Re: [tip: locking/core] lockdep: Fix lockdep recursion

[Paul E. McKenney]

On Tue, Oct 13, 2020 at 01:25:44PM +0200, Peter Zijlstra wrote:
> On Tue, Oct 13, 2020 at 12:44:50PM +0200, Peter Zijlstra wrote:
> > On Tue, Oct 13, 2020 at 12:34:06PM +0200, Peter Zijlstra wrote:
> > > On Mon, Oct 12, 2020 at 02:28:12PM -0700, Paul E. McKenney wrote:
> > > > It is certainly an accident waiting to happen.  Would something like
> > > > the following make sense?
> > > 
> > > Sadly no.
> > > 
> > > > ------------------------------------------------------------------------
> > > > 
> > > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > > index bfd38f2..52a63bc 100644
> > > > --- a/kernel/rcu/tree.c
> > > > +++ b/kernel/rcu/tree.c
> > > > @@ -4067,6 +4067,7 @@ void rcu_cpu_starting(unsigned int cpu)
> > > >  
> > > >  	rnp = rdp->mynode;
> > > >  	mask = rdp->grpmask;
> > > > +	lockdep_off();
> > > >  	raw_spin_lock_irqsave_rcu_node(rnp, flags);
> > > >  	WRITE_ONCE(rnp->qsmaskinitnext, rnp->qsmaskinitnext | mask);
> > > >  	newcpu = !(rnp->expmaskinitnext & mask);
> > > > @@ -4086,6 +4087,7 @@ void rcu_cpu_starting(unsigned int cpu)
> > > >  	} else {
> > > >  		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
> > > >  	}
> > > > +	lockdep_on();
> > > >  	smp_mb(); /* Ensure RCU read-side usage follows above initialization. */
> > > >  }
> > > 
> > > This will just shut it up, but will not fix the actual problem of that
> > > spin-lock ending up in trace_lock_acquire() which relies on RCU which
> > > isn't looking.
> > > 
> > > What we need here is to supress tracing not lockdep. Let me consider.
> > 
> > We appear to have a similar problem with rcu_report_dead(), it's
> > raw_spin_unlock()s can end up in trace_lock_release() while we just
> > killed RCU.
> 
> So we can deal with the explicit trace_*() calls like the below, but I
> really don't like it much. It also doesn't help with function tracing.
> This is really early/late in the hotplug cycle and should be considered
> entry, we shouldn't be tracing anything here.
> 
> Paul, would it be possible to use a scheme similar to IRQ/NMI for
> hotplug? That seems to mostly rely on atomic ops, not locks.

The rest of the rcu_node tree and the various grace-period/hotplug races
makes that question non-trivial.  I will look into it, but I have no
reason for optimism.

But there is only one way to find out...  ;-)

