Skip to contentSkip to conversation

rcu: v3 add lockdep-based diagnostics to rcu_dereference()

PATCH In review Paul E. McKenney 2 participants 12 messages Jan 15, 2010 – Jan 19, 2010

Subsystems (18)
Networking Documentation VFS Library code +14 more subsystems
lore
Share & tools
lore

Paul E. McKenney

author
#0 lore raw

cover rcu: v3 add lockdep-based diagnostics to rcu_dereference()

Hello!

This patch series adds lockdep-based checking to the rcu_dereference() primitive in order to flag misuses of RCU. The first three patches put the RCU infrastructure in place, the next five use this infrastructure in the net, sched, vfs, radix-tree, and idr subsystems, and the last patch documents how to use the new lockdep-checked rcu_dereference() primitives.

The new (and default-disabled) CONFIG_PROVE_RCU makes this safe to include in tip/core/rcu -- unlike v2, your system won't complain about RCU misuse unless you ask it to. There are still several situations that trigger on systems I have access to (and I have email out on a couple of them), some of the remaining will be triggered only by hardware I don't have access to and kernel features I am unfamiliar with. So this is ready for some more-general testing.

Changes since v2 (http://lkml.org/lkml/2010/1/4/496):

o Removed references to the check condition for the non-lockdep
version of rcu_dereference_check(), thus preventing numerous build errors.

o Create a CONFIG_PROVE_RCU (default disabled) so that developers
who want to use lockdep only to debug locks aren't bothered by RCU-specific lockdep complaints that will no doubt persist for a little bit while some of the more obscure uses of RCU are reverse-engineered.

o Fix a few more issues identified by the checking.

o Add documentation in Documentation/RCU/lockdep.txt.

Changes since v1 (http://lkml.org/lkml/2009/12/15/553):

o Made SRCU have per-srcu_struct lockdep maps to prevent lockdep
from conflating independent uses of SRCU, and updated the rcu_dereference() checking in rcutorture to match.

o Fix a few issues identified by the checking.

o Disable checking within RCU list macros, RCU radix trees, and
IDR.

author
#1 lore raw

PATCH 1/9 rcu: introduce lockdep-based checking to RCU read-side primitives

Inspection is proving insufficient to catch all RCU misuses, which is understandable given that rcu_dereference() might be protected by any of four different flavors of RCU (RCU, RCU-bh, RCU-sched, and SRCU), and might also/instead be protected by any of a number of locking primitives. It is therefore time to enlist the aid of lockdep.

This set of patches is inspired by earlier work by Peter Zijlstra and Thomas Gleixner, and takes the following approach:

o Set up separate lockdep classes for RCU, RCU-bh, and RCU-sched.

o Set up separate lockdep classes for each instance of SRCU.

o Create primitives that check for being in an RCU read-side
critical section. These return exact answers if lockdep is fully enabled, but if unsure, report being in an RCU read-side critical section. (We want to avoid false positives!) The primitives are:

For RCU: rcu_read_lock_held(void)

For RCU-bh: rcu_read_lock_bh_held(void)

For RCU-sched: rcu_read_lock_sched_held(void)

For SRCU: srcu_read_lock_held(struct srcu_struct *sp)

o Add rcu_dereference_check(), which takes a second argument
in which one places a boolean expression based on the above primitives and/or lockdep_is_held().

o A new kernel configuration parameter, CONFIG_PROVE_RCU, enables
rcu_dereference_check(). This depends on CONFIG_PROVE_LOCKING, and should be quite helpful during the transition period while CONFIG_PROVE_RCU-unaware patches are in flight.

The existing rcu_dereference() primitive does no checking, but upcoming patches will change that.

Signed-off-by Paul E. McKenney <[email protected]>
---
 include/linux/rcupdate.h |  126 ++++++++++++++++++++++++++++++++++++++++++----
 include/linux/srcu.h     |   87 +++++++++++++++++++++++++++++++-
 kernel/rcupdate.c        |   10 ++++
 kernel/rcutorture.c      |   12 ++++-
 kernel/srcu.c            |   50 ++++++++++++------
 lib/Kconfig.debug        |   12 ++++
 lib/debug_locks.c        |    1 +
 7 files changed, 267 insertions(+), 31 deletions(-)
include/linux/rcupdate.h +116 -10
index 24440f4..e3d37ef 100644
@@ -78,14 +78,120 @@ extern void rcu_init(void);
 } while (0)
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
+
 extern struct lockdep_map rcu_lock_map;
-# define rcu_read_acquire()	\
-			lock_acquire(&rcu_lock_map, 0, 0, 2, 1, NULL, _THIS_IP_)
+# define rcu_read_acquire() \
+		lock_acquire(&rcu_lock_map, 0, 0, 2, 1, NULL, _THIS_IP_)
 # define rcu_read_release()	lock_release(&rcu_lock_map, 1, _THIS_IP_)
-#else
-# define rcu_read_acquire()	do { } while (0)
-# define rcu_read_release()	do { } while (0)
-#endif
+
+extern struct lockdep_map rcu_bh_lock_map;
+# define rcu_read_acquire_bh() \
+		lock_acquire(&rcu_bh_lock_map, 0, 0, 2, 1, NULL, _THIS_IP_)
+# define rcu_read_release_bh()	lock_release(&rcu_bh_lock_map, 1, _THIS_IP_)
+
+extern struct lockdep_map rcu_sched_lock_map;
+# define rcu_read_acquire_sched() \
+		lock_acquire(&rcu_sched_lock_map, 0, 0, 2, 1, NULL, _THIS_IP_)
+# define rcu_read_release_sched() \
+		lock_release(&rcu_sched_lock_map, 1, _THIS_IP_)
+
+/**
+ * rcu_read_lock_held - might we be in RCU read-side critical section?
+ *
+ * If CONFIG_PROVE_LOCKING is selected and enabled, returns nonzero iff in
+ * an RCU read-side critical section.  In absence of CONFIG_PROVE_LOCKING,
+ * this assumes we are in an RCU read-side critical section unless it can
+ * prove otherwise.
+ */
+static inline int rcu_read_lock_held(void)
+{
+	if (debug_locks)
+		return lock_is_held(&rcu_lock_map);
+	return 1;
+}
+
+/**
+ * rcu_read_lock_bh_held - might we be in RCU-bh read-side critical section?
+ *
+ * If CONFIG_PROVE_LOCKING is selected and enabled, returns nonzero iff in
+ * an RCU-bh read-side critical section.  In absence of CONFIG_PROVE_LOCKING,
+ * this assumes we are in an RCU-bh read-side critical section unless it can
+ * prove otherwise.
+ */
+static inline int rcu_read_lock_bh_held(void)
+{
+	if (debug_locks)
+		return lock_is_held(&rcu_bh_lock_map);
+	return 1;
+}
+
+/**
+ * rcu_read_lock_sched_held - might we be in RCU-sched read-side critical section?
+ *
+ * If CONFIG_PROVE_LOCKING is selected and enabled, returns nonzero iff in an
+ * RCU-sched read-side critical section.  In absence of CONFIG_PROVE_LOCKING,
+ * this assumes we are in an RCU-sched read-side critical section unless it
+ * can prove otherwise.  Note that disabling of preemption (including
+ * disabling irqs) counts as an RCU-sched read-side critical section.
+ */
+static inline int rcu_read_lock_sched_held(void)
+{
+	int lockdep_opinion = 0;
+
+	if (debug_locks)
+		lockdep_opinion = lock_is_held(&rcu_sched_lock_map);
+	return lockdep_opinion || preempt_count() != 0;
+}
+
+#else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
+
+# define rcu_read_acquire()		do { } while (0)
+# define rcu_read_release()		do { } while (0)
+# define rcu_read_acquire_bh()		do { } while (0)
+# define rcu_read_release_bh()		do { } while (0)
+# define rcu_read_acquire_sched()	do { } while (0)
+# define rcu_read_release_sched()	do { } while (0)
+
+static inline int rcu_read_lock_held(void)
+{
+	return 1;
+}
+
+static inline int rcu_read_lock_bh_held(void)
+{
+	return 1;
+}
+
+static inline int rcu_read_lock_sched_held(void)
+{
+	return preempt_count() != 0;
+}
+
+#endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
+
+#ifdef CONFIG_PROVE_RCU
+
+/**
+ * rcu_dereference_check - rcu_dereference with debug checking
+ *
+ * Do an rcu_dereference(), but check that the context is correct.
+ * For example, rcu_dereference_check(gp, rcu_read_lock_held()) to
+ * ensure that the rcu_dereference_check() executes within an RCU
+ * read-side critical section.  It is also possible to check for
+ * locks being held, for example, by using lockdep_is_held().
+ */
+#define rcu_dereference_check(p, c) \
+	({ \
+		if (debug_locks) \
+			WARN_ON_ONCE(!(c)); \
+		rcu_dereference(p); \
+	})
+
+#else /* #ifdef CONFIG_PROVE_RCU */
+
+#define rcu_dereference_check(p, c)	rcu_dereference(p)
+
+#endif /* #else #ifdef CONFIG_PROVE_RCU */
 
