OpenROUTE 2.1b
Release Notes
(Part Number 42-555011-00 Rev A)

1.0 New Features and Enhancements

The following table identifies the new functionality in OpenROUTE 2.1. This release is available on all OpenROUTE Networks internetworking platforms including the GlobeTrotter, Corporate Network Exchange (CNX), Departmental Network Exchange (DNX), and the Remote Branch Exchange (RBX) series.

Table 1 OpenROUTE 2.1 Functionality
Feature	Description
BOSS SNMP Agent	Provides integrated SNMP management of the DNX 300m and DNX 350 routers, the Series 90 network concentrator, and the BOSS 3-slot hub.
Data Compression (STAC®)	Lets the router software pack more user data into the packets moving through router interfaces.
Frame Relay over ISDN	Provides a dial-on-demand interface to the Frame Relay network to be used as a WAN reroute alternative in case of a failure of the primary Frame Relay connection. The dial-on-demand circuit provides the same functionality as a dedicated Frame Relay interface.
IPX Filters	There are two types of IPX filters: Global filters-existing filters which apply to all IPX interfaces on the router. Interface filters-new filters which apply only to the IPX interfaces that you specify.
IPX Keepalive and Serialization Filtering	Prevents keepalive and serialization packets from continually activating a dial-on-demand link or to minimize traffic over a dial-on-demand link.
LAN Network Manager (LNM) over DLSw	Allows you to manage your Token Ring networks interconnected by source route bridges. This allows you to monitor the operation of rings, bridges and individual ring stations over Data Link Switching (DLSw).
OSI Ping	Supports RFC 1139. Ping-1139 causes the router to send an echo request to a destination and watch for a reply. Ping-1139 specifies this as an OSI function and not as a DECnet function. It supports short-term and long term-term echoes.
DLSw SDLC Secondary	Supports local secondary station support in the router, as opposed to the currently supported remote secondary station, which is DLSw SDLC primary from the router's perspective.
User Datagram Protocol (UDP) Broadcast	Allows the router to forward UDP broadcast frames to specific IP addresses.
X.25 Transport Protocol (XTP)	Works with X.25 devices that have locations at multiple remote sites. In such environments, XTP can eliminate the use of X.25 packet-switched networks for communication with servers at one or more centralized locations. To enable this, use routers at the server and remote locations to encapsulate the data and deliver the X.25 packets between the clients and the server via TCP/IP.

2.0 Corrected Problems

The following problems have been corrected since the OpenROUTE 2.1a software release.

2.1 General

When toggling aliases in the OSI protocol in talk 5, the router would crash with the message FVS: Bad buffer addr.
The router crashed when an attempt was made to ping over an Ethernet interface with packets larger than 1464 bytes.
When running Telnet from the console of a router, the router intermittently crashed with a bus access error. The command line wasn't being parsed correctly.
Local conversion was not working properly starting with release 16.1a.
If a dial circuit received a Disconnect Indication and the idle timer had not expired, a netdwn was being reported to the routing protocols which caused all routing tables to be deleted.
The router sent TEST frames with source and destination SAPs of zero for the router interfaces when Source Routing was enabled. This caused a performance problem for large configurations. The patch variable no-xidexp-frames has been added which circumvents the sending of XID explorer frames. If this variable is set to 1, the problem is avoided.
The router crashed with BUGHLT message FVS: Attempting to release memory twice. The dlsf_dlc_xid function had called FVS according to the stack trace.
Various modem dial-up problems were discovered. The static representation of modem signal size was fixed. The order of modem commands (Ring-Wait to Wait-Ring) was also changed. For service ports, wait for either a Carrier Detect or Ring Indicator on external modems.
Information wasn't being exchanged between our router and Cisco. This was an intermittent problem. A fix was provided to update the Advertising Length for the Link State header of the Database description packet and the Link State Acknowledgment packet.
Various XID-3 frames were corrupted by DLS/LLCIM, which caused endstations to fail to establish connections.
IA router crashed immediately after Telneting to a remote router or host, with a BUGHLT message of either Loop in IORBs or MOS: no free kernel block for timer or signal.
The talk 5 command was not displaying the line speed. Instead, it showed a fixed value of 1 Mbps. This functionality was removed in a prior release. The feature has been re-inserted with the addition that if the line speed is configured to anything other than 0 (zero), the configured value is displayed and used by protocols which care about such things. If the line speed is configured as 0 (zero), and clocking is set to external, the router attempts to calculate the line speed.
Certain LLC endstations (namely, Siemens cluster controllers) issue FRMR frames when DLSw routers send RNR frames to them. Our LLC protocol implementation is not at fault here, but we have added a patch variable, llc-compatibility, which when the least significant bit is set, alters the way the Nr field is updated when sending RNR frames. This patch works around the bug in Siemens' LLC implementation. The other bits in the llc-compatibility flags are reserved for future use, so as of this release, only the values 0 (zero) and 1 are valid.
When logging in to a Windows NT server from a Windows 95 client running NetBIOS over Frame Relay, the router at the client side would crash with trailing guard word corruption.
When running a file transfer over a PPP dial-up link, the echo packets were not making it out the interface, and this resulted in the dropping of the line. The transmission of echo packets is now configurable through a patch variable. The default setting for ppp-echo is 1 which enabled the echo feature. The echo feature may be disabled by setting ppp-echo to 0 (zero).
When the Internal IP Address was the same as an interface IP address, and the interface was disabled, then the internal address was also disabled. The internal IP address is no longer disabled in this case.
The router crashed when the network range was not accessible. This was caused when a static route backed up a network range route and the network range went away.
TFTP time-outs were occurring because of erroneous indications of low input buffers. The solution was to reset the Input Buffer Low flag in the MOS Receive Buffer Handler logic.
An "Illegal instruction" fault was seen following a fault condition. This could cause data for that fault to be useless.
If PPP compression and CCP were enabled, and the packet that was being compressed fit the buffer exactly, the packet was being discarded.
An enhancement was included which causes DTR and RTS to drop for a configurable duration when a link is lost. This affects the CNX/GTAM platforms only. The number of ticks is set by setting the qsl_tran_time patch variable to a non-zero value.
An enhancement was made to OpenROUTE that implements configuration, monitoring, and ELS reporting modifications to provide the missing interface utilization measurement functionality. All data-link types can now dynamically calculate line speed.
The i_cmd field of the IORB was set to an invalid value because Proteon Serial Line (PSL) failed to initialize it with the d_cmd field of the DCT. This caused the ISDN driver to bughalt.
A problem with negotiating a start with DLS occurs when the no-test-frames patch is enabled.
As part of the upgrade to RFC 1795, the ability to modify the receive buffer sizes transitioned from a global ability to a per-interface ability. To this end, a global receive buffer size was removed. The global receive buffer size feature was re-inserted.
A patch variable (dls-glob-tcprcv) has been added to restore the global TCP receive buffer capability. To use the patch, the recommended sequence is: 1) set the patch variable to a non-zero value, 2) do talk 6 set tcp, 3) restart. Any pre-R16 value stored for the global receive buffer size is recovered when this patch variable is activated.
CNX routers infrequently crashed with a data access error while executing in the tm_poll or tm_pollbr function.
When performing transfers (IP or IPX) over a PPP WAN link, the router would crash with the message QTFI: Buffer freed twice in a
row !!. This only happened if compression (CCP) was enabled on the WAN link.
The router was always advertising static routes even when it was configured not to send static routes. The rip_sta_sup patch variable was removed, so that the advertising of static routes would be under the control of the user.
Interface Utilization Measurement commands should be present only in software loads for OpenROUTE Networks routers. They are not present in resellers' loads.
The router sometimes crashed with a watchdog timer expiration when it was executing the bgp_withdraw function.

