Release Note for B57DIAG Engineering Diagnostics. ================================================== Due to interrupt sharing problem on some systems with older BIOS, interrupt test and all loopback test will fail on some of these rare systems. If so, it is a bios problem and the way to get around this is to assign the Device Under Test to another IRQ. version 9.00 1. Problem: (CQ 22892) Upgfrm -b sb5787M2.01 and/or upgfrm -b sb5787c2.01 do not upgrade the sb patch. However, seprg -f worked fine. Cause: For Selfboot firmware, "upgfrm -b" command only support Format0 to Format1 upgrade and Format1 to Format1 patch upgrade. However, in case of Format0 to Format0 and Format1 to Format0 update, Diag has a bug to handle these cases. Fix: Diag will warn user to use "seprg" command to handle Format0 to Format0 and Format1 to Format0 Selfboot firmware update. 2. Problem: (CQ 22893) B57diag v8.34 switcg -I allows illegal character as argument (e.g. -I xyz). Fix: Check for illegal characters function has been added to filter out illegal characters for -I (iteration) option switch. 3. Enhancement: Added code to display "format0" in stead of version string for Format0 Selfboot firmware. Only Format1 Selfboot firmware contains version string. Diag will display "V0.00" for Format0 Selfboot firmware. Diag has been modified to display "format0" instead of "V0.00" for Format0 Selfboot firmware. 4. Enhancement: Added ISCSI boot firmware programming support to B57diag. "iscsiprg" command will allow user to program ISCSI boot firmware, ISCSI Config block and ISCSI Config program to NVRAM. 5. Enhancement: Modified Diag to display "0 byte" instead of "16 bytes" NVRAM size when Diag is not able to detect the size of the NVRAM for 5787 and 5754. 6. Enhancement: Added new support for UMP firmware to enable UMP Echo function. The "umpecho" command will allow user to enable of disable the Echo Test function of the UMP firmware. 7. Enhancement: (CQ#22851) Added code to prevent user to program A0/A1 Selfboot patch into A2 device. 8. Enhancement: "-stan", "-jade" and "-hm" has been set to enable by default. User do not need to enter such option switches to enable support of 5755, 5754 5787, 5714, 5715, 5780. version 8.34 1. Problem: (CQ 22660 ???) The engineering mode "secfg" command presented the cable sense option (i.e. option 53) for the 5753M part (it should not). Cause: The option added in v8.20 has a bug that displays the option for all parts due to a typo in the code. Fix: Fixed the typo in the code so that cable sense is available for the parts noted below in point 2 of version 8.20. 2. Problem: The engineering mode "seprg" command produces an incorrect warning when executed on a 5753 part. Cause: Identification of the 5753 part was not coded into the error checking portion of the "seprg" command. Fix: Added error checking for the 5753 part in the "seprg" cmd. 3. Problem: (CQ 22701) Mac address of 2nd port is incorrect if b57diag is used with parameter "-fmac" Cause: In the function util_AccessMacAddressFile(), after b57diag finishes programming the mac address of 1st port, b57diag increases mac address by one and saved it to file and global.mac_address for 2nd port usage. The first zero will be gone after processing the code. Fix: Make sure 6 hexadecimal digits are always saved in global.mac_address[6] and this is the reason why b57diag doesn't fail if only one NIC is programmed at a time. b57diag always saves the correct format in macaddr.txt and it reads the mac address from macaddr.txt every time. 4. Enhancement: Modified code to read TxMac_status (0x460) bit 3 to check link status for 5714S/5715 step A2 and up. 5. Enhancement: Bit 8 of MII register 0x1e for 5755 will not be test in MII Register Test to prevent Energy_Det signal to go low during MII Register Test. 6. Enhancement: (CQ#22638) Added Atmel 24C02 device support for Jade in Diag. 7. Enhancement: Added "-d" option to "seprg" command in to skip ID check when programming ASF/IPMI code to let engineers has more flexibilities when debugging ASF/IPMI firmware. 8. Enhancement: Added "asfmbox" command in b5diag engineering mode to create and view ASF Off-line Mailbox in NVRAM. The "dir", "sechksum" and "semap" command had also been modified to support ASF Off-line Mailbox. 9. Enhancement: Item 14 in the "ASF Settings" submenu (Item 1 of "ASF CFG" menu) has been changed to "Send 'ASD Ready' SMBus Msg". version 8.33 1. Enhancement: Added ASF firmware support for 5755, 5754 and 5787. Version 8.32: 1. Enhancement: CQ14458 Modified code to restore MAC Address and NVRAM config data between Bootcode and Selfboot firmware in "seprg" command. Fix: Added -p option to "seprg" command to restore some of the NVRAM config data between Boot code and selfboot firmware. -p option will restore NVRAM configuration data between Bootcode and selfboot firmware. The configuration data that will be restored are listed below. a. Wol_enable b. Wol_Limit_10 c. Driver_wol_enable d. Lom_design e. Phy_auto_power_down f. Reverse_nway g. Disable_power_saving h. Phy_led_mode i. Shasta_ext_led_mode j. Cable_sense 2. Enhancement: Added 5714A3 and 5787/5754A2 bond IDs support to b57diag. version 8.31: 1. Problem: (CQ 14518) Upgrading selfboot patch shows incorrect version format. Fix: Extended enhancement CQ 14406 to "upgfrm" command. 2. Enhancement: Added HOT SWAP support for 5714/5715/5780 devices. 3. Enhancement: (CQ 14605) Removed the warning of using the default VPD value when programming the blank nvram with "seprg" command. 4. Enhancement: (CQ 14573) Extended the "-firmall" option with b57diag to also support UMP firmware. 5. Enhancement: (CQ 14466) Extended the "-geneep" option with b57diag to also support selfboot firmware. 6. Enhancement: Extended the "-chksecfg" option with b57diag to also support selfboot firmware. 7. Enhancement: Added "sbfcfg" command in engineering mode. This function will allow user to edit the configuration of of a selfboot file image. 8. Enhancement: Added 5 more keywords in "-chksecfg" option to check firmware versions. bootver = pxever = adfwver = adfwcfgver = iscsiver = 9. Enhancement:(CQ 14098) Added validation to check if the firmware to be updated is right for the chip. Diag will not only check device id but also device type, such as M and F. This check will only be applied to devices after 5705. 10. Enhancement: Modified C1 NVRAM Test to work with small size nvram devices. 11. Enhancement: Applied new Phy script to 5787 and 5755 after Phy reset. version 8.30: 1. Problem: (CQ 14426) 5715s A2 fails D2 Phy Loopback Test. Cause: ECD 13216 Fix has been applied which cause the test to fail. When in PHY loopback mode signal detect is not generated (as there is no activity on the wire, loop is closed within the serdes) so when fix is enabled link (which is up) is logically and-ed with signal detect (which is down) so result is link down. Fix: To resolve this issue, ECD 13216 Fix will be disable when running Phy loopback test on 5714/5714 A2 Serdes devices. 2. Enhancement: (CQ 14406) New code has been added to support the new selfboot version string format. major, one digit minor, two digits build, convert to a, b, c...., When build = 0, do not display any letter. Otherwise, 1=a, 2=b....26=z, 27=A, 28=B... major=1, minor 2, build=0, display as "sb 1.02" major=1, minor 2, build=3, display as "sb 1.02c" 3. Problem: (CQ 14216) Intermittently fails b57diag internal MAC loopback test on 5704S Cause: Software halted the internal cpu while the cpu was autoneging to establish link with the ESM partner. The autoneg process is automatically done by bootcode after GRC reset. Halting the CPU also caused the auto-negotiation process got halted, such that there was no signal detected by the MAC. As a result, no data could be received and the test failed. Fix: Software will wait if the autoneg process is going on before halting the CPU to avoid this issue. version 8.29: 1. Enhancement: Renamed "-iscsi" option switch to "-piscsi1". This option is used to execute a field upgrade of NVRAM to add ISCSI boot firmware. The firmware is programmed into a/the device/s specified by “-c” option switch. 2. Enhancement: Added "-pump1" option switch. This option is used to execute a field upgrade of NVRAM to add UMP firmware. The firmware is programmed into a/the device/s specified by “-c” option switch. 3. Enhancement: (cQ #14351) Added "-uump" option switch. This option is used to execute a field upgrade UMP firmware in NVRAM. The firmware is programmed into a/the device/s specified by “-c” option switch if UMP firmware is preexisted in NVRAM. 4. Enhancement: (cQ #14350) Added "-dir" option switch. This option is used to display file directory in NVRAM. version 8.28: 1. Enhancement: Modified 5755 NVRAM workaround code, so that it only applied to the A0. 2. Enhancement: Limited the maximum values programmed to the "Replenish Threshold" register for 5755 and 5787 for improved performance. 3. Enhancement: Added support for 5755 A1. 4. Enhancement: Added the "cable_sense" keyword to the words recognized by diagnostics in the EEPROM.TXT file. Details in the *.PDF. 5. Enhancement: In the "secfg" and "secfgsb1" command for selfboot firmware, added the Cable Sense option (enable/disable) selfboot supported devices. 6. Enhancement: Added access to GPIO3 via "gpioread" and "gpiowrite". GPIO3 is only valid for 5752 and new devices. 7. Problem: (CQ 14312) SMBus address does not display properly in secfg Cause: If the SMB address is set by older version of diag, the SMB address setting in "asfcfg" of such nvram image will not sync up with the SMB address in "secfg". Fix: Added "util_syncSMBaddr()" function to sync up the SMB address in both "asfcfg" and "secfg". 8. Problem: (CQ 14228) B57Diag: 5755 with ASF enabled fails C2, C5, C8 Cause: C2 and C5 test failed when diag reset the device. The reset failed because ASF firmware did not deposit "~Kevt" to sram 0xb50 after GRC reset. C8 test will wait for "0x1d" in Reg 0x5034 after CPU reset. Since it is ASF firmware will write "0x1d" in Reg 0x5034 instead of 1st phase of bootcode, the wait time is longer. Hence C8 test will fail. Fix: Next release of ASF firmware will have C2 and C5 issues resolved. The solution for C8 issue is to increase the wait time when ASF is enabled on 5755. 9. Problem: (CQ 14306) B57Diag: "Cannot gain access to flash, SwArbit: 3100" error message with 5752 Cause: The issue to the last cpu reset. In which, the cpu has been reset without halting the cpu. It will cause the software arbiter to lock up. Fix: To resolve this, the cpu will be halted before issue the cpu reset. version 8.27: 1. Problem: (CQ 13932) Device cannot link up with partner when entering "loadd" command. Cause: The "loadd" command will autoneg link speed by default. If the link partner is set to "force link", the UUT device and the link partner will not link up. Fix: Added "-f" option switch in "loadd" command. User can use "-f" switch to "force link" instead of "autoneg" for link speed. 2. Enhancement: Enabled CQ13779 fix. 3. Enhancement: Increased wait time for C8 test for 5755 if ASF firmware is enabled since it takes longer to load all firmwares. version 8.26: 1. Enhancement: Finished 5755 A0 NVRAM bug work around. Update flshdgsj.bin with the bug work abound in v3.3. 2. Enhancement: (CQ 14153) Removed option 14 from ASFCFG. That option is now configured via SECFG. 3. Problem: (CQ 14174) B57DIAG Eng Fails on exit with 5701s PXE enabled. Cause: This issue can also be found in V8.24. It is introduced when adding selfboot support. There is a immature read nvram function call when bootcode is loading PXE after reset during exit. Such call only is needed for devices support selfboot. Fix: Limit this read nvram function call only to devices that support selfboot. 4. Problem: (CQ 14094) sedump command does not work in q57diag v 8.23 and 8.24. Correctly specified sedump command gives error: Cannot open then displays usage message. Cause: A modification has been done on sedump command. Since most of the commands are using a format that don't need a space after options, the sedump command has been modified to follow the same format. Fix: Modified the code to change back to support the old input format to support customer script files. 5. Problem: Diag will not show 5787 bootcode version number. Cause: There is a bug in the selfboot firmware detection function which fail to detect if the nvram contains selfboot firmware or legacy bootcode. Fix: Modified the code to fix the bug in the function. version 8.25: 1. Enhancement: Started adding code for 5755 A0 NVRAM bug work around. 2. Problem: VPD Test failed on 5787A0, A1 with NIC card using EEPROM. Cause: The bug was introduced when adding 5755 support since V8.24. A workaround for 5755 has been incorrectly applied to 5787. Fix: The code has been modified to apply the workaround only to 5755. 3. Problem: "txpkt" command will hang system with 5705 device. Cause: The bug was introduced when adding 5755 support since V8.20. A workaround for 5755 and 5787 has been incorrectly applied on 5705. Fix: The code has been modified to apply the workaround only to 5755 and 5787. version 8.24: 1. Enhancement: Improved the script file (*.do) "write" function for NVRAM access by byte swapping the write value. 2. Enhancement: Enhanced the "dids" engineering mode command and the "-dids" DOS command line option to identify 5752 A2 silent revs. 3. Enhancement: Update Jade A1 Rev ID. 4. Enhancement: Enabled ECD 13216 Fix for 5714S/5715S A2. 5. Enhancement: (CQ13951) Added 3 commands to let user config selfboot firmware without entering the secfg mode. The 3 command are "secfgsb1", "secfgsb2" and "secfgsb3" 6. Enhancement: (CQ13927) Due to chip timing issue, diag skipped MAC loopback test for 5780. 7. Enhancement: Updated 5755/5755M Bond ID. 8. Problem: (CQ #13953) 5714/5715 A2 with ST M45PE10 Flash gets NVRAM corrupted after upgrading bootcode Cause: This issue was introduced since V8.22 when adding support for STM25P10A no-buffer flash on Jade. Function programNoBufferFlash() got corrupted. Fix: Fixed function programNoBufferFlash() and upgrading bootcode did not show any problem. 9. Problem: (CQ 13945) B57Diag: "Reading current NVRAM ... OK" line should not be present Fix: Removed such message. version 8.23: 1. Enhancement: V3.2 of flshdgsj.bin released. Improved NVRAM strapping options decoding to account all possible NVRAM configurations. 2. Enhancement: Modified code to support selfboot firmware in C8 CPU Fetch Test. version 8.22: 1. Enhancement: Enabled two tests for 5751Cx/5755/5787 family by default. The "serial" test, tests the serial number capabilities registers and is now run as nictest A6. The "power" test, tests the power capabilities registers and is now run nictest A7. 2. Enhancement: Modified the "cputest" (test C2) for the 5714/5715/5780 so that it will now pass. A bug in the chip is worked around by setting a bit that causes the hardware to use a path that is not effected by the chip bug. 3. Enhancement: Added iSCSI Boot support in diag. Impacted command: upgfrm (Added -i option switch) iscsiprg (new added) secfg (added option 3 in item 26) secfg3 (added new parameter for -m option). Impacted option switch: -piscsi (new added) -iscsi (new added) 4. Problem: (CQ #13904) B57Diag: Updating MAC address with "-m" switch erases VID, DID, SVID, SDID and MAC address. Cause: The issue has been introduced since V8.18. When using -m and -mac along with -f, it works fine. Once -f is not entered, -m and -mac will not work and corrupted the nvram. The reason of the nvram corruption is that software is writing an uninitialized buffer data to the nvram. Fix: This issue has been fixed by initializing the buffer before writing to nvram. 5. Enhancement: Added "-putil" option switch. This option switch should be used along with "-firmall" option switch. When "-putil" was entered, diag will call another application "b57putil.exe". "b57putil.exe" will unload PXE and such that diag can have full control over the device. "-putil" should be used when updating firmware via PXE connection. version 8.21: 1. Enhancement: Added support for the 5755/5787 in the "cputest" and "vpdtest". New bin files have been added. (cpusj.bin and flshdgsj.bin). 2. Enhancement: Enable Gigabit Mode for 5787 and 5755. Old versions of diag only support 10/100 speed with 5787/5755 FPGA board. 3. Enhancement: Updated Register Map for 5787 according to the most updated Register spec. 4. Enhancement: Added "-M" option in txpkt command to enable CQ 9139 Fix in 5787and 5755. 5. Enhancement: Modified to allow programming 5787M image to NVRAM in 5787. 6. Enhancement: Added programming Selfboot firmware to NVRAM via "upgfrm" command. 7. Problem: (CQ #13773, #13825) System with 5700 and 5701 hang when running Diag. Cause: The issue was introduced since Diag V8.18. Diag has accessed the 0x7000 block register when adding support for 5755 and 5787. In 5700 and 5701, the 0x7000 block is not accessible and cause the system to hang. Fix: Diag will not access the 0x7000 block registers when the target is 5700 or 5701. 8. Problem: (CQ #13822) 5752A2 with ST Flash shows incorrect firmware/configuration info when loading PXE or ASF. Cause: Bug has been introduced since V8.17 when adding support for 5755 and 5787. The problem only occurs when other NVRAM except Atmel is used. The function that detects if Atmel Flash has been used is broken and cause the software to read the version string at a different location. Fix: Fixed the function in question. 9. Problem: (CQ #13768) B57Diag: "sedump -l -f" command dumps incorrect NVRAM size Cause: When -l option is not entered, the sedump file default length will be 8K. If -l option had been entered but did not specify the dump size, the entire NVRAM will be copied to file. Fix: When -l option has been entered but missing the dump size, a warning will be issued to prompt user to enter the dump size or the entire NVRAM will be copied to file. version 8.20: 1. Enhancement (CQ # 11370): Added the engineering mode command "aspm" for setting PCI-E link states. 2. Enhancement: In the "secfg" and "secfg5" command added the Cable Sense option (enable/disable) for the 5752M/5755M/5787M/5787FM chips. 3. Enhancement: Enabled C8 CPU Fetch Test in NICTEST by default. 4. Enhancement: Enhancement Register Test for 5755 and 5787. 5. Enhancement: Enabled 5714 and 5715 support by default. 6. Enhancement: Removed ECD 13216 fix on 5714/5715 A2 SERDES devices for re-evaluation. version 8.19: 1. Problem: (CQ #13760) B57diag V8.17 "sedump" command was broken. Cause: When adding selfboot firmware support for 5787, a bug was introduced. Output file name cannot be enter correctly, such that nvram data cannot be dumped to file. Fix: Corrected the issue by fixing the file name input function. Impact: The issue was introducing since version 8.12. 2. Enhancement: Added 5714/5715 A2 support. 3. Enhancement: Added 5752 A2 support. 4. Enhancement: Enable ECD 13216 on 5714/5715 A2 SERDES devices. 5. Enhancement: Enable ECD 13332 on 5714/5715 CUPPER and SERDES devices. 6. Enhancement: Set Bit 15 and clear Bit 14 on Reg 0x6C of 5714/5714 to maximize performance. version 8.18: 1. Enhancement: Enhanced the "-dids" command line option and the "dids" engineering mode command so that when they are run after a new device ID has been programmed to the NVRAM and the device has been reset the newly programmed ID of the device will be displayed via "dids". 