LKML Archive on lore.kernel.org
 help / Atom feed
* [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints
@ 2018-06-28 18:21 Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 1/7] srcu: Add notrace variants of srcu_read_{lock,unlock} Joel Fernandes
                   ` (7 more replies)
  0 siblings, 8 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

This is a posting of v9 preempt/irq tracepoint clean up series rebased
onto v4.18-rc2. No changes in the series, just a rebase + repost.

All patches have a Reviewed-by tags now from reviewers. This series has
been well tested and is a simplification/refactoring of existing code,
along with giving a speed-up for tracepoints using the rcu-idle API.
With this our users will find it easier to use tools depending on
existing preempt tracepoints since it simplifies the configuration for
them.

Future enhancements/fixes I am developing for preempt-off tracer will
depend on these patches, so I suggest prioritizing these well reviewed
and tested patches for that reason as well.

Introduction to the series:
The preempt/irq tracepoints exist but not everything in the kernel is using it
whenever they need to be notified that a preempt disable/enable or an irq
disable/enable has occurred.  This makes things not work simultaneously (for
example, only either lockdep or irqsoff trace-events can be used at a time).

This is particularly painful to deal with, since turning on lockdep breaks
tracers that install probes on IRQ events, such as the BCC atomic critical
section tracer [1]. This constraint also makes it impossible to use synthetic
events to trace irqsoff sections with lockdep simulataneously turned on.

This series solves that, and also results in a nice clean up of relevant parts
of the kernel. Several ifdefs are simpler, and the design is more unified and
better. Also as a result of this, we also speeded performance all rcuidle
tracepoints since their handling is simpler.

[1] https://github.com/iovisor/bcc/blob/master/tools/criticalstat_example.txt

v8->v9:
- Small style changes to tracepoint code (Mathieu)
- Minor style fix to use PTR_ERR_OR_ZERO (0-day bot)
- Minor fix to test_atomic_sections to use unsigned long.
- Added Namhyung's, Mathieu's Reviewed-by to some patches.
- Added Acks from Matsami

v7->v8:
- Refactored irqsoff tracer probe defines (Namhyung)

v6->v7:
- Added a module to simulate an atomic section, a kselftest to load and
  and trigger it which verifies the preempt-tracer and this series.

- Fixed a new warning after I rebased in early boot, this is because
early_boot_irqs_disabled was set too early, I moved it after the lockdep
initialization.

- added back the softirq fix since it appears it wasn't picked up.

- Ran Ingo's locking API selftest suite which are passing with this
  series.

- Mathieu suggested ifdef'ing the tracepoint_synchronize_unregister
  function incase tracepoints aren't enabled, did that.


Joel Fernandes (Google) (6):
  srcu: Add notrace variant of srcu_dereference
  trace/irqsoff: Split reset into separate functions
  tracepoint: Make rcuidle tracepoint callers use SRCU
  tracing: Centralize preemptirq tracepoints and unify their usage
  lib: Add module to simulate atomic sections for testing preemptoff
    tracers
  kselftests: Add tests for the preemptoff and irqsoff tracers

Paul McKenney (1):
  srcu: Add notrace variants of srcu_read_{lock,unlock}

 include/linux/ftrace.h                        |  11 +-
 include/linux/irqflags.h                      |  11 +-
 include/linux/lockdep.h                       |   8 +-
 include/linux/preempt.h                       |   2 +-
 include/linux/srcu.h                          |  22 ++
 include/linux/tracepoint.h                    |  49 +++-
 include/trace/events/preemptirq.h             |  23 +-
 init/main.c                                   |   5 +-
 kernel/locking/lockdep.c                      |  35 +--
 kernel/sched/core.c                           |   2 +-
 kernel/trace/Kconfig                          |  22 +-
 kernel/trace/Makefile                         |   2 +-
 kernel/trace/trace_irqsoff.c                  | 253 ++++++------------
 kernel/trace/trace_preemptirq.c               |  71 +++++
 kernel/tracepoint.c                           |  16 +-
 lib/Kconfig.debug                             |   8 +
 lib/Makefile                                  |   1 +
 lib/test_atomic_sections.c                    |  77 ++++++
 tools/testing/selftests/ftrace/config         |   3 +
 .../test.d/preemptirq/irqsoff_tracer.tc       |  73 +++++
 20 files changed, 453 insertions(+), 241 deletions(-)
 create mode 100644 kernel/trace/trace_preemptirq.c
 create mode 100644 lib/test_atomic_sections.c
 create mode 100644 tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc

-- 
2.18.0.rc2.346.g013aa6912e-goog



^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 1/7] srcu: Add notrace variants of srcu_read_{lock,unlock}
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 2/7] srcu: Add notrace variant of srcu_dereference Joel Fernandes
                   ` (6 subsequent siblings)
  7 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Paul McKenney, Joel Fernandes, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: Paul McKenney <paulmck@linux.vnet.ibm.com>

This is needed for a future tracepoint patch that uses srcu, and to make
sure it doesn't call into lockdep.

tracepoint code already calls notrace variants for rcu_read_lock_sched
so this patch does the same for srcu which will be used in a later
patch. Keeps it consistent with rcu-sched.

[Joel: Added commit message]
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Paul McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/srcu.h | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index 91494d7e8e41..3e72a291c401 100644
--- a/include/linux/srcu.h
+++ b/include/linux/srcu.h
@@ -195,6 +195,16 @@ static inline int srcu_read_lock(struct srcu_struct *sp) __acquires(sp)
 	return retval;
 }
 
+/* Used by tracing, cannot be traced and cannot invoke lockdep. */
+static inline notrace int
+srcu_read_lock_notrace(struct srcu_struct *sp) __acquires(sp)
+{
+	int retval;
+
+	retval = __srcu_read_lock(sp);
+	return retval;
+}
+
 /**
  * srcu_read_unlock - unregister a old reader from an SRCU-protected structure.
  * @sp: srcu_struct in which to unregister the old reader.
@@ -209,6 +219,13 @@ static inline void srcu_read_unlock(struct srcu_struct *sp, int idx)
 	__srcu_read_unlock(sp, idx);
 }
 
+/* Used by tracing, cannot be traced and cannot call lockdep. */
+static inline notrace void
+srcu_read_unlock_notrace(struct srcu_struct *sp, int idx) __releases(sp)
+{
+	__srcu_read_unlock(sp, idx);
+}
+
 /**
  * smp_mb__after_srcu_read_unlock - ensure full ordering after srcu_read_unlock
  *
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 2/7] srcu: Add notrace variant of srcu_dereference
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 1/7] srcu: Add notrace variants of srcu_read_{lock,unlock} Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 3/7] trace/irqsoff: Split reset into separate functions Joel Fernandes
                   ` (5 subsequent siblings)
  7 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In the last patch in this series, we are making lockdep register hooks
onto the irq_{disable,enable} tracepoints. These tracepoints use the
_rcuidle tracepoint variant. In this series we switch the _rcuidle
tracepoint callers to use SRCU instead of sched-RCU. Inorder to
dereference the pointer to the probe functions, we could call
srcu_dereference, however this API will call back into lockdep to check
if the lock is held *before* the lockdep probe hooks have a chance to
run and annotate the IRQ enabled/disabled state.

For this reason we need a notrace variant of srcu_dereference since
otherwise we get lockdep splats. This patch adds the needed
srcu_dereference_notrace variant.

Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/srcu.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/linux/srcu.h b/include/linux/srcu.h
index 3e72a291c401..67135d4a8a30 100644
--- a/include/linux/srcu.h
+++ b/include/linux/srcu.h
@@ -169,6 +169,11 @@ static inline int srcu_read_lock_held(const struct srcu_struct *sp)
  */
 #define srcu_dereference(p, sp) srcu_dereference_check((p), (sp), 0)
 
+/**
+ * srcu_dereference_notrace - no tracing and no lockdep calls from here
+ */
+#define srcu_dereference_notrace(p, sp) srcu_dereference_check((p), (sp), 1)
+
 /**
  * srcu_read_lock - register a new reader for an SRCU-protected structure.
  * @sp: srcu_struct in which to register the new reader.
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 3/7] trace/irqsoff: Split reset into separate functions
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 1/7] srcu: Add notrace variants of srcu_read_{lock,unlock} Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 2/7] srcu: Add notrace variant of srcu_dereference Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
                   ` (4 subsequent siblings)
  7 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

Split reset functions into seperate functions in preparation
of future patches that need to do tracer specific reset.

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 kernel/trace/trace_irqsoff.c | 22 +++++++++++++++++++---
 1 file changed, 19 insertions(+), 3 deletions(-)

diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
index 03ecb4465ee4..f8daa754cce2 100644
--- a/kernel/trace/trace_irqsoff.c
+++ b/kernel/trace/trace_irqsoff.c
@@ -634,7 +634,7 @@ static int __irqsoff_tracer_init(struct trace_array *tr)
 	return 0;
 }
 
-static void irqsoff_tracer_reset(struct trace_array *tr)
+static void __irqsoff_tracer_reset(struct trace_array *tr)
 {
 	int lat_flag = save_flags & TRACE_ITER_LATENCY_FMT;
 	int overwrite_flag = save_flags & TRACE_ITER_OVERWRITE;
@@ -665,6 +665,12 @@ static int irqsoff_tracer_init(struct trace_array *tr)
 
 	return __irqsoff_tracer_init(tr);
 }
+
+static void irqsoff_tracer_reset(struct trace_array *tr)
+{
+	__irqsoff_tracer_reset(tr);
+}
+
 static struct tracer irqsoff_tracer __read_mostly =
 {
 	.name		= "irqsoff",
@@ -697,11 +703,16 @@ static int preemptoff_tracer_init(struct trace_array *tr)
 	return __irqsoff_tracer_init(tr);
 }
 
+static void preemptoff_tracer_reset(struct trace_array *tr)
+{
+	__irqsoff_tracer_reset(tr);
+}
+
 static struct tracer preemptoff_tracer __read_mostly =
 {
 	.name		= "preemptoff",
 	.init		= preemptoff_tracer_init,
-	.reset		= irqsoff_tracer_reset,
+	.reset		= preemptoff_tracer_reset,
 	.start		= irqsoff_tracer_start,
 	.stop		= irqsoff_tracer_stop,
 	.print_max	= true,
@@ -731,11 +742,16 @@ static int preemptirqsoff_tracer_init(struct trace_array *tr)
 	return __irqsoff_tracer_init(tr);
 }
 
+static void preemptirqsoff_tracer_reset(struct trace_array *tr)
+{
+	__irqsoff_tracer_reset(tr);
+}
+
 static struct tracer preemptirqsoff_tracer __read_mostly =
 {
 	.name		= "preemptirqsoff",
 	.init		= preemptirqsoff_tracer_init,
-	.reset		= irqsoff_tracer_reset,
+	.reset		= preemptirqsoff_tracer_reset,
 	.start		= irqsoff_tracer_start,
 	.stop		= irqsoff_tracer_stop,
 	.print_max	= true,
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
                   ` (2 preceding siblings ...)
  2018-06-28 18:21 ` [PATCH v9 3/7] trace/irqsoff: Split reset into separate functions Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-07-11 12:49   ` Peter Zijlstra
                     ` (2 more replies)
  2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
                   ` (3 subsequent siblings)
  7 siblings, 3 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In recent tests with IRQ on/off tracepoints, a large performance
overhead ~10% is noticed when running hackbench. This is root caused to
calls to rcu_irq_enter_irqson and rcu_irq_exit_irqson from the
tracepoint code. Following a long discussion on the list [1] about this,
we concluded that srcu is a better alternative for use during rcu idle.
Although it does involve extra barriers, its lighter than the sched-rcu
version which has to do additional RCU calls to notify RCU idle about
entry into RCU sections.

In this patch, we change the underlying implementation of the
trace_*_rcuidle API to use SRCU. This has shown to improve performance
alot for the high frequency irq enable/disable tracepoints.

Test: Tested idle and preempt/irq tracepoints.

Here are some performance numbers:

With a run of the following 30 times on a single core x86 Qemu instance
with 1GB memory:
hackbench -g 4 -f 2 -l 3000

Completion times in seconds. CONFIG_PROVE_LOCKING=y.

No patches (without this series)
Mean: 3.048
Median: 3.025
Std Dev: 0.064

With Lockdep using irq tracepoints with RCU implementation:
Mean: 3.451   (-11.66 %)
Median: 3.447 (-12.22%)
Std Dev: 0.049

With Lockdep using irq tracepoints with SRCU implementation (this series):
Mean: 3.020   (I would consider the improvement against the "without
	       this series" case as just noise).
Median: 3.013
Std Dev: 0.033

[1] https://patchwork.kernel.org/patch/10344297/

Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/tracepoint.h | 49 +++++++++++++++++++++++++++++++-------
 kernel/tracepoint.c        | 16 ++++++++++++-
 2 files changed, 56 insertions(+), 9 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 19a690b559ca..beeb01e147f8 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -15,6 +15,7 @@
  */
 
 #include <linux/smp.h>
+#include <linux/srcu.h>
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/cpumask.h>
@@ -33,6 +34,8 @@ struct trace_eval_map {
 
 #define TRACEPOINT_DEFAULT_PRIO	10
 
+extern struct srcu_struct tracepoint_srcu;
+
 extern int
 tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data);
 extern int
@@ -75,10 +78,16 @@ int unregister_tracepoint_module_notifier(struct notifier_block *nb)
  * probe unregistration and the end of module exit to make sure there is no
  * caller executing a probe when it is freed.
  */
+#ifdef CONFIG_TRACEPOINTS
 static inline void tracepoint_synchronize_unregister(void)
 {
+	synchronize_srcu(&tracepoint_srcu);
 	synchronize_sched();
 }
+#else
+static inline void tracepoint_synchronize_unregister(void)
+{ }
+#endif
 
 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
 extern int syscall_regfunc(void);
@@ -129,18 +138,38 @@ extern void syscall_unregfunc(void);
  * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just
  * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto".
  */
-#define __DO_TRACE(tp, proto, args, cond, rcucheck)			\
+#define __DO_TRACE(tp, proto, args, cond, rcuidle)			\
 	do {								\
 		struct tracepoint_func *it_func_ptr;			\
 		void *it_func;						\
 		void *__data;						\
+		int __maybe_unused idx = 0;				\
 									\
 		if (!(cond))						\
 			return;						\
-		if (rcucheck)						\
-			rcu_irq_enter_irqson();				\
-		rcu_read_lock_sched_notrace();				\
-		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\
+									\
+		/*							\
+		 * For rcuidle callers, use srcu since sched-rcu	\
+		 * doesn't work from the idle path.			\
+		 */							\
+		if (rcuidle) {						\
+			if (in_nmi()) {					\
+				WARN_ON_ONCE(1);			\
+				return; /* no srcu from nmi */		\
+			}						\
+									\
+			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
+			it_func_ptr =					\
+				srcu_dereference_notrace((tp)->funcs,	\
+						&tracepoint_srcu);	\
+			/* To keep it consistent with !rcuidle path */	\
+			preempt_disable_notrace();			\
+		} else {						\
+			rcu_read_lock_sched_notrace();			\
+			it_func_ptr =					\
+				rcu_dereference_sched((tp)->funcs);	\
+		}							\
+									\
 		if (it_func_ptr) {					\
 			do {						\
 				it_func = (it_func_ptr)->func;		\
@@ -148,9 +177,13 @@ extern void syscall_unregfunc(void);
 				((void(*)(proto))(it_func))(args);	\
 			} while ((++it_func_ptr)->func);		\
 		}							\
-		rcu_read_unlock_sched_notrace();			\
-		if (rcucheck)						\
-			rcu_irq_exit_irqson();				\
+									\
+		if (rcuidle) {						\
+			preempt_enable_notrace();			\
+			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
+		} else {						\
+			rcu_read_unlock_sched_notrace();		\
+		}							\
 	} while (0)
 
 #ifndef MODULE
diff --git a/kernel/tracepoint.c b/kernel/tracepoint.c
index 6dc6356c3327..955148d91b74 100644
--- a/kernel/tracepoint.c
+++ b/kernel/tracepoint.c
@@ -31,6 +31,9 @@
 extern struct tracepoint * const __start___tracepoints_ptrs[];
 extern struct tracepoint * const __stop___tracepoints_ptrs[];
 
+DEFINE_SRCU(tracepoint_srcu);
+EXPORT_SYMBOL_GPL(tracepoint_srcu);
+
 /* Set to 1 to enable tracepoint debug output */
 static const int tracepoint_debug;
 
@@ -67,16 +70,27 @@ static inline void *allocate_probes(int count)
 	return p == NULL ? NULL : p->probes;
 }
 
-static void rcu_free_old_probes(struct rcu_head *head)
+static void srcu_free_old_probes(struct rcu_head *head)
 {
 	kfree(container_of(head, struct tp_probes, rcu));
 }
 
+static void rcu_free_old_probes(struct rcu_head *head)
+{
+	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
+}
+
 static inline void release_probes(struct tracepoint_func *old)
 {
 	if (old) {
 		struct tp_probes *tp_probes = container_of(old,
 			struct tp_probes, probes[0]);
+		/*
+		 * Tracepoint probes are protected by both sched RCU and SRCU,
+		 * by calling the SRCU callback in the sched RCU callback we
+		 * cover both cases. So let us chain the SRCU and sched RCU
+		 * callbacks to wait for both grace periods.
+		 */
 		call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
 	}
 }
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
                   ` (3 preceding siblings ...)
  2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-07-06 22:06   ` Steven Rostedt
                     ` (2 more replies)
  2018-06-28 18:21 ` [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers Joel Fernandes
                   ` (2 subsequent siblings)
  7 siblings, 3 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

This patch detaches the preemptirq tracepoints from the tracers and
keeps it separate.

Advantages:
* Lockdep and irqsoff event can now run in parallel since they no longer
have their own calls.

* This unifies the usecase of adding hooks to an irqsoff and irqson
event, and a preemptoff and preempton event.
  3 users of the events exist:
  - Lockdep
  - irqsoff and preemptoff tracers
  - irqs and preempt trace events

The unification cleans up several ifdefs and makes the code in preempt
tracer and irqsoff tracers simpler. It gets rid of all the horrific
ifdeferry around PROVE_LOCKING and makes configuration of the different
users of the tracepoints more easy and understandable. It also gets rid
of the time_* function calls from the lockdep hooks used to call into
the preemptirq tracer which is not needed anymore. The negative delta in
lines of code in this patch is quite large too.

In the patch we introduce a new CONFIG option PREEMPTIRQ_TRACEPOINTS
as a single point for registering probes onto the tracepoints. With
this,
the web of config options for preempt/irq toggle tracepoints and its
users becomes:

 PREEMPT_TRACER   PREEMPTIRQ_EVENTS  IRQSOFF_TRACER PROVE_LOCKING
       |                 |     \         |           |
       \    (selects)    /      \        \ (selects) /
      TRACE_PREEMPT_TOGGLE       ----> TRACE_IRQFLAGS
                      \                  /
                       \ (depends on)   /
                     PREEMPTIRQ_TRACEPOINTS

One note, I have to check for lockdep recursion in the code that calls
the trace events API and bail out if we're in lockdep recursion
protection to prevent something like the following case: a spin_lock is
taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
and then sets lockdep_recursion, and then calls __lockdep_acquired. In
this function, a call to get_lock_stats happens which calls
preempt_disable, which calls trace IRQS off somewhere which enters my
tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
This flag is then never cleared causing lockdep paths to never be
entered and thus causing splats and other bad things.

Other than the performance tests mentioned in the previous patch, I also
ran the locking API test suite. I verified that all tests cases are
passing.

I also injected issues by not registering lockdep probes onto the
tracepoints and I see failures to confirm that the probes are indeed
working.