2.2 ARP

ARP cache entries stayed in the "pending" state forever.
AppleTalk ARP was erroneously adding a MAC address to the cache even though the MAC address was for a host that had sent a Probe message. The Probe message is used by a host that is trying to use a specific node address. In this case, the cache was updated, so all traffic from the existing AppleTalk session started going to the prober. The solution is to handle the Probe message differently. When the Probe message is being handled by ARP, no ARP entry is created or updated.
ARP traffic was not accepted from some vendors' boxes when they didn't include the hardware address type. We have relaxed the hardware type check that handles the receipt of an ARP message.
When you add a manual ARP entry for an IP address, where the entry is on a Token Ring, and source-routing end station behavior is enabled on that Token Ring, the packets do not go out. An IP cache entry gets built, but the reference count does not go beyond 1. No ELS message is produced. Permanent ARP entries are only for use in testing environments. They should never be used in a production network.

2.3 BGP

When adding a subnetted network to the IP routing table, the cost was always set to 1. The customer wanted to be able to set the cost to a higher value. The customer was using BGP4. This new feature is now available when the ip-def-agg-cost patch variable is set to something greater than 0 (zero).
BGP4 didn't allow a TCP connection to be established to a neighbor router unless the two routers shared a direct interface. The customer wanted the ability to configure BGP4 to work as Cisco routers do. They can support multi-hopping. This feature is now available when the multihop-bgp patch variable is set to 1.
There were BGP4 interoperability problems with Bay Networks routers. The BGP code no longer uses the extended length packet format for packets that are less than 255 octets in length.
When running BGP, a change in the route table is not reflected in the routing cache.
A BGP natural net was not being accepted from a BGP neighbor. BGP routes are now allowed to overwrite T_SBN route under all conditions.

2.4 BRS

The router was discarding IP packets, but there was no evidence of this when looking at the interface statistics. The router now increments the "Output Discards" statistic when a packet is discarded by BRS. This statistic may be viewed by using the error command from talk 5.
The router crashed in the srread function when the WatchDog timer expired. The router had an extremely large Frame Relay configuration that took a long time to read. The watchdog timer is now reset periodically when reading the BRS SRAM records.

