Revision Notes for Router Software v8.10 rev 2
Date: 5/4/95
Subj: Revision Notes for 8.10 fix 2
===========================================================================
SITE MANAGER COMPATIBILITY
===========================================================================
Router version is managed by Site Manager version
v8.10 2.10
8.10 fix 1, 2 2.10, 2.10 fix 1
======================== Version 8.10 fix 2 ==============================
The image for FRE BN hardware is: 2577512 May 3 16:03 bn.exe
The image for FRE ASN hardware is: 2467141 May 3 16:04 asn.exe
The image for VME & AFN Floppy hardware is: 2362938 May 3 16:03 ace.out
The image for AN hardware is: 2307242 May 3 16:01 an.exe
The image for IN Flash hardware is: 2476796 May 3 16:29 in.exe
The image for AFN Flash hardware is: 2246611 May 3 16:29 afn.exe
The image for BNX BN hardware is: 1646820 May 3 16:28 bn.exe
The image for BNX ASN hardware is: 1526055 May 3 16:29 asn.exe
___________________________________________________________________________
Revision fix/8.10/2 corrects the following problems:
===========================================================================
CR 13920: IP.
A Slot may reset because of an "out of memory" or "insufficient
memory to continue => GAME restarts" condition. The loss of memory
occurs in the Ip Routing Table Manager and can be triggered by an
unstable environment where Synchronous links are going down/up.
The leak may continue after the environment stabilizes possibly
leading to the above mentioned slot resets.
CR 13529: OSI.
OSI router bus error panics when configured with an OSI External
(interdomain) interface going over a PPP line.
CR 13517: DLSw.
In a DLSw network that has AN hubs, if end stations send out LLC
test frames with a data length larger than 2KB, the AN may
experience memory corruption and/or PANIC.
CR 13522: Bridge.
When the Learning Bridge forwarding table is full traffic from
end station moved from one interface to another is not bridged.
CR 13432: LLC.
Receipt of an SSP contact message for a session that is in the
LLC2 disconnect phase (ie N2 *T1 not expired) causes a Bus Error
in the receiving router.
CR 13534: PPP.
When using the SHOW PPP LQR STATS and SHOW PPP LQR config scripts
the circuit names are displayed incorrectly.
CR 13493: IP.
Ip router can reset with 'PANIC in ip_rtm_misc.c at line 220'
while IP router on another slot is resetting.
CR 13571: Bridge.
While bridging traffic over an ATM network occasionally the
bridge would erroneously detect that a station had moved onto (or
behind) the ATM network. As a result the bridge forwarding table
would be incorrect and bridge packets on the ATM interface would
start to be dropped (incrementing wfBrTpInterfaceInDiscards).
Any connections that had previously been established would be lost.
CR 13373: DLSw.
In a heavily redundant DLSw network, DLSw may cause a panic
in mib_ent.c when creating a MIB instance.
CR 13460: AppleTalk.
AppleTalk ARP sub-system does not always recover after period
of AppleTalk router instability.
CR 13480: TCP.
Watchdog expiration slot resets can occur upon receipt of TCP
packet having corrupted TCP option (0 length of unknown option).
CR 13443: IP.
AFN router can hang following receipt of ARP request for one of
the IP addresses defined on that multi-netted interface.
CR 13538: IP.
Invalid Arps are logged erroneously when configured for ip
multinetting.
CR 13368: IP.
IP router generates checksum of -0 (0xffff) instead of +0 (0x0000)
for received packets with checksum of 0xfeff.
CR 12184: Source Route Spanning Tree Bridge.
A slot may incorrectly consider it's Bridge ID to be Root.
CR None: SYNC.
The granularity of the LAPB T1 ack timer has been changed from
seconds to tenths of a second. Note that this change affects
existing config files such that what the network administrator
previously configured in seconds is now a value in tenths of
a second. The MIB variables effected are;
wfSyncEntry.wfSyncRetryTimer and wfLabpEntry.wfLapbT1AckTimer.
CR 12355: IPX.
