[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: New jffs2 files stay uncompressed?



> joakim.tjernlund@xxxxxxx.se said:
> > In my tree, the most time is spent scanning empty flash and doing CRC
> > on the inodes data part. If there were a way to avoid scanning the
> > empty nodes(that is, play with the ACCURATE bit in the  CLEANMARKER
> > node) and avoid CRC(CRC is checked at every read request anyway so
> > maybe we could just skip the extra CRC check at mount) at mount on
> > every files data, the mount would be very fast.
> 
> If we can make that work, I'd prefer it to the special case in
> jffs2_mark_node_obsolete(). Until now, the ACCURATE bit has been strictly 
> an optimisation - you can't do it on NAND flash or even some NOR chips. 
> That situation has to work too.

 How? I figure that iff ACCURATE bit is set in a CLEANMARKER, then 
 that erase block is empty. On first write to such a EB, clear ACCURATE bit in the
 CLEANMARKER node and then write you data to flash. Can't see how to do that
 in jffs2_mark_node_obsolete()

> 
> > BTW, today the CLEANMARKER is optional, is that intentional? If it was
> > mandatory it would make life simpler when playing with the erase
> > blocks.
> 
> It wasn't implemented in the beginning. Making it mandatory would have 
> involved a break in compatibility. And mkfs.jffs2 still doesn't emit them.

Changing from CRC32 to Adler32 would also break compatibility, but we
could at least have it as an compile option?

 Jocke



To unsubscribe from this list: send the line "unsubscribe jffs-dev" in
the body of a message to majordomo@xxxxxxx.com