 /**
  * rcu_read_lock - mark the beginning of an RCU read-side critical section.

@@ -160,7 +266,7 @@ static inline void rcu_read_lock_bh(void)
 {
 	__rcu_read_lock_bh();
 	__acquire(RCU_BH);
-	rcu_read_acquire();
+	rcu_read_acquire_bh();
 }
 
 /*

@@ -170,7 +276,7 @@ static inline void rcu_read_lock_bh(void)
  */
 static inline void rcu_read_unlock_bh(void)
 {
-	rcu_read_release();
+	rcu_read_release_bh();
 	__release(RCU_BH);
 	__rcu_read_unlock_bh();
 }

@@ -188,7 +294,7 @@ static inline void rcu_read_lock_sched(void)
 {
 	preempt_disable();
 	__acquire(RCU_SCHED);
-	rcu_read_acquire();
+	rcu_read_acquire_sched();
 }
 
 /* Used by lockdep and tracing: cannot be traced, cannot call lockdep. */

@@ -205,7 +311,7 @@ static inline notrace void rcu_read_lock_sched_notrace(void)
  */
 static inline void rcu_read_unlock_sched(void)
 {
-	rcu_read_release();
+	rcu_read_release_sched();
 	__release(RCU_SCHED);
 	preempt_enable();
 }
include/linux/srcu.h +85 -2

103 lines not on this page · raw message

kernel/rcupdate.c +10 -0
index 9b7fd47..033cb55 100644
@@ -50,6 +50,16 @@ static struct lock_class_key rcu_lock_key;
 struct lockdep_map rcu_lock_map =
 	STATIC_LOCKDEP_MAP_INIT("rcu_read_lock", &rcu_lock_key);
 EXPORT_SYMBOL_GPL(rcu_lock_map);
+
+static struct lock_class_key rcu_bh_lock_key;
+struct lockdep_map rcu_bh_lock_map =
+	STATIC_LOCKDEP_MAP_INIT("rcu_read_lock_bh", &rcu_bh_lock_key);
+EXPORT_SYMBOL_GPL(rcu_bh_lock_map);
+
+static struct lock_class_key rcu_sched_lock_key;
+struct lockdep_map rcu_sched_lock_map =
+	STATIC_LOCKDEP_MAP_INIT("rcu_read_lock_sched", &rcu_sched_lock_key);
+EXPORT_SYMBOL_GPL(rcu_sched_lock_map);
 #endif
 
 /*
kernel/rcutorture.c +10 -2
index adda92b..5f43f30 100644
@@ -796,7 +796,11 @@ static void rcu_torture_timer(unsigned long unused)
 
 	idx = cur_ops->readlock();
 	completed = cur_ops->completed();
-	p = rcu_dereference(rcu_torture_current);
+	p = rcu_dereference_check(rcu_torture_current,
+				  rcu_read_lock_held() ||
+				  rcu_read_lock_bh_held() ||
+				  rcu_read_lock_sched_held() ||
+				  srcu_read_lock_held(&srcu_ctl));
 	if (p == NULL) {
 		/* Leave because rcu_torture_writer is not yet underway */
 		cur_ops->readunlock(idx);

@@ -853,7 +857,11 @@ rcu_torture_reader(void *arg)
 		}
 		idx = cur_ops->readlock();
 		completed = cur_ops->completed();
-		p = rcu_dereference(rcu_torture_current);
+		p = rcu_dereference_check(rcu_torture_current,
+					  rcu_read_lock_held() ||
+					  rcu_read_lock_bh_held() ||
+					  rcu_read_lock_sched_held() ||
+					  srcu_read_lock_held(&srcu_ctl));
 		if (p == NULL) {
 			/* Wait for rcu_torture_writer to get underway */
 			cur_ops->readunlock(idx);
kernel/srcu.c +33 -17

93 lines not on this page · raw message

lib/Kconfig.debug +12 -0
index 6bf97d1..6af20a8 100644
@@ -499,6 +499,18 @@ config PROVE_LOCKING
 
 	 For more details, see Documentation/lockdep-design.txt.
 
+config PROVE_RCU
+	bool "RCU debugging: prove RCU correctness"
+	depends on PROVE_LOCKING
+	default n
+	help
+	 This feature enables lockdep extensions that check for correct
+	 use of RCU APIs.  This is currently under development.  Say Y
+	 if you want to debug RCU usage or help work on the PROVE_RCU
+	 feature.
+
+	 Say N if you are unsure.
+
 config LOCKDEP
 	bool
 	depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
lib/debug_locks.c +1 -0
index bc3b117..5bf0020 100644
@@ -23,6 +23,7 @@
  * shut up after that.
  */
 int debug_locks = 1;
+EXPORT_SYMBOL_GPL(debug_locks);
 
 /*
  * The locking-testsuite uses <debug_locks_silent> to get a
Signature · 1 line
1.5.2.5
author
#2 lore raw

PATCH 2/9 rcu: add lockdep-enabled variants of rcu_dereference()

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Make rcu_dereference() check for being in an RCU read-side critical section, and create rcu_dereference_bh(), rcu_dereference_sched(), and srcu_dereference() to check for the other flavors of RCU. Also create rcu_dereference_raw() to avoid checking, and make rcu_dereference_check() use rcu_dereference_raw().

