Carefully examine each fix for applicability before using it.

• Download a file with all fixes

Download Fix File

The CLAW link driver is designed to use multiple input and output buffers for greater efficiency. However, there is an operating systems incompatibility with the required channel programs. For this reason, the default number of input buffers is set to "1" (Output buffers are not subject to the problem). In release 1.5D, the default was incorrectly changed to "2".

This fix returned the default number of input buffers to its previous value of "1".

Please note: This problem can be easily circumvented by adding "INBUFFERS=1" to the "DEFINE LINK,TYPE=CLAW" command.

Download Fix File

Attempting to load the DBCS portion of a double-byte characters set returns an error message.

This fix corrects the problem AFFECTS PHASE=IPNET ALTER 08350A 1004:2008 ALTER 083514 1004:2008

Download Fix File

The message formatting routine is using an incorrect offset when displaying less that a full-word binary value. This fix corrects the problem.

Download Fix File

During shutdown, an SVC 25 is being issued while in 31-bit amode. This fix ensures 24-bit mode prior to issuing the SVC

This problem only affects those using an integrated CETI Ethernet adapter.

If you need this fix, it will be apparent the first time you shut down TCP/IP.

Download Fix File

Issuance of the pending expiration message destroys pointer to control area.

This fix preserves the pointer.

Download Fix File

During shutdown, an SVC 25 may be issued while in 31-bit amode. This fix ensures 24-bit mode prior to issuing the SVC

This fix only affects those using a 3172 controller, or equivalent.

Download Fix File

The DUMP processor is using an incorrect offset to obtain the next entry on a control block chain.

This fix corrects the problem.

Download Fix File

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

• This fix replaces ZP15D202, which included an "old" version of an IBM-provided CSECT in $EDCTCPV.

Download Fix File

For reasons that are not clear, inbound socket requests may include a parametric SVA subpool ID that ends in "00" rather than the ID= value of the destination stack. If additional socket blocks or data areas are required to process, the request, they are obtained from the storage pool "owned" by the stack with ID=00.

This causes no problem unless and until the "00" stack is terminated. During shutdown, each stack frees its own SVA-based storage. Due to the incorrect allocation, stack "00" will release some storage belonging to other stacks. This will then cause applications associated with the other stacks to lose their connections.

Additionally, if another partition obtains SVA storage following the shutdown of stack 00 and before the remaining stacks are terminated, then that storage may be improperly freed or corrupted by residual socket processing.

This fix causes the parametric specification to be ignored and the correct subpool to always be used.

Applications need NOT be recompiled or relinked for this fix to take effect, although all stack partitions must be recycled.

Download Fix File

Idle connections are probed at intervals to ensure that the remote host is still active and available. This is done by retransmitting a byte of data, requiring the remote stack to resend an ACK.

When setting up the probe, the IP headers fields are not always being cleared. Since most fields are explicitly set, this is seldom a problem. However, if any residual data appears in the "fragmentation" flags or offset, this may inhibit the remote host from sending the required ACK.

This fix ensures that all IP header fields are cleared at the start of processing.

Download Fix File

A problem has been found in the method used for computing checksums. In very rare instances, the result will be incorrect. Although the odds of this happening are only 1 in (1 gig / length of datagram). However, due to the number of times that checksums are computed, the cummulative effect may be noticeable.

We recommend that this fix be applied by everyone.

Download Fix File

Error and reason codes from failing VTAM RPL-based requests are being incorrectly reported in TEL972E, TEL970E, TEL974E, and TEL976E.

This fix corrects the display.

Download Fix File

The "IPT317I Response time-out resetting connection" message is documented as being "warning level" but is being issued as "informational". Installations that direct TCP/IP informational messages to SYSLST only are incorrectly prevented from seeing IPT317 messages on the console.

This fix changes the message to "warning level".

• Installations that do not wish this message to be displayed may use the MESSAGE command to suppress it specifically.

Download Fix File

The two optional values are incorrectly stored. They are being scaled by a factor of 32,768.

This fix corrects the problem.