This series + lockdep probes not registered (just to inject errors):
[    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
[    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
[    0.000000]        sirq-safe-A => hirqs-on/12:FAILED|FAILED|  ok  |
[    0.000000]        sirq-safe-A => hirqs-on/21:FAILED|FAILED|  ok  |
[    0.000000]          hard-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
[    0.000000]          soft-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
[    0.000000]          hard-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
[    0.000000]          soft-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
[    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
[    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |

With this series + lockdep probes registered, all locking tests pass:

[    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
[    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
[    0.000000]        sirq-safe-A => hirqs-on/12:  ok  |  ok  |  ok  |
[    0.000000]        sirq-safe-A => hirqs-on/21:  ok  |  ok  |  ok  |
[    0.000000]          hard-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
[    0.000000]          soft-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
[    0.000000]          hard-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
[    0.000000]          soft-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
[    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
[    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/ftrace.h            |  11 +-
 include/linux/irqflags.h          |  11 +-
 include/linux/lockdep.h           |   8 +-
 include/linux/preempt.h           |   2 +-
 include/trace/events/preemptirq.h |  23 +--
 init/main.c                       |   5 +-
 kernel/locking/lockdep.c          |  35 ++---
 kernel/sched/core.c               |   2 +-
 kernel/trace/Kconfig              |  22 ++-
 kernel/trace/Makefile             |   2 +-
 kernel/trace/trace_irqsoff.c      | 231 ++++++++----------------------
 kernel/trace/trace_preemptirq.c   |  71 +++++++++
 12 files changed, 194 insertions(+), 229 deletions(-)
 create mode 100644 kernel/trace/trace_preemptirq.c

diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 8154f4920fcb..f32e3c81407e 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -709,16 +709,7 @@ static inline unsigned long get_lock_parent_ip(void)
 	return CALLER_ADDR2;
 }
 
-#ifdef CONFIG_IRQSOFF_TRACER
-  extern void time_hardirqs_on(unsigned long a0, unsigned long a1);
-  extern void time_hardirqs_off(unsigned long a0, unsigned long a1);
-#else
-  static inline void time_hardirqs_on(unsigned long a0, unsigned long a1) { }
-  static inline void time_hardirqs_off(unsigned long a0, unsigned long a1) { }
-#endif
-
-#if defined(CONFIG_PREEMPT_TRACER) || \
-	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
+#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
   extern void trace_preempt_on(unsigned long a0, unsigned long a1);
   extern void trace_preempt_off(unsigned long a0, unsigned long a1);
 #else
diff --git a/include/linux/irqflags.h b/include/linux/irqflags.h
index 9700f00bbc04..50edb9cbbd26 100644
--- a/include/linux/irqflags.h
+++ b/include/linux/irqflags.h
@@ -15,9 +15,16 @@
 #include <linux/typecheck.h>
 #include <asm/irqflags.h>
 
-#ifdef CONFIG_TRACE_IRQFLAGS
+/* Currently trace_softirqs_on/off is used only by lockdep */
+#ifdef CONFIG_PROVE_LOCKING
   extern void trace_softirqs_on(unsigned long ip);
   extern void trace_softirqs_off(unsigned long ip);
+#else
+# define trace_softirqs_on(ip)	do { } while (0)
+# define trace_softirqs_off(ip)	do { } while (0)
+#endif
+
+#ifdef CONFIG_TRACE_IRQFLAGS
   extern void trace_hardirqs_on(void);
   extern void trace_hardirqs_off(void);
 # define trace_hardirq_context(p)	((p)->hardirq_context)
@@ -43,8 +50,6 @@ do {						\
 #else
 # define trace_hardirqs_on()		do { } while (0)
 # define trace_hardirqs_off()		do { } while (0)
-# define trace_softirqs_on(ip)		do { } while (0)
-# define trace_softirqs_off(ip)		do { } while (0)
 # define trace_hardirq_context(p)	0
 # define trace_softirq_context(p)	0
 # define trace_hardirqs_enabled(p)	0
diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index 6fc77d4dbdcd..a8113357ceeb 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -266,7 +266,8 @@ struct held_lock {
 /*
  * Initialization, self-test and debugging-output methods:
  */
-extern void lockdep_info(void);
+extern void lockdep_init(void);
+extern void lockdep_init_early(void);
 extern void lockdep_reset(void);
 extern void lockdep_reset_lock(struct lockdep_map *lock);
 extern void lockdep_free_key_range(void *start, unsigned long size);
@@ -406,7 +407,8 @@ static inline void lockdep_on(void)
 # define lock_downgrade(l, i)			do { } while (0)
 # define lock_set_class(l, n, k, s, i)		do { } while (0)
 # define lock_set_subclass(l, s, i)		do { } while (0)
-# define lockdep_info()				do { } while (0)
+# define lockdep_init()				do { } while (0)
+# define lockdep_init_early()			do { } while (0)
 # define lockdep_init_map(lock, name, key, sub) \
 		do { (void)(name); (void)(key); } while (0)
 # define lockdep_set_class(lock, key)		do { (void)(key); } while (0)
@@ -532,7 +534,7 @@ do {								\
 
 #endif /* CONFIG_LOCKDEP */
 
-#ifdef CONFIG_TRACE_IRQFLAGS
+#ifdef CONFIG_PROVE_LOCKING
 extern void print_irqtrace_events(struct task_struct *curr);
 #else
 static inline void print_irqtrace_events(struct task_struct *curr)
diff --git a/include/linux/preempt.h b/include/linux/preempt.h
index 5bd3f151da78..c01813c3fbe9 100644
--- a/include/linux/preempt.h
+++ b/include/linux/preempt.h
@@ -150,7 +150,7 @@
  */
 #define in_atomic_preempt_off() (preempt_count() != PREEMPT_DISABLE_OFFSET)
 
-#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_PREEMPT_TRACER)
+#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_TRACE_PREEMPT_TOGGLE)
 extern void preempt_count_add(int val);
 extern void preempt_count_sub(int val);
 #define preempt_count_dec_and_test() \
diff --git a/include/trace/events/preemptirq.h b/include/trace/events/preemptirq.h
index 9c4eb33c5a1d..9a0d4ceeb166 100644
--- a/include/trace/events/preemptirq.h
+++ b/include/trace/events/preemptirq.h
@@ -1,4 +1,4 @@
-#ifdef CONFIG_PREEMPTIRQ_EVENTS
+#ifdef CONFIG_PREEMPTIRQ_TRACEPOINTS
 
 #undef TRACE_SYSTEM
 #define TRACE_SYSTEM preemptirq
@@ -32,7 +32,7 @@ DECLARE_EVENT_CLASS(preemptirq_template,
 		  (void *)((unsigned long)(_stext) + __entry->parent_offs))
 );
 
-#ifndef CONFIG_PROVE_LOCKING
+#ifdef CONFIG_TRACE_IRQFLAGS
 DEFINE_EVENT(preemptirq_template, irq_disable,
 	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
 	     TP_ARGS(ip, parent_ip));
@@ -40,9 +40,14 @@ DEFINE_EVENT(preemptirq_template, irq_disable,
 DEFINE_EVENT(preemptirq_template, irq_enable,
 	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
 	     TP_ARGS(ip, parent_ip));
+#else
+#define trace_irq_enable(...)
+#define trace_irq_disable(...)
+#define trace_irq_enable_rcuidle(...)
+#define trace_irq_disable_rcuidle(...)
 #endif
 
-#ifdef CONFIG_DEBUG_PREEMPT
+#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
 DEFINE_EVENT(preemptirq_template, preempt_disable,
 	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
 	     TP_ARGS(ip, parent_ip));
@@ -50,22 +55,22 @@ DEFINE_EVENT(preemptirq_template, preempt_disable,
 DEFINE_EVENT(preemptirq_template, preempt_enable,
 	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
 	     TP_ARGS(ip, parent_ip));
+#else
+#define trace_preempt_enable(...)
+#define trace_preempt_disable(...)
+#define trace_preempt_enable_rcuidle(...)
+#define trace_preempt_disable_rcuidle(...)
 #endif
 
 #endif /* _TRACE_PREEMPTIRQ_H */
 
 #include <trace/define_trace.h>
 
-#endif /* !CONFIG_PREEMPTIRQ_EVENTS */
-
-#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || defined(CONFIG_PROVE_LOCKING)
+#else /* !CONFIG_PREEMPTIRQ_TRACEPOINTS */
 #define trace_irq_enable(...)
 #define trace_irq_disable(...)
 #define trace_irq_enable_rcuidle(...)
 #define trace_irq_disable_rcuidle(...)
-#endif
-
-#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || !defined(CONFIG_DEBUG_PREEMPT)
 #define trace_preempt_enable(...)
 #define trace_preempt_disable(...)
 #define trace_preempt_enable_rcuidle(...)
diff --git a/init/main.c b/init/main.c
index 3b4ada11ed52..44fe43be84c1 100644
--- a/init/main.c
+++ b/init/main.c
@@ -648,6 +648,9 @@ asmlinkage __visible void __init start_kernel(void)
 	profile_init();
 	call_function_init();
 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
+
+	lockdep_init_early();
+
 	early_boot_irqs_disabled = false;
 	local_irq_enable();
 
@@ -663,7 +666,7 @@ asmlinkage __visible void __init start_kernel(void)
 		panic("Too many boot %s vars at `%s'", panic_later,
 		      panic_param);
 
-	lockdep_info();
+	lockdep_init();
 
 	/*
 	 * Need to run this when irqs are enabled, because it wants
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 5fa4d3138bf1..b961a1698e98 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -55,6 +55,7 @@
 
 #include "lockdep_internals.h"
 
+#include <trace/events/preemptirq.h>
 #define CREATE_TRACE_POINTS
 #include <trace/events/lock.h>
 
@@ -2845,10 +2846,9 @@ static void __trace_hardirqs_on_caller(unsigned long ip)
 	debug_atomic_inc(hardirqs_on_events);
 }
 
-__visible void trace_hardirqs_on_caller(unsigned long ip)
+static void lockdep_hardirqs_on(void *none, unsigned long ignore,
+				unsigned long ip)
 {
-	time_hardirqs_on(CALLER_ADDR0, ip);
-
 	if (unlikely(!debug_locks || current->lockdep_recursion))
 		return;
 
@@ -2887,23 +2887,15 @@ __visible void trace_hardirqs_on_caller(unsigned long ip)
 	__trace_hardirqs_on_caller(ip);
 	current->lockdep_recursion = 0;
 }
-EXPORT_SYMBOL(trace_hardirqs_on_caller);
-
-void trace_hardirqs_on(void)
-{
-	trace_hardirqs_on_caller(CALLER_ADDR0);
-}
-EXPORT_SYMBOL(trace_hardirqs_on);
 
 /*
  * Hardirqs were disabled:
  */
-__visible void trace_hardirqs_off_caller(unsigned long ip)
+static void lockdep_hardirqs_off(void *none, unsigned long ignore,
+				 unsigned long ip)
 {
 	struct task_struct *curr = current;
 
-	time_hardirqs_off(CALLER_ADDR0, ip);
-
 	if (unlikely(!debug_locks || current->lockdep_recursion))
 		return;
 
@@ -2925,13 +2917,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
 	} else
 		debug_atomic_inc(redundant_hardirqs_off);
 }
-EXPORT_SYMBOL(trace_hardirqs_off_caller);
-
-void trace_hardirqs_off(void)
-{
-	trace_hardirqs_off_caller(CALLER_ADDR0);
-}
-EXPORT_SYMBOL(trace_hardirqs_off);
 
 /*
  * Softirqs will be enabled:
@@ -4338,7 +4323,15 @@ void lockdep_reset_lock(struct lockdep_map *lock)
 	raw_local_irq_restore(flags);
 }
 
-void __init lockdep_info(void)
+void __init lockdep_init_early(void)
+{
+#ifdef CONFIG_PROVE_LOCKING
+	register_trace_prio_irq_disable(lockdep_hardirqs_off, NULL, INT_MAX);
+	register_trace_prio_irq_enable(lockdep_hardirqs_on, NULL, INT_MIN);
+#endif
+}
+
+void __init lockdep_init(void)
 {
 	printk("Lock dependency validator: Copyright (c) 2006 Red Hat, Inc., Ingo Molnar\n");
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 78d8facba456..4c956f6849ec 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3192,7 +3192,7 @@ static inline void sched_tick_stop(int cpu) { }
 #endif
 
 #if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
-				defined(CONFIG_PREEMPT_TRACER))
+				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
 /*
  * If the value passed in is equal to the current preempt count
  * then we just disabled preemption. Start timing the latency.
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index dcc0166d1997..8d51351e3149 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -82,6 +82,15 @@ config RING_BUFFER_ALLOW_SWAP
 	 Allow the use of ring_buffer_swap_cpu.
 	 Adds a very slight overhead to tracing when enabled.
 
+config PREEMPTIRQ_TRACEPOINTS
+	bool
+	depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
+	select TRACING
+	default y
+	help
+	  Create preempt/irq toggle tracepoints if needed, so that other parts
+	  of the kernel can use them to generate or add hooks to them.
+
 # All tracer options should select GENERIC_TRACER. For those options that are
 # enabled by all tracers (context switch and event tracer) they select TRACING.
 # This allows those options to appear when no other tracer is selected. But the
@@ -155,18 +164,20 @@ config FUNCTION_GRAPH_TRACER
 	  the return value. This is done by setting the current return
 	  address on the current task structure into a stack of calls.
 
+config TRACE_PREEMPT_TOGGLE
+	bool
+	help
+	  Enables hooks which will be called when preemption is first disabled,
+	  and last enabled.
 
 config PREEMPTIRQ_EVENTS
 	bool "Enable trace events for preempt and irq disable/enable"
 	select TRACE_IRQFLAGS
-	depends on DEBUG_PREEMPT || !PROVE_LOCKING
-	depends on TRACING
+	select TRACE_PREEMPT_TOGGLE if PREEMPT
+	select GENERIC_TRACER
 	default n
 	help
 	  Enable tracing of disable and enable events for preemption and irqs.
-	  For tracing preempt disable/enable events, DEBUG_PREEMPT must be
-	  enabled. For tracing irq disable/enable events, PROVE_LOCKING must
-	  be disabled.
 
 config IRQSOFF_TRACER
 	bool "Interrupts-off Latency Tracer"
@@ -203,6 +214,7 @@ config PREEMPT_TRACER
 	select RING_BUFFER_ALLOW_SWAP
 	select TRACER_SNAPSHOT
 	select TRACER_SNAPSHOT_PER_CPU_SWAP
+	select TRACE_PREEMPT_TOGGLE
 	help
 	  This option measures the time spent in preemption-off critical
 	  sections, with microsecond accuracy.
diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile
index e2538c7638d4..84a0cb222f20 100644
--- a/kernel/trace/Makefile
+++ b/kernel/trace/Makefile
@@ -35,7 +35,7 @@ obj-$(CONFIG_TRACING) += trace_printk.o
 obj-$(CONFIG_TRACING_MAP) += tracing_map.o
 obj-$(CONFIG_CONTEXT_SWITCH_TRACER) += trace_sched_switch.o
 obj-$(CONFIG_FUNCTION_TRACER) += trace_functions.o
-obj-$(CONFIG_PREEMPTIRQ_EVENTS) += trace_irqsoff.o
+obj-$(CONFIG_PREEMPTIRQ_TRACEPOINTS) += trace_preemptirq.o
 obj-$(CONFIG_IRQSOFF_TRACER) += trace_irqsoff.o
 obj-$(CONFIG_PREEMPT_TRACER) += trace_irqsoff.o
 obj-$(CONFIG_SCHED_TRACER) += trace_sched_wakeup.o
diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
index f8daa754cce2..770cd30cda40 100644
--- a/kernel/trace/trace_irqsoff.c
+++ b/kernel/trace/trace_irqsoff.c
@@ -16,7 +16,6 @@
 
 #include "trace.h"
 
-#define CREATE_TRACE_POINTS
 #include <trace/events/preemptirq.h>
 
 #if defined(CONFIG_IRQSOFF_TRACER) || defined(CONFIG_PREEMPT_TRACER)
@@ -450,66 +449,6 @@ void stop_critical_timings(void)
 }
 EXPORT_SYMBOL_GPL(stop_critical_timings);
 
-#ifdef CONFIG_IRQSOFF_TRACER
-#ifdef CONFIG_PROVE_LOCKING
-void time_hardirqs_on(unsigned long a0, unsigned long a1)
-{
-	if (!preempt_trace() && irq_trace())
-		stop_critical_timing(a0, a1);
-}
-
-void time_hardirqs_off(unsigned long a0, unsigned long a1)
-{
-	if (!preempt_trace() && irq_trace())
-		start_critical_timing(a0, a1);
-}
-
-#else /* !CONFIG_PROVE_LOCKING */
-
-/*
- * We are only interested in hardirq on/off events:
- */
-static inline void tracer_hardirqs_on(void)
-{
-	if (!preempt_trace() && irq_trace())
-		stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
-}
-
-static inline void tracer_hardirqs_off(void)
-{
-	if (!preempt_trace() && irq_trace())
-		start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
-}
-
-static inline void tracer_hardirqs_on_caller(unsigned long caller_addr)
-{
-	if (!preempt_trace() && irq_trace())
-		stop_critical_timing(CALLER_ADDR0, caller_addr);
-}
-
-static inline void tracer_hardirqs_off_caller(unsigned long caller_addr)
-{
-	if (!preempt_trace() && irq_trace())
-		start_critical_timing(CALLER_ADDR0, caller_addr);
-}
-
-#endif /* CONFIG_PROVE_LOCKING */
-#endif /*  CONFIG_IRQSOFF_TRACER */
-
-#ifdef CONFIG_PREEMPT_TRACER
-static inline void tracer_preempt_on(unsigned long a0, unsigned long a1)
-{
-	if (preempt_trace() && !irq_trace())
-		stop_critical_timing(a0, a1);
-}
-
-static inline void tracer_preempt_off(unsigned long a0, unsigned long a1)
-{
-	if (preempt_trace() && !irq_trace())
-		start_critical_timing(a0, a1);
-}
-#endif /* CONFIG_PREEMPT_TRACER */
-
 #ifdef CONFIG_FUNCTION_TRACER
 static bool function_enabled;
 
@@ -659,15 +598,34 @@ static void irqsoff_tracer_stop(struct trace_array *tr)
 }
 
 #ifdef CONFIG_IRQSOFF_TRACER
+/*
+ * We are only interested in hardirq on/off events:
+ */
+static void tracer_hardirqs_on(void *none, unsigned long a0, unsigned long a1)
+{
+	if (!preempt_trace() && irq_trace())
+		stop_critical_timing(a0, a1);
+}
+
+static void tracer_hardirqs_off(void *none, unsigned long a0, unsigned long a1)
+{
+	if (!preempt_trace() && irq_trace())
+		start_critical_timing(a0, a1);
+}
+
 static int irqsoff_tracer_init(struct trace_array *tr)
 {
 	trace_type = TRACER_IRQS_OFF;
 
+	register_trace_irq_disable(tracer_hardirqs_off, NULL);
+	register_trace_irq_enable(tracer_hardirqs_on, NULL);
 	return __irqsoff_tracer_init(tr);
 }
 
 static void irqsoff_tracer_reset(struct trace_array *tr)
 {
+	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
+	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
 	__irqsoff_tracer_reset(tr);
 }
 
@@ -690,21 +648,34 @@ static struct tracer irqsoff_tracer __read_mostly =
 	.allow_instances = true,
 	.use_max_tr	= true,
 };
-# define register_irqsoff(trace) register_tracer(&trace)
-#else
-# define register_irqsoff(trace) do { } while (0)
-#endif
+#endif /*  CONFIG_IRQSOFF_TRACER */
 
 #ifdef CONFIG_PREEMPT_TRACER
+static void tracer_preempt_on(void *none, unsigned long a0, unsigned long a1)
+{
+	if (preempt_trace() && !irq_trace())
+		stop_critical_timing(a0, a1);
+}
+
+static void tracer_preempt_off(void *none, unsigned long a0, unsigned long a1)
+{
+	if (preempt_trace() && !irq_trace())
+		start_critical_timing(a0, a1);
+}
+
 static int preemptoff_tracer_init(struct trace_array *tr)
 {
 	trace_type = TRACER_PREEMPT_OFF;
 
+	register_trace_preempt_disable(tracer_preempt_off, NULL);
+	register_trace_preempt_enable(tracer_preempt_on, NULL);
 	return __irqsoff_tracer_init(tr);
 }
 
 static void preemptoff_tracer_reset(struct trace_array *tr)
 {
+	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
+	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
 	__irqsoff_tracer_reset(tr);
 }
 
@@ -727,23 +698,29 @@ static struct tracer preemptoff_tracer __read_mostly =
 	.allow_instances = true,
 	.use_max_tr	= true,
 };
-# define register_preemptoff(trace) register_tracer(&trace)
-#else
-# define register_preemptoff(trace) do { } while (0)
-#endif
+#endif /* CONFIG_PREEMPT_TRACER */
 