Signed-off-by Paul E. McKenney <[email protected]>
---
 include/linux/rcupdate.h |   41 ++++++++++++++++++++++++++++++++++-------
 include/linux/srcu.h     |    8 ++++++++
 2 files changed, 42 insertions(+), 7 deletions(-)
include/linux/rcupdate.h +34 -7
index e3d37ef..839d296 100644
@@ -184,12 +184,12 @@ static inline int rcu_read_lock_sched_held(void)
 	({ \
 		if (debug_locks) \
 			WARN_ON_ONCE(!(c)); \
-		rcu_dereference(p); \
+		rcu_dereference_raw(p); \
 	})
 
 #else /* #ifdef CONFIG_PROVE_RCU */
 
-#define rcu_dereference_check(p, c)	rcu_dereference(p)
+#define rcu_dereference_check(p, c)	rcu_dereference_raw(p)
 
 #endif /* #else #ifdef CONFIG_PROVE_RCU */
 

@@ -325,22 +325,49 @@ static inline notrace void rcu_read_unlock_sched_notrace(void)
 
 
 /**
- * rcu_dereference - fetch an RCU-protected pointer in an
- * RCU read-side critical section.  This pointer may later
- * be safely dereferenced.
+ * rcu_dereference_raw - fetch an RCU-protected pointer
+ *
+ * The caller must be within some flavor of RCU read-side critical
+ * section, or must be otherwise preventing the pointer from changing,
+ * for example, by holding an appropriate lock.  This pointer may later
+ * be safely dereferenced.  It is the caller's responsibility to have
+ * done the right thing, as this primitive does no checking of any kind.
  *
  * Inserts memory barriers on architectures that require them
  * (currently only the Alpha), and, more importantly, documents
  * exactly which pointers are protected by RCU.
  */
-
-#define rcu_dereference(p)     ({ \
+#define rcu_dereference_raw(p)	({ \
 				typeof(p) _________p1 = ACCESS_ONCE(p); \
 				smp_read_barrier_depends(); \
 				(_________p1); \
 				})
 
 /**
+ * rcu_dereference - fetch an RCU-protected pointer, checking for RCU
+ *
+ * Makes rcu_dereference_check() do the dirty work.
+ */
+#define rcu_dereference(p) \
+	rcu_dereference_check(p, rcu_read_lock_held())
+
+/**
+ * rcu_dereference_bh - fetch an RCU-protected pointer, checking for RCU-bh
+ *
+ * Makes rcu_dereference_check() do the dirty work.
+ */
+#define rcu_dereference_bh(p) \
+		rcu_dereference_check(p, rcu_read_lock_bh_held())
+
+/**
+ * rcu_dereference_sched - fetch RCU-protected pointer, checking for RCU-sched
+ *
+ * Makes rcu_dereference_check() do the dirty work.
+ */
+#define rcu_dereference_sched(p) \
+		rcu_dereference_check(p, rcu_read_lock_sched_held())
+
+/**
  * rcu_assign_pointer - assign (publicize) a pointer to a newly
  * initialized structure that will be dereferenced by RCU read-side
  * critical sections.  Returns the value assigned.
include/linux/srcu.h +8 -0
index adbe167..3084f80 100644
@@ -106,6 +106,14 @@ static inline int srcu_read_lock_held(struct srcu_struct *sp)
 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
 /**
+ * srcu_dereference - fetch SRCU-protected pointer with checking
+ *
+ * Makes rcu_dereference_check() do the dirty work.
+ */
+#define srcu_dereference(p, sp) \
+		rcu_dereference_check(p, srcu_read_lock_held(sp))
+
+/**
  * srcu_read_lock - register a new reader for an SRCU-protected structure.
  * @sp: srcu_struct in which to register the new reader.
  *
Signature · 1 line
1.5.2.5
author
#3 lore raw

PATCH 3/9 rcu: disable lockdep checking in RCU list-traversal primitives

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

The theory is that use of bare rcu_dereference() is more prone to error than use of the RCU list-traversal primitives. Therefore, disable lockdep RCU read-side critical-section checking in these primitives for the time being. Once all of the rcu_dereference() uses have been dealt with, it may be time to re-enable lockdep checking for the RCU list-traversal primitives.

Signed-off-by Paul E. McKenney <[email protected]>
---
 include/linux/rculist.h       |   14 +++++++-------
 include/linux/rculist_nulls.h |    4 ++--
 2 files changed, 9 insertions(+), 9 deletions(-)
include/linux/rculist.h +7 -7
index 1bf0f70..779d707 100644
@@ -208,7 +208,7 @@ static inline void list_splice_init_rcu(struct list_head *list,
  * primitives such as list_add_rcu() as long as it's guarded by rcu_read_lock().
  */
 #define list_entry_rcu(ptr, type, member) \
-	container_of(rcu_dereference(ptr), type, member)
+	container_of(rcu_dereference_raw(ptr), type, member)
 
 /**
  * list_first_entry_rcu - get the first element from a list

@@ -225,9 +225,9 @@ static inline void list_splice_init_rcu(struct list_head *list,
 	list_entry_rcu((ptr)->next, type, member)
 
 #define __list_for_each_rcu(pos, head) \
-	for (pos = rcu_dereference((head)->next); \
+	for (pos = rcu_dereference_raw((head)->next); \
 		pos != (head); \
-		pos = rcu_dereference(pos->next))
+		pos = rcu_dereference_raw(pos->next))
 
 /**
  * list_for_each_entry_rcu	-	iterate over rcu list of given type

@@ -257,9 +257,9 @@ static inline void list_splice_init_rcu(struct list_head *list,
  * as long as the traversal is guarded by rcu_read_lock().
  */
 #define list_for_each_continue_rcu(pos, head) \
-	for ((pos) = rcu_dereference((pos)->next); \
+	for ((pos) = rcu_dereference_raw((pos)->next); \
 		prefetch((pos)->next), (pos) != (head); \
-		(pos) = rcu_dereference((pos)->next))
+		(pos) = rcu_dereference_raw((pos)->next))
 
 /**
  * list_for_each_entry_continue_rcu - continue iteration over list of given type

@@ -418,10 +418,10 @@ static inline void hlist_add_after_rcu(struct hlist_node *prev,
  * as long as the traversal is guarded by rcu_read_lock().
  */
 #define hlist_for_each_entry_rcu(tpos, pos, head, member)		 \
-	for (pos = rcu_dereference((head)->first);			 \
+	for (pos = rcu_dereference_raw((head)->first);			 \
 		pos && ({ prefetch(pos->next); 1; }) &&			 \
 		({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \
-		pos = rcu_dereference(pos->next))
+		pos = rcu_dereference_raw(pos->next))
 
 #endif	/* __KERNEL__ */
 #endif
include/linux/rculist_nulls.h +2 -2
index 589a409..b70ffe5 100644
@@ -101,10 +101,10 @@ static inline void hlist_nulls_add_head_rcu(struct hlist_nulls_node *n,
  *
  */
 #define hlist_nulls_for_each_entry_rcu(tpos, pos, head, member) \
-	for (pos = rcu_dereference((head)->first);			 \
+	for (pos = rcu_dereference_raw((head)->first);			 \
 		(!is_a_nulls(pos)) &&			\
 		({ tpos = hlist_nulls_entry(pos, typeof(*tpos), member); 1; }); \
-		pos = rcu_dereference(pos->next))
+		pos = rcu_dereference_raw(pos->next))
 
 #endif
 #endif
Signature · 1 line
1.5.2.5
author
#4 lore raw

PATCH 4/9 net: add checking to rcu_dereference() primitives

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Update rcu_dereference() primitives to use new lockdep-based checking.

