Re: [PATCH tip/core/rcu 4/6] rcu: Silence compiler array out-of-bounds false positive

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

On Mon, Jan 07, 2013 at 09:16:02AM -0800, Paul E. McKenney wrote:
> On Mon, Jan 07, 2013 at 07:50:02AM -0800, Josh Triplett wrote:
> > On Sat, Jan 05, 2013 at 09:09:36AM -0800, Paul E. McKenney wrote:
> > > From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
> > > 
> > > It turns out that gcc 4.8 warns on array indexes being out of bounds
> > > unless it can prove otherwise.  It gives this warning on some RCU
> > > initialization code.  Because this is far from any fastpath, add
> > > an explicit check for array bounds and panic if so.  This gives the
> > > compiler enough information to figure out that the array index is never
> > > out of bounds.
> > > 
> > > However, if a similar false positive occurs on a fastpath, it will
> > > probably be necessary to tell the compiler to keep its array-index
> > > anxieties to itself.  ;-)
> > > 
> > > Markus Trippelsdorf <markus@trippelsdorf.de>
> > > Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > > ---
> > >  kernel/rcutree.c |    4 ++++
> > >  1 files changed, 4 insertions(+), 0 deletions(-)
> > > 
> > > diff --git a/kernel/rcutree.c b/kernel/rcutree.c
> > > index d145796..e0d9815 100644
> > > --- a/kernel/rcutree.c
> > > +++ b/kernel/rcutree.c
> > > @@ -2938,6 +2938,10 @@ static void __init rcu_init_one(struct rcu_state *rsp,
> > >  
> > >  	BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf));  /* Fix buf[] init! */
> > >  
> > > +	/* Silence gcc 4.8 warning about array index out of range. */
> > > +	if (rcu_num_lvls > RCU_NUM_LVLS)
> > > +		panic("rcu_init_one: rcu_num_lvls overflow");
> > 
> > Why not write this as BUG_ON(rcu_num_lvls > RCU_NUM_LVLS)?  Given that
> > the condition in question can never happen, you don't really need an
> > explanatory message.
> 
> Good point, will do!

Ah, wait, BUG_ON() sometimes compiles to nothingness:

#ifndef HAVE_ARCH_BUG_ON
#define BUG_ON(condition) do { if (condition) ; } while(0)
#endif

So I do need the explicit "if".  :-(

							Thanx, Paul

> > I do find it surprising, though, that the compiler can't figure this one
> > out, given that rcu_num_lvls gets initialized right before this in the
> > same file (and likely inlined into the same function).  I wonder if it
> > thought some other code might change it unexpectedly, since rcu_num_lvls
> > doesn't get declared as static?  Unfortunately, the loop macros in
> > rcutree.h make it difficult to make rcu_num_lvls static, but as far as I
> > can tell only one use of those macros ever gets expanded outside of
> > rcutree.c: the one in rcutree_trace.c.  If you compile out tracing, and
> > declare rcu_num_lvls static, does the warning go away?
> 
> I found it quite surprising also, hence the "array-index anxieties" above.
> 
> I added Marcus on CC for his thoughts on this.
> 
> 							Thanx, Paul