IBM Peer Fixpak (IP08602) - Italy

IBM OS/2 Warp Server for e-business Fix Pack IPI8602 Italy This Fix Pack can be applied to: - IBM OS/2 Warp Server for e-business This Fix Pack contains fixes for the product listed above and for the following components of these products: OS/2 LAN Server, OS/2 LAN Requester, UPM, HPFS, RIPL and Fault Tolerance. WARNING: 1 - This Fix Pack must be installed using Corrective Service Facility (CSF) 2-B Version 1.42 or above. WARNING: 2 - Archiving is not enabled for this product. IP08602.INF, located on the first Fix Pack diskette, contains installation, and other useful information about this Fix Pack. Please review before installation of service. Diskettes can be created using the LOADDSKF utility. There are 3 files/images included in this package: ============================================================== CSFIT142 DSK Fix Pack Tool Diskette 3.50" IPI8602 1DK LS DISKETTE #1 3.50 1.44MB IPI8602 2DK LS DISKETTE #2 3.50 1.44MB INCLUDED APARS Note: This section describes Warp Server for e-business APARs that are corrected in IPx8602. o APAR = IC24478 IF THE BIT NO. 10 OF SRVHEURISTICS IN IBMLAN.INI IS SET TO A VALUE >= 6,THE SCAVENGER WAITS TOO LONG BEFORE WAKING UP. If the bit 10 of srvhueristics in ibmlan.ini is set to a value >=6, the scavenger thread of netservr.exe waits too long before waking up. Hence listens are not posted on all the adapters and a 'net view' command from certain clients fails. o APAR = IC24817 INCORRECT DIR LISTINGS IN A DOS WINDOW FROM A REDIRECTED DRIVE CONNECTED TO WARP SERVER FOR E-BUSINESS Incorrect dir listings in a DOS window on a redirected drive of the client connected to Warp Server for e-business. o APAR = IC24950 TRAP BY COPY FILES ON WINNT TO WARP SERVER SMP OR DIR COMMAND When issuing COPY x:\path\*.* c:\temp while x: is network shared drive of Windows-NT (both fat and ntfs) and c: is the 386hpfs local drive, if those files are too many (we usually have it with 400 files), it causes trap at CS:\EIP=33b8:\0000268f on Warp Server smp machine. This problem also happens with dir /ad command (around 10%), or simply dir command (a few %) also. o APAR = IC24960 NETAPI.DLL TRAPS BECAUSE OF STACK OVERFLOW. When a 32bit API is called, it traps in 16 bit NETAPI.dll because of low stack space. The test exe runs successfully on W3 and W4, but traps on WSeB. o APAR = IC25155 PENDING ASYNC CLOSES CAUSE THE REDIRECTOR /CLIENT TO HANG. Async closes are sent out by the redirector. These calls are completed on the next request sent to the server. Until then the call remains pending in the redirector table. Given a large number of servers this could lead to the client hanging when the table gets full. For example, copying a file from a client to over 25 servers using UNC style redirection could cause the client to hang. o APAR = IC25503 LANINST.EXE TRAPS IN NETAPI.DLL ON WARP SERVER FOR E-BUSINESS WHEN JFS VOLUME IS PRESENT. A Trap occurs when LANINST is run on Warp Server for e-business machine which also has JFS volume. Even though the trap screen shows trap in DOSCALL1.DLL, actually NETAPI.DLL traps due to non-existing semaphore. o APAR = IC26138 MEMORY LEAK IN LANPROBE.DLL Logon to a BDC fails with access denied but logon to PDC works fine. Net error on BDC gives: NET3226: An error occurred while synchronizing with the primary domain controller, PDC. The hex data 08 00 indicates a lack of memory on the system. o APAR = IC26281 MUG_GET_UID INTERFACE FUNCTION NEVER CHECKS FOR THE USERID AND PASSWORD OF AN API LOGON DONE BY ANOTHER PROCESS IN SAME SESSION The UPM interface function MUG_GET_UID never checks for the userid logged on the same session using a logon API like UPMELGN, UPMEULGN, U32EULGN etc. That is, if a process A does a UPMELGN and then it spawns another process B, a MUG_GET_UID call in B never returns the API logon done by A. The return of userid from B is a must for external programs like DB2 which use the interface function MUG_GET_UID. o APAR = IC26416 NET USER USERID /ASSIGN PUBLIC:APPNAME DOES NOT SAVE THE UPDATE. The NET USER USERID /ASSIGN PUBLIC:APPNAME command does not save the update. If we issue NET USER USERID to get the information about the user after issuing the above command, it shows no application is assigned to that user. o APAR = IC26424 TRAP IN NETWKSTA WHEN DOING A RAW READ AND LOGOFF IS FORCED BEFORE THE READ IS COMPLETED. We observed a trap in netwksta when it is doing a network read. For example, copy a remote file to local machine using RAW READ, and a logoff is forced before the RAW read is completed. o APAR = IC26425 REMOTE LAN ADMINISTRATION USING CLI AUDITS "***" AS A USERID IN AUDIT.LOG FILE INSTEAD OF ACTUAL USERID. o APAR = IC26426 NET SESS SHOWS NULL USERID WHEN LOGGED ON FROM PLC OR CLC CLIENTS FOR WINNT 1) Logon to OS2 server from WinNT with CLC or PLC. 2) Run NET SESS command on OS/2 server. You will find NULL user id displayed on the screen. NT opens a NULL session on OS/2, and OS/2 server is not aware of that. o APAR = IC26427 TRAP IN HPFS386 FINDNEXT ROUTINE DUE TO BAD SMB. Trap occurs in hpfs386 FindNext routine. Reason: The SMB has a path name whose terminating character is a DBCS lead byte which confuses hpfs386 leading to a trap. o APAR = IC26430 ALL THE COMMANDS IN THE *.LST FILE FOR THE LOCKED FILE DEVICE DRIVER IBMLANLK.SYS AND IBMLANLK.EXE ARE NOT EXECUTED. All the commands in the *.lst file for the locked file device driver ibmlanlk.sys and ibmlanlk.exe are not executed in Warp v4. Generally the process fails to complete when the *.lst file has more than 1000 command lines. For example, when the command MOVE is used to move more than 1500 files, only first 1216 MOVE commands statements are executed and the remaining files remain in the source directory itself. o APAR = IC26433 PRINT SERVER TRAPS Print Server traps in NETAPI NB_OPEN_R_2. o APAR = IC26438 TRAP OR HANG IN HPFS386 Trap or hang in hpfs386. Failing routines are: FindDir, TrimSBDir, MHF. Heap corruption and other traps can also result. Get trap 0e in hpfs386 due to race condition when multiple threads open a new file. o APAR = IC26439 TRAP IN NETWKSTA WHEN DOING A RAW READ AND LOGOFF IS FORCED BEFORE THE READ IS COMPLETED. We observed a trap in netwksta when it is doing a network read. For example, copy a remote file to local machine using RAW READ, and a logoff is forced before the RAW read is completed. o APAR = IC26440 API U32EUSRL, REFERS NETWKSTA.SYS TO DETERMINE LANROOT The api U32EUSRL when used to determine the users logged on to the domain, refers to netwksta.sys to determine the lanroot. Since there is no file netwksta.sys (netwksta.200), this search fails and the default search for muglib dir is searched. In the WSOD environment since all request from the client go to the server, this causes a lot of unnecessary traffic. o APAR = IC26441 HPFS386 SWAPPER FILE CORRUPTION CAUSING TRAP 0E IN REXX.DLL The swapper.dat file on HPFS386 drive gets corrupted if there is growth failure. This causes traps in applications, rexx.dll in this case, due to failed swap in requests. Problem is that HPFS386 fails to reset the file size in OFT correctly after a growth failure, say due to lack of requisite amount of free space. o APAR = IC26442 HFS0019 AFTER AUTOCHECK IS RUN ON LARGE HPFS386 DRIVES When HPFS386 runs autocheck, chkdsk may invoke CHKDSK.SYS if the drive is large and there is not enough physical memory below 16MB. But then, after the autocheck is done, HPFS386 may not be able to complete the init (not enough mem for heap/cache) there by hanging the system with HFS0019. Note that if we complete the init, then system will be rebooted by CHKDSK.SYS later on in its INIT_COMPLETE processing. So, to avoid this hang, we must reboot the system from HPFS386 itself, through a new IOCtl provided by CHKDSK.SYS (PJ26974). o APAR = IC26465 HANG IN HPFS386 UNDER HEAVY STRESS DUE TO CORRUPTED B_LCNT Hang in HPFS386 under heavy stress due to corrupted value in B_LCNT. Symptoms will be: - number of threads blocked waiting for PBUFs (Path buffers) - theads holding PBUFs blocked on a different buffer whose B_LCNT happens to be 0xff, which clearly is invalid.