Signed-off-by Paul E. McKenney <[email protected]>
---
 net/core/dev.c         |    2 +-
 net/core/filter.c      |    6 +++---
 net/core/sock.c        |    2 +-
 net/decnet/dn_route.c  |   14 +++++++-------
 net/ipv4/route.c       |   14 +++++++-------
 net/packet/af_packet.c |    2 +-
 6 files changed, 20 insertions(+), 20 deletions(-)
net/core/dev.c +1 -1
index be9924f..0d0ff82 100644
@@ -2041,7 +2041,7 @@ gso:
 	rcu_read_lock_bh();
 
 	txq = dev_pick_tx(dev, skb);
-	q = rcu_dereference(txq->qdisc);
+	q = rcu_dereference_bh(txq->qdisc);
 
 #ifdef CONFIG_NET_CLS_ACT
 	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
net/core/filter.c +3 -3
index 08db7b9..3541aa4 100644
@@ -86,7 +86,7 @@ int sk_filter(struct sock *sk, struct sk_buff *skb)
 		return err;
 
 	rcu_read_lock_bh();
-	filter = rcu_dereference(sk->sk_filter);
+	filter = rcu_dereference_bh(sk->sk_filter);
 	if (filter) {
 		unsigned int pkt_len = sk_run_filter(skb, filter->insns,
 				filter->len);

@@ -521,7 +521,7 @@ int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk)
 	}
 
 	rcu_read_lock_bh();
-	old_fp = rcu_dereference(sk->sk_filter);
+	old_fp = rcu_dereference_bh(sk->sk_filter);
 	rcu_assign_pointer(sk->sk_filter, fp);
 	rcu_read_unlock_bh();
 

@@ -536,7 +536,7 @@ int sk_detach_filter(struct sock *sk)
 	struct sk_filter *filter;
 
 	rcu_read_lock_bh();
-	filter = rcu_dereference(sk->sk_filter);
+	filter = rcu_dereference_bh(sk->sk_filter);
 	if (filter) {
 		rcu_assign_pointer(sk->sk_filter, NULL);
 		sk_filter_delayed_uncharge(sk, filter);
net/core/sock.c +1 -1
index e1f6f22..9309cc9 100644
@@ -1073,7 +1073,7 @@ static void __sk_free(struct sock *sk)
 	if (sk->sk_destruct)
 		sk->sk_destruct(sk);
 
-	filter = rcu_dereference(sk->sk_filter);
+	filter = rcu_dereference_bh(sk->sk_filter);
 	if (filter) {
 		sk_filter_uncharge(sk, filter);
 		rcu_assign_pointer(sk->sk_filter, NULL);
net/decnet/dn_route.c +7 -7
index a032840..af96613 100644
@@ -1155,8 +1155,8 @@ static int __dn_route_output_key(struct dst_entry **pprt, const struct flowi *fl
 
 	if (!(flags & MSG_TRYHARD)) {
 		rcu_read_lock_bh();
-		for(rt = rcu_dereference(dn_rt_hash_table[hash].chain); rt;
-			rt = rcu_dereference(rt->u.dst.dn_next)) {
+		for(rt = rcu_dereference_bh(dn_rt_hash_table[hash].chain); rt;
+			rt = rcu_dereference_bh(rt->u.dst.dn_next)) {
 			if ((flp->fld_dst == rt->fl.fld_dst) &&
 			    (flp->fld_src == rt->fl.fld_src) &&
 			    (flp->mark == rt->fl.mark) &&

@@ -1618,9 +1618,9 @@ int dn_cache_dump(struct sk_buff *skb, struct netlink_callback *cb)
 		if (h > s_h)
 			s_idx = 0;
 		rcu_read_lock_bh();
-		for(rt = rcu_dereference(dn_rt_hash_table[h].chain), idx = 0;
+		for(rt = rcu_dereference_bh(dn_rt_hash_table[h].chain), idx = 0;
 			rt;
-			rt = rcu_dereference(rt->u.dst.dn_next), idx++) {
+			rt = rcu_dereference_bh(rt->u.dst.dn_next), idx++) {
 			if (idx < s_idx)
 				continue;
 			skb_dst_set(skb, dst_clone(&rt->u.dst));

@@ -1654,12 +1654,12 @@ static struct dn_route *dn_rt_cache_get_first(struct seq_file *seq)
 
 	for(s->bucket = dn_rt_hash_mask; s->bucket >= 0; --s->bucket) {
 		rcu_read_lock_bh();
-		rt = dn_rt_hash_table[s->bucket].chain;
+		rt = rcu_dereference_bh(dn_rt_hash_table[s->bucket].chain);
 		if (rt)
 			break;
 		rcu_read_unlock_bh();
 	}
-	return rcu_dereference(rt);
+	return rt;
 }
 
 static struct dn_route *dn_rt_cache_get_next(struct seq_file *seq, struct dn_route *rt)

@@ -1674,7 +1674,7 @@ static struct dn_route *dn_rt_cache_get_next(struct seq_file *seq, struct dn_rou
 		rcu_read_lock_bh();
 		rt = dn_rt_hash_table[s->bucket].chain;
 	}
-	return rcu_dereference(rt);
+	return rcu_dereference_bh(rt);
 }
 
 static void *dn_rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
net/ipv4/route.c +7 -7
index e446496..3476b3b 100644
@@ -287,12 +287,12 @@ static struct rtable *rt_cache_get_first(struct seq_file *seq)
 		if (!rt_hash_table[st->bucket].chain)
 			continue;
 		rcu_read_lock_bh();
-		r = rcu_dereference(rt_hash_table[st->bucket].chain);
+		r = rcu_dereference_bh(rt_hash_table[st->bucket].chain);
 		while (r) {
 			if (dev_net(r->u.dst.dev) == seq_file_net(seq) &&
 			    r->rt_genid == st->genid)
 				return r;
-			r = rcu_dereference(r->u.dst.rt_next);
+			r = rcu_dereference_bh(r->u.dst.rt_next);
 		}
 		rcu_read_unlock_bh();
 	}

@@ -314,7 +314,7 @@ static struct rtable *__rt_cache_get_next(struct seq_file *seq,
 		rcu_read_lock_bh();
 		r = rt_hash_table[st->bucket].chain;
 	}
-	return rcu_dereference(r);
+	return rcu_dereference_bh(r);
 }
 
 static struct rtable *rt_cache_get_next(struct seq_file *seq,

@@ -2687,8 +2687,8 @@ int __ip_route_output_key(struct net *net, struct rtable **rp,
 	hash = rt_hash(flp->fl4_dst, flp->fl4_src, flp->oif, rt_genid(net));
 
 	rcu_read_lock_bh();
-	for (rth = rcu_dereference(rt_hash_table[hash].chain); rth;
-		rth = rcu_dereference(rth->u.dst.rt_next)) {
+	for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth;
+		rth = rcu_dereference_bh(rth->u.dst.rt_next)) {
 		if (rth->fl.fl4_dst == flp->fl4_dst &&
 		    rth->fl.fl4_src == flp->fl4_src &&
 		    rth->fl.iif == 0 &&

