Re: [PATCH RFC tip/core/rcu 03/41] rcu: Add lockdep-RCU checks forsimple self-deadlock

From: Paul E. McKenney
Date: Thu Feb 02 2012 - 11:21:26 EST


On Wed, Feb 01, 2012 at 04:55:54PM -0800, Josh Triplett wrote:
> On Wed, Feb 01, 2012 at 11:41:21AM -0800, Paul E. McKenney wrote:
> > From: "Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx>
> >
> > It is illegal to have a grace period within a same-flavor RCU read-side
> > critical section, so this commit adds lockdep-RCU checks to splat when
> > such abuse is encountered. This commit does not detect more elaborate
> > RCU deadlock situations. These situations might be a job for lockdep
> > enhancements.
>
> Since doing so also violates the prohibition on blocking within an RCU
> read-side critical section, wouldn't it suffice to call might_sleep() or
> equivalent, which also detects other problems? (Obviously this doesn't
> apply to SRCU, but it applies to the other variants of RCU.)

Yes, but...

The advantage of the lockdep-RCU splat is that it gives you a better
hint as to where the RCU read-side critical section was entered, which
is very helpful when tracking these down, especially when they are
intermittent.

On of the downsides of the Linux kernel community being more RCU-savvy
is that the errors they now tend to commit are more complex. ;-)

And yes, I should also well check for the other variants of RCU read-side
critical section (other than RCU). Done.

I also glued the strings together to promote grepability as you suggest
later. (But I leave it to you to get checkpatch.pl upgraded -- it currently
warns about long lines, but not about strings split across lines.)

> > --- a/kernel/rcutiny.c
> > +++ b/kernel/rcutiny.c
> > @@ -319,6 +319,9 @@ static void rcu_process_callbacks(struct softirq_action *unused)
> > */
> > void synchronize_sched(void)
> > {
> > + rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),
> > + "Illegal grace period in RCU read-side "
> > + "critical section");
>
> This message doesn't seem entirely obvious to me. A grace period didn't
> occur; a synchronize call did, which tried to request a grace period
> that can never happen.

I suppose I might as well make it consistent with the other messages. ;-)

> > --- a/kernel/rcutree.c
> > +++ b/kernel/rcutree.c
> > @@ -1816,6 +1816,9 @@ EXPORT_SYMBOL_GPL(call_rcu_bh);
> > */
> > void synchronize_sched(void)
> > {
> > + rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),
> > + "Illegal synchronize_sched() in RCU-sched "
> > + "read-side critical section");
> > if (rcu_blocking_is_gp())
> > return;
> > wait_rcu_gp(call_rcu_sched);
> > @@ -1833,6 +1836,9 @@ EXPORT_SYMBOL_GPL(synchronize_sched);
> > */
> > void synchronize_rcu_bh(void)
> > {
> > + rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map),
> > + "Illegal synchronize_sched() in RCU-bh "
> > + "read-side critical section");
>
> Copy-paste problem here: this should say synchronize_sched_bh. (Or
> perhaps it should say __func__. :) )

Fixed, but will pass on __func__ for the moment. Cool though it might
be to exercise varargs. ;-)

> > --- a/kernel/srcu.c
> > +++ b/kernel/srcu.c
> > @@ -172,6 +172,10 @@ static void __synchronize_srcu(struct srcu_struct *sp, void (*sync_func)(void))
> > {
> > int idx;
> >
> > + rcu_lockdep_assert(!lock_is_held(&sp->dep_map),
> > + "Illegal SRCU grace period in same-type "
> > + "SRCU read-side critical section");
>
> Same issue with the message: a grace period didn't occur, and it never
> will; a call to synchronize_srcu requesting a grace period occurred.

Good catch, fixed!

Thanx, Paul

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/