-#if defined(CONFIG_IRQSOFF_TRACER) && \
-	defined(CONFIG_PREEMPT_TRACER)
+#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
 
 static int preemptirqsoff_tracer_init(struct trace_array *tr)
 {
 	trace_type = TRACER_IRQS_OFF | TRACER_PREEMPT_OFF;
 
+	register_trace_irq_disable(tracer_hardirqs_off, NULL);
+	register_trace_irq_enable(tracer_hardirqs_on, NULL);
+	register_trace_preempt_disable(tracer_preempt_off, NULL);
+	register_trace_preempt_enable(tracer_preempt_on, NULL);
+
 	return __irqsoff_tracer_init(tr);
 }
 
 static void preemptirqsoff_tracer_reset(struct trace_array *tr)
 {
+	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
+	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
+	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
+	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
+
 	__irqsoff_tracer_reset(tr);
 }
 
@@ -766,115 +743,21 @@ static struct tracer preemptirqsoff_tracer __read_mostly =
 	.allow_instances = true,
 	.use_max_tr	= true,
 };
-
-# define register_preemptirqsoff(trace) register_tracer(&trace)
-#else
-# define register_preemptirqsoff(trace) do { } while (0)
 #endif
 
 __init static int init_irqsoff_tracer(void)
 {
-	register_irqsoff(irqsoff_tracer);
-	register_preemptoff(preemptoff_tracer);
-	register_preemptirqsoff(preemptirqsoff_tracer);
-
-	return 0;
-}
-core_initcall(init_irqsoff_tracer);
-#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
-
-#ifndef CONFIG_IRQSOFF_TRACER
-static inline void tracer_hardirqs_on(void) { }
-static inline void tracer_hardirqs_off(void) { }
-static inline void tracer_hardirqs_on_caller(unsigned long caller_addr) { }
-static inline void tracer_hardirqs_off_caller(unsigned long caller_addr) { }
+#ifdef CONFIG_IRQSOFF_TRACER
+	register_tracer(&irqsoff_tracer);
 #endif
-
-#ifndef CONFIG_PREEMPT_TRACER
-static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
-static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
+#ifdef CONFIG_PREEMPT_TRACER
+	register_tracer(&preemptoff_tracer);
 #endif
-
-#if defined(CONFIG_TRACE_IRQFLAGS) && !defined(CONFIG_PROVE_LOCKING)
-/* Per-cpu variable to prevent redundant calls when IRQs already off */
-static DEFINE_PER_CPU(int, tracing_irq_cpu);
-
-void trace_hardirqs_on(void)
-{
-	if (!this_cpu_read(tracing_irq_cpu))
-		return;
-
-	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
-	tracer_hardirqs_on();
-
-	this_cpu_write(tracing_irq_cpu, 0);
-}
-EXPORT_SYMBOL(trace_hardirqs_on);
-
-void trace_hardirqs_off(void)
-{
-	if (this_cpu_read(tracing_irq_cpu))
-		return;
-
-	this_cpu_write(tracing_irq_cpu, 1);
-
-	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
-	tracer_hardirqs_off();
-}
-EXPORT_SYMBOL(trace_hardirqs_off);
-
-__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
-{
-	if (!this_cpu_read(tracing_irq_cpu))
-		return;
-
-	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
-	tracer_hardirqs_on_caller(caller_addr);
-
-	this_cpu_write(tracing_irq_cpu, 0);
-}
-EXPORT_SYMBOL(trace_hardirqs_on_caller);
-
-__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
-{
-	if (this_cpu_read(tracing_irq_cpu))
-		return;
-
-	this_cpu_write(tracing_irq_cpu, 1);
-
-	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
-	tracer_hardirqs_off_caller(caller_addr);
-}
-EXPORT_SYMBOL(trace_hardirqs_off_caller);
-
-/*
- * Stubs:
- */
-
-void trace_softirqs_on(unsigned long ip)
-{
-}
-
-void trace_softirqs_off(unsigned long ip)
-{
-}
-
-inline void print_irqtrace_events(struct task_struct *curr)
-{
-}
+#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
+	register_tracer(&preemptirqsoff_tracer);
 #endif
 
-#if defined(CONFIG_PREEMPT_TRACER) || \
-	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
-void trace_preempt_on(unsigned long a0, unsigned long a1)
-{
-	trace_preempt_enable_rcuidle(a0, a1);
-	tracer_preempt_on(a0, a1);
-}
-
-void trace_preempt_off(unsigned long a0, unsigned long a1)
-{
-	trace_preempt_disable_rcuidle(a0, a1);
-	tracer_preempt_off(a0, a1);
+	return 0;
 }
-#endif
+core_initcall(init_irqsoff_tracer);
+#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
diff --git a/kernel/trace/trace_preemptirq.c b/kernel/trace/trace_preemptirq.c
new file mode 100644
index 000000000000..dc01c7f4d326
--- /dev/null
+++ b/kernel/trace/trace_preemptirq.c
@@ -0,0 +1,71 @@
+/*
+ * preemptoff and irqoff tracepoints
+ *
+ * Copyright (C) Joel Fernandes (Google) <joel@joelfernandes.org>
+ */
+
+#include <linux/kallsyms.h>
+#include <linux/uaccess.h>
+#include <linux/module.h>
+#include <linux/ftrace.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/preemptirq.h>
+
+#ifdef CONFIG_TRACE_IRQFLAGS
+/* Per-cpu variable to prevent redundant calls when IRQs already off */
+static DEFINE_PER_CPU(int, tracing_irq_cpu);
+
+void trace_hardirqs_on(void)
+{
+	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
+		return;
+
+	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
+	this_cpu_write(tracing_irq_cpu, 0);
+}
+EXPORT_SYMBOL(trace_hardirqs_on);
+
+void trace_hardirqs_off(void)
+{
+	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
+		return;
+
+	this_cpu_write(tracing_irq_cpu, 1);
+	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
+}
+EXPORT_SYMBOL(trace_hardirqs_off);
+
+__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
+{
+	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
+		return;
+
+	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
+	this_cpu_write(tracing_irq_cpu, 0);
+}
+EXPORT_SYMBOL(trace_hardirqs_on_caller);
+
+__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
+{
+	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
+		return;
+
+	this_cpu_write(tracing_irq_cpu, 1);
+	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
+}
+EXPORT_SYMBOL(trace_hardirqs_off_caller);
+#endif /* CONFIG_TRACE_IRQFLAGS */
+
+#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
+
+void trace_preempt_on(unsigned long a0, unsigned long a1)
+{
+	trace_preempt_enable_rcuidle(a0, a1);
+}
+
+void trace_preempt_off(unsigned long a0, unsigned long a1)
+{
+	trace_preempt_disable_rcuidle(a0, a1);
+}
+#endif
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
                   ` (4 preceding siblings ...)
  2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-07-11  0:47   ` Steven Rostedt
  2018-06-28 18:21 ` [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers Joel Fernandes
  2018-07-03 14:15 ` [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
  7 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Andy Shevchenko, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Steven Rostedt, Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In this patch we introduce a test module for simulating a long atomic
section in the kernel which the preemptoff or irqsoff tracers can
detect. This module is to be used only for test purposes and is default
disabled.

Following is the expected output (only briefly shown) that can be parsed
to verify that the tracers are working correctly. We will use this from
the kselftests in future patches.

For the preemptoff tracer:

echo preemptoff > /d/tracing/current_tracer
sleep 1
insmod ./test_atomic_sections.ko atomic_mode=preempt atomic_time=500000
sleep 1
bash-4.3# cat /d/tracing/trace
preempt -1066    2...2    0us@: atomic_sect_run <-atomic_sect_run
preempt -1066    2...2 500002us : atomic_sect_run <-atomic_sect_run
preempt -1066    2...2 500004us : tracer_preempt_on <-atomic_sect_run
preempt -1066    2...2 500012us : <stack trace>
 => kthread
 => ret_from_fork

For the irqsoff tracer:

echo irqsoff > /d/tracing/current_tracer
sleep 1
insmod ./test_atomic_sections.ko atomic_mode=irq atomic_time=500000
sleep 1
bash-4.3# cat /d/tracing/trace
irq dis -1069    1d..1    0us@: atomic_sect_run
irq dis -1069    1d..1 500001us : atomic_sect_run
irq dis -1069    1d..1 500002us : tracer_hardirqs_on <-atomic_sect_run
irq dis -1069    1d..1 500005us : <stack trace>
 => ret_from_fork

Co-developed-by: Erick Reyes <erickreyes@google.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 lib/Kconfig.debug          |  8 ++++
 lib/Makefile               |  1 +
 lib/test_atomic_sections.c | 77 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 86 insertions(+)
 create mode 100644 lib/test_atomic_sections.c

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8838d1158d19..622c90e1e066 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1956,6 +1956,14 @@ config TEST_KMOD
 
 	  If unsure, say N.
 
+config TEST_ATOMIC_SECTIONS
+	tristate "Simulate atomic sections for tracers to detect"
+	depends on m
+	help
+	  Select this option to build a test module that can help test atomic
+	  sections by simulating them with a duration supplied as a module
+	  parameter. Preempt disable and irq disable modes can be requested.
+
 config TEST_DEBUG_VIRTUAL
 	tristate "Test CONFIG_DEBUG_VIRTUAL feature"
 	depends on DEBUG_VIRTUAL
diff --git a/lib/Makefile b/lib/Makefile
index 90dc5520b784..7831e747bf72 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -44,6 +44,7 @@ obj-y += string_helpers.o
 obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.o
 obj-y += hexdump.o
 obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
+obj-$(CONFIG_TEST_ATOMIC_SECTIONS) += test_atomic_sections.o
 obj-y += kstrtox.o
 obj-$(CONFIG_FIND_BIT_BENCHMARK) += find_bit_benchmark.o
 obj-$(CONFIG_TEST_BPF) += test_bpf.o
diff --git a/lib/test_atomic_sections.c b/lib/test_atomic_sections.c
new file mode 100644
index 000000000000..1eef518f0974
--- /dev/null
+++ b/lib/test_atomic_sections.c
@@ -0,0 +1,77 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Atomic section emulation test module
+ *
+ * Emulates atomic sections by disabling IRQs or preemption
+ * and doing a busy wait for a specified amount of time.
+ * This can be used for testing of different atomic section
+ * tracers such as irqsoff tracers.
+ *
+ * (c) 2018. Google LLC
+ */
+
+#include <linux/delay.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/kthread.h>
+#include <linux/ktime.h>
+#include <linux/module.h>
+#include <linux/printk.h>
+#include <linux/string.h>
+
+static ulong atomic_time = 100;
+static char atomic_mode[10] = "irq";
+
+module_param_named(atomic_time, atomic_time, ulong, S_IRUGO);
+module_param_string(atomic_mode, atomic_mode, 10, S_IRUGO);
+MODULE_PARM_DESC(atomic_time, "Period in microseconds (100 uS default)");
+MODULE_PARM_DESC(atomic_mode, "Mode of the test such as preempt or irq (default irq)");
+
+static void busy_wait(ulong time)
+{
+	ktime_t start, end;
+	start = ktime_get();
+	do {
+		end = ktime_get();
+		if (kthread_should_stop())
+			break;
+	} while (ktime_to_ns(ktime_sub(end, start)) < (time * 1000));
+}
+
+int atomic_sect_run(void *data)
+{
+	unsigned long flags;
+
+	if (!strcmp(atomic_mode, "irq")) {
+		local_irq_save(flags);
+		busy_wait(atomic_time);
+		local_irq_restore(flags);
+	} else if (!strcmp(atomic_mode, "preempt")) {
+		preempt_disable();
+		busy_wait(atomic_time);
+		preempt_enable();
+	}
+
+	return 0;
+}
+
+static int __init atomic_sect_init(void)
+{
+	char task_name[50];
+	struct task_struct *test_task;
+
+	snprintf(task_name, sizeof(task_name), "%s_test", atomic_mode);
+
+	test_task = kthread_run(atomic_sect_run, NULL, task_name);
+	return PTR_ERR_OR_ZERO(test_task);
+}
+
+static void __exit atomic_sect_exit(void)
+{
+	return;
+}
+
+module_init(atomic_sect_init)
+module_exit(atomic_sect_exit)
+MODULE_LICENSE("GPL v2");
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
                   ` (5 preceding siblings ...)
  2018-06-28 18:21 ` [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers Joel Fernandes
@ 2018-06-28 18:21 ` Joel Fernandes
  2018-07-11  0:49   ` Steven Rostedt
  2018-07-03 14:15 ` [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
  7 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-06-28 18:21 UTC (permalink / raw)
  To: linux-kernel
  Cc: Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

Here we add unit tests for the preemptoff and irqsoff tracer by using a
kernel module introduced previously to trigger atomic sections in the
kernel.

Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 tools/testing/selftests/ftrace/config         |  3 +
 .../test.d/preemptirq/irqsoff_tracer.tc       | 73 +++++++++++++++++++
 2 files changed, 76 insertions(+)
 create mode 100644 tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc

diff --git a/tools/testing/selftests/ftrace/config b/tools/testing/selftests/ftrace/config
index b01924c71c09..29588b328345 100644
--- a/tools/testing/selftests/ftrace/config
+++ b/tools/testing/selftests/ftrace/config
@@ -4,3 +4,6 @@ CONFIG_FUNCTION_PROFILER=y
 CONFIG_TRACER_SNAPSHOT=y
 CONFIG_STACK_TRACER=y
 CONFIG_HIST_TRIGGERS=y
+CONFIG_PREEMPT_TRACER=y
+CONFIG_IRQSOFF_TRACER=y
+CONFIG_TEST_ATOMIC_SECTIONS=m
diff --git a/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc b/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
new file mode 100644
index 000000000000..1806d340035d
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
@@ -0,0 +1,73 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+# description: test for the preemptirqsoff tracer
+
+MOD=test_atomic_sections
+
+fail() {
+    reset_tracer
+    rmmod $MOD || true
+    exit_fail
+}
+
+unsup() { #msg
+    reset_tracer
+    rmmod $MOD || true
+    echo $1
+    exit_unsupported
+}
+
+modprobe $MOD || unsup "$MOD module not available"
+rmmod $MOD
+
+grep -q "preemptoff" available_tracers || unsup "preemptoff tracer not enabled"
+grep -q "irqsoff" available_tracers || unsup "irqsoff tracer not enabled"
+
+reset_tracer
+
+# Simulate preemptoff section for half a second couple of times
+echo preemptoff > current_tracer
+sleep 1
+modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
+rmmod $MOD || fail
+modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
+rmmod $MOD || fail
+modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
+rmmod $MOD || fail
+
+cat trace
+
+# Confirm which tracer
+grep -q "tracer: preemptoff" trace || fail
+
+# Check the end of the section
+egrep -q "5.....us : <stack trace>" trace || fail
+
+# Check for 500ms of latency
+egrep -q "latency: 5..... us" trace || fail
+
+reset_tracer
+
+# Simulate irqsoff section for half a second couple of times
+echo irqsoff > current_tracer
+sleep 1
+modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
+rmmod $MOD || fail
+modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
+rmmod $MOD || fail
+modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
+rmmod $MOD || fail
+
+cat trace
+
+# Confirm which tracer
+grep -q "tracer: irqsoff" trace || fail
+
+# Check the end of the section
+egrep -q "5.....us : <stack trace>" trace || fail
+
+# Check for 500ms of latency
+egrep -q "latency: 5..... us" trace || fail
+
+reset_tracer
+exit 0
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints
  2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
                   ` (6 preceding siblings ...)
  2018-06-28 18:21 ` [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers Joel Fernandes
@ 2018-07-03 14:15 ` Joel Fernandes
  2018-07-03 14:23   ` Steven Rostedt
  7 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-03 14:15 UTC (permalink / raw)
  To: linux-kernel
  Cc: Boqun Feng, Byungchul Park, Ingo Molnar, Julia Cartwright,
	linux-kselftest, Masami Hiramatsu, Mathieu Desnoyers,
	Namhyung Kim, Paul McKenney, Peter Zijlstra, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Thu, Jun 28, 2018 at 11:21:42AM -0700, Joel Fernandes wrote:
> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> 
> This is a posting of v9 preempt/irq tracepoint clean up series rebased
> onto v4.18-rc2. No changes in the series, just a rebase + repost.
> 
> All patches have a Reviewed-by tags now from reviewers. This series has
> been well tested and is a simplification/refactoring of existing code,
> along with giving a speed-up for tracepoints using the rcu-idle API.
> With this our users will find it easier to use tools depending on
> existing preempt tracepoints since it simplifies the configuration for
> them.

Steve, all patches of this tracing series have been reviewed and/or acked and
there haven't been additional changes for a couple of reposts. Are you Ok
with it?

Thanks,

-Joel


> Future enhancements/fixes I am developing for preempt-off tracer will
> depend on these patches, so I suggest prioritizing these well reviewed
> and tested patches for that reason as well.
> 
> Introduction to the series:
> The preempt/irq tracepoints exist but not everything in the kernel is using it
> whenever they need to be notified that a preempt disable/enable or an irq
> disable/enable has occurred.  This makes things not work simultaneously (for
> example, only either lockdep or irqsoff trace-events can be used at a time).
> 
> This is particularly painful to deal with, since turning on lockdep breaks
> tracers that install probes on IRQ events, such as the BCC atomic critical
> section tracer [1]. This constraint also makes it impossible to use synthetic
> events to trace irqsoff sections with lockdep simulataneously turned on.
> 
> This series solves that, and also results in a nice clean up of relevant parts
> of the kernel. Several ifdefs are simpler, and the design is more unified and
> better. Also as a result of this, we also speeded performance all rcuidle
> tracepoints since their handling is simpler.
> 
> [1] https://github.com/iovisor/bcc/blob/master/tools/criticalstat_example.txt
> 
> v8->v9:
> - Small style changes to tracepoint code (Mathieu)
> - Minor style fix to use PTR_ERR_OR_ZERO (0-day bot)
> - Minor fix to test_atomic_sections to use unsigned long.
> - Added Namhyung's, Mathieu's Reviewed-by to some patches.
> - Added Acks from Matsami
> 
> v7->v8:
> - Refactored irqsoff tracer probe defines (Namhyung)
> 
> v6->v7:
> - Added a module to simulate an atomic section, a kselftest to load and
>   and trigger it which verifies the preempt-tracer and this series.
> 
> - Fixed a new warning after I rebased in early boot, this is because
> early_boot_irqs_disabled was set too early, I moved it after the lockdep
> initialization.
> 
> - added back the softirq fix since it appears it wasn't picked up.
> 
> - Ran Ingo's locking API selftest suite which are passing with this
>   series.
> 
> - Mathieu suggested ifdef'ing the tracepoint_synchronize_unregister
>   function incase tracepoints aren't enabled, did that.
> 
> 
> Joel Fernandes (Google) (6):
>   srcu: Add notrace variant of srcu_dereference
>   trace/irqsoff: Split reset into separate functions
>   tracepoint: Make rcuidle tracepoint callers use SRCU
>   tracing: Centralize preemptirq tracepoints and unify their usage
>   lib: Add module to simulate atomic sections for testing preemptoff
>     tracers
>   kselftests: Add tests for the preemptoff and irqsoff tracers
> 
> Paul McKenney (1):
>   srcu: Add notrace variants of srcu_read_{lock,unlock}
> 
>  include/linux/ftrace.h                        |  11 +-
>  include/linux/irqflags.h                      |  11 +-
>  include/linux/lockdep.h                       |   8 +-
>  include/linux/preempt.h                       |   2 +-
>  include/linux/srcu.h                          |  22 ++
>  include/linux/tracepoint.h                    |  49 +++-
>  include/trace/events/preemptirq.h             |  23 +-
>  init/main.c                                   |   5 +-
>  kernel/locking/lockdep.c                      |  35 +--
>  kernel/sched/core.c                           |   2 +-
>  kernel/trace/Kconfig                          |  22 +-
>  kernel/trace/Makefile                         |   2 +-
>  kernel/trace/trace_irqsoff.c                  | 253 ++++++------------
>  kernel/trace/trace_preemptirq.c               |  71 +++++
>  kernel/tracepoint.c                           |  16 +-
>  lib/Kconfig.debug                             |   8 +
>  lib/Makefile                                  |   1 +
>  lib/test_atomic_sections.c                    |  77 ++++++
>  tools/testing/selftests/ftrace/config         |   3 +
>  .../test.d/preemptirq/irqsoff_tracer.tc       |  73 +++++
>  20 files changed, 453 insertions(+), 241 deletions(-)
>  create mode 100644 kernel/trace/trace_preemptirq.c
>  create mode 100644 lib/test_atomic_sections.c
>  create mode 100644 tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
> 
> -- 
> 2.18.0.rc2.346.g013aa6912e-goog
> 
> 

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints
  2018-07-03 14:15 ` [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
@ 2018-07-03 14:23   ` Steven Rostedt
  0 siblings, 0 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-03 14:23 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Tue, 3 Jul 2018 07:15:37 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> On Thu, Jun 28, 2018 at 11:21:42AM -0700, Joel Fernandes wrote:
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > 
> > This is a posting of v9 preempt/irq tracepoint clean up series rebased
> > onto v4.18-rc2. No changes in the series, just a rebase + repost.
> > 
> > All patches have a Reviewed-by tags now from reviewers. This series has
> > been well tested and is a simplification/refactoring of existing code,
> > along with giving a speed-up for tracepoints using the rcu-idle API.
> > With this our users will find it easier to use tools depending on
> > existing preempt tracepoints since it simplifies the configuration for
> > them.  
> 
> Steve, all patches of this tracing series have been reviewed and/or acked and
> there haven't been additional changes for a couple of reposts. Are you Ok
> with it?
> 

I'm currently chasing down some fires, but I'll try to get to it this
week.

