==================================== Hewlett-Packard November 15, 2002 ==================================== Fabric OS V3.0.2k Patch Release Notes Patch Release Notes contain descriptions of bug fixes from release v3.0.2a to v3.0.2k =================================== BUGS FIXED IN FABRIC OS 3.0.2a =================================== Defect ID: 10109 Priority: 1 Problem Area: Zoning Found in Release: v3.0.2 Problem description: SID/DID CAMs are not setup properly when a HBA issues a PDISC instead of PLOGI LOGO frames do not arrive at PDISC source port Description of Fix: Zoning task rebuids the SID/DID CAMs when a PDISC is detected. =================================== BUGS FIXED IN FABRIC OS 3.0.2c =================================== Defect ID: 10504 Priority: 1 Problem Area: QuickLoop Found in Release: v3.0.1b Problem description: In a configuration with two hosts attached in 2Gbit\sec QuickLoop mode, when hba sends an OPN, it does not get forwarded to the storage device, it instead is sent back to the hba, resulting in the hba being unable to access storage. Description of Fix: Save bitmap of looplets that failed to initialize during first pass and use it to reinitialize them at the end of pass2, if timing deems this necessary. Defect ID: 10577 Priority: 3 Problem Area: SNMP Found in Release: v3.0.1b Problem description: E-port display in FA-MIB and trunking only shows Master Trunk Port. Description of Fix: swNbTable displayed the entries for master ports only. It now displays the entries for slave ports also. Defect ID: 10743 Priority: 1 Problem Area: QL Found in Release: v3.0.2 Problem description: Port becomes In_Sync when switch is rebooted with JNI FCE2-6410. This issue will only occur with the dual port HBA FCE2- 6410. Description of Fix: Under certain loop environment, firmware handles the OFFLINE scn twice by calling qlLoopletDown. It is now handle propetly. Defect ID: 11077 Priority: 3 Problem Area: Web Tools Found in Release: v2.4.1c Problem description: Hyphen in the switch node name in DNS disables access to Zone admin & Name server windows in WebTools. Description of Fix: Modified the HTTPd task to accept hyphen in the switch node name. Defect ID: 11295 Priority: 2 Problem Area: Command Found in Release: v3.0.2b Description of Fix: The command nsAliasShow has been added to this patch release. This command is a duplicate of the command nsShow with the added feature of displaying the defined configuration aliases that the device belongs to. Defect ID: 11638 Priority: 1 Problem Area: Name Server Found in Release: v3.0.2b Problem Area: NS was requiring another probing to be completed before sending RSCN. Description of Fix: If probing has failed with NOT_SUPPORTED as the reason code, probing will discontinue and the NS will send an RSCN once the device has been registered. =================================== BUGS FIXED IN FABRIC OS 3.0.2d =================================== Defect ID: 12129 Priority: 1 Problem Area: Frame Filtering Found in Release: v3.0.2c Problem description: In a 20 switch fabric VMS hosts loses path to storage targets after the storage has been rebooted. Description of Fix: Change to process Port login of Open VMS hosts can now maintain path to storage targets after storage has been rebooted. Defect ID: 12136 Priority: 3 Problem Area: SNMP Found in Release: Problem Description: CERT test suite case #1421 causes memory leak in switch. Patch by WindRiver. Description of Fix: Issue found by WindRiver when running CERT test #1421. WindRiver fixed the issue and provided a patch to Brocade. Defect ID: 12198 Priority: 1 Problem Area: Compatibility Found in Release: 3.0.2 Problem Description: Frames discarded during remote copy. Description of Fix: The root cause of this is the loop deadlock. In order to fix this problem, the switch will periodically checks the port for a loop idle condition where frames are not tx'd and rx'd for at least 16ms. At which point, the loop tenancy is broken by generating the CLS Defect ID: 12018 Priority: 1 Problem Area: QL Problem Description: Switch hangs when receiving unexpected LISM storms from device. Description of Fix: Fixed tReceive task to handle unexpected LISM storms correctly. Defect ID: 12027 Priority: 1 Problem Area: Zoning Problem Description: Port zoning all 