@@ -3006,8 +3006,8 @@ int ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb)
 		if (!rt_hash_table[h].chain)
 			continue;
 		rcu_read_lock_bh();
-		for (rt = rcu_dereference(rt_hash_table[h].chain), idx = 0; rt;
-		     rt = rcu_dereference(rt->u.dst.rt_next), idx++) {
+		for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt;
+		     rt = rcu_dereference_bh(rt->u.dst.rt_next), idx++) {
 			if (!net_eq(dev_net(rt->u.dst.dev), net) || idx < s_idx)
 				continue;
 			if (rt_is_expired(rt))
net/packet/af_packet.c +1 -1
index f126d18..939471e 100644
@@ -508,7 +508,7 @@ static inline unsigned int run_filter(struct sk_buff *skb, struct sock *sk,
 	struct sk_filter *filter;
 
 	rcu_read_lock_bh();
-	filter = rcu_dereference(sk->sk_filter);
+	filter = rcu_dereference_bh(sk->sk_filter);
 	if (filter != NULL)
 		res = sk_run_filter(skb, filter->insns, filter->len);
 	rcu_read_unlock_bh();
Signature · 1 line
1.5.2.5
author
#5 lore raw

PATCH 5/9 sched: use lockdep-based checking on rcu_dereference()

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Update the rcu_dereference() usages to take advantage of the new lockdep-based checking.

Signed-off-by Paul E. McKenney <[email protected]>
---
 include/linux/cgroup.h |    5 ++++-
 include/linux/cred.h   |    2 +-
 init/main.c            |    2 ++
 kernel/cgroup.c        |   12 ++++++++++++
 kernel/exit.c          |   14 +++++++++++---
 kernel/fork.c          |    1 +
 kernel/notifier.c      |    6 +++---
 kernel/pid.c           |    2 +-
 kernel/sched.c         |   11 ++++++++---
 9 files changed, 43 insertions(+), 12 deletions(-)
include/linux/cgroup.h +4 -1
index 0008dee..c9bbcb2 100644
@@ -28,6 +28,7 @@ struct css_id;
 extern int cgroup_init_early(void);
 extern int cgroup_init(void);
 extern void cgroup_lock(void);
+extern int cgroup_lock_is_held(void);
 extern bool cgroup_lock_live_group(struct cgroup *cgrp);
 extern void cgroup_unlock(void);
 extern void cgroup_fork(struct task_struct *p);

@@ -486,7 +487,9 @@ static inline struct cgroup_subsys_state *cgroup_subsys_state(
 static inline struct cgroup_subsys_state *task_subsys_state(
 	struct task_struct *task, int subsys_id)
 {
-	return rcu_dereference(task->cgroups->subsys[subsys_id]);
+	return rcu_dereference_check(task->cgroups->subsys[subsys_id],
+				     rcu_read_lock_held() ||
+				     cgroup_lock_is_held());
 }
 
 static inline struct cgroup* task_cgroup(struct task_struct *task,
include/linux/cred.h +1 -1
index 4e3387a..4db09f8 100644
@@ -280,7 +280,7 @@ static inline void put_cred(const struct cred *_cred)
  * task or by holding tasklist_lock to prevent it from being unlinked.
  */
 #define __task_cred(task) \
-	((const struct cred *)(rcu_dereference((task)->real_cred)))
+	((const struct cred *)(rcu_dereference_check((task)->real_cred, rcu_read_lock_held() || lockdep_is_held(&tasklist_lock))))
 
 /**
  * get_task_cred - Get another task's objective credentials
init/main.c +2 -0
index dac44a9..ea6280c 100644
@@ -416,7 +416,9 @@ static noinline void __init_refok rest_init(void)
 	kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
 	numa_default_policy();
 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
+	rcu_read_lock();
 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
+	rcu_read_unlock();
 	unlock_kernel();
 
 	/*
kernel/cgroup.c +12 -0
index 1fbcc74..1b1373c 100644
@@ -166,6 +166,18 @@ static DEFINE_SPINLOCK(hierarchy_id_lock);
  */
 static int need_forkexit_callback __read_mostly;
 
+#ifdef CONFIG_PROVE_LOCKING
+int cgroup_lock_is_held(void)
+{
+	return lockdep_is_held(&cgroup_mutex);
+}
+#else /* #ifdef CONFIG_PROVE_LOCKING */
+int cgroup_lock_is_held(void)
+{
+	return mutex_is_locked(&cgroup_mutex);
+}
+#endif /* #else #ifdef CONFIG_PROVE_LOCKING */
+
 /* convenient tests for these bits */
 inline int cgroup_is_removed(const struct cgroup *cgrp)
 {
kernel/exit.c +11 -3
index 546774a..45ed043 100644
@@ -85,7 +85,9 @@ static void __exit_signal(struct task_struct *tsk)
 	BUG_ON(!sig);
 	BUG_ON(!atomic_read(&sig->count));
 
-	sighand = rcu_dereference(tsk->sighand);
+	sighand = rcu_dereference_check(tsk->sighand,
+					rcu_read_lock_held() ||
+					lockdep_is_held(&tasklist_lock));
 	spin_lock(&sighand->siglock);
 
 	posix_cpu_timers_exit(tsk);

@@ -170,8 +172,10 @@ void release_task(struct task_struct * p)
 repeat:
 	tracehook_prepare_release_task(p);
 	/* don't need to get the RCU readlock here - the process is dead and
-	 * can't be modifying its own credentials */
+	 * can't be modifying its own credentials. But shut RCU-lockdep up */
+	rcu_read_lock();
 	atomic_dec(&__task_cred(p)->user->processes);
+	rcu_read_unlock();
 
 	proc_flush_task(p);
 

@@ -473,9 +477,11 @@ static void close_files(struct files_struct * files)
 	/*
 	 * It is safe to dereference the fd table without RCU or
 	 * ->file_lock because this is the last reference to the
-	 * files structure.
+	 * files structure.  But use RCU to shut RCU-lockdep up.
 	 */
+	rcu_read_lock();
 	fdt = files_fdtable(files);
+	rcu_read_unlock();
 	for (;;) {
 		unsigned long set;
 		i = j * __NFDBITS;

@@ -521,10 +527,12 @@ void put_files_struct(struct files_struct *files)
 		 * at the end of the RCU grace period. Otherwise,
 		 * you can free files immediately.
 		 */
+		rcu_read_lock();
 		fdt = files_fdtable(files);
 		if (fdt != &files->fdtab)
 			kmem_cache_free(files_cachep, files);
 		free_fdtable(fdt);
+		rcu_read_unlock();
 	}
 }
kernel/fork.c +1 -0
index 5b2959b..e01ec3e 100644
@@ -86,6 +86,7 @@ int max_threads;		/* tunable limit on nr_threads */
 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
 
