Skip to contentSkip to conversation

Documentation and rcutorture changes for 3.9

PATCH In review Paul E. McKenney 2 participants 5 messages Jan 5, 2013 – Jan 7, 2013

Subsystems (5)
Documentation Tracing RCU Atomic infrastructure +1 more subsystems
lore
Share & tools
lore

Paul E. McKenney

author
#0 lore raw

cover Documentation and rcutorture changes for 3.9

Hello!

The following are changes to documentation and rcutorture:

1. Export trace_clock_local() in order to allow rcutorture event
tracing to emit the time of the beginning of the RCU read-side critical section at the point where a failure is detected.
2. Reduce rcutorture's read-side tracing to include only failures,
thus cutting the number of events down to something reasonable (normally zero, in fact!).
3. Add atomic_xchg() to the list of atomic operations and memory
barriers, courtesy of Richard Braun.

Thanx, Paul

 Documentation/atomic_ops.txt      |    2 ++
 Documentation/memory-barriers.txt |    1 +
 include/linux/rcupdate.h          |   13 ++++++++++---
 include/trace/events/rcu.h        |   19 ++++++++++++++-----
 kernel/rcupdate.c                 |    9 ++++++---
 kernel/rcutorture.c               |   31 ++++++++++++++++++++++++-------
 kernel/trace/trace_clock.c        |    1 +
 7 files changed, 58 insertions(+), 18 deletions(-)
author
#3 lore raw

PATCH 3/3 Documentation: Memory barrier semantics of atomic_xchg()

In reply to Paul E. McKenney

From: Richard Braun <[email protected]>

Add atomic_xchg() to documentation for atomic operations and memory barriers.

Signed-off-by Richard Braun <[email protected]>
Signed-off-by Paul E. McKenney <[email protected]>
---
 Documentation/atomic_ops.txt      |    2 ++
 Documentation/memory-barriers.txt |    1 +
 2 files changed, 3 insertions(+), 0 deletions(-)
Documentation/atomic_ops.txt +2 -0
index 27f2b21..d9ca5be 100644
@@ -253,6 +253,8 @@ This performs an atomic exchange operation on the atomic variable v, setting
 the given new value.  It returns the old value that the atomic variable v had
 just before the operation.
 
+atomic_xchg requires explicit memory barriers around the operation.
+
 	int atomic_cmpxchg(atomic_t *v, int old, int new);
 
 This performs an atomic compare exchange operation on the atomic value v,
Documentation/memory-barriers.txt +1 -0
index 3c4e1b3..fa5d8a9 100644
@@ -1685,6 +1685,7 @@ explicit lock operations, described later).  These include:
 
 	xchg();
 	cmpxchg();
+	atomic_xchg();
 	atomic_cmpxchg();
 	atomic_inc_return();
 	atomic_dec_return();
Signature · 1 line
1.7.8
author
#2 lore raw

PATCH 2/3 rcu: Reduce rcutorture tracing

In reply to Paul E. McKenney

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

Currently, rcutorture traces every read-side access. This can be problematic because even a two-minute rcutorture run on a two-CPU system can generate 28,853,363 reads. Normally, only a failing read is of interest, so this commit traces adjusts rcutorture's tracing to only trace failing reads. The resulting event tracing records the time and the ->completed value captured at the beginning of the RCU read-side critical section, allowing correlation with other event-tracing messages.

Signed-off-by Paul E. McKenney <[email protected]>
Signed-off-by Paul E. McKenney <[email protected]>
---
 include/linux/rcupdate.h   |   13 ++++++++++---
 include/trace/events/rcu.h |   19 ++++++++++++++-----
 kernel/rcupdate.c          |    9 ++++++---
 kernel/rcutorture.c        |   31 ++++++++++++++++++++++++-------
 4 files changed, 54 insertions(+), 18 deletions(-)
include/linux/rcupdate.h +10 -3
index 275aa3f..7f89cea 100644
@@ -53,7 +53,10 @@ extern int rcutorture_runnable; /* for sysctl */
 extern void rcutorture_record_test_transition(void);
 extern void rcutorture_record_progress(unsigned long vernum);
 extern void do_trace_rcu_torture_read(char *rcutorturename,
-				      struct rcu_head *rhp);
+				      struct rcu_head *rhp,
+				      unsigned long secs,
+				      unsigned long c_old,
+				      unsigned long c);
 #else
 static inline void rcutorture_record_test_transition(void)
 {

