Dell PowerEdge RAID Controller 3/QC, 3/DC, 3/DCL and 3/SC Firmware Update Version: 196T TABLE OF CONTENTS 1.0 Introduction 2.0 Installation 3.0 Fixes or Enhancements Since last Release 4.0 Important Information 5.0 Change History 1.0 Introduction This firmware update delivers bug fixes for following PowerEdge Expandable RAID Controllers. PERC 3/QC PERC 3/DC PERC 3/DCL PERC 3/SC NOTE: Your controller should be running firmware ver 1.57 or higher for this firmware update. If you have a PERC 3/DCL running firmware version 1.48 then please update to firmware v1.57 before this update. 2.0 Installation Performing the Firmware Upgrade Dell Webpack: This program creates a bootable diskette to upgrade the firmware of the controllers. You will need a formatted 3 1/2 inch floppy diskette for the following procedure. 1) Download the file to a directory on your hard drive. The file downloaded is a self extracting executable. 2) Insert a blank diskette into the system's diskette drive. 3) Double click on the file that you downloaded and follow the instructions on the screen to extract the files to the diskette. 4) Insert the diskette into the system with the RAID controller for which you want to update the firmware and reboot the system. 5) Follow the instructions on the screen to update the firmware. NOTICE: pflash.exe will be executed automatically and will report existing code level and prompt you to reconfirm before starting the firmware upgrade process. To perform this process in non-interactive mode please execute, pflash /? and read about possible command line switches. 6) Remove the diskette and reboot the system. Dell Update Package Instructions: Download 1. Click the "Download Now" link to download the file. 2. If the Export Compliance Disclaimer window appears, click the "I agree" link to accept the agreement. 3. When the File Download window appears, click "Save" to save the file to your hard drive. Installation 1. Browse to the location where you downloaded the file and double-click the new file. 2. Read over the release information presented in the dialog window. 3. Download and install any prerequisites identified in the dialog window before proceeding. 4. Install any necessary Embedded Systems Management firmware prior to this BIOS update. 5. Click the Install button. 6. Follow the remaining prompts to perform the update. The Update Package can also be used with Dell OpenManage Server Administrator. See the Dell Update Package User's Guide for more information. 3.0 Fixes or Enhancements Component Current Version Previous Version Firmware 196T 1.92 BIOS 3.33 3.31 Ctrl-M U812 7.09 Boot Block 1.7.1 1.6.7 Qlogic 12160t Firmware 11.12.19 11.12.19 Firmware Enhancements 1. Improved hard drive status logic when virtual disks are spanned across channels. 2. Firmware will automatically change cache settings to WriteThru when a battery problem is detected to preserve data. If WriteBack was originally selected, firmware will switch back once battery problem has been corrected. 3. In the Rebuild function of CRTL-M, a new status called “INCON” will be displayed after a rebuild operation completes if any unrecoverable errors are found in any of the hard disk drives. To identify the location of the error and if any data was present, run a check disk from the O/S (e.g. chkdsk in windows.) Note: Dell recommends that a check disk or consistency check should be run after any RAID rebuild operation is completed by the controller. Firmware Bug Fixes 1. Manual initialization calls will be rejected if a background initialization or consistency check is in progress on the logical drive. 2. Address potential controller lock-ups that could occur during either degraded, rebuilding or multiple drive pull and reinsert scenarios. 3. Addressed potential controller freezing during driver failure on possible Raid configurations 4. Consistency checks are aborted after reboot in cluster or stand alone mode. User must restart consistency check manually if needed. 5. When multiple hard drive media errors are encountered, array rebuild operation will complete. 