Thanks for the patience ;-)

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
@ 2018-07-06 22:06   ` Steven Rostedt
  2018-07-07  4:20     ` Joel Fernandes
  2018-07-10 14:20   ` Steven Rostedt
  2018-07-11 13:12   ` Peter Zijlstra
  2 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-06 22:06 UTC (permalink / raw)
  To: Joel Fernandes, Peter Zijlstra
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi


Peter,

Want to ack this? It touches Lockdep.

Joel,

I got to this patch and I'm still reviewing it. I'll hopefully have my
full review done by next week. I'll make it a priority. But I still
would like Peter's ack on this one, as he's the maintainer of lockdep.

Thanks,

-- Steve


On Thu, 28 Jun 2018 11:21:47 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> 
> This patch detaches the preemptirq tracepoints from the tracers and
> keeps it separate.
> 
> Advantages:
> * Lockdep and irqsoff event can now run in parallel since they no longer
> have their own calls.
> 
> * This unifies the usecase of adding hooks to an irqsoff and irqson
> event, and a preemptoff and preempton event.
>   3 users of the events exist:
>   - Lockdep
>   - irqsoff and preemptoff tracers
>   - irqs and preempt trace events
> 
> The unification cleans up several ifdefs and makes the code in preempt
> tracer and irqsoff tracers simpler. It gets rid of all the horrific
> ifdeferry around PROVE_LOCKING and makes configuration of the different
> users of the tracepoints more easy and understandable. It also gets rid
> of the time_* function calls from the lockdep hooks used to call into
> the preemptirq tracer which is not needed anymore. The negative delta in
> lines of code in this patch is quite large too.
> 
> In the patch we introduce a new CONFIG option PREEMPTIRQ_TRACEPOINTS
> as a single point for registering probes onto the tracepoints. With
> this,
> the web of config options for preempt/irq toggle tracepoints and its
> users becomes:
> 
>  PREEMPT_TRACER   PREEMPTIRQ_EVENTS  IRQSOFF_TRACER PROVE_LOCKING
>        |                 |     \         |           |
>        \    (selects)    /      \        \ (selects) /
>       TRACE_PREEMPT_TOGGLE       ----> TRACE_IRQFLAGS
>                       \                  /
>                        \ (depends on)   /
>                      PREEMPTIRQ_TRACEPOINTS
> 
> One note, I have to check for lockdep recursion in the code that calls
> the trace events API and bail out if we're in lockdep recursion
> protection to prevent something like the following case: a spin_lock is
> taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> this function, a call to get_lock_stats happens which calls
> preempt_disable, which calls trace IRQS off somewhere which enters my
> tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> This flag is then never cleared causing lockdep paths to never be
> entered and thus causing splats and other bad things.
> 
> Other than the performance tests mentioned in the previous patch, I also
> ran the locking API test suite. I verified that all tests cases are
> passing.
> 
> I also injected issues by not registering lockdep probes onto the
> tracepoints and I see failures to confirm that the probes are indeed
> working.
> 
> This series + lockdep probes not registered (just to inject errors):
> [    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> [    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> [    0.000000]        sirq-safe-A => hirqs-on/12:FAILED|FAILED|  ok  |
> [    0.000000]        sirq-safe-A => hirqs-on/21:FAILED|FAILED|  ok  |
> [    0.000000]          hard-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
> [    0.000000]          soft-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
> [    0.000000]          hard-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
> [    0.000000]          soft-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
> [    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> [    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> 
> With this series + lockdep probes registered, all locking tests pass:
> 
> [    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> [    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> [    0.000000]        sirq-safe-A => hirqs-on/12:  ok  |  ok  |  ok  |
> [    0.000000]        sirq-safe-A => hirqs-on/21:  ok  |  ok  |  ok  |
> [    0.000000]          hard-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
> [    0.000000]          soft-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
> [    0.000000]          hard-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
> [    0.000000]          soft-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
> [    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> [    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> 
> Reviewed-by: Namhyung Kim <namhyung@kernel.org>
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> ---
>  include/linux/ftrace.h            |  11 +-
>  include/linux/irqflags.h          |  11 +-
>  include/linux/lockdep.h           |   8 +-
>  include/linux/preempt.h           |   2 +-
>  include/trace/events/preemptirq.h |  23 +--
>  init/main.c                       |   5 +-
>  kernel/locking/lockdep.c          |  35 ++---
>  kernel/sched/core.c               |   2 +-
>  kernel/trace/Kconfig              |  22 ++-
>  kernel/trace/Makefile             |   2 +-
>  kernel/trace/trace_irqsoff.c      | 231 ++++++++----------------------
>  kernel/trace/trace_preemptirq.c   |  71 +++++++++
>  12 files changed, 194 insertions(+), 229 deletions(-)
>  create mode 100644 kernel/trace/trace_preemptirq.c
> 
> diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> index 8154f4920fcb..f32e3c81407e 100644
> --- a/include/linux/ftrace.h
> +++ b/include/linux/ftrace.h
> @@ -709,16 +709,7 @@ static inline unsigned long get_lock_parent_ip(void)
>  	return CALLER_ADDR2;
>  }
>  
> -#ifdef CONFIG_IRQSOFF_TRACER
> -  extern void time_hardirqs_on(unsigned long a0, unsigned long a1);
> -  extern void time_hardirqs_off(unsigned long a0, unsigned long a1);
> -#else
> -  static inline void time_hardirqs_on(unsigned long a0, unsigned long a1) { }
> -  static inline void time_hardirqs_off(unsigned long a0, unsigned long a1) { }
> -#endif
> -
> -#if defined(CONFIG_PREEMPT_TRACER) || \
> -	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
> +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
>    extern void trace_preempt_on(unsigned long a0, unsigned long a1);
>    extern void trace_preempt_off(unsigned long a0, unsigned long a1);
>  #else
> diff --git a/include/linux/irqflags.h b/include/linux/irqflags.h
> index 9700f00bbc04..50edb9cbbd26 100644
> --- a/include/linux/irqflags.h
> +++ b/include/linux/irqflags.h
> @@ -15,9 +15,16 @@
>  #include <linux/typecheck.h>
>  #include <asm/irqflags.h>
>  
> -#ifdef CONFIG_TRACE_IRQFLAGS
> +/* Currently trace_softirqs_on/off is used only by lockdep */
> +#ifdef CONFIG_PROVE_LOCKING
>    extern void trace_softirqs_on(unsigned long ip);
>    extern void trace_softirqs_off(unsigned long ip);
> +#else
> +# define trace_softirqs_on(ip)	do { } while (0)
> +# define trace_softirqs_off(ip)	do { } while (0)
> +#endif
> +
> +#ifdef CONFIG_TRACE_IRQFLAGS
>    extern void trace_hardirqs_on(void);
>    extern void trace_hardirqs_off(void);
>  # define trace_hardirq_context(p)	((p)->hardirq_context)
> @@ -43,8 +50,6 @@ do {						\
>  #else
>  # define trace_hardirqs_on()		do { } while (0)
>  # define trace_hardirqs_off()		do { } while (0)
> -# define trace_softirqs_on(ip)		do { } while (0)
> -# define trace_softirqs_off(ip)		do { } while (0)
>  # define trace_hardirq_context(p)	0
>  # define trace_softirq_context(p)	0
>  # define trace_hardirqs_enabled(p)	0
> diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
> index 6fc77d4dbdcd..a8113357ceeb 100644
> --- a/include/linux/lockdep.h
> +++ b/include/linux/lockdep.h
> @@ -266,7 +266,8 @@ struct held_lock {
>  /*
>   * Initialization, self-test and debugging-output methods:
>   */
> -extern void lockdep_info(void);
> +extern void lockdep_init(void);
> +extern void lockdep_init_early(void);
>  extern void lockdep_reset(void);
>  extern void lockdep_reset_lock(struct lockdep_map *lock);
>  extern void lockdep_free_key_range(void *start, unsigned long size);
> @@ -406,7 +407,8 @@ static inline void lockdep_on(void)
>  # define lock_downgrade(l, i)			do { } while (0)
>  # define lock_set_class(l, n, k, s, i)		do { } while (0)
>  # define lock_set_subclass(l, s, i)		do { } while (0)
> -# define lockdep_info()				do { } while (0)
> +# define lockdep_init()				do { } while (0)
> +# define lockdep_init_early()			do { } while (0)
>  # define lockdep_init_map(lock, name, key, sub) \
>  		do { (void)(name); (void)(key); } while (0)
>  # define lockdep_set_class(lock, key)		do { (void)(key); } while (0)
> @@ -532,7 +534,7 @@ do {								\
>  
>  #endif /* CONFIG_LOCKDEP */
>  
> -#ifdef CONFIG_TRACE_IRQFLAGS
> +#ifdef CONFIG_PROVE_LOCKING
>  extern void print_irqtrace_events(struct task_struct *curr);
>  #else
>  static inline void print_irqtrace_events(struct task_struct *curr)
> diff --git a/include/linux/preempt.h b/include/linux/preempt.h
> index 5bd3f151da78..c01813c3fbe9 100644
> --- a/include/linux/preempt.h
> +++ b/include/linux/preempt.h
> @@ -150,7 +150,7 @@
>   */
>  #define in_atomic_preempt_off() (preempt_count() != PREEMPT_DISABLE_OFFSET)
>  
> -#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_PREEMPT_TRACER)
> +#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_TRACE_PREEMPT_TOGGLE)
>  extern void preempt_count_add(int val);
>  extern void preempt_count_sub(int val);
>  #define preempt_count_dec_and_test() \
> diff --git a/include/trace/events/preemptirq.h b/include/trace/events/preemptirq.h
> index 9c4eb33c5a1d..9a0d4ceeb166 100644
> --- a/include/trace/events/preemptirq.h
> +++ b/include/trace/events/preemptirq.h
> @@ -1,4 +1,4 @@
> -#ifdef CONFIG_PREEMPTIRQ_EVENTS
> +#ifdef CONFIG_PREEMPTIRQ_TRACEPOINTS
>  
>  #undef TRACE_SYSTEM
>  #define TRACE_SYSTEM preemptirq
> @@ -32,7 +32,7 @@ DECLARE_EVENT_CLASS(preemptirq_template,
>  		  (void *)((unsigned long)(_stext) + __entry->parent_offs))
>  );
>  
> -#ifndef CONFIG_PROVE_LOCKING
> +#ifdef CONFIG_TRACE_IRQFLAGS
>  DEFINE_EVENT(preemptirq_template, irq_disable,
>  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
>  	     TP_ARGS(ip, parent_ip));
> @@ -40,9 +40,14 @@ DEFINE_EVENT(preemptirq_template, irq_disable,
>  DEFINE_EVENT(preemptirq_template, irq_enable,
>  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
>  	     TP_ARGS(ip, parent_ip));
> +#else
> +#define trace_irq_enable(...)
> +#define trace_irq_disable(...)
> +#define trace_irq_enable_rcuidle(...)
> +#define trace_irq_disable_rcuidle(...)
>  #endif
>  
> -#ifdef CONFIG_DEBUG_PREEMPT
> +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
>  DEFINE_EVENT(preemptirq_template, preempt_disable,
>  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
>  	     TP_ARGS(ip, parent_ip));
> @@ -50,22 +55,22 @@ DEFINE_EVENT(preemptirq_template, preempt_disable,
>  DEFINE_EVENT(preemptirq_template, preempt_enable,
>  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
>  	     TP_ARGS(ip, parent_ip));
> +#else
> +#define trace_preempt_enable(...)
> +#define trace_preempt_disable(...)
> +#define trace_preempt_enable_rcuidle(...)
> +#define trace_preempt_disable_rcuidle(...)
>  #endif
>  
>  #endif /* _TRACE_PREEMPTIRQ_H */
>  
>  #include <trace/define_trace.h>
>  
> -#endif /* !CONFIG_PREEMPTIRQ_EVENTS */
> -
> -#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || defined(CONFIG_PROVE_LOCKING)
> +#else /* !CONFIG_PREEMPTIRQ_TRACEPOINTS */
>  #define trace_irq_enable(...)
>  #define trace_irq_disable(...)
>  #define trace_irq_enable_rcuidle(...)
>  #define trace_irq_disable_rcuidle(...)
> -#endif
> -
> -#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || !defined(CONFIG_DEBUG_PREEMPT)
>  #define trace_preempt_enable(...)
>  #define trace_preempt_disable(...)
>  #define trace_preempt_enable_rcuidle(...)
> diff --git a/init/main.c b/init/main.c
> index 3b4ada11ed52..44fe43be84c1 100644
> --- a/init/main.c
> +++ b/init/main.c
> @@ -648,6 +648,9 @@ asmlinkage __visible void __init start_kernel(void)
>  	profile_init();
>  	call_function_init();
>  	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
> +
> +	lockdep_init_early();
> +
>  	early_boot_irqs_disabled = false;
>  	local_irq_enable();
>  
> @@ -663,7 +666,7 @@ asmlinkage __visible void __init start_kernel(void)
>  		panic("Too many boot %s vars at `%s'", panic_later,
>  		      panic_param);
>  
> -	lockdep_info();
> +	lockdep_init();
>  
>  	/*
>  	 * Need to run this when irqs are enabled, because it wants
> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> index 5fa4d3138bf1..b961a1698e98 100644
> --- a/kernel/locking/lockdep.c
> +++ b/kernel/locking/lockdep.c
> @@ -55,6 +55,7 @@
>  
>  #include "lockdep_internals.h"
>  
> +#include <trace/events/preemptirq.h>
>  #define CREATE_TRACE_POINTS
>  #include <trace/events/lock.h>
>  
> @@ -2845,10 +2846,9 @@ static void __trace_hardirqs_on_caller(unsigned long ip)
>  	debug_atomic_inc(hardirqs_on_events);
>  }
>  
> -__visible void trace_hardirqs_on_caller(unsigned long ip)
> +static void lockdep_hardirqs_on(void *none, unsigned long ignore,
> +				unsigned long ip)
>  {
> -	time_hardirqs_on(CALLER_ADDR0, ip);
> -
>  	if (unlikely(!debug_locks || current->lockdep_recursion))
>  		return;
>  
> @@ -2887,23 +2887,15 @@ __visible void trace_hardirqs_on_caller(unsigned long ip)
>  	__trace_hardirqs_on_caller(ip);
>  	current->lockdep_recursion = 0;
>  }
> -EXPORT_SYMBOL(trace_hardirqs_on_caller);
> -
> -void trace_hardirqs_on(void)
> -{
> -	trace_hardirqs_on_caller(CALLER_ADDR0);
> -}
> -EXPORT_SYMBOL(trace_hardirqs_on);
>  
>  /*
>   * Hardirqs were disabled:
>   */
> -__visible void trace_hardirqs_off_caller(unsigned long ip)
> +static void lockdep_hardirqs_off(void *none, unsigned long ignore,
> +				 unsigned long ip)
>  {
>  	struct task_struct *curr = current;
>  
> -	time_hardirqs_off(CALLER_ADDR0, ip);
> -
>  	if (unlikely(!debug_locks || current->lockdep_recursion))
>  		return;
>  
> @@ -2925,13 +2917,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
>  	} else
>  		debug_atomic_inc(redundant_hardirqs_off);
>  }
> -EXPORT_SYMBOL(trace_hardirqs_off_caller);
> -
> -void trace_hardirqs_off(void)
> -{
> -	trace_hardirqs_off_caller(CALLER_ADDR0);
> -}
> -EXPORT_SYMBOL(trace_hardirqs_off);
>  
>  /*
>   * Softirqs will be enabled:
> @@ -4338,7 +4323,15 @@ void lockdep_reset_lock(struct lockdep_map *lock)
>  	raw_local_irq_restore(flags);
>  }
>  
> -void __init lockdep_info(void)
> +void __init lockdep_init_early(void)
> +{
> +#ifdef CONFIG_PROVE_LOCKING
> +	register_trace_prio_irq_disable(lockdep_hardirqs_off, NULL, INT_MAX);
> +	register_trace_prio_irq_enable(lockdep_hardirqs_on, NULL, INT_MIN);
> +#endif
> +}
> +
> +void __init lockdep_init(void)
>  {
>  	printk("Lock dependency validator: Copyright (c) 2006 Red Hat, Inc., Ingo Molnar\n");
>  
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 78d8facba456..4c956f6849ec 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -3192,7 +3192,7 @@ static inline void sched_tick_stop(int cpu) { }
>  #endif
>  
>  #if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
> -				defined(CONFIG_PREEMPT_TRACER))
> +				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
>  /*
>   * If the value passed in is equal to the current preempt count
>   * then we just disabled preemption. Start timing the latency.
> diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
> index dcc0166d1997..8d51351e3149 100644
> --- a/kernel/trace/Kconfig
> +++ b/kernel/trace/Kconfig
> @@ -82,6 +82,15 @@ config RING_BUFFER_ALLOW_SWAP
>  	 Allow the use of ring_buffer_swap_cpu.
>  	 Adds a very slight overhead to tracing when enabled.
>  
> +config PREEMPTIRQ_TRACEPOINTS
> +	bool
> +	depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
> +	select TRACING
> +	default y
> +	help
> +	  Create preempt/irq toggle tracepoints if needed, so that other parts
> +	  of the kernel can use them to generate or add hooks to them.
> +
>  # All tracer options should select GENERIC_TRACER. For those options that are
>  # enabled by all tracers (context switch and event tracer) they select TRACING.
>  # This allows those options to appear when no other tracer is selected. But the
> @@ -155,18 +164,20 @@ config FUNCTION_GRAPH_TRACER
>  	  the return value. This is done by setting the current return
>  	  address on the current task structure into a stack of calls.
>  
> +config TRACE_PREEMPT_TOGGLE
> +	bool
> +	help
> +	  Enables hooks which will be called when preemption is first disabled,
> +	  and last enabled.
>  
>  config PREEMPTIRQ_EVENTS
>  	bool "Enable trace events for preempt and irq disable/enable"
>  	select TRACE_IRQFLAGS
> -	depends on DEBUG_PREEMPT || !PROVE_LOCKING
> -	depends on TRACING
> +	select TRACE_PREEMPT_TOGGLE if PREEMPT
> +	select GENERIC_TRACER
>  	default n
>  	help
>  	  Enable tracing of disable and enable events for preemption and irqs.
> -	  For tracing preempt disable/enable events, DEBUG_PREEMPT must be
> -	  enabled. For tracing irq disable/enable events, PROVE_LOCKING must
> -	  be disabled.
>  
>  config IRQSOFF_TRACER
>  	bool "Interrupts-off Latency Tracer"
> @@ -203,6 +214,7 @@ config PREEMPT_TRACER
>  	select RING_BUFFER_ALLOW_SWAP
>  	select TRACER_SNAPSHOT
>  	select TRACER_SNAPSHOT_PER_CPU_SWAP
> +	select TRACE_PREEMPT_TOGGLE
>  	help
>  	  This option measures the time spent in preemption-off critical
>  	  sections, with microsecond accuracy.
> diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile
> index e2538c7638d4..84a0cb222f20 100644
> --- a/kernel/trace/Makefile
> +++ b/kernel/trace/Makefile
> @@ -35,7 +35,7 @@ obj-$(CONFIG_TRACING) += trace_printk.o
>  obj-$(CONFIG_TRACING_MAP) += tracing_map.o
>  obj-$(CONFIG_CONTEXT_SWITCH_TRACER) += trace_sched_switch.o
>  obj-$(CONFIG_FUNCTION_TRACER) += trace_functions.o
> -obj-$(CONFIG_PREEMPTIRQ_EVENTS) += trace_irqsoff.o
> +obj-$(CONFIG_PREEMPTIRQ_TRACEPOINTS) += trace_preemptirq.o
>  obj-$(CONFIG_IRQSOFF_TRACER) += trace_irqsoff.o
>  obj-$(CONFIG_PREEMPT_TRACER) += trace_irqsoff.o
>  obj-$(CONFIG_SCHED_TRACER) += trace_sched_wakeup.o
> diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
> index f8daa754cce2..770cd30cda40 100644
> --- a/kernel/trace/trace_irqsoff.c
> +++ b/kernel/trace/trace_irqsoff.c
> @@ -16,7 +16,6 @@
>  
>  #include "trace.h"
>  
> -#define CREATE_TRACE_POINTS
>  #include <trace/events/preemptirq.h>
>  
>  #if defined(CONFIG_IRQSOFF_TRACER) || defined(CONFIG_PREEMPT_TRACER)
> @@ -450,66 +449,6 @@ void stop_critical_timings(void)
>  }
>  EXPORT_SYMBOL_GPL(stop_critical_timings);
>  
> -#ifdef CONFIG_IRQSOFF_TRACER
> -#ifdef CONFIG_PROVE_LOCKING
> -void time_hardirqs_on(unsigned long a0, unsigned long a1)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		stop_critical_timing(a0, a1);
> -}
> -
> -void time_hardirqs_off(unsigned long a0, unsigned long a1)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		start_critical_timing(a0, a1);
> -}
> -
> -#else /* !CONFIG_PROVE_LOCKING */
> -
> -/*
> - * We are only interested in hardirq on/off events:
> - */
> -static inline void tracer_hardirqs_on(void)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
> -}
> -
> -static inline void tracer_hardirqs_off(void)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
> -}
> -
> -static inline void tracer_hardirqs_on_caller(unsigned long caller_addr)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		stop_critical_timing(CALLER_ADDR0, caller_addr);
> -}
> -
> -static inline void tracer_hardirqs_off_caller(unsigned long caller_addr)
> -{
> -	if (!preempt_trace() && irq_trace())
> -		start_critical_timing(CALLER_ADDR0, caller_addr);
> -}
> -
> -#endif /* CONFIG_PROVE_LOCKING */
> -#endif /*  CONFIG_IRQSOFF_TRACER */
> -
> -#ifdef CONFIG_PREEMPT_TRACER
> -static inline void tracer_preempt_on(unsigned long a0, unsigned long a1)
> -{
> -	if (preempt_trace() && !irq_trace())
> -		stop_critical_timing(a0, a1);
> -}
> -
> -static inline void tracer_preempt_off(unsigned long a0, unsigned long a1)
> -{
> -	if (preempt_trace() && !irq_trace())
> -		start_critical_timing(a0, a1);
> -}
> -#endif /* CONFIG_PREEMPT_TRACER */
> -
>  #ifdef CONFIG_FUNCTION_TRACER
>  static bool function_enabled;
>  
> @@ -659,15 +598,34 @@ static void irqsoff_tracer_stop(struct trace_array *tr)
>  }
>  
>  #ifdef CONFIG_IRQSOFF_TRACER
> +/*
> + * We are only interested in hardirq on/off events:
> + */
> +static void tracer_hardirqs_on(void *none, unsigned long a0, unsigned long a1)
> +{
> +	if (!preempt_trace() && irq_trace())
> +		stop_critical_timing(a0, a1);
> +}
> +
> +static void tracer_hardirqs_off(void *none, unsigned long a0, unsigned long a1)
> +{
> +	if (!preempt_trace() && irq_trace())
> +		start_critical_timing(a0, a1);
> +}
> +
>  static int irqsoff_tracer_init(struct trace_array *tr)
>  {
>  	trace_type = TRACER_IRQS_OFF;
>  
> +	register_trace_irq_disable(tracer_hardirqs_off, NULL);
> +	register_trace_irq_enable(tracer_hardirqs_on, NULL);
>  	return __irqsoff_tracer_init(tr);
>  }
>  
>  static void irqsoff_tracer_reset(struct trace_array *tr)
>  {
> +	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
> +	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
>  	__irqsoff_tracer_reset(tr);
>  }
>  
> @@ -690,21 +648,34 @@ static struct tracer irqsoff_tracer __read_mostly =
>  	.allow_instances = true,
>  	.use_max_tr	= true,
>  };
> -# define register_irqsoff(trace) register_tracer(&trace)
> -#else
> -# define register_irqsoff(trace) do { } while (0)
> -#endif
> +#endif /*  CONFIG_IRQSOFF_TRACER */
>  
>  #ifdef CONFIG_PREEMPT_TRACER
> +static void tracer_preempt_on(void *none, unsigned long a0, unsigned long a1)
> +{
> +	if (preempt_trace() && !irq_trace())
> +		stop_critical_timing(a0, a1);
> +}
> +
> +static void tracer_preempt_off(void *none, unsigned long a0, unsigned long a1)
> +{
> +	if (preempt_trace() && !irq_trace())
> +		start_critical_timing(a0, a1);
> +}
> +
>  static int preemptoff_tracer_init(struct trace_array *tr)
>  {
>  	trace_type = TRACER_PREEMPT_OFF;
>  
> +	register_trace_preempt_disable(tracer_preempt_off, NULL);
> +	register_trace_preempt_enable(tracer_preempt_on, NULL);
>  	return __irqsoff_tracer_init(tr);
>  }
>  
>  static void preemptoff_tracer_reset(struct trace_array *tr)
>  {
> +	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
> +	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
>  	__irqsoff_tracer_reset(tr);
>  }
>  
> @@ -727,23 +698,29 @@ static struct tracer preemptoff_tracer __read_mostly =
>  	.allow_instances = true,
>  	.use_max_tr	= true,
>  };
> -# define register_preemptoff(trace) register_tracer(&trace)
> -#else
> -# define register_preemptoff(trace) do { } while (0)
> -#endif
> +#endif /* CONFIG_PREEMPT_TRACER */
>  
> -#if defined(CONFIG_IRQSOFF_TRACER) && \
> -	defined(CONFIG_PREEMPT_TRACER)
> +#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
>  
>  static int preemptirqsoff_tracer_init(struct trace_array *tr)
>  {
>  	trace_type = TRACER_IRQS_OFF | TRACER_PREEMPT_OFF;
>  
> +	register_trace_irq_disable(tracer_hardirqs_off, NULL);
> +	register_trace_irq_enable(tracer_hardirqs_on, NULL);
> +	register_trace_preempt_disable(tracer_preempt_off, NULL);
> +	register_trace_preempt_enable(tracer_preempt_on, NULL);
> +
>  	return __irqsoff_tracer_init(tr);
>  }
>  
>  static void preemptirqsoff_tracer_reset(struct trace_array *tr)
>  {
> +	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
> +	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
> +	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
> +	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
> +
>  	__irqsoff_tracer_reset(tr);
>  }
>  
> @@ -766,115 +743,21 @@ static struct tracer preemptirqsoff_tracer __read_mostly =
>  	.allow_instances = true,
>  	.use_max_tr	= true,
>  };
> -
> -# define register_preemptirqsoff(trace) register_tracer(&trace)
> -#else
> -# define register_preemptirqsoff(trace) do { } while (0)
>  #endif
>  
>  __init static int init_irqsoff_tracer(void)
>  {
> -	register_irqsoff(irqsoff_tracer);
> -	register_preemptoff(preemptoff_tracer);
> -	register_preemptirqsoff(preemptirqsoff_tracer);
> -
> -	return 0;
> -}
> -core_initcall(init_irqsoff_tracer);
> -#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
> -
> -#ifndef CONFIG_IRQSOFF_TRACER
> -static inline void tracer_hardirqs_on(void) { }
> -static inline void tracer_hardirqs_off(void) { }
> -static inline void tracer_hardirqs_on_caller(unsigned long caller_addr) { }
> -static inline void tracer_hardirqs_off_caller(unsigned long caller_addr) { }
> +#ifdef CONFIG_IRQSOFF_TRACER
> +	register_tracer(&irqsoff_tracer);
>  #endif
> -
> -#ifndef CONFIG_PREEMPT_TRACER
> -static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
> -static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
> +#ifdef CONFIG_PREEMPT_TRACER
> +	register_tracer(&preemptoff_tracer);
>  #endif
> -
> -#if defined(CONFIG_TRACE_IRQFLAGS) && !defined(CONFIG_PROVE_LOCKING)
> -/* Per-cpu variable to prevent redundant calls when IRQs already off */
> -static DEFINE_PER_CPU(int, tracing_irq_cpu);
> -
> -void trace_hardirqs_on(void)
> -{
> -	if (!this_cpu_read(tracing_irq_cpu))
> -		return;
> -
> -	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> -	tracer_hardirqs_on();
> -
> -	this_cpu_write(tracing_irq_cpu, 0);
> -}
> -EXPORT_SYMBOL(trace_hardirqs_on);
> -
> -void trace_hardirqs_off(void)
> -{
> -	if (this_cpu_read(tracing_irq_cpu))
> -		return;
> -
> -	this_cpu_write(tracing_irq_cpu, 1);
> -
> -	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> -	tracer_hardirqs_off();
> -}
> -EXPORT_SYMBOL(trace_hardirqs_off);
> -
> -__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
> -{
> -	if (!this_cpu_read(tracing_irq_cpu))
> -		return;
> -
> -	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
> -	tracer_hardirqs_on_caller(caller_addr);
> -
> -	this_cpu_write(tracing_irq_cpu, 0);
> -}
> -EXPORT_SYMBOL(trace_hardirqs_on_caller);
> -
> -__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
> -{
> -	if (this_cpu_read(tracing_irq_cpu))
> -		return;
> -
> -	this_cpu_write(tracing_irq_cpu, 1);
> -
> -	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
> -	tracer_hardirqs_off_caller(caller_addr);
> -}
> -EXPORT_SYMBOL(trace_hardirqs_off_caller);
> -
> -/*
> - * Stubs:
> - */
> -
> -void trace_softirqs_on(unsigned long ip)
> -{
> -}
> -
> -void trace_softirqs_off(unsigned long ip)
> -{
> -}
> -
> -inline void print_irqtrace_events(struct task_struct *curr)
> -{
> -}
> +#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
> +	register_tracer(&preemptirqsoff_tracer);
>  #endif
>  
> -#if defined(CONFIG_PREEMPT_TRACER) || \
> -	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
> -void trace_preempt_on(unsigned long a0, unsigned long a1)
> -{
> -	trace_preempt_enable_rcuidle(a0, a1);
> -	tracer_preempt_on(a0, a1);
> -}
> -
> -void trace_preempt_off(unsigned long a0, unsigned long a1)
> -{
> -	trace_preempt_disable_rcuidle(a0, a1);
> -	tracer_preempt_off(a0, a1);
> +	return 0;
>  }
> -#endif
> +core_initcall(init_irqsoff_tracer);
> +#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
> diff --git a/kernel/trace/trace_preemptirq.c b/kernel/trace/trace_preemptirq.c
> new file mode 100644
> index 000000000000..dc01c7f4d326
> --- /dev/null
> +++ b/kernel/trace/trace_preemptirq.c
> @@ -0,0 +1,71 @@
> +/*
> + * preemptoff and irqoff tracepoints
> + *
> + * Copyright (C) Joel Fernandes (Google) <joel@joelfernandes.org>
> + */
> +
> +#include <linux/kallsyms.h>
> +#include <linux/uaccess.h>
> +#include <linux/module.h>
> +#include <linux/ftrace.h>
> +
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/preemptirq.h>
> +
> +#ifdef CONFIG_TRACE_IRQFLAGS
> +/* Per-cpu variable to prevent redundant calls when IRQs already off */
> +static DEFINE_PER_CPU(int, tracing_irq_cpu);
> +
> +void trace_hardirqs_on(void)
> +{
> +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> +	this_cpu_write(tracing_irq_cpu, 0);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_on);
> +
> +void trace_hardirqs_off(void)
> +{
> +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	this_cpu_write(tracing_irq_cpu, 1);
> +	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_off);
> +
> +__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
> +{
> +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
> +	this_cpu_write(tracing_irq_cpu, 0);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_on_caller);
> +
> +__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
> +{
> +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	this_cpu_write(tracing_irq_cpu, 1);
> +	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_off_caller);
> +#endif /* CONFIG_TRACE_IRQFLAGS */
> +
> +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
> +
> +void trace_preempt_on(unsigned long a0, unsigned long a1)
> +{
> +	trace_preempt_enable_rcuidle(a0, a1);
> +}
> +
> +void trace_preempt_off(unsigned long a0, unsigned long a1)
> +{
> +	trace_preempt_disable_rcuidle(a0, a1);
> +}
> +#endif


