The information in this article applies to:
SUMMARYMicrosoft LAN Manager 2.2cThis article describes the bugs that are fixed in Microsoft LAN Manager 2.2c patch. This description includes only those fixes not included in previous LAN Manager 2.2 patches (2.2a and 2.2b). For a description of all fixes in LAN Manager 2.2c or for information on how to obtain the patch, query on the following word (without the spaces) in the Microsoft Knowledge Base: LMPATCHNOTE: The LAN Manager 2.2c patch contains all components that have been fixed since the original release of LAN Manager 2.2, however this list describes only those bug fixes that are new in this release. Bug Fixes Added in the LAN Manager 2.2c Release
Text of Bug FixesComm Server MS-DOS Client Hangs When Comm Server Logs Off NetworkCSD: LM22.CSD00.077SYMPTOMS
Your DCA/Microsoft Communication Server (Comm Server) MS-DOS Client
computer stops responding (hangs) when the Comm Server service is removed
from the network. The hang occurs when exiting the Comm Server MS-DOS
Client 3270 emulator by pressing <SHIFT-SHIFT>.
Binary AffectedTINYRFC.EXEMS-DLC and Attachmate Extra Hang WindowsCSD: LM22.CSD00.094Your computer running Attachmate Extra for Windows v3.22 and LM2.2 with MSDLC displays a "510" error message after about 8 session opens and closes and if you try to close the session again your computer will stop responding (hang). Receiving a frame on an Ethernet network with the source routing bit erroneously enabled can cause a machine running NetBEUI to propagate a network storm. LAN Manager 2.2 clients may stop responding (hang) after receiving more than one response to a single name query sent out over TCP/IP. Scripts may get cut off when you transfer large amounts of data during a logon script or run a script over a slow link. When an OS/2 version 1.3 LAN Manager 2.2 server has more than 5000 users in its NET.ACC, RASADMIN reports this error when you select users and permissions while in the RASADMIN utility: You are prevented from adding users to RAS access. On a LAN Manager 2.2 enhanced MS-DOS workstation running Windows for Workgroups 3.1 or 3.11, trying to load Microsoft TCP/IP on top of a Novell Netware ODI MLID and ODINSUP fails with this error: Microsoft NetBEUI loads and functions in this configuration. Using a Microsoft LAN Manager enhanced workstation loaded into EMS may cause an MS-DOS critical error under Windows 3.1. This problem, which is intermittent and may occur during workstation stress, is specific to the enhanced redirector loaded into EMS running under Windows. If you abort when a critical error is reported during a large-file copy under Windows, the machine hangs if the workstation service is loaded in EMS. If the machine doesn't hang immediately, a network error occurs. Upon disconnection, the system can retry successfully. The Domain Name Resolution (DNR) in LM TCP/IP does not resolve a non- authoritative response from a Domain Name Service (DNS) server with consistency. It works correctly the first time, but on subsequent responses it does not work. LAN Manager servers running any version up to and including 2.2b on complex token ring networks may experience a Trap-D.
(CSLIM and IP may vary slightly on different hotfix levels)
When a remoteboot workstation is starting the redirector or when it is actively accessing the FIT translated RPL volume (drive C:), and a multicast packet is received, the system sometimes stops responding (hangs). The Microsoft LAN Manager client workstation redirector (NETWKSTA.EXE) can be loaded in the upper memory area under MS-DOS 5.0 and 6.0, but not if you upgrade to MS-DOS 6.2. An application written to use OS/2 2.1 DosSFileLock functions for unlocking files causes a Trap000D on the workstation when unlocking files on a redirected drive. This does not occur when running an application that is unlocking files on the local workstation. Users may get an intermittent sharing violation when their logon script is running and SHARE.EXE is running. An application can fail because DosReadMailslot sometimes returns a message that was not addressed to it. Generally this occurs when mailslot traffic is heavy. Attempting to ping another host by name from an OS/2 or MS-DOS LAN Manager computer using the Domain Name Resolution (DNR) service may return the message: In some instances of this problem, Windows for Workgroups 3.1 workstations received this error when they attempted to ping a remote node by using its alias; Windows NT machines did not. After the Microsoft LAN