HP TCP/IP Services for OpenVMS
Release Notes

4.6.16 Passive mode on a multihomed system

Problem:

When using passive mode on a multihomed system, the FTP client fails to ensure that the source IP address for the data connection matches the IP address used for the control connection. Many FTP servers reject such connections for security reasons.

Solution:

This problem is corrected in this release.

4.6.17 Sends the incorrect file version

Problem:

The FTP server sends the incorrect file version in the 150 info message to the FTP client.

Solution:

This problem is corrected in this release.

4.6.18 Display of files residing on second and subsequent disks

Problem:

When a DIRECTORY command is executed on a search list pointed to by a concealed logical, the list contains information about files only on the first disk and fails to display the files residing on second and subsequent disks.

Solution:

This problem is corrected in this release.

4.6.19 Transferring files greater than 2GB

Problem:

While transferring huge files, greater than 2GB, from a disk, all the other operations on this disk will hang until the transfer of file is accomplished.

Solution:

This problem is corrected in this release.

4.7 IMAP problems fixed in this release

The following sections describe IMAP problems fixed in this release.

4.7.1 IMAP server allows potential attackers

Problem:

IMAP server allows potential attackers with unlimited guess of username and password combinations.

Solution:

This problem is fixed in this release.

4.7.2 Listing of more than hundred empty folders fails

Problem:

The IMAP server crashes while listing more than hundred empty folders.

Solution:

This problem is corrected in this release.

4.7.3 IMAP server process hang in the exception handler

Problem:

An IMAP server process may hang in the exception handler.

Solution:

This problem is corrected in this release.

4.8 INETDRIVER problems fixed in this release

The following sections describe INETDRIVER problems fixed in this release.

4.8.1 System crash in the KVCI$$GENERATE_ASSOC_ID routine

Problem:

Users of the SRI QIO interface (INETDRIVER) experience a system crash with INVEXCEPTN in the KVCI$$GENERATE_ASSOC_ID routine.

Solution:

This problem is corrected in this release.

4.9 IPC (socket library) problems fixed in this release

The following sections describe IPC (socket library) problems fixed in this release.

4.9.1 TCPIP$INETACP process uses 100% CPU

Problem:

In a multithreaded customer application in which thousands of threads call select(), TCPIP$INETACP process uses 100% CPU.

Solution:

This problem is fixed in this release.

4.9.2 Alignment faults in TCPIP$ACCESS_SHR.EXE image

Problem:

Alignment faults are observed in TCPIP$ACCESS_SHR.EXE image. The most common PC range is below:
TCPIP$ACCESS_SHR + 54230
TCPIP$ACCESS_SHR + 54264

Solution:

This problem is fixed in this release.

4.9.3 Definitions for TCP socket

Problem:

Some of the "definitions" are not available for certain TCP socket options in SYS$SHARE:TCPIP$INETDEF.*.

Solution:

The following definitions are added with this release of TCP/IP:

#DEFINE INET$C_TCP_TSOPTENA 16 /* time stamp option */ #DEFINE INET$C_TCP_PAWS 32 /* PAWS option */ #DEFINE INET$C_TCP_SACKENA 64 /* SACK enabled */

Counterparts with the TCPIP$ prefix used instead of the INET$ prefix are also added.

4.9.4 getnameinfo( ) returns "unknown name or service" error

Problem:

The getnameinfo() function returns an unknown name or service error if the specified address is not found. The RFC defines that getnameinfo() must return the address. The routine also fails to honor the NI_NAMEREQD or NI_NOFQDN flags in all cases.

Solution:

This problem is corrected in this release.

4.9.5 freeaddrinfo( ) causes an ACCVIO

Problem:

freeaddrinfo() causes an ACCVIO condition when a NULL pointer is passed for freeaddrinfo() .

Solution:

This problem is corrected in this release.

4.9.6 IPv6 address queried before IPv4 address

Problem:

The BIND9 Resolver sends queries for IPv6 addresses before querying for IPv4 addresses, even when no local IPv6 addresses are configured.

Solution:

This problem is corrected in this release.

4.9.7 BIND9 Resolver flags for getaddrinfo are inadvertently shifted

Problem:

The BIND9 Resolver AI_ALL and AI_V4MAPPED flags for getaddrinfo are inadvertently shifted, preventing the IPv6 application build against the previous versions of TCPIP from working on TCPIP V5.6.

Solution:

Because the previous flag values are restored, some IPv6 applications built for the original TCPIP V5.6 release will no longer function correctly following the installation of this kit.

The relevant header file is netdb.h . Application developers having trouble with these flags must ensure that they are using a "netdb.h" file with the old (and recently restored) values.

4.9.8 Delay when communicating between socket pair

Problem:

The socketpair() call returns a pair of TCP sockets, which are connected through a localhost ephemeral port numbers. When communicating between this socket pair, a 200ms delay is encountered while receiving and acknowledging the data.

Solution:

This problem is corrected in this release.

