[U-Boot-Users] NAND and bad blocks
Hi,
I'm trying to figure out what my options are for booting a system completely from NAND flash. Googling around hasn't yielded much info, but browsing "drivers/nand/" I've gathered that U-Boot handles bad blocks (factory or worn-out) & bit-flips to some degree. It's not entirely clear how things work by glancing at the code, though, so I have a few questions:
1. How does U-Boot handle a bad block in the area where the environment is stored? I can create a second environment via CFG_ENV_OFFSET_REDUND, but that's not very flexible (what if the primary and the backup are both bad blocks?) I also need to write to the environment from within Linux via the "fw_setenv" utility, but it doesn't seem to have any NAND awareness.
2. It's not obvious to me how U-Boot reacts when it encounters a bad block. If I'm copying a kernel image out of NAND, and a bad block is encountered, does it skip to the next block? If so, do I need to compensate by padding each partition with unused blocks, or how else does addressing work? Similarly for writes - if I do "nand write <memaddr> <location> 100000", and the first block is bad, will U-Boot write all 0x100000 bytes _after_ that block?
3. Can Linux handle bad blocks in the same manner? Specifically, I need to use 'nandwrite' to copy a kernel image to an MTD partition from within Linux, and have U-Boot read the image correctly on bootup, both in the presence of bad blocks and bit-flips.
The only references I found were at the openmoko site, but I'm not sure how much of what they discuss is native in U-Boot & Linux, vs. being added on for openmoko. Thanks in advance for the help!
In message DC8113368EDCB444B74B22F57D9D807E6E33AA@CLEOHSMB05.napa.ad.etn.com you wrote:
I'm trying to figure out what my options are for booting a system completely from NAND flash. Googling around hasn't yielded much info,
The first question you shouls answer to yourself - and probably let us know about the answer - is which sort of CPU you are using, which sort of NAND controller it has and if the CPU supports booting from NAND.
Without that knowledge, all speculation is just a waste of time.
Best regards,
Wolfgang Denk
wd@denx.de wrote:
The first question you shouls answer to yourself - and probably let us know about the answer - is which sort of CPU you are using, which sort of NAND controller it has and if the CPU supports booting from NAND.
Without that knowledge, all speculation is just a waste of time.
Sorry, I meant to include that. The processor (Freescale MPC8313) has a flash controller that _does_ support booting from NAND - I've built and loaded "u-boot-nand.bin" (with the split 16kB IPL and SPL), and it boots fine on my development board. Now I'm trying to figure out the software side of things.
Thanks!
Hi Matthew,
On Tuesday 18 September 2007, MatthewLCreech@eaton.com wrote:
- How does U-Boot handle a bad block in the area where the environment
is stored? I can create a second environment via CFG_ENV_OFFSET_REDUND, but that's not very flexible (what if the primary and the backup are both bad blocks?) I also need to write to the environment from within Linux via the "fw_setenv" utility, but it doesn't seem to have any NAND awareness.
You could be right here. The bad block handling in the NAND environment is not perfect. If both blocks of the environment are bad (primary & redundant) then we really have a problem.
There was a patch sent to the list a few months ago (IIRC, from the Openmoko project), that addressed this problem. Unfortunately I didn't find the time till now to take a deeper look at it.
- It's not obvious to me how U-Boot reacts when it encounters a bad
block. If I'm copying a kernel image out of NAND, and a bad block is encountered, does it skip to the next block?
There are both option:
a) Bad blocks are not skipped and writted as 0xff in memory b) Bad blocks are skipped upon read from NAND
I personally never used a) and always use b).
If so, do I need to compensate by padding each partition with unused blocks, or how else does addressing work?
Yes, you have to add some reserve to your partitions size for bad blocks that could occur.
Similarly for writes - if I do "nand write <memaddr> <location> 100000", and the first block is bad, will U-Boot write all 0x100000 bytes _after_ that block?
Yes.
- Can Linux handle bad blocks in the same manner? Specifically, I need
to use 'nandwrite' to copy a kernel image to an MTD partition from within Linux, and have U-Boot read the image correctly on bootup, both in the presence of bad blocks and bit-flips.
Sure, this works the same way.
Best regards, Stefan
===================================================================== DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office@denx.de =====================================================================
participants (3)
-
MatthewLCreech＠eaton.com -
Stefan Roese -
Wolfgang Denk