[Cluster-devel] Curious lockdep report

[Steven Whitehouse]

Hi,

On Thu, 2014-01-09 at 11:43 +0000, Steven Whitehouse wrote:
> Hi,
> 
> I'm trying to track down the cause of the lockdep report which I've
> included below. It is a bit odd since it doesn't say which lock it
> refers to, although my suspicions are that it is the mapping's
> tree_lock. The other thing I know is that it goes away if I revert this
> patch:
> 
> https://git.kernel.org/cgit/linux/kernel/git/steve/gfs2-3.0-nmw.git/commit/?id=70d4ee94b370c5ef54d0870600f16bd92d18013c
> 
> However, it is not immediately obvious why, if the lockdep code is
> certain that there are no errors in the code, it needs to warn about
> missing annotation, nor is it clear what that annotation should be.
> 
> The reproducer is simple - just mount a gfs2 filesystem with the latest
> gfs2 -nmw tree and with lockdep turned on (can be local, doesn't need to
> be clustered) and do mkdir.
> 
> Any suggestions as to whats going wrong gratefully received :-)
> 
> Steve.
> 

Andy Price has solved the mystery... seems just to be some missing
initialisation for the address space in question. I'll get a patch sortd
out shortly,

Steve.