Download Fix File

If you specify the optional IP address when defining a Daemon, incoming requests are tested against this address and are assigned to the Daemon only if there is a match.

In addition to an exact match, generic matching is performed, using "zero-host" notation, in conjunction with a "subnet mask".

The problem occurs when multiple subnet masks are used (DEFINE MASK). Current processing only considers the single mask specified via the SET MASK command.

This fix causes the correct mask to be used, based on network number.

Download Fix File

If an application issues a socket request using the same ECB area used to close a previous connection, it may be rejected as a "duplicate".

This fix prevents the "duplicate" test from being applied in this instance. This will not cause other problems because the "old" ECB is already marked as invalid and will never be used.

Another scenario occurs when a program running in a batch partition terminates without properly closing its sockets. If the same program is started in the same partition, it is possible that the same storage addresses will be used for socket control areas and a "duplicate ECB" conflict will arise.

This fix causes TCP/IP to compare the "job start time" of the new request against that of the existing request. If this value is not identical, then it indicates that the program that owned the original request has terminated. The ECB address passed with the new request is considered unique and the "duplicate ECB" failure is bypassed. Note that the "old" request will not cause problems because "phase name" and "job start time" are always validated prior to using any external data areas. Additionally, normal cleanup operations detect and eliminate these orphan control blocks at a nominal interval of 45 seconds.

Download Fix File

TCP/IP for VSE-based Daemons and applications perform passive OPENs to establish "listen" connections. External clients then issue active OPENs to link with a "listen" connection, thus establishing a working connection.

Normally, an external stack will send an initial SYN packet, indicating a connection request. This request is ACKed and a SYN is returned. When this SYN is acknowledged by the remote stack, the connection is established.

The problem occurs when the connecting stack transmits duplicate SYN packets in rapid succession.

Like all TCP datagrams, the initial SYN must be acknowledged in a timely fashion. If no ACK is forthcoming, the packet is retransmitted.

The makers of at least one largely-used stack have, apparently, begun transmitting multiple SYN packets, without waiting for and ACK. We conjecture that this is to allow for packet-loss on the network.

Once a connection is established, duplicate datagrams are not a problem and are disposed of properly. However, when two or more SYN requests arrive in the same physical data block (from a network adapter), they can each be assigned to a separate listen connection, although only the first connection reaches "established" status. Following connection CLOSE, the "extra" connection control blocks should fall into error recovery and reset to a "listen" state.

This fix allows detection of a duplicate SYN and causes it to be queued to the same connection as the original.

Download Fix File

This fix removes PDUMPS from a variety of phases. This will not affect PDUMPS that are produces as part of STXIT processing or via an operator DUMP command.

Download Fix File

This fix removes the test for a file name that begins with a period.

Download Fix File

API socket processing for both the COBOL and PL/1 interfaces contain an error that results in socket blocks being obtained in an SVA subpool assigned to the stack with ID=00.

During shutdown, stack "00" releases its entire storage subpool, causing socket requests queued to other stacks to fail.

This fix causes the correct storage subpools to be used, based upon the destination stack's ID.

• To make this fix effective, it will be necessary to relink all phases containing COBOL or PL/1 API socket requests.
• Also see fix ZP15D111 that provides a similar correction for applications using the BSD/C and assembler API's.
• This fix also incorporates an "eyecatcher" that consists of the module name, TCP/IP version/service pack, assembly date and assembly time. This fix will contain the date: "02/02/05"

Download Fix File

The file I/O driver for the VSAM catalog will sometimes attempt an improper storage release.

This appears to occur following a failed directory list request.

This fix ensures that, following a failed directory list request, a storage pointer is cleared prior to invoking the cleanup routine.

Download Fix File

If an empty variable is assigned to another variable or used as part of a string being assigned to a variable, the assignment will fail and storage corruption may occur.

This fix causes an empty variable to function as a null placeholder.

Download Fix File

This fix is applicable only if you password-protect operator functions.

When SET PASSWORD= has been used to establish a password, the operator is prompted for the password before a command can be entered. However, if the command is entered via the DATA= parameter of the MSG command, the prompt is skipped.