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-06 22:06   ` Steven Rostedt
@ 2018-07-07  4:20     ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-07  4:20 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Peter Zijlstra, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Fri, Jul 06, 2018 at 06:06:10PM -0400, Steven Rostedt wrote:
> 
> Peter,
> 
> Want to ack this? It touches Lockdep.
> 
> Joel,
> 
> I got to this patch and I'm still reviewing it. I'll hopefully have my
> full review done by next week. I'll make it a priority. But I still
> would like Peter's ack on this one, as he's the maintainer of lockdep.

Thanks a lot Steven.

Peter, the lockdep calls are just small changes to the calling of the irq
on/off hooks and minor clean ups. Also I ran full locking API selftests with
all tests passing. I hope you are Ok with this change. Appreciate an Ack for
the lockdep bits and thanks.

-Joel
 

> Thanks,
> 
> -- Steve
> 
> 
> On Thu, 28 Jun 2018 11:21:47 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > 
> > This patch detaches the preemptirq tracepoints from the tracers and
> > keeps it separate.
> > 
> > Advantages:
> > * Lockdep and irqsoff event can now run in parallel since they no longer
> > have their own calls.
> > 
> > * This unifies the usecase of adding hooks to an irqsoff and irqson
> > event, and a preemptoff and preempton event.
> >   3 users of the events exist:
> >   - Lockdep
> >   - irqsoff and preemptoff tracers
> >   - irqs and preempt trace events
> > 
> > The unification cleans up several ifdefs and makes the code in preempt
> > tracer and irqsoff tracers simpler. It gets rid of all the horrific
> > ifdeferry around PROVE_LOCKING and makes configuration of the different
> > users of the tracepoints more easy and understandable. It also gets rid
> > of the time_* function calls from the lockdep hooks used to call into
> > the preemptirq tracer which is not needed anymore. The negative delta in
> > lines of code in this patch is quite large too.
> > 
> > In the patch we introduce a new CONFIG option PREEMPTIRQ_TRACEPOINTS
> > as a single point for registering probes onto the tracepoints. With
> > this,
> > the web of config options for preempt/irq toggle tracepoints and its
> > users becomes:
> > 
> >  PREEMPT_TRACER   PREEMPTIRQ_EVENTS  IRQSOFF_TRACER PROVE_LOCKING
> >        |                 |     \         |           |
> >        \    (selects)    /      \        \ (selects) /
> >       TRACE_PREEMPT_TOGGLE       ----> TRACE_IRQFLAGS
> >                       \                  /
> >                        \ (depends on)   /
> >                      PREEMPTIRQ_TRACEPOINTS
> > 
> > One note, I have to check for lockdep recursion in the code that calls
> > the trace events API and bail out if we're in lockdep recursion
> > protection to prevent something like the following case: a spin_lock is
> > taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> > and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> > this function, a call to get_lock_stats happens which calls
> > preempt_disable, which calls trace IRQS off somewhere which enters my
> > tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> > This flag is then never cleared causing lockdep paths to never be
> > entered and thus causing splats and other bad things.
> > 
> > Other than the performance tests mentioned in the previous patch, I also
> > ran the locking API test suite. I verified that all tests cases are
> > passing.
> > 
> > I also injected issues by not registering lockdep probes onto the
> > tracepoints and I see failures to confirm that the probes are indeed
> > working.
> > 
> > This series + lockdep probes not registered (just to inject errors):
> > [    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> > [    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> > [    0.000000]        sirq-safe-A => hirqs-on/12:FAILED|FAILED|  ok  |
> > [    0.000000]        sirq-safe-A => hirqs-on/21:FAILED|FAILED|  ok  |
> > [    0.000000]          hard-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
> > [    0.000000]          soft-safe-A + irqs-on/12:FAILED|FAILED|  ok  |
> > [    0.000000]          hard-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
> > [    0.000000]          soft-safe-A + irqs-on/21:FAILED|FAILED|  ok  |
> > [    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> > [    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> > 
> > With this series + lockdep probes registered, all locking tests pass:
> > 
> > [    0.000000]      hard-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> > [    0.000000]      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
> > [    0.000000]        sirq-safe-A => hirqs-on/12:  ok  |  ok  |  ok  |
> > [    0.000000]        sirq-safe-A => hirqs-on/21:  ok  |  ok  |  ok  |
> > [    0.000000]          hard-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
> > [    0.000000]          soft-safe-A + irqs-on/12:  ok  |  ok  |  ok  |
> > [    0.000000]          hard-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
> > [    0.000000]          soft-safe-A + irqs-on/21:  ok  |  ok  |  ok  |
> > [    0.000000]     hard-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> > [    0.000000]     soft-safe-A + unsafe-B #1/123:  ok  |  ok  |  ok  |
> > 
> > Reviewed-by: Namhyung Kim <namhyung@kernel.org>
> > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > ---
> >  include/linux/ftrace.h            |  11 +-
> >  include/linux/irqflags.h          |  11 +-
> >  include/linux/lockdep.h           |   8 +-
> >  include/linux/preempt.h           |   2 +-
> >  include/trace/events/preemptirq.h |  23 +--
> >  init/main.c                       |   5 +-
> >  kernel/locking/lockdep.c          |  35 ++---
> >  kernel/sched/core.c               |   2 +-
> >  kernel/trace/Kconfig              |  22 ++-
> >  kernel/trace/Makefile             |   2 +-
> >  kernel/trace/trace_irqsoff.c      | 231 ++++++++----------------------
> >  kernel/trace/trace_preemptirq.c   |  71 +++++++++
> >  12 files changed, 194 insertions(+), 229 deletions(-)
> >  create mode 100644 kernel/trace/trace_preemptirq.c
> > 
> > diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> > index 8154f4920fcb..f32e3c81407e 100644
> > --- a/include/linux/ftrace.h
> > +++ b/include/linux/ftrace.h
> > @@ -709,16 +709,7 @@ static inline unsigned long get_lock_parent_ip(void)
> >  	return CALLER_ADDR2;
> >  }
> >  
> > -#ifdef CONFIG_IRQSOFF_TRACER
> > -  extern void time_hardirqs_on(unsigned long a0, unsigned long a1);
> > -  extern void time_hardirqs_off(unsigned long a0, unsigned long a1);
> > -#else
> > -  static inline void time_hardirqs_on(unsigned long a0, unsigned long a1) { }
> > -  static inline void time_hardirqs_off(unsigned long a0, unsigned long a1) { }
> > -#endif
> > -
> > -#if defined(CONFIG_PREEMPT_TRACER) || \
> > -	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
> > +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
> >    extern void trace_preempt_on(unsigned long a0, unsigned long a1);
> >    extern void trace_preempt_off(unsigned long a0, unsigned long a1);
> >  #else
> > diff --git a/include/linux/irqflags.h b/include/linux/irqflags.h
> > index 9700f00bbc04..50edb9cbbd26 100644
> > --- a/include/linux/irqflags.h
> > +++ b/include/linux/irqflags.h
> > @@ -15,9 +15,16 @@
> >  #include <linux/typecheck.h>
> >  #include <asm/irqflags.h>
> >  
> > -#ifdef CONFIG_TRACE_IRQFLAGS
> > +/* Currently trace_softirqs_on/off is used only by lockdep */
> > +#ifdef CONFIG_PROVE_LOCKING
> >    extern void trace_softirqs_on(unsigned long ip);
> >    extern void trace_softirqs_off(unsigned long ip);
> > +#else
> > +# define trace_softirqs_on(ip)	do { } while (0)
> > +# define trace_softirqs_off(ip)	do { } while (0)
> > +#endif
> > +
> > +#ifdef CONFIG_TRACE_IRQFLAGS
> >    extern void trace_hardirqs_on(void);
> >    extern void trace_hardirqs_off(void);
> >  # define trace_hardirq_context(p)	((p)->hardirq_context)
> > @@ -43,8 +50,6 @@ do {						\
> >  #else
> >  # define trace_hardirqs_on()		do { } while (0)
> >  # define trace_hardirqs_off()		do { } while (0)
> > -# define trace_softirqs_on(ip)		do { } while (0)
> > -# define trace_softirqs_off(ip)		do { } while (0)
> >  # define trace_hardirq_context(p)	0
> >  # define trace_softirq_context(p)	0
> >  # define trace_hardirqs_enabled(p)	0
> > diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
> > index 6fc77d4dbdcd..a8113357ceeb 100644
> > --- a/include/linux/lockdep.h
> > +++ b/include/linux/lockdep.h
> > @@ -266,7 +266,8 @@ struct held_lock {
> >  /*
> >   * Initialization, self-test and debugging-output methods:
> >   */
> > -extern void lockdep_info(void);
> > +extern void lockdep_init(void);
> > +extern void lockdep_init_early(void);
> >  extern void lockdep_reset(void);
> >  extern void lockdep_reset_lock(struct lockdep_map *lock);
> >  extern void lockdep_free_key_range(void *start, unsigned long size);
> > @@ -406,7 +407,8 @@ static inline void lockdep_on(void)
> >  # define lock_downgrade(l, i)			do { } while (0)
> >  # define lock_set_class(l, n, k, s, i)		do { } while (0)
> >  # define lock_set_subclass(l, s, i)		do { } while (0)
> > -# define lockdep_info()				do { } while (0)
> > +# define lockdep_init()				do { } while (0)
> > +# define lockdep_init_early()			do { } while (0)
> >  # define lockdep_init_map(lock, name, key, sub) \
> >  		do { (void)(name); (void)(key); } while (0)
> >  # define lockdep_set_class(lock, key)		do { (void)(key); } while (0)
> > @@ -532,7 +534,7 @@ do {								\
> >  
> >  #endif /* CONFIG_LOCKDEP */
> >  
> > -#ifdef CONFIG_TRACE_IRQFLAGS
> > +#ifdef CONFIG_PROVE_LOCKING
> >  extern void print_irqtrace_events(struct task_struct *curr);
> >  #else
> >  static inline void print_irqtrace_events(struct task_struct *curr)
> > diff --git a/include/linux/preempt.h b/include/linux/preempt.h
> > index 5bd3f151da78..c01813c3fbe9 100644
> > --- a/include/linux/preempt.h
> > +++ b/include/linux/preempt.h
> > @@ -150,7 +150,7 @@
> >   */
> >  #define in_atomic_preempt_off() (preempt_count() != PREEMPT_DISABLE_OFFSET)
> >  
> > -#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_PREEMPT_TRACER)
> > +#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_TRACE_PREEMPT_TOGGLE)
> >  extern void preempt_count_add(int val);
> >  extern void preempt_count_sub(int val);
> >  #define preempt_count_dec_and_test() \
> > diff --git a/include/trace/events/preemptirq.h b/include/trace/events/preemptirq.h
> > index 9c4eb33c5a1d..9a0d4ceeb166 100644
> > --- a/include/trace/events/preemptirq.h
> > +++ b/include/trace/events/preemptirq.h
> > @@ -1,4 +1,4 @@
> > -#ifdef CONFIG_PREEMPTIRQ_EVENTS
> > +#ifdef CONFIG_PREEMPTIRQ_TRACEPOINTS
> >  
> >  #undef TRACE_SYSTEM
> >  #define TRACE_SYSTEM preemptirq
> > @@ -32,7 +32,7 @@ DECLARE_EVENT_CLASS(preemptirq_template,
> >  		  (void *)((unsigned long)(_stext) + __entry->parent_offs))
> >  );
> >  
> > -#ifndef CONFIG_PROVE_LOCKING
> > +#ifdef CONFIG_TRACE_IRQFLAGS
> >  DEFINE_EVENT(preemptirq_template, irq_disable,
> >  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
> >  	     TP_ARGS(ip, parent_ip));
> > @@ -40,9 +40,14 @@ DEFINE_EVENT(preemptirq_template, irq_disable,
> >  DEFINE_EVENT(preemptirq_template, irq_enable,
> >  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
> >  	     TP_ARGS(ip, parent_ip));
> > +#else
> > +#define trace_irq_enable(...)
> > +#define trace_irq_disable(...)
> > +#define trace_irq_enable_rcuidle(...)
> > +#define trace_irq_disable_rcuidle(...)
> >  #endif
> >  
> > -#ifdef CONFIG_DEBUG_PREEMPT
> > +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
> >  DEFINE_EVENT(preemptirq_template, preempt_disable,
> >  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
> >  	     TP_ARGS(ip, parent_ip));
> > @@ -50,22 +55,22 @@ DEFINE_EVENT(preemptirq_template, preempt_disable,
> >  DEFINE_EVENT(preemptirq_template, preempt_enable,
> >  	     TP_PROTO(unsigned long ip, unsigned long parent_ip),
> >  	     TP_ARGS(ip, parent_ip));
> > +#else
> > +#define trace_preempt_enable(...)
> > +#define trace_preempt_disable(...)
> > +#define trace_preempt_enable_rcuidle(...)
> > +#define trace_preempt_disable_rcuidle(...)
> >  #endif
> >  
> >  #endif /* _TRACE_PREEMPTIRQ_H */
> >  
> >  #include <trace/define_trace.h>
> >  
> > -#endif /* !CONFIG_PREEMPTIRQ_EVENTS */
> > -
> > -#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || defined(CONFIG_PROVE_LOCKING)
> > +#else /* !CONFIG_PREEMPTIRQ_TRACEPOINTS */
> >  #define trace_irq_enable(...)
> >  #define trace_irq_disable(...)
> >  #define trace_irq_enable_rcuidle(...)
> >  #define trace_irq_disable_rcuidle(...)
> > -#endif
> > -
> > -#if !defined(CONFIG_PREEMPTIRQ_EVENTS) || !defined(CONFIG_DEBUG_PREEMPT)
> >  #define trace_preempt_enable(...)
> >  #define trace_preempt_disable(...)
> >  #define trace_preempt_enable_rcuidle(...)
> > diff --git a/init/main.c b/init/main.c
> > index 3b4ada11ed52..44fe43be84c1 100644
> > --- a/init/main.c
> > +++ b/init/main.c
> > @@ -648,6 +648,9 @@ asmlinkage __visible void __init start_kernel(void)
> >  	profile_init();
> >  	call_function_init();
> >  	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
> > +
> > +	lockdep_init_early();
> > +
> >  	early_boot_irqs_disabled = false;
> >  	local_irq_enable();
> >  
> > @@ -663,7 +666,7 @@ asmlinkage __visible void __init start_kernel(void)
> >  		panic("Too many boot %s vars at `%s'", panic_later,
> >  		      panic_param);
> >  
> > -	lockdep_info();
> > +	lockdep_init();
> >  
> >  	/*
> >  	 * Need to run this when irqs are enabled, because it wants
> > diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> > index 5fa4d3138bf1..b961a1698e98 100644
> > --- a/kernel/locking/lockdep.c
> > +++ b/kernel/locking/lockdep.c
> > @@ -55,6 +55,7 @@
> >  
> >  #include "lockdep_internals.h"
> >  
> > +#include <trace/events/preemptirq.h>
> >  #define CREATE_TRACE_POINTS
> >  #include <trace/events/lock.h>
> >  
> > @@ -2845,10 +2846,9 @@ static void __trace_hardirqs_on_caller(unsigned long ip)
> >  	debug_atomic_inc(hardirqs_on_events);
> >  }
> >  
> > -__visible void trace_hardirqs_on_caller(unsigned long ip)
> > +static void lockdep_hardirqs_on(void *none, unsigned long ignore,
> > +				unsigned long ip)
> >  {
> > -	time_hardirqs_on(CALLER_ADDR0, ip);
> > -
> >  	if (unlikely(!debug_locks || current->lockdep_recursion))
> >  		return;
> >  
> > @@ -2887,23 +2887,15 @@ __visible void trace_hardirqs_on_caller(unsigned long ip)
> >  	__trace_hardirqs_on_caller(ip);
> >  	current->lockdep_recursion = 0;
> >  }
> > -EXPORT_SYMBOL(trace_hardirqs_on_caller);
> > -
> > -void trace_hardirqs_on(void)
> > -{
> > -	trace_hardirqs_on_caller(CALLER_ADDR0);
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_on);
> >  
> >  /*
> >   * Hardirqs were disabled:
> >   */
> > -__visible void trace_hardirqs_off_caller(unsigned long ip)
> > +static void lockdep_hardirqs_off(void *none, unsigned long ignore,
> > +				 unsigned long ip)
> >  {
> >  	struct task_struct *curr = current;
> >  
> > -	time_hardirqs_off(CALLER_ADDR0, ip);
> > -
> >  	if (unlikely(!debug_locks || current->lockdep_recursion))
> >  		return;
> >  
> > @@ -2925,13 +2917,6 @@ __visible void trace_hardirqs_off_caller(unsigned long ip)
> >  	} else
> >  		debug_atomic_inc(redundant_hardirqs_off);
> >  }
> > -EXPORT_SYMBOL(trace_hardirqs_off_caller);
> > -
> > -void trace_hardirqs_off(void)
> > -{
> > -	trace_hardirqs_off_caller(CALLER_ADDR0);
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_off);
> >  
> >  /*
> >   * Softirqs will be enabled:
> > @@ -4338,7 +4323,15 @@ void lockdep_reset_lock(struct lockdep_map *lock)
> >  	raw_local_irq_restore(flags);
> >  }
> >  
> > -void __init lockdep_info(void)
