This chapter describes the following network interfaces:
· Ethernet
· Fiber Distributed Data Interface (FDDI)
· Token Ring
· LAN Emulation over ATM
· Classical IP over ATM
· HYPERchannel
· IP-over-DECnet
· proNET
· HP Wide Area Network (WAN) device drivers
· Pseudo devices
TCPware for OpenVMS supports all HP Ethernet, FDDI, Token Ring, LAN Emulation over Asynchronous Transfer Mode (ATM), and Classical IP over ATM (CLIP) interfaces, so that you can send IP datagrams over these types of LANs.
These interfaces generally include the use of the Address Resolution Protocol (ARP) and the Reverse Address Resolution Protocol (RARP), except CLIP, which does not support RARP.
For details on configuring LAN network devices and their addresses, see the Installation & Configuration Guide, Chapter 3, Configuring the TCP/IP Core Environment.
ARP dynamically maps between internet and physical addresses. TCPware provides an ARP table of mappings that it keeps in cache. If a mapping is not in this cache when a datagram is transmitted, TCPware queues the datagram and broadcasts an ARP request over the network. When a host responds with an internet-to-physical address mapping, TCPware adds it to its cache and transmits the queued datagrams.
TCPware’s implementation of ARP does not probe for imposters. However, it may generate the following OPCOM message, which includes the physical address of the system with a duplicate internet address:
Duplicate IP address: Sent from physical address...
In the rare case where the remote system does not support ARP, you can use the ADD ARP and REMOVE ARP commands in TCPware’s Network Control Utility (NETCU) to add or remove entries for the system. You can also find specific ARP entries using FIND ARP, and show the entire ARP table using the SHOW ARP command.
You can also set various ARP parameters using /ARP_x qualifiers with the SET INTERFACE command. The ARP qualifiers control when to check the age of an ARP entry, how long to keep it in cache, how long to wait for an unresolved entry to be removed from cache, and the maximum size of the cache. TCPware provides defaults for these parameters, so you do not normally need to use these commands. For example, it normally removes entries from its ARP table if it does not receive a packet for an entry within 10 minutes, or if the table is more than 512 entries long.
The Reverse Address Resolution Protocol (RARP) enables a diskless client to find its IP address at startup from a RARP server. The diskless client broadcasts a request that contains its physical hardware address. The server maps the hardware address to the IP address corresponding to the physical address of the client. The TCPware system only responds to RARP requests for permanent address entries in its ARP cache. (Note that BOOTP provides the same type of services.)
RARP support is enabled by default for all Ethernet, FDDI, and Token Ring interfaces. RARP support is disabled for LAN Emulation over Asynchronous Transfer Mode (ATM) and Classical IP over ATM (CLIP-n) lines. You can explicitly disable RARP support using the NETCU STARTUP command with the /FLAGS=NORARP qualifier.
TCPware can operate with trailer packets enabled or disabled. Trailer packets have some of the packet headers at the end of the packet rather than at the beginning. TCPware disables trailer packet support by default on Ethernet lines. However, some implementations use trailer packets, such as those running under UNIX. TCPware can receive and process trailer packets, but it will never transmit them.
To disable trailer packets on UNIX systems, use the ifconfig command with the -trailers option. To disable trailer packets on OpenVMS systems, use the NETCU START/IP /FLAGS=NOTRAILERS qualifier.
The START/IP command supports several qualifiers that you can use with Ethernet, FDDI, Token Ring, LAN Emulation over Asynchronous Transfer Mode (ATM), and Classical IP over ATM lines. See Table 2-11 and Table 2-13 in the NETCU Command Reference.
TCPware supports VMS Communications Interfaces (VCIs). VCI is a high speed interface to the LAN drivers. If you want to disable VCI support for some reason, use the /FLAGS=NOVCI qualifier to the NETCU START/IP command.
TCPware supports placing a limit on the number of receive packets it processes per second. If a limit is set on an interface and that limit is exceeded, TCPware may issue the following OPCOM message:
Warning - maximum receive packet rate exceeded on line line-id (rate packets/second).
This indicates that the interface specified by line-id received more packets than were allowed. This may indicate that either the receive packet rate limit is too low or that a flood of packets arrived at the system and a network problem exists that should be corrected. If the limit is too low, raise it using the NETCU SET INTERFACE /RECEIVE_LIMIT command. If a network problem exists, investigate it and correct it.
TCPware supports any HYPERchannel interface supported by Network System Corporation’s H269 device driver, including the UNIBUS, QBUS, MASSBUS, and BIBUS interfaces.
The HYPERchannel interface support includes the use of the Address Resolution Protocol (ARP). Use ARP to automatically map an internet address to a physical address.