2. Enhancement: Added 5903M and 5903F support according to 8.3 MRD. 3. Enhancement: Added NVRAM devices support for 5755 and 5787. 4. Enhancement: Added new PHY ID support for 5755 and 5787. 5. Enhancement: Added -K option in txpkt command to enable CRC-32 check on RX path. It is part of the IPV6 support for 5787. 6. Enhancement: (CQ #13546) Enhance display format of "seread" for LITTLE ENDIAN Fix: Added -l option in seread and sewrite commands to support LITTLE ENDIAN format. 7. Enhancement: Added code to support NVRAM image which is larger than 128K byte in size. NVRAM image with size larger than 128K can be program to NVRAM from both command line mode or engineering mode if the NVRAM is large enough. 8. Enhancement: Added 5780 B0 support. version 8.17: 1. Problem: (CQ #13510) After upgrading pxe version, Diag sets incorrect pxe speed on 2nd port of dual port. Cause: The issue is due to diag incorrectly update secfg when programming PXE. After programming PXE, diag will need to update Expansion ROM size in "secfg" for both port. During the update, software has updated the wrong field. Fix: Fixed the bug. 2. Enhancement: Added UMP Firmware support. New added commands and option switches. umpcfg, seprg -u, -pump. 3. Enhancement: Removed C8, CPU Fetch test from NICTEST for re-evaluation. version 8.16: 1. Problem: Secfg menu upate phase 2. In secfg menu, some of the settings need to be applied independently to selected port in dual port devices. However, the current diag only apply the setting to the primary port. Fix: Modified code to correct just that. Impact: It will have impact on 5704, 5714, 5715 and HTLE. Commands get affected are: secfg, secfg2, secfg3, secfg4 and -meep. The settings get affected are: Item 33: Reversed Nway Item 37: Phy Auto Powerdown Item 44: SERDES TX Drvr Pre-Emp - ENABLE (5704 only) 2. Enhancement: Updated Diagnostic User's Guide PDF file. version 8.15: 1. Problem: (CQ#13038) After disabling MAC_A of dual port devices and then reboot the system, the MAC_B no longer able to pass traffic. Cause: When user disable MAC_A with the current diag , only NVRAM offset 0xc8[11:10] has been set. NVRAM offset 0xD8[11:10] remain unchange. Therefore 0xc8[11:10] and 0xd8[11:10] are NOT the same. For bootcode to correctly disable MAC_A both 0xc8[11:10] and 0xd8[11:10] must be the same. Fix: Modified code to make sure that 0xc8[11:10] and 0xd8[11:10] have the same setting when disabling any port. Impact: This issue is introduced in diag V8.02. It will have impact on 5704, 5714, 5715 and HTLE. 2. Problem: In secfg menu, some of the settings need to be applied the same to both ports for dual port devices. However, the current diag does not do that. Fix: Modified code to correct just that. Impact: This issue is introduced in diag V8.02. It will have impact on 5704, 5714, 5715 and HTLE. Commands get affected are: secfg, secfg2, secfg3, secfg4 and -meep. The settings get affected are: Item 19: Force PCI Mode Item 23: Max PCI Retry Item 25: Dual Mac Mode Item 29: Expansion ROM size Item 30: Design Type Item 33: Reversed Nway version 8.14: 1. Problem: (CQ#13415) DIAG does not display NVRAM settings correctly that have been set by previous DIAG Cause: Source code to access EEPROM has been corrupted when adding support to new nvram. The issue is introduced in version 8.13. Fix: Resolved issue by correcting corrupted source code. version 8.13: 1. Problem: (CQ#12718) A2 control Register failed on 5714, 5715 and HTLE. Read/write bit 1 did not get cleared after writing zero at offset 0x468. Cause: Phy detected False Carrier and cause Bit 1 to go high during the test and it stay high for a short amount of time. Fix: Modified diag to put the phy in ISOLATED Mode before A2 test and reset Phy at the end of the A2 Test. 2. Enhancement: Updated Stanford Pin Strap Table for NVRAM 3. Enhancement: Change SBD Hrdlen field from 8 bits long to 6 bits long for IPV6. version 8.12: 1. Problem: (CQ# 13077) According to B57Diag.pdf, tests B1 and B3 are enabled by default, but they are not ran when executing B57diag.exe. When using the command "-T" to include those tests, they still do not run. Cause: B1 test is skipped by mistake for Hamilton. B3 test will only test on device up to 5704. Section of PDF file is outdated. Fix: Including B1 Test for Hamilton. For test that is not applicable to target device will have N/A labeled. PDF file has been updated. 2. Problem: User is not able to edit firmware image using -meep if image only contain bootcode. Cause: When adding dual port IPMI support, bugs has been introduced. User can only edit firmware image that contain bootcode and IPMI code. Fix: This has been fixed. 3. Enhancement: Added Register Test support for Stanford and Jade. 4. Enhancement: Continue to add selfboot support. 5. Enhancement: Continue to add IPV4 and IPV6 support for Stanford. 6. Enhancement: Added IPV4 and IPV6 LSO support for Stanford. 7. Enhancement: Added IPV4 and IPV6 new BD format support for Stanford and Jade. 8. Enhancement: Added IPV4 and IPV6 IP/TCP/UDP checksum off load support for Stanford and Jade. 9. Enhancement: Added RSS support for Stanford and Jade. 10. Enhancement: Added new Flash device support for Stanford and Jade. version 8.11: 1. Problem: (CQ# 13028) In asfcfg command, user could configure the link to speed to a invalid configuration. Cause: There is no such configuration: Force 10_100 speed or force all_speed. When force speed is selected, one of 10, 100, or 1000 speed must be used. Fix: When an invalid conbination is found, it will change to force 100 speed as default. 2. Problem: D4 test failed on all good chips. Cause: This issue was introduced since V8.09. D4 test failed was due to "SND_BD_FLAG_DONT_GEN_CRC" bit was not set. When modifying code to use small size packet in loop back, the "SND_BD_FLAG_DONT_GEN_CRC" bit was not set and it will let the device to pad zeros and generate CRC for the packet. However, not setting the "SND_BD_FLAG_DONT_GEN_CRC" bit causes D4 test to fail since it use the software to generate CRC and the "SND_BD_FLAG_DONT_GEN_CRC" bit need to be set. Fix: Modified the code to set "SND_BD_FLAG_DONT_GEN_CRC" bit in D4 test. 3. Problem: (CQ# 13027) In asfcfg command, user could not change the speed to 1000 Cause: The code was erronously coded; when use try to force 1000, the code will configure to force to all_speed. Fix: Fixed the bug. 4. Enhancement: Enhanced IPv6 extension headers support for Stanford. Hop-by-Hop and Destination header now supported. 5. Enhancement: Modified memory test for the Stanford memory map. 6. Enhancement: Modified to support new NVRAM pin strap tables for Stanford and Jade. 7. Enhancement: Modified register test for Stanford and Jade with the new register spec. version 8.10 : 1. Problem: (CQ# 12946) System under test would be hung when save secfg data via b57diag v8.08 and V8.09. Cause: This issue was introduce during adding IPMI support for dual-port devices in diag v8.08. The cause is a flag used for detecting if IPMI firmware checksum was corrupted had not been set correctly when IPMI firmware was not loaded. Fix: Modified code such that the flag will be set correctly. 2. Enhancement: (CQ12909) Enabled HTLE (5780) support by default and added bin files for HTLE (5780) to the b57diag release directory. version 8.09 : 1. Problem: Diag failed to filter out chips that did not pad small size packet (< 60 bytes) to 60 bytes with zeros. Cause: Diag loopback test did not test with packet that has less than 60 bytes in length. The start_packet_size in loopback test is 60. Fix: Changed the start_packet_size to 22 bytes in loopback test. With packet size less than 60 bytes, the chip should pad it to 60 bytes. For those devices that fail to do so will be filter out. 2. Enhancement: Finished coding first take of IPv6 for Stanford. No extension headers at supported at present. 3. Enhancement: Added support to PCI and PCI-E PXE firmware programming. Impacted commands and option switches: pxeprg, pxeprga, upgfrm, -pxe, -ppxe. version 8.08 : 1. Problem: When NVRAM is corrupted or is blank, diag will warn user to reprogram NVRAM and Mac address. "Warnung: NVRAM content corrupted" "Reprogrammning NVRAM and Mac Addr may correct this problem" When user only try to program the blank NVRAM without programming the MAC address. Diag will issue the same warning message. However, user usually ignore the warning. Fixed: Improved the warning messages when the NVRAM is corrupted or is blank. ******************************************************************************* *WARNING: NVRAM content is corrupted or NVRAM is blank. * * Please reprogram the following: * * 1. NVRAM by using -f option. * * 2. MAC ADDRESS by using -m, -mac (for single port) or -fmac options * ******************************************************************************* Improved the warning messages when user only progame the corrupted/blank NVRAM without programming the MAC address. ******************************************************************************* *WARNING: NVRAM content is corrupted or NVRAM is blank. * * Please reprogram the following: * * 1. MAC ADDRESS by using -m, -mac (for single port) or -fmac options * ******************************************************************************* 2. Problem: (CQ # 12887) sedump with '-a' option fails OEM reported that when using the command "sedump" with the '-a' option, it will fail the Checksum test. Cause: A checksum test has been added in "sedump" command to check the data being read back from NVRAM. Before NVRAM data being dump to a file without using “-a” option, the start address of the firmware will be changed. Then the checksum test function will take care of the change and calculate the checksum correctly. However, in case of using “-a” option, there is no start address modification. That is why the checksum function failed. Fix: Diag will perform the checksum test to a duplicate of the NVRAM data buffer. If the test pass, diag will either perform the start address modification depends on if the “-a” option present. 3. Problem: (CQ # 12570) If a flash image is created with the wrong length the image still can be programmed using the b57diag command line without failure. Cause: Normally, diag will only check the magic value before programming the NVRAM. Fix: Diag will test the checksum of the NVRAM image before programming to NVRAM to make sure the NVRAM image is in good condition. 4. Enhancement: The "-exe" option was enhanced so that when multiple devices are present and the "-exe" option is used then the "@" symbol can be used to execute a tests on multiple devices. 5. Enhancement: Continued adding code for Stanford IPv6 support. 6. Enhancement: Added Smbus address programming support for IPMI on dual-port devices. In case of IPMI firmware on dual-port devices, SMBUS Address can be configurated via "secfg" command, item 51 and 52. User can only configurate SMBUS Address on the 1st port via "asfcfg" command. In case of IPMI firmware and ASF firmware on single-port devices, SMBUS Address can be configurated via both "asfcfg" command and "secfg" command, item 51. version 8.07 : 1. Problem: The -pnchk option checks that an entered part number matches the value stored in the image file that is to be programmed. If the entered string was longer than the string in the file image, then no error was flagged (incorrect action). Cause: The code has a bug that did not flag an error when the entered string was longer than the string in the image file. Fix: Fixed the code so that if the string length of the entered value is longer than the expected value in the file then the characters up to the expected string length are compared. 2. Enhancement: Decreased the flash arbitration time out value because the delay on failure was too long. 3. Enhancement: (Hamilton) - Read the PCI bus speed value from PCI register 0x74 for revision A3-on instead of using the fixed speed of 133 MHz. - "seprg" now does not over write the second MAC address - "umpecho" command added to put the ump firmware in loopback mode so that every thing seen on the input is echo'ed to the output (supported on A1 and newer devices) / ump14b.bin is updated (version 1.3) 4. Enhancement: Added the following changes for Stanford: - added support for variable ATMEL NVRAM sizes - fixed a number NVRAM access bug - disabled FastBoot - disabled VPD test (not supported) - disable voltage variation because it is fixed - added multiple ring support - added device ID for the mobile version - IPv6 implementation started version 8.06 : 1. Enhancement: Modified "regtest" so that it runs on the Stanford FPGA when the -fpga55 option is used. 2. Enhancement: Added SelfBoot Support to Diag (Phase 3). Impacted the following option switches and Commands. -w -asf -x -fmac -m -mac -mba -mbas -pxe -pipmi -dids -pasf -pxes -pupdateasfcfg -firm -firmall -updatesecfg -updateumpcfg -chksecfg -ppxe -smbaddr -vpxe -f -u nictest dids setwol 3. Enhancement: Started adding Stanford support for multiple rings. 4. Enhancement: Added the "pciwr" command and improved the "pcird" command's help. 5. Enhancement: (PCDR) Added 5714 support to PCDR. New Bin files need to be added as well. version 8.05 : 1. Enhancement: Removed "Hot Plug" secfg options for Baxter family of chips because the chip family does not support those options. 2. Enhancement: Modified Jade device ID to 0x7f7f. 3. Enhancement: Added engineering mode commands txcrcip, txcrcport, txdstip, txdstport, txtype, txprot and txipprot which set up the Tx packet information without using the txcfg command. 4. Enhancement: The blast engineering mode command can now specify an absolute timeout value after which the blast command automatically returns to the command prompt via -Q. 5. Enhancement: (CQ #12619) The "nvsize" commands allows entry and display of data in hex and decimal. 6. Enhancement: (CQ #12620) The "tpm" size was misinterpreted due to a bug in the "nvsize" command which was fixed above in CQ #12619. 7. Enhancement: Added SelfBoot Support to Diag (Phase 2). Impacted the following Commands. seprg upgfrm pxeprg asfprg setpxe setwol setump setasf dir dirw erase semap secfg setipmi sechksum sedump sever semap secfg secfg1 secfg2 secfg3 secfg4 secfg5 asf asfcfg umpcfg binchksum vpdwrite vpdinfo userblock tpm. 8. Problem: (CQ #12545) A use did not like the display when programming at ST NVRAM on the Shasta chip family. Cause: Too much intermediate status send to screen. Fix: Modified code to remove verbose output. version 8.04 : 1. Enhancement: For Stanford support: - Modified the msi_test according to new register locations - Added Stanford register definitions for the PCI cfg reg - Memory testing now uses the Stanford memory map. 2. Enhancement: Added initial Jade support. 3. Enhancement: Added Shasta C1 support. 4. Enhancement: Replaced CPU14.BIN with CPU14A.BIN and CPU14B.BIN. CPU14A.BIN will have no-op workaround. CPU14B.BIN will NOT have no-op workaround. Hamilton A0, HTLE A0-A3 will need to use CPU14A.BIN. Hamilton A1 will need to use CPU14B.BIN 5. Enhancement: Replaced FLSHDG14.BIN with FLSHD14A.BIN and FLSHD14B.BIN. FLSHD14A.BIN will have no-op workaround. FLSHD14B.BIN will NOT have no-op workaround. Hamilton A0, HTLE A0-A3 will need to use FLSHD14A.BIN. Hamilton A1 will need to use FLSHD14B.BIN 6. Enhancement: Replaced UMP14.BIN with UMP14A.BIN and UMP14B.BIN. UMP14A.BIN will have no-op workaround. UMP14B.BIN will NOT have no-op workaround. Hamilton A0, HTLE A0-A3 will need to use UMP14A.BIN. Hamilton A1 will need to use UMP14B.BIN version 8.03 : 1. Problem: (CQ #12424). Baxter displays voltage variation on test B7 and C8, even though the voltage of the chip is fixed. Cause: Oversight. Fix: Voltage variation for test B7 and C8 removed. 2. Enhancement: Run the Open/Short Test (E1) only at 1000 Mbits/s. Experimental data shows no advantage to running at 10/100. 3. Enhancement: Baxter support enable by default. 4. Enhancement: Modified to ban bin-bang interface to NVRAM. version 8.02 : 1. Enhancement: Performed changing in device ids for HT-LE Copper and Serdes devices so that it will be different than the device ID for Hamilton. The new device IDs for the 5780 (aka HT-LE) are: 5780 (HT-LE) copper: 0x166A 5780 (HT-LE) serdes: 0x166B 2. Enhancement: Added code in "binchksum" command to handle invalid input file. Function will return "Invalid Magic Value" message if invalid file has been used. 3. Enhancement: -chksecfg | Check Boot code configuration. -chksecfg will enable the NVRAM Secfg Test. Boot code configuration will be checked against input files. The input files are in the same format as EEPROM.TXT. is for the primary port and is for the secondary port. Software will read command from input files and compare with the boot code configuration and return “Passed” or “Failed” accordingly. Example: b57diag –chksecfg file_p.txt (for single port devices) b57diag –chksecfg file_p.txt file_s.txt (for dual port devices) 4. Enhancement: Added "vpdinfo" command in engineering mode to Show VPD Information. vpdinfo cmd: vpdinfo Description: Show VPD Information. Syntax: vpdinfo version 8.01 : 1. Enhancement: Added the engineering mode command "dmashasta" that tests basic DMA functionality (valid for the Shasta family only). 2. Enhancement: Baxter A1 bond ID test and seprg device identification enhanced/fixed. 3. Enhancement: Added "binchksum" command in Diag engineering mode. The "binchksum" command will take a firmware file and test the checksum of each piece of firmware stored in the file. It is simular to the "sechksum" command which is testing the checksum of each piece of firmware stored in nvram. syntax: 0:>binchksum -f 4. Enhancement: Added code to the "sedump" command to verify the firmware checksum of the extracted data from nvram to ensure the qulity of the output image of the "sedump" command. version 8.00 : 1. Enhancement: Modified diagnostics to not touch register 0x2018 for chips that belong to the Shasta family. 2. Enhancement: Added back the Open/Short test that was removed in release v7.01 of diagnostics, with some algorithm improvements. The test is available under the new "Group E" tests. Group E is used for tests under development. 3. Enhancement: (CQ# 12180) Allow enable of IPMI on both ports for the dual-port chips. Fix: Modified diag to let user to enable IPMI firmware in both ports of the dual port devices. However, user will be able to enable ASF firmware only in one port of the dual port devices. Impact: Impacted commands and option switches are list below. 1. secfg 2. secfg3 -a 3. setasf 4. setipmi 5. setwol 6. setpxe 7. setmba 8. -asf 9. -mba 10. -mbap 11. -mbas 12. -pxes 13. -w 14. -x 15. -ipmi 16. -meep 4. Enhancement: (CQ# 12323) Enhancement request to display the slot type in Diag engineering mode. Fix: Modified diag to show slot type and speed in the following format. For PCI-E Devices display changed from: PCI = E 2 Spd = 500 to: PCI = Ex1 (for pci-e x1 slot type) Spd = 250 (2.5G) version 7.60 : 1. Problem: (CQ# 12252) System would lock up when testing b57diag v7.59 on adapters with EEPROM Cause: Diag write to reg 0x7020 when devices do not support memory arbitration. Fix: To fix this, diag will check if devices support memory arbitration before writing to Reg 0x7020. 