2.5 DLSw

When running DLSw OpenROUTE 2.0 or above between two routers when one of the routers is a DNX, it is sometimes impossible to establish and maintain a TCP connection between the two routers. Once contact is established, both routers terminate their read TCP connection. DLSw had a problem with byte swapping of the internal IP address. The code was not taking the byte swapping into account. This resulted in the fact that both ends of a TCP connection might think that they had the higher IP address, so they both terminate their read TCP connection.
The router crashed with an Access Fault in dls_con_pr+88 just after the talk 5 DLSw list dls session all command was issued. The code was accessing an SCB after a printf caused a stack swap. This resulted in changes to the scb and cookie which caused the structures to be unreliable.
There was a DLSw problem with frame size compatibility between the router and 3270 PC equipment.
DLSw routers on very busy rings consumed large amounts of memory when there was a lot of explorer traffic on the ring. This was due to DLSw's allocating and broadcasting many CANUREACH_ex frames. DLSw was behaving exactly as it should, but its heavy memory consumption could have interfered with the operation of other subsystems in the router. One can tell if this problem is occurring by using the list dls memory command and checking the Global transmit pool bytes used field. If it is grossly over the granted field, then enabling the patch described below helps. The patch variable named dls-cur-limit was added. When it is set to 1, the DLSw stops sending CANUREACH frames when its global transmit pool is depleted. It resumes sending frames when memory from other outstanding CANUREACH frames (and other DLSw traffic) is returned to the pool.
DLSw no longer crashes the router when an attempt is made to remove an entry that is not in the cache.
A crash occurred in the DLSw console when a DLSw session or transport went away while displaying sessions via any of the list dlsw commands.
DLSw was not handling session startup correctly between a PC and a server. XID-3 packets seemed to be getting discarded during the negotiations. XIDR handling was added to the LLC retry logic.
DLSw sometimes failed to establish a TCP connection to its neighbor after the neighbor router was restarted or reloaded. The router that was not restarted was trying to perform the CAPEX exchange, but the two routers never managed to see one another because the neighbor router that was not restarted never actually sent the CAPEX requests or responses even though the ELS messages indicated that they were being sent.
On a 2-port Ethernet RBX 200 model router (24E), when both transparent bridging and DLSw were configured on both Ethernet segments, DLSw interfered with the ability of local bridged SNA sessions to establish a connection. When the remote DLSw router was powered off or its WAN link severed, the local bridge session could not be established. The explorer flows were modified so that the indication of whether DLSw handles the SABME is dependent also on whether a MAC/IP address association is found.
DLSw SDLC secondary interfaces which support multiple link stations occasionally start sending RNR frames for all link stations. When an attached host initiated a session with an XID frame instead of a TEST frame, it caused DLSw's SDLC interface module to drop the buffer which contained the XID frame, which is owned by SDLC. Eventually, all of SDLC's input buffers are lost in this fashion, so it has to send RNR frames to tell the host that it cannot receive any further data. Since the buffers are lost, the RNR condition never exits. The buffer leak problem was fixed.
If DLSw was in local busy state, it did not respond to I-frames. There were some cases when DLSw needed to respond to a poll bit. This problem seemed to occur with a Microsoft Win 32 implementation. The result was that the workstation sent a Frame Reject (FRMR) before the router sent an ACK.
When using the DLSw list priority command in either talk 5 or talk 6, garbled strings for the SAP Session Types were displayed, and the router intermittently crashed with an Access fault. A similar problem was also visible in the DLSw set priority command.

2.6 DVMRP

The DVMRP implementation did not support the Pruning feature. Pruning was added to the DVMRP code since this is a requirement to be on the MBONE. This brings the code up to DVMRP, version 3 level, as specified by internet-drafts/draft-ietf-idmr-dvmrp-v3-03.txt.

2.7 GlobeTrotter 60/62

Attempting to use Host Services on a GlobeTrotter 60/62 router that has no multicast IP code caused the router to crash.
The GlobeTrotter 60/62 Async driver failed to work properly when CHAT was being used. When the Async driver switches from HDLC mode to UART mode, the receive buffer descriptors were left in a bad state.
The GlobeTrotter 62 crashed with a panic when reloaded/restarted if it was configured with Host Services and Router Discovery enabled. On other routers, the destination address of the ICMP Router Solicitation frame was incorrectly set to 224.0.0.1.

2.8 HSSI

The HSSI interface card was receiving frames which were discarded by IP/ICMP/TCP due to checksum errors. This occurred most often when running PPP with address and control field compression enabled. The bad checksums occurred on almost all packets with an odd number of bytes.

2.9 IPX

IPX packets were not forwarded across X.25 after a period of time had elapsed. The problem occurred when an SVC required ARP entries that had sidecard information. The use of SVCs for IPX traffic now causes permanent ARP entries to be added to the ARP cache.
IPX SAP entries were being aged out and deleted and then immediately added again. This was happening because of the large number of SAPs and a one minute SAP update interval. A patch variable was added (ipx-age-time) that allows an additional amount of time before SAP entries are aged out or deleted. The default value is 0 (zero) minutes.
IPX watchdog request packets and serialization packets generated by 4.1 servers were not being handled within IPX spoofing/serialization packet filtering. The software now handles watchdog request packets and serialization packets with multiple packets types.
IPX SAP broadcasts were not being received by the router when the interface encapsulation frame type was configured for TOKEN-RING_SNAP. Existing SAP entries were being aged out.
The router was crashing when trying to forward IPX packets out an Ethernet interface configured for 802.2 encapsulation. The IPX packets were 1497 bytes and were generated by an NT client.
When receiving an IPX echo packet on CNX 500/CNX 600, the router crashed with a BUGHLT message of Unexpected trap #016.

2.10 ISDN

When displaying the ISDN accounting statistics under talk 5, the router crashed as the result of a bus error. The problem occurred only when the ISDN address exceeded 10 digits.
ISDN subaddressing wasn't working properly. Some calls were being accepted when they shouldn't have been. Also, when using the INS64 switch, a setup request with an invalid called party subaddress was being released. It should have been ignored.
Warm/cold start SNMP traps sometimes do not reach the other side of an ISDN connection. The ISDN link sometimes takes up to ten seconds to come up, and the SNMP code was not waiting long enough, so it never sent the trap. The code now waits up to thirty seconds before giving up.
The ISDN console display circuits option did not display information about active inbound circuits.
ISDN calls indicating an unsupported calling plan were accepted regardless of whether the call was destined for the router. When an ISDN call is received from a Net3 switch, the calling plan provided in the Called Party Information Element was checked for supported values. This check is outdated and has been removed.
After configuring many dial circuits and ISDN addresses on an RBX, the router will bughalt upon restart with the message Watchdog timer expired. The software was modified to allow the maximum number of 64 dial circuits to be configured without crashing.

