Hi all,
> YAFFS2 uses OOB area more intensively than JFFS2, so in order to
> read/write things it issues more NAND commands generally, which turns
> to be a performance drop for GPIO/bitbanging NAND controller
> emulation.
I'm somewhat puzzled by this statement. I'm not very familiar with
JFFS2, but if it writes to the OOB at all, then it issues exactly as
many write commands as yaffs2 - ie one command to write a whole page,
including the OOB area.
As for reading the flash, I thought that Yaffs was pretty good for
reading too - ie, you must read all OOBs in used blocks once at scan
time, along with a page for each file. I imagine that JFFS2 does a
similar scan at mount time. From memory, the only time I can think of
that Yaffs hits the flash more often than absolutely necessary is when
you need to get detailed information about a file. To keep ram usage
down, Yaffs only stores an abbreviated file description structure in
ram. To get more detailed info, you need to read the object header
(one page) from flash.
I'd be really interested in hearing what the NAND usage pattern of JFFS is.
Cheers
Brad
(text/plain)