Changes/Fixes between FW 7.4.0.16.00 and FW 7.4.0.22.00

PR 113271

- When a QLogic QLE2562 HBA in a server running Windows Server 2003 is put in loop mode and connected to a SANbox 5802 switch, it gets "Decode Errors" and "Invalid CRCs".

- Was fixed by changing the SERDES training on the 8Gb loop to leave the SERDES in 'standby' mode when the loop is up and running to avoid problems with decode and CRC errors.

PR 114313

- Java 1.6u15-u16 crashes while QuickTools is coming up.

- Was fixed by implementing a single large jar file for QuickTools to work around the Java issue.

PR 115483

- SANbox 5602 switch with FW 7.4.0.16 does not respond to Broadcast Ping. This is different behavior than FW 5.x because the default behavior changed for Linux between the kernel versions associated with FW 5.x and FW 7.4.x.

- Was fixed by adding support to the CLI (and associated applications) to show and allow the setting of the IPv4IgnBcastPing config value. Requires the user to be in advanced mode to show/set the value.

PR 115649

- When an SBOD drive is pulled from the storage enclosure and re-inserted, the fabric switch may down the switch port with the message … [Port with device violating frame source identification addressing is DOWNED.] . This occurs because nameserver advertises a device before the FLOGI ACC sent.

- Was fixed by delaying the addition of the device to the database until the FLOGI ACC is sent.

PR 112804/115392

- Snmpmain service is stopping unexpectedly.

- Was fixed by changing the snmp libraries to detect when they are using a stale, previously deallocated session structure when processing a reply message.

PR 116314

- Snmpmain (pid 24272) died from signal 11.

- Was resolved by restarting the SNMP daemon when the snmpmain process dies. It can be restarted without detrimental effect to the rest of the switch. Also, we now simply report an error and restart the SNMP daemons instead of faulting the switch or just leaving things in a bad state (depending on platform). This approach was already used for the snmpd process on all platforms; we have just extended it to include the snmpmain process.

PR 116314

- Snmpmain (pid 24272) died from signal 11. The SNMP "set" procedure consists of many steps; for each step the subagent gets a message from the agent. The memory leak occurs because sometimes the subagent gets the 'first step' message from the agent twice; when it receives this message, it creates a record of the request and puts it on a list. Later, it removes only the one record it expects, leaving the second there until the end of process.

- Was fixed by scanning the list for an already existing record (identified by session and transaction id) when creating a new record, and if one is found, do not create a duplicate record.

PR 117001

- /itasca/bin/ens (pid 796) died from signal 6] resulting in internal system error. The issue was caused by misconfiguring NTP to point at a non-existent server which resulted in a slow memory leak.

- Was fixed by closing a window where we can lose flow credit if the receiver of a message exits after the message is dequeued for delivery, but before the message can be copied to user space.

PR 117113

- Time of the SANbox 5802 switch is incorrectly shifted 1 hour when set to Tokyo time because old timezone rules are filtered by our tools to shrink the size of the timezones.zic file.

- Was fixed by keeping the default rule for time zone ones like Asia/Tokyo that no longer use DST.

PR 117927

- Add partial support for TxWait counters.

- Currently fully implemented for box products

- These are available in advanced mode in clish

-  The changes made to add support of displaying TxWaits1, TxWaits2, TxWaits3 were incorrectly made and resulted in clish crashing during a 'show port <#>' display

-  Made the changes correct and modified the code so it no longer needs to be in sync between the two files (thus making it less error prone and improving maintainability).

PR 117787

-Zoning is crashing in Brocade Mode

-Remove the additional delete of the TransmitSequence.