 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock);  /* outer */
+EXPORT_SYMBOL_GPL(tasklist_lock);
 
 int nr_processes(void)
 {
kernel/notifier.c +3 -3
index acd24e7..2488ba7 100644
@@ -78,10 +78,10 @@ static int __kprobes notifier_call_chain(struct notifier_block **nl,
 	int ret = NOTIFY_DONE;
 	struct notifier_block *nb, *next_nb;
 
-	nb = rcu_dereference(*nl);
+	nb = rcu_dereference_raw(*nl);
 
 	while (nb && nr_to_call) {
-		next_nb = rcu_dereference(nb->next);
+		next_nb = rcu_dereference_raw(nb->next);
 
 #ifdef CONFIG_DEBUG_NOTIFIERS
 		if (unlikely(!func_ptr_is_kernel_text(nb->notifier_call))) {

@@ -309,7 +309,7 @@ int __blocking_notifier_call_chain(struct blocking_notifier_head *nh,
 	 * racy then it does not matter what the result of the test
 	 * is, we re-check the list after having taken the lock anyway:
 	 */
-	if (rcu_dereference(nh->head)) {
+	if (rcu_dereference_raw(nh->head)) {
 		down_read(&nh->rwsem);
 		ret = notifier_call_chain(&nh->head, val, v, nr_to_call,
 					nr_calls);
kernel/pid.c +1 -1
index 2e17c9c..b08e697 100644
@@ -367,7 +367,7 @@ struct task_struct *pid_task(struct pid *pid, enum pid_type type)
 	struct task_struct *result = NULL;
 	if (pid) {
 		struct hlist_node *first;
-		first = rcu_dereference(pid->tasks[type].first);
+		first = rcu_dereference_check(pid->tasks[type].first, rcu_read_lock_held() || lockdep_is_held(&tasklist_lock));
 		if (first)
 			result = hlist_entry(first, struct task_struct, pids[(type)].node);
 	}
kernel/sched.c +8 -3
index c535cc4..ad419d9 100644
@@ -645,6 +645,11 @@ static inline int cpu_of(struct rq *rq)
 #endif
 }
 
+#define for_each_domain_rd(p) \
+	rcu_dereference_check((p), \
+			      rcu_read_lock_sched_held() || \
+			      lockdep_is_held(&sched_domains_mutex))
+
 /*
  * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
  * See detach_destroy_domains: synchronize_sched for details.

@@ -653,7 +658,7 @@ static inline int cpu_of(struct rq *rq)
  * preempt-disabled sections.
  */
 #define for_each_domain(cpu, __sd) \
-	for (__sd = rcu_dereference(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
+	for (__sd = for_each_domain_rd(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
 
 #define cpu_rq(cpu)		(&per_cpu(runqueues, (cpu)))
 #define this_rq()		(&__get_cpu_var(runqueues))

@@ -1531,7 +1536,7 @@ static unsigned long target_load(int cpu, int type)
 
 static struct sched_group *group_of(int cpu)
 {
-	struct sched_domain *sd = rcu_dereference(cpu_rq(cpu)->sd);
+	struct sched_domain *sd = rcu_dereference_sched(cpu_rq(cpu)->sd);
 
 	if (!sd)
 		return NULL;

@@ -4877,7 +4882,7 @@ static void run_rebalance_domains(struct softirq_action *h)
 
 static inline int on_null_domain(int cpu)
 {
-	return !rcu_dereference(cpu_rq(cpu)->sd);
+	return !rcu_dereference_sched(cpu_rq(cpu)->sd);
 }
 
 /*
Signature · 1 line
1.5.2.5
author
#6 lore raw

PATCH 6/9 vfs: apply lockdep-based checking to rcu_dereference() uses

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Add lockdep-ified RCU primitives to alloc_fd(), files_fdtable() and fcheck_files().

Cc Alexander Viro <[email protected]>
Signed-off-by Paul E. McKenney <[email protected]>
---
 fs/file.c               |    2 +-
 fs/proc/array.c         |    2 ++
 fs/proc/base.c          |    6 +++++-
 include/linux/fdtable.h |    8 ++++++--
 4 files changed, 14 insertions(+), 4 deletions(-)
fs/file.c +1 -1
index 87e1290..38039af 100644
@@ -478,7 +478,7 @@ repeat:
 	error = fd;
 #if 1
 	/* Sanity check */
-	if (rcu_dereference(fdt->fd[fd]) != NULL) {
+	if (rcu_dereference_raw(fdt->fd[fd]) != NULL) {
 		printk(KERN_WARNING "alloc_fd: slot %d not NULL!\n", fd);
 		rcu_assign_pointer(fdt->fd[fd], NULL);
 	}
fs/proc/array.c +2 -0
index 13b5d07..18e20fe 100644
@@ -270,7 +270,9 @@ static inline void task_sig(struct seq_file *m, struct task_struct *p)
 		blocked = p->blocked;
 		collect_sigign_sigcatch(p, &ignored, &caught);
 		num_threads = atomic_read(&p->signal->count);
+		rcu_read_lock();  /* FIXME: is this correct? */
 		qsize = atomic_read(&__task_cred(p)->user->sigpending);
+		rcu_read_unlock();
 		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
 		unlock_task_sighand(p, &flags);
 	}
fs/proc/base.c +5 -1
index 18d5cc6..db19ca8 100644
@@ -1095,8 +1095,12 @@ static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
 	if (!capable(CAP_AUDIT_CONTROL))
 		return -EPERM;
 
-	if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
+	rcu_read_lock();
+	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
+		rcu_read_unlock();
 		return -EPERM;
+	}
+	rcu_read_unlock();
 
 	if (count >= PAGE_SIZE)
 		count = PAGE_SIZE - 1;
include/linux/fdtable.h +6 -2
index a2ec74b..144412f 100644
@@ -57,7 +57,11 @@ struct files_struct {
 	struct file * fd_array[NR_OPEN_DEFAULT];
 };
 
-#define files_fdtable(files) (rcu_dereference((files)->fdt))
+#define files_fdtable(files) \
+	(rcu_dereference_check((files)->fdt, \
+			       rcu_read_lock_held() || \
+			       lockdep_is_held(&(files)->file_lock) || \
+			       atomic_read(&files->count) == 1))
 
 struct file_operations;
 struct vfsmount;

@@ -78,7 +82,7 @@ static inline struct file * fcheck_files(struct files_struct *files, unsigned in
 	struct fdtable *fdt = files_fdtable(files);
 
 	if (fd < fdt->max_fds)
-		file = rcu_dereference(fdt->fd[fd]);
+		file = rcu_dereference_check(fdt->fd[fd], rcu_read_lock_held() || lockdep_is_held(&files->file_lock) || atomic_read(&files->count) == 1);
 	return file;
 }
Signature · 1 line
1.5.2.5
author
#7 lore raw

PATCH 7/9 radix-tree: disable RCU lockdep checking in radix tree

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Because the radix tree is used with many different locking designs, we cannot do any effective checking without changing the radix-tree APIs. It might make sense to do this later, but only if the RCU lockdep checking proves itself sufficiently valuable.