2.11 Macintosh

MacIntosh systems could not establish sessions with AS400 using two 2210s connected by the Frame Relay. The MacIntosh systems were on one Ethernet LAN at 2210 #1 (remote) and AS400 was on another Ethernet on 2210 #2 (local). MacRumba software was used and the AppleTalk traffic was being routed. The initial screen appeared on the MacIntosh and then everything hung. Very rarely could you move to the next screen and hang there. Mac-to-Mac worked fine. Locally-connected MacIntosh systems talked to AS400 without any problem. The Ethernet handler was changed to not do even-octet padding for AppleTalk.

2.12 NetBIOS

IBM-RT NetBIOS clients were unable to establish connections to a NetBIOS server over a DLSw connection. ELS messages indicated that the NetBIOS UI pool had been depleted. The enhancement made to support the IBM-RT (80D5) Ethernet type was not applied to all of the necessary code. NetBIOS and LLC type I messages were not fully accounted for. DLSw and LLC modules now adjust the LLC header for IBM-RT Ethernet type and pass the correct message length to the higher layers.
When using NetBIOS name filters on an Ethernet port with a default of exclude, Supervisory frames (RRs) were being filtered. This was causing problems establishing LLC type 2 sessions. LLC Supervisory frames are no longer processed by the NetBIOS name and byte filters since they contain no data.
When using NetBIOS Name and Byte filters, the determination of whether to forward or drop NetBIOS packets was not being handled correctly when multiple filter lists were applied to a single input or output filter. The results of matching the first filter list were not being taken into account.
When running NetBIOS over DLSw, the Best path remote IP address was never aging out. If a NetBIOS client was moved to another DLSw partner, all traffic for that client was being sent to the original DLSw partner.
When attempting to pass NetBIOS traffic over a DLS connection on a DNX 350, the router intermittently crashed with the message TMFO+61; Data access error.
The router intermittently crashed after NetBIOS filters were configured.
When NetBIOS name caching was enabled, Spanning Tree Explorer frames were not always converted properly when a cache entry existed for the destination NetBIOS name.

2.13 OSPF

OSPF broadcast/multicast was not working with 802.3 framing. Ethernet II framing worked fine. The solution involved changing hardware Type 6 to Type 1 when an ARP packet is received on an Ethernet version 2 interface.
Parallel P2MP/PPP OSPF equal-cost topologies could cause routing to fail.

2.14 RIP

If the rip-extensions patch variable was enabled, and RIP was configured to send subnets, and the route was an aggregated subnet, we were incorrectly sending the route. Sending aggregates when we are also sending subnets is incorrect. If the rip-extensions flag is turned on and the s_sub (send subnets) indicator is on, and the route is an aggregated subnet, then we no longer send this route along.
RIP broadcasts were coming from the IP Internal Address instead of the interface IP address.

2.15 SDLC

Remote SDLC stations were unable to establish/re-establish connection with the host. The RR/RNR transition handling was disabled in DLSw; leave it to the relevant SDLC busy/unbusy logic.
SDLC secondary didn't handle a received SNRM when in DISC_PENDING state. It responded with a DISC/RD and remained in DISC_PENDING. This is usually acceptable, but some older hosts ignored our response and retransmitted the SNRM. This went on forever, and the only way out was to disable and then enable the end station with the disable/enable station sdlc monitor commands, or by disabling/testing the interface, or restarting the router. SDLC secondary now still responds with the DISC/RD, but it enters the LINK_CLOSED state. The SDLC state machine then responds properly to either a SNRM or DISC frame.
An SDLC POS device was failing intermittently. The POS device was sending a Frame Relay Reject (FRMR) in response to a poll with an incorrect N(R). When checking the transmit queue for data to transmit, we no longer release I frames for sending/resending if a poll time-out has occurred. Instead, an RR is transmitted.
When configuring the maximum packet size for a station on an SDLC link, the router would bughalt with a stack overflow. This would only happen if the actual packet size was not specified on the initial command line when the user entered the set command. The user is now prompted for the packet-size, if it is not entered on the set command line.
Performance on SDLC links was poor. A change was made that causes the poll delay timer to be used only if no I frames have been transmitted. If I frames are transmitted, the poll gets sent immediately. This allows transmission of around 6.54 windows worth of data per second, for an effective throughput of about 7000 bytes per second (56000 bps).
SDLC devices intermittently failed to work properly when the SDLC link poll time-out value was set to 0.5 seconds. 0.5 seconds was the default value, and this did not agree with Cisco's default value of 2 seconds. The SDLC link poll time-out value now has a default setting of 2 seconds. If a customer already has SDLC configured, the set link poll timeout 2 command may be used to force the timer to the more acceptable value.

