Hi Scott, Chin,
On Wed, 23 Apr 2014 02:15:10 -0500 Chin Liang See clsee@altera.com wrote:
On Tue, 2014-04-22 at 14:12 -0500, Scott Wood wrote:
On Tue, 2014-04-22 at 10:04 +0900, Masahiro Yamada wrote:
Hi Scott,
It is really really painful to wait more than 10 seconds just for bad block scanning to boot Linux.
Making bad block scans faster is a good thing, but why do you need to scan them just to boot Linux? Aren't you using an on-flash BBT?
I did not know that. I thought all blocks must be scanned.
Could you teach me the better way?
If you use NAND_BBT_USE_FLASH, and NAND_BBT_CREATE is present in the bbt descriptor (this is true of the default descriptors), then the scanning should only need to happen on first use. On subsequent boots only the bad block table should need to be read.
Yup, I agreed with this statement :) I believe this bad block table can be used by kernel in later stage. Probably someone can comment if I am wrong.
It worked for me and it improved bbm checking. Thanks!
BTW, NAND_BBT_USE_FLASH in denali driver did not work at first.
Finally I figured out it. denali.c v7 adds the option to nand->options nand->options |= NAND_BBT_USE_FLASH;
I had to fix it like this nand->bbt_options |= NAND_BBT_USE_FLASH;
Best Regards Masahiro Yamada