> > +void __init lockdep_init_early(void)
> > +{
> > +#ifdef CONFIG_PROVE_LOCKING
> > +	register_trace_prio_irq_disable(lockdep_hardirqs_off, NULL, INT_MAX);
> > +	register_trace_prio_irq_enable(lockdep_hardirqs_on, NULL, INT_MIN);
> > +#endif
> > +}
> > +
> > +void __init lockdep_init(void)
> >  {
> >  	printk("Lock dependency validator: Copyright (c) 2006 Red Hat, Inc., Ingo Molnar\n");
> >  
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index 78d8facba456..4c956f6849ec 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -3192,7 +3192,7 @@ static inline void sched_tick_stop(int cpu) { }
> >  #endif
> >  
> >  #if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
> > -				defined(CONFIG_PREEMPT_TRACER))
> > +				defined(CONFIG_TRACE_PREEMPT_TOGGLE))
> >  /*
> >   * If the value passed in is equal to the current preempt count
> >   * then we just disabled preemption. Start timing the latency.
> > diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
> > index dcc0166d1997..8d51351e3149 100644
> > --- a/kernel/trace/Kconfig
> > +++ b/kernel/trace/Kconfig
> > @@ -82,6 +82,15 @@ config RING_BUFFER_ALLOW_SWAP
> >  	 Allow the use of ring_buffer_swap_cpu.
> >  	 Adds a very slight overhead to tracing when enabled.
> >  
> > +config PREEMPTIRQ_TRACEPOINTS
> > +	bool
> > +	depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
> > +	select TRACING
> > +	default y
> > +	help
> > +	  Create preempt/irq toggle tracepoints if needed, so that other parts
> > +	  of the kernel can use them to generate or add hooks to them.
> > +
> >  # All tracer options should select GENERIC_TRACER. For those options that are
> >  # enabled by all tracers (context switch and event tracer) they select TRACING.
> >  # This allows those options to appear when no other tracer is selected. But the
> > @@ -155,18 +164,20 @@ config FUNCTION_GRAPH_TRACER
> >  	  the return value. This is done by setting the current return
> >  	  address on the current task structure into a stack of calls.
> >  
> > +config TRACE_PREEMPT_TOGGLE
> > +	bool
> > +	help
> > +	  Enables hooks which will be called when preemption is first disabled,
> > +	  and last enabled.
> >  
> >  config PREEMPTIRQ_EVENTS
> >  	bool "Enable trace events for preempt and irq disable/enable"
> >  	select TRACE_IRQFLAGS
> > -	depends on DEBUG_PREEMPT || !PROVE_LOCKING
> > -	depends on TRACING
> > +	select TRACE_PREEMPT_TOGGLE if PREEMPT
> > +	select GENERIC_TRACER
> >  	default n
> >  	help
> >  	  Enable tracing of disable and enable events for preemption and irqs.
> > -	  For tracing preempt disable/enable events, DEBUG_PREEMPT must be
> > -	  enabled. For tracing irq disable/enable events, PROVE_LOCKING must
> > -	  be disabled.
> >  
> >  config IRQSOFF_TRACER
> >  	bool "Interrupts-off Latency Tracer"
> > @@ -203,6 +214,7 @@ config PREEMPT_TRACER
> >  	select RING_BUFFER_ALLOW_SWAP
> >  	select TRACER_SNAPSHOT
> >  	select TRACER_SNAPSHOT_PER_CPU_SWAP
> > +	select TRACE_PREEMPT_TOGGLE
> >  	help
> >  	  This option measures the time spent in preemption-off critical
> >  	  sections, with microsecond accuracy.
> > diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile
> > index e2538c7638d4..84a0cb222f20 100644
> > --- a/kernel/trace/Makefile
> > +++ b/kernel/trace/Makefile
> > @@ -35,7 +35,7 @@ obj-$(CONFIG_TRACING) += trace_printk.o
> >  obj-$(CONFIG_TRACING_MAP) += tracing_map.o
> >  obj-$(CONFIG_CONTEXT_SWITCH_TRACER) += trace_sched_switch.o
> >  obj-$(CONFIG_FUNCTION_TRACER) += trace_functions.o
> > -obj-$(CONFIG_PREEMPTIRQ_EVENTS) += trace_irqsoff.o
> > +obj-$(CONFIG_PREEMPTIRQ_TRACEPOINTS) += trace_preemptirq.o
> >  obj-$(CONFIG_IRQSOFF_TRACER) += trace_irqsoff.o
> >  obj-$(CONFIG_PREEMPT_TRACER) += trace_irqsoff.o
> >  obj-$(CONFIG_SCHED_TRACER) += trace_sched_wakeup.o
> > diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
> > index f8daa754cce2..770cd30cda40 100644
> > --- a/kernel/trace/trace_irqsoff.c
> > +++ b/kernel/trace/trace_irqsoff.c
> > @@ -16,7 +16,6 @@
> >  
> >  #include "trace.h"
> >  
> > -#define CREATE_TRACE_POINTS
> >  #include <trace/events/preemptirq.h>
> >  
> >  #if defined(CONFIG_IRQSOFF_TRACER) || defined(CONFIG_PREEMPT_TRACER)
> > @@ -450,66 +449,6 @@ void stop_critical_timings(void)
> >  }
> >  EXPORT_SYMBOL_GPL(stop_critical_timings);
> >  
> > -#ifdef CONFIG_IRQSOFF_TRACER
> > -#ifdef CONFIG_PROVE_LOCKING
> > -void time_hardirqs_on(unsigned long a0, unsigned long a1)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		stop_critical_timing(a0, a1);
> > -}
> > -
> > -void time_hardirqs_off(unsigned long a0, unsigned long a1)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		start_critical_timing(a0, a1);
> > -}
> > -
> > -#else /* !CONFIG_PROVE_LOCKING */
> > -
> > -/*
> > - * We are only interested in hardirq on/off events:
> > - */
> > -static inline void tracer_hardirqs_on(void)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
> > -}
> > -
> > -static inline void tracer_hardirqs_off(void)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
> > -}
> > -
> > -static inline void tracer_hardirqs_on_caller(unsigned long caller_addr)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		stop_critical_timing(CALLER_ADDR0, caller_addr);
> > -}
> > -
> > -static inline void tracer_hardirqs_off_caller(unsigned long caller_addr)
> > -{
> > -	if (!preempt_trace() && irq_trace())
> > -		start_critical_timing(CALLER_ADDR0, caller_addr);
> > -}
> > -
> > -#endif /* CONFIG_PROVE_LOCKING */
> > -#endif /*  CONFIG_IRQSOFF_TRACER */
> > -
> > -#ifdef CONFIG_PREEMPT_TRACER
> > -static inline void tracer_preempt_on(unsigned long a0, unsigned long a1)
> > -{
> > -	if (preempt_trace() && !irq_trace())
> > -		stop_critical_timing(a0, a1);
> > -}
> > -
> > -static inline void tracer_preempt_off(unsigned long a0, unsigned long a1)
> > -{
> > -	if (preempt_trace() && !irq_trace())
> > -		start_critical_timing(a0, a1);
> > -}
> > -#endif /* CONFIG_PREEMPT_TRACER */
> > -
> >  #ifdef CONFIG_FUNCTION_TRACER
> >  static bool function_enabled;
> >  
> > @@ -659,15 +598,34 @@ static void irqsoff_tracer_stop(struct trace_array *tr)
> >  }
> >  
> >  #ifdef CONFIG_IRQSOFF_TRACER
> > +/*
> > + * We are only interested in hardirq on/off events:
> > + */
> > +static void tracer_hardirqs_on(void *none, unsigned long a0, unsigned long a1)
> > +{
> > +	if (!preempt_trace() && irq_trace())
> > +		stop_critical_timing(a0, a1);
> > +}
> > +
> > +static void tracer_hardirqs_off(void *none, unsigned long a0, unsigned long a1)
> > +{
> > +	if (!preempt_trace() && irq_trace())
> > +		start_critical_timing(a0, a1);
> > +}
> > +
> >  static int irqsoff_tracer_init(struct trace_array *tr)
> >  {
> >  	trace_type = TRACER_IRQS_OFF;
> >  
> > +	register_trace_irq_disable(tracer_hardirqs_off, NULL);
> > +	register_trace_irq_enable(tracer_hardirqs_on, NULL);
> >  	return __irqsoff_tracer_init(tr);
> >  }
> >  
> >  static void irqsoff_tracer_reset(struct trace_array *tr)
> >  {
> > +	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
> > +	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
> >  	__irqsoff_tracer_reset(tr);
> >  }
> >  
> > @@ -690,21 +648,34 @@ static struct tracer irqsoff_tracer __read_mostly =
> >  	.allow_instances = true,
> >  	.use_max_tr	= true,
> >  };
> > -# define register_irqsoff(trace) register_tracer(&trace)
> > -#else
> > -# define register_irqsoff(trace) do { } while (0)
> > -#endif
> > +#endif /*  CONFIG_IRQSOFF_TRACER */
> >  
> >  #ifdef CONFIG_PREEMPT_TRACER
> > +static void tracer_preempt_on(void *none, unsigned long a0, unsigned long a1)
> > +{
> > +	if (preempt_trace() && !irq_trace())
> > +		stop_critical_timing(a0, a1);
> > +}
> > +
> > +static void tracer_preempt_off(void *none, unsigned long a0, unsigned long a1)
> > +{
> > +	if (preempt_trace() && !irq_trace())
> > +		start_critical_timing(a0, a1);
> > +}
> > +
> >  static int preemptoff_tracer_init(struct trace_array *tr)
> >  {
> >  	trace_type = TRACER_PREEMPT_OFF;
> >  
> > +	register_trace_preempt_disable(tracer_preempt_off, NULL);
> > +	register_trace_preempt_enable(tracer_preempt_on, NULL);
> >  	return __irqsoff_tracer_init(tr);
> >  }
> >  
> >  static void preemptoff_tracer_reset(struct trace_array *tr)
> >  {
> > +	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
> > +	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
> >  	__irqsoff_tracer_reset(tr);
> >  }
> >  
> > @@ -727,23 +698,29 @@ static struct tracer preemptoff_tracer __read_mostly =
> >  	.allow_instances = true,
> >  	.use_max_tr	= true,
> >  };
> > -# define register_preemptoff(trace) register_tracer(&trace)
> > -#else
> > -# define register_preemptoff(trace) do { } while (0)
> > -#endif
> > +#endif /* CONFIG_PREEMPT_TRACER */
> >  
> > -#if defined(CONFIG_IRQSOFF_TRACER) && \
> > -	defined(CONFIG_PREEMPT_TRACER)
> > +#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
> >  
> >  static int preemptirqsoff_tracer_init(struct trace_array *tr)
> >  {
> >  	trace_type = TRACER_IRQS_OFF | TRACER_PREEMPT_OFF;
> >  
> > +	register_trace_irq_disable(tracer_hardirqs_off, NULL);
> > +	register_trace_irq_enable(tracer_hardirqs_on, NULL);
> > +	register_trace_preempt_disable(tracer_preempt_off, NULL);
> > +	register_trace_preempt_enable(tracer_preempt_on, NULL);
> > +
> >  	return __irqsoff_tracer_init(tr);
> >  }
> >  
> >  static void preemptirqsoff_tracer_reset(struct trace_array *tr)
> >  {
> > +	unregister_trace_irq_disable(tracer_hardirqs_off, NULL);
> > +	unregister_trace_irq_enable(tracer_hardirqs_on, NULL);
> > +	unregister_trace_preempt_disable(tracer_preempt_off, NULL);
> > +	unregister_trace_preempt_enable(tracer_preempt_on, NULL);
> > +
> >  	__irqsoff_tracer_reset(tr);
> >  }
> >  
> > @@ -766,115 +743,21 @@ static struct tracer preemptirqsoff_tracer __read_mostly =
> >  	.allow_instances = true,
> >  	.use_max_tr	= true,
> >  };
> > -
> > -# define register_preemptirqsoff(trace) register_tracer(&trace)
> > -#else
> > -# define register_preemptirqsoff(trace) do { } while (0)
> >  #endif
> >  
> >  __init static int init_irqsoff_tracer(void)
> >  {
> > -	register_irqsoff(irqsoff_tracer);
> > -	register_preemptoff(preemptoff_tracer);
> > -	register_preemptirqsoff(preemptirqsoff_tracer);
> > -
> > -	return 0;
> > -}
> > -core_initcall(init_irqsoff_tracer);
> > -#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
> > -
> > -#ifndef CONFIG_IRQSOFF_TRACER
> > -static inline void tracer_hardirqs_on(void) { }
> > -static inline void tracer_hardirqs_off(void) { }
> > -static inline void tracer_hardirqs_on_caller(unsigned long caller_addr) { }
> > -static inline void tracer_hardirqs_off_caller(unsigned long caller_addr) { }
> > +#ifdef CONFIG_IRQSOFF_TRACER
> > +	register_tracer(&irqsoff_tracer);
> >  #endif
> > -
> > -#ifndef CONFIG_PREEMPT_TRACER
> > -static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
> > -static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
> > +#ifdef CONFIG_PREEMPT_TRACER
> > +	register_tracer(&preemptoff_tracer);
> >  #endif
> > -
> > -#if defined(CONFIG_TRACE_IRQFLAGS) && !defined(CONFIG_PROVE_LOCKING)
> > -/* Per-cpu variable to prevent redundant calls when IRQs already off */
> > -static DEFINE_PER_CPU(int, tracing_irq_cpu);
> > -
> > -void trace_hardirqs_on(void)
> > -{
> > -	if (!this_cpu_read(tracing_irq_cpu))
> > -		return;
> > -
> > -	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> > -	tracer_hardirqs_on();
> > -
> > -	this_cpu_write(tracing_irq_cpu, 0);
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_on);
> > -
> > -void trace_hardirqs_off(void)
> > -{
> > -	if (this_cpu_read(tracing_irq_cpu))
> > -		return;
> > -
> > -	this_cpu_write(tracing_irq_cpu, 1);
> > -
> > -	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> > -	tracer_hardirqs_off();
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_off);
> > -
> > -__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
> > -{
> > -	if (!this_cpu_read(tracing_irq_cpu))
> > -		return;
> > -
> > -	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
> > -	tracer_hardirqs_on_caller(caller_addr);
> > -
> > -	this_cpu_write(tracing_irq_cpu, 0);
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_on_caller);
> > -
> > -__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
> > -{
> > -	if (this_cpu_read(tracing_irq_cpu))
> > -		return;
> > -
> > -	this_cpu_write(tracing_irq_cpu, 1);
> > -
> > -	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
> > -	tracer_hardirqs_off_caller(caller_addr);
> > -}
> > -EXPORT_SYMBOL(trace_hardirqs_off_caller);
> > -
> > -/*
> > - * Stubs:
> > - */
> > -
> > -void trace_softirqs_on(unsigned long ip)
> > -{
> > -}
> > -
> > -void trace_softirqs_off(unsigned long ip)
> > -{
> > -}
> > -
> > -inline void print_irqtrace_events(struct task_struct *curr)
> > -{
> > -}
> > +#if defined(CONFIG_IRQSOFF_TRACER) && defined(CONFIG_PREEMPT_TRACER)
> > +	register_tracer(&preemptirqsoff_tracer);
> >  #endif
> >  
> > -#if defined(CONFIG_PREEMPT_TRACER) || \
> > -	(defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_PREEMPTIRQ_EVENTS))
> > -void trace_preempt_on(unsigned long a0, unsigned long a1)
> > -{
> > -	trace_preempt_enable_rcuidle(a0, a1);
> > -	tracer_preempt_on(a0, a1);
> > -}
> > -
> > -void trace_preempt_off(unsigned long a0, unsigned long a1)
> > -{
> > -	trace_preempt_disable_rcuidle(a0, a1);
> > -	tracer_preempt_off(a0, a1);
> > +	return 0;
> >  }
> > -#endif
> > +core_initcall(init_irqsoff_tracer);
> > +#endif /* IRQSOFF_TRACER || PREEMPTOFF_TRACER */
> > diff --git a/kernel/trace/trace_preemptirq.c b/kernel/trace/trace_preemptirq.c
> > new file mode 100644
> > index 000000000000..dc01c7f4d326
> > --- /dev/null
> > +++ b/kernel/trace/trace_preemptirq.c
> > @@ -0,0 +1,71 @@
> > +/*
> > + * preemptoff and irqoff tracepoints
> > + *
> > + * Copyright (C) Joel Fernandes (Google) <joel@joelfernandes.org>
> > + */
> > +
> > +#include <linux/kallsyms.h>
> > +#include <linux/uaccess.h>
> > +#include <linux/module.h>
> > +#include <linux/ftrace.h>
> > +
> > +#define CREATE_TRACE_POINTS
> > +#include <trace/events/preemptirq.h>
> > +
> > +#ifdef CONFIG_TRACE_IRQFLAGS
> > +/* Per-cpu variable to prevent redundant calls when IRQs already off */
> > +static DEFINE_PER_CPU(int, tracing_irq_cpu);
> > +
> > +void trace_hardirqs_on(void)
> > +{
> > +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> > +		return;
> > +
> > +	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> > +	this_cpu_write(tracing_irq_cpu, 0);
> > +}
> > +EXPORT_SYMBOL(trace_hardirqs_on);
> > +
> > +void trace_hardirqs_off(void)
> > +{
> > +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> > +		return;
> > +
> > +	this_cpu_write(tracing_irq_cpu, 1);
> > +	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> > +}
> > +EXPORT_SYMBOL(trace_hardirqs_off);
> > +
> > +__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
> > +{
> > +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> > +		return;
> > +
> > +	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
> > +	this_cpu_write(tracing_irq_cpu, 0);
> > +}
> > +EXPORT_SYMBOL(trace_hardirqs_on_caller);
> > +
> > +__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
> > +{
> > +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> > +		return;
> > +
> > +	this_cpu_write(tracing_irq_cpu, 1);
> > +	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
> > +}
> > +EXPORT_SYMBOL(trace_hardirqs_off_caller);
> > +#endif /* CONFIG_TRACE_IRQFLAGS */
> > +
> > +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
> > +
> > +void trace_preempt_on(unsigned long a0, unsigned long a1)
> > +{
> > +	trace_preempt_enable_rcuidle(a0, a1);
> > +}
> > +
> > +void trace_preempt_off(unsigned long a0, unsigned long a1)
> > +{
> > +	trace_preempt_disable_rcuidle(a0, a1);
> > +}
> > +#endif
> 

