On Wed, 8 Aug 2012 15:38:41 +0200, David Sterba wrote:

On Sat, Aug 04, 2012 at 01:53:28PM +0800, Miao Xie wrote:

quoted

But I'm not sure these two bugs is the same, so I need the test tool
of David to look into it.

Attached. It's a set of scripts and has a few assumptions hardcoded,
like where the tar srouce is and the name of extracted directory, so
it'll need a few tweaks. Also the actions are started inside a tmux
session for convenience.

The expected stress load is to generate lots of files and directories,
snapshot create and delete and rm of the untarred directory. The 'rm'
step is delayed a few minutes so the tar generates enough data. On some
hosts the rm phase is fast and will clean the untar directory too
quickly.

Thanks.

I find this bug you reported is not introduced by my patch, it is a old bug.
I have sent a RFC patch, I'll be very appreciated if you can get any comment.

Regards
Miao