4.0 Important Information Firmware upgrade while controller is doing background I/O task Firmware upgrade process will fail if any background I/O task (i.e. RAID-5 background consistency check or power on auto-resume rebuild and reconstruction) is being performed. The program will stop at the "Flash Download" screen with the following message: "Consistency Check/Rebuild of logical drive(s) in progress for adapter-X. Failed Writing Image on Adapter X with Baseport ..." Firmware and configuration on the adapter will not be modified by the program. A reboot can be performed to return to the Operating System. Wait until the background task completes before updating the firmware. Installing the PERC 3/DC and 3/QC in a PowerEdge 8450 System If the PE8450 has a four channel PERC 2 controller installed along with the PERC 3/QC or PERC 3/DC then PERC 2 Ctrl-A utility may not work during system boot up. This is due to the limited amount of option ROM space for that system. In this situation, configure PERC 2 controller using OS level application or remove other controllers. Migrating from the PERC 2 to the PERC 3 controllers If you are migrating from the PERC 2 to the PERC 3 controllers you should back up your files to tape and then install the PERC 3. Logical Drive parameters for PERC 3 SC If you have a PERC 3 SC in a platform with 4 GB or more of system memory, do not have the following combination of virtual disk parameters; Direct I/O, No Read- Ahead, and Write-Through on a RAID 0, 1 or 10 virtual disk. If you have a configuration with the settings listed above and you do not want to use cache during I/O operations, change your logical drive parameters to Direct I/O, Adaptive Read-Ahead, and Write-Through. Universal CRTL-M: Multiple adapters can be configured from CRTL-M, but currently selected adapter should be verified before performing any configuration changes using the Select Adapter option. Drive Migration: When hard drives are moved from one controller to another, the target controller configuration must be cleared first to ensure a configuration mismatch does not occur. FlexRAID: FlexRAID Virtual Sizing is no longer supported by the current firmware (v.196T) and should be disabled prior to firmware flash. This can be performed from CRTL-M by accessing Objects -> Logical drive -> View/Update Parameters -> FlexRAID Virtual Sizing and will need to be performed for each logical drive present. 5.0 Change History Change History from 1.80 to 1.92 Firmware Enhancements 1) Enable ability to save debug information across reboots. Firmware Bug Fixes 1) The required byte count for TTY_HIST_READ is not calculated properly. Its been fixed to calculate properly now. 2) The Consistency Check Progress of the LD is read from the local copy of Consistency Array(which is not preserved across boots). Now it's made to read from NVRAM copy of Consistency Array 3) Fixed issue where misbehaving drive would cause the system to hang. 4) Fixed issue where BGI would run during an NVRAM mismatch. 5) Sense Info pointer now increments correctly. 6) Fixed issue where rebuild might not restart after reboot. BIOS Changes None Ctrl-M Changes 1) Fixed Misspellings Change History from 1.78 to 1.80 Component Current version Previous version Firmware 1.80 1.78 BIOS 3.29 3.29 Ctrl-M 5.54 5.54 WebBIOS (Ctrl-H) 1.60 1.60 Firmware Enhancements None Firmware Bug Fixes 1. When media errors are detected during Rebuild operations, the rebuild is unable to complete because the data cannot be reconstructed. A Media Error detected during a Rebuild operation cannot be recovered, and can result in loss of data. All redundant logical drives will be initialized. More specifically, this means that a background initialization (BGI) will also be performed on 3 and 4 drive RAID 5 arrays, and all RAID 1 arrays (and RAID 10/50 variants). 2. Improper reassigning of bad sectors when Write_and_Verify command fails with Media Error (03 11 01). Firmware configures disks to automatically reassign sectors by setting AWRE bit in mode page 1. However some drives do not reassign properly when Write_and_Verify fails & typically return Media Error. Recovery state machine modified and 2 new states added to perform reassign & then reissue original command. 3. Data under run is detected by firmware and after multiple retries; the drive is failed. During SCSI READ, the target completes its data transfer, then disconnects from the SCSI bus without issuing a Save Data Pointers message. When the target reselects the initiator, it sends the final (successful) status. Following a very specific (and uncommon) sequence of SCSI events, the MegaRAID firmware misinterprets the returned command status, and incorrectly simulates a Data Under Run error. The simulated Data Under Run condition will not be generated when the Residual Transfer Count is zero (normal completion). BIOS Changes None Ctrl-M Changes None WebBIOS changes None Change History from 1.74 to 1.78 Component Current ver 1.74 version Firmware 1.78 1.74 BIOS 3.29 3.27 Ctrl-M 5.54 5.54 WebBIOS (Ctrl-H) 1.60 1.47 Firmware Bug Fixes: 1. Hard drive drops off line after power cycling hard drive the fix was to renegotiate for all request sense commands. BIOS Changes 1. Fixed minor bug in Bios. 2. There