4 ports on a VMS host does not see all ports from storage device. Description of Fix: Clear the appropriate CAMs at the same time the CAMs in the host is cleared. =================================== BUGS FIXED IN FABRIC OS 3.0.2f =================================== Defect ID: 12215 Priority: 1 Problem Area: Problem Description: Requesting bloomDataShow on an invalid port causes inconsistent behavior. Description of Fix: An error message 'Invalid Port' will be displayed if the port requested for bloomDataShow is out of range. Defect ID:: 12228 Priority: 2 Problem Area: Auto Negotiation Problem Description: Changing switch speed from AU to 1GB causes an HBA to go into loopback state Description of Fix: Changed timing between portenable and portdisable to prevent an HBA from going into loopback state when switch speed is changed from AU to 1GB. Defect ID: 12324 Priority: 2 Problem Area: Frame Filtering Problem Description: BA_ACC frame to an ABTS is being dropped. Description of Fix: BA_ACC frame is now handled properly when an ABTS is received. Defect ID: 12367 Priority: 1 Problem Area: Interop Problem Description: GA_NXT request times out from McData to Brocade. Description of Fix: Brocade switch returns next device in local name server. Defect ID: 12477 Priority: Problem Area: Quick Loop Description: LILP Frame did not contain all phantom ALPAs in the ports zone. This resulted in ports not seeing all devices in its zone. Description of Fix: Phantom map now includes all phantom ALPAs in ports zone. Defect ID: 12575 Priority: 1 Problem Area: Port Initialization Problem Description: Host does not recognize storage after reboot. Description of Fix: F_port setup has changed. Defect ID: 12879, 12739 Priority: 1 Problem Area: Zoning. Problem Description: When zoning is enabled, a cluster node cannot see it's storage device after reboot. Description of Fix: Cluster node can now see storage devices after reboot when zoning is enabled. Defect ID: 12588 Priority: 2 Problem Area: Fabric Watch Problem Description: tThad disable configure option is only available to root user Description of Fix: tThad disable configure option now is available at the Admin level. Defect ID: 13023 Priority: 2 Problem Area: Web Tools Problem Description: Buttons at the bottom of screen are not visible at 800x600 resolution, slider bar is unusable. Description of Fix: Buttons are now visible at 800x600 resolution and slider bar is fully accessible. =================================== BUGS FIXED IN FABRIC OS 3.0.2g =================================== Defect ID: 12330 Priority: 1 Problem Area: Zoning Problem Description: HDS MSCS linkup problem Description of Fix: Set TFR_SEARCH to force another timeout search to clear the timeout flag Defect ID: 12350 Priority: 1 Problem Area: Extended Fabric Problem Description: Switch reboot after adding extended fabric license causes problem with spinsilk Description of Fix: Spinsilk does not have problem anymore with this version Defect ID: 13043 Priority: 1 Problem Area: Fabric Watch Problem Description: Fabric Watch Warning is returning (NULL) when the timebase is modified from default Description of Fix: Correction was made to return the unit string to display the value "Error(s)" Defect ID: 13044 Priority: 2 Problem Area: Port configuration Problem Description: More than one port in a quad may be designated as a long distance port. Description of Fix: Once a port is set as long distance port, trying to designate another port in the same quad to long distance will generate error message Defect ID: 13045 Priority: 2 Problem Area: Zoning Problem Description: Zoning interface allows duplicate entries to be added to an alias Description of Fix: Duplicate entries will be detected and not added to the list Defect ID: 13303 Priority: 2 Problem Area: Diagnostics Problem Description: LOGO frame was dropped at the switch Description of Fix: Trap the LOGO frame also in PLOGI trap and forward it Defect ID: 13304 Priority: 2 Problem Area: QuickLoop Problem Description: Switch is dropping SCSI command frames when a Mylex controller does a failover Description of Fix: Continue creation of additional phantoms regardless of the error. During the LISA phase, detect the error and resolve the error condition Defect ID: 13326 Priority: 1 