Manager 2.2b update is applied to a server running Microsoft SQL Server or DCA/Microsoft Communications Server, client workstations may not be able to receive any data from the server. A SQL Server client running SAF (or other front end) will not receive any information in response to a query. Some named pipe clients (including ISQL) fail although the READPIPE/MAKEPIPE test programs function properly. A Communications Server client will no longer be able to open an emulation session (such as 3270 or 5250). When the user tries to open an emulation session, the host sign on screen will never appear and the emulator will hang. When you run DCA/Microsoft Communications Server (Comm Server) version 1.2 or greater on LM 2.1 or 2.2 under heavy client stress, "broken pipe" errors start occurring at the server, requiring a reboot to recover. When the failure occurs Comm Server clients will no longer be able to open a session with the server, the server console will be very slow to respond to local commands, and the "NET FILES" command at the server will show multiple open named pipes (\PIPE\COMNODE) for a given client. NOTE: The failures are preceded by the LAN Manager server returning a named pipe file handle of 0000 in response to a Server Message Block (SMB) Open request, indicating that the server is out of internal RPQ data structures. Receiving a frame on an Ethernet network with the source routing bit erroneously enabled can cause a machine running NetBEUI to propagate a network storm. Using an ICL EtherTeam 16i adapter and the protected mode redirector on Windows for Workgroups 3.11 can cause the LAN Manager 2.2b TCP/IP (TCP 1.0a) to fail under certain circumstances. After starting Windows for Workgroups, you should be able to ping in a command window. If you are using protected mode NetBEUI, however, and start a large software package (such as Excel) over the net, an attempt to ping in a window fails with a DGN0209 error. NetBEUI works, but TCP/IP fails. On boot, a LAN Manager 2.2b OS/2 server fails with this error:
CAUSEThis is because the Pipe callback code is never called if there is an error on the session. Binary AffectedNETWKSTA.*Network Servers Running TCP/IP Fail with a Trap DCSD: LM22.CSD00.098This is caused by an incorrect User Datagram Protocol (UDP) broadcast frame that is sent to all servers on the network. Binary AffectedTCPDRV.OS2UNIX Server Drops Connection to LAN Manager TCP/IP ClientCSD: LM22.CSD00.067If there isn't a RECEIVE or RECEIVE ANY command outstanding, or if the application doesn't issue a RECEIVE or RECEIVE ANY in a short period of time, the RFC KEEP ALIVE messages sent from a UNIX server to an MS-DOS client use up all of the NetBIOS receive buffers, because messages coming from the line are managed only when a RECEIVE command is issued. In this case the RFC KEEP ALIVE messages are stored in the buffer, until it is full, at which point TCP can't send any more data and requests retransmission of the frame. After a certain number of retransmissions the UNIX server drops the connection. This happens if the user is inactive for 10 to 15 minutes. Binary AffectedTINYRFC.EXETinyrfc Problem in RECEIVE Command Returning "MORE DATA"CSD: LM22.CSD00.066On some networks (with bridges), some token ring frames have 90 hex in the AC field (normal is 10 hex). The 80 hex bit is reserved. So when a SABME frame has the 80 hex bit in the AC field turned on, MSDLC copies the AC field as is to the outgoing frame and the MAC driver cannot not send the frame. Binary AffectedMSDLC.EXENetBEUI Network Storm Caused by Erroneous Source Routing Bit on EthernetCSD: LM22.CSD00.092When a link is established and the SABME frame has the source routing bit set erroneously, the routing length field is accounted for in the link table entry as well as in the original routing header info. This causes all subsequent frames on this link (including the following UA) to contain the original source routing info, but the source routing bit is ignored on the receive end, causing the frame to be misinterpreted. When the original frame contains a SABME with the source routing bit set incorrectly, the two machines enter a loop broadcasting SABME frames back and forth. Binary AffectedNETBEUI.OS2TCP/IP Should Handle Duplicate ResponseCSD: LM22.CSD00.067The second response is being received either because there are duplicate names on the network or because a third-party machine is sending more than a single response. Rather than discard the second response, the LAN Manager client is attempting