2. Problem: (CQ# 12176) B57diag did not get NvRAM grant before issuing GRC_RESET. Fix: B57diag will get NvRAM grant before issuing GRC_RESET. 3. Problem: (CQ# 12179) b57diag C2 & D1 test fails Cause: C2 test fails has been fixed in v7.59.(CQ# 12143) The cause of D1 fails is due to diag did not wait long enough after resetting the RX Engine. Diag continued to its rx setup when the rx engine was still in reset mode. This only happened to the two CIOBE ports. Fix: Diag will increase the wait after resetting the RX engine. 4. Enhancement: Remove support for Baxter device. version 7.59 : 1. Problem: (CQ# 12184) DOS DIAG: Not able to use 'Reset -c' command. System return error message "Cannot gain access to flash". Cause: The issue is due to nvram arbitration lock up. Firware was granted arbitration to access nvram after cpu was halted by diag. Fix: diag cleared firware arbitration after halting CPU. 2. Problem: (CQ# 12143) Diag C2. Fails on 5704, 5703, 5702 cards. Cause: Tx cpu test time out loop has been shorted when debuging other issue and did not get restored. Fix: Restored tx cpu test waiting time and issue got fixed. version 7.58 : 1. Enhancement: Modified MSI test for Hamilton. After each MSI the mailbox register is written to clear the interrupt. 2. Enhancement: Added an option to setup hot plug power values for Baxter chips and for Shasta C0 devices and newer. The engineering mode command "secfg" or "secfg5" and the DOS command "-meep" now support this feature. 3. Enhancement: Modified "-firm" and "-firmall" option switches and "upgfrm" command to support new function of hot plug power values for Baxter. version 7.57 : 1. Problem: The VPD test failed on a system (CQ #12013). Cause: Static function are handled incorrectly by the compiler. Fix: Removed static function calls and recompiled the following binary images (release 3.0) cpu05.bin, cpu.bin, cpudg05.bin, cpudiag.bin, flshdg05.bin, flshdg5x.bin, and flshdiag.bin. 2. Problem: The same system send/receive test fails. Cause: The interrupt fix (CQ# 11107) of point 1 of release v7.47 below introduced an error which caused the above failure. The error is that on some interrupts an EOI was not sent. Fix: Modified the code to send an EOI for all interrupts destined for the BRCM controller. 3. Problem: The embedded CPU test (cputest or nictest c2) fails when the host CPU is throttled. Cause: The modification in version 7.53 point 5 caused this problem and additional delay is required in order for the embedded CPU to return the pass indication to the host CPU. Fix: Increased the time out amount the host CPU waits for the embedded CPU before a failure is declared. version 7.56 : 1. Problem: PCI cfg reg test failed on Shasta C0 at reg 0x6000. Cause: For Shasta C0 the functionality of register 0x6000 (the MSI register) has changed. Fix: Modified code to correctly test register 0x6000. 2. Problem: (CQ #11960) 'Cannot gain access to flash' error on OEM 5751 A1 LOM Cause: The issue happened during diag unloading driver and only when ASF was enabled. When diag was tring to gain NVRAM arbitration before stop CPU, CPU was still loading ASF firmware. Since the LOM was using EEPROM instead of flash, it took longer time to load ASF from NVRAM. As a result, daig failed to gain access to NVRAM. Fix: To resolve this issue, diag will wait until CPU completed loading firmware from NVRAM before accessing NVRAM. 3. Enhancement: (CQ 11902) Add Command Line Option for Programming ASF SMBus Address Fix: Added -smbaddr | …| option switch to diag. -smbaddr can take mutiple parameters. Example: a. -smbaddr 45 (SMB Address = 0x45) b. -smbaddr 45 6c (SMB Address for 1st device=0x45 and 2nd device = 0x6c) -smbaddr must be used along with -c option. Number of parameters for -smbaddr must be the same as number of devices selected by -c. Example: b57diag -c 0 3 6 -smbaddr 6c 6e 70 -t abcd. b57diag –c 0 –e b57kia –pasf asf.bin –smbaddr 66 –t abcd version 7.55 : 1. Problem: (CQ #11944) An EEPROM 5721 LOM chip fails the NVRAM test (nictest C1). Cause: During modification of the NVRAM (Baxter) code a bug was introduced. Fix: Modified code to remove above bug. 2. Problem: The 5701 chip family fails the bond ID test. Cause: Fixing the issue in version 7.53 point 1 caused this issue because the default case in the bond ID test which allows all zero bond ID devices to pass was allowing the 5701 family of chips to pass the bond ID test. Fix: Added a case to the code to specifically test the bond ID for the 5701 family of chips. 3. Enhancement: Added support for Shasta C0. 4. Enhancement: Added new revision ID support for Hamilton A1 and HTLE A2 on. 5. Enhancement: Added new revision ID support for Baxter A1 on. 6. Enhancement: (CQ 1159) Add support to diag manufacturing mode for user using PCI bus, dev and function to select devices. Fix: Modified -bus option switch to accept dev and func input. e.g. -bus 1 (select all Broadcom devices on bus #1) -bus 1:2 (select all Broadcom devices with bus #1 and dev #2) -bus 1:2:3 (select Broadcom device with bus #1, dev #2 and func #3) User can use either -c or -bus to select devices. version 7.54 : 1. Problem: The Baxter chip and the ST M45PE10/20/40 NVRAMs had the following issues . the NVRAM size displays incorrectly . the method used for programming was slow . NVRAMs with a zero value for the FLASH bit were treated as EEPROMs (the FLASH bit is a jumper option that is reflected in reg 7014) . failed the VPD write test Cause: The causes of the above issues were identified as: . diagnostics did not have a routine to size ST NVRAMs and hence displayed the incorrect size . the ST part was programmed as a un-buffered device and should be handled as a buffered device . for Baxter flash devices with a zero FLASH bit can be flashes . VPD write are not supported for Baxter Fix: The fixes for the above issue are: . added a NVRAM sizing routine for ST parts . treated the ST parts as buffered devices . allow the FLASH bit to be zero for flashes . do not run the VPD write test on Baxter chips 2. Problem: When FastBoot on the Baxter chip is enable and bootcode is updated the first and second phase of bootcode can get out of sync and can cause a number of issues. Cause: Problems are caused on FastBoot because in memory phase one can be out of sync with the NVRAM phase two bootcode. Fix: When bootcode is updated in NVRAM force the normal boot procedure. version 7.53 : 1. Problem: (CQ# 11401) The 5789-B1 passed the bond ID test in v7.34 unexpectedly. Cause: Diagnostic allows devices that have a bond ID of zero to pass the bond ID test (the bond ID of 5789-B1 is zero) even if the device is not supported by that particular version of diagnostic. Fix: Modified diagnostics to always check for supported bond IDs even if the bond ID is zero. 2. Problem: (CQ# 11737) DIAG C2 test failed (CPU Test) on Baxter in PCI-E X16 slot. Cause: Compiler has bug that it compiles 'static' routine incorrectly. Fix: By replacing all static routine to public routine, it fixed the problem. CPU05.bin V2.1 will fixe the issue. 3. Problem: (CQ# 11499) 5751A1 NIC is not recgonized by DIAG in OEM system. Cause: Normally, diag detects devices by passing vendor ID and device ID to BIOS and BIOS will return information about the device if the device is present. In this system, BIOS did not return valid data to diag. As a result, Diag could not see the device. Fix: The "missing" device can only be found by scanning all PCI devices in the system. So to working around the "missing" device issue, a new command line switch (-sc) has been added to diag. It will change the way diag detect devices. By entering "-sc" diag will scan all the PCI devices in the system instead of let BIOS detecting the target devices. e.g. b57diag -sc 4. Enhancement: Update labels for year 2005. 5. Enhancement: Modified the embedded CPU reset routine to only halt the CPU and not reset the CPU before the halt. Rename the function t3_reset_cpu() to t3_halt_cpu(). version 7.52 : 1. Enhancement: Added the "-nosz" DOS command line option that will disable the storing of the TPM and NVRAM size in NVRAM when diagnostics is run from the DOS prompt. Since V7.50 the NVRAM is written to automatically when diagnostic testing is executed from the DOS command line. Use the "-nosz" option when repeatedly power cycling a system in order not to exhaust the total NVRAM write cycles. 2. Enhancement: Replaced the following bin files with the latest version. File Name Latest Version CPU.bin V2.0 CPU05.bin V2.0 FLSHDG5X.bin V2.6 version 7.51: 1. Problem: When user create TPM block in NVRAM, this block may overlap with other firmware and result in firmware corruption. Cause : "tpm" cmd allow user to specify the location of the TPM block. However, diag did not check if the new TPM block will overlap with other firmware. Beside, user is also able to specify a TPM block that is larger than the maximum available space of the NVRAM. Fix: Diag will check for the location of the tpm block to make sure it is not overlapping with other firmware. Diag will also make sure the size of the TPM block will not larger that the maximun available space of the NVRAM. version 7.50 : 1. Enhancement: The "nvsize" command now runs automatically when the following commands are run: - "-f", "-firm", "-firmall" in DOS command mode, and - "seprg", "upgfrm -b" in engineering mode. version 7.49 : 1. Problem: (CQ# 11542) The "-firmall" command fails with the SA25F020 flash chip. Cause: Diagnostics is code for maximum of 131Kbytes of NVRAM and the above chip provides 256Kbytes. Fix: Updated maximum NVRAM constant size to allow for 1Mbyte. 2. Problem: The "nvsize" command corrupts the checksum in NVRAM. Cause: A bug existed in the manner the checksum routine was called. Fix: Called the checksum routine with the correct values. version 7.48 : 1. Problem: (CQ# 11311) Missing carriage return when "sechksum -v0" is run. Cause: Oversight. Fix: Added carriage return. 2. Enhancement: Added an option to enable/disable L0s performance enhancement for the Shasta devices. The engineering mode command "secfg" or "secfg5" and the DOS command "-meep" now support this en/dis feature. 3. Enhancement: Store the NVRAM and TPM size to NVRAM. The engineering mode command "nvsize" preforms this function. version 7.47 : 1. Problem: (CQ# 11107) Diagnostics crashes on certain systems. Cause: The interrupt handling fails on certain systems. Diagnostics EOI's all interrupts that it sees. This causes problems for some chained ISRs and produces the diagnostics crash observed. Fix: Modified diagnostic to only EOI interrupts destined for it. Interrupts destined for chained ISRs are not EOIed. 2. Problem: (CQ# 11256) Diagnostics ROM loader fails when the host CPU is throttled. Cause: There is a bug in the diagnostic code that tracks the ROM loader that is observable only during CPU throttling. Fix: Modified the ROM loader tracking code to allow for the case seen during CPU throttling. version 7.46 : 1. Enhancement: Add support to configure the 5704 SERDES TX Pre-Emphasis. The engineering mode "secfg" and "secfg4" commands, and the "-meep" DOS command line option now have options to enable the Pre-Emphasis and set the values for both ports. Fix: Added support as specified above. 1. Problem: (CQ# 10966) b57diag -all option position dependent in command string Cause: "-firm" and "-firmall" are single task functions. They only take a few option switches which are entered after "-firm" and "-firmall" commands. "-firm" will take: "-sil" and "-log" "-firmall" will take: "-sil", "-log", "-updatesecfg" and "-updateasfcfg". Fix: "-all" has been added to "-firm" and "-firmall" option switches list. verson 7.45 : 1. Enhancement: Enabled force VPD write in vpdtest command. Cause: By default, diag will do vpdwrite test when vpdtest command gets call the first time to presurve the life of the NVRAM. User cannot force diag to do vpdwrite test again. Fix: Enabled force VPD write in vpdtest command. 2. Enhancement: Update flshdg05.bin, flshdg5x.bin and flshdiag.bin from V2.24 to V2.25. 1. Problem : Nictest C2 (cputest) failed with version 7.44 when run with the new NVRAMs that are supported as of version 7.44. Cause: Support for the programming of the new NVRAMs was added to diagnostics but the support for running the NVRAM test (C1 or setest) was not added which corrupted the NVRAM and caused the C2 (cputest) to fail. Fix: Added support for the C1 test or setest for the NVRAMs that support was added for in version 7.44. version 7.44: 1. Enhancement: Add support for following flash devices. - SaiFun SA25F0xx where xx=05/10/20 (512Kbit-2Mbit) - ST M45PEx0 where x=1/2/4/8 (1Mbit-8Mbit) Fix: Support for flash devices specified above added. 2. Enhancement : Remove the "-ctpm" command line option added in v7.42. Fix: Removed the "-ctpm" command line option. 3. Enhancement : Enhance the engineering mode command "log " so that output is flushed often, so that if the system hangs a log file is still produced. Fix: Added the "-logfl" DOS command line startup option that enable frequent log file flushes (aka write do disk). 4. Enhancement : Change the PXE programming code so that it also changes the Expansion ROM size parameter in the NVRAM based on the PXE image size. Fix: Added code to do so. 5. Problem: (CQ# 11130) Diags -firmall errors out if file size doesn't match device size Cause: The error in fact is because of mismatch in file size and device size. Diag checks file size and nvram size before checking device ID. Fix: Diag will be modified to move size checking after device ID check. 6. Enhancement: Updated flshdiag.exe, flshdg05.exe and flshdg5x.exe to V2.24. version 7.43 1. Problem : (CQ# 11066) Shasta 12x12 device identification bug causes diagnostics to fail on nictest A5 and C7. Cause: Bug in device identification code added in v7.41 item 8. Fix: Added new device ID to device identification code. version 7.42 1. Enhancement : (CQ# 11045) Add a DOS command line option that will erase the data stored in the TPM NVRAM block and maintain the block directory entry. Fix: Added the "-ctpm" command option that erases the TPM data. 2. Problem : Shasta 12x12 device identification text was incorrect when added in v7.41 (item 8). Cause: Shasta 12x12 device identification text has been changed. Fix: Added support for the new text (names) for the devices. 3. Problem : (CQ# 10988) Need diag command to update IPMI only if IPMI is present Cause: Need an option or a flag in the q57udiag command to update IPMI if and only if IPMI is already present. Currently, -pipmi will program IPMI code to all devices or all devices that are selected. Fix: Added -uipmi option switch to support IPMI firmware updata on devices that IPMI is already present. version 7.41 1. Problem : (CQ# 10905) Diagnostic fails the send/respond test when run on device 1 and 2 and there is no link on device 0. Cause: Diagnostic incorrectly checks for link-up on all devices in the system. Fix: Modified diagnostic to check for link only devices being tested. 2. Problem : (CQ# 10888) The "dids" command BOND REV column values do not align correctly when a PXE version number is displayed. Cause: Diagnostic incorrectly aligns the BOND REV values. Fix: Modified diagnostic to correctly align the BOND REV value. 3. Problem : When the help screen is displayed at the DOS command prompt some commands are not visible. Cause: The page breaks are set incorrectly. Fix: Modified/added page breaks so all diagnostic command's help information can be seen when displayed at the command line. 4. Problem : Immediately after starting diagnostics in the engineering mode when "memtest" is run the CPU GPR test is executed by default but on all following runs of "memtest" the CPU GPR test is not run and is disabled. Cause: During the execution of memtest the default setup is stored and all memory test are then enable and run; after the run the original memtest setup is restored but is done so incorrectly due to a typo in the code. Fix: Modified code so that now the restored values for memtest are the same as the defaults saved before running the test. 5. Enhancement : Started added skip on error functionality to engineering mode and manufacturing mode. Fix: Skip on error, skips the remaining part of the test when the test starts to fail. At present skip on error functionality present in the external loopback test. Skip on error functionality will be added to all test over time. 6. Enhancement : Added the "-fail2" DOS command line option that causes the failed message to be offset from the present location by six characters (i.e. not aligned with pass messages). This feature makes it easier to identify failures when scanning through log file outputs. Fix: As specified above. 7. Problem : (CQ# 10979) The cputest fails when the HOST cpu is throttled. Cause: The embedded cputest has a bug in the reset sequencing that is only visible when the host cpu is throttled. Fix: Modified reset sequencing order so it now works with host cpu throttling. 8. Enhancement : Added support for 12x12 mm Shasta devices. Fix: Added support for above. version 7.40 1. Problem : Diagnostic would crash when the BIOS would not initialize the base address register (BAR). Cause: Diagnostic did not check to see if BIOS initialized the bar to a non-zero value. Fix: If diagnostic sees a device that has a zero BAR it will ignore the device and present a message to the user noting the offending device. 2. Enhancement : Modify diagnostic so that engineering mode commands can be executed when there are NO Broadcom devices present. Fix: Added the command line startup option "-exe". When this flag is set any engineering mode diagnostic command will execute without a device present. 3. Enhancement : Improved error message when the cputest memory test fails. Fix: Now print out address, expected and read data. Initially on failure printed our address and expected. Version 1.9 or new is required for cpu.bin and cpu05.bin. 4. Problem : (CQ10857) b57diag -firmall option requires 2 updates to get Checksum correct Cause: -firmall option will preserve the ASF CFG code from the NVRAM. However the ASF CFG code offset of the NVRAM has been used, when preserving the ASF cfg block. It causes the ASF CFG block being copied to a different location of the image buffer and damage the ASF CPUB block which is below the ASF CFG block in the buffer. This is the reason why ASF CFG and ASF CPUB failed the checksum check at the first After the firmware being updated the first time, the ASF CFG code offsets of the NVRAM is the same as the buffer of the image. In addition, a new checksum of the ASF CFG block has been re-calculated before updating the buffer and programming the NVRAM. It cause ASF CFG checksum to pass. However, even though the checksum is good, the content of the ASF CFG block is not. Fix: The whole issue is due to an incorrect code offset has been use. By selecting the right code offset, the issue has been resolved. 