[PATCH][JFFS2] Fix garbage collector block search

[David Woodhouse]

On Wed, 2008-04-23 at 17:08 +0400, Alexander Belyakov wrote:
> On Wed, Apr 23, 2008 at 3:32 PM, David Woodhouse <dwmw2 at infradead.org> wrote:
> >
> >  I'm playing with it now, with the intent of committing it. Can you show
> >  me how to reproduce the problem?
> >
> 
> The most easy way to reproduce it is to fill jffs2 volume with single
> large file until there is no empty space, then remove this file and
> immediately try to write another file.
> 
> Something like:
> ---
> #!/bin/sh
> rm -rf /mnt/vol1/*
> dd if=/dev/urandom of=/mnt/vol1/test.file bs=4096 count=32768
> rm /mnt/vol1/test.file
> dd if=/dev/urandom of=/mnt/vol1/test_new.file bs=2048 count=1024
> -------
> 
> The bug is better reproduced on small volumes (for me it is just
> 4Mbytes). The first dd should exit with "No space left on device"
> error to be sure no free space left. The last dd should always
> succeed. If it fails - you catch the bug. Sometimes it is necessary to
> run this script in several time in a row before last dd exits with an
> error.

Hm, I can't make it fail. I'm using nandsim, and have enabled the
'Museum IDs' so that I can have a 4MiB device (I've also tried 2MiB).

I never even see the 'Synching wbuf in order to reuse
erasable_pending_wbuf_list blocks' message, whether I have debugging
enabled or whether I just unable that one message.

-- 
dwmw2