Emulex 4Gb fibre channel expansion card firmware v2.70a5 - IBM BladeCenter

Download Description

• New IBM xSeries firmware and device driver file naming convention
  

Version 2.70a5 - 01 Nov 2007

• With VNP enabled and no registered VNPs, firmware will now function as if VNPs are not enabled. (CR 23132)
• Port number generation for virtual ports has been corrected. (CR 23051, 23131)
• Corrected handling of AHB parity errors. (CR 23049, 23050)
• A timeout is prevented by allowing the firmware to process fatal DMA engine errors even if the DMA is not complete, such as during a PCI master abort. (CR 22563)
• The R_A_TOV field is now properly expressed in milliseconds. (CR 22562)
• Any attentions pending during link down are now sent. (CR 22510)
• Firmware now seperates BDL DMA requests so that one is stopped before another is started. (CR 22162, 22179)
• QDR memory POST was modified to establish proper ECC and prevent false memory read errors. (CR 22519, 22520)
• Corrected processing of BDL with unexpected offset to properly handle incoming data frame. (CR 22564)
• Incorporated new Emulex SLI-3 protocol. (CR 20834, 20843)
• SLI-3 coding inconsistencies were found during in-house test with no reported instances of failure.
  • Proper wakeup parameters have been generated (CR 21597)
  • Mailbox response interlock correction (CR 22051)
  • READ_REV command had to be corrected (CR 22067, 22114)
  • Corrected data abort by assuring all data structures are initialized to zero before allocation. (CR 22513)
• Allow VPD offset to be outside of specification for greater flexibility with some systems. (CR 21548)
• Corrected PCI configuration to advertise VPD access for Zephyr 1.1 ASICs. (CR 21391)
• Protocol testing with invalid conditions revealed protocol response errors which have had no reported instances of failure in the field. (CR 17199, 17200, 17203, 20788)
• An I/O timeout during tape operation was eliminated by delaying processing of any FCP2 errors during discovery. (CR 18287)
• Improved DMA error handling of split errors where host memory has been made unavailable. (CR 19348)
• Improved the dynamic coalescing algorithm for better auto-tuning performance. (CR 20208)
• The HBA no longer transmits RRDY during buffer clean-up before FLOGI is sent. Transmitting RRDY during this time is allowed and does not cause any error, but was seen by lab analysis and was regarded as undesirable behavior. (CR 18341, 18342)
• A DMA attempt and error immediately after unconfiguring memory no longer causes the HBA to ignore further commands. (CR 19349)
• Improved management of unavailable memory by limiting the number of retries. The HBA now reports an error attention after retries have been exhausted. (CR 19351)
• Improved transmit channel reset handling to await ASIC memory access completion. (CR 19606)
• Various concurrent FLASH programming issues were resolved by improving the FLASH interlocking/unlocking algorithm, implementing Intel errata, and implementing erase failure recovery code. (CR 18828, 19855, 20480, 20541, 21103, 20151)
• Changed DMA engine to quiesce and shutdown so that a channel reset does not affect other operations of the HBA. (CR 20436, 20438)
• Improved handling of outstanding PCI queue entries on failing DMA attempts. The PCI entry which failed is now invalidated properly. (CR 20439)
• Fixed a recovery issue where a corrupted REC frame with a mismatching offset would cause the HBA to loop, continually requesting recovery. (CR 20792)
• Modified initialization of configuration region to avoid errors if host reads it before the initial configuration is complete. (CR 20795)
• We now default to allowing the PCI bridge to reorder data frames as necessary, which may result in a system performance increase for some systems. (CR 20798)
• Corrected the reinitialization of PCI Config register on PCI Reset. (CR 20864)
• Improved FCP2 recovery to avoid the dropping of frames (and the resulting delays) during REC ELS command. (CR 20935)
• To avoid dropped receive frames in the case where an OXID is reused for a new exchange during the ABTS/RRQ processing window, we now skip any matching "new" exchanges which are in the RRQ state. (CR 20962)
• Commands are now timed out properly when in FCP-2 mode. (CR 21170)
• Corrected SAN boot problem on a specific PCI-Express system by treating SERR as a non-fatal error in SLI-1 as well as SLI-2 mode. (CR 21213)