Essential Software for the QLogic iSCSI TOE Expansion card - IBM BladeCenter

Download Description

New IBM xSeries firmware and device driver file naming convention

Version 2.00.00.62

Changes between Version 02.00.00.61 and Version 02.00.00.62:

• ER55563 - INITIATOR ONLY - workaround for Solaris driver for future QLogic Initiator HBA management software enhancements to SendTargets discovery.

Changes between Version 02.00.00.45 and Version 02.00.00.61:

• ER55563 - INITIATOR ONLY - added firmware support for future QLogic Initiator HBA management software enhancements to SendTargets discovery.
• ER50410,ER39279 - For the QLogic Initiator HBAs, added support to reassemble Login and Text PDUs having the continue flag set. The reassembly does not apply to passthru IOCBs, which are the host's responsibility to reassemble. Changed Discovery sessions to advertise a 16k or smaller MaxRecvDataSegmentLength. Maximum total byte count when reassembling continuation PDUs is 64k.
• ER34312 - TARGET ONLY - fixed problem with received PDU counts in target connection mode for Login, Text, Logout, and Task Management Request PDUs received with the Immediate flag set.
• ER55208 - For the QLogic Initiator HBAs, fixed parsing problem with large Text PDUs, such as SendTargets responses.
• ER54579 - INITIATOR ONLY - changed default ISID value for DDB Entries, so each port as a unique default ISID value, based on the MAC address. Added code to upgrade existing DDBs that were configured with the old default ISID, to use the new default ISID instead. DDBs with non-Default ISID values are unchanged. The new Default ISID has the OUI format, with the QLogic OUI, and a 24-bit qualifier whose least significant 21 bits match those in the port's MAC address, and most significant 3 bits are reserved for MPIO.
• ER54602 - Remove previous workaround for fiber link negotiation issue (ER46846) because the workaround was causing link issues.
• ER46569 - Dropped polling timer from 100ms to 25ms in firmware to catch retransmission scenarios which need to be transitioned into New Reno.
• ER52216 - The ISP40xx previously supported configuring DataPDUInOrder and DataSequenceInOrder. The default configuration is “No” for both. With DataPDUInOrder, No is used to indicate that the data PDUs within sequences can be in any order. Yes is used to indicate that data PDUs within sequences have to be at continuously increasing addresses and overlays are forbidden. With DataSequenceInOrder, No is used to indicate that Data Sequences (a sequence of Data-In or Data-Out PDUs that end with F bit set to 1) may be transferred in any order. Yes is used to indicate that data sequences must be transferred using continuously non-decreasing sequence offsets. Prior to this fix, if the ISP40xx received data in any order, with non-increasing addresses or overlays, the data would not be placed correctly into host memory. To fix this problem, firmware was changed to: (1) always set the firmware configuration to Yes for DataPDUInOrder and DataSequenceInOrder (regardless of what the driver sets these values to via the Initialize Firmware mailbox command 0060h, Set Device Database mailbox command 0063h, or Flash); and (2) detect any out-of-order data PDUs and fail the corresponding CTIO Type 3 IOCB or Command Type 3 IOCB back to the driver with a status error 0036h (PDU out of order).
• ER53478 - INITIATOR ONLY - Firmware was falsely generating an 0x8002 ASSERT when parsing a large SendTargets response. This problem can occur during the parsing of a large Text PDU, such as a large SendTargets response, which the firmware loads into a chain of multiple internal buffers. The problem was caused after the final key=value was copied from the last buffer in the chain. The firmware freed the buffer and set its pointer to NULL, but then mistakenly attempted to validate some of the fields in the buffer header using the NULL pointer. Modified the firmware not to make any assertions after the buffer pointer has been set to NULL. The ASSERTs should only be performed on active buffers.
• ER52744 - Fixed issue where SYN retransmit timers were occurring too frequently. This issue was introduced in build 2.0.0.38 with the fix for ER46569.
• ER52492 - Fixes a potential case where a buffer was not being freed in target connection mode when the target driver closes a new connection immediately after receiving an ATIO6 (for login cmd) and posting a CTIO6 (to retrieve the login payload). This problem was seen while attempting to generate 128 simultaneous logins with the sequence described above. In this case, the firmware stopped sending the Connection Closed AEN indicating connection closure.
• ER51753. performance improvement... removed iqBufStats flag for pendingRP, and changed code to check RP Count against threshold instead of checking flag.
• ER51598 - Fixed case where iscsi keepalive (nopin) timer wasn't getting reset when receiving a tcp-ack for a datain pdu or scsi response pdu in target mode.
• ER051097 - Near simultaneous ping requests on both ports exposed a timing window which could cause one of the requests to not complete (With the fix, this situation will cause the second request to return with a 0x4005 error since only one request is allowed to be outstanding at a time).