Signed-off-by Paul E. McKenney <[email protected]>
---
 lib/radix-tree.c |   24 ++++++++++++------------
 1 files changed, 12 insertions(+), 12 deletions(-)
lib/radix-tree.c +12 -12
index 92cdd99..6b9670d 100644
@@ -364,7 +364,7 @@ static void *radix_tree_lookup_element(struct radix_tree_root *root,
 	unsigned int height, shift;
 	struct radix_tree_node *node, **slot;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (node == NULL)
 		return NULL;
 

@@ -384,7 +384,7 @@ static void *radix_tree_lookup_element(struct radix_tree_root *root,
 	do {
 		slot = (struct radix_tree_node **)
 			(node->slots + ((index>>shift) & RADIX_TREE_MAP_MASK));
-		node = rcu_dereference(*slot);
+		node = rcu_dereference_raw(*slot);
 		if (node == NULL)
 			return NULL;
 

@@ -568,7 +568,7 @@ int radix_tree_tag_get(struct radix_tree_root *root,
 	if (!root_tag_get(root, tag))
 		return 0;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (node == NULL)
 		return 0;
 

@@ -602,7 +602,7 @@ int radix_tree_tag_get(struct radix_tree_root *root,
 			BUG_ON(ret && saw_unset_tag);
 			return !!ret;
 		}
-		node = rcu_dereference(node->slots[offset]);
+		node = rcu_dereference_raw(node->slots[offset]);
 		shift -= RADIX_TREE_MAP_SHIFT;
 		height--;
 	}

@@ -711,7 +711,7 @@ __lookup(struct radix_tree_node *slot, void ***results, unsigned long index,
 		}
 
 		shift -= RADIX_TREE_MAP_SHIFT;
-		slot = rcu_dereference(slot->slots[i]);
+		slot = rcu_dereference_raw(slot->slots[i]);
 		if (slot == NULL)
 			goto out;
 	}

@@ -758,7 +758,7 @@ radix_tree_gang_lookup(struct radix_tree_root *root, void **results,
 	unsigned long cur_index = first_index;
 	unsigned int ret;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (!node)
 		return 0;
 

@@ -787,7 +787,7 @@ radix_tree_gang_lookup(struct radix_tree_root *root, void **results,
 			slot = *(((void ***)results)[ret + i]);
 			if (!slot)
 				continue;
-			results[ret + nr_found] = rcu_dereference(slot);
+			results[ret + nr_found] = rcu_dereference_raw(slot);
 			nr_found++;
 		}
 		ret += nr_found;

@@ -826,7 +826,7 @@ radix_tree_gang_lookup_slot(struct radix_tree_root *root, void ***results,
 	unsigned long cur_index = first_index;
 	unsigned int ret;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (!node)
 		return 0;
 

@@ -915,7 +915,7 @@ __lookup_tag(struct radix_tree_node *slot, void ***results, unsigned long index,
 			}
 		}
 		shift -= RADIX_TREE_MAP_SHIFT;
-		slot = rcu_dereference(slot->slots[i]);
+		slot = rcu_dereference_raw(slot->slots[i]);
 		if (slot == NULL)
 			break;
 	}

@@ -951,7 +951,7 @@ radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
 	if (!root_tag_get(root, tag))
 		return 0;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (!node)
 		return 0;
 

@@ -980,7 +980,7 @@ radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
 			slot = *(((void ***)results)[ret + i]);
 			if (!slot)
 				continue;
-			results[ret + nr_found] = rcu_dereference(slot);
+			results[ret + nr_found] = rcu_dereference_raw(slot);
 			nr_found++;
 		}
 		ret += nr_found;

@@ -1020,7 +1020,7 @@ radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
 	if (!root_tag_get(root, tag))
 		return 0;
 
-	node = rcu_dereference(root->rnode);
+	node = rcu_dereference_raw(root->rnode);
 	if (!node)
 		return 0;
Signature · 1 line
1.5.2.5
author
#8 lore raw

PATCH 8/9 idr: apply lockdep-based diagnostics to rcu_dereference() uses

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Because idr can be used with any of a number of locks or with any flavor of RCU, just disable the lockdep-based diagnostics. If idr needs diagnostics, the check expression will need to be passed into the relevant idr primitives as an additional argument.

Signed-off-by Paul E. McKenney <[email protected]>
---
 lib/idr.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
lib/idr.c +4 -4
index 1cac726..4877fce 100644
@@ -502,7 +502,7 @@ void *idr_find(struct idr *idp, int id)
 	int n;
 	struct idr_layer *p;
 
-	p = rcu_dereference(idp->top);
+	p = rcu_dereference_raw(idp->top);
 	if (!p)
 		return NULL;
 	n = (p->layer+1) * IDR_BITS;

@@ -517,7 +517,7 @@ void *idr_find(struct idr *idp, int id)
 	while (n > 0 && p) {
 		n -= IDR_BITS;
 		BUG_ON(n != p->layer*IDR_BITS);
-		p = rcu_dereference(p->ary[(id >> n) & IDR_MASK]);
+		p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
 	}
 	return((void *)p);
 }

@@ -550,7 +550,7 @@ int idr_for_each(struct idr *idp,
 	struct idr_layer **paa = &pa[0];
 
 	n = idp->layers * IDR_BITS;
-	p = rcu_dereference(idp->top);
+	p = rcu_dereference_raw(idp->top);
 	max = 1 << n;
 
 	id = 0;

@@ -558,7 +558,7 @@ int idr_for_each(struct idr *idp,
 		while (n > 0 && p) {
 			n -= IDR_BITS;
 			*paa++ = p;
-			p = rcu_dereference(p->ary[(id >> n) & IDR_MASK]);
+			p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
 		}
 
 		if (p) {
Signature · 1 line
1.5.2.5
maintainer
#9 lore raw

Re: [PATCH tip/core/rcu 8/9] idr: apply lockdep-based diagnostics to rcu_dereference() uses

Quoted 8 lines · Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Because idr can be used with any of a number of locks or with any flavor of RCU, just disable the lockdep-based diagnostics. If idr needs diagnostics, the check expression will need to be passed into the relevant idr primitives as an additional argument.

Signed-off-by Paul E. McKenney <[email protected]>

The subject line of this patch seems wrong. Like the preceding patches for radix tree and list operations, this *disables* the checking by using _raw.

- Josh Triplett

author
#10 lore raw

Re: [PATCH tip/core/rcu 8/9] idr: apply lockdep-based diagnostics to rcu_dereference() uses

Quoted 8 lines · Josh Triplett
Quoted 7 lines · Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Because idr can be used with any of a number of locks or with any flavor of RCU, just disable the lockdep-based diagnostics. If idr needs diagnostics, the check expression will need to be passed into the relevant idr primitives as an additional argument.

Quoted 5 lines
Quoted 1 line
Signed-off-by Paul E. McKenney <[email protected]>

The subject line of this patch seems wrong. Like the preceding patches for radix tree and list operations, this *disables* the checking by using _raw.

It does indeed use _raw. If we start having enough issues with these interfaces, this will need to change. At the moment, the obvious change is to provide multiple interfaces as was done for rcu_dereference(). For example, idr_find(), idr_find_bh(), idr_find_sched(), idr_find_srcu(), idr_find_rcu(), and idr_find_check().

