[ECOS] JFFS2 fragment corruption during gc?

[Alf Nilsson]

Hi all,

We've noticed some odd behaviors from the JFFS2 file system and would 
like to know if anyone else has had the same type of problems.
Our problem is that when we fill up the file system until it returns 
ENOSPC and then start to remove files, the remaining files seem to be 
corrupted.
The files are the correct size, but it seems like some fragments are 
cleared and set to zero within the file.
We've run tests with the D2 option enabled, but this generates a rather 
large amount of  debug info, but I can make is accessible in case anyone 
is interested.

I have attached a source code file which causes the error. Depending on 
which configuration that is used it takes a different amount of time 
before it fails.
In my current repository the fileio2.c file resides in my 
fs/jffs2/current/tests directory, with a slightly altered cdl. In case 
It would be preferred to send this test as a patch instead let me know.

To make the test run without failing lower the value of FILE_LIMIT, it 
seems like if the file system isn't filled up completely the corruption 
does not occur.

We have noticed this behavior on 2 different flash circuits and on the 
synthetic target.
Our setup has been without compression and CMODE set to NONE default.

The test is based upon writing files that are 75% of the block size, and 
filling the file system completely at least 10 times.

I've also attached my exported settings from the ecos.ecc file, in case 
anyone want to view which settings I've been using.

Thanks,
Alf Nilsson

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: settings_jffs2_test.txt
URL: <http://sourceware.org/pipermail/ecos-discuss/attachments/20041005/a189fd07/attachment.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: fileio2.c
URL: <http://sourceware.org/pipermail/ecos-discuss/attachments/20041005/a189fd07/attachment.c>
-------------- next part --------------
-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss