Re: [PATCH] rcu: Use the proper lockdep annotation in dump_blkd_tasks()

[Paul E. McKenney]

On Fri, Mar 09, 2018 at 09:32:18AM +0800, Boqun Feng wrote:
> Sparse reported this:
> 
> | kernel/rcu/tree_plugin.h:814:9: warning: incorrect type in argument 1 
> (different modifiers)
> | kernel/rcu/tree_plugin.h:814:9:expected struct lockdep_map const *lock
> | kernel/rcu/tree_plugin.h:814:9:got struct lockdep_map [noderef] 
> *
> 
> This is caused by using vanilla lockdep annotations on rcu_node::lock,
> and that requires accessing ->lock of rcu_node directly. However we need
> to keep rcu_node::lock __private to avoid breaking its extra ordering
> guarantee. And we have a dedicated lockdep annotation for
> rcu_node::lock, so use it.
> 
> Signed-off-by: Boqun Feng 

Good eyes!  Queued both this and v3 for testing and further review.

Thanx, Paul

> ---
>  kernel/rcu/tree_plugin.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 9accacffd138..ea33f57abae5 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -811,7 +811,7 @@ static void dump_blkd_tasks(struct rcu_node *rnp, int 
> ncheck)
>   int i;
>   struct list_head *lhp;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   pr_info("%s: grp: %d-%d level: %d ->qamask %#lx ->gp_tasks %p 
> ->boost_tasks %p ->exp_tasks %p &->blkd_tasks: %p offset: %u\n", __func__, 
> rnp->grplo, rnp->grphi, rnp->level, rnp->qsmask, rnp->gp_tasks, 
> rnp->boost_tasks, rnp->exp_tasks, >blkd_tasks, (unsigned 
> int)offsetof(typeof(*rnp), blkd_tasks));
>   pr_cont("\t->blkd_tasks");
>   i = 0;
> -- 
> 2.16.2
> 

Re: [PATCH] rcu: Use the proper lockdep annotation in dump_blkd_tasks()

[Paul E. McKenney]

On Fri, Mar 09, 2018 at 09:32:18AM +0800, Boqun Feng wrote:
> Sparse reported this:
> 
> | kernel/rcu/tree_plugin.h:814:9: warning: incorrect type in argument 1 
> (different modifiers)
> | kernel/rcu/tree_plugin.h:814:9:expected struct lockdep_map const *lock
> | kernel/rcu/tree_plugin.h:814:9:got struct lockdep_map [noderef] 
> *
> 
> This is caused by using vanilla lockdep annotations on rcu_node::lock,
> and that requires accessing ->lock of rcu_node directly. However we need
> to keep rcu_node::lock __private to avoid breaking its extra ordering
> guarantee. And we have a dedicated lockdep annotation for
> rcu_node::lock, so use it.
> 
> Signed-off-by: Boqun Feng 

Good eyes!  Queued both this and v3 for testing and further review.

Thanx, Paul

> ---
>  kernel/rcu/tree_plugin.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 9accacffd138..ea33f57abae5 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -811,7 +811,7 @@ static void dump_blkd_tasks(struct rcu_node *rnp, int 
> ncheck)
>   int i;
>   struct list_head *lhp;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   pr_info("%s: grp: %d-%d level: %d ->qamask %#lx ->gp_tasks %p 
> ->boost_tasks %p ->exp_tasks %p &->blkd_tasks: %p offset: %u\n", __func__, 
> rnp->grplo, rnp->grphi, rnp->level, rnp->qsmask, rnp->gp_tasks, 
> rnp->boost_tasks, rnp->exp_tasks, >blkd_tasks, (unsigned 
> int)offsetof(typeof(*rnp), blkd_tasks));
>   pr_cont("\t->blkd_tasks");
>   i = 0;
> -- 
> 2.16.2
>