Probably only a few of these are actually in use, but even so, I decided to see how rcu_dereference() turned out before pushing up to the higher-level APIs. My guess would be that the list APIs would be most pressing, but we will see. ;-)

Thanx, Paul

author
#11 lore raw

PATCH 9/9 rcu: documentation update for CONFIG_PROVE_RCU

In reply to Paul E. McKenney

From: Paul E. McKenney <[email protected]>

Adds a lockdep.txt file and updates checklist.txt and whatisRCU.txt to reflect the new lockdep-enabled capabilities of RCU.

Signed-off-by Paul E. McKenney <[email protected]>
---
 Documentation/RCU/00-INDEX      |    2 +
 Documentation/RCU/checklist.txt |   34 +++++++++++--------
 Documentation/RCU/lockdep.txt   |   67 +++++++++++++++++++++++++++++++++++++++
 Documentation/RCU/whatisRCU.txt |   13 +++++---
 4 files changed, 97 insertions(+), 19 deletions(-)
 create mode 100644 Documentation/RCU/lockdep.txt
Documentation/RCU/00-INDEX +2 -0
index 0a27ea9..71b6f50 100644
@@ -6,6 +6,8 @@ checklist.txt
 	- Review Checklist for RCU Patches
 listRCU.txt
 	- Using RCU to Protect Read-Mostly Linked Lists
+lockdep.txt
+	- RCU and lockdep checking
 NMI-RCU.txt
 	- Using RCU to Protect Dynamic NMI Handlers
 rcubarrier.txt
Documentation/RCU/checklist.txt +20 -14
index 767cf06..cbc180f 100644
@@ -127,10 +127,14 @@ over a rather long period of time, but improvements are always welcome!
 		perfectly legal (if redundant) for update-side code to
 		use rcu_dereference() and the "_rcu()" list-traversal
 		primitives.  This is particularly useful in code that
-		is common to readers and updaters.  However, neither
-		rcu_dereference() nor the "_rcu()" list-traversal
-		primitives can substitute for a good concurrency design
-		coordinating among multiple updaters.
+		is common to readers and updaters.  However, lockdep
+		will complain if you access rcu_dereference() outside
+		of an RCU read-side critical section.  See lockdep.txt
+		to learn what to do about this.
+
+		Of course, neither rcu_dereference() nor the "_rcu()"
+		list-traversal primitives can substitute for a good
+		concurrency design coordinating among multiple updaters.
 
 	b.	If the list macros are being used, the list_add_tail_rcu()
 		and list_add_rcu() primitives must be used in order

@@ -249,7 +253,9 @@ over a rather long period of time, but improvements are always welcome!
 	must be protected by appropriate update-side locks.  RCU
 	read-side critical sections are delimited by rcu_read_lock()
 	and rcu_read_unlock(), or by similar primitives such as
-	rcu_read_lock_bh() and rcu_read_unlock_bh().
+	rcu_read_lock_bh() and rcu_read_unlock_bh(), in which case
+	the matching rcu_dereference() primitive must be used in order
+	to keep lockdep happy, in this case, rcu_dereference_bh().
 
 	The reason that it is permissible to use RCU list-traversal
 	primitives when the update-side lock is held is that doing so

@@ -302,15 +308,15 @@ over a rather long period of time, but improvements are always welcome!
 	not the case, a self-spawning RCU callback would prevent the
 	victim CPU from ever going offline.)
 
-14.	SRCU (srcu_read_lock(), srcu_read_unlock(), synchronize_srcu(),
-	and synchronize_srcu_expedited()) may only be invoked from
-	process context.  Unlike other forms of RCU, it -is- permissible
-	to block in an SRCU read-side critical section (demarked by
-	srcu_read_lock() and srcu_read_unlock()), hence the "SRCU":
-	"sleepable RCU".  Please note that if you don't need to sleep
-	in read-side critical sections, you should be using RCU rather
-	than SRCU, because RCU is almost always faster and easier to
-	use than is SRCU.
+14.	SRCU (srcu_read_lock(), srcu_read_unlock(), srcu_dereference(),
+	synchronize_srcu(), and synchronize_srcu_expedited()) may only
+	be invoked from process context.  Unlike other forms of RCU, it
+	-is- permissible to block in an SRCU read-side critical section
+	(demarked by srcu_read_lock() and srcu_read_unlock()), hence the
+	"SRCU": "sleepable RCU".  Please note that if you don't need
+	to sleep in read-side critical sections, you should be using
+	RCU rather than SRCU, because RCU is almost always faster and
+	easier to use than is SRCU.
 
 	Also unlike other forms of RCU, explicit initialization
 	and cleanup is required via init_srcu_struct() and
Documentation/RCU/lockdep.txt new file +67 -0

67 lines not on this page · raw message

Documentation/RCU/whatisRCU.txt +8 -5
index 469a58b..1dc00ee 100644
@@ -323,15 +323,17 @@ used as follows:
 	Defer			Protect
 
 a.	synchronize_rcu()	rcu_read_lock() / rcu_read_unlock()
-	call_rcu()
+	call_rcu()		rcu_dereference()
 
 b.	call_rcu_bh()		rcu_read_lock_bh() / rcu_read_unlock_bh()
+				rcu_dereference_bh()
 
 c.	synchronize_sched()	rcu_read_lock_sched() / rcu_read_unlock_sched()
 				preempt_disable() / preempt_enable()
 				local_irq_save() / local_irq_restore()
 				hardirq enter / hardirq exit
 				NMI enter / NMI exit
+				rcu_dereference_sched()
 
 These three mechanisms are used as follows:
 

@@ -781,9 +783,8 @@ Linux-kernel source code, but it helps to have a full list of the
 APIs, since there does not appear to be a way to categorize them
 in docbook.  Here is the list, by category.
 
-RCU pointer/list traversal:
+RCU list traversal:
 
-	rcu_dereference
 	list_for_each_entry_rcu
 	hlist_for_each_entry_rcu
 	hlist_nulls_for_each_entry_rcu

@@ -809,7 +810,7 @@ RCU:	Critical sections	Grace period		Barrier
 
 	rcu_read_lock		synchronize_net		rcu_barrier
 	rcu_read_unlock		synchronize_rcu
-				synchronize_rcu_expedited
+	rcu_dereference		synchronize_rcu_expedited
 				call_rcu
 
 

@@ -817,7 +818,7 @@ bh:	Critical sections	Grace period		Barrier
 
 	rcu_read_lock_bh	call_rcu_bh		rcu_barrier_bh
 	rcu_read_unlock_bh	synchronize_rcu_bh
-				synchronize_rcu_bh_expedited
+	rcu_dereference_bh	synchronize_rcu_bh_expedited
 
 
 sched:	Critical sections	Grace period		Barrier

@@ -826,12 +827,14 @@ sched:	Critical sections	Grace period		Barrier
 	rcu_read_unlock_sched	call_rcu_sched
 	[preempt_disable]	synchronize_sched_expedited
 	[and friends]
+	rcu_dereference_sched
 
 
 SRCU:	Critical sections	Grace period		Barrier
 
 	srcu_read_lock		synchronize_srcu	N/A
 	srcu_read_unlock	synchronize_srcu_expedited
+	srcu_dereference
 
 SRCU:	Initialization/cleanup
 	init_srcu_struct
Signature · 1 line
1.5.2.5