MultiNet ECO kit announcement

The following ECO kit is now available for MultiNet:

ECO: KERNEL-UPDATE-090_A050
Description: Disable arp info OPCOM msgs; fix Telnet user ID flags problem
Release date: 11-MAY-2005
Ranking: 3
Max ranking: 1
Versions: V5.0A

ftp://ftp.multinet.process.com/patch...e-090_a050.zip

To search the MultiNet ECO database, please visit the following URL:

http://www.multinet.process.com/eco.html

For more information, contact Process Software via:

E-mail: support@process.com
Phone: 1-800-394-8700

The ECO kit README contents are below.

-----------------------------------------------------------
KERNEL-UPDATE-090_A050 ECO kit rev 9.0 For Multinet V5.0 Rev A 11-May-2005

Copyright (c) 2004, 2005, Process Software LLC

************** NOTE **************

KERNEL-UPDATE-021_A050 and later REQUIRES the following UCXDRIVER kit:
- for MultiNet v5.0: UCXDRIVER-012_A050 or later

**********************************

Overall ECO Rank 1 [KERNEL-UPDATE-070_A050]
Version 9.0 Rank 3

This ECO kit provides a new version of the MultiNet kernel,
MULTINET.EXE, for Multinet V5.0 Rev A. This kernel includes the
following changes:

KERNEL-UPDATE-090_A50 - ECO Rank 3
-----------------------------------
- Added the ability to disable "arp info overwritten..." OPCOM
messages by setting the kernel variable ARP_OVERWRITTEN_LOGS
to 0 (zero). (DE 9948)

- Fixed problem with /encr & /auth flags that were being added
arbitrarily to the end of the user id string on incoming telnet
connections. (DE 9881)


----------------------------------------

This kit also includes the following changes from previous ECO kits:

KERNEL-UPDATE-080_A50 - ECO Rank 3
-----------------------------------
- Add the kernel variables TCP_DO_NAGLE, TCP_USER_TTL, TCP_USER_TOS,
TCP_TOS, and TCP_TTL that were in previous versions of MultiNet to
MultiNet 5.0. TCP_DO_NAGLE can be set to 0 (zero) to disable the
Nagle algorithm (RFC 896) which can delay sending TCP data by up to
200 milliseconds in the hopes that multiple sends can be combined
into a single transmission. TCP_USER_TTL and TCP_USER_TOS can be
set to 0 (zero) to cause the value of TCP_TTL or TCP_TOS
(respectively) to be used for the TTL or TOS fields instead of the
normal values. (DE 9971)

KERNEL-UPDATE-070_A50 - ECO Rank 1
-----------------------------------
- Correct a potential system crash when flushing routes with
MULTINET SET/ROUTE/FLUSH. (DE 9950)

- Correct a synchronization problem that can cause a system crash
when setting a new address on a PD device that currently has an
address. (DE 9873)

KERNEL-UPDATE-063_A050 - ECO Rank 2
-----------------------------------
- Correct a problem with how connections in TIME_WAIT state are
processed that can cause other connections to get erroneously
deleted. (D/E 9936)

- Correct a problem with selects on sockets with non-blocking writes
completing before there is buffer space available. (D/E 9919)

- Correct a problem which allowed a bind to a port on a specific
address after one had already been done on the wildcard address.
If the process is binding a specific port, then processes that have
SYSPRV, OPER, BYPASS or a group less than or equal to EXE$GL_SYSUIC
can bind to ports that are already bound unless the MultiNet kernel
variable ALLOW_SUSER_BIND_HIJACK is set to 0 (zero) with:
MULTINET SET/KERNEL ALLOW_SUSER_BIND_HIJACK 0
Processes that are binding an unspecified port will always get a
port that is not bound to either the specific address or the
wildcard address. (D/E 9922)

KERNEL-UPDATE-050_A050 - ECO Rank 3
-----------------------------------
- Correct a problem with not retransmitting responses to connection
open requests when the acknowledgement of the response is not
received. (D/E 9801)

- Performance improvements.

- The ability to bypass most of the IPSEC code by setting the kernel
variable NO_IPSEC to a non-zero value. This can improve throughput
for sites that do not use IPSEC. If you desire to set this
variable, it should be done very soon after starting MultiNet
(MULTINET:LOCAL_ROUTES.COM is a good place to put it).
Once the variable has been set to a non-zero value you should NOT
set it to 0 (zero). Changing the variable from non-zero to zero
while the system is running may result in a system crash due to
expected data structures that are created at one point for a
connection being missing at later times.

- Provided a means of controlling the memory and I/O usage while
using DECNET-over-IP (Phase V). A new kernel variable
PWIP_IO_LIMIT has been provided. (D/E 7128)

- Allow read and write attention ASTs to be canceled from the
BG device. (D/E 9860)

KERNEL-UPDATE-040_A050 - ECO Rank 2
-----------------------------------
- Correct an error which would cause MultiNet to drop TCP packets
when the last part of the host IP address is between 224 and 239.
(D/E 9760)

KERNEL-UPDATE-030_A050 - ECO Rank 3
-----------------------------------
- Provide a method of controlling the range for ephemeral ports.
Use the commands:
MULTINET SET/KERNEL IPPORT_EPHEMERAL_MIN nnnn
to set the minimume value.
MULTINET SET/KERNEL IPPORT_USERRESERVED nnnn
to set the maximum value. The default values are 1025 and 5000,
which reserves ports between 1025 and 5000 as ephemeral ports. It
is recommended that this be done early in the MultiNet startup;
MULTINET:LOCAL_ROUTES.COM would be a good place to put the
command. (D/E 9753)

- Remove a debugging statement that was accidentily left in.
(D/E 9749)

KERNEL-UPDATE-021_A050 - ECO Rank 1
-----------------------------------
- Correct a problem where an attempt to perform a set or get
options on a socket that is mostly closed can cause a system
crash. (DE 9674)

- Corrects a problem where a $QIO with nowait, would return
an error status of SYSTEM-F-SHUT (0x208c) instead of
SYSTEM-F-SUSPENDED (0x03a4). (DE 9677)

- Corrected a potential crash for an ICMP message received over
an interface that does not have an IP address. (DE 9700)

KERNEL-UPDATE-010_A050 - ECO Rank 3
-----------------------------------
- The kernel keeps a list of pending data transactions to be sent
down to the interface. If that queue gets too large, the kernel
will drop a packet and retransmit it after a calculated retransmit
timer has expired. That queue has been lengthened to help prevent
an unnecessary dependence on retransmissions in high activity times.
(D/E 9626)

- Correct a problem where multiple processes can bind to the same
port when SO_REUSEADDR is set and INADDR_ANY is used
(D/E 9631) ECO Rank 3

************** NOTE **************

KERNEL-UPDATE-021_A050 and later REQUIRES the following UCXDRIVER kit:
- for MultiNet v5.0: UCXDRIVER-012_A050 or later

**********************************

You must reboot after installing this ECO kit for the changes to
take effect.


[End of ECO announcement]