatic by including the suds_ctrl command in an appropriate rc2.d(ADM) script.

  5. Certain database applications require the administrator to also configure the database for use with SUDS or asynchronous I/O. This should be described by the documentation provided with the application.

  6. You must shut down the system and reboot for the changes to take effect.

Intel Optimizing C Compiler no longer supported
The Intel Optimizing C Compiler (icc) is no longer supported in SCO OpenServer, and the software has been removed from the product.

The SCO OpenServer Release 5.0.7 documentation might still mention this compiler in certain places, but this is incorrect; SCO OpenServer no longer includes icc.


Requirements for using the GNU Compiler Collection on SCO OpenServer
To use the GNU Compiler Collection (gcc) from the GNU Development Tools package, you must first install one of the following:

NOTE: None of the other tools provided in the GNU Development Tools package require that this software be installed.

Both the SCO OpenServer Development System and the SCO OpenServer Linker and Application Development Libraries package are located on the SCO OpenServer CD-ROM. See ``Installing and managing software components'' for information on how to install software using the Software Manager.

© 2003 Caldera International, Inc. All rights reserved.