RACORE 100VG TECH NOTES
Support Bulletin: 3070 Rev: G
CTL-ALT-DEL RESET: A warm boot (CTL-ALT-DEL) will not initialize 100VG adapters; a hardware reset is required. This is only an issue if you have previously initialized the adapter via diagnostics or by loading a driver; performing a warm boot under these circumstances may not properly initialize the adapter. To ensure proper operation, perform a power-off reboot or press the hardware reset button.
With Plug-n-Play (PNP) BIOS: If you experience a system hang after a CTL-ALT-DEL reset but not after a hard reset, you need to select the FLASH option of your BIOS instead of the PNP option. See Plug-n-Play BIOS below.
DIAGNOSTICS BEFORE LOADING DRIVER: If you run diagnostics, it is recommended that you reset the machine (as described above) before attempting to load a driver. Failure to do so may result in an error such as driver trying to initialize at an impossible interrupt, i.e., INT 99 or INT 33.
EMM386: Do not use v3.48 (11-1-93) or earlier, they don't have the ability to handle PCI calls; PCI machines didn't exist as of this date. Use the current version of EMM386.
PLUG-N-PLAY BIOS: If your machine employs a PLUG-N-PLAY BIOS (PNP), you must select the FLASH option instead of the PNP option, otherwise a warm boot (CTL-ALT-DEL) will hang the machine; however, a hard reset (power-off or reset button) will function properly.
I/O ADDRESS (PORT)
It is recommended that the I/O address (keyword is PORT) be defined in either the NET.CFG or PROTOCOL.INI file for all client drivers. If a driver fails to load properly (no adapter found at specified address) or other error (TRAP D with OS/2, etc.), try passing the adapters true PORT address to the appropriate configuration file (NET.CFG, PROTOCOL.INI).
Version: Use version 2.15 or higher. The current release is 2.16.
Diagnostics is only functional if no network adapter drivers have been loaded. If an adapter driver has first been loaded, you will get an error message instructing you to re-boot the computer.
It is not always obvious that a driver has been loaded. For example, if you boot to Windows For Workgroups which is configured for network use with a Racore adapter, you will load the Racore driver but you will not see any prompts indicating that a driver has been loaded. It is recommended to boot directly to DOS without loading anything else, i.e., do not use a CONFIG.SYS or AUTOEXEC.BAT file.
Do not power-down or re-boot the computer while diagnostics is running. This could permanently erase the data contained in the adapter's on-board EEPROM.
Reset to defaults: it is possible for some of the data in the EEPROM to be corrupted in such a way that diags value passes but it won't load a driver. If this happens, you'll need to re-program the flash as follows:
in diags, choose DEFAULTS, answer YES to the warning screen. It will then re-program the EEPROM with the right stuff except for the node address and MFG date. You can change the node address by selecting EDIT, move to node address field and hit ENTER; you'll be given the message "Node address is a fixed "; press "N" and you'll be allowed to change it.
Use v1.18 RVG386.LAN or higher: No longer uses keyword PCI=XXXX; instead, use: PORT (required) and SHARED (if adapter is set to shared memory mode). Fixes hub link error if adapter is set to 100MB/s .
Example: LOAD RVG386 PORT=4000 SHARED
Windows for Workgroups
If your PCI BIOS assigns the adapter an I/O address (PORT) other than default (FC00), it is necessary to modify the PROTOCOL.INI file to include the adapters port address as reported by diagnostics (VGDIAG).
This is done by adding the PORT=XXXX line to the Racore Adapter Driver section of the ROTOCOL.INI
file following the first line DriverName=R100VG$. Example:
With Other PCI Devices
The BIOS of some PCI machines will assign a common interrupt to more than one PCI device, for example, a Dell machine might assign INT 15 to both the Racore adapter and Adaptec PCI SCSI controller.
Because these devices support shared interrupts (level triggered), there should be no conflict in most cases. However, not all PCI machines handle this properly and it may be necessasry to assign the common interrupt to a phantom ISA device, if your systems BIOS gives you that flexibility.
There are several different implementations of the PCI specification among the various makers of motherboards. PCI was supposed to be plug-and-play, but in fact each machine presents a different combination of configuration parameters to be juggled. Below is a list of some of the quirks encountered.
In short, pay very close attention to CMOS setup. Make sure PCI slots are truly active and have resource assignments that make sense. Look for anything out of the ordinary like unusual SHADOW, WAIT STATE, etc.
type assignments. As always, carefully check resource assignments.
On some machines, the PCI slots are dead unless you activate them and assign them interrupts via CMOS Setup. Note that there may be more than one place in CMOS Setup to Enable/Disable these slots.
Diagnostics would fail with "Interrupt Generation Failure" unless CMOS was properly configured.
One machine required jumpers to be set in order to activate and assign interrupts to the PCI slots. The CMOS Setup erroneously reported that the PCI slots were activated and assigned INT 15 before the jumpers were set. Diagnostics would fail with "Interrupt Generation Failure" unless the jumpers were properly set.
The Shadow Ram CMOS configuration in one machine defaulted to "NOT PRESENT" for all address locations. This essentially turned out to be the same as "SHADOW ON", which interfered with the Racore adapter if set to SHARED MEMORY mode. Changing the configuration to "SHADOW OFF" for the memory map location used by the adapter fixed the problem.
Most PCI/ISA machines automatically prevent PCI resources from being used by ISA slots, even if the PCI slot is empty. For example, if you manually configure PCI slot 1 (empty) to use INT 10, your ISA slots will not be able to use INT 10. If you have an ISA adapter that quit working after having adjusted the PCI resources or installed a PCI adapter, this could be the cause.
The CMOS setup of one machine had a setting for READ WAIT STATE. When this was set to default (AUTO), the machine's performance was extremely slow (approximately 1/7 normal). Changing this option to 0 brought the performance back in line.
Full Insertion into Slot:
Failure to fully insert an EISA
device into the slot is very common because of the added depth
of the connector. A typical symptom of this occurs when running
the machines EISA configuration program: you will be prompted
to provide a configuration file for the adapter that doesnt
match the file on the distribution diskette. For example, it
may ask for !RAC1929.CFG, but the actual file is !RAC1020.CFG.
In this case, the EISA BIOS cannot properly read the adapters
ID because there is an improper connection at the bus.
This particular adapter requires that its interrupt be set through diagnostics rather than through the EISA configuration utility. It is therefore possible to have an interrupt conflict with other EISA devices, including multiple 100VG EISA adapters, unless INT assignments are chosen carefully.
You must identify the slot number on the command line that loads the driver. This is done in the PORT parameter. Example: LOAD RVG386 PORT = XC38 (X=slot number, Port = C38)
This is set in diagnostics and does not require any key word for either workstation or server use. IMPORTANT NOTE: when using multiple adapters set to Shared Memory mode, be sure that the Memory ranges of the adapters (as set by diagnostics) do not conflict with each other.