This fix causes the password prompt to be enforced.

Download Fix File

TCP/IP incorporates a test for data buffers to prevent accidental writing into the supervisor. This test is based on the knowledge that all data being passed externally is via 31-bit SVA.

The problem occurs when 31-bit SVA is either exhausted or fragmented and VSE returns 24-bit SVA for a 31-bit request. The storage address then later fails the "reasonable address" test.

This fix relaxes the validity checking to allow use of 24-bit SVA.

Download Fix File

The "expected ACK" value is being set to the wrong value.

This fix sets the "expected ACK" value the to sequence number of the retransmitted (Pulse) byte.

Download Fix File

The NL character (X'15') is being used to determine the end of the command line. However, the parsed commands may contain binary fields with an embedded X'15' byte.

This fix causes the command interface to use the entire passed data string rather than its "parsed" length.

Download Fix File

Because LPR streams may originate from a wide variety of non-VSE hosts, the LP Daemon attempts to ensure that a meaningful 8-character POWER name is assigned to the file. If the LPR client does not provide a suitable name, then the LPD generates a unique value and assigns it.

The problem occurs because the LPD does not have enough latitude in its name edit routine. POWER permits a wide range of characters --including non-printable characters-- in file names.

This fix removes the file name edit check.

• The POWER manual cautions against using non-displayable characters in file names. It may prove difficult to enter commands for specific files if the name is not permissible in an operator command.

Download Fix File

Due to changes designed to speed CLOSE processing, the return code for CLOSE is not initialized when WAIT(YES) is specified.

This fix retrofits a correction made in Sevice Pack E.

• After installing this fix it is necessary to recompile and relink the affected programs.

Download Fix File

Normal CLOSE processing by TELNETD includes waiting for the remote client's stack to respond with a FIN packet. Although this normally occurs within a few milliseconds, there is no protocol requirement.

Since the close response may take significant time and the TN3270 protocol does not include post-CLOSE error recovery, this fix eliminates the WAIT and immediately makes the Daemon available for another connection (CLOSE processing continues and is completed by the stack without any further involvement by the Telnet Daemon).

This fix also removes a redundant CLOSE. When a CLOSE is issued against an already-closed connection, an extraneous TCP/IP error message is issued.

Download Fix File

On a "busy" system (many concurrent socket requests), it is possible for the stack to improperly dispose of a socket block prior to its conclusion. This allows the possibility of overlap with the next socket request.

This error has two thresholds: At the most common level --which is still rare-- a recoverable chaining error occurs. This will usually result in a message referring to a "SOSBNEXT" or "socket queue" error.

These conditions are automatically diagnosed and corrected.

At the second threshold, data corruption can occur. This can take the form of missing, extra, or replaced data on a connection. We have very few reports of this condition, since several unlikely events must occur simultaneously.

This fix ensures that all required information is extracted from the socket block prior to its release for reuse.

Download Fix File

At regular intervals, TCP/IP examines connections to determine if they are still viable. The problem occurs when an application (eg, FTPBATCH) specifies a time limit on a connection and then does not receive sufficient CPU time to make the connection appear to be active.

This fix eliminates the test for "stalled" connections and allows as much time as required.

Connections will still be terminated if the remote host quits responding or if the batch program goes to EOJ.

Download Fix File

The Automation Daemon is only passing a 2-digit POWER segment number to the LPR client.

This fix ensures that three digits are passed.

Download Fix File

Service Pack E changes the Socket Interface linkage to be ESA compliant. Specifically, the external interface uses BASSM / BSM instructions instead of BALR / BR.

This fix allows the Socket Interface to determine the linkage technique being used and act accordingly. This permits applications assembled, compiled, or linkedited with Service Pack E macros and routines to execute with a 1.5D-level stack.

Download Fix File

When an attempt is made to access a non-existant POWER LST queue entry, a temporary work area is left allocated.

This fix ensures that the work area is always released.

