Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Paul E. McKenney]

On Wed, Jan 17, 2018 at 04:41:43PM -0800, Matthew Wilcox wrote:
> On Wed, Jan 17, 2018 at 03:29:13PM -0800, Paul E. McKenney wrote:
> > On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> > > 
> > > From: Matthew Wilcox 
> > > 
> > > Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> > > by accessing rcu_node->lock directly and ignoring the __private
> > > marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> > > in srcutree.c introduced by a3883df3935e.
> > > 
> > > Signed-off-by: Matthew Wilcox 
> > 
> > Good catch!  Applied for review and testing.
> > 
> > For some reason, I was expecting 0day to catch this sort of thing...
> 
> Funny you should say, it was 0day which pointed it out to me!  I changed
> lockdep to take a const struct lockdep_map *lock instead of a plain
> struct lockdep_map *lock and I got a whinge from 0day that I'd changed
> some of the warnings, so I looked into it in case it was my fault.

Nary a peep at me.  Regardless, glad you caught it, and thank you for
the fix!

Thanx, Paul

Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Paul E. McKenney]

On Wed, Jan 17, 2018 at 04:41:43PM -0800, Matthew Wilcox wrote:
> On Wed, Jan 17, 2018 at 03:29:13PM -0800, Paul E. McKenney wrote:
> > On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> > > 
> > > From: Matthew Wilcox 
> > > 
> > > Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> > > by accessing rcu_node->lock directly and ignoring the __private
> > > marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> > > in srcutree.c introduced by a3883df3935e.
> > > 
> > > Signed-off-by: Matthew Wilcox 
> > 
> > Good catch!  Applied for review and testing.
> > 
> > For some reason, I was expecting 0day to catch this sort of thing...
> 
> Funny you should say, it was 0day which pointed it out to me!  I changed
> lockdep to take a const struct lockdep_map *lock instead of a plain
> struct lockdep_map *lock and I got a whinge from 0day that I'd changed
> some of the warnings, so I looked into it in case it was my fault.

Nary a peep at me.  Regardless, glad you caught it, and thank you for
the fix!

Thanx, Paul

Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Matthew Wilcox]

On Wed, Jan 17, 2018 at 03:29:13PM -0800, Paul E. McKenney wrote:
> On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> > 
> > From: Matthew Wilcox 
> > 
> > Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> > by accessing rcu_node->lock directly and ignoring the __private
> > marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> > in srcutree.c introduced by a3883df3935e.
> > 
> > Signed-off-by: Matthew Wilcox 
> 
> Good catch!  Applied for review and testing.
> 
> For some reason, I was expecting 0day to catch this sort of thing...

Funny you should say, it was 0day which pointed it out to me!  I changed
lockdep to take a const struct lockdep_map *lock instead of a plain
struct lockdep_map *lock and I got a whinge from 0day that I'd changed
some of the warnings, so I looked into it in case it was my fault.

Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Matthew Wilcox]

On Wed, Jan 17, 2018 at 03:29:13PM -0800, Paul E. McKenney wrote:
> On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> > 
> > From: Matthew Wilcox 
> > 
> > Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> > by accessing rcu_node->lock directly and ignoring the __private
> > marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> > in srcutree.c introduced by a3883df3935e.
> > 
> > Signed-off-by: Matthew Wilcox 
> 
> Good catch!  Applied for review and testing.
> 
> For some reason, I was expecting 0day to catch this sort of thing...

Funny you should say, it was 0day which pointed it out to me!  I changed
lockdep to take a const struct lockdep_map *lock instead of a plain
struct lockdep_map *lock and I got a whinge from 0day that I'd changed
some of the warnings, so I looked into it in case it was my fault.

Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Paul E. McKenney]

On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> 
> From: Matthew Wilcox 
> 
> Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> by accessing rcu_node->lock directly and ignoring the __private
> marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> in srcutree.c introduced by a3883df3935e.
> 
> Signed-off-by: Matthew Wilcox 

Good catch!  Applied for review and testing.

For some reason, I was expecting 0day to catch this sort of thing...