2.16 SNA

SNA sessions were being dropped due to segment errors. After transferring about four billion bytes of data, the router's TCP code lost track of the proper TCP sequence number.
It appeared that we could not use BAN with SNA and a backup line. When we used one BAN link, both SDLC and SNA worked, but when a second BAN line was defined for type 2, the SNA PU traffic didn't flow. The problem is that the second 5394 on a multidrop line takes the first line down. The BAN port was modified to use the original MAC address instead of a target MAC address.
SNA SABME was not getting sent using SRB through the router when the router had DLSw enabled.

2.17 SNMP

SNMP traps were generated with the internal IP address as the source. SNMP traps now use the IP address of the interface generating the trap.
The SNMP query for snmpFddiMACSMTAddress was returning a MAC address value of 00-00-00-00-93-00-26-6F instead of 00-00-93-00-64-F6. MAC addresses were being shared with Station ID object which is an 8 octet value and returned as such without the last two bytes being cut. MAC addresses are now being correctly converted and returned as 6 octet values.
If the destination address of a received SNMP message was the router's internal IP address, the response source address was set to the address of the interface it was received on. This caused a problem with some SNMP applications. The code now sets the address of the inbound interface within the IORB to NULL if the destination address is the router's internal IP address. This is checked in the UDP response processing and, if NULL, the source address is set to the router's internal address.

2.18 TCP

The router crashed occasionally with an "out of memory" condition. ELS messages showed unexplained TCP resets (TCP.40) with pending states for TCP partners. In addition, DLSw ELS events indicated an unrecognized message type.

2.19 Token Ring

The Token Ring interface failed maintenance and self-test because there was a lack of input buffers. DLSw was hanging onto the buffers because it was running out of heap memory. The router had to be restarted to get the interface back. The solution is to have DLSw discard the message and allow it to be retransmitted later.
When two routers were on the same Token Ring and were configured for DLSw using the same virtual ring number, the NetBIOS Name Query frames put onto the ring by one router were picked up by the other router, and the names were cached as being local, before the frame was discarded. Since both the routers cached the name as local, DLSw would not support further connections to that name. Logic was added to prevent the update of the NetBIOS cache if the DLSw segment is already in RI.
When pinging from Host Services over a Frame Relay WAN with Source Routing Bridge configured, ping failed if the remote Token Ring interface went down or was disabled.
LNM was always reporting the Token Ring speed to be 16 Mbps (megabits per second) even though the speed was set to 4 Mbps.
MAC frames sometimes left the Token Ring with the monitor bit set since the router had not cleared the bit. When the frame arrived at the destination Token Ring, the second monitor packet on the ring caused the monitor station to purge the ring. Purging occurred frequently.

2.20 WAN Reroute

WAN Reroute was not learning new routes. Two routers were running DLSw over a PPP link with WAN Reroute configured. When the primary connection was broken, the alternate link came up fine, but when the primary link was restored, the IP routes were not relearned.

2.21 X.25

IP connectivity over X.25 was lost intermittently. Pinning between two routers did not work. The problem corrected itself eventually.
Router crashed with message Bughlt: Unhandled trap #0x01 near address 0x103008EC while running X.25 API forwarder tests.
When using an SNMP MIB browser to look at the physical interfaces, the X.25 interface reported its MAC address as garbage. It should have been a null string. The MAC address field in the net structure is now set to 0 (zero) instead of 8.
Updating the T1 and T2 parameters on an X.25 interface caused these parameters on other X.25 interfaces to revert to default values.

2.22 XTP

XTP did not work with X.25 devices that did not provide the calling DTE address in the call request packet. This is actually an enhancement to the code.
XTP intermittently held onto X.25 input buffers if a PVC went down and came back up. Eventually, the X.25 interface locked up because it had no more input buffers. Also, the XTP talk 5 and talk 6 menus did not conform to OpenROUTE Networks' standards. XTP now sends any queued X.25 input buffers to the peer router when the PVC comes up (becomes REMOTE ACTIVE). Also, the XTP talk 5 and talk 6 menus now conform to OpenROUTE Networks standards.
The XTP forwarder didn't work with certain ATM devices that are not able to recognize their own DTE address in a call request packet. The devices are only capable of recognizing the last few digits of the DTE address. The X.25 code was modified to support a new feature that allows the user to cause X.25 to truncate the called DTE address in an outbound call request packet. The national enable truncate-called-addresses command can be used to enable the feature. Also, the national set truncate-called-addr-size command can be used to set the size of the truncated DTE address. The truncation of digits starts from left. This feature is only available with XTP forwarder call request packets.
Bandwidth Reservation didn't support filtering of XTP traffic.
The XTP forwarder's handling of the X.25 Q-bit was not working. The Q-bit was always equal to zero when a packet was sent.
The customer did not have the option of configuring an SVC inactivity time-out for the XTP protocol. What this meant is that a locked-up or inactive X.25 device could hang on to an SVC until the router was rebooted. The SVC inactivity timer is now configurable from within the talk 6 X.25 configuration menu when the add prot xtp or change prot xtp commands are used. The default value for XTP remains at 0 (zero) seconds which indicates that the timer is disabled. If a customer encounters problems with getting X.25 to free up an XTP controlled X.25 device, this timer value can be set to something like 30 seconds or 60 seconds. The timer value should be set to a number that is much greater than the expected transaction response time. This prevents an SVC from coming down before the transaction completes.

3.0 Known Deficiencies, Limitations, and/or Clarifications

This section explains known deficiencies in OpenROUTE 2.1b and indicates limitations with the software.