to establish two TCP/IP connections, resulting in a hang. Binary AffectedTINYRFC.EXEWindows Logon Script Time-out Values Too ShortCSD: LM22.CSD00.068This is caused by a hard coded timeout in code. It was 10 seconds in LAN Manager 2.1 and 30 seconds in LAN Manager 2.2. The RAS utility was not written to accommodate this many user accounts. TCP/IP issues an OpenAdapter NDIS call to determine if a token ring network card is available and functional. If this call fails, TCP/IP fails to load. Novell's ODINSUP NDIS support layer for ODI drivers does not provide support for the OpenAdapter call. Instead it performs its own adapter initialization and sets bit 4 of the MAC status table s flag field to indicate that the adapter is already open. When you abort out of the critical error handler, control is not passed back to code that called the error handler because the workstation must switch back the original LIM paged memory to pass control back. It cannot do this, however, because the necessary code is in the part of the workstation that lies in LIM. Sockets was specifically requiring an authoritative response. This occurs only on token ring. The size of the Source Routing Information buffer allowed for all protocol headers is inadequate in some cases. This was caused by a routine being called by means of a dispatch table to low memory when the code resided in low memory and should have been called directly. Binary AffectedNETWKSTA.*LAN Manager Client NETWKSTA.EXE Does Not Load in UMA on MS-DOS 6.2CSD: LM22.CSD00.073The problem is that the BL register is not initialized. LAN Manager assumes that the version number is "X.00." The "X" number is loaded into the BX register, while the BL register is left unchanged and entered as "00." This has worked until now because recent MS-DOS versions have been "X.00" releases. NOTE: NET.EXE is for LAN Manager Basic workstations only. Binaries AffectedNETWKSTA.500NET.EXE OS/2 2.1 Application File Locking Causes Trap DCSD: LM22.CSD00.075The OS/2 2.1 client redirector used uninitialized global variables instead of passed arguments in the FS_FILELOCKS function. Binary AffectedNETWKSTA.SYSLMSCRIPT.EXE Causes Sharing ViolationCSD: LM22.CSD00.081The sharing mode used to open the script file was not compatible with multiple user access when workstations run SHARE.EXE. Binary AffectedNETAPI.DLLDosReadMailSlot Reads Wrong Mailslot Under StressCSD: LM22.CSD00.079The code used when writing mailslot data to a local mailslot is not being marked as busy early enough in the routine to keep other mailslot datagrams from affecting the data. Binary AffectedNETWKSTA.*TCP/IP Remote Name Cannot Be Resolved - Dgn0217 ErrorCSD: LM22.CSD00.084When LAN Manager calls the DNR it explicitly demands an authoritative response for a cname. If a DNR request is sent from a system in another domain and that request hasn't been performed previously, the local DNR contacts the authoritative Domain Name Server (DNS) and receives an authoritative cname reply, showing the cname and the actual name (and IP address) of the target system. However, if another request is sent for the same cname afterwards, the local DNS system replies with the same answer from cache, this time marking it as non-authoritative. This lookup fails, returning an error message such as "unable to resolve host" or "unknown hostname." Comm Server clients successfully open a named pipe to the server, then post a read request, and successfully write data down the pipe to request a session from Comm Server. Then the Comm Server node fails to respond to the client read request. In the Comm Server internal trace, the write pipe messages appear to fail, causing the server to retry the send on the pipe. Binary AffectedNETWKSTA.SYSBroken Pipe Errors Under Comm Server StressCSD: LM22.CSD00.085Due to a timing problem in the LAN Manager redirector, internal RPQ data structures are not freed when server side named pipes are closed very quickly. This eventually leads to broken pipe errors, which require Comm Server to be rebooted. Binary AffectedNETWKSTA.SYSIncorrect Source Routing Bit Causes NetBEUI Network StormCSD: LM22.CSD00.080On an Ethernet network an incorrectly formatted frame from one Network Interface Card (NIC) causes an incorrect response frame from another NIC. This process repeats itself and causes a storm on the network. Binaries AffectedNETBEUI.EXENETBEUI.OS2 TCP/IP Fails on WfW 3.11 with Protect Mode NetBEUI and ICL 16i AdapterCSD: LM22.CSD00.083If the MAC driver for the ICL card fails to copy