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
  2018-07-06 22:06   ` Steven Rostedt
@ 2018-07-10 14:20   ` Steven Rostedt
  2018-07-10 17:33     ` Joel Fernandes
  2018-07-11 13:12   ` Peter Zijlstra
  2 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-10 14:20 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Thu, 28 Jun 2018 11:21:47 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> diff --git a/kernel/trace/trace_preemptirq.c b/kernel/trace/trace_preemptirq.c
> new file mode 100644
> index 000000000000..dc01c7f4d326
> --- /dev/null
> +++ b/kernel/trace/trace_preemptirq.c
> @@ -0,0 +1,71 @@

Can you send a patch on top of this, that adds a SPDX header here. Just
add another patch, no need to resend this one.

I need to go through all the files in kernel/trace/* and add SPDX
headers. I don't want to add more files that don't have them.

I'm still playing around with this patch, and testing it.

-- Steve

> +/*
> + * preemptoff and irqoff tracepoints
> + *
> + * Copyright (C) Joel Fernandes (Google) <joel@joelfernandes.org>
> + */
> +
> +#include <linux/kallsyms.h>
> +#include <linux/uaccess.h>
> +#include <linux/module.h>
> +#include <linux/ftrace.h>
> +
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/preemptirq.h>
> +
> +#ifdef CONFIG_TRACE_IRQFLAGS
> +/* Per-cpu variable to prevent redundant calls when IRQs already off */
> +static DEFINE_PER_CPU(int, tracing_irq_cpu);
> +
> +void trace_hardirqs_on(void)
> +{
> +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	trace_irq_enable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> +	this_cpu_write(tracing_irq_cpu, 0);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_on);
> +
> +void trace_hardirqs_off(void)
> +{
> +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	this_cpu_write(tracing_irq_cpu, 1);
> +	trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_off);
> +
> +__visible void trace_hardirqs_on_caller(unsigned long caller_addr)
> +{
> +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	trace_irq_enable_rcuidle(CALLER_ADDR0, caller_addr);
> +	this_cpu_write(tracing_irq_cpu, 0);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_on_caller);
> +
> +__visible void trace_hardirqs_off_caller(unsigned long caller_addr)
> +{
> +	if (lockdep_recursing(current) || this_cpu_read(tracing_irq_cpu))
> +		return;
> +
> +	this_cpu_write(tracing_irq_cpu, 1);
> +	trace_irq_disable_rcuidle(CALLER_ADDR0, caller_addr);
> +}
> +EXPORT_SYMBOL(trace_hardirqs_off_caller);
> +#endif /* CONFIG_TRACE_IRQFLAGS */
> +
> +#ifdef CONFIG_TRACE_PREEMPT_TOGGLE
> +
> +void trace_preempt_on(unsigned long a0, unsigned long a1)
> +{
> +	trace_preempt_enable_rcuidle(a0, a1);
> +}
> +
> +void trace_preempt_off(unsigned long a0, unsigned long a1)
> +{
> +	trace_preempt_disable_rcuidle(a0, a1);
> +}
> +#endif


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-10 14:20   ` Steven Rostedt
@ 2018-07-10 17:33     ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-10 17:33 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Tue, Jul 10, 2018 at 10:20:50AM -0400, Steven Rostedt wrote:
> On Thu, 28 Jun 2018 11:21:47 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > diff --git a/kernel/trace/trace_preemptirq.c b/kernel/trace/trace_preemptirq.c
> > new file mode 100644
> > index 000000000000..dc01c7f4d326
> > --- /dev/null
> > +++ b/kernel/trace/trace_preemptirq.c
> > @@ -0,0 +1,71 @@
> 
> Can you send a patch on top of this, that adds a SPDX header here. Just
> add another patch, no need to resend this one.
> 
> I need to go through all the files in kernel/trace/* and add SPDX
> headers. I don't want to add more files that don't have them.

Sure, I'll send a patch ontop of this.

> I'm still playing around with this patch, and testing it.

Ok, thanks.

-Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers
  2018-06-28 18:21 ` [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers Joel Fernandes
@ 2018-07-11  0:47   ` Steven Rostedt
  2018-07-11  5:26     ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11  0:47 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Andy Shevchenko, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Thu, 28 Jun 2018 11:21:48 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> 
> In this patch we introduce a test module for simulating a long atomic
> section in the kernel which the preemptoff or irqsoff tracers can
> detect. This module is to be used only for test purposes and is default
> disabled.
> 
> Following is the expected output (only briefly shown) that can be parsed
> to verify that the tracers are working correctly. We will use this from
> the kselftests in future patches.
> 
> For the preemptoff tracer:
> 
> echo preemptoff > /d/tracing/current_tracer
> sleep 1
> insmod ./test_atomic_sections.ko atomic_mode=preempt atomic_time=500000
> sleep 1
> bash-4.3# cat /d/tracing/trace
> preempt -1066    2...2    0us@: atomic_sect_run <-atomic_sect_run
> preempt -1066    2...2 500002us : atomic_sect_run <-atomic_sect_run
> preempt -1066    2...2 500004us : tracer_preempt_on <-atomic_sect_run
> preempt -1066    2...2 500012us : <stack trace>
>  => kthread
>  => ret_from_fork  
> 
> For the irqsoff tracer:
> 
> echo irqsoff > /d/tracing/current_tracer
> sleep 1
> insmod ./test_atomic_sections.ko atomic_mode=irq atomic_time=500000
> sleep 1
> bash-4.3# cat /d/tracing/trace
> irq dis -1069    1d..1    0us@: atomic_sect_run
> irq dis -1069    1d..1 500001us : atomic_sect_run
> irq dis -1069    1d..1 500002us : tracer_hardirqs_on <-atomic_sect_run
> irq dis -1069    1d..1 500005us : <stack trace>
>  => ret_from_fork  
> 
> Co-developed-by: Erick Reyes <erickreyes@google.com>
> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> ---
>  lib/Kconfig.debug          |  8 ++++
>  lib/Makefile               |  1 +
>  lib/test_atomic_sections.c | 77 ++++++++++++++++++++++++++++++++++++++

I think this code should reside in kernel/trace directory. I already
have modules there. See the ring_buffer_benchmark code and the test
module for mmio tracer.

>  3 files changed, 86 insertions(+)
>  create mode 100644 lib/test_atomic_sections.c
> 
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 8838d1158d19..622c90e1e066 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1956,6 +1956,14 @@ config TEST_KMOD
>  
>  	  If unsure, say N.
>  
> +config TEST_ATOMIC_SECTIONS
> +	tristate "Simulate atomic sections for tracers to detect"

Hmm, I don't like this title. It's not very obvious to what it is
about. What about "Preempt / IRQ disable delay thread to test latency
tracers" ? Or something along those lines.


> +	depends on m
> +	help
> +	  Select this option to build a test module that can help test atomic
> +	  sections by simulating them with a duration supplied as a module
> +	  parameter. Preempt disable and irq disable modes can be requested.

"If unsure say N"

> +
>  config TEST_DEBUG_VIRTUAL
>  	tristate "Test CONFIG_DEBUG_VIRTUAL feature"
>  	depends on DEBUG_VIRTUAL
> diff --git a/lib/Makefile b/lib/Makefile
> index 90dc5520b784..7831e747bf72 100644
> --- a/lib/Makefile
> +++ b/lib/Makefile
> @@ -44,6 +44,7 @@ obj-y += string_helpers.o
>  obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.o
>  obj-y += hexdump.o
>  obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
> +obj-$(CONFIG_TEST_ATOMIC_SECTIONS) += test_atomic_sections.o
>  obj-y += kstrtox.o
>  obj-$(CONFIG_FIND_BIT_BENCHMARK) += find_bit_benchmark.o
>  obj-$(CONFIG_TEST_BPF) += test_bpf.o
> diff --git a/lib/test_atomic_sections.c b/lib/test_atomic_sections.c
> new file mode 100644
> index 000000000000..1eef518f0974
> --- /dev/null
> +++ b/lib/test_atomic_sections.c
> @@ -0,0 +1,77 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Atomic section emulation test module
> + *
> + * Emulates atomic sections by disabling IRQs or preemption
> + * and doing a busy wait for a specified amount of time.
> + * This can be used for testing of different atomic section
> + * tracers such as irqsoff tracers.
> + *
> + * (c) 2018. Google LLC
> + */
> +
> +#include <linux/delay.h>
> +#include <linux/interrupt.h>
> +#include <linux/irq.h>
> +#include <linux/kernel.h>
> +#include <linux/kthread.h>
> +#include <linux/ktime.h>
> +#include <linux/module.h>
> +#include <linux/printk.h>
> +#include <linux/string.h>
> +
> +static ulong atomic_time = 100;
> +static char atomic_mode[10] = "irq";
> +
> +module_param_named(atomic_time, atomic_time, ulong, S_IRUGO);
> +module_param_string(atomic_mode, atomic_mode, 10, S_IRUGO);
> +MODULE_PARM_DESC(atomic_time, "Period in microseconds (100 uS default)");

It's not a "Period", it's a delay. "Length of time in critical section"

> +MODULE_PARM_DESC(atomic_mode, "Mode of the test such as preempt or irq (default irq)");

  "Mode of the test: preempt or irq disabled (default irq)"

> +
> +static void busy_wait(ulong time)
> +{
> +	ktime_t start, end;
> +	start = ktime_get();
> +	do {
> +		end = ktime_get();
> +		if (kthread_should_stop())
> +			break;
> +	} while (ktime_to_ns(ktime_sub(end, start)) < (time * 1000));
> +}
> +
> +int atomic_sect_run(void *data)
> +{
> +	unsigned long flags;
> +
> +	if (!strcmp(atomic_mode, "irq")) {
> +		local_irq_save(flags);
> +		busy_wait(atomic_time);
> +		local_irq_restore(flags);
> +	} else if (!strcmp(atomic_mode, "preempt")) {
> +		preempt_disable();
> +		busy_wait(atomic_time);
> +		preempt_enable();
> +	}

So this is a one shot deal? That should be explained somewhere,
probably in the config help message. In fact, I think the config help
message should show how to use this.

-- Steve

> +
> +	return 0;
> +}
> +
> +static int __init atomic_sect_init(void)
> +{
> +	char task_name[50];
> +	struct task_struct *test_task;
> +
> +	snprintf(task_name, sizeof(task_name), "%s_test", atomic_mode);
> +
> +	test_task = kthread_run(atomic_sect_run, NULL, task_name);
> +	return PTR_ERR_OR_ZERO(test_task);
> +}
> +
> +static void __exit atomic_sect_exit(void)
> +{
> +	return;
> +}
> +
> +module_init(atomic_sect_init)
> +module_exit(atomic_sect_exit)
> +MODULE_LICENSE("GPL v2");


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers
  2018-06-28 18:21 ` [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers Joel Fernandes
@ 2018-07-11  0:49   ` Steven Rostedt
  2018-07-11  5:27     ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11  0:49 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Thu, 28 Jun 2018 11:21:49 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> 
> Here we add unit tests for the preemptoff and irqsoff tracer by using a
> kernel module introduced previously to trigger atomic sections in the
> kernel.
> 
> Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
> Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>

This looks fine. The only patch that needs to be changed and resent is
patch 6 and 7. Just send 6, and this one again because it depends on
patch 6.

I'll go ahead and apply 1-5 and kick off my other tests.

Thanks!