Download Fix File

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

• THIS FIX HAS BEEN SUPERCEDED BY ZP15D109, which corrects a problem with $EDCTCPV.PHASE.
• The replacement fix can be applied on top of, or instead of, ZP15D202.
• If you have already applied ZP15D202 AND relinked your BSD/C routines, you skip the relinks after applying ZP15D109.

Download Fix File

The put of a empty VSAM KSDS dataset should be allowed when the ZEROERR=NO option, which is the default setting, is used.

Download Fix File

AutoFTP should be able to ignore an error and execute the next FTP command, but the "error occurred" flag was not being reset.

This fix corrects the problem.

Download Fix File

A protection exception may occur when invoking ftpbatch from a IBM REXX program because of access register 2 not being cleared.

Download Fix File

This fix removes a PDUMP that occurrs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

Download Fix File

This failure occurs when FTPBATCH is invoked more than 255 times from a single REXX programm.

This fix will correct the problem.

Note: Fix ZP15D209 MUST also be applied following application of this fix.

Download Fix File

Use of an empty (null) variable can casue an ABEND to occur. This includes POWER-provided variables that are not defined for the listing being processed.

This fix will cause an empty variable to function as a null string. // EXEC MSHP,SIZE=900K

Download Fix File

When FTPBATCH is all done it releases storage all allocated partition getvis storage. This is meant to allow REXX to invoke ftpbatch multiple times without storage increasing, but is not necessary if ftpbatch is not invoked from Rexx. This zap removes the cleanup storage problem that was causing a abend. This zap also corrects a problem caused by ZP15D207 where FTPBATCH failes with the following message: FTP326E VSE extract information failed rs=PRODAUTH

Download Fix File

Note: This is a cumulative fix for the BSD/C API interface. It supercedes all previous fixes for the BSD/C API. To complete installation of the fix, you will need to relink BSD/C applications with the included object decks.

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix also corrects the problem of storage not being freed following a failed connection request.

This fix removes a PDUMP that occurrs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

This fix also corrects a problem where TakeSocket and an asynchronous I/O write both return the same socket number for a newly-created connection.

Download Fix File

This fix will prevent a very rare abend that can occur in a batch partition executing the FTP phase.

Download Fix File

This fix directs additional diagnostic messages to the console following an FTPBATCH error.

Download Fix File

This is a cumulative fix for the BSD/C API interface. It supercedes all previous fixes for the BSD/C API. To complete installation of the fix, you will need to relink BSD/C applications with the included object decks.

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix also corrects the problem of storage not being freed following a failed connection request.

This fix removes a PDUMP that occurs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

This fix also corrects a problem where TakeSocket and an asynchronous I/O write both return the same socket number for a newly-created connection.

This corrects a problem where Socket allocation does not recognize CICS/TS, which can cause a duplicate socket number to be allocate.

This corrects a problem where a non-blocking SEND fails with an "ewouldblock" errno. Although this behavior is correct, the documentation currently states than non-blocking only applies to connects and, in fact, worked that way prior to 1.5D. For compatability, a $SOCKOPT option must be set to allow non-blocking SENDs.

Download Fix File

The BSD/C GiveSocket(_) and TakeSocket(_) functions allow sockets to be passed between routines. However, a "null receive" remains in effect and can cause a "duplicate ECB" error.

This fix causes the unneeded "null receive" to be canceled.

Download Fix File

FTPBATCH should terminate with a non-zero return could when SYSIPT commands are flushed. This zap will force a non-zero return code when commands are flushed.

Download Fix File

This fix changes the path taken by this specific method and allows exploitation of OPEN/CLOSE efficiencies used by other calling interfaces.

Download Fix File

Note: This is a cumulative fix for the BSD/C API interface. It supercedes all previous fixes for the BSD/C API. To complete installation of the fix, you will need to relink BSD/C applications with the included object decks.

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix corrects the problem of storage not being freed following a failed connection request.

This fix removes a PDUMP that occurrs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

This fix also corrects a problem where TakeSocket and an asynchronous I/O write both return the same socket number for a newly-created connection.

This corrects a problem where Socket allocation does not recognize CICS/TS, which can cause a duplicate socket number to be allocate.

