Intense JFFS2 testcase fails after 500 times

[Vipin Malik]

At 01:23 PM 6/30/2001 -0400, Frederic Giasson wrote:

>The first time I fill the flash, df command reports a 88% disk uasge.

Hmm, there is something weird right here. Why does it say 88% full
when you say you "fill flash".

The JFFS2 statfs takes into account the number of sectors required
for GC and should report 100% full when it really is full. Except of course,
the 18% empty flash is so small that it cannot fit even one of your "small 
files".

>   It
>remains like that until about 200 test case loops have executed.

You mean that when you erase all the files, disk space goes back to
x% (not 0 of course due to the reserved sectors for GC) but each time
the max full is >88%?

>   But after
>that, disk usage begins to grow slowly, and at about 520 loops, my testcase
>fails, reporting a full disk.

Do you mean even if you delete all the files, df still says that disk is full?
Is your GC task running?

>   After the test, I checked the flashes chips
>and they are not burned out, they are still very good.  Did anyone ever ran
>such tests?  If so, did it gave the same results?  If so, could there be
>some sort of memory leaks with JFFS2 filesystem?

By memory leak, I guess you mean media (flash) space leak, as you didn't say
anything about loosing RAM?

Vipin