[Sheepdog] race window accessing current_vnode_info?

[MORITA Kazutaka]

At Thu, 03 May 2012 15:06:46 +0800,
Liu Yuan wrote:
> 
> On 05/03/2012 02:16 PM, tao.peng at emc.com wrote:
> 
> > Hi list,
> > 
> > I was reading sheepdog code and noticed that in update_vnode_info(),
> > if one thread calls get_vnode_info() between
> > put_vnode_info(current_vnode_info) and current_vnode_info =
> > vnode_info, it then will get an invalid pointer. So it looks like a
> > race window there?
> > 
> > Best, Tao
> 
> 
> Oops, it seems that we have brought in a big monster, we don't have any
> lock mechanism to protect current_vnode_info, which is referenced and
> calculated by worker threads and main thread scattered everywhere in the
> code and it is really a dilemma to introduce such a ubiquitous lock
> between worker and main threads.
> 
> Should we revert the patch introduce current_vnode_info?

The old code also has the completely same problem; worker threads call
get_ordered_sd_vnode_list()/free_ordered_sd_vnode_list().  The correct
approach is to modify the code so that all
get_vnode_info()/put_vnode_info() are called only in the main thread.
I'll do the work if no one is trying to fix it now.

Thanks,

Kazutaka