Thanx, Paul

> diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
> index 59c471de342a..30db0a581628 100644
> --- a/kernel/rcu/rcu.h
> +++ b/kernel/rcu/rcu.h
> @@ -360,7 +360,7 @@ do {  
> \
>  } while (0)
> 
>  #define raw_spin_unlock_irqrestore_rcu_node(p, flags)
> \
> - raw_spin_unlock_irqrestore(_PRIVATE(p, lock), flags) \
> + raw_spin_unlock_irqrestore(_PRIVATE(p, lock), flags)
> 
>  #define raw_spin_trylock_rcu_node(p) \
>  ({   \
> @@ -371,6 +371,9 @@ do {  
> \
>   ___locked;  \
>  })
> 
> +#define raw_lockdep_assert_held_rcu_node(p)  \
> + lockdep_assert_held(_PRIVATE(p, lock))
> +
>  #endif /* #if defined(SRCU) || !defined(TINY_RCU) */
> 
>  #ifdef CONFIG_TINY_RCU
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index 6d5880089ff6..b7bdd3ff4611 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -412,7 +412,7 @@ static void srcu_gp_start(struct srcu_struct *sp)
>   struct srcu_data *sdp = this_cpu_ptr(sp->sda);
>   int state;
> 
> - lockdep_assert_held(>lock);
> + lockdep_assert_held(_PRIVATE(sp, lock));
>   WARN_ON_ONCE(ULONG_CMP_GE(sp->srcu_gp_seq, sp->srcu_gp_seq_needed));
>   rcu_segcblist_advance(>srcu_cblist,
> rcu_seq_current(>srcu_gp_seq));
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index f9c0ca2ccf0c..aba8dd0cc1f8 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -1245,7 +1245,7 @@ static int rcu_is_cpu_rrupt_from_idle(void)
>   */
>  static void rcu_gpnum_ovf(struct rcu_node *rnp, struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   if (ULONG_CMP_LT(READ_ONCE(rdp->gpnum) + ULONG_MAX / 4, rnp->gpnum))
>   WRITE_ONCE(rdp->gpwrap, true);
>   if (ULONG_CMP_LT(rdp->rcu_iw_gpnum + ULONG_MAX / 4, rnp->gpnum))
> @@ -1712,7 +1712,7 @@ void rcu_cpu_stall_reset(void)
>  static unsigned long rcu_cbs_completed(struct rcu_state *rsp,
>  struct rcu_node *rnp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /*
>* If RCU is idle, we just wait for the next grace period.
> @@ -1759,7 +1759,7 @@ rcu_start_future_gp(struct rcu_node *rnp, struct 
> rcu_data *rdp,
>   bool ret = false;
>   struct rcu_node *rnp_root = rcu_get_root(rdp->rsp);
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /*
>* Pick up grace-period number for new callbacks.  If this
> @@ -1887,7 +1887,7 @@ static bool rcu_accelerate_cbs(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>  {
>   bool ret = false;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* If no pending (not yet ready to invoke) callbacks, nothing to do. */
>   if (!rcu_segcblist_pend_cbs(>cblist))
> @@ -1927,7 +1927,7 @@ static bool rcu_accelerate_cbs(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>  static bool rcu_advance_cbs(struct rcu_state *rsp, struct rcu_node *rnp,
>   struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* If no pending (not yet ready to invoke) callbacks, nothing to do. */
>   if (!rcu_segcblist_pend_cbs(>cblist))
> @@ -1955,7 +1955,7 @@ static bool __note_gp_changes(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>   bool ret;
>   bool need_gp;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* Handle the ends of any preceding grace periods first. */
>   if (rdp->completed == rnp->completed &&
> @@ -2380,7 +2380,7 @@ static bool
>  rcu_start_gp_advanced(struct rcu_state *rsp, struct rcu_node *rnp,
> struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   if (!rsp->gp_kthread || !cpu_needs_another_gp(rsp, rdp)) {
>   /*
>* Either we have not 

Re: [PATCH] rcu: Use wrapper for lockdep asserts

[Paul E. McKenney]

On Wed, Jan 17, 2018 at 06:24:30AM -0800, Matthew Wilcox wrote:
> 
> From: Matthew Wilcox 
> 
> Commits c0b334c5bfa9 and ea9b0c8a26a2 introduced new sparse warnings
> by accessing rcu_node->lock directly and ignoring the __private
> marker.  Introduce a new wrapper and use it.  Also fix a similar problem
> in srcutree.c introduced by a3883df3935e.
> 
> Signed-off-by: Matthew Wilcox 

Good catch!  Applied for review and testing.

