0

In our system we use jffs2 to write kernel and initramfs into NAND memory. We would like to use stronger ECC soon. But new ECC format requires whole OOB area, and does not work with JFFS2.

I am not sure yet, why do we need to use JFFS2 at all. AFAIK kernel and initramfs are copied into RAM at startup, and runs from RAM. They are not changed very often (maybe few times in year for update).

Bill the Lizard
  • 398,270
  • 210
  • 566
  • 880
Bartlomiej Grzeskowiak
  • 1,009
  • 5
  • 16
  • 35
  • 1
    "*In our system we use jffs2 to write kernel and initramfs into NAND memory"* -- Then you are storing your kernel image within a filesystem. Other systems store the image(s) in a raw flash "partition". It's your choice (assuming the bootloader can read the filesystem). Handling multiple versions/copies of the kernel is much easier with a filesystem than a dedicated partition. – sawdust Feb 23 '14 at 23:20
  • I see now that nand read.jffs2 command confused me. according to: http://blackfin.uclinux.org/doku.php?id=bootloaders:u-boot:nand-flash "The write.jffs2 command has nothing to do with jffs2 and is poorly named - it just means skip bad blocks". It is more like: In our system we use BBT write and maybe cleanmarkers from jffs2 to read/write kernel and initramfs. – Bartlomiej Grzeskowiak Mar 05 '14 at 10:00

1 Answers1

2

jffs2 is designed as filesystem for flash storage, usually use as rootfs. There's no any advantage for using it for kernel and initramfs image.

andycjw
  • 1,021
  • 1
  • 8
  • 11