diff options
Diffstat (limited to 'doc')
| -rw-r--r-- | doc/README.mpc8349emds.ddrecc | 156 | ||||
| -rw-r--r-- | doc/README.nand | 34 |
2 files changed, 190 insertions, 0 deletions
diff --git a/doc/README.mpc8349emds.ddrecc b/doc/README.mpc8349emds.ddrecc new file mode 100644 index 0000000..401c0b6 --- /dev/null +++ b/doc/README.mpc8349emds.ddrecc @@ -0,0 +1,156 @@ +Overview +======== + +The overall usage pattern for ECC diagnostic commands is the following: + + * (injecting errors is initially disabled) + + * define inject mask (which tells the DDR controller what type of errors + we'll be injecting: single/multiple bit etc.) + + * enable injecting errors - from now on the controller injects errors as + indicated in the inject mask + +IMPORTANT NOTICE: enabling injecting multiple-bit errors is potentially +dangerous as such errors are NOT corrected by the controller. Therefore caution +should be taken when enabling the injection of multiple-bit errors: it is only +safe when used on a carefully selected memory area and used under control of +the 'ecc test' command (see example 'Injecting Multiple-Bit Errors' below). In +particular, when you simply set the multiple-bit errors in inject mask and +enable injection, U-Boot is very likely to hang quickly as the errors will be +injected when it accesses its code, data etc. + + +Use cases for DDR 'ecc' command: +================================ + +Before executing particular tests reset target board or clear status registers: + +=> ecc captureclear +=> ecc errdetectclr all +=> ecc sbecnt 0 + + +Injecting Single-Bit Errors +--------------------------- + +1. Set 1 bit in Data Path Error Inject Mask + +=> ecc injectdatahi 1 + +2. Run test over some memory region + +=> ecc test 200000 10 + +3. Check ECC status + +=> ecc status +... +Memory Data Path Error Injection Mask High/Low: 00000001 00000000 +... +Memory Single-Bit Error Management (0..255): + Single-Bit Error Threshold: 255 + Single Bit Error Counter: 16 +... +Memory Error Detect: + Multiple Memory Errors: 0 + Multiple-Bit Error: 0 + Single-Bit Error: 0 +... + +16 errors were generated, Single-Bit Error flag was not set as Single Bit Error +Counter did not reach Single-Bit Error Threshold. + +4. Make sure used memory region got re-initialized with 0xcafecafe pattern + +=> md 200000 +00200000: cafecafe cafecafe cafecafe cafecafe ................ +00200010: cafecafe cafecafe cafecafe cafecafe ................ +00200020: cafecafe cafecafe cafecafe cafecafe ................ +00200030: cafecafe cafecafe cafecafe cafecafe ................ +00200040: cafecafe cafecafe cafecafe cafecafe ................ +00200050: cafecafe cafecafe cafecafe cafecafe ................ +00200060: cafecafe cafecafe cafecafe cafecafe ................ +00200070: cafecafe cafecafe cafecafe cafecafe ................ +00200080: deadbeef deadbeef deadbeef deadbeef ................ +00200090: deadbeef deadbeef deadbeef deadbeef ................ + + +Injecting Multiple-Bit Errors +----------------------------- + +1. Set more than 1 bit in Data Path Error Inject Mask + +=> ecc injectdatahi 5 + +2. Run test over some memory region + +=> ecc test 200000 10 + +3. Check ECC status + +=> ecc status +... +Memory Data Path Error Injection Mask High/Low: 00000005 00000000 +... +Memory Error Detect: + Multiple Memory Errors: 1 + Multiple-Bit Error: 1 + Single-Bit Error: 0 +... + +Observe that both Multiple Memory Errors and Multiple-Bit Error flags are set. + +4. Make sure used memory region got re-initialized with 0xcafecafe pattern + +=> md 200000 +00200000: cafecafe cafecafe cafecafe cafecafe ................ +00200010: cafecafe cafecafe cafecafe cafecafe ................ +00200020: cafecafe cafecafe cafecafe cafecafe ................ +00200030: cafecafe cafecafe cafecafe cafecafe ................ +00200040: cafecafe cafecafe cafecafe cafecafe ................ +00200050: cafecafe cafecafe cafecafe cafecafe ................ +00200060: cafecafe cafecafe cafecafe cafecafe ................ +00200070: cafecafe cafecafe cafecafe cafecafe ................ +00200080: deadbeef deadbeef deadbeef deadbeef ................ +00200090: deadbeef deadbeef deadbeef deadbeef ................ + + +Test Single-Bit Error Counter and Threshold +------------------------------------------- + +1. Set 1 bit in Data Path Error Inject Mask + +=> ecc injectdatahi 1 + +2. Enable error injection + +=> ecc inject en + +3. Let u-boot run for a with Single-Bit error injection enabled + +4. Disable error injection + +=> ecc inject dis + +4. Check status + +=> ecc status + +... +Memory Single-Bit Error Management (0..255): + Single-Bit Error Threshold: 255 + Single Bit Error Counter: 60 + +Memory Error Detect: + Multiple Memory Errors: 1 + Multiple-Bit Error: 0 + Single-Bit Error: 1 +... + +Observe that Single-Bit Error is 'on' which means that Single-Bit Error Counter +reached Single-Bit Error Threshold. Multiple Memory Errors bit is also 'on', that +is Counter reached Threshold more than one time (it wraps back after reaching +Threshold). + + diff --git a/doc/README.nand b/doc/README.nand index 0f2bdc5..f2d6a5b 100644 --- a/doc/README.nand +++ b/doc/README.nand @@ -1,5 +1,9 @@ NAND FLASH commands and notes + +See NOTE below!!! + + # (C) Copyright 2003 # Dave Ellis, SIXNET, dge@sixnetio.com # @@ -173,3 +177,33 @@ More Definitions: #define NAND_ChipID_UNKNOWN 0x00 #define NAND_MAX_FLOORS 1 #define NAND_MAX_CHIPS 1 + + +NOTE: +===== + +We now use a complete rewrite of the NAND code based on what is in +2.6.12 Linux kernel. + +The old NAND handling code has been re-factored and is now confined +to only board-specific files and - unfortunately - to the DoC code +(see below). A new configuration variable has been introduced: +CFG_NAND_LEGACY, which has to be defined in the board config file if +that board uses legacy code. If CFG_NAND_LEGACY is defined, the board +specific config.mk file should also have "BOARDLIBS = +drivers/nand_legacy/libnand_legacy.a". For boards using the new NAND +approach (PPChameleon and netstar at the moment) no variable is +necessary, but the config.mk should have "BOARDLIBS = +drivers/nand/libnand.a". + +The necessary changes have been made to all affected boards, and no +build breakage has been introduced, except for NETTA and NETTA_ISDN +targets from MAKEALL. This is due to the fact that these two boards +use JFFS, which has been adopted to use the new NAND, and at the same +time use NAND in legacy mode. The breakage will disappear when the +board-specific code is changed to the new NAND. + +As mentioned above, the legacy code is still used by the DoC subsystem. +The consequence of this is that the legacy NAND can't be removed from +the tree until the DoC is ported to use the new NAND support (or boards +with DoC will break). |