IN particular circumstances (misconfiguration in an IPX network)
IPX Directed broadcast(rip/sap) gets routed out same interface
it was received on.
CR 13366: OSPF.
Networks are not installed after moving ospf pt-to-pt link to
another router. If you dynamically reconfigure two routers,
deleting an ospf pt-to-pt link on router_A, and adding (moving)
the ospf pt-to-pt link to router_B. The pt-to-pt networks will
not be installed in router_A's routing table. Router_B is
properly advertising the routes, other routers (router_C) in the
network learn of the pt-to-pt networks from router_B. It is only
the pt-to-pt networks which are not installed in router_A's
routing table, other networks residing on the remote router
(router_D) are installed in router_A's routing table.
CR 13198: X.25.
When configuring an X25 Closed User Group with a value of 0 the
router does not create a facility in the Call Request Packet for
the CUG. Any other value from 1 - 999 will create this facility.
Site Manager says that all values from 0 - 999 are valid.
CR 12647: IPX.
In IPX networks with large numbers of end stations with many
clients attempting to open connections to one server a system
task and its associated memory can be lost.
CR 6432: System.
Network packets normally routed (if configured) are momentarily
bridged while routing sub-system (IP, IPX, AppleTalk...) is being
restarted. The packets are bridged permanently if the routing
sub-system is configured but disabled. This behavior as designed
in the series 7 and 8 architecture has created some rather
pernicious traffic conditions in looped IPX topologies as an
example.
This enhancement changes the router's behavior as described in
the chart below which makes distinctions between a routing
sub-system that is operating, configured but disabled, and
deleted.
SITUATION ACTION WHY
1. Box up - protocol Drop protocol. The way 5 series
disabled. platform works.
2. Boot with protocol Drop protocol. The way 5 series
disabled. platform works.
3. Box up - protocol Drop protocol. Stop routing loops.
fails.
4. Box up - protocol Bridge protocol. Protocol no longer
deleted. registered should in a
dynamic system act as
if never there.
5. Box up - protocol added Drop protocol. Similar to #2.
but disabled.
6. Box up - protocol Route protocol. Expected.
enabled.
Prior to this enhancement #1 thru #5 would bridge the protocol.
CR 13363: OSI.
When a Ping is issued from TI for OSI the field only allows you
to enter 38 Hex digits for the value of the address. The problem
is that you should be able to enter the Selector field with the
address which would be a total of 40 Hex digits.
CR 13297: DLSw.
The receipt of a CANUREACH with different source and destination
SAP's where the destination SAP is one defined in the DLS SAP
table but the source one is not, will cause a BUS error. If both
SAP's are defined in the SAP table the problem doesn't happen.
There is also no problem if neither is defined in the SAP table.
CR 13119: IP.
Snap Encapsulated 802.3 ARP requests incorrectly have the Hardware
Type Field set to 1 (Ethernet). This field should be set to
6 (IEEE 802 Network). It seems that certain End-stations, IBM
RS6000s for one, will not respond to our SNAP Encapsulated ARP
Requests over Enet when the endstation is configured for an 802.3
Network Interface. They will only respond if this field is set to 6.
CR 13210: Switched Services.
Hold down timer for dial backup does not work. If it is configured
the primary line may never come up on boot.
CR 13200: Bridge.
Configuring Bridge Filters that filter on DST MAC address has
higher precedence than the filter rule#1 which should be
triggered first. This appeared when two filters were configured
on one circuit. The first filter is a FWD filter to circuit E22
with field of ETYPE 0x6004. The second Filter is a drop filter
with DST MAC of 0x000000000000-0xffffffffffff. The second filter
has lower precedence and should be applied after the first filter
but the packets with ETYPE 0x6004 were dropped.
CR 13299: MIB.
The Mib tree does not support the case of two instances of the
same object, when one instance is a prefix of the other one
(variable length instances) the Mib will insert both of them in
the tree, but the longer one is then inaccessible and if one
deletes the shorter instances then both instances will be deleted
(creating a memory leak).