3.1 For CNX 600 Only

There is a known anomaly with the CNX 600 that prevents the user from accessing reading or writing diskettes while the router is experiencing heavy system activity. The diskette I/O operations fail because the CPU is unable to extract data from the 16 byte DMAQ buffer in a timely fashion. The loss of data results in unreliable disk operations. The workaround is to access diskettes only during periods of low system activity. Another alternative is to transfer configurations and loads via the flashcard that has no problems.
There is a known problem where the CNX 600 floppy drive is supposed to support 720 KB, 1.44 MB, and 2.88 MB diskettes, but the current Boot PROM and router software can only accommodate 1.44 MB diskettes. The workaround is to use the flashcard.

3.2 For CNX 500 Only

Version 3.0 Boot PROMs do not properly reset the Dual Serial card after a failed boot attempt (long ACK time-out, line not coming up, and so on). Successive attempts then fail with the message unable to load microcode.
When the Boot PROMs are started, the serial card is reset and everything runs fine. If you restart the Boot PROMs using Ctrl C or the Reset button, the serial card resets and everything works fine. The reload command from the gateway starts up the Boot PROMs. The first load attempt succeeds but successive attempts without a Ctrl C or a reset fail.

3.3 For DNX 350 Only

Note: This is the last release that supports the DNX family of routers.

There is a problem where Carrier Detect and RXD LEDs remain illuminated after a WAN port test has been completed and the cable removed. (This has no impact on WAN software or normal DNX operations).

3.4 For RBX 200 Series Only

The RS-232D specification indicates that Pin 1 (EIA pin AA) should go to protective ground. The RBX 200 Series RS-232 cables do not have a Pin 1. With certain communications hardware, (Microcom V.25 bis modem Model QX/4232bis+ 14.4 kbps, Codex 334, V.I.R. mer232 19.2Kb RS232 modem eliminator, NEC Mmtpm TDM, and probably others), the router discards a significant percentage of packets with reports of (byte length) errors. When needed, you can order a replacement RBX 200 Series RS232 DTE cable, EC# 13H8755. In some cases, you may need an additional filter in the cable. The part number for the filter cable is 55H7756. Call customer services for price and availability.
A workaround is to ground Pin 1 on the modem to the DB25 shell on the RS-232 cable or use a different modem (Microcom DeskPorteFAST, for example).

3.5 For DNX 300 and DNX 350 Only

Note: This is the last release to support these platforms.

A buffer leak has been observed in the router when the DNX 300 or the DNX 350 is powered up but not connected to anything or all the interfaces are down and no traffic is present.
Media type selection of Auto-Config is always translated to STP, even if media is UTP. You can see this in two ways:
- Select Auto-Config as media type when booting the load.
- Via OneVIEW Router PSM software (Interfaces>Current Interface>Specifics...).
When you use the breakpoint command with autoboot (AB) and watchdog (WD) switches on, you get the error message:
```
BPT+4; BPT
DDTGET+34; NMI - Watchdog Timer Fired
```
and the router hangs until you press <CR>; then the router reboots. When just the AB switch is ON, the router goes to the > prompt.

3.6 ASRT Bridging

In Transparent and Source Route Bridging, the client ignores STEs that are converted to AREs when participating in an IP tunnel configuration using multicast addressing. Bridging over an IP tunnel using unicast addressing functions correctly.
When setting up bridging, it is important to understand that some protocols do not work well, if at all, in some bridged environments. VINES, for example, is not well suited to bridging over WAN links since the VINES clients and servers do not know that the bridging has occurred, and expect responses at the higher speeds of the LANs.
In a bridged network in which there are transparent and source-route paths between stations (with translational bridging), it is not possible to configure bridges whose largest frame size is less than 1470. Because there is no way to inform stations on the transparent bridge networks of a different frame size, 1470 is assumed. This is also true when working with DLSw.
When using the IP tunnel, you should avoid IP fragmentation when tunneling any protocol that cannot tolerate misordering of packets, such as SNA and NetBIOS. If the router that is doing the re-assembly receives both fragmented large packets and non-fragmented small packets, it can misorder them. While most protocols tolerate receiving packets out of order, the LLC-2 link layer fails if re-ordering occurs. If fragmentation is occurring, contact OpenROUTE Networks customer service for assistance in reconfiguring the router to avoid fragmentation.
For bridge configurations where the accumulative bridge personality is of Source Routing Transparent, that is, at least one interface has both source routing and transparent bridge behavior, the setting of group addresses for the IEEE 802.1d Spanning Tree Protocol on the source-routing-only Token Ring interface fails. This interferes with Spanning Tree Protocol execution on these Token Ring interfaces. The end result is that the associated bridge ports always remain in the forwarding state, allowing multiple paths, if they exist, for the Spanning Tree Explorer frames.

3.7 Bandwidth Reservation

Allocating less than 10% of the available bandwidth to the local class can cause the router to discard the routing packets of some protocols. This can destabilize those protocols, leading to sporadic reachability.
When a WAN restoral secondary link is an ISDN-dialed circuit, the primary BRS count command does not track packets the router redirects to the secondary link. You can see these packet counts if the BRS classes are configured on the secondary dialed circuit.
Enabling bandwidth reservation reduces the number of packets on the output device queue. This is done in order to make prioritization of output packets very timely because you cannot re-order packets in the device output queue. This may reduce small-packet transmit performance when you use bandwidth reservation. The bandwidth reservation configuration queue-length command can be used to increase the length of the BRS priority queues to help alleviate the small-packet transmit performance problem, but do it only if it is essential to do so. Setting the values for queue length too high can seriously degrade the performance of the router.