5. Enhancement : Add support for Shasta B1. Fix: As specified above. version 7.39 1. Problem : When the "cpudiag" test is run mutliple time the floppy drive fails. Cause: In the diagnostis command prompt, command history code a index varible wrote in one location passed the allocated memory space. Allocate an array of size 20 and wrote to indexs 0-20 (i.e. 21) Fix: Now index goes from 0-20 and does not over-write other data. 2. Problem : When a "*.do" is executed the diagnostic error control value for the command prompt is not preserved but over written. Cause: Value not saved and restored in code. Fix: Code now save the command prompt error control value before executing the do script and restores it after. 3. problem: (CQ10688) Diag allow IPMI enable on both ports of 5704 Cause: Diag did not check the current status of the IPMP enable setting before enabling IPMI firmware. Fix: Diag has been modified such that ASF/IPMI will not be enabled in both ports of the dual ports devices. In case of no port is specified to enable ASF/IPMI, Diag will enable ASF/IPMI in Primary port only. In case of ASF/IPMI is enabled in one port and user specified to enable ASF/IPMI in another port, Diag will enable the specified port and disable the other port. 4. problem: (CQ10689) Diag does not report error if features cannot be enabled. Fix: 2 Error Level have been added. In case of enabling PXE while there is no PXE present in NVRAM. Diag will return ERR_MISS_PXE (160) Error Level. In case of enabling ASF/IPMI while there is no ASF/IPMI present in NVRAM. Diag will return ERR_MISS_ASF (161) Error Level. Diag will also has error messages print out on screen in both cases. version 7.38 1. Problem (CQ#10699): System hangs when using -firmall feature of b57diag with a bin file containing bootcode but no ASF. Cause: The issue is caused by a unbounded for loop which hangs the system This for loop is used for detecting if the input file for -firmall contain ASF firmware image. Fix: Keeping the for loop within boundary, system does not hang any more. version 7.37 1. Problem : When there are no devices in the system and diagnostic is started up, diagnostics will hang. Cause: The ODI16 interrupt fix implemented between release 7.34 and 7.35 did not check for the presence of a device before writing to the device which causes the hang. Fix: Only apply the above mentioned fix if a device is present. 2. Problem: The 5701 fiber fails nictest A1. Cause: On the 5701 fiber the ASIC comes up with statistics enable. The ASIC updates the statistic in part of the chip while this test is trying to test the same part of the chip. Fix: Disable statistic gathering at the start of test A1 since the 5701 fiber comes up with statistics enabled. 1. Enhancement (CQ#9395): Modify b57diag to Allow Full NVRAM Configuration Without Requiring a Card to be Installed Fix: New modified version of -meep command line option switch allows user to change the setting of the configuration of the input bin file without requiring a card to be installed. The input file can be a bootcode only file or a file that contains mutiple firmwares. "-meep" will allow user to modify configuration of the primary port. The configuration of the secondary port can also be modified if the input file is for a dual port device. Besides, user can also modify ASF/IMPI configuration of the input bin file if ASF/IPMI firmware is detected in the file. 2. Enhancement : Modified -firmall to preserve secfg setting and asfcfg setting by default Fix: Also add -updatesecfg and -updateasfcfg to force secfg and asfcfg blocks to be updated with the input file for -firmall -updatesecfg will update the NVRAM with the Secfg setting in the input file. -updateasfcfg will update the NVRAM with the ASF Cfg setting in the input file. version 7.36 1. Problem : (CQ 10618) -firmall will perform programming of the NVRAM even the NVRAM size is smaller than the file image size. Cause: Diag did not check for NVRAM size vs the file image size. Fix: Added code to check NVRAM size VS file image size before program. If NVRAM size is less than file image size, error code will be issued. 1. Enhancement : Improve ASF displayed error when the file is too big for available NVRAM. Fix: Print out file to big, instead of invalid file. version 7.35 1. Problem : VPD Test failed on 5704 when IPMI enabled. Cause: per Duc Tran When diag started, it will HALT the internal CPU. Before running the VPDTEST, it will halt the internal CPU and then proceed to download the "flshdiag.bin", and the kick-start the CPU to launch the "flshdiag.bin". This "flasdiag.bin" is part of the b57diag vpd test component that runs inside the internal cpu that services VPD events. However when the b57diag supposed to halt the internal CPU, it only halts CPUA, and left the CPUB running. Furthermore, in the later IPMI PT firmware, there is a WHILE loop that monitors the Expansion ROM and VPD event bits was implemented in CPUB. So when the b57diag started the VPDTEST, it load-n-starts the "flshdiag.bin" on CPUA, while CPUB is continuously running, monitoring and servicing the VPD and Expansion ROM, hence the conflict and leads to b57diag VPD test hung. Fix: Stop both CPUs before running vpdtest and also after GRC reset to avoid conflict with IPMI firware. 2. Problem (CQ#10470): Broadcom Manufacturing Diagnostics Hanging on Interrupt Tests. Step to reproduce: 1. load ODI16 driver 2. unload ODI16 driver 3. load diag 4. run interrupt test. Cause: This issue can also be reproduce on 570x and without load and unload the ODI-16 driver: in Diagnostic set BIT 9 of Miscellaneous Host Control Register (Reg 0x68) then run "intrtest"; system hangs. On 570x, when issue a "reset" in Diagnostic, this bit gets cleared but for the PCI-E NICs it's still set. Hence, causes the Interrupt to continuously to trigger because this bit is use to enable the Tag Status Mode between the firmware and the driver. The reason for Bit 9 of reg 0x68 will not get clear in PCI-E device is when implemented the ECO in A1, PCI-E block will not get reset to maintain hte link if there is a GRC reset. Reg 0x68 is inside the PCI-E block. Fix: The ODI-16 driver clears BIT 9 when it's unloads. Also, in diag, diag will clear Bit 9 of Reg 0x68 before enbale interrupt and before running interrupt test. 1. Enhancement (CQ# 9277): Enhanced IRQ testing for 5705 and 575x devices. Fix: Details of the enhancement is provided in the CQ entry noted above. Basically bits 10 and 19 at PCI register offset 4 are tested. 2. Enhancement: Diag did not show device which has IRQ greater than 15. Cause: Diag skip devices which have IRQ greater than 15 to avoid user has access to these devices. Diag only support devices have IRQ less than 16. Fix: Diag will issue warning to use if any device has invalid IRQ. 3. Enhancement (CQ#10446): Extend the -firm option to update IPMI code also, in addition to update the boot code. Fix: Added -firmall option switch to download image which contain mutiple firmwares to NVRAM. Diag will match sid, vid, ssid and svid of the image to the those in NVRAM before updating NVRAM. Diag will only preserve the Mac address. version 7.34 1. Problem (CQ#10409): B57diag lock-up on OEM system when programming ASF. Cause: The issue is due to a variable pointer not getting correctly initialized. Fix: After initialized the variable pointer, issue has been resolved. 1. Enhancement (CQ10397): Need way to disable IPMI on Broadcom devices for LSO Functionality Fix: Added "setipmi" command in diag manufacturing mode and "-ipmi" as a command line option switch to enable or disable the IPMI firmware. It works the same way as "setasf" and "-asf" for ASF firmware. 2. Enhancement (CQ10041): Add support for Capactive Coupling. The diagnostic software needs to support a new hardware configuration called Capactive Coupling (CC). This feature is enabled/disabled using a new bit in the EEPROM, This bit is also exported by the bootcode to applications running on the host processort using shared memory. Fix: Diag has added Capacitive Coupling feature support. This support is added in "secfg" and "secfg4" commands in diag engineering mode. In manufacturing mode, "-geneep" option switch also have this support. version 7.33 1. Problem (CQ #10323): When attempting to enable Wake On LAN via the b57diag utility with the -b57eng parameter, using the setwol -e command, getting the following warning: "Boot Firmware version 2.6 or later is required to support this command" Cause: Diag uses the same function to enable and disable pxe, asf and wol. This function will detect if pxe and asf code are loaded before changing the setting of pxe, asf AND wol, even wol has nothing to do with bootcode version. In order to detect the present of pxe and asf, the bootcode must be 2.6 and later. Since it is using the same funciton, the same boot code requirement will apply when changing wol setting. In fact changing WOL setting does NOT require boot code 2.6 and later. Fix: Diag has been modified to resolve this issue. Diag will not check the present of firmwares when changing the WOL setting. 2. Problem (CQ #10256): B57diag "write" Command Fails When Using "m" Address Modifier Cause: "m" address modifier in "Write" Command has never been implemented. In order to read from and write to mii registers, "mread" and "mwrite" will normally be used. Fix: Added code to support "m" address modifier in "write" command. 1. Enhancement (CQ #10306): Customer requests to have an option to invoke 'cpufetch' test from the command line in Manufacturing mode. Fix: Added 'cpufetch' test to nictest (C8) and set the default to disable. 2. Enhancement: Enabled dword access on non-dword boundaries for Shasta B0 and newer during the Expansion ROM test (romtest or test C7). Fix: Modified diagnostic to do the above. Shasta based devices now uses flshdg5x.bin as the image file for diagnostics ROM test and VPD test. With diagnostic v7.33 flshdg5x.bin version 2.3 or newer has to be used or a failure will be seen. 3. Enhancement (CQ #10207): Need user diag feature to update IPMI the same as updating PXE. IPMI update will only be allowed when the NVRAM already contains IPMI. Fix: Added "-pipmi" to program ipmi firmware. It works the same way as -pxe. version 7.32 1. Enhancement: Added support for the 5751F and 5789 Shasta variants. 2. Problem (CQ# 10179, 10166): Diag will attempt to load ASF FW into insufficient EEPROM space Cause: Diag did not check for free NVRAM memory before load firmware to NVRAM from file. Fix: Check free NVRAM memory space before programming ASF firmware to avoid this issue. 3. Problem:(CQ: 10099) b57diag secfg menu PXE enable option is not clear. The b57diag secfg menu item for enable/disable PXE is not clear for applications that have the PXE code in the system ROM. Fix: Added "does not apply to PXE in system ROM" text to the following items in the secfg manual. Item 8. pxe Item 29. Expansion ROM size version 7.31 1. Problem: 5751M not supported by diagnostic. Cause: 5751M WAS not supported due to lack of test systems. SQA now has 5751M based systems for testing. Fix: 5751M supported added back into diagnostic. 2. Problem (CQ# 9249): When executing the send/respond test via: b57diag -rf 0 -c 1 -t abcd -T d3 with two cards in the same system the test sometimes failed. Cause: There is a race condition that occurs in the interrupt handler with chained interrupts. Fix: Modified diagnostic to avoid the race condition. 3. Problem:(CQ: 10099) b57diag secfg menu PXE enable option is not clear. The b57diag secfg menu item for enable/disable PXE is not clear for applications that have the PXE code in the system ROM. Fix: Added "does not apply to PXE in system ROM" text to the following items in the secfg manual. Item 8. pxe Item 9. pxe link speed Item 26. MBA Boot Protocol Item 27. MBA Bootstrap Type Item 28. MBA Delay Time Item 29. Expansion ROM size Item 39. Hide MBA Setup Prompt Item 40. MBA Setup Hot Key version 7.30 1. Problem (CQ #10025): The 5701 fiber fails the cputest. Cause: In version 1.3 of cpu.bin the memory testing was expanded. On the 5701 fiber the ASIC comes up with statistics enable. The ASIC updates the statistic part of memory while the memory testing is going on. Fix: Disable statistic gathering at the start of the cputest since the 5701 fiber comes up with statistics enabled. 2. Problem (CQ #10006): Upgrade bootcode cause TPM broadsafe to change. This has not been fully fixed in V7.29. Cause: The block size of the TPM block registered in the NVRAM directory is in byte count, while other firmware sizes registered in NVRAM directory are in word count. diag has got incorrect free NVRAM space and offset when try to program or update firmwares. As a result some firmware may get corrected when updating firmware. Fix: The TPM block is always located on the upper part of the NVRAM address. Diag is using the MAX NVRAM Size to calculate the block offset and free space in the NVAM. When diag has detected the present of the TPM block, it will reduct the MAX NVRAM Sizee from 128K to the starting address of the TPM block. This way diag does not need to due with the fact that the block size of the TPM black is registered in byte in the NVRAM directory. version 7.29 1. Problem (CQ #10006): Upgrade bootcode cause TPM broadsafe to change. Cause: When diag updating or adding firmware to NVRAM, diag will move the firmware in NVRAM around in order to gain more space. Since TMP block is located in the top part of the NVRAM. Once diag try to update the bootcode, TPM block will be moved to another location. Fix: Modified firmware programming tools in diag, such that TPM block will not be moved when programming or updating firmwares. Only "-f" option switch and command seprg will erase TPM block automatically. The following commands and option switches have been modified. erase sechk semap upgfrm -b upgfrm -p pxeprg asfprg seprg -a -firm -pxe -ppxe -pasf version 7.28 1. Problem (CQ #9901): Access to register 0x4c during register testing fails. This issue is Shasta specific. Cause: A IC issue exists which has a work around for revision A3 and newer. Fix: Inserted code that avoids the problem by enabling the fix by writing 0x300 to 0x7e2c. Also after register 0x4c is accessed a delay of 10us seconds is added before any other action is preformed. 2. Problem: The "-mfct" option can fail. Cause: If a user enters something other than the SSID and SVID in the input text file the CRC will be incorrect. Fix: Modified code so options other than the SSID and SVID are ignored when entered in the input text file. 3. Problem (CQ #9984): 5704s EEPROM GPIO/WP issue, not able to write to NVRAM after single MAC enablement Cause: This is due to Set GPIO funtion does not setup GPIO setting correctly. Fix: Modified Set GPIO function to correct the issue. version 7.27 1. Problem (CQ #9955): 5751M part are supported by diagnostic and they should not be. Cause: Since SQA does not have any system with the 5751M part no SQA testing has been done on this part. Fix: Remove support for the M parts from diagnostic. 2. Problem: Bond ID test fails on 5721 A3. Cause: Code not programmed to recognize the above chip and rev. Fix: Modified code to recognize above chip and rev. 3 Problem: Shasta B0 fails PCI config reg test and Expansion ROM Test Cause: Register 0x4c access during PCI register test is problematic. The "Expansion ROM Test" firmware has a bug, when accessing non-dword boundaries, which will be fix on the next diagnostic release. Fix: By-passed the above failing testing on revision B0. 4 Problem (CQ #9952): b57udiag support the -e option and should not. Cause: The -e option is not IFDEFed out for b57udiag in the code. Fix: Remove (IFDEFed out) the -e option for b57udiag. version 7.26 1. Problem (CQ# 8989): Incorrect voltage values for A1 Shasta chips are produced during nictest B7 (CPU GPR Test) Cause: Voltage characterization data for A0 and A1 is different. Fix: Added code to setup voltages based on chip revision. 2 Problem: Shasta A3 fails PCI config reg test and Expansion ROM Test Cause: Shasta A3 did not address the following issues: - Register 0x4c access during PCI register test (power reg) - Non-dword boundary access during "Expansion ROM Test". Fix: By-passed the above failing testing on revision A3. 3. Enhancement: Customer requests a quiet mode of operation for the "-firm" command. Fix: Added the "-sil" command to disable warning messages when the "-firm" command is used. 4. Enhancement: Customer requests a new command to only program the SSID and SVID in the NVRAM (the CRC also gets updated). Fix: Added the "-mfct " command to accomplish the above. The file contains text strings describing the new SSID and SVID. The format for is specified in the diagnostics user manual "EEPROM.TXT FORMAT" section. version 7.25 1. Problem : B4. Memory test takes 4 min longer to complete. Cause: Due to the debug code that was added delay the memory read process. Fixed: After cleaning up debug code, test time reduced to normal. version 7.24 1. Problem (CQ# 9281): When the "mdev" command is used to change the PHY's MII address in order to access SERDES (where available) and then the chip is reset via "reset" an error message is produced and the PHY is not reset. Cause: The "reset" routine uses the MII address to reset the PHY and when a user changes this address the reset routine can not access the PHY and produces an error. Fixed: Since our PHY address is fixed, modified the code so that the "mdev" command does not change this PHY address. Also modified the "mread" and "mwrite" user commands to access the device specified by the "mdev" variable. 2. Problem (CQ# 9281): BCM95701A10 A3 FW:2.2 & FW: 2.04 Fails Diag PHY LoopBack Cause: Phy type detection function returned incorrect phy type and as a result failed the link test. Fixed: Modified Phy type detection function to return correct phy type. The issue has been resolved. 3. Request (CQ# 9249): Create a same system Send/Response test. Fixed: Test D3 run against a reference card is the send/response test. - added the "-rf" option to identify a reference card - modified the "-lbspd" option so that the line speed of reference card can be set during test d3 Documentation on using this feature was added the manual. Basic Usage: "b57diag -rf 0 -c 1 -t abcd -T d3" 4. Enhancements Added the "-addref" option to append a reference to the refinfo.txt file. Added the "-blink" option to blink the LED of a specified device. version 7.23 1. Request: Add command to create a TPM directory in NVRAM. Fixed: Coded above request (created the "tpm" command). 2. Problem: "seprg" cmd cannot program IMPI ASF firmware. Cause: Diag function that detect the device type is not reliable. As a result it fails the chip type check when programming the firmware. Fixed: Modified the check device type function such that it can accurately detect the device type. version 7.22 1. Problem: Voltages not set correctly for Shasta during test B7. Cause: Code for setting voltages is different for Shasta than the code that previous 57xx family of chips. Fixed: Modified code to setup required 1.1, 1.2 & 1.3 volts based on data gathered during chip characterization. 