On Fri, Apr 24, 2009 at 02:57:52PM +0900, Kyungmin Park wrote:
Actually, I don't like the CONFIG_SYS_MONITOR_LEN approaches, now you are consider the bad block at 1. But we can also consider the bad block 2, if there two consecutive 2 bad block at block 1, 2, we should define the CONFIG_SYS_MONITOR_LEN as 128KiB * 4 and reserved block block 4 always even though we use 2 blocks.
There are two separate constants -- the length of data excluding bad blocks, and the size of the region set aside for that data including bad blocks. CONFIG_SYS_MONITOR_LEN is the former.
We may need to do something special when writing back the environment to find its location if there were any bad blocks, though.
Right, we should consider bad block at IPL, so my recommendation is IPL + u-boot should be fit at block 0
We wouldn't be having this discussion if someone didn't have a u-boot that didn't fit in block 0.
128KiB at OneNAND, 256KiB at Flex-OneNAND. In case of apollon. we should reduce the size as 128KiB for OneNAND side. If the IPL + u-boot size under 128KiB, Flex-OneNAND also can boot it.
For comparison, powerpc u-boots are pretty much never less than 128KiB, and with the NAND subsystem are typically larger than 256KiB. We may need to boot from OneNAND some day.
For simplicity How about to use block scheme? we always use block 0 for boot (IPL + u-boot) whatever block size.
NACK.
-Scott