Thomas pointed me here.

On Mon, Oct 31, 2011 at 10:32:46AM -0700, Simon Kirby wrote:

[104661.244767] 
[104661.244767]  Possible unsafe locking scenario:
[104661.244767]        
[104661.244767]        CPU0                    CPU1
[104661.244767]        ----                    ----
[104661.244767]   lock(slock-AF_INET);
[104661.244767]                                lock(slock-AF_INET);
[104661.244767]                                lock(slock-AF_INET);
[104661.244767]   lock(slock-AF_INET);
[104661.244767] 
[104661.244767]  *** DEADLOCK ***
[104661.244767] 

Bah, I used the __print_lock_name() function to show the lock names in
the above, which leaves off the subclass number. I'll go write up a
patch that fixes that.

Thanks,

-- Steve