[Openais] [lck] patches segfault, lock_algo, orphan locks

Steven Dake sdake at redhat.com
Thu Sep 27 18:40:28 PDT 2007


Pascal,

I think it is ok to add an EX lock to the wait list so I'll merge the
first patch.

Good catch on the 2nd patch not an obvious problem to see.  I had to
stare at it a good 15 mins to see why it was needed :)  merged

Orphan locks should be recovered but they aren't today.  I won't merge
this patch because it has comments in it and I'm not sure what its
intent is.

good work thanks for the patches

regards
-steve

On Mon, 2007-09-24 at 14:19 +0200, Pascal Bouchareine wrote:
> Hi, 
> 
> Three patches attached : 
> 
> lock_algo queues lock requests rather than ignoring them if
> an exclusive lock is held on a resource. This led to stale
> processes trying to acquire an exclusive lock forever.
> 
> segfault does not use source.conn unless the originating
> request is local. The actual code tries to dereference pointers
> created on another node.
> 
> orphan_locks is a guess that, with the "lock_fd/close" hack used
> today, in conjunction with the lck_lib_exit_fn called whenever
> the fd is closed, the actual code should lead to orphan locks.
> 
> This last one is tested but quite unsure to me.
> 
> Do they seem correct to you ?
> 
> Thanks,
> Pascal
> 
> -- 
> \o/   Pascal Bouchareine - Gandi 
>  g    0170393757           15, place de la Nation - 75011 Paris       



More information about the Openais mailing list