2. Problem (CQ# 9580): Invalid bond id on 5721A1 Cause: ASIC Rev ID for 5721A1 has been added to diag, as a result, it failed bond id test. Fixed: Added 5721A1 support to diag. version 7.21 1. Problem: The "asfprg" command fails in b57diag v7.20 with an "ASF firmware version error" and if you "Continue" or "Abort", then b57diag crashes out with a PMODE/W Exception. The same firmware binaries can be programmed just fine with b57diag v7.19 or earlier. Cause: Invalid structure pointer had been passed to funtion and cause software to crash. Fixed: Instead of passing structure pointer, local Variable has been use. 2. Program: -mf option switch cannot handle large file size. Cause: Diag is using a 16K byte size buffer to store the data from the input file. As a result, data get corrupted. Fixed: Enable buffer size up to 128K to handle input file with large size. 1. Request: When diagcfg option is used to specify the test to be run do not revert back to the default tests after one run when using the nictest command. Fixed: Coded above request. 2. Request: Reverse Shared Traffic Mode and Shasta Mac mode parameter of Item 21 (LED MODE) in secfg menu according to Shasta spec. User used to enter 3 to select Shared Traffic Mode and 4 to select Shasta Mac mode. Start from Diag version 7.21, user will need to enter 4 to select Shared Traffic Mode and 3 to select Shasta Mac mode. version 7.20 1. Request: Enable the -rsttm option for all PCI-E devices (not only A0). Fixed: Coded above request. 2. Request: Modify reset for PCI-E Shasta devices POST A0. Used to be 6804<=1, now 6804<=20000000, then 6804<=20000001 Fixed: Coded above request. 3. Request: Disabled testing of register 0x4c during the PCICfg rg test for Shasta A1 (used to be only A0). A OEM system fails the PCICfg rg test with this enable. Another OEM is fine (looks like a HP BIOS issue). Fixed: Coded above request. 4. Request: Improve the memory test done during the cputest (test C2). Include walking 1's/0's and incrementing data. Fixed: Coded above request. 5. Request: Added asf combined firmware file programming via "seprg -a" command and "-pasf" option switch. Fixed: Coded above request. 6. Request: (CQ# 9421) b57diag: Need to add an additional test which will test CPU instruction fetch logic. Fixed: Added "cpufetch" command to diag to support request. version 7.19 1. Problem: (CQ# 9419) NVRAM corruption when enabling ASF via b57diag. User got error message: ROM LOADER Failed With ffffffff.data =100000 A reset of the fash is required to get out of thes state Cause: When asf is enabled, diag will issue a GRC reset to put asf running in a loop and will not interrupt diag activities. The issue memtioned above is due to the 500ms delay after GRC reset has been skipped. This 500 ms delay is need only if it is a PCIE device. Normally diag will detect if this is a PCIE device. However, in this particular case, the detection came in after the GRC reset. As a result, the 500ms had been skipped. Fixed: The fix is move the detection before GRC reset. 1. Request: Expanded expansion ROM test for PCI-E devices. Access dwords on byte and word boundaries. Fixed: Coded above request. 2. Request: Expanded reset sequence to support Shasta post A0. Change the write to GRC register 6804 to a read/modify/write. Fixed: Coded above request. version 7.18 1. Problem: (CQ# 9221) Cannot upgarde ASF 1.0 with 7.15 Diag. Diag v7.15 when trying to upgrade gives the following message when user upgrades ASF1.0 firmware to v2.04 (for 5702-5704) and v5.05 (for 5705): "Please use B57diag v7.12 or earlier to program ASF v2.01-v5.x. Abort?" Cause: The old diag code structure only support the lastest ASF platform (ASF 2.0). Fixed: Modified diag to support both ASF 1.0 and ASF 2.0 platform. New Addition: ======== 1. Request:(CQ#9076) Add ability to script secfg commands Fixed: Added 4 more engineering mode commands (secfg1, secfg2, secfg3, secfg4) to the diag. Each command will take care 10 secfg fields. Each command will have 10 parameters for user to input. Users can write their own scrip to program all secfg fields. The example below will program the mac address and the vendor id. 0:> secfg1 -m=00:10:18:12:34:56 -v=14e4 version 7.17 1. Problem: (CQ# 9252) The "-rsttm " option does not work in b57udiag. Cause: The "-rsttm " option was not include in the udiag #ifdef. Fixed: Add the "-rsttm " into the udiag #ifdef statement. 2. Request: Remove the "-shasta" option and use "-rsttm 500" as the default for the "-rsttm ..." option. Fixed: Coded above request. version 7.16 1. Problem: (CQ# 9146) RX DMA error text swaps the expected and got (observed) value. Cause: Same printf is used for Tx and Rx errors where source and destination are switched. Fixed: Added a printf for RX DMA error so that the expected and got values are not swapped. 2. Problem: (CQ# 9165) b57udiag: Cannot upgrade firmware on specific adapter using "b57udiag -c 1 -firm a:\ee5702c2.28" from command line. Cause: -c command line option is not implemented to work with -firm command line option. Fixed: Added -c command line option to work with -firm command line option. User can specify card to be updated. 3. Problem: Diag is not able to program ASF+IPMI firmware. Cause: Diag is not aware of ASF+IPMI firmware. Diag will take ASF+IPMI firmware as basic ASF firmware and failed config table check. Fixed: Added support to ASF+IPMI firmware and config table check. version 7.15 1. Updated 2003 labels to 2004. versjon 7.14 1. Request: Added a new test to perform Sequential register and memory writes and reads. Fixed: Added a "wseq" command in engineering mode in Diag. This test will cover Mac Registers, Mail Box Registers and internal memory. A. Sequential Register Write and Read Test. Selected Register for read write: 0x414, 0x41c, 0x424 and 0x42c Data Pattern: Increment. Iteration (loop count): 0x2000 B. Sequential Mail Box Register Write and Read Test. Selected Mail Box Register for read write: 0x200, 0x268, 0x280, 0x300 and 0x380 Data Pattern: Increment. Iteration (loop count): 0x2000 C. Sequential Memory Write and Read Test. Selected Internal Memory for read write: 0x8000 to 0x8020 Data Pattern: Increment. Iteration (loop count): 0x2000 version 7.13 Bug Fix: ======== 1. Preblem: When program 5703A3 NIC with boot code version 2.36, diag issue chip family type mismatched with boot code image. Cause: Since the latest boot code is using a new device id (0x16C7) and diag is not aware that new device id has been use. As a result, diag issue family type mismatch error message. Fixed: Diag will allow boot code with new device id be programmed. version 7.12 Bug Fix: ======== 1. Problem: The log file generate when the cpu test is run had a line greater than 80 characters and was messy, due to the newly added "cpu reset" message. Cause: Stated above. Fix: Display "cpu reset" message on screen but do not include it in the log file. Impact: As described above. version 7.11 Bug Fix: ======== 1. Problem: (CQ 8932) phyctrl command not setting speed/duplex settings correctly Cause: The issue is due to incorrect bits in MII offset 0x00 defined in diag. Fix: Correction has been made to the diag code. Was: Auto Neg Enbled = Bit 9. Speed Selection MSB = Bit 13 Speed Selection LSB = Bit 6 Should be: Auto Neg Enbled = Bit 12. Speed Selection MSB = Bit 6 Speed Selection LSB = Bit 13 Impact: As described above. 2. Problem: (CQ 8926) Cannot update NVRAM image w/ b57diag64 when ASF is enabled. User getting NVRAM image not matching with chip type error. Cause: When ASF code is enabled, it will block diag to access Phy MII registers. Without the access to PHY MII Register, diag will see the phy as a fiber phy running in TBI mode. As a result, diag will think the bin file for copper phy does not match with the phy type of the chip and issue an error. Fix: If ASF is enabled, Diag will put ASF running in a loop before diag getting the type of the Phy by writing DEVICE_RESET_MAGIC to offset 0xb50 and 0 to the last_driver_state of the NIC_SHARED_DATA in SRAM. This way, ASF code will not block the Phy access and diag will not see the copper phy as fibler phy. Impact: As described above. New Addition: ======== 1. Enhencement: Added changes to make sure that diag will download the correct bin files, such as flash.bin, cpu.bin and cpu05.bin, when doing tests. b57diag64 will display bin file versions when loading. New bin files also modified to support version string. Those new files are included in this release. version 7.10 Bug Fix: ======== 1. Problem: Diagnostic will not run from a CD ROM. Cause: At startup the DIAGCFG.BIN file is automatically rewritten. Fix: At startup do not re-write the DIAGCFG.BIN file. A user HAS to use the "diagcfg" command to write the file. Impact: As described above. 2. Problem: User try to program flash/eeprom via the command listed below get error message. "b57diag64 -c 0 -e b57kia -f ee5703c3.03" "Error #67 Invalid Magic Value" "Warning: NVRAM content corrupted. Reprogramming NVRAM and Mac Addr may correct this problem." Cause: Before diag program the flash, it will check to see if the flash is blank or corrupted. It case of the flash is blank or corrupted and user is not going to program both flash and mac address, diag will give out the error messages listed above. "b57diag64 -c 0 -e b57kia -f ee5703c3.03" is only asked diag to program the flash only. Therefore diag issue error message. 3. Problem: Current diag will only accept -f and -m options when user try to program both flash and mac for a blank or corrupted flash. If user use other options such as -mac and -fmac along with -f option to program the blank flash It will so get error messages same as those listed in problem #2 Fixed: Diag v7.10 and above will also let user to use -mac and -fmac along with -f option to program blank and corrupted flash. 4. Problem: When user update mac address via mac address text file for the 2nd time, user may get "In valid mac address" error message. Cause: When user update mac address via mac address text file, the mac address text file will be updated with a new mac start address. However, there is an issue that the mac address fiel is not updated correctly due to data conversion within diag. Fixed: Diag will check the data before updating the mac address text file. version 7.09 New Addition: ======== 1. Request: (CQ 8832) the 'rxcfg' and 'txcfg' commands in the diag allow the user to specify receiver and transmit parameters/properties for various testing, but they are both interactive; i.e., user has to select an option and flag, then exit. for automation/scripting purposes, we'd like to have the capability to set the parameters at the diag command prompt; i.e., for example, instead of having to enter 'txcfg, then enter '12=1' (to enable ip checksum offload), then enter '0' (to exit), we'd like to have the option to just set this option at the command prompt like this: 'txcfg 12=1'. Enhencement: 2 option switches (-txcfg & -rxcfg) have been added. User can set receive and transmit parameters/properties for various testing via command line. user can enter these options multiple times to set different parameters. eg: b57diag64 -b57eng -txcfg 12=1 -txcfg 1=00:11:22:33:44:55 -rxcfg 5=1518 -rxcfg 1=1 The parameter format for both options is item=data. No space needed. User will enter the parameter in the same format as they will enter when running the command "txcfg" and "rxcfg" in engineering mode with a minor exception. In case of rxcfg option, if user want to set RX Mask (item 11), user need to enter the data in HEX. Below is how to construct the data. data = LM_ACCEPT_UNICAST | LM_ACCEPT_MULTICAST | LM_ACCEPT_ALL_MULTICAST | LM_ACCEPT_BROADCAST | LM_ACCEPT_ERROR_PACKET | LM_PROMISCUOUS_MODE LM_ACCEPT_UNICAST Bit 0 LM_ACCEPT_MULTICAST Bit 1 LM_ACCEPT_ALL_MULTICAST Bit 2 LM_ACCEPT_BROADCAST Bit 3 LM_ACCEPT_ERROR_PACKET Bit 4 LM_PROMISCUOUS_MODE Bit 16 eg: b57diag64 -b57eng -rxcfg 11=1001E (Enabled all except LM_ACCEPT_UNICAST) In addition to these 2 option switches, both commands (txcfg and rxcfg) in engineering mode have also been modified. In engineering mode, user can enter the follow to enable ip checksum offload. 0:>txcfg 12=1 It will be more useful in automation/scripting purposes. 2. Request: (CQ 8831) Need test to do looping of reset test to screen intermittent chip errors at users discretion. Enhencement: From command line, "-cpurstlp" option switch has been added. It will set the limit how many times boot test within cpu test will perform. "boot test" will reset the chip and wait for the completion of bootcode loading. eg: b57diag64 -cpurstlp 10 (It will running boot test 10 times as part of the cpu test) Form engineering mode, user is able to do the same with the following command. 0:>cputest -r10 version 7.08 New Addition: ======== 1. Problem: When CPU test (test C2) fails it displays "CPU test failed" only. Enhencement: 1. Changed cpu.bin/cpu05.bin to encode the error 2. Changed b57diag64 to decode the error message to text format. New possible error messages: 1. %cx CPU failed on %s bit access to address %08X 2. %cx CPU failed on memory pattern %08X at address %08X 3. %cx CPU Instruction test failed Impact: This requires new cpu.bin/cpu05.bin firmware version 7.07 Bug Fix: ======== 1. Problem: (CQ# 8766) Diags bustest test error reported on the 5705. Cause: The bustest is not supported on the 5705 family of devices. The 5705 family of device does not provide direct access to DMA engines. The 5700/01/02/03/04 did provide direct access to the DMA engines. Fix: When the bustest is now run a NA is printed out to note that the bustest is not applicable to the 5705 family of devices. Originally running the bustest on the 5705 family of devices would produce an error. Impact: As described above version 7.06 Bug Fix: ======== 1. Problem: (CQ# 8669) After running nictest, diag fails to reset phy and display the DFE locks up message when exit to DOS. Cause: The issue is due to ASF and Diag are fighting for mii access. When start Diag, Diag will put ASF running in a loop by writing DEVICE_RESET_MAGIC to offset 0xb50 and 0 to the last_driver_state of the NIC_SHARED_DATA in SRAM. When exit Diag, Diag will issue a GRC reset and release ASF. ASF is activated and fighting with Diag to access mii. As a result, Diag will fail the DEF Taps tests when resetting phy. Fix: The workaround is to reset the phy before release ASF. Impact: As described above version 7.05 Bug Fix: ======== 1. Problem: User is not able to program blank eeprom/flash and getting NVRAM corrupted message. Cause: Diag will perform a NVRAM check before programming bootcode. If NVRAM image is blank or corrupted, diag will display NVRAM corrupted message and exit to DOS. Fix: In case of NVRAM is blank or corrupted, if user specified -m, -mac or -fmac together with -f option, diag will let user reprogram the NVRAM. Impact: It will have impact when user tries to program a blank eeprom/flash or recover a corrupted NVRAM image. 2. Problem: When user try to edit bootcode image using -meep command line options, diag will take "\n" character as part of the input parameter. Cause: This issue was introduced since version v7.00. Since gets() lib function in diag has been replaced by fgets() lib function, "\n" character will be included as part of the input string. Fix: Filter "\n" out from the input string before process the data parameter. Impact: As described above. version 7.04 Bug Fix: ======== 1. Problem: "V" Option in dmaw and dmar commands is not working when user specified -b (byte swap) option. Cause: "V" Option in dmaw and dmar commands will cause diag to verify data between SRAM and Host memory. Wehn user enter -b optioin, diag did not byte swap data before comparing data from SRAM and Host memory. Fix: Wehn user enter -b optioin, diag will byte swap data before comparing data from SRAM and Host memory. Impact: As described above. 2. Problem: reset -w did not for bootcode to complete loading. Cause: This issue was introduced since v7.01. When user enter -w options, diag did not wait for bootcode to complete loading. It is dur to the origanal logic was broken by the new added debug code. Fix: The issue has been fixed such that diag will wait for bootcode to complete loading if -w option has been used. Impact: As described above. version 7.03 Bug Fix: ======== 1. Problem: (CQ 8606) Diag allow 1000FD PXE speed on copper cards. Cause: Diag allow user entering invalid speed selection for PXE and MBA. Speed of 1000FD is only valid for fiber card. Diag did not check card type before executing selected option. Fix: Diag will check for card type before executing selected speed option for PXE and MBA. It will only executing 1000FB speed option for PXE and MBA if target is a fiber card. Impact: As described above. version 7.02 Bug Fix: ======== 1. Problem: (CQ 8461) Customer requests to modify the way diag and udiag handle corrupted NVRAM case. Cause: In command line mode, diag and udiag will try to recover NVRAM content when it detects incorrect Magic Value or incorrect checksum of the NVRAM contain. In case of diag running in engineering mode or -cmd option switch has been enter when running udiag, both softwares will not try to recover NVRAM content. Fix: Diag and udiag will display error messages when corrupted NVRAM content is detected. Then diag will exit and no action will be taken. In case of diag running in command line mode, diag will suggest user to reprograme NVRAM. Impact: As described above. version 7.01 New Addition: ============= 1. Request: Added incremental length option in dmaw and dmar cmds. Change: Added -i option in cmd dmaw and dmar. With this option, diag will dma read/write start with 60 bytes to a maximum bytes count specified by -l option. -l option default to 256. 2. Request: Remove 1G lines open/show loop back test (D4) from nic test suite. Change: Test D4 (screening for 1G Tx/Rx lines open/short) has been removed from nic test suite. D5 test (1G False Carrier) will become D4. version 7.00 New Addition: (Support 5705F) ============= 1. Request: Support 5705F Change: Added 5705F Device ID to diag. version 6.31 Bug Fix: ======== 1. Problem: (CQ 8445) 5701 fiber card failed B2 (DB test) when running diag v6.30 Cause: In 5701 fiber card, bootcode will monitoring the link change. If there is link change, bootcode will initialize the mac to negotiate for link and modify the descriptors. During DB test, even cpu has been stopped, mac may still try to modify descriptors. As a result, DB test failed. Fix: Diag will stop cpu and reset mac block before running test to prevent mac accessing the DB block. Impact: As described above. version 6.30 Bug Fix: ======== 1. Problem: Expansion ROM test failed during ikos emulation. Cause: Emulator was responding to slow and incorrect data is read. Fix: Added a delay before each expansion ROM read for emulation. Impact: As described above. 