the frame on TransferData (more likely while loading Excel from the server over NetBEUI), the TCP/IP driver fails to release the buffer, resulting in a memory leak that eventually makes the TCP/IP stack non-functional. Binary AffectedTCPTSR.EXETCP/IP Fails with Error PR00022 During Bind on LM2.2b OS/2 ServerCSD: LM22.CSD00.088An OS/2 server running TCP/IP may fail if it is accessed by another client during the boot process, because the server's TCP/IP stack is busy dealing with numerous SYNs from the client to bind. RESOLUTIONThe default timeout has been left at 30 seconds, although new code allows you to set the timeout by adding an entry to the LANMAN.INI file. To change the default timeout, add a section to LANMAN.INI, [windows], and insert a logintimeout value between 1 and 32000 (seconds). Binary AffectedNETAP.DLLRASADMIN Reports Not Enough Memory If UAS Has 5000 UsersCSD: DIALIN.CSD01.005New MS-DOS and OS/2 RASUSER.EXE utility that can be run from command line to allow setting / viewing and clearing of RAS permissions. Type RASUSER for help on usage. Binary AffectedRASUSER.EXETCP/IP Load Failure with ODI Under Token RingCSD: LM22.CSD00.071The TCPTSR driver was changed so that it does not attempt an OpenAdapter NDIS call until it checks bit 4 of the MAC Service Specific Status Table to see if the adapter is already open. Binary AffectedTCPTSR.EXELAN Manager Enhanced Redirector Loaded into EMS Causes Critical ErrorCSD: LM22.CSD00.076A new routine in the low code segment now is called in place of the critical error handler. It switches LIM pages and then calls the original critical error handler. Binary AffectedNETWKSTA.*DNR: Non-Authoritative Response of an Alias Address Is Not ResolvedCSD: LM22.CSD00.070This was changed so that upon receiving a non-authoritative CNAME reply, the client sends a second request, this time for the real name. When the second request comes back, (non-authoritative again) and is accepted. Binary AffectedSOCKETS.EXETrap D in TCPDRV.OS2 CSLIM=9F60 IP=162FCSD: LM22.CSD00.082Two more bytes are allocated to the Source Routing Information buffer. Binary AffectedTCPDRV.OS2Remote Boot Workstation Locks Up When Server Sends Netlogon MulticastCSD: LM22.CSD00.078A fix is available that modifies the NMTSR.EXE and NMDRV.OS2 drivers so that they do not demand authoritative responses. Binary AffectedNMTSR.EXELAN Manager 2.2b Server Update Hangs SQL & Comm ServerCSD: LM22.CSD00.085Binary AffectedTCPDRV.OS2LM 2.2/2.2b Server Fails With CSLIM=8A1E/CSLIM=8A0ECSD: LM22.CSD00.089SYMPTOMSThe failure CSLIM=8A1E is reported on LAN Manager 2.2 systems; CSLIM=8A0E on LAN Manager 2.2b systems. Serious network problems may be observed at the time of failure, including frame misalignments, corrupt packets and excessively high network utilization. Multiple traps across multiple servers may occur simultaneously, since all servers receive the same corrupt packets.CAUSEThe network protocol driver receives a truncated DLC administrative frame (RR) because an adapter or bridge malfunctions under network stress and generates a truncated packet that contains a valid cyclical redundancy check (CRC) but incomplete NetBEUI data.RESOLUTIONTruncated frames are now rejected even if they pass a CRC check by the network adapter driver. However, the network device responsible for creating the truncated packets must be repaired to assure proper network performance.Binary AffectedNETBEUI.OS2Unable to Unload When Sockets and DNR Are Loaded in LM 2.2bCSD: LM22.CSD00.099SYMPTOMSTCP/IP utilities such as Sockets, DNR, Telnet or BAPI cannot be unloaded after they are loaded on a LAN Manager 2.2b client running TCP/IP stack 1.0a. If you try to unload one of these utilities, an error message states that the utility is not loaded.Changing load order causes symptoms to differ but does not correct this problem. In the most drastic case, loading Sockets last causes the machine to hang when you try to unload it. The TCPTSR.EXE and TCPDRV.DOS are dated October 1993. SOCKETS.EXE and DNR.EXE are dated June 1993. Telnet and BAPI are dated February 1992. CAUSESockets returns an invalid buffer size, which creates the problem.Binary AffectedSOCKETS.EXESome products included here are manufactured by vendors independent of Microsoft; we make no warranty, implied or otherwise, regarding these products' performance or reliability. Additional query words: wfw wfwg prodlm patch update fix CSD
Keywords : |
Last Reviewed: November 9, 1999 © 2000 Microsoft Corporation. All rights reserved. Terms of Use. |