							Thanx, Paul

> ---
> diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
> index d05db575f60f..22e3a3523ad3 100644
> --- a/include/linux/lockdep.h
> +++ b/include/linux/lockdep.h
> @@ -159,7 +159,7 @@ extern void lockdep_init_task(struct task_struct *task);
>   */
>  #define LOCKDEP_RECURSION_BITS	16
>  #define LOCKDEP_OFF		(1U << LOCKDEP_RECURSION_BITS)
> -#define LOCKDEP_RECURSION_MASK	(LOCKDEP_OFF - 1)
> +#define LOCKDEP_TRACE_MASK	(LOCKDEP_OFF - 1)
>  
>  /*
>   * lockdep_{off,on}() are macros to avoid tracing and kprobes; not inlines due
> @@ -176,6 +176,16 @@ do {							\
>  	current->lockdep_recursion -= LOCKDEP_OFF;	\
>  } while (0)
>  
> +#define lockdep_trace_off()				\
> +do {							\
> +	current->lockdep_recursion++;			\
> +} while (0)
> +
> +#define lockdep_trace_on()				\
> +do {							\
> +	current->lockdep_recursion--			\
> +} while (0)
> +
>  extern void lockdep_register_key(struct lock_class_key *key);
>  extern void lockdep_unregister_key(struct lock_class_key *key);
>  
> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> index 3e99dfef8408..2df98abee82e 100644
> --- a/kernel/locking/lockdep.c
> +++ b/kernel/locking/lockdep.c
> @@ -87,7 +87,7 @@ static inline bool lockdep_enabled(void)
>  	if (raw_cpu_read(lockdep_recursion))
>  		return false;
>  
> -	if (current->lockdep_recursion)
> +	if (current->lockdep_recursion >> LOCKDEP_RECURSION_BITS)
>  		return false;
>  
>  	return true;
> @@ -5410,7 +5410,8 @@ void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
>  {
>  	unsigned long flags;
>  
> -	trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
> +	if (!(current->lockdep_recursion & LOCKDEP_TRACE_MASK))
> +		trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
>  
>  	if (!debug_locks)
>  		return;
> @@ -5450,7 +5451,8 @@ void lock_release(struct lockdep_map *lock, unsigned long ip)
>  {
>  	unsigned long flags;
>  
> -	trace_lock_release(lock, ip);
> +	if (!(current->lockdep_recursion & LOCKDEP_TRACE_MASK))
> +		trace_lock_release(lock, ip);
>  
>  	if (unlikely(!lockdep_enabled()))
>  		return;
> @@ -5662,7 +5664,8 @@ void lock_contended(struct lockdep_map *lock, unsigned long ip)
>  {
>  	unsigned long flags;
>  
> -	trace_lock_acquired(lock, ip);
> +	if (!(current->lockdep_recursion & LOCKDEP_TRACE_MASK))
> +		trace_lock_acquired(lock, ip);
>  
>  	if (unlikely(!lock_stat || !lockdep_enabled()))
>  		return;
> @@ -5680,7 +5683,8 @@ void lock_acquired(struct lockdep_map *lock, unsigned long ip)
>  {
>  	unsigned long flags;
>  
> -	trace_lock_contended(lock, ip);
> +	if (!(current->lockdep_recursion & LOCKDEP_TRACE_MASK))
> +		trace_lock_contended(lock, ip);
>  
>  	if (unlikely(!lock_stat || !lockdep_enabled()))
>  		return;
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index edeabc232c21..dbd56603fc0a 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -4047,6 +4047,11 @@ void rcu_cpu_starting(unsigned int cpu)
>  
>  	rnp = rdp->mynode;
>  	mask = rdp->grpmask;
> +
> +	/*
> +	 * Lockdep will call tracing, which requires RCU, but RCU isn't on yet.
> +	 */
> +	lockdep_trace_off();
>  	raw_spin_lock_irqsave_rcu_node(rnp, flags);
>  	WRITE_ONCE(rnp->qsmaskinitnext, rnp->qsmaskinitnext | mask);
>  	newcpu = !(rnp->expmaskinitnext & mask);
> @@ -4064,6 +4069,7 @@ void rcu_cpu_starting(unsigned int cpu)
>  	} else {
>  		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
>  	}
> +	lockdep_trace_on();
>  	smp_mb(); /* Ensure RCU read-side usage follows above initialization. */
>  }
>  
> @@ -4091,6 +4097,11 @@ void rcu_report_dead(unsigned int cpu)
>  
>  	/* Remove outgoing CPU from mask in the leaf rcu_node structure. */
>  	mask = rdp->grpmask;
> +
> +	/*
> +	 * Lockdep will call tracing, which requires RCU, but we're switching RCU off.
> +	 */
> +	lockdep_trace_off();
>  	raw_spin_lock(&rcu_state.ofl_lock);
>  	raw_spin_lock_irqsave_rcu_node(rnp, flags); /* Enforce GP memory-order guarantee. */
>  	rdp->rcu_ofl_gp_seq = READ_ONCE(rcu_state.gp_seq);
> @@ -4101,8 +4112,10 @@ void rcu_report_dead(unsigned int cpu)
>  		raw_spin_lock_irqsave_rcu_node(rnp, flags);
>  	}
>  	WRITE_ONCE(rnp->qsmaskinitnext, rnp->qsmaskinitnext & ~mask);
> +	/* RCU is off, locks must not call into tracing */
>  	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
>  	raw_spin_unlock(&rcu_state.ofl_lock);
> +	lockdep_trace_on();
>  
>  	rdp->cpu_started = false;
>  }
> diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
> index 39334d2d2b37..403b138f7cd4 100644
> --- a/kernel/rcu/update.c
> +++ b/kernel/rcu/update.c
> @@ -275,8 +275,8 @@ EXPORT_SYMBOL_GPL(rcu_callback_map);
>  
>  noinstr int notrace debug_lockdep_rcu_enabled(void)
>  {
> -	return rcu_scheduler_active != RCU_SCHEDULER_INACTIVE && debug_locks &&
> -	       current->lockdep_recursion == 0;
> +	return rcu_scheduler_active != RCU_SCHEDULER_INACTIVE && __lockdep_enabled;
> +
>  }
>  EXPORT_SYMBOL_GPL(debug_lockdep_rcu_enabled);
>