2. Problem (CQ#8288): CPU hangs post Expansion ROM testing. Cause: The memory arbiter is reset during CPU execution. Fix: Halt the CPU, reset the memory arbiter and restart the CPU. Impact: As described above. New Addition: (CQ#8350) ============= 1. Request: Add a field syntax for "ENABLE PHY AUTO POWERDOWN" in eeprom.txt file. Change: Diag has been modified to allow user to change the "ENABLE PHY AUTO POWERDOWN" in input text file. The keyword is: enable_auto_powerdown = yes/no version 6.29 Bug Fix: ======== 1. Problem: B5 test (mbuf sram test via dma) screen displayed incorrect pattern info. As the screen display diag was testing all zeros pattern, in fact it was test 16 zeros and 16 FFs pattern. Cause: Pattern was selected by passing a pattern index to the pattern generator. However, incorrect index has been used. Fix: Diag will pass a correct pattern select index to the pattern generator such that the screen will not display incorrect pattern info. Impact: As described above. version 6.28 Bug Fix: ======== 1. Problem(CQ#8285): Commands in B57UDAIG cause system to freeze at bios boot. After entering the following commands and reboot system, system hang after memory check. The system is a lom system and pxe code may be in BIOS. b57udiag -mba 1 b57udiag -mbas 4 Cause: When user enter -mba 1, -mbas 4 option switch, diag will update nvram and enable mba and set the speed to 4. In case of no mba code in nvram, firmware may provide incorrect info to BIOS and case system to hang. Fix: Diag will check if mba/pxe code is loaded in nvram before enable the bits in nvram. Impact: As described above. 2. Problem(CQ#8287): elog functionality not work per spec. When user enter 'b57udiag -elog out.txt' b57udiag.exe is not performed and only Help command is displayed. Cause: elog option switch was not enabled for b57udiag.exe. Therefore, b57udiag will not take 'elog' as a valid option. Such that Help command was displayed. Fix: elog option switch has been added to b57udiag. Impact: As described above. 3. Problem(CQ#8211): The -mbap option prints out an incorrect error message and executed correctly. Cause: In the code the wrong bit field was being used for the -mbap option error checking. Fix: Changed the code to use the correct bit field. Impact: As described above. version 6.27 Bug Fix: ======== 1. Problem: (CQ #8259, #8273) The CPU test and VPD fails on some 5701 systems. Cause: Debug code was left in by mistake on v6.26. Fix: Removed the debug code. Impact: As described above. 2. Problem: (CQ #8258) The secfg command had labels swapped for the options 2 & 3. The Power Dissipated (2) and Consumed (3) "D" states were swapped. Cause: The printf statement has always printed these options swapped. Fix: Modified the printf statement so the "D" states are correct. Impact: As described above. 3. Problem: (CQ #8237) Add device id to -geneep or expose -meep to customers. Cause: There is no "official" mechanism to update the device id within a bootcode image Fix: Diag has been modified to allow user to change vendor id and device id in input text file. The keyword is: system_vendor_id = xxxx system_device_id = xxxx Impact: As described above. 4. Problem: (CQ #8276) DOS Diag: Diag needs to return ErrorLevel for invalid cmd line switch. Cause: There is no error level has been returned for invalid cmd line switch. Fix: Diag.exe will return 140 when an invalid command line argument is entered. The same has been done for b44diag. Impact: As described above. 5. Problem: (CQ #8276) Diag does not exit test when NVRAM Contect test fails. Cause: Exit diag when NVRAM Contect test fails has been removed for some special test, but never put back after that. Fix: The b57diag64 will not continue testing eventhough the NVRAM content check has failed. diag will exit properly with error code. Impact: As described above. version 6.26 Bug Fix: ======== 1. Problem: Hardware fixes in 5704 B0 silicon did not get enabled for the target port but the partner get enabled in verson 6.25 Cause: After GRC reset on target port, diag will enable HW fixes for the target port and the partner port. PCI config Reg 0x4 of the target port got clear by reset, therefore, diag lost memory access. As a result, hardware fixes did not get enabled on the target port. Hardware fixes only got enabled on the partner port. Fix: After GRC reset, diag will restore memory access (write 0x6 to PCI Cfg Reg 0x4) before enabling hardware fixes. Also, Bit 27 of Reg 0X64 will not be set. Impact: As described above. 2. Problem: (CQ8242,8234) 5704SB0 Fiber fails C5 VPD: Read Test intermittently. 5704SB0 Fiber fails C5 VPD: Loading Firmware test intermittently. Cause: During VPD test, diag did not stop cpu when loading VPD firmware code. Since CPU was in the unknow state, loading VPD firmware without stopping the CPU might have unexpected result. I had found the PCI config Reg 0x4 has been clear and memory access has been disabled. From PCI bus trace, there was not any write to PCI config space register 0x4 to clear this register or any unexpected GRC reset during VPD test. However, the unexpected cpu activities may cause chip to reset itself such that Diag lost memory access as a result. Fix: Stop CPU before loading VPD firmware code can prevent unexpected cpu activities. Issue cannot be reproduced after 6800 Iterations. Impact: As described above. 3. Problem: (CQ8242,8234) b57diag64 IPMI pass-thru code programming problem. Using firmware 3.19, IPMI pass-through 2.08, and b57diag64 v6.18. The following issues had been found. 1. Chip/file mismatch: Getting the error message that says the file doesn't match the hardware (i.e. fiber vs. copper mis-match). Firmware type had been verified. 2. VPD write test failure: VPD write failure on the first iteration of the VPD test (the only time it does the write test). 3. The option to force a write test doesn't work with the vpdtest command; even with that option, it only attempts to write once. 4. Programming failure: Intermittently fail to program the SEPROM properly. Getting a subsequent checksum test failure. Cause: 1. When the IPMI pass-thru code was origenally running, it would enterfere the mii access from diag. As a result, diag would get incorrect phy type and thinking the chip was a fiber chip. As a result, diag will fail code/image type matching When programming new code. This will not happen if there is no IPMI pass-thru running origenally. 2. During VPD test, diag did not stop cpu when loading VPD firmware code. Since CPU was in the unknow state, loading VPD firmware without stopping the CPU might have unexpected result. VPD write test failure can be related to not stopping CPU before loading VPD firmware code. 3. Forcing write test in VPDtest will write to NVRAM. If running force write test in VPDtest in many iterations will kill NVRAM. As a result, force write test in VPDtest has been removed. 4. Same as item #2. Fix: Stop CPU before loading VPD firmware code and boot code can prevent unexpected cpu activities. Also, stop CPU activities after programming firmware can also prevent CPU interferes diag's normal funcitons. Therefore, CPU will be stopped before and after firmware programming under normal conditions. This will not be applied to any test, such as CPU test and VPD test, that require cpu running. Impact: As described above. version 6.25 New Addition: (Support 5704 B0) ============= 1. Request: Enable hardware fixes in 5704_Bx parts for known 5704 Ax errata. Change: Added code to driver init routines to enable hardware fixes in 5704 B0 silicon. See 5704 errata for more details info about the changes/fixes in B0 silicon. version 6.24 Bug Fix: ======== 1. Problem: (CQ8109) While running the MII test (C4) in the Broadcom diagnostics on a 5700 the test will eventually either fail or the system will hang. Cause: Delate in between polling for mii read/write completion has been reduce to 25us. It will increase polling frequency. As a result, system may fail MII test. Fix: Increase delay from 25us to 50us, problem goes away. Impact: As described above. version 6.23 Bug Fix: ======== 1. Problem: (CQ #8083) In the "-n" manufacturing mode diagnotics can print out an incorrect VID. Cause: When the NVRAM is programmed in manufacturing the VID, DID, SVID, and SDID are updated in the NVRAM and the device and not in the global memory copy that diagnostics keeps. Fix: Before the VID, etc is printed out diagnostics now reads the devices value and updates the global memory copy. Impact: As described above. New Addition: ============= 1. Request: Want the false carrier test to be executable in engineering mode via "carriertest" after the "loadd" command is executed - this is done to allow the voltage register to be changed after the "loadd" and then execute the false carrier test - initially "nictest d5" was used to run the false carrier test but the "loadd" and the false carrier test are automatically executed one after the other and that would clear any voltages that were setup before the execution of the "nictest d5" command Change: Added code to achieve the above. version 6.22 Bug Fix: ======== 1. Problem: (6167) The 'blast -q' option, which adds software CRC-32 on trasmitting packets, creates oversized packets when trying to send 1518 byte sized packets making the packet size 1522 bytes instead. This only happened to the 1st 100 packet. Cause: -q option did not get invoked in the 1st 100 packet and software did not generate the CRC. It is a software bug. Fix: Software will generate the CRC for every packet if -q option is entened. Impact: As described above. New Addition: ============= 1. Request: (CQ #7657) Added an external loopback test with the following features: - check for a false carrier at the end of the test - allow a user selected number of packets to be transmitted - allow the user to display the packet count - allow the user to set an error limit The user manual documents the details of these options Change: Added code to achieve the above. 2. Request: (CQ #8012) Added the -ckdev option x = check chip rev y = metal rev If xy does not match the device being accessed diagnostics exits. Use this option on a production line to verify that only chips of a specific rev are run with diagnostics and all others fail. Change: Added code to achieve the above. 3. Request: (CQ #8041) The '-did' command now also displays the PXE version. Change: Added code to achieve the above. 4. Request: (CQ #8041) When diagnostic is run in the manufacturing mode, continually looping testing new devices, the "-dids" information is printed at the end of testing each device. This display is useful so that manufacturing personnel can verify the information of the device is correct. Change: Added code to achieve the above. version 6.21 Bug Fix: ======== 1. Problem: (7785) diag display speed UNKNOWN when set pxe speed to 1000FD. Cause: Since PXE code did not support 1000FD before, diag is not able to decode the new pxe speed. Fix: Added support of 1000FD PXE speed in diag. User is able to display and set PXE/MBA speed to 1000FD for fiber mode Also modified the few error messages labels. Impact: Secfg, setpxe and setmba cmds are modified to support 1000FD (fiber Mode) PXE/MBA speed setting. New Addition: ============= 1. Problem: Secfg cmd provides user a list of item in the EEPROM that user can modily. Item 18 (voltage setting) is only apply to 5701. However this Item has still been shown as a valid item for other chips. Change: Secfg will only show Item 18 as a valid item for 5701. version 6.20 Bug Fix: ======== 1. Problem: (7785) diag display speed UNKNOWN when set pxe speed to 1000FD. Cause: Since PXE code did not support 1000FD before, diag is not able to decode the new pxe speed. Fix: Added support of 1000FD PXE speed in diag. User is able to display and set PXE/MBA speed to 1000FD for fiber mode Impact: Secfg, setpxe and setmba cmds are modified to support 1000FD (fiber Mode) PXE/MBA speed setting. version 6.18 Bug Fix: ======== 1. Problem: (7764) -meep switch does not support boot code larger than 8K byte size. Cause: The issue is due to a software array be limited to 8K. In case of the fiber boot code that has 12K size, not the whole file be stored. Fix: Once the this array size increase to 16K, it allow the the whole file be stored and the issue is fixed. Impact: All boot code has size larger than 8K byte will be affected. version 6.17 (May 09, 2003) Bug Fix: ======== 1. Problem: In engineering mode, when blast 1514 byte packet with a 64 bit 00000000FFFFFFFF pattern System will hang. Cause: When the diag fill up the memory with the 64 bit pattern, the software byte counter of the for loop did not increment correctly and course memory overflow and hang the system. Fix: Corrected for loop index increment. Impact: It hang system when running with any 57xx chip. New Addition: ============= 1. Problem: (CQ 7754) Customer has requested to add a -n parameter to the sechk parameter within b57diag64 in engineering mode. This would allow the test to run with the following behavior: - Supress any prompts resulting from executing 'sechk'. - Do not automatically correct for an invalid checksum - The ability to display and log the sechk results should remain intact. Change: Added code to achieve the above. version 6.16 (May 01, 2003) Bug Fix: ======== 1. Problem: (CQ 7676) Diag show DEF lock up error message when testing Fiber cards (TBI_MODE). Cause: Diag applied Phy reset work around to 5703, 5704 in TBI_MODE. This will cause the phy reset work around to fail and display DEF lock up error message. Fix: Apply the Phy reset work around to 5703, 5704 and 5705 in Copper Mode only. Impact: As described above. version 6.15 (April 29, 2003) Bug Fix: ======== 1. Problem: 1000BT external loopback test fail on 5704/CIOBE due to no link. Cause: 5704/CIOB-E phy reset occasionally will cause Phy to lock up. Once the phy has been locked up, it will not link up at 1000BT. Phy reset is needed to unlock the Phy. The same can be found in 5703 and 5705 Fix: Apply the same work around that has been introduce in V6.14 to 5703 and 5705 Impact: As described above. 2. Problem: (CQ #7652) When there are multiple different devices in a system and an NVRAM update is done, the update stops at the first device that does not match the ID in the image file. There can be other devices that match the image file ID that do not get updated. Cause: When a device ID does not match a file image ID diagnostics exit instead of continuing. Change: Changed code to continue checking for device ID that match the ID of the NVRAM file ID after an unmatched device ID is found. Impact: As described above. version 6.14 Bug Fix: ======== 1. Problem: (CQ #7583) The NVRAM and VPD test will fail if diagnostics is run for many iterations. Cause: The life cycle of the NVRAM will be exhausted if the NVRAM test (C1) is run for an extended period of time. Change: Modified the code so that write to the NVRAM is done only on the first iteration when many iterations are executed. Impact: For life cycle testing use the diagnostics command "setest" in a loop. The diagnostics command "nictest c1 -n=0" will only run the test write operation on the first iteration. 2. Problem: 1000BT external loopback test fail on 5704/CIOBE due to no link. Cause: 5704/CIOB-E phy reset occasionally will cause Phy to lock up. Once the phy has been locked up, it will not link up at 1000BT. Phy reset is needed to unlock the Phy. Fix: New routine has been added to do phy reset for 5704/CIOBE. This routine will issue PHY_RESET and check if the reset is successful. If not, another PHY RESET will be issued, until max "retry" reaches. Once the max retry has been reached, a warning will be displayed and test will continue. Impact: Once a Phy lock up has happened and cannot be unlocked by issuing phy reset, chip will not link up at 1000BT. It will fail the 1000BT external loopback test. New Addition: ============= 1. Request: (CQ #7398) When an invalid DOS command line option is entered, diagnostics displays help screens showing valid command line options. A customer requested that the diagnostics banner be printed before the display of the valid commands. Change: Added code to achieve the above. version 6.13 Bug Fix: ======== 1. Problem: (ClearQuest #7429, #7464) The CPU test fails on some 5701 systems. Cause: Diagnostic does not wait long enough for the CPU to finish setup. Fix: Modified the code the increase the CPU wait time. Impact: As described above. 2. Problem: (ClearQuest #7428) Diag fails on 5700 - Mac loopback Cause: Diagnostic does not wait long enough for the CPU to finish setup during load driver. Fix: Modified the code the increase the CPU wait time. Impact: As described above. New Addition: ============= 1. Request: (ClearQuest #7439) When external loopback test D3 and D4 test fail provide different error codes for failures at 10/100/1000 Mbit/s. Fix: Added code to achieve the above and documented new error codes. version 6.12 Bug Fix: ======== 1. Problem: Some tests, such as ASF Test, MSI Test and Reg Test has been skipped for all b57xx except b5788. Also HW_init will initialize chip the same way as 5788. Cause: The logic that indecate 5788 is the target device will always return 1. Fix: Diagnostics is modified such that the logic will truely indecate if 5788 is really the target device. Impact: As described above. version 6.11 Bug Fix: ======== 1. Problem: (ClearQuest #7217) On some machines the 5901 is incorrectly classified and the gigabit loopback test is run (should be bypassed) and fails. Cause: During a driver load the chip is reset but the diagnostics does not wait for the firmware to complete (setup the PCI registers) and this causes the 5901 to be classified incorrectly. Fix: Diagnostics is modified so it now waits for the firmware to signal that it has completed execution (setup). Impact: As described above. 2. Problem: (ClearQuest #7264) CPU Test of Dos Diag fails for 5700B3 LOM on Dell 6600. The failure displays, CPU reset failed, register 5034 is 0x1d. 1)The problem does not occur in b57diag64 v6.07 2)The problem occurs in b57diag64 v6.08, v6.09, and v6.10 3)The problem is seen on both 5700B3 LOM Cause: Diagnostics does not wait long enough for the firmware to complete during the setup phase of the CPU test. Fix: Increase FIRMWARE_WAIT_TIME from 150 to 170 ms to allow firmware has longer wait time to load bootcode. Impact: As described above. New Addition: ============= 1. Request: (ClearQuest #7288) A customer requested that an error message is printed and an error code of 1 is returned when an user enters an incorrect number when selecting a card to run tests against from the DOS command line. Fix: Add code to achieve the above. version 6.10 Bug Fix: ======== 1. Problem: (ClearQuest #7217) A 5901 is recognized as a 5705 during manufacturing and the gigabit loopback test is run and fail because the device is a 5901. Cause: In manufacturing diagnostics programs the device's firmware and then runs the manufacturing tests. Before firmware is programmed the 5901 is classified as a 5705. After firmware is programmed diag. did not re-classify the device as a 5901. Fix: Added code re-classify the device after the firmware is programmed. Impact: As described above. version 6.09 Bug Fix: ======== 1. Problem: Software mistaking 5705 device as 5788 and not allow user to update 5705 bootcode. Cause: The logic that indecate 5788 is the target device reported incorrect info. Fix: Diagnostics is modified such that the logic will truely indecate if 5788 is really the target device. Impact: This has been introduce in version 6.08 version 6.08 1. Added support to bcm5788. Bug Fix: ======== 2. Problem: (ClearQuest #7155) User requested the following modification to the -elog option: Create the error log file only if an error occurs. If the error log file exists append to it. Cause: Modification Request. Fix: Added code to fulfill the user request. Impact: None. New Addition: ============= 1. Request: (ClearQuest #7171) User requested that the ability to print out the DID,VID, SDID, VID, firmware, PXE, PXESpd, WOL, ASF and MBA options from a DOS command line option. Fix: Added the DOS command line option -dids to provide the feature. version 6.07 1. The space between the file name an -f option in 'upgfm' is no longer needed. Bug Fix: ======== 2. Problem: (ClearQuest #7155) When viewing a log file there is not way to tell if a user stopped the testing procedure via a . Cause: When the user presses logging is stopped. Fix: When a user presses the string "User Aborted Testing" is logged to the log file. Impact: None. 3. Problem: (ClearQuest #7171) When a log file is created and the flash update option is selected then an incrementing address counter is displayed on screen and all these incrementing values are written into the log file. Cause: The code does not disable logging of the flash address counter. Fix: Disable logging of flash address counter. Impact: None. New Addition: ============= 1. Request: (ClearQuest #7155) User requested that an error log file is generated when an error occurs. Only error information is logged (NOT pass information). Fix: Added DOS command line option "-elog " to provide the feature. version 6.06 Bug Fix: ======== 1. Problem: (ClearQuest #6748) The diagnostics command 'memtest -c" fails on the 5705. Cause: The 5705 memory map is different than the 5700/01/02/03/04. The present code does not take this into account. Fix: Use the 5705 memory map NOT the 5700/01/02/03/04. Impact: This test failed on the 5705 on previous version of diagnostics. 