Sean McNeil wrote: > What does dynpart refer to then when it says is reads from BBT.
Hmm, when does it say something like this ? It does read the BBT, that's true. It does so to find out how much space it has to leave for bad blocks. (By the way, all this is mainly about blocks marked as "bad" at the factory.) > And why doesn't dynpart obey what is written in the env? Dynpart creates the mtdparts information. Its own input, i.e., partition names and sizes, is hard-coded in u-boot, in dynpart_size and dynpart_names in board/neo1973/gta02/gta02.c As Andy has already pointed out, we don't quite like this approach anymore. But it's a legacy that has survived until GTA02. It's interesting to note that, had we sized partitions statically such that there some "slack" blocks in each to accommodate bad blocks, it seems that we would have had to spend 1% of the NAND size for slack blocks if we'd wanted to aim for a <10^-6 per-partition risk of not having enough slask, or 1.5% of the NAND for a <10^-9 risk. Assuming that bad blocks are evenly distributed and that my calculations are correct: svn.openmoko.org/developers/werner/bb_slack.pl - Werner