CR None: System:
TI 'save config' command fails on large config files even though
free memory available on the slot running TI is more than twice
the size of the config file. When the save fails, although an
informative error is recorded in the system log, NO error is
displayed to the TI session.
Note: The configuration save module now requires much less memory to
save the file to flash but it is important to note that the
memory used comes from a single segment which must be smaller
then the value recorded in the slot's MIB attribute
wfKernelMemoryMaxSegFree.
If the config save does fail for lack of a single memory segment
large enough there will now be an error displayed to the TI
session like; 'Config. file retrieval error.'
CR 11692: System.
Slot resets infrequently occur with an 'out of GID' error in the log.
CR 13066: Bridge.
Unable to reach Host Only interfaces after disabling Bridge.
CR 13179: Vines.
Vines router tends to PANIC during momentary period of all packet
buffers on a slot being in use.
CR 12751: IPX.
With a looped environment and number of paths greater than 1,
IPX never loses routes that should be aged out.
CR 13000: Source Route Bridge.
Source Route Traffic Filter to Forward to Circuit Group only,
also forwards packets to IP Encaps subsystem.
CR 13057: IP.
Slot resets with Tag violations can occur on receipt of
re-ordered IP fragments when IP header of the first fragment
exceeds 20 bytes.
CR 12916: PPP.
Slots running IPX router can panic when changing IPX network
address on a PPP interface.
CR 12807: Bridge.
Bridge SNAP filters do not work over PPP or Frame Relay interfaces.
CR 13131: System.
When the operating system's mechanism for single threading access
to a shared resource is utilized and one process waits behind
another, a 32 byte memory block is lost. Over time, this memory
leak that does not show up in the wfKernelEntry stats other than
a lack of free memory, can cause 'out of memory' slot resets.
CR 11762: DLSw.
A particular Netbios SQL application fails when going through
WF DLSw because we only allow for one connection per data link.
This application needs the client and server to have 2 sessions
opened, for 2 different Netbios names, at the same time. The first
session is opened OK but when the second tries to come up the WF
routers will terminate the data link (DISC at both ends).
CR 10670: OSPF.
OSPF adjacencies do not form when using the dial backup line after
the primary line is disabled.
CR 12895: Vines.
The Vines client program SEND.EXE does not work on serverless
segments.
CR 13021: SNMP.
SNMP Get Request (45 Objects) from SunNet Mgr fails with
Error - Get Request Too Big. It used to work in 7.71 fix 8 and it
fails with 7.80 fix 5 and up.
CR 11433: Token Ring.
The ring data rate is incorrect in the 8502 frame sent from the
router to the IBM Lan Network Manager. Even though the ring is
configured for 4Mbps it still sends the ring data rate as 16Mbps.
CR 12836: OSPF.
When filtering OSPF events to keep only INFO and STATE with
SM/Sun 2.0, we still have OSPF TRACE messages in the log after
OSPF initializes.
CR 12555: LLC.
After having opened and closed around 270 LLC sessions through a
router, it is no longer possible to establish new sessions. The
typical error messages in the log are:
933: 01/20/95 12:12:40.271 DEBUG SLOT 2 LLC Code: 11
Context not found or unavailable.
934: 01/20/95 12:12:40.271 DEBUG SLOT 2 LLC Code: 11
LLC ERROR: No Context Available
CR 12532: DLSw.
When DLSw and LNM are both configured on the same slot either
may fail to operate.
CR 12830: IP.
IP Router responds to a UDP echo request with the same DST and
SRC ports as the requesting packet.
CR 13514: Bridge.
Changed Bridge 'Forwarding Table Full' message from debug to
warning.
CR 13483: OSPF.
Adding a second non-backbone area (with no 0.0.0.0) can cause
Tag Violation slot resets.
CR 13286: IP.
IP packets with bad headers (fails checksum) where the header's
total length field contains a large value (larger than the size
of a buffer), can cause Tag Violation slot resets.
CR 13407: IP.