4.9.9 Alignment faults in gethostbyname()

Problem:

Alignment faults are detected in the gethostbyname() call and friends.

Solution:

This problem is corrected in this release.

4.9.10 Documentation for getaddrinfo() and gai_strerror() - EAI_BADHINTS

Problem:

Along with the other getaddrinfo() error codes documented in the HP TCP/IP Services for OpenVMS Sockets API and System Services Programming guide, getaddrinfo() may also return:

EAI_BADHINTS "Invalid value for hints"

This error is returned if the hints parameter in the call to getaddrinfo() is not correctly initialized.

Solution:

This problem is corrected in this release.

4.10 Load Broker problems fixed in this release

The following section describes Load Broker problems fixed in this release.

4.10.1 Load Broker memory leak

Problem:

If you are running several OpenVMS Clusters and using several dynamic cluster aliases for balancing the workload across the cluster members with each load broker maintaining 10 cluster aliases, after few days of operation the load broker dies with the %SYSTEM-F-OPCCUS error.

Solution:

This problem is corrected in this release.

4.11 LPD problems fixed in this release

The following sections describe LPD problems fixed in this release.

4.11.1 Incorrect job status in the mail message

Problem:

LPD printing with the /PARAMETERS=MAIL qualifier includes an incorrect job status in the resulting mail message.

Solution:

This problem is corrected in this release.

4.11.2 Printing to an LPD queue with a large setup module is inefficient

Problem:

Printing to an LPD queue with a large (over 1024 characters) setup module is inefficient. Although, correct output is printed, the logfile shows that for each job, there is a series of attempts to read the setup module into increasingly large buffers.

Solution:

A new configuration parameter, "Setup-Buffer-Size", in the TCPIP$LPD.CONF file allows the system manager to specify the initial setup module buffer size. The default value is 1024 bytes.

4.11.3 "TCPIP-E-LPD_REQREJECT" message displayed multiple times

Problem:

The TCPIP-E-LPD_REQREJECT message is displayed many times when attempting to deliver LPD jobs to a printer that is not in service.

Solution:

This problem is corrected in this release.

4.11.4 Latent coding defect within the LPD symbiont

Problem:

A latent coding defect within the LPD symbiont led the symbiont to exit with an ACCVIO error after the VMS83A_RMS V8.0 (or later) patch was installed on an OpenVMS 8.3 Alpha system.

Solution:

LPD users should install the latest TCP/IP kit along with the RMS V8.0 or V9.0 patch.

4.12 Management Utilities problems fixed in this release

The following sections describe Management Utilities problems fixed in this release.

4.12.1 TCPIP$CONFIG does not create an alias IP address

Problem:

TCPIP$CONFIG does not create an alias IP address, which is a substring of the primary address. For example, if the primary address is 10.1.1.100, then it is not possible to add an alias address 10.1.1.10.

Solution:

This problem is fixed in this release.

4.12.2 Large number of packets are sent when using the flood functionality

Problem:

In some instances, a large number of packets are sent, when using the flood functionality of the PING utility (-f) in combination with the option for sending fixed number of packets(-c).

Solution:

This problem is fixed in this release.

4.12.3 netstat -i fails to display the network names correctly

Problem:

netstat -i fails to display the network names correctly.

Solution:

This problem is corrected in this release.

4.12.4 Misleading and unsightly error message when the BIND resolver is not enabled

Problem:

Attempting to use the "dig" utility results in a misleading and unsightly error message when the BIND resolver is not enabled.

Solution:

This problem is corrected in this release.

4.12.5 TCPIP$CONFIG.COM fails to see devices

Problem:

TCPIP$CONFIG.COM fails to see devices when the controller letter does not begin with "A". For example, if EIB exists but EIA does not, then the EI controller does not appear in the Interface menu.

Solution:

This problem is corrected in this release.

4.12.6 Missing argument for the ip6hoplimit value

Problem:

Executing the $ IFCONFIG WE0 INET6 IP6HOPLIMIT command results in an ACCVIO because of the missing argument for the ip6hoplimit value.

Solution:

This problem is corrected in this release.

4.12.7 Errors when executing netstat -z

Problem:

When executing netstat -z , the following message is displayed:

netstat: -z is not implemented on this operating system

Solution:

Netstat will now zero the counters. In addition, if you attempt to use the -z option without privileges, netstat will no longer attempt to display the counters, but rather displays the following message:

netstat: must be root to zero counters

4.13 NET (Kernel) problems fixed in this release

The following sections describe NET (Kernel) problems fixed in this release.

4.13.1 TCP/IP routine that services I/O CANCEL and DEASSIGN requests does not restore the entry IPL

Problem:

Some processes, such as CIMSERVER, were found hanging in the RWINS state. This happened because the TCP/IP routine that services I/O CANCEL and DEASSIGN for the BG devices was not restoring the Interrupt Priority Level (IPL) to the entry IPL before returning.

Solution:

The entry IPL is now saved on the stack at the beginning and restored from the stack before returning.