3.8 BGP4

BGP4 does not run on 4 MB routers. The router requires at least 8 MB, if you want to run any other capabilities concurrently with BGP4. BGP4 is very memory intensive.
BGP4 is not supported over unnumbered serial lines.

3.9 Data Link Switching (DLSw)

When LLC enters a local busy state, it sends an extraneous RR frame after it sends the initial RNR, and sometimes causes a remote LLC to send another I-frame. The local LLC always drops this I-frame, since it is in a local busy state. When the local busy condition is exited, the remote LLC must always retransmit this dropped I-frame. Although the LLC protocol is not being violated, it is inefficient and may possibly cause problems with other less-robust LLC implementations.
On 4 MB RBX 200 Series routers, it is important to ensure that there is enough memory for proper operation of DLSw. It is recommended that you use the set global-buffer command at the Config> prompt to lower the number of global buffers to 200. This frees up more memory for DLSw usage. This is very critical in the RBX 200 Series loads containing ISDN code that includes ISDN.
Certain versions of the IBM PC/3270 emulator hang for no apparent reason, ignoring all received frames. Some PC/3270 emulators running on slower PC's cannot handle more than one received frame at a time. To prevent this problem, set the SDLC transmit window to 1.

3.10 DNA Phase IV

The router cannot send level 1 routing updates to adjacencies that advertise their buffer size as less than 514 bytes. If you encounter the following ELS message
```
DN.090: Cannot bld lvl 1 rte on cir 1 net FR/0, blk sz too small -254
```
then adjust the buffer size on the adjacent IS to clear this problem.

3.11 DNA Phase V

When doing a Phase V to Phase IV packet translation, never set the internal bit.
DECnet/OSI VAX clusters are not supported. (This does not apply to Phase IV VAX clusters, which are supported.)

3.12 DVMRP

When running DVMRP, make sure the routing table size is adequate. Otherwise, you may start losing routes. Under certain circumstances, this may lead to router crashes.

3.13 EasyStart

EasyStart only displays ELS messages on the primary console port by default. For the CNX 600, the primary port is known as the DTE port. On the RBX platforms, it is known as port A. If you use the alternate console port, you must explicitly enable the display of ELS messages by typing talk 2.
EasyStart waits for the full indicated time (6, 30, 60, or 100 seconds), if all detected interfaces indicate down or up (as opposed to testing). For a Frame Relay interface, if there is no cable connected, the interface reports down.
When configuring on the RBX Series, the BNC connector is specified as a valid 10baseT connector even though the RBX Series router does not support the BNC connector. The EasyStart interface operates normally in this configuration. You can change the connector to the AUI only at the Eth Config> prompt.
When configuring EasyStart, over a WAN link, you require a MAC hardware type of 6 for Frame Relay and a MAC address value according to the following format: the first bit is a 1, bits 42-32 being the DLCI, and bits 31-0 being the IP address of the other end of the serial link.
For EasyStart to work correctly, you must enable BOOTP forwarding on the RBX Series for interoperability with IBM 6611 systems.
BOOTP extensions for EasyStart:
- New vendor-specific text field, with a tag value of 200, which specifies the configuration filename. If this field is absent, then the router uses the boot filename (standard field) as the configuration filename. If both are absent, the router ignores the reply.
- PPP, PSL, and Frame Relay behave differently because no MAC address is available. The BOOTP server entry must have a MAC hardware type of 6 and a MAC address value according to the following format:
Bit 47 must be set to 1, bits 46 to 42 are un-used, bits 42 to 32 are Frame Relay DLCI, and bits 31 to 0 are IP address.
- The IP address is the IP address of the other end (usually the central site) of the serial link. If using PSL or PPP, the router ignores the Frame Relay DLCI field. For example, using the CMU BOOTP server found on most Sun workstations, the following BOOTP tab file entry is for a router named d3-sl, using vendor extension format according to rfc1048, subnet mask of 255.255.255.0, hardware type of 6, hardware address 800080b9a301 (the 80b0a301 is the IP address of the other end: 128.185.163.1), an IP address of 128.185.163.3 and the configuration filename in the vendor extension field.

3.14 Ethernet

If you copy an Ethernet-based configuration into a Token Ring-based RBX 200 (or vice-versa), upon restart, the router clears the entire SRAM configuration to an unconfigured state. You can accomplish this form or configuration clone by loading the router with an r15.1 load, copying the configuration into SRAM, restarting the router, and then reloading with the r15.2 load.
On the GlobeTrotter 100, the Ethernet port always registers in the Up state, unless you explicitly disabled the port, or the hardware is broken.

3.15 Frame Relay

For Dual serial cards only (CNX Dual, DNX 300 Dual), overloaded Frame Relay does not allow for passing of routing updates. This causes the software to delete the routing table.