When starting a HYPERchannel line, you must specify line-specific-information. This is the local HYPERchannel address in 32-bit address. The format is aa-bb-cc-dd, where aa, bb, cc, and dd are hexadecimal values representing each byte of the address:
· aa is the global network address domain (if none, specify 00)
· bb is the global network address network (if none, specify 00)
· cc is the physical unit
· dd is the logical unit
If you are using 16-bit addresses, specify the address as 00-00-cc-dd.
The H269 driver’s IO$_ATTACH function uses the cc-dd portion of the local HYPERchannel address as the path address. You should always specify the local HYPERchannel address parameter as the 32-bit HYPERchannel address.
TCPware needs to map the 32-bit internet addresses into 32-bit HYPERchannel addresses. These two address families are not related. That is, there is no mathematical formula that you can use to convert from one address family to the other. Instead, you must either configure an ARP server or pre-load the address resolution table with the mappings.
|
Note: You must properly configure the server with address mappings for all systems on the HYPERchannel network.
|
To pre-load the address resolution table, use NETCU ADD ARP commands (see the NETCU Command Reference). The ADD ARP command requires that you specify a 48-bit value (in hexadecimal format). The syntax is:
aa-bb-cc-dd-ee-ff
where the additional ee is the HYPERchannel trunks-to-try mask (typically FF) and ff is the HYPERchannel flags mask (typically 00). For ease of use, the ADD ARP command allows you to specify most addresses using aa-bb-cc-dd-00-00. It supplies the proper trunks-to-try mask of FF. See the HYPERchannel documentation for information on other values for the trunks-to-try and flags fields.
Note that if you load the address resolution table through
the ADD ARP command, specify
/PERMANENT. Otherwise, TCPware removes the added
entries after a short time.
The START/IP command supports several qualifiers that you can use with HYPERchannel lines. See Table 2-11 in the NETCU Command Reference.
TCPware provides support for DECnet interface implementations so that you can send IP datagrams over DECnet links. This lets you connect separate TCP/IP LANs over DECnet WAN links.
To configure a DECnet line:
1. Enter the DECnet line identification, internet address, and host name for the local internet address in the response to the applicable prompts in the CNFNET network configuration utility. Note that you need a different IP address for the DECnet line.
2. Enter the appropriate line-specific information.
The START/IP command line-specific-information parameter provides the required DECnet link information. Enter the line-specific-information in the following format:
node-name::”TASK=object-name”
|
Parameter |
Identifies the... |
|
node-name |
listener node when the master mode issues it. It identifies the master node when the listener mode issues it. |
|
object-name |
object used on the listener node. Both the master and listener nodes must specify the same object-name. |
An IP-over-DECnet line has a master node at one end and a listener node at the other end.
The below example shows selections from a sample IP-over-DECnet configuration in TCPware’s CNFNET.
Line Id Network Device
QNA-n for HP’s DELQA, DESQA, or DEQNA (UQDRIVER)
UNA-n for HP’s DELUA or DEUNA (XEDRIVER)
Enter the line identifications [LPB-0, ISA-0]: LPB-0, ISA-0, DECNET-0
What is the local host’s INTERNET ADDRESS for line ISA-0 [10.16.1.1]: RETURN
What is the NAME for line ISA-0 [BVA2]: BVA2
What is the SUBNET MASK for the line SVA-0 [255.255.0.0]: RETURN
Do you want to enable TRAILER packet support for line ISA-0 [NO]: RETURN
Do you want to enable RARP (Reverse ARP) support for line ISA-0 [YES]: RETURN
What is the local host’s INTERNET ADDRESS for line ISA-0 [10.16.1.2]: RETURN
What is the NAME for line DECNET-0 [BVA2]: BVA2
What is the SUBNET MASK for the line DECNET-0 [255.255.0.0]: RETURN
What is the DECnet link information for line DECNET-0: ONA1::”TASK=OZONE”
Is this the LISTENER end of the DECnet link for line DECNET-0 [NO]: RETURN
The network devices are configured as follows:
Line Address Name Options
LPB-0 127.0.0.1 LOOPBACK
SVA-0
10.16.1.1 BVA2 /MASK=255.255.0.0
/FLAGS=(NOTRAILERS)
DECNET-0
10.16.1.2 ONA1 /MASK=255.255.0.0
ONA1::”task=OZONE” /FLAGS=LISTENER
The START/IP command supports several qualifiers that you can use with DECnet lines. See Table 2-11 in the NETCU Command Reference.
The proNET-10 and proNET-80 token ring controllers produced by Proteon, Inc. form a link between the hardware devices, the token ring, and TCPware.
When you configure proNET lines, the D component of the internet address (using the standard A.B.C.D convention) must match the node address of the proNET controller.
The START/IP command supports a qualifier that you can use with proNET lines. See Table 2-11 in the NETCU Command Reference.
The HP WAN Device Drivers are synchronous interfaces that form a link between the hardware devices and TCPware. TCPware for OpenVMS supports the DSV11, DSB32, and DST32 HP WAN interfaces.
