vsauder at users.sourceforge.net
Thu Feb 24 05:57:31 EST 2005
I am working on a "failed" PPC system with a 2.4.25 kernel. I saw a
lockup where the JFFS2 GC took all the CPU time and the application was
hung. All writes to the file system failed causing catastrophic problems.
On the way to fixing the application to not completely choke and go
haywire, I updated the MTD and JFFS2 drivers to the snapshot as of 16
Feb and now I am having different problems. One thing was that the patch
did not apply cleanly; I had to modify several files to fit it into my
kernel. I noticed that the mtd driver seems much faster (writes) which
was one of my main goal. I was testing the FLASH to verify that the
lockup wasn't the MTD failing to write to the FLASH. There were several
instances on the FLASH that seemed to show a one bit difference (always
a zero that should have been a one). The new drivers sped up the
testing. I fixed some utilities/scripts I was using and it looks like
the FLASH is writing and reading fine now.
I seem to have lost ZLIB compression. It is enabled but the JFFS2 debug
says it doesn't know anything about compression 6. It fails to mount it
as the root fs but it will mount it later but the data is not right
then. I thought I saw somewhere that there was a /proc entry that showed
the jffs2 compressors currently loaded. I can't find it now. Is the
JFFS2_PROC meaningless for the time being?
I am ultimately looking for complete stats on the JFFS2 to tell what is
going on in terms of fragmentation and GC. Our application may be
overloading the JFFS2 by writing too fast or something like that. I
accidentally found significant info from the U-Boot prompt that looks at
first glance like it is a good start. Does anyone else have such a tool?
Is it in JFFS2 and I haven't found it yet?
More information about the linux-mtd