Problem Area: Fabric Watch Problem Description: fwConfigReload not working properly Description of Fix: Added ALPA, end-to-end and filter class configreload Defect ID: 13519 Priority: 0 Problem Area: Backward/Forward Compatibility Problem Description: During core PID upgrade in a redundant fabric, if the host mistakenly sends a PLOGI with bad D_ID, and if hardware zoning is configured, filtering logic will use the wrong port derived from the bad D_ID Description of Fix: Check the port attached before using it =================================== BUGS FIXED IN FABRIC OS 3.0.2h =================================== Defect ID: 11606 Priority 2 Problem Area: Web Tools Problem Description: v3.0.2x warning message is not consistent with 2.6.0x when enabling zoning configurations. Description of Fix: The following warning message in FOS 2.6.0x now exists in 3.0.2g as a reminder that when enabling it is only for the defined configuration: "You are about to save Zoning Configuration changes. This action will only save the defined configurations. If you have changed the Currently Enabled Config, the changes will not be in until you re-enable it. Do you want to save zoning changes only?" If changes have been made to the effective configuration it will nee to be re-enabled for those changes to be effective. Defect ID: 14220 Priority: 2 Problem Area: Zoning Problem Area: AL_PA's are being return in the LILP that belong to Ports that are in different zone. When two switches are QL partnered together and have two zones, the AL_PA's from the first zone are being reported in the LILP in the second zone. Fix: QuickLoop code was functioning as designed. Zoning code was not enforcing zoning configurations as defined. Zone configurations are now enforced. Defect ID: 13558 Priority: 2 Problem Area:Kernal Problem Description: Switch is not discarding invalid frames at destination at hardware speed. Description of Fix: Created function so invalid frames can be dropped at hardware speed. Defect ID: 15719 Priority: 1 Problem Area: QuickLoop Problem Description: Frame hold time was changed in 3.0.2d from 500ms to 200ms. As a result compatibility issues with certain storage devices have occurred. Description of Fix: Frame hold time was increased back to 500ms to ensure compatibility with storage devices. Defect ID: 15463,15677 Priority: 2 Problem Area: Kernel Problem Description: Additional changes to watchdog instrumentation in 3.0.2x code base. Description of Fix: Instrumentation added to capture task activity on the switch at time of reboot and setTaskLogMode to capture second level instrumentation can now be run by admin to capture further watchdog data capture. When 2nd level instrumentation is enabled switch reboot is no longer required. Defect ID: 13820 Priority: 1 Problem Area: Name Server Problem Description: While trying to replicate a virtual disk the storage device does not respond to the other similar storage devices which causes a time out and aborts the operation. Fix: Now allow RNID to be accepted. =================================== BUGS FIXED IN FABRIC OS 3.0.2j =================================== Defect ID: 12895 Please note this fix was incorporated in 3.0.2g and due to clerical errors did not appear in the README file. Priority: High Problem Description: The problem occurs when getting the active zone configuration name and the active zone database contents. Each time the switch is polled for this information, the length of the active configuration name bytes on the switch is no longer visible. So if the active configuration is "myzone1", 8 bytes are lost (7 characters plus NULL terminator) each call. Description of Fix: Free the the bytes allocated for the Active Configuration name. Defect ID: 14517 Priority: High Problem Description: With FCP device probing turned off on the switch, Fabric controller does not send an RSCN after a target device registers with the name server with a RFT_ID Description of Fix: Using a new bitmap to indicate NS to send RSCNs. Defect ID: 14993 Priority: High Problem Description: Switch not sending enough LIPs to transition from AL-PA sequence to Old_Port Description of Fix: After sending out LIP for LIP_RETRY by tFcph, the device responded by sending out LIP followed by IDLE in the same micro second, then LIP