Routing table missing IP routes when parallel rip networks are
configured between 2 routers. If new rip advertisements
containing info on the same route(s) appear simultaneously on
two interfaces which reside on different slots, the routing table
can become corrupted so that no slot owns the just-learned routes.
CR 13392: XMODEM.
XMODEM - YMODEM is padding files transferred from the Router.
This is not a problem. Both XMODEM and YMODEM send padded files
as they should because frame size in these protocols is
constant - 128 bytes (128,1024 for YMODEM). The issue raised
in this CR was caused by TI's help saying 'Use the -y option to
send binary files WITHOUT being padded', which is incorrect.
The repair for this CR is to change the TI's help to read:
' ... With XYmodem files are padded with Control-Z (0x1A) such
that the size is a multiple of the frame size (128 or 1024
bytes). Use the -y option to receive binary files WITHOUT being
padded.'
' ... The padding (0x1A) will be removed on a receive side. '
CR 13367: AppleTalk.
On receiving an NBP broadcast request from a node on a directly
connected network, NBP lookups with the multicast address
different than what is sent in ZIP GetNetInfo Reply is sent from
AppleTalk router. As a result of this we do not get any NBP
replies from the servers on the same zone on the different
networks.
CR 13204: IP.
IP All Subnet Broadcasts not forwarded when ASB is enabled on
outgoing interface when ASB is initiated from an end-station
more than one hop away.
CR 13277: OSPF.
OSPF can reset with 'Error in ospf_rtab.c at line 399' when WAN
link is disabled.
CR 13188: System.
The event log capacity for a VME based router has been increased
from 32 KB to 64 KB.
CR 13399: IP.
Version 8.01 doesn't handle IP Rip updates correctly from a
multinetted interface within the same class address with the
remote router having more subnets defined, and configured for
explicit broadcast.
CR 12776: DLSw.
Disabling LLc2 on a router can cause Panic Vector 2 Bus Errors
if Sap 0 is configured.
======================== Version 8.10 fix 1 ==============================
The image for FRE BN hardware is: 2563507 Mar 10 18:37 bn.exe
The image for FRE ASN hardware is: 2490217 Mar 10 18:37 asn.exe
The image for VME & AFN Floppy hardware is: 2363445 Mar 10 18:44 ace.out
The image for AN hardware is: 2306663 Mar 10 18:40 an.exe
The image for IN Flash hardware is: 2333448 Mar 10 18:42 in.exe
The image for AFN Flash hardware is: 2248720 Mar 10 18:42 afn.exe
The image for BNX BN hardware is: 1584296 Mar 10 18:39 bn.exe
The image for BNX ASN hardware is: 1488650 Mar 10 18:39 asn.exe
___________________________________________________________________________
CR 13253: ATM.
Disconnecting then reconnecting the fiber into the receive
interface on the OC3 SingleMode interface can place the receiver
into a reset state, not recovering until the driver has been
reinitialized.
CR 13228: ATM.
When running multiple PVCs over ATM, the router may display an
ERROR in atmalc_xmt.c. This seems to occur when traffic is
transmitted out all VCs simultaneously.
CR 13240: ATM.
An ATM interface has received the following message:
# 1372: 02/22/95 12:30:58.590 DEBUG SLOT 3 ATMINTF Code: 96
Port 1: coprocessor driver fault 23.
This occurred while running IPX Vines and IP over 80 group
mode VCs. This error is not easily reproducible and it has
only been seen twice.
CR 13177: ATM.
At random times, the ATM driver will log "coprocessor driver fault
26" or "coprocessor driver fault 27". In addition, the driver
will reset at atmalc_rcv.c at line 323 when this happens. This
anomaly occurs very infrequently and the driver recovers quickly.
CR 13220: IP.
DVMRP may cause IP on the slot to restart if a network with a
non-contiguous subnet mask is advertised to DVMRP by a neighbor.

Feedback | Service Search
| ERC
Copyright
© Bay Networks,Inc., 1997. All rights reserved.
|
/tr> |