• ER51458 - Problem was being seen where a 0x4002 completion was occurring on a mailbox command IOCB. To address this, a fix was made to clean up semaphore handling between the request queue handler and the mailbox handler.
• ER46296/51463 - Fixed race condition in session mode where "slowpath" inbound pdus (e.g., NOP-outs, TaskMgmtCmds, TextCmds) may get processed out of order, causing the pdu to be dropped by the firmware because the cmdSN of the inbound pdu does not match the expCmdSN. Also fixed case (caused by the above fix) where the FW was rejecting a taskmgmt command by sending an iSCSI Reject PDU.
• ER50991 - Fixed an issue where an initiator driver was getting transport errors and/or mailbox timeouts when using host DSD lists due to firmware running out of internal RP structures. Modified firmware to pre-allocate continuation RPs for host DSD lists. The RPs are pre-allocated whenever a Cmd Type3, CTIO3 with host DSD lists (instead of Continuation IOCBs), or a Passthru Type 1 IOCB is received from the host.
• ER50729 - Initiator abortTask response may contain invalid mbox[1] value, causing driver to misinterpret mbox[2] and [3] as maxCmdSn and expCmdSn values.
• ER50267 - If target session mode receives a taskmgmt command with the immediate flag set in the pdu header, the firmware does not set the immediate bit in the ImmediateNotifyAck IOCB posted to the target driver.
• ER49634 - Initiator posting taskmgmt with immed.flag set may cause command to get queued forever and eventually timeout. The change in 2.0.0.52 is a supplement to the 2.0.0.51 change. This change guarantees that taskmgmt mailbox commands are sent in the same order as they are queued to the SCSI Request Queue (when sent via mailbox IOCBs).
• ER 49346 - If we get a NULL local ncb on a create, load the NCB via the hash parameters for debugging.
• ER49634 - Initiator posting taskmgmt with immed.flag set may cause command to get queued forever and eventually timeout.
• ER N/A - Fixed issue with improper results for Ping command.
• ER48688 - FW was hitting an assert due to running out of buffers. This was in a stress test environment when the Initiator task had a lot of things queued up and was getting in a state where it was blocking other tasks from running. Made a change to allow the CPU to be given up in this scenario to allow other tasks to run which would free resources.
• ER48852 - An issue was hit where I/Os were running between 2 ports. The cable was removed for 5 seconds then plugged in. After that the driver received a Status (Entry Type 3) IOCB with a handle which was already completed, and there was no Command (Entry Type 0x19) with that handle currently outstanding to FW. It turns out that a mailbox IOCB was incorrectly being completed to the host as a status type 0 response, in handling a TKM mailbox IOCB.
• ER48465 - Modified Flash Write mailbox command to allow up to 512KB when writing BIOS section (instead of 128KB previously)
• ER48483 - return an error if the host issues a second Ping mailbox command, while the first ping request is still outstanding.
• ER48530 - Fixed issue with Marvell Phy where the Firmware was not properly negotiating to half duplex mode.
• ER46357 - Clean up for ER46278 fixed in revision 2.0.0.32. Changed logic that could have given a "false postive".
• ER48782 - Fixed case where Firmware stopped responding to the host after the ASIC received immediate data which was not in a multiple of 4-byte lengths.
• ER48078 - Fixed problem introduced in build 2.0.0.47 with the fix for ER48078. The fix caused the Set Device Database mailbox command to return a bad status in certain cases.
• ER48148 - ATIO3 could have been posted to wrong host function in SNACK and error cases (e.g., data digest error, payload too large, etc.).
• ER48327 - When DHCP server was unreachable, DHCP was failing at the rebinding phase resulting in lease lost.
• ER48446 - Added intermediate completion option to Ping. This will allow the Ping mailbox command to return immediately rather than waiting for a Ping response.
• ER48078 - Fixes AEN 8002 in target connection mode, which can occur if the target driver is in the process of logging in the iscsi session (i.e., the connection is up, but not logged in yet), and driver issues a connection close mailbox command (0x56) to send a TCP-RST, followed by issuing another connection close mailbox comand (0x56) to send a FIN.
• ER47969 - Handle mal-formed tcp options (with an invalid TCP Option length of 0 or a length larger than the overall frame) on TCP SYN's
• ER48036/ER39930 - Handle AHS spanning more than 64 bytes when digests are enabled.
• ER47968 - FreeDDB mailbox command outgoing mailbox 4 register should return connState, not deviceState

Changes between Version 02.00.00.44 and Version 02.00.00.45:

• ER47629 - Invalid firmware condition was occurring when driver was loaded after BIOS reset card to Factory Defaults. This problem actually occurred due to the fact that there was a problem introduced in the DHCP initiatization code (so the invalid condition could be seen when the user enables DHCP).