was a bug where the PERC Bios would hang waiting for a response from the firmware. This would occur if there was data in the cache on boot up. This is fixed. This only affected PERC 3/DC and PERC 3/DCL. Ctrl-M Changes None WebBIOS changes 1. There was a bug when a user created a logical drive greater than 1 TB. This is fixed. Change History from 1.73 to 1.74 Component 1.74 ver Previous version Firmware 1.74 1.73 BIOS 3.27 3.27 Ctrl-M 5.54 5.54 WebBIOS (Ctrl-H) 1.47 1.47 Firmware Changes: 1. Dellconf had a bug where it would hang on launching diagnostics. The fix is that diagnostics commands are not getting properly incremented in the firmware. This is fixed. BIOS Changes None Ctrl-M Changes None WebBIOS changes None Change History from 1.72 to 1.73 Component 1.73 ver 1.72 version Firmware 1.73 1.72 BIOS 3.27 3.27 Ctrl-M 5.54 5.54 WebBIOS (Ctrl-H) 1.47 1.47 Firmware Changes: 1. An issue has been seen with some of the Fujitsu U160 drives. During I/O stress, some Fujitsu drives intermittently report write media error with sense key "0C 03" (recommend re-assign). PERC 3 controllers (firmware version 161n/1.63 and before) set drive to do re-assignment automatically and do not expect this return. They do not handle the error well and cause systems to hang. This issue is fixed in this firmware. The change is limited and minor, that in case drives report write media error with sense key "0C 03", 4 retries will attempt before failing the drive. This change recovers the specific error reported by Fujitsu drives and I/O process continues. BIOS Changes None Ctrl-M Changes None WebBIOS changes None Change History from 161n/1.63 to 1.72 Component 1.72 ver 161n/1.63 version Firmware 1.72 161n/1.63 BIOS 3.27 3.17 Ctrl-M 5.54 5.39 WebBIOS (Ctrl-H) 1.47 1.41 Firmware Changes: 1. When a new Logical Drive has more than 4 drives, the firmware started a consistency check at the newly added data. 2. The System hangs due to failed PCI Transactions generated by Host. Added Firmware work around to guarantee Command Integrity. 3. Data miscompare problem due to misbehaving signal on i960RN after Power- Cycle. 4. Changed DRAM Signature to work around HW problem and properly recover dirty data in Cache. 5. Random Deletion feature of logical drives is enabled. 6. Check is put in for SAF-TE presence when looking for predictive failures. 7. Firmware now reduces the number of outstanding commands to the drives in a case of drive time-outs. 8. Chip timeout and chip reset could occur during recovery of write commands in write back mode. It is now fixed. 9. Reduced group size for caching to support stripe size of 2K. 10. The difference in the number of pending commands is detected to improve FASTIO R1 load balancing. 11. The firmware now determines the disk whose head is nearer to the requested LBA. 12. Memory allocating for different cache pools in the ratio of logical drives associated with each cache poll. BIOS Changes 1. More time delay introduced for invoking CTRL M & CTRL H. 2. To display in mother board BIOS BBS menu as PERC 3/DCL. 3. BIOS Grammatical Errors figured out by DELL are fixed. 4. A bug in the code, which prevented the invoking of CTRL-M, is fixed in this release. Ctrl-M Changes 1. Disallow spanning across partially configured arrays. 2. To allow different memory segments usage by Ctrlm to make newer bigger Ctrlm run on many DELL motherboards. 3. Logical drives with number larger than zero being in background initialization were detected as being in check consistency state. This has been fixed. 4. Distinguish background initialization and background check consistency. 5. "SCSI Standard" string in physical drive information window was misspelled as "SCSI Sdandard". 6. Added support for PERC 4/DI. 7. Added support for the CERC ATA100/4Ch Controller. 8. Support added for NAS features. 9. Added En/Disable Write cache option in physical device menu if device is SCSI DISK. WebBios Changes 1. When we span 8 arrays during create LD, WebBIOS takes 9 and creates a LD. This was happening only in the cluster-enabled mode. This is fixed now. 2. If user selects a filled array (i.e one with no empty space) during create, we were running into problems. This is fixed. 