2. Problem: (ClearQuest #7171) If a user used the diagnostics command line option "-m" to program the MAC address the user HAS to enter the MAC address. If a user ESCapes out of entering the MAC address the NVRAM gets corrupted. Cause: The MAC address entry routine was not coded to handle the ESCape sequence correctly and corrupts the NVRAM. Fix: Modified the MAC address entry routine so that when a MAC address is not entered (ESC pressed) the NVRAM is not corrupted. Impact: This "-m" corrupted the NVRAM on previous version of diagnostics. 3. Problem: (ClearQuest #7171) When "b57diag64 -c 0 1 -e b57kia -m -n -t abcd" is run on a 5704 the user is not allowed to exit out of diagnostics after the first execution loop. The user is force to run the diagnostics loop two time on each device before he is allowed to exit. Cause: Since the 5704 is a dual device the command above should run twice (once for each device). Due to a coding error the user is forced to run four times (twice for each device). Fix: Modified the code so the above test is run once per device. Impact: Described above. New Addition: ============= 4. Request: (SQA request) Users would like to update the PXE code from the DOS command line using b57udiag.exe. Fix: Added the DOS command line -pxe option to diagnostics so that the PXE code can be updated. Impact: None version 6.05 1. Enable swapXBAR option in SECFG when 5404 B0 is detected. 2. Added 5703B0 ASIC Rev ID Bug Fix: ======== 3. Problem: (ClearQuest #7087) The diagnostic DOS command line command option "-firm" fails when programming multiple cards. Cause: In order to program the NVRAM a mapping has to be done from the file image to an NVRAM image. When programming multiple cards the mapping routine was mistakenly called multiple time. The first call correctly preformed the mapping and the first card is programmed correctly. All remaining call corrupted the NVRAM image and the following card are programmed incorrectly. Fix: Call the file image to NVRAM image mapping routine once. Impact: Previous version of diagnostics will incorrectly program the 2nd-nth card. version 6.04 Bug Fix: ======== 1. Problem: (ClearQuest #5898) Warning text is added to the D4 test. The warning text reads "(** ZERO LEN LB RJ45 **)". Cause: Test D4 (screening for 1G Tx/Rx lines open/short) only works with zero length (ie short) RJ45 loopback connectors. Fix: Text change in code. Impact: None. version 6.03 Bug Fix: ======== 1. Problem: (ClearQuest #6619) When the NVRAM is updated via the DOS command line "-firm" command the NVRAM update fails. Cause: The NVRAM image file was incorrectly mapped to the NVRAM. Fix: The NVRAM image file is correctly mapped to the NVARM. Impact: Image loaded to NVRAM with the with previous version of the "-firm" command would not work correctly. New Addition: ============= 1. Request: (ClearQuest #5898) Customer requests to test for 1G Tx/Rx lines open/short. Fix: Inserted and coded test D4 for 1G Tx/Rx lines open/short testing. Impact: Original D4 and D5 test now change to test D5 and D6. version 6.02 Bug Fix: ======== 1. Problem: (ClearQuest #5898) D3 (external loopback) test was modified so the it could check for shorts and opens on the 1G lines of short RJ45 cable. This test failed with long cables. Cause: Screening for open and short can be done only on short cables. Fix: Open/short screening has been removed from the D3 test so that the D3 test would pass with long cables. Impact: Open/short testing not supported. 2. Change copyright string from Broadcom to OEM. version 6.01 1. Fixed setorture command failure. Bug added in code since v5.03 where the firmware bootup handshaking is tested for twice, which causes the incorrect failure. 2. Bounds checking for udiag command line options added. 3. Added Support to 5705 A2/A3, 5901 A2/A3, 5705M A2/A3. 4. Added OEM NC7761 Gigabit Server Adapter support. 5. Modified Phy Interrupt test. Since the purpose of this test is to check if phy can update our interrupt status bit (to generate interrupt) and we do not want to use actually interrupt to test this feature we were polling the interrupt status. Since some systems will randomly call our interrupt service routine and clears the interrupt status, caused our polling routine to fail to see the interrupt status. The workaround was not only polling for interrupt status in the register, but also checks if actually interrupt service routine has cleared the status. If the interrupt service routine clears the status, it will increment a counter to indicate that the status was cleared by interrupt service routine. 6. Removed "Disable Expansion Rom Test" from Expansion Rom test. Once the Expansion ROM is "disabled" on the system, when the diag is accessing to the Expansion ROM, it in fact, is accessing to "non-existing" memory address. Depending on the system design, the system reaction to the "invalid memory access" may be different, and it is very possible to hang the system. version 6.00 1. Customer requested that DOS command line execution of the loopback test D3 allow selection of line speed. Line speeds for this test can be set via the command line option "-lbspd t/h/g", where t/h/g is ten/hundred/giga line speeds. To run ten and giga LB "b57diag64 -t abcd T D3 -lbspd tg" 2. A customer requested a DOS command line option to program the firmware on a device, only if the did, vid, sdid, & svid of the device matches the information in the firmware file. Use "b57udiag -firm abc.bin" to program all device on a system where the device did, vid, sdid, & svid matches the same information in the firmware file "abc.bin". 3. 5704 support has been added to q57udiag. version 5.18 1. Tx/Rx 32 bit packet counters are extended to 64 bits, to allow for longer test runs without a counter rollover. 2. -w and -b options take no effect in dmaw cmd in engineering mode. These had been fixed. 3. HP NC7761 Gigabit Server Adapter PCI_ID has been added. 4. The 5705 would not catch a open on gigabit Tx/Rx lines during loopback. Tap2, mean square error, on the DSP is used to screen for opens. 5. Stopping "blast -t -r -n=0" via ESC would stop the Tx/Rx simultaneously. If packets are loopedback (LB) some packets would not be RXed due to the LB delay. Fixed by first stopping Tx, then waiting, then stopping Rx. version 5.17 1. Added 5704 A3 bond ID and Rev ID. version 5.16 1. Extended 5705 status block non-error counters in software. 2. Modified 5705 reset sequence required for A2 chips. 3. Fix a bug in accessing flash while flash is not supported which will cause system hang. In bug is at the end of the CPU test. 4. Modified enabling Clock Run function version 5.15 1. Fixed incorrect qstat cctet count for 2nd-nth 5705. 2. In q57udiag, -mbas speed option will only support auto for Fiber cards 3. In q57udiag, setmba and setpxe cmds will only support auto for Fiber cards 4. Fixed multiple do with parameter-passing bug 5. All loopback tests are using hw counters instead of software interrupts counters. version 5.14 1. Mask out the all W2c bits in offset 0x6 of PCI CFG registers test. version 5.13 1. Increase tx wait time in external loop back to allow all packets can be tx. version 5.12 1. Fixed CPU test failure with correct reset sequence version 5.11 1. Enable wol En/Dis options in q57udiag 2. Removed 5704 support in q5udiag. 3. After reset, software will enable clock-run bit in PCI cfg space if 5705 detected. version 5.10 1. Added delay at the end of cpu test to let cpu fully complete loading fw. 2. Reduce the frequency of polling register during reset to avoid fighting between CPU and host. version 5.09 1. "loadd" cmd will config link speed depend on card/chip type (FE/GE). 2. Fixed bug when running "blast" consecutively without a chip reset. The bug was introduced in 5.07. 3. Removed cpu_mem_access test per Tak 4. Modified CPU test util_chip_global_reset to wait for CPU loadd fw completed. 5. Fixed bug when running "blast -t" where TXing eventaully stops. 6. Modified PCI cfg read mask to skip Bit 15 of offset 0x4c since this bit is w2c. 7. Enhance program ASF FW function in both engineering and manufacturing mode. Version 5.08 1. Added 10 and 100 Mbit/s operational mode to the command line options -tr and -hlb (pg. 13 in manual) 2. Added 5782 support. 3. Added new PCI Device for 5702 A3, 5703 A3 Version 5.07 1. Added command line option -tr (pg. 13 in manual) 2. Added command line option -hlb (pg. 13 in manual) 3. Made changes to the blast command to support above options 4. Modified the way that the diag checks for ASF version. (fixes ClearQuest 6113) 5. Modified the way to reset the chip. 6. Increase Phy reset time out. 7. Modified 5705 PCI cfg register 0x74 read mask. 8. Fixed standalone_exe flag. 9. Unset DMA once bit in case of 5701 running in PCXI mode. 10. Added support to 5705 Version 5.02 1. Added write function 2. Added option -j(jumbo packet) and -t(tcpseg) for loaddrv command 3. Overhauled t3_mem_pool_init routine for more memory allocation options 4. Fixed 5705 ROM Size Error 5. Added Intelligence Mechanism to fw / PXE updrades. Version 5.01 1. Added 5704 A2 NVRAM write enable support 2. Added Configuratin check for TT mode to very secondary port ASF/PXE/WOL enable 3. Added Cofiguration check for NIC mode to warn user for both port set to 100mpbs 4. Finalized FE->GE feature with the latest spec. 5. Fixed checksum in sechk command for ASF. 6. Fixed cheksum in dir command for ASF. Version 3.24 1. Fixed LED mode when loading driver 2. Added CPU debbuging command 3. Added Internal CPU GPR test 4. Added delay for both dmar & dmaw & bustest done polling 5. Added support for both CPU on Internal CPUGPR test. Version 3.23 1. Fixed Indirect register test on oem system 2. Fixed asfcfg backward compatibility 3. Added delay after voltage change. 4. Fixed 5704S voltage setting value. 5. Added lowWaterMarkMaxRxFrame for txpkt and blast command Version 3.22 1. Added multiple device DMA (bustest, dmar, dmaw) 2. Fixed dmaw/dmar -c option bug 3. Added detection for 5704 in oem products. 4. Fixed A5 PCI config. register test on 5704. Version 3.21 1. Fixed 5700/5701 failure on oem machine. 2. Added delay for cpugprtest 3. Change voltage control in cpu general purpose register test. 4. Added 5704S Bond id support 5. Added Updated Bond id check for 5700 6. Fixed Voltage regular control correctly for 5704 secondary device 7. Removed debugging code. 8. Modified Mac source address during tx/rx 9. Added 5703SA2 Bond Id fix Version 3.20 1. Added loopback packet options. 2. Added multiple boot agent options (-mba). 3. Added command line options for b57udiag. Version 3.19 1. Fixed bond id checking for CIOB Version 3.18 1. Enhanced cpugprtest to test under v1.1, v1.2, and v1.3 2. Added workaround for 5704 A1 for NVRAM access locking when NVRAM was invalid 3. Fixed a problem where clock control register is overwritten when driver is loadded. 4. Added support for 5704S A1 5. Added support for 5705 (non-mobile version) 6. Added 5705 cputest failure workaround Version 3.17 1. Inverted polarity for Auto Powerdown mode so that default Autopower down is disabled. Version 3.16 1. Removed some debug break points. 2. Fixed 5704 eeprom programming. 3. Fixed a bug causes 5703A3 to always read 133MHz PCI speed. 4. Added more statistics support for qstats commands. 5. Fixed an issue where Tx flow control cannot be set with drvrcfg command. 6. Added 5901 support. 7. Added Auto powerdown mode support. Version 3.15 1. Changed so that Phy interrupt test is skipped for BCM5705 A0. 2. Added Mini-PCI configuration in secfg command. Version 3.14 1. Fixed memory allocation in pxe programming code. 2. Added 5702fe exclusion for gig external lpbk. Version 3.13 1. Added 5705 support. 2. Fixed a memory allocation error on some systems. Version 3.12 1. Embedded bist data into codes to reduce number of released files. 2. Corrected some spelling errors. 3. Fixed -f option to program eeprom image with size greater than 8k. 4. Added ASF command line to program ASF firmware. 5. ASF firmware files to be read from a text file. 6. VPD Write test to run only once if in continuous loop. Version 3.11.0 1. Added commands "dir", "asfprg" for end user purpose. 2. Fixed b57udiag freezing problem. Version 3.10.0 1. Fixed Link polarity problem in loopback modes. 2. Implement Power Saving option for firmware and driver use. Version 3.09 1. Fixed a bug when both -m -f are entered in 5704. 2. Fixed output messages when -fmac option is in use. 3. Fixed problem with EFI hanging when clearing nic statistics Version 3.08 1. Automatically skips serdes loopback if external link is detected in fiber. 2. Fixed help command in engineering mode. 3. Fixed -fmac command line. Version 3.07 1. Enhenced setest command. 2. Fixed reset -w option bug 3. Skipped ROM/PCI config regi. test for secondary device in Daul MAC mode 3 for 5704. 4. Create the first version of short diag -- b57sdiag 5. Changed Altima indirect access method workaround only specific to A2 revision 6. Changed to dynamic firmware debug printf 7. Fixed bustest in ikos system (need more time before checking data) 8. Added help menu and miscellaneous small changes per oem's request Version 3.06 1. Fixed serdes loopback failures on some product 2. Added filename support for -geneep option Version 3.05 1. Changed (shorten) delay time for 5704 NVRAM arbitration timeout 2. Fixe Fiber internal loopback problem. Version 3.04 1. Fixed serdes loopback failures on 5701 2. Fixed Phy loopback failure on 5700 Version 3.03 1. Fixed Altima 1011/1012 screening process power on/off problem 2. Added tx/rx octal display on qstat command. 3. Changed stsblk command display format from 10-digit-number to 15-digit-number 4. Added manufacturing mode, concurrent nvram programming and pxe update capability Version 3.02 1. Added workaround for 1011/1012 2. Fixed some indirect memory access bugs Version 3.01 1. Added Fiber Voltage workaround. 2. Added -fmac option to program mac address from a file. Version 3.00 1. Added compile option for b57udiag (user diag) 2. Code size reduction Version 2.39 1. Added phy crc counter to engineering blast 2. Fixed multiple interrupt problem on 5700 B2/B1/B0 3. Fixed failure when WOL is enabled 4. Added prevention to program wrong eeprom.bin to device Version 2.38 1. Added cread/cwrite command 2. Fixed broken phy workaround created in v2.37 3. Fixed sedump command to support ATMEL Flash Version 2.37 1. Added 5704 support 2. Added 5704 A0 workarounds 3. Added geneep new configuration commands 4. Fixed bug for chicken and egg problem for LOM write protection Version 2.36 1. Fixed flash corruption failure. 2. Fixed indirect register test failure. Version 2.35 This version does not support more than 3 devices on a PCIX system. It would result in register test A2 failure if 4 or more devices are in the system. 1. Fixed dma test problem in 5701/pcix. 2. Fixed a failure on phy interrupt test. 3. Fixed a failure on Rx Cpu test when loaded from the floppy drive. Version 2.34 1. Added secondary device (for 5704) NVRAM configaration support Cold reset instead of warm reset before exit 2. Fixed bustest problem with new DOS extender 3. Added pciinit function call before reset if engineering mode was entered be noinit option to prevent reset & flash access fail problem. 4. Fixed NVRAM test failure 5. Fixed Register Faiure 6. Added legacy eeprom test for sedvt command 7. Added dynamic directory support 8. Fixed some bugs in seeprom cfg which did not save cfg completely. 9. Fixed loopback failure on 5700 for backward compatibility. Version 2.33 1. Added seclock command 2. Added seinit command 3. Changed setorture command display 4. Added -e, -a, -w option to sedvt command Version 2.32 1. Added coverage for 5702FE 2. Handle appropriate link polarity for OEM in PHY loopback test Version 2.31 1. Added memory test patterns to memory test 2. Replaced dos4gw dos extender to pmodew 3. Added fix to eeprom corruption problem 4. Fixed IRQ9 hanging problem Version 2.30 1. Restored the original prompt :> instead of > 2. Fixed function/do command parameter passing bug 3. Fixed version string problem 4. Extended wait time for firmware after reset 5. Fixed ROM_test/VPD_test failure 6. Fixed upgfrm command bug 7. Added -r option for reset/pcii command to not to halt CPU 8. Changed sechksum command Version 2.29 1. Fixed function printf() parameter bug 2. Appended CRC for ASF image programming 3. Changed sechksum command to display every block checksum in the directory 4. Support two phase boot code 5. Changed upgfrm algorithm 6. Changed secfg to support more configurations 7. Added user define function 8. Added parameter passing for user defined function and do commands 9. Added auto malloc()/DPMI memory detection algorithm Version 2.28 1. Fixed problem with firmware response 2. Fixed problem with eeprom corruption 3. Fixed MSI test to avoid device string corruption Version 2.27 1. Fixed atmel bitbang, seread bug 2. Fixed -i option for 5703A1 3. Fixed bug for Gigabit link issue. -- not to force Master all the time. 4. Added 'skips' 5. Fixed EEPROM size detection bug 6. Fixed Rainer 1000 link problem 7. Removed test engineering test B8 from test menu 8. Fixed error control on test D3 9. Add loading escape if ASF hangs 10. Add ASF cold and warm boot simulation 11. Add WOL configuation in menu 12. Fix smbout parameter issue 13. Added Error Control 14. Removed flshprg command 15. In manufacturing mode: ROM_test (C7) is skipped for 5700/5701 running on PCIX-bus. BIST_test (A4) is skipped for 5700/5701. MSI_test (D4) is skipped for 5700/5701. 