@@ -63,9 +66,13 @@ static inline void rcutorture_record_progress(unsigned long vernum)
 }
 #ifdef CONFIG_RCU_TRACE
 extern void do_trace_rcu_torture_read(char *rcutorturename,
-				      struct rcu_head *rhp);
+				      struct rcu_head *rhp,
+				      unsigned long secs,
+				      unsigned long c_old,
+				      unsigned long c);
 #else
-#define do_trace_rcu_torture_read(rcutorturename, rhp) do { } while (0)
+#define do_trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
+	do { } while (0)
 #endif
 #endif
include/trace/events/rcu.h +14 -5
index d4f559b..09af021 100644
@@ -523,22 +523,30 @@ TRACE_EVENT(rcu_batch_end,
  */
 TRACE_EVENT(rcu_torture_read,
 
-	TP_PROTO(char *rcutorturename, struct rcu_head *rhp),
+	TP_PROTO(char *rcutorturename, struct rcu_head *rhp,
+		 unsigned long secs, unsigned long c_old, unsigned long c),
 
-	TP_ARGS(rcutorturename, rhp),
+	TP_ARGS(rcutorturename, rhp, secs, c_old, c),
 
 	TP_STRUCT__entry(
 		__field(char *, rcutorturename)
 		__field(struct rcu_head *, rhp)
+		__field(unsigned long, secs)
+		__field(unsigned long, c_old)
+		__field(unsigned long, c)
 	),
 
 	TP_fast_assign(
 		__entry->rcutorturename = rcutorturename;
 		__entry->rhp = rhp;
+		__entry->secs = secs;
+		__entry->c_old = c_old;
+		__entry->c = c;
 	),
 
-	TP_printk("%s torture read %p",
-		  __entry->rcutorturename, __entry->rhp)
+	TP_printk("%s torture read %p %luus c: %lu %lu",
+		  __entry->rcutorturename, __entry->rhp,
+		  __entry->secs, __entry->c_old, __entry->c)
 );
 
 /*

@@ -608,7 +616,8 @@ TRACE_EVENT(rcu_barrier,
 #define trace_rcu_invoke_kfree_callback(rcuname, rhp, offset) do { } while (0)
 #define trace_rcu_batch_end(rcuname, callbacks_invoked, cb, nr, iit, risk) \
 	do { } while (0)
-#define trace_rcu_torture_read(rcutorturename, rhp) do { } while (0)
+#define trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
+	do { } while (0)
 #define trace_rcu_barrier(name, s, cpu, cnt, done) do { } while (0)
 
 #endif /* #else #ifdef CONFIG_RCU_TRACE */
kernel/rcupdate.c +6 -3
index a2cf761..303359d 100644
@@ -404,11 +404,14 @@ EXPORT_SYMBOL_GPL(rcuhead_debug_descr);
 #endif /* #ifdef CONFIG_DEBUG_OBJECTS_RCU_HEAD */
 
 #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU) || defined(CONFIG_RCU_TRACE)
-void do_trace_rcu_torture_read(char *rcutorturename, struct rcu_head *rhp)
+void do_trace_rcu_torture_read(char *rcutorturename, struct rcu_head *rhp,
+			       unsigned long secs,
+			       unsigned long c_old, unsigned long c)
 {
-	trace_rcu_torture_read(rcutorturename, rhp);
+	trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c);
 }
 EXPORT_SYMBOL_GPL(do_trace_rcu_torture_read);
 #else
-#define do_trace_rcu_torture_read(rcutorturename, rhp) do { } while (0)
+#define do_trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
+	do { } while (0)
 #endif
kernel/rcutorture.c +24 -7
index 31dea01..a583f1c 100644
@@ -46,6 +46,7 @@
 #include <linux/stat.h>
 #include <linux/srcu.h>
 #include <linux/slab.h>
+#include <linux/trace_clock.h>
 #include <asm/byteorder.h>
 
 MODULE_LICENSE("GPL");

@@ -1028,7 +1029,6 @@ void rcutorture_trace_dump(void)
 		return;
 	if (atomic_xchg(&beenhere, 1) != 0)
 		return;
-	do_trace_rcu_torture_read(cur_ops->name, (struct rcu_head *)~0UL);
 	ftrace_dump(DUMP_ALL);
 }
 

@@ -1042,13 +1042,16 @@ static void rcu_torture_timer(unsigned long unused)
 {
 	int idx;
 	int completed;
+	int completed_end;
 	static DEFINE_RCU_RANDOM(rand);
 	static DEFINE_SPINLOCK(rand_lock);
 	struct rcu_torture *p;
 	int pipe_count;
+	unsigned long long ts;
 
 	idx = cur_ops->readlock();
 	completed = cur_ops->completed();
+	ts = trace_clock_local();
 	p = rcu_dereference_check(rcu_torture_current,
 				  rcu_read_lock_bh_held() ||
 				  rcu_read_lock_sched_held() ||

@@ -1058,7 +1061,6 @@ static void rcu_torture_timer(unsigned long unused)
 		cur_ops->readunlock(idx);
 		return;
 	}
-	do_trace_rcu_torture_read(cur_ops->name, &p->rtort_rcu);
 	if (p->rtort_mbtest == 0)
 		atomic_inc(&n_rcu_torture_mberror);
 	spin_lock(&rand_lock);