and LLI interrupt was raised at almost the same time. Software always processes LLI interrupt before LPSM interrupt, handling IDLE with a LIP interrupt pending. In this case the IDLE was just sent out to flush the LIP as required by FCPH, not really a protocol to try old port. Ignore the IDLEs and go to loop init directly. Defect ID: 15061 (secondary of 14827) Priority: Critical Problem Description: Switch lockup due to shell task running indefinitely in the case of telnet/shell out of sync. Description of Fix: Add taskdelays so that shelltask doesn't run indefinitely. Defect ID: 15422 Priority: High Problem Description: Switch does not prompt for a Reboot after configuration change. Description of Fix: Remove the reboot requiremnt of the configuration parameter "Enable Close on Open". Setting takes effect as soon as switchEnable. Defect ID: 15577 Priority: Medium Problem Description: PRLI being dropped Description of Fix: Always reset the port before excluding zoning in it. Defect ID: 15757 Priority: Medium Problem Description: If user specifies telnet session timeout value at say 5 minutes and then telnets to a switch BUT does not login, the session will NOT timeout. Additional users will see that the switch is "IN USE...." Description of Fix: Add the timeout mechanism in the portion of code prior to telnet login. Then telnet will also timeout before login. Defect ID: 15799 Priority: Medium Problem Description: lbit was cleared in ALPAmap incorrectly Description of Fix: Setting the lbit in the ALPAmap in LIFA phase Defect ID: 16113 Priority: Low Problem Description: Need to add Fabric Watch E-mail notification Description of Fix: Added E-mail notification. Defect ID: 16154 Priority: High Problem Description: Device sent LINIT(CD,00), but the switch sends a LIP(F7,F7) when it should have sent a LIP(CD,00) Description of Fix: Pass the arg2 as the third parameter and value of 1 as 4th parameter when calling bloomStartLoop() in bloom_ioctl.c. 4th parameter allows F8 in byte3. Add a condition to check if the reason of calling bloopStartLoop is caused by LINIT. If it is, uses passed value of byte3 and byte4 to do the lip. Defect ID: 16850 Priority: High Problem Description: RJT was being used to setup CAMS which caused RJTs to be dropped. Description of Fix: Not using RJT for setting up CAMs Defect ID: 16865 Priority: High Problem Description: Auto-negotiation failure with CPQ switch. Description of Fix: After auto-negotiate complete, FC_AL_RESET was done twice. Both FC_AL_RESET are not necessary. Will do FC_AL_RESET in the following three situations: 1. Gain SYnc 2. Change Speed 3. FIFO hits some problem During auto-nego, after entering NEGOTIATE_COMPLETE, we have already been in sync for more tan 271 ms in NEGOTIATE_FOLLOW state. Defect ID: 16903 Priority: High Problem Description: Fabric Manager accesses the page FabricInfo2.html while the fabric is being re-configured. Description of Fix: Serialize accesses to the domain database to prevent intermittent values while information in the database is being computed. Defect ID: 17313 Priority: High Problem Description: The ls_port_name field in the ADISC accept payload was not consistent with the values returned in the PLOGI and PDISC accept payloads when the ADISC was sent to a well-known address (e.g., Management Server). Brocade switch returns a different WWN in the ADISC Accept than the one it returns in the PLOGI Accept for the Management Server. The PLOGI and PDISC handlers take into account whether the ELS is addressed to a well-known address (e.g., Management Server) or to some other port address. The ADISC handler did not. Description of fix: Changing the ADISC ELS handler: Get the d_id field from the inbound ADISC ELS: Make the port name field in the ADISC accept response consistent with the PLOGI and PDISC responses (i.e., pick the appropriate port name based on the d_id field in the ADISC ELS). =================================== BUGS FIXED IN FABRIC OS 3.0.2k =================================== Defect ID: 18058 Priority: Critcal Problem Description: If a host is rebooted with an Emulex LP952 HBA and the topology setting is (2) auto-sensing then the HBA will be logged in as G-port. Description of Fix: Changed loop initialization process back to previous specification.