16. Removed ROM disable test from 5703_A0 device. Version 2.26 1. Changed MBUF water mark for BCM5702/BCM5703. Version 2.25 1. Fixed 5703 register 0x4c00 bit 31/30 error (should be no problem with 5703 A1) 2. Fixed external loop hang problem Version 2.24 1. Added support for bridge detection and testing 2. Changed nictest command. It take "test list" parameter. By defaut, disable EEPROM content check in engineering mode 3. Separated register test for each board. 4. Added pcicfgtest 5. Added bitbang mode for atmel/stt FLASH device 6. Added passthru mode for all atmel/att Flash and seeprom 7. Fixed @ sign (multiple device command parameter bug) 8. Added ROM disable test 9. Fixed scratchpad memory test in conjunction with runing firmware bug. Version 2.23 1. Updated 5703 register test 2. Updated DMA test 3. Added individual register test with back-to-back access 4. Fixed Interrupt test failing on BCM5700 Version 2.22 1. Fixed eeprom programming 4k/8k problem 2. Fixed 5704 register test 3. Fixed receive mask error for 5704 XBar mode 4. Fixed MIItest failure 5. Fixed 5704 xbar mode, interrupt line bug 6. Unconditionally enables RX/TX cpu clock at start up (it may be possible disabled by bootcode for WOL) 7. Added Bist test 8. Added back-to-back access register test option Version 2.21 1. Moved changes from bcmdiag version 2.6 - version 2.9 2. Changed t_writemem16, t_writemem8, t_readmem16, t_readmem8 from big endian to little endian 3. Separated regtest for 5700, 5703, and 5704 Version 2.20 1. Fixed blast -q option bug 2. Fixed ASF eeprom programming gap filling bug Version 2.19 1. Fixed Register test 2. Restore letency timer/cache line value after reset 3. Fixed sleep command 4. Fixed rs232 echoing 5. Fixed register test: regster 690,694 -> 590,594 6. Fixed Broadcast config. 7. Changed IKOS flash clock devider to zero Version 2.18 1. Fixed VPD test 2. 5703.bin update 3. flshdiag.bin update Version 2.17 1. Stablizing tx/rx test 2. Fixed memory allocation bug 3. C5/D4 tests still have to be disabled 4. D2 test has to be disabled due to board design Version 2.16.2 1. command clearbit bug fix 2. added checksum report for bootstrap on sechksum command 3. Added MSI test. Due to chip bug, this test will not pass on 5703 A0 chips 4. Fixed setest failure after running romtest. Version 2.16.1 1. Memory Leak fix Version 2.16 1. Added support commands for engineering mode 2. Added auto-polling test to mii_test 3. Added 10MB loopback test 4. Fixed "Stack Overflow!" bug 5. Fixed romtest/vpdtest 6. Added Backword bcmediag compatibility syntax for upgfrm command 7. Merged board to board test with Fiber Test 8. Fixed command line option -i to seach for invalid device id. 9. Stablized link in TBI/engineering mode. Version 2.14 1. Finalize external memory test coding 2. Added external memory test option 3. Added external memory code execution test into external mem test 4. Fixed wol/pxe enabling/disabling problem 5. Enhanced -x & -w options to accept inputs 0/1 for disabling/enabling Version 2.13 1. Correct FCS error statistic read Version 2.12 1. Fixed a bug in PXE programming code Version 2.11 1. Fixed DMA failure for specific OEM 2. Fixed failure on an OEM PCIX bridge Version 2.10 1. Added detection for specific OEM Version 2.09 1. Removed un-neccessary command line options Version 2.08 1. Correct Silicon revision Version 2.07 1. Fixed external loopback failure in version 2.06 2. Added command line option to view eeprom binary file 3. Disable WOL update function Version 2.06 1. Fixed TBI mode failures on phy_type problem 2. Fixed Specific OEM message 3. Enter OEM manufacture code to handle link polarity in loopback tests Version 2.05 1. Fixed DMA on PCIX failure 2. Take out packet # on loopback tests 3. Change order of tests per customer request 4. Fixed external loopback failure 5. Changed VPD test 6. Added -all option 7. Added -noinit option Version 2.04 1. Fixed copper mac loopback hang problem 2. Fixed D3 loopback link problem Version 2.03 1. Added Phy type/led mode/voltage source/part revision selectability in eeprom.txt 2. Added real/customer rev options 3. Speed up prog. pxe 4. Mac I/O reset Version 2.02 1. Added TBI support Version 2.01 1. Added FORCE_PCI and VOLTAGE_SOURCE to EEPROM.TXT file. 2. Changed secfg command to support FORCE_PCI and VOLTAGE_SOURCE 3. Added -coe to continue on error Version 2.00 1. Fixed IRQ9 problem on specific OEM 2. Fixed Interrupt test failure reported from QSA 3. Fixed CPU test problem reported from Systems Engineers Version 1.29 1. Skip test on NIC card that has ODI loaded 2. Fixed loopback hanging problem by adding delay time to ISR Version 1.28 1. Added generic key code for eeprom generation Version 1.27.1 1. Fixed a bug in manufacturing mode 2. Added compile option for specific oem Version 1.27 1. Added sramtest command to test external memory in engineering mode 2. Added dbgmode to set a debug mode flag in engineering mode 3. Removed xmtest option which is not really needed 4. Added patch in loopback test to lower expectation on reception in case of chip testing 5. Modified external memory test routine 6. Added options debug/nocrc/chip/mac 7. Modified LoadPatchFirmware routine 8. Modified testfailed routine to enter engineering mode if fails and debug option entered 9. Fixed bug in eeprom programming Version 1.26 1. Checked-in the fix for HP shared 9 interrupt issue (enabled back irq9 interrupt) 2. Fixed a bug in loopback tests 3. Fixed IRQ sharing with another device, such as Intel NIC, on memory read problem 4. Added CRC patch from Engineering for 5701 Version 1.25 (Internal Verion) 1. Added -debug option, when it fails jumps to debug prompt 2. Reduced timeout wait time Version 1.24 1. Fixed bug in the load_firmware routine 2. Disabel IRQ9 chaining Version 1.23 1. Added support for BCM5701-A0 2. Fixed interrupt chaining with IRQ9 Version 1.20 1. Changed to jump into command mode when no card detected with option -b57eng 2. Revise options -x and -w as following: When nothing entered, always disable PXE/WOL When -x or -w is entered, always enable PXE/WOL When -f is entered, use eeprom.bin's content for PXE/WOL/PXE speed setting When -f is entered and -x or -w is also entered, force PXE or WOL to enable Version 1.19.3 1. Changed -f option to use eeprom.bin PXE and WOL setting instead of forcing to disable Version 1.19.2 1. Changed software to exit directly to dos when no card was detected. No ESC needed Version 1.19.1: 1. Added Altima/5411 support 2. Added loop command 3. Added mrloop command 4. Added option -geneep, -geneepkey 5. Added option -ver to display current version number 6. Removed mac_pref.txt input file for mac prefix -> now should use eeprom.txt to update eeprom.bin. The eeprom.bin prefix is used if the mac input is less than 12 digit. 7. Added RJ45 external loopback for 100/1000 Version 1.19: 1. Add -xmtest option to increase memory test coverage 2. Reverse some coding on eeprom file dumping that is no long needed 3. Change eeprom programming counter display from hex to decimal 4. Added B7 MBUF memory special test 5. Put back interrupt chaining except irq9 6. Increased MAX_DEVICE from 8 to 16. Now it can support up to 16 devices 7. Fixed bug to restore phy state from isolated mode after test Version 1.18e: 1. Fixed EEPROM access hang when loopback connector is connector is connected Version 1.18: 1. Fixed IRQ 9 test D1 hang problem. 2. Restored LED to LINK3 LEDMODE after phy loopback test. 3. Added command rm. 4. Fixed MAC address display bug. 5. Added Link speed 10 test A1,A2 failure problem with workaround. 6. Changed -bus option to test specified slot only. This option is invalid if -c option is used. 7. Fixed intermitent failure/system hang problem when loopback connector is connected Version 1.17: 1. Change MAC address reading to accept variable length instead fixing to 7 digits. Version 1.16: 1. Changed the parameter misspelling "-dmpi" to "-dpmi" 2. Fixed the EEPROM size 0x204 programming bug due to the structure size define error. 3. Display not only device number but also bus:dev on error messages 4. Fixed Test D1, "Cannot Tx Packet" bug. 5. No sharing on IRQ 9 Version 1.15: 1. Added command line option -dpmi to use DMPI method memory allocation 2. Added command line option -bus bb:dd:ff to test against specified bus & device number 3. Displays device number on error message Version 1.14: 1. Fixed PCI-X eeprom access bug. 2. Invalidate the default xx-xx-xx-00-00-00 mac address. However, it allow user to enter mac address as xx-xx-xx-00-00-00. Version 1.13: 1. Fixed Memory Allocation Problem for Dell 8450. 2. Added PCI-X support 3. Programmed MAC (in PCI-X mode) to not allow set the Relaxed Ordering Bit in the Requestor Attributes of transactions it initiates that do not require strong write ordering. Version 1.12: The new options -n, -m, and -f now requires -e option to be activated 1. New Option -n With this option, the program will run in manufacturing loop. It supports power on/off of UUT. This will ignore the option -I, the iteration option. 2. New Option -m Prompt user for the MAC address. With this option, user must enter/scan the mac address before testing. The program also checks for the file "mac_pref.txt" This is a text file should contain six digits of ASCII MAC three-byte-prefix address. Any of the following format is supported: Example: 000467 00 04 67 00 04 67 If this file exists, user has option to enter/scan 7 digit hex number. The first digit will be ignored and the last 6 digits will be used for the lower part of MAC address. Combine with the prefix, it creates 6 byte (12 digit) hex number. If this file does not exist, the whole 12 digit number must be entered for the MAC address. When entering MAC address, a space character is allow between each byte. For example, any of the following is valid. 000467010203 00 04 67 01 02 03 1010203 (currently, the scanner uses this format) 3. New Option -f Program "eeprom.bin" into EEPROM before testing. The file "eeprom.bin" must be there or the program will not run with this option. The -m and -f combination will create the following behavior: a) -f and -m: Program will not validate the eeprom content and go ahead to prompt user for the MAC address. It programs MAC address and EEPROM content and then checks the validity of eeprom content at the end of programming. Loading EEPROM content from eeprom.bin: passed -- enter MAC address after pass here Programming EEPROM from eeprom.bin....: passed Checking EEPROM content...............: passed b) -f only Program will check the validity of eeprom. If it is not valid, it will act as a), -f -m option. If it is good, it saves the MAC address from eeprom, program new eeprom.bin content into EEPROM and then restores the original MAC address. It checks the validity of eeprom content once more at the end of programming. Checking EEPROM content...............: passed Loading EEPROM content from eeprom.bin: passed Programming EEPROM from eeprom.bin....: passed Checking EEPROM content...............: passed or Checking EEPROM content...............: invalid Loading EEPROM content from eeprom.bin: passed -- enter MAC address after pass here Programming EEPROM from eeprom.bin....: passed Checking EEPROM content...............: passed c) -m only Program will check the validity of eeprom. If it is not valid, it will act as a), -f -m option. If it is good, the program will prompt the user for a new MAC address and program the MAC address only. It checks the validity of eeprom content once more at the end of programming. Checking EEPROM content...............: passed -- enter MAC address after pass here Programming MAC address...............: passed Checking EEPROM content...............: passed or Checking EEPROM content...............: invalid Loading EEPROM content from eeprom.bin: passed -- enter MAC address after pass here Programming EEPROM from eeprom.bin....: passed Checking EEPROM content...............: passed d) no -m and -f options Program will check the validity of eeprom. If it is not valid, it will act as a), -f -m option. If it is good, it proceed to normal diagnostics. Checking EEPROM content...............: passed or Checking EEPROM content...............: invalid Loading EEPROM content from eeprom.bin: passed -- enter MAC address after pass here Programming EEPROM from eeprom.bin....: passed Checking EEPROM content...............: passed 4. Fixed -x -w option checksum false error report bug Version 1.11: 1. Added features to secfg command to allow users to program VPD-R and VPD-W information. 2. Removed all debugging information to reduce file size. 3. Manufacture test includes EEPROM content testing. If manufacture checkusm is bad, it prompts user to enter MAC address. If the checksum it good, it will save the mac address before programming eeprom.bin if user choose to do so. 4. Changed Internal MAC loopback test from 100 to 200 packets. Changed Phy loopback test from 100 to 2000 packets. Changed RJ45 loopback test from 100 to 2000 packets. 5. Removed pattern and altenate pattern test from memory test to increase speed. 6. Removed VPD test to increase speed. Since both eeprom test and vpd test exercise hardware eeprom. There is no needs for extensive test for both. Version 1.10: 1. Removed bit 13 register 11 read only bit test for MII test The MDI Crossover state bit changes internally when there is no cable connected. It is not reliable to be tested as read only bit and it is not an error. 2. Suppressed the error dialog when run without -p option Version 1.09: 1. Moved over bcmediag.exe version 1.24 - 1.30 changes. bcmediag.exe Version 1.30: a. Added bustest to test out PCI bus in the worst case scenarios. The goal is to have every data/control lines toggled. Usage : bustest -a=address -n=iteration -l=min_len -h=max_len -i=transaction -s=start_case -e=end_case -a : NIC address to DMA data to. -l : Minimum length (default = 256). -h : Maximum length (default = 1024). -n : iteration (default = 1) -i : Number of transactions per pattern (default = 10) -s : Start of test case (default = 0) -e : End of test case (default = 259) There are total 260 test cases (258 unique tests cases) which are described as follows: Test case# Pattern ========== ================================== 0 ffffffff ffffffff 00000000 00000000 1 ffffffff fffffffe 00000000 00000000 2 ffffffff fffffffd 00000000 00000000 . . . . . . 64 7fffffff ffffffff 00000000 00000000 65 00000000 00000000 ffffffff ffffffff 66 00000000 00000000 ffffffff fffffffe 67 00000000 00000000 ffffffff fffffffd . . . . . . 129 00000000 00000000 7fffffff ffffffff 130 00000000 00000000 ffffffff ffffffff (repeat) 131 00000000 00000001 ffffffff ffffffff 132 00000000 00000002 ffffffff ffffffff . . . . . . 194 80000000 00000000 ffffffff ffffffff 195 ffffffff ffffffff 00000000 00000000 (repeat) 196 ffffffff ffffffff 00000000 00000001 197 ffffffff ffffffff 00000000 00000002 . . . . . . 259 ffffffff ffffffff 80000000 00000000 If you run bustest command without any parameters, it will perform DMA testing on all 260 patterns with 10 iterations per pattern and different data length in each iteration. First eight bytes of data are used to store the following info for debug: byte 0-4 : length byte 5-6 : iteration# byte 6-7 : test case# b. Added PCI-X support and work-around for B1 silicon. This version diagnostics detects if the NIC is in PCI-X slot. If yes, work-around is implemented. However, user still can overwrite this by issuing debug -c=5 to toggle PCI-X work-around before loading driver. bcmediag.exe Version 1.28: a. Made minor changes to support B1 silicon due to changes in ASIC revision format. bcmediag.exe 1.27: a. Added new commands to utilize BCM5500 built-in BERT features. Three new commands are added: * lbertram: load BERT pattern into BERT TX&RX RAM, and enable BERT. * dbertram: Dump BERT TX& RX RAM. * bertstats: Display BERT statistics. bcmediag.exe Version 1.26 a. Changed initialization of BCM5500. bcmediag.exe version 1.25 a. Added support for Serdes PHY (BCM5500) and TBI interface. b. Added code to allow transmit jumbo frames up to 9022 bytes/packet with txpkt command. (Note that you have to delete diagcfg.bin before run bcmediag if you want to transmit jumbo frames over 4Kbyte/packet). c. Added PCI-X support and work-around for B0 silicon. This version diagnostics detects if the NIC is in PCI-X slot. If yes, work-around is implemented. However, user still can overwrite this by issuing debug -c=5 to toggle PCI-X work-around before loading driver. bcmediag.exe version 1.24: a. Enhanced txpkt/txcfg command to allow users to: * Transmit random IP headers including IP options. * Transmit random TCP headers including TCP options. * Transmit random data payload. b. Added code to check IP/TCP/UDP checksums of incoming packets. 2. Changed command user interface routines. This changes some of command syntax 3. Fixed WOL spelling 4. Change command help output. 5. Added -p option to print on Error 6. Display error on Red dialog with option to continue or print Version 1.08: 1. Put delay in test C3 to avoid low priority DMA test error 2. Replaced original nictest with manufacturing tests 3. Fixed domument error code error, error code number 28 and greater was shifted by 1. 4. Added error code 39,40,41,42 5. Changed test B5. External RAM test to detect the existance of external RAM first for testing Version 1.07: 1. Changed ESC to CTL-C as break key 2. Changed -t and -T option to allow multiple group specification in one option 3. Changed test D3 to be disabled by default 4. Added "Unit Under Test: NIC" line at error output 5. Unload all drivers upon exit Version 1.06: Added -I option for iteration Version 1.05: 1. Rewrote all tests to comply to OEM's manufacture error message requirement 2. Added -t to disable tests and -T to enable tests format: -t where g is the group letter can be A,B,C, or D n is the subtest number it should be '1' to '9' example: -t A345 -t C13 -T b12 this will disable test A3,A4,A5,C1,C3 and enable test B1,B2. Version 1.04: 1. Added -m option to disable RJ45Loopback Test Version 1.03: 1. Fixed multiple card interrupt problem. Original call had only one variable to save original interrupt vector. If multiple cards use different interrupts, some origianal interrupt vector get lost. This will cause problem after exiting program. Code is changed to support shared interrupt as well. 2. Comment out some unreachable code in order to reduce code size. Version 1.02: 1. Added changes due to changes in boot code firmware 0.6. Before issuing CHIP reset (or GRC reset), it writes a siginature to SRAM so that firmware wouldn't perform code boot initialization such as PHY H/W reset, PXE loading, etc... 2. Added new command to upgrade boot code firmware without corrupting manufacturing information. upgfrm -b -p -f= -b: upgrade boot code. -p: upgrade PXE. -f: input filename. 3. Changed command line option -m to -e to run engineering mode 4. Added command line -i for force_program_device_id option restored option -a for setting base memory address 5. Added 5701 support 6. Added multiple card support This version change is the same as bcmediag.exe v1.20 changes to v1.23. Version 1.01: 1. Fixed and enabled RJ45 loopback test The test requires to reset the system each time replace the adapter. Version 1.00: 1. Fixed Phy Loopback test 2. Fixed ASIC version number display 3. Disabled RJ45 loopback test The program disables PXE and WOL by default. To enable PXE: type q57diag -x To enable WOL: type q57diag -w