@@ -1071,10 +1073,16 @@ static void rcu_torture_timer(unsigned long unused)
 		/* Should not happen, but... */
 		pipe_count = RCU_TORTURE_PIPE_LEN;
 	}
-	if (pipe_count > 1)
+	completed_end = cur_ops->completed();
+	if (pipe_count > 1) {
+		unsigned long __maybe_unused ts_rem = do_div(ts, NSEC_PER_USEC);
+
+		do_trace_rcu_torture_read(cur_ops->name, &p->rtort_rcu, ts,
+					  completed, completed_end);
 		rcutorture_trace_dump();
+	}
 	__this_cpu_inc(rcu_torture_count[pipe_count]);
-	completed = cur_ops->completed() - completed;
+	completed = completed_end - completed;
 	if (completed > RCU_TORTURE_PIPE_LEN) {
 		/* Should not happen, but... */
 		completed = RCU_TORTURE_PIPE_LEN;

@@ -1094,11 +1102,13 @@ static int
 rcu_torture_reader(void *arg)
 {
 	int completed;
+	int completed_end;
 	int idx;
 	DEFINE_RCU_RANDOM(rand);
 	struct rcu_torture *p;
 	int pipe_count;
 	struct timer_list t;
+	unsigned long long ts;
 
 	VERBOSE_PRINTK_STRING("rcu_torture_reader task started");
 	set_user_nice(current, 19);

@@ -1112,6 +1122,7 @@ rcu_torture_reader(void *arg)
 		}
 		idx = cur_ops->readlock();
 		completed = cur_ops->completed();
+		ts = trace_clock_local();
 		p = rcu_dereference_check(rcu_torture_current,
 					  rcu_read_lock_bh_held() ||
 					  rcu_read_lock_sched_held() ||

@@ -1122,7 +1133,6 @@ rcu_torture_reader(void *arg)
 			schedule_timeout_interruptible(HZ);
 			continue;
 		}
-		do_trace_rcu_torture_read(cur_ops->name, &p->rtort_rcu);
 		if (p->rtort_mbtest == 0)
 			atomic_inc(&n_rcu_torture_mberror);
 		cur_ops->read_delay(&rand);

@@ -1132,10 +1142,17 @@ rcu_torture_reader(void *arg)
 			/* Should not happen, but... */
 			pipe_count = RCU_TORTURE_PIPE_LEN;
 		}
-		if (pipe_count > 1)
+		completed_end = cur_ops->completed();
+		if (pipe_count > 1) {
+			unsigned long __maybe_unused ts_rem =
+					do_div(ts, NSEC_PER_USEC);
+
+			do_trace_rcu_torture_read(cur_ops->name, &p->rtort_rcu,
+						  ts, completed, completed_end);
 			rcutorture_trace_dump();
+		}
 		__this_cpu_inc(rcu_torture_count[pipe_count]);
-		completed = cur_ops->completed() - completed;
+		completed = completed_end - completed;
 		if (completed > RCU_TORTURE_PIPE_LEN) {
 			/* Should not happen, but... */
 			completed = RCU_TORTURE_PIPE_LEN;
Signature · 1 line
1.7.8
author
#1 lore raw

PATCH 1/3 tracing: Export trace_clock_local()

In reply to Paul E. McKenney

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

The rcutorture tests need to be able to trace the time of the beginning of an RCU read-side critical section, and thus need access to trace_clock_local(). This commit therefore adds a the needed EXPORT_SYMBOL_GPL().

Signed-off-by Paul E. McKenney <[email protected]>
---
 kernel/trace/trace_clock.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
kernel/trace/trace_clock.c +1 -0
index 3947835..1bbb1b2 100644
@@ -44,6 +44,7 @@ u64 notrace trace_clock_local(void)
 
 	return clock;
 }
+EXPORT_SYMBOL_GPL(trace_clock_local);
 
 /*
  * trace_clock(): 'between' trace clock. Not completely serialized,
Signature · 1 line
1.7.8
maintainer Reviewed-by on the v1 series
#4 lore raw
In reply to Paul E. McKenney
Quoted 4 lines · Paul E. McKenney

Hello!

The following are changes to documentation and rcutorture:

Quoted 8 lines

1. Export trace_clock_local() in order to allow rcutorture event
tracing to emit the time of the beginning of the RCU read-side critical section at the point where a failure is detected.
2. Reduce rcutorture's read-side tracing to include only failures,
thus cutting the number of events down to something reasonable (normally zero, in fact!).
3. Add atomic_xchg() to the list of atomic operations and memory
barriers, courtesy of Richard Braun.

For the whole series:

Reviewed-by Josh Triplett <[email protected]>