The NETCU START/IP command’s line-specific-information parameter provides the required DECnet link information. The line-specific-information is a quoted string of the line configuration options shown in Chapter 4.
For details on these parameters, such as the possible values for the line speed, CRC, and so on, see the VAX Wide Area Network Device Driver’s Programmer’s Guide.
An example of line-specific-information is
“PROTOCOL DDCMP POINT CLOCK INTERNAL LINE SPEED 64000”
When specifying line-specific-information for HP WAN device drivers lines, be aware of the following:
· You must enclose line-specific-information in quotes for these lines.
· You can use keyword abbreviations.
The START/IP command supports several qualifiers you can use with HP WAN device driver lines. See Table 2-11 in the NETCU Command Reference.
|
Parameter |
Takes... |
Description |
|
PROTOCOL |
DDCMP
POINT |
Line protocol used |
|
DUPLEX |
HALF |
Line operation used |
|
CLOCK |
INTERNAL |
Line clocking used |
|
CRC |
type |
Type of CRC used (not recommended) |
|
LINE SPEED |
baud |
Line speed (only useful with CLOCK INTERNAL) |
|
RECEIVE BUFFERS |
number |
Number of receive buffers |
|
RETRANSMIT TIMER |
time |
Retransmission tine (for PROTOCOL, DDCMP POINT only) |
Pseudo devices are a way to configure a physical device to have multiple Internet addresses. Pseudo devices are typically used when a system is connected to a network that needs extra network numbers assigned to it. Pseudo devices can also be used in place of secondary addresses (for example, when a system has multiple addresses on the same network).
When starting a pseudo device, you specify the local Internet address, network mask, and the physical device to which the pseudo device is connected.
CNFNET allows the configuration of pseudo devices.
The TCPware line-id for a pseudo device is PSD-n, and n is 0 to 255.
To configure one or more pseudo devices via CNFNET, include the line-id or line-ids for the pseudo devices when prompted to enter the line identifications for all the network devices. Be sure to enter the pseudo device line-id after the physical device line-id. CNFNET prompts you for the standard information (Internet address, host name, and subnet mask) and for the physical device line-id for the pseudo device.
You can also start pseudo devices by using the NETCU START/IP command. See START/IP in Chapter 2 of the NETCU Command Reference.
Pseudo devices are interchangeable and usable just like physical devices.
However, there are a few special characteristics that are important to point out:
· Multicast joins/leaves are redirected to the physical device.
· NETCU SHOW NETWORK shows no transmit/receive counts for pseudo devices. The physical device reflects the transmit/receive activity.
· Pseudo devices are removed automatically whenever the physical device is removed (such as by a NETCU STOP/IP line-id command.
· Starting a pseudo device on a pseudo device (by specifying a pseudo device line-id as the Real-Line-ID for a NETCU START/IP command) is allowed; however, the underlying physical device is used.
· Packet filtering is not available for pseudo devices as these devices never receive any packets (the physical device does). Therefore, you must do all packet filtering on the physical device and must take this into consideration when creating the packet filter list. Attempting to issue a NETCU SET FILTER or NETCU SHOW FILTER command on a pseudo device returns an error message.
·
Once a pseudo device is started, the command NETCU SHOW INTERFACE line-id
can be utilized to display the physical device information. For example:
$ NETCU SHOW INTERFACE PSD-1
For Network Line PSD-1:
On Physical Line EWA-0:
No receive packet rate limit has been set.
The maximum receive packet rate was 0 packets/second.
The ARP entry limit is 512 entries.
The ARP age check interval is 30 seconds.
The ARP entry age limit is 600 seconds.
The ARP entry wait limit is 20 seconds.
· The line-id value for pseudo devices is 00nn0042 (hex), where n is the unit number (PSD-n).
· For proper operation of pseudo devices (and TCPware is general), the LPB-0 (loopback) device must exist. The LPB-0 device is technically not optional.
· Pseudo devices cannot be started on unnumbered interfaces.
TCPware continues to support secondary addresses (NETCU ADD SECONDARY) and interface routes (NETCU ADD ROUTE) in addition to the new pseudo devices. Some recommendations as to which method to use and the conditions under which to use them are described next.
· If a TCPware system is connected to a network via a single interface that has multiple networks numbers assigned to it:
o Use a pseudo device for each network number on which the TCPware system has an Internet address (other than the one that is used to start the physical device).
o Use an interface route for each network number on which the TCPware system does not have an Internet address. For an interface route, specify the line-id of the physical device in place of the gateway address parameter.
· If a TCPware system has multiple addresses on a single network number:
o Use either pseudo devices or secondary addresses for the additional addresses. Using a pseudo device has some advantages and is recommended (especially if a DNS server is running on the system).
· To use the cluster alias failover support:
o The secondary address feature must be used.
|
Note: If your site is using secondary addresses you might want to consider whether switching to pseudo devices makes sense for these addresses.
|