4.13.2 Entering the username and password in binary mode

Problem:

When a user enters ((Ctrl+U) and username) followed by ((Ctrl+U) and password) to the telnet server in binary mode, the user authorization fails.

Note

A new logical TCPIP$TELNET_BINARY_IGNORE is defined, which when enabled on binary negotiation, will not set the TT$M_PASSALL bit in the terminal characteristics.

Solution:

This problem is fixed in this release.

4.13.3 TELNET server does not accept new connections

Problem:

Occasionally, the TELNET server does not accept new connections whereas other services such as FTP and SSH appear to accept new connections.

Solution:

This problem is fixed in this release.

4.13.4 RLogin fails

Problem:

Rlogin to a remote system crashes and displays the following message:

SSRVEXCEPT, Unexpected system service exception

Solution:

This problem is corrected in this release.

4.13.5 Corruption of non-paged pool

Problem:

Various system crashes are reported, involving corruption of non-paged pool.

Solution:

This problem is corrected in this release.

4.13.6 SACK retransmission transmits more data

Problem:

SACK retransmission resulted in too much data being retransmitted; that is, it retransmitted beyond the SACK Left Edge, SLE.

Solution:

This problem is corrected in this release.

4.13.7 Fail to sense SHARE and FULL_DUPLEX_CLOSE

Problem:

While it was possible to set the socket options, it is not possible to sense SHARE and FULL_DUPLEX_CLOSE.

Solution:

This problem is corrected in this release.

4.13.8 System crash after failing to start TCPIP

Problem:

The system crashes after a mysterious failure to start TCPIP, displaying the SPLIPLHIGH error.

Solution:

This problem is corrected in this release.

4.13.9 Setting the inet sysconfig parameter may cause a crash

Problem:

Setting the inet sysconfig parameter, ovms_printf_to_opcom may cause a crash at TCPIP start, on any version of the scaling kernel. The crash happens if the startup code attempts to print something, for example:

sysconfigtab: attribute sobacklog_hiwat in subsystem socket can't be configured

Solution:

This problem is corrected in this release.

4.13.10 System crash because of coded bugcheck in m_copym( )

Problem:

A system crash occurs due to a coded bugcheck in the m_copym() routine because an unexpected negative offset is calculated during selective acknowledgement (SACK) processing.

Solution:

This problem is corrected in this release.

4.13.11 System crash while processing select()

Problem:

A system crash within the TCPIP$INTERNET_SERVICES execlet occurs while processing a select() call.

Solution:

This problem is corrected in this release.

4.13.12 System crash during Packet loss and SACK processing

Problem:

The system occasionally crashed or created an ACK storm during some circumstances involving packet loss and SACK processing.

Solution:

This problem is corrected in this release.

4.13.13 Impossible to disable error message display

Problem:

It is not possible to disable certain error messages displayed via OPCOM or directly to the operator console.

Solution:

Messages such as the following:

arp: local IP address nn.nn.nn.nn in use by hardware address mm-mm-mm-mm-mm-mm

can now be displayed in the following ways:

$ sysconfig -r inet ovms_printf_to_opcom=1 ! On OPCOM $ sysconfig -r inet ovms_printf_to_opcom=0 ! On OPA0: $ sysconfig -r inet log_open=1 ! No display

In the future, the final setting may send messages to the SYSLOG facility if and when it is implemented.

4.13.14 System crash during a select() operation

Problem:

The system crashes during a select() operation, or immediately after the operation is complete.

Solution:

This problem is corrected in this release.

4.13.15 Debug code to verify MBAG free list

Problem:

Add debug code to verify MBAG free list during get and free .

Solution:

This problem is corrected in this release.

4.13.16 Process in RWAST state during process rundown

Problem:

The Process goes into RWAST state during process rundown.

Solution:

This problem is corrected in this release.

4.13.17 use of select() results in a non-paged pool memory leak

Problem:

Under certain conditions, use of the select() function results in a non-paged pool memory leak.

Solution:

This problem is corrected in this release.

4.13.18 Issuing process in the RWAST state

Problem:

A select() operation with certain parameters can cause the issuing process to enter the RWAST state.

Solution:

This problem is corrected in this release.

4.13.19 Multicast traffic can be lost

Problem:

Multicast traffic can be lost when aggressive IGMP snooping is enabled on a switch. This is the result of OpenVMS delaying IGMP reports when the IGMP query specified a maximum response time less than 10 seconds.

Solution:

This problem is corrected in this release.

4.13.20 Extensive use of Out Of Band data can cause system crash

Problem:

Extensive use of Out Of Band data by applications can trigger a system crash at offset PANIC_C+00330 (On V5.6-9, Integrity servers).

Solution:

This problem is corrected in this release.

4.13.21 INETACP process experiences a deadlock

Problem:

The INETACP process experiences a deadlock, frequently stuck in the RWAST state. The internal AQB (work queue) would be non-empty, with perhaps hundreds of outstanding requests.

Solution:

This problem is corrected in this release.