3. Support added for Random Deletion of Logical drives. Change History from 161j to 1.63/161n Component 161n ver Previous version Firmware 161n 161j BIOS 3.17 3.17 Ctrl-M 5.39 5.37 WebBIOS (Ctrl-H) 1.41 1.41 Firmware changes: 1. A customer experienced system hangs due to failed PCI transactions generated by Host 2. Added firmware work around to guarantee command integrity 3. Data miscompare problem due to misbehaving signal on i960RN after power-Cycle 4. Changed DRAM signature to work around HW problem and properly recover dirty data in cache. BIOS Changes None Ctrl-M Changes 1. Fixed problem related to configuration of logical drive, logical drive size was not being displayed properly. WebBIOS Changes None Change History from 1.57 to 161j Component 161j ver Previous version Firmware 161j 1.57 BIOS 3.17 3.13 Ctrl-M 5.37 5.35 WebBIOS (Ctrl-H) 1.41 1.37 Firmware Changes: 1. If a bad drive failing START_UNIT command is attached to the controller then controller appeared to be hung during scan as firmware attempted retries. This is fixed. 2. If a system was rebooted during reconstruction (RAID level migration or Online Capacity Expansion) then controller was hanging at power ON if drive(s) under reconstruction is now missing. There was a problem in handling R/W calls if the drive does not exist. It is fixed. 3. If a drive fails and rebuild starts the beep starts, but beep does not stop if the configuration is cleared. It is fixed. 4. Rebuilding/Check Consistency logic is changed. Before it used to hog the controller resources and other (host) I/O's were given less time thus causing sever system performance degradation. The new logic is: - If Rebuild Rate is between 21 and 90 then the rebuild/chk-con for the stripe is done after 92/(RebuildRate + 2) milliseconds. - If Rebuild Rate is between 1 and 20 then the rebuild/chkcon for the stripe is done after 200/RebuildRate milliseconds. - If Rebuild Rate is 0 then the rebuild/chkcon for the stripe is done after 800 milliseconds. - Otherwise, it behaves as previous firmwares. 5. On heavily loaded RAID-10 with WriteBack setting, there was so much I/O that SCSI chip was not handling it well. This is fixed by throttling mechanism. 6. The chip timeout handling is enhanced to take care of various limiting conditions. 7. Norton Anti-Virus software was issuing reads for LBA out of range. The firmware was not handling it properly resulting in failing of drives. This is fixed. If the LBA is out of range the command is simply rejected. 8. Device driver timeout was logged in the system event log under heavy IO load when certain heavily loaded drives were repeatedly timed out by firmware. This is fixed by reducing the maximum load on such drives after first timeout. 9. U160M hard drives that report SCSI-4 in inquiry command were being negotiated at 80MB/s. This is fixed, now the firmware treats it as a 160M drive. 10. When initialization is aborted the "background init" and "auto resume" NVRAM variables were not set to aborted. They were still set to progress. This is fixed. 11. Added mechanism to synchronize with OS driver. This prevents the hang when Windows 2000 "new device found wizard" is installing driver on an existing Windows 2000 system. 12. Performance Enhancements by better utilization of controller processor resources and load balancing during R1 and R10 read I/Os. 13. Added new code to handle battery monitoring. - Voltage level is reported only during state transition(fast charge to monitoring, monitoring to fast charge) - Trickle charging is now turned on at all times. This should not affect previous DELL battery interface to utilities. 14. Enhanced drive Pull recovery. - Now a retry is done before failing all I/O from the drive. This will avoid failing drive on spurious selection timeout which can occur under abnormal bus condition, resulting in good drives to be reported as failed. - Earlier, it used to take a lot of time to fail the drive if too many IOs were pending. Now, the IOs are completed and drive is failed much faster. 15. Updated Cable pull algorithm. It will work in non cluster environment too. So if the cable is pulled, all the drives in the box are not failed immediately (though the I/O is failed). On connecting the cable back and rebooting the machine, "Disk/Nvram Mismatch" will not be shown. 16. Added support for disk enclosures with more than 12 drives. BIOS Changes 1. Support added to return device path information in Get Device Parameter call (Ext Int 13h func 48h). The SCSI ID is the logical drive number for the drive on a particular adapter. Ctrl-M Changes 1. Changes in 'Objects' - 'Physical Drive' menu, where - 'Remove HSP' option is added. - 'synchronous negotiation' option is removed. WebBIOS Changes 1. NVRAM mismatch occurs when a Change Logical Drive Property operation is done. This is fixed. 2. On Physical Drive properties page, SCSI level 4 being displayed for some drives. This is filtered off and shown as level 3. 3. The previous WebBIOS was allowing to "add" migration with smaller drive in an array. This is fixed. 4. The WebBIOS hung if controllers are scanned during rebuild. This is fixed.