Re: [PATCH tip/core/rcu 0/21] v6 add lockdep-based diagnostics to rcu_dereference()

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Tue, Feb 23, 2010 at 08:15:44AM -0500, Mathieu Desnoyers wrote:
> * Arnd Bergmann (arnd@arndb.de) wrote:
> > On Tuesday 23 February 2010, Paul E. McKenney wrote:
> > > This patch series adds lockdep-based checking to the rcu_dereference()
> > > primitive in order to flag misuses of RCU.
> > 
> > While I haven't looked much at this series, I've been thinking about
> > adding static diagnostics for rcu_dereference misuse, in the form of
> > an __rcu address space qualifier for pointers. Such a patch would
> > obviously conflict with this series, so I'd wait for yours to go
> > in first, but maybe you like the idea enough to do it yourself ;-).
> > 
> > The observation is that all accesses to an RCU protected pointer
> > are either through rcu_dereference, rcu_assign_pointer or one of their
> > variants. so it should be possible to add a new address space like we
> > have for __iomem, __user and soon __percpu and have sparse check that
> > we use RCU consistently on pointers that need it.
> 
> Just to make myself the devil's advocate: how should we consider
> initialization of RCU pointers at boot time that happens before any
> possible concurrent reader is allowed to run ? I think this case is an
> example of valid RCU-pointer access that is not done through the RCU
> primitives. It seems valid to perform these RCU-pointer accesses when
> serialized by a different exclusion mechanism, in this case being the
> guarantee that no concurrent reader are running at early boot. The same
> applies to stop_machine(), and, as I come to think of it, we could
> probably think of a scheme that dynamically switch from an RCU read-lock
> to, e.g., a mutex for all users, wait for RCU quiescent state, and then
> serialize all users on the mutex during the update. So while some of
> these cases are a bit far-fetched, I think they are valid, and I wonder
> how the address space validation would take them into account.

And this is an excellent exposition of a few of the initialization
issues I referred to in my earlier email!

						Thanx, Paul