3.16 IEEE 802.5 Token Ring

For DNX 300m only, if a failure is introduced just downstream in another router or end station from the router, the router leaves the Token Ring and does not come back.
Intermittently, DNX 300 hangs when booting over Token Ring. You must press reset and restart.
When adding a manual ARP entry for an IP address, where the entry is on a Token Ring, and source-routing end station behavior is enabled on that Token Ring, the packets do not go out and the software does not generate an ELS message. Disabling the source-routing end-station works.
Token Ring now supports all valid packet sizes up to 18 K (17749). However, the larger the packet sizes, the fewer buffers there are in the router. At 18 K, there are only 28 buffers in a .5MB router. If there are multiple protocols enabled and more than three interfaces, the interfaces may not be able to come up. OpenROUTE Networks recommends that you use packets with a size above 5 K with caution.
IEEE 802.5 Token Ring interfaces running protocols that set the MAC address (DNA Phase IV or Apollo Domain) always fail their first self-test after a restart or reload. This includes testing a statically disabled interface. This is because the IEEE 802.5 does not allow a station to change its MAC address while on the Token Ring, and the device cannot learn its MAC address from the forwarder until it comes up.
When source-routing end-station behavior is disabled on an IEEE 802.5 interface (TKR config>disable source-routing), it does not respond to Ethernet loopback protocol messages at SNAP PID 00-00-00-90-00.
When an STP cable is connected to the Token Ring connector and the other end is not connected to a MAU, it is possible for the Token Ring interface to come up. However, it soon fails maintenance and goes down. It continues this up and down sequence as long as you do not plug the cable into the MAU. OpenROUTE Networks recommends that you disable the interface when this occurs.

3.17 IP

In MOSPF, a potential router fault can occur during downstream multicast route selection when the calculating router is an area border router and two or more border routers attached to the same area are advertising equal cost to the target multicast group member. OpenROUTE Networks recommends that you assign different link costs to each area border router interface. In the event the border router is directly attached to the backbone area, you should assign the least cost of all attached area interfaces to the interface towards the backbone.
When using IP, RIP, OSPF, or DLSw, the internal IP address belongs to the router as a whole and in many cases it is preferable to set an address that does not match any of the IP interface addresses, except in the case of RIP where it must match the IP address of a physical port/interface. Depending on the class of address you are using, there is one convention that OpenROUTE Networks recommends for all protocols, including RIP, that results in an address matching the physical port address and still remaining unique to the router. Match the Net ID and Host ID to a physical interface and set the subnet to 255.
If DVMRP is misconfigured to run on an interface that OSPF reserved as a secondary route, OSPF does not recalculate its routes upon disabling DVMRP, although the enabling/disabling of DVMRP is dynamic. Make sure you restart the router for it to take effect.
OpenROUTE Networks routers keep direct routing information in their IP routing tables. Packets originating at one end of a serial line to a nonexistent destination sharing the same subnet as the serial line interface may result in a routing loop when the two ends of the serial line belong to the same subnet. The loop continues until the TTL expires. This situation may also occur if the network has a default route.
When you run OSPF over Frame Relay, the software has a restriction on the number of neighbors that you can configure on a point-to-multipoint interface. If the number of neighbors on all point-to-multipoint interfaces for any particular area exceeds 120, the router may crash.
The router crashes if you assign a second IP address to a serial line with PPP encapsulation.
Routers running the EGP protocol should not be using OSPF area ranges. When using EGP and OSPF area ranges simultaneously, routers crash.
The OSPF implementation does not include an implementation of
TOS-based routing.

CAUTION:
Enable multicast on all OSPF routers on a multicast subnet. At a minimum, you must enable OSPF multicast on the designated router.

3.18 IPX

For IPX interface filters configuring greater than 300 total list-items over the four possible filter types (RIP router/RIP/SAP/IPX) may cause the router to crash during initialization with the following error:
```
Bughlt: MOS: no free kernel block for timer or signal 
```
300 appears to be the limits for the RBX 200 Series only. The CNX platform appears to be higher,

3.19 ISDN on RBX 205

For certain North American ISDN installations, it may be necessary to support 56 Kbps data service, as opposed to the default 64 Kbps data service. OpenROUTE 2.0 supports the automatic fallback to 56 Kbps data service in these cases; however, some early RBX 205 platforms do not. There are two ways to determine if the router hardware supports 56 Kbps ISDN data service. One way is to monitor the ISDN ELS messages when attempting to connect a site that only supports 56 Kbps data service. If the router hardware is unable to support the 56 Kbps data connection, the following ELS message is logged:

Unable to connect at 56 K bps - unsupported in hardware

Another way to check whether or not the hardware supports 56 Kbps ISDN data service is to check the hardware configuration under the Bootstrap Monitor:

Bootstrap Monitor V1.20
Copyright 1990-1994 Proteon, Inc.
uc

Boot Device types available:
IBD
Token Ring
WAN

Console Baud Rate: 9600
Main memory size 8 MB
IBD (flash memory) size 4 MB
IBD (flash memory) size 4 MB
Base MAC address: 10005A630437
System Part Number 1
System Serial Number 1
System EC Level (see below)
System Card Part Number (see below)
System Card Serial Number 05GFXQ
System Card EC Level (see below)

If the following values appear, then the hardware is not capable of supporting 56 Kbps ISDN data service, and you need to upgrade it:

System EC Level
System Card Part Number
TBS (2 MB IBD Ethernet)
TBS (4 MB IBD Ethernet)
TBS (2 MB IBD Token Ring)
TBS (4 MB IBD Token Ring)
TBS (4 MB IBD Token Ring)
System Card EC Level

However, if the following values appear, then the hardware is capable of supporting 56 Kbps ISDN data service:

System EC Level            E43940
System Card Part Number    72H2725 (2 MB IBD Ethernet)
                           72H2726 (4 MB IBD Ethernet)
                           72H2727 (2 MB IBD Token Ring)
                           72H2728 (4 MB IBD Token Ring)
System Card EC Level       E43940