For some reason, I was expecting 0day to catch this sort of thing...

Thanx, Paul

> diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
> index 59c471de342a..30db0a581628 100644
> --- a/kernel/rcu/rcu.h
> +++ b/kernel/rcu/rcu.h
> @@ -360,7 +360,7 @@ do {  
> \
>  } while (0)
> 
>  #define raw_spin_unlock_irqrestore_rcu_node(p, flags)
> \
> - raw_spin_unlock_irqrestore(_PRIVATE(p, lock), flags) \
> + raw_spin_unlock_irqrestore(_PRIVATE(p, lock), flags)
> 
>  #define raw_spin_trylock_rcu_node(p) \
>  ({   \
> @@ -371,6 +371,9 @@ do {  
> \
>   ___locked;  \
>  })
> 
> +#define raw_lockdep_assert_held_rcu_node(p)  \
> + lockdep_assert_held(_PRIVATE(p, lock))
> +
>  #endif /* #if defined(SRCU) || !defined(TINY_RCU) */
> 
>  #ifdef CONFIG_TINY_RCU
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index 6d5880089ff6..b7bdd3ff4611 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -412,7 +412,7 @@ static void srcu_gp_start(struct srcu_struct *sp)
>   struct srcu_data *sdp = this_cpu_ptr(sp->sda);
>   int state;
> 
> - lockdep_assert_held(>lock);
> + lockdep_assert_held(_PRIVATE(sp, lock));
>   WARN_ON_ONCE(ULONG_CMP_GE(sp->srcu_gp_seq, sp->srcu_gp_seq_needed));
>   rcu_segcblist_advance(>srcu_cblist,
> rcu_seq_current(>srcu_gp_seq));
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index f9c0ca2ccf0c..aba8dd0cc1f8 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -1245,7 +1245,7 @@ static int rcu_is_cpu_rrupt_from_idle(void)
>   */
>  static void rcu_gpnum_ovf(struct rcu_node *rnp, struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   if (ULONG_CMP_LT(READ_ONCE(rdp->gpnum) + ULONG_MAX / 4, rnp->gpnum))
>   WRITE_ONCE(rdp->gpwrap, true);
>   if (ULONG_CMP_LT(rdp->rcu_iw_gpnum + ULONG_MAX / 4, rnp->gpnum))
> @@ -1712,7 +1712,7 @@ void rcu_cpu_stall_reset(void)
>  static unsigned long rcu_cbs_completed(struct rcu_state *rsp,
>  struct rcu_node *rnp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /*
>* If RCU is idle, we just wait for the next grace period.
> @@ -1759,7 +1759,7 @@ rcu_start_future_gp(struct rcu_node *rnp, struct 
> rcu_data *rdp,
>   bool ret = false;
>   struct rcu_node *rnp_root = rcu_get_root(rdp->rsp);
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /*
>* Pick up grace-period number for new callbacks.  If this
> @@ -1887,7 +1887,7 @@ static bool rcu_accelerate_cbs(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>  {
>   bool ret = false;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* If no pending (not yet ready to invoke) callbacks, nothing to do. */
>   if (!rcu_segcblist_pend_cbs(>cblist))
> @@ -1927,7 +1927,7 @@ static bool rcu_accelerate_cbs(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>  static bool rcu_advance_cbs(struct rcu_state *rsp, struct rcu_node *rnp,
>   struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* If no pending (not yet ready to invoke) callbacks, nothing to do. */
>   if (!rcu_segcblist_pend_cbs(>cblist))
> @@ -1955,7 +1955,7 @@ static bool __note_gp_changes(struct rcu_state *rsp, 
> struct rcu_node *rnp,
>   bool ret;
>   bool need_gp;
> 
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
> 
>   /* Handle the ends of any preceding grace periods first. */
>   if (rdp->completed == rnp->completed &&
> @@ -2380,7 +2380,7 @@ static bool
>  rcu_start_gp_advanced(struct rcu_state *rsp, struct rcu_node *rnp,
> struct rcu_data *rdp)
>  {
> - lockdep_assert_held(>lock);
> + raw_lockdep_assert_held_rcu_node(rnp);
>   if (!rsp->gp_kthread || !cpu_needs_another_gp(rsp, rdp)) {
>   /*
>* Either we have not yet spawned the grace-period
> @@ -2442,7