JFFS2 not persistent?

cpope schrieb:

Depending on your embedded distribution you may need to specify a "flash save" or similar.

Martin

I'd suggest ensuring the JFFS2 partition is mounted with -osync. This should ensure that all write transactions block when they are committed to the flash, and are not held off in the cache.

copy

new

This did seem to help but now I have a very strange problem. If I create a file with a period in the name, e.g. reboot.sh, it appears correct but when I unmount and remount it I get a CRC error and the file is missing (I suspect it's still there but the inode is screwed up). However, if I create a file without a period everything works fine.

So, why would a period in the filename clobber JFFS2?

Thanks, Clark

In your system must be strange things happen. I'm working since years with JFFS2 filesystems on various targets. No problems like yours yet. What kind of memory do you use (NOR, NAND)? What kernel revision? How large is your flash memory (whole flash and page size)? How does the CPU accesses the memory (linear, paged)? Is the CPU cache disabled in the flash area? What CPU do you are using?

Juergen

the

(I

large

accesses

I think the period in the filename may be a coincidence. I've since had files without the period throw the same error.

nor

All accesses are through MTD so I assume it configures the accesses to flash and disables cache in that are, right?

Thnaks, Clark

Some flash chips allow for memory block protection on individual block by block basis, within the same flash chip. I am wondering whether you have some of the blocks write protected and others not. If that was the case you may get this kind of strange behavior.