This corrects a problem where a send on a non-blocking fails with an EWouldBlock errno. Although this behavior is correct, the documentation currently states than non-blocking only applies to connects, and worked that way prior to 1.5D. For compatability, a $SOCKOPT option must now be set to allow non-blocking for sends.

This corrects a problem where an Abend occurs during a socket call in a non-CICS partition.

Download Fix File

This zap will allow a maximum of 100 characters on a single line for a foreign directory list. This restores previous behavior and recognizes that directory lines are constructed to meet an arbitrary 100 character maximum imposed by some GUI clients.

Download Fix File

Note: This is a cumulative fix for the BSD/C API interface. It supercedes all previous fixes for the BSD/C API. To complete installation of the fix, you will need to relink BSD/C applications with the included object decks.

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix corrects the problem of storage not being freed following a failed connection request.

This fix removes a PDUMP that occurrs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

This fix also corrects a problem where TakeSocket and an asynchronous I/O write both return the same socket number for a newly-created connection.

This corrects a problem where Socket allocation does not recognize CICS/TS, which can cause a duplicate socket number to be allocate.

This corrects a problem where a non-blocking SEND fails with an EWouldBlock errno. Although this behavior is correct, the documentation currently states than non-blocking only applies to connects, and worked that way prior to 1.5D. For compatibility, a $SOCKOPT option must now be set to allow non-blocking for SENDs.

This corrects a problem where an abend occurs during a socket call in a non-CICS partition.

This corrects a problem where an abend occurs in DFHSOCK during CICS startup.

Download Fix File

If the name of the previously-specified FCB contains more than 5 characters, then "*NULL" is not recognized and an attempt is made to load an FCB of name "*NULLxxx", where "xxx" is taken from the previous FCB name.

This fix ensures that *NULL is treated as a 5-byte special word.

Download Fix File

If the SITE REXX command is passed to the FTP demon, it will load the REXX program, execute it, receive the data, and will be in the wrong AMODE for returning the data to the terminal and so it will blow up.

This fix ensures that the correct AMODE is in place.

Download Fix File

When sending a datagram that is less than 1K and not on a 256/byte boundary, the REXX program fails.

This fix corrects the problem.

Download Fix File

This fix has been withdrawn.

Download Fix File

Since messages produced with the "WTO" site command carry an internal attribute of "informational", use of the "MESSAGE INFO=PRINTER" command affects their display.

This fix changes the display attribute from "informational" to "command response".

Download Fix File

The default setting of BSDCFG1 in the SOCKOPT macro should be $OPTMECB. If $SOCKOPT is reassembled and BSDCFG1 flag is not specified, the default option of $OPTMECB will not be set.

Download Fix File

FTP BIM-EDIT interface was only showing members with security of LIST, EDIT, DEF and DEFS. With this zap, it will show everything BIM-EDIT returns allowing BIM-EDIT to control what the user sees.

Download Fix File

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix also corrects the problem of storage not being freed following a failed connection request.

Download Fix File

The security exit is not called for the last part of a directory change (CD). For example, CD /PRD1/BASE, the security exit would only see PRD1. With this fix, the exit will be called for the last directory with a type value of x'1A' (SXTYCWDL).

Download Fix File

Note: This is a cumulative fix for the BSD/C API interface. It supercedes all previous fixes for the BSD/C API. To complete installation of the fix, you will need to relink BSD/C applications with the included object decks.

SELECTEX processing limits an application to a maximum of 64 ECBs. This fix increases the limit to 512.

This fix also corrects the problem of storage not being freed following a failed connection request.

This fix removes a PDUMP that occurrs when a BSD/C TakeSocket call detects a non-connected socket after a successful GiveSocket.

This fix ensures that the correct return code is set when a socket request fails during TCP/IP shutdown. VSE/POWER uses the BSD/C API and may go into a message loop when TCP/IP is shutdown.

This page and all materials residing on this site Copyright 1996, 2012 by CSI International. You may read, download, and copy these materials only for the purpose of evaluating or publicizing CSI International, its services and products or in accordance with a license agreement with CSI International.