-- Steve


> ---
>  tools/testing/selftests/ftrace/config         |  3 +
>  .../test.d/preemptirq/irqsoff_tracer.tc       | 73 +++++++++++++++++++
>  2 files changed, 76 insertions(+)
>  create mode 100644 tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
> 
> diff --git a/tools/testing/selftests/ftrace/config b/tools/testing/selftests/ftrace/config
> index b01924c71c09..29588b328345 100644
> --- a/tools/testing/selftests/ftrace/config
> +++ b/tools/testing/selftests/ftrace/config
> @@ -4,3 +4,6 @@ CONFIG_FUNCTION_PROFILER=y
>  CONFIG_TRACER_SNAPSHOT=y
>  CONFIG_STACK_TRACER=y
>  CONFIG_HIST_TRIGGERS=y
> +CONFIG_PREEMPT_TRACER=y
> +CONFIG_IRQSOFF_TRACER=y
> +CONFIG_TEST_ATOMIC_SECTIONS=m
> diff --git a/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc b/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
> new file mode 100644
> index 000000000000..1806d340035d
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test.d/preemptirq/irqsoff_tracer.tc
> @@ -0,0 +1,73 @@
> +#!/bin/sh
> +# SPDX-License-Identifier: GPL-2.0
> +# description: test for the preemptirqsoff tracer
> +
> +MOD=test_atomic_sections
> +
> +fail() {
> +    reset_tracer
> +    rmmod $MOD || true
> +    exit_fail
> +}
> +
> +unsup() { #msg
> +    reset_tracer
> +    rmmod $MOD || true
> +    echo $1
> +    exit_unsupported
> +}
> +
> +modprobe $MOD || unsup "$MOD module not available"
> +rmmod $MOD
> +
> +grep -q "preemptoff" available_tracers || unsup "preemptoff tracer not enabled"
> +grep -q "irqsoff" available_tracers || unsup "irqsoff tracer not enabled"
> +
> +reset_tracer
> +
> +# Simulate preemptoff section for half a second couple of times
> +echo preemptoff > current_tracer
> +sleep 1
> +modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
> +rmmod $MOD || fail
> +modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
> +rmmod $MOD || fail
> +modprobe $MOD atomic_mode=preempt atomic_time=500000 || fail
> +rmmod $MOD || fail
> +
> +cat trace
> +
> +# Confirm which tracer
> +grep -q "tracer: preemptoff" trace || fail
> +
> +# Check the end of the section
> +egrep -q "5.....us : <stack trace>" trace || fail
> +
> +# Check for 500ms of latency
> +egrep -q "latency: 5..... us" trace || fail
> +
> +reset_tracer
> +
> +# Simulate irqsoff section for half a second couple of times
> +echo irqsoff > current_tracer
> +sleep 1
> +modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
> +rmmod $MOD || fail
> +modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
> +rmmod $MOD || fail
> +modprobe $MOD atomic_mode=irq atomic_time=500000 || fail
> +rmmod $MOD || fail
> +
> +cat trace
> +
> +# Confirm which tracer
> +grep -q "tracer: irqsoff" trace || fail
> +
> +# Check the end of the section
> +egrep -q "5.....us : <stack trace>" trace || fail
> +
> +# Check for 500ms of latency
> +egrep -q "latency: 5..... us" trace || fail
> +
> +reset_tracer
> +exit 0


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers
  2018-07-11  0:47   ` Steven Rostedt
@ 2018-07-11  5:26     ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-11  5:26 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: linux-kernel, Andy Shevchenko, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, kernel-team, Namhyung Kim, Paul McKenney,
	Peter Zijlstra, Thomas Glexiner, Tom Zanussi

On Tue, Jul 10, 2018 at 08:47:07PM -0400, Steven Rostedt wrote:
> On Thu, 28 Jun 2018 11:21:48 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > 
> > In this patch we introduce a test module for simulating a long atomic
> > section in the kernel which the preemptoff or irqsoff tracers can
> > detect. This module is to be used only for test purposes and is default
> > disabled.
> > 
> > Following is the expected output (only briefly shown) that can be parsed
> > to verify that the tracers are working correctly. We will use this from
> > the kselftests in future patches.
> > 
> > For the preemptoff tracer:
> > 
> > echo preemptoff > /d/tracing/current_tracer
> > sleep 1
> > insmod ./test_atomic_sections.ko atomic_mode=preempt atomic_time=500000
> > sleep 1
> > bash-4.3# cat /d/tracing/trace
> > preempt -1066    2...2    0us@: atomic_sect_run <-atomic_sect_run
> > preempt -1066    2...2 500002us : atomic_sect_run <-atomic_sect_run
> > preempt -1066    2...2 500004us : tracer_preempt_on <-atomic_sect_run
> > preempt -1066    2...2 500012us : <stack trace>
> >  => kthread
> >  => ret_from_fork  
> > 
> > For the irqsoff tracer:
> > 
> > echo irqsoff > /d/tracing/current_tracer
> > sleep 1
> > insmod ./test_atomic_sections.ko atomic_mode=irq atomic_time=500000
> > sleep 1
> > bash-4.3# cat /d/tracing/trace
> > irq dis -1069    1d..1    0us@: atomic_sect_run
> > irq dis -1069    1d..1 500001us : atomic_sect_run
> > irq dis -1069    1d..1 500002us : tracer_hardirqs_on <-atomic_sect_run
> > irq dis -1069    1d..1 500005us : <stack trace>
> >  => ret_from_fork  
> > 
> > Co-developed-by: Erick Reyes <erickreyes@google.com>
> > Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > ---
> >  lib/Kconfig.debug          |  8 ++++
> >  lib/Makefile               |  1 +
> >  lib/test_atomic_sections.c | 77 ++++++++++++++++++++++++++++++++++++++
> 
> I think this code should reside in kernel/trace directory. I already
> have modules there. See the ring_buffer_benchmark code and the test
> module for mmio tracer.

Ok, I'll move it to there.

> >  3 files changed, 86 insertions(+)
> >  create mode 100644 lib/test_atomic_sections.c
> > 
> > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> > index 8838d1158d19..622c90e1e066 100644
> > --- a/lib/Kconfig.debug
> > +++ b/lib/Kconfig.debug
> > @@ -1956,6 +1956,14 @@ config TEST_KMOD
> >  
> >  	  If unsure, say N.
> >  
> > +config TEST_ATOMIC_SECTIONS
> > +	tristate "Simulate atomic sections for tracers to detect"
> 
> Hmm, I don't like this title. It's not very obvious to what it is
> about. What about "Preempt / IRQ disable delay thread to test latency
> tracers" ? Or something along those lines.

Sure, I'll change it to that. I agree its better. I'll change the text to
that and call the config TEST_PREEMPT_IRQ_DISABLE_DELAY.

> > +	depends on m
> > +	help
> > +	  Select this option to build a test module that can help test atomic
> > +	  sections by simulating them with a duration supplied as a module
> > +	  parameter. Preempt disable and irq disable modes can be requested.
> 
> "If unsure say N"

Sure, sounds good.

> > +
> >  config TEST_DEBUG_VIRTUAL
> >  	tristate "Test CONFIG_DEBUG_VIRTUAL feature"
> >  	depends on DEBUG_VIRTUAL
> > diff --git a/lib/Makefile b/lib/Makefile
> > index 90dc5520b784..7831e747bf72 100644
> > --- a/lib/Makefile
> > +++ b/lib/Makefile
> > @@ -44,6 +44,7 @@ obj-y += string_helpers.o
> >  obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.o
> >  obj-y += hexdump.o
> >  obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
> > +obj-$(CONFIG_TEST_ATOMIC_SECTIONS) += test_atomic_sections.o
> >  obj-y += kstrtox.o
> >  obj-$(CONFIG_FIND_BIT_BENCHMARK) += find_bit_benchmark.o
> >  obj-$(CONFIG_TEST_BPF) += test_bpf.o
> > diff --git a/lib/test_atomic_sections.c b/lib/test_atomic_sections.c
> > new file mode 100644
> > index 000000000000..1eef518f0974
> > --- /dev/null
> > +++ b/lib/test_atomic_sections.c
> > @@ -0,0 +1,77 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Atomic section emulation test module
> > + *
> > + * Emulates atomic sections by disabling IRQs or preemption
> > + * and doing a busy wait for a specified amount of time.
> > + * This can be used for testing of different atomic section
> > + * tracers such as irqsoff tracers.
> > + *
> > + * (c) 2018. Google LLC
> > + */
> > +
> > +#include <linux/delay.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/irq.h>
> > +#include <linux/kernel.h>
> > +#include <linux/kthread.h>
> > +#include <linux/ktime.h>
> > +#include <linux/module.h>
> > +#include <linux/printk.h>
> > +#include <linux/string.h>
> > +
> > +static ulong atomic_time = 100;
> > +static char atomic_mode[10] = "irq";
> > +
> > +module_param_named(atomic_time, atomic_time, ulong, S_IRUGO);
> > +module_param_string(atomic_mode, atomic_mode, 10, S_IRUGO);
> > +MODULE_PARM_DESC(atomic_time, "Period in microseconds (100 uS default)");
> 
> It's not a "Period", it's a delay. "Length of time in critical section"

Sure.

> > +MODULE_PARM_DESC(atomic_mode, "Mode of the test such as preempt or irq (default irq)");
> 
>   "Mode of the test: preempt or irq disabled (default irq)"

Ok.

> > +
> > +static void busy_wait(ulong time)
> > +{
> > +	ktime_t start, end;
> > +	start = ktime_get();
> > +	do {
> > +		end = ktime_get();
> > +		if (kthread_should_stop())
> > +			break;
> > +	} while (ktime_to_ns(ktime_sub(end, start)) < (time * 1000));
> > +}
> > +
> > +int atomic_sect_run(void *data)
> > +{
> > +	unsigned long flags;
> > +
> > +	if (!strcmp(atomic_mode, "irq")) {
> > +		local_irq_save(flags);
> > +		busy_wait(atomic_time);
> > +		local_irq_restore(flags);
> > +	} else if (!strcmp(atomic_mode, "preempt")) {
> > +		preempt_disable();
> > +		busy_wait(atomic_time);
> > +		preempt_enable();
> > +	}
> 
> So this is a one shot deal? That should be explained somewhere,
> probably in the config help message. In fact, I think the config help
> message should show how to use this.

Sounds good, I'll clarify it better. Thanks!

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers
  2018-07-11  0:49   ` Steven Rostedt
@ 2018-07-11  5:27     ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-11  5:27 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Thomas Glexiner, Tom Zanussi

On Tue, Jul 10, 2018 at 08:49:58PM -0400, Steven Rostedt wrote:
> On Thu, 28 Jun 2018 11:21:49 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > 
> > Here we add unit tests for the preemptoff and irqsoff tracer by using a
> > kernel module introduced previously to trigger atomic sections in the
> > kernel.
> > 
> > Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
> > Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
> > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> 
> This looks fine. The only patch that needs to be changed and resent is
> patch 6 and 7. Just send 6, and this one again because it depends on
> patch 6.
> 
> I'll go ahead and apply 1-5 and kick off my other tests.

Sounds good, I'll resend those shortly with the changes you suggested.

Thanks!

- Joel

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
@ 2018-07-11 12:49   ` Peter Zijlstra
  2018-07-11 13:00     ` Steven Rostedt
  2018-07-11 12:53   ` Peter Zijlstra
  2018-07-11 12:56   ` Peter Zijlstra
  2 siblings, 1 reply; 52+ messages in thread
From: Peter Zijlstra @ 2018-07-11 12:49 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\

I would convert to rcu_dereference_raw() to appease sparse. The fancy
stuff below is pointless if you then turn off all checking.

> +									\
> +		/*							\
> +		 * For rcuidle callers, use srcu since sched-rcu	\
> +		 * doesn't work from the idle path.			\
> +		 */							\
> +		if (rcuidle) {						\
> +			if (in_nmi()) {					\
> +				WARN_ON_ONCE(1);			\
> +				return; /* no srcu from nmi */		\
> +			}						\
> +									\
> +			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
> +			it_func_ptr =					\
> +				srcu_dereference_notrace((tp)->funcs,	\
> +						&tracepoint_srcu);	\
> +			/* To keep it consistent with !rcuidle path */	\
> +			preempt_disable_notrace();			\
> +		} else {						\
> +			rcu_read_lock_sched_notrace();			\
> +			it_func_ptr =					\
> +				rcu_dereference_sched((tp)->funcs);	\
> +		}							\

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
  2018-07-11 12:49   ` Peter Zijlstra
@ 2018-07-11 12:53   ` Peter Zijlstra
  2018-07-12  2:32     ` Joel Fernandes
  2018-07-11 12:56   ` Peter Zijlstra
  2 siblings, 1 reply; 52+ messages in thread
From: Peter Zijlstra @ 2018-07-11 12:53 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> -		rcu_read_lock_sched_notrace();				\
> -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\
> +									\
> +		/*							\
> +		 * For rcuidle callers, use srcu since sched-rcu	\
> +		 * doesn't work from the idle path.			\
> +		 */							\
> +		if (rcuidle) {						\
> +			if (in_nmi()) {					\
> +				WARN_ON_ONCE(1);			\
> +				return; /* no srcu from nmi */		\
> +			}						\
> +									\
> +			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
> +			it_func_ptr =					\
> +				srcu_dereference_notrace((tp)->funcs,	\
> +						&tracepoint_srcu);	\
> +			/* To keep it consistent with !rcuidle path */	\
> +			preempt_disable_notrace();			\
> +		} else {						\
> +			rcu_read_lock_sched_notrace();			\
> +			it_func_ptr =					\
> +				rcu_dereference_sched((tp)->funcs);	\
> +		}							\
> +									\
>  		if (it_func_ptr) {					\
>  			do {						\
>  				it_func = (it_func_ptr)->func;		\
> @@ -148,9 +177,13 @@ extern void syscall_unregfunc(void);
>  				((void(*)(proto))(it_func))(args);	\
>  			} while ((++it_func_ptr)->func);		\
>  		}							\
> -		rcu_read_unlock_sched_notrace();			\
> -		if (rcucheck)						\
> -			rcu_irq_exit_irqson();				\
> +									\
> +		if (rcuidle) {						\
> +			preempt_enable_notrace();			\
> +			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
> +		} else {						\
> +			rcu_read_unlock_sched_notrace();		\
> +		}							\
>  	} while (0)

In fact, I would write the thing like:

		preempt_disable_notrace();
		if (rcuidle)
			idx = srcu_read_lock_notrace(&tracepoint_srcu);

		it_func_ptr = rcu_dereference_raw((tp)->funcs);

		/* ... */

		if (rcu_idle)
			srcu_read_unlock_notrace(&tracepoint_srcu, idx);
		preempt_enable_notrace();

Much simpler and very much the same.

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
  2018-07-11 12:49   ` Peter Zijlstra
  2018-07-11 12:53   ` Peter Zijlstra
@ 2018-07-11 12:56   ` Peter Zijlstra
  2018-07-11 13:06     ` Steven Rostedt
  2 siblings, 1 reply; 52+ messages in thread
From: Peter Zijlstra @ 2018-07-11 12:56 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
>  static inline void tracepoint_synchronize_unregister(void)
>  {
> +	synchronize_srcu(&tracepoint_srcu);
>  	synchronize_sched();
>  }

Given you below do call_rcu_sched() and then call_srcu(), isn't the
above the wrong way around?

Also, does the above want to be barrier instead of synchronize, so as to
guarantee completion of the callbacks.

> +static void srcu_free_old_probes(struct rcu_head *head)
>  {
>  	kfree(container_of(head, struct tp_probes, rcu));
>  }
>  
> +static void rcu_free_old_probes(struct rcu_head *head)
> +{
> +	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
> +}
> +
>  static inline void release_probes(struct tracepoint_func *old)
>  {
>  	if (old) {
>  		struct tp_probes *tp_probes = container_of(old,
>  			struct tp_probes, probes[0]);
> +		/*
> +		 * Tracepoint probes are protected by both sched RCU and SRCU,
> +		 * by calling the SRCU callback in the sched RCU callback we
> +		 * cover both cases. So let us chain the SRCU and sched RCU
> +		 * callbacks to wait for both grace periods.
> +		 */
>  		call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
>  	}
>  }

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 12:49   ` Peter Zijlstra
@ 2018-07-11 13:00     ` Steven Rostedt
  2018-07-11 14:27       ` Paul E. McKenney
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 13:00 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 14:49:54 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\  
> 
> I would convert to rcu_dereference_raw() to appease sparse. The fancy
> stuff below is pointless if you then turn off all checking.

The problem with doing this is if we use a trace event without the
proper _idle() or whatever, we wont get a warning that it is used
incorrectly with lockdep. Or does lockdep still check if "rcu is
watching" with rcu_dereference_raw()?

-- Steve

> 
> > +									\
> > +		/*							\
> > +		 * For rcuidle callers, use srcu since sched-rcu	\
> > +		 * doesn't work from the idle path.			\
> > +		 */							\
> > +		if (rcuidle) {						\
> > +			if (in_nmi()) {					\
> > +				WARN_ON_ONCE(1);			\
> > +				return; /* no srcu from nmi */		\
> > +			}						\
> > +									\
> > +			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
> > +			it_func_ptr =					\
> > +				srcu_dereference_notrace((tp)->funcs,	\
> > +						&tracepoint_srcu);	\
> > +			/* To keep it consistent with !rcuidle path */	\
> > +			preempt_disable_notrace();			\
> > +		} else {						\
> > +			rcu_read_lock_sched_notrace();			\
> > +			it_func_ptr =					\
> > +				rcu_dereference_sched((tp)->funcs);	\
> > +		}							\  


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 12:56   ` Peter Zijlstra
@ 2018-07-11 13:06     ` Steven Rostedt
  2018-07-11 15:17       ` Peter Zijlstra
  2018-07-12  0:31       ` Joel Fernandes
  0 siblings, 2 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 13:06 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 14:56:47 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> >  static inline void tracepoint_synchronize_unregister(void)
> >  {
> > +	synchronize_srcu(&tracepoint_srcu);
> >  	synchronize_sched();
> >  }  
> 
> Given you below do call_rcu_sched() and then call_srcu(), isn't the
> above the wrong way around?

Good catch!

	release_probes()
		call_rcu_sched()
			---> rcu_free_old_probes() queued

	tracepoint_synchronize_unregister()
		synchronize_srcu(&tracepoint_srcu);
			< finishes right away >
		synchronize_sched()
			--> rcu_free_old_probes()
				--> srcu_free_old_probes() queued
	
Here tracepoint_synchronize_unregister() returned before the srcu
portion ran.


> 
> Also, does the above want to be barrier instead of synchronize, so as to
> guarantee completion of the callbacks.

Not sure what you mean here.

-- Steve

> 
> > +static void srcu_free_old_probes(struct rcu_head *head)
> >  {
> >  	kfree(container_of(head, struct tp_probes, rcu));
> >  }
> >  
> > +static void rcu_free_old_probes(struct rcu_head *head)
> > +{
> > +	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
> > +}
> > +
> >  static inline void release_probes(struct tracepoint_func *old)
> >  {
> >  	if (old) {
> >  		struct tp_probes *tp_probes = container_of(old,
> >  			struct tp_probes, probes[0]);
> > +		/*
> > +		 * Tracepoint probes are protected by both sched RCU and SRCU,
> > +		 * by calling the SRCU callback in the sched RCU callback we
> > +		 * cover both cases. So let us chain the SRCU and sched RCU
> > +		 * callbacks to wait for both grace periods.
> > +		 */
> >  		call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
> >  	}
> >  }  


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
  2018-07-06 22:06   ` Steven Rostedt
  2018-07-10 14:20   ` Steven Rostedt
@ 2018-07-11 13:12   ` Peter Zijlstra
  2018-07-11 13:19     ` Steven Rostedt
  2018-07-12  0:44     ` Joel Fernandes
  2 siblings, 2 replies; 52+ messages in thread
From: Peter Zijlstra @ 2018-07-11 13:12 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Thu, Jun 28, 2018 at 11:21:47AM -0700, Joel Fernandes wrote:
> One note, I have to check for lockdep recursion in the code that calls
> the trace events API and bail out if we're in lockdep recursion

I'm not seeing any new lockdep_recursion checks...

> protection to prevent something like the following case: a spin_lock is
> taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> this function, a call to get_lock_stats happens which calls
> preempt_disable, which calls trace IRQS off somewhere which enters my
> tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> This flag is then never cleared causing lockdep paths to never be
> entered and thus causing splats and other bad things.

Would it not be much easier to avoid that entirely, afaict all
get/put_lock_stats() callers already have IRQs disabled, so that
(traced) preempt fiddling is entirely superfluous.

---
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 5fa4d3138bf1..8f5ce0048d15 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -248,12 +248,7 @@ void clear_lock_stats(struct lock_class *class)
 
 static struct lock_class_stats *get_lock_stats(struct lock_class *class)
 {
-	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
-}
-
-static void put_lock_stats(struct lock_class_stats *stats)
-{
-	put_cpu_var(cpu_lock_stats);
+	return &this_cpu_ptr(&cpu_lock_stats)[class - lock_classes];
 }
 
 static void lock_release_holdtime(struct held_lock *hlock)
@@ -271,7 +266,6 @@ static void lock_release_holdtime(struct held_lock *hlock)
 		lock_time_inc(&stats->read_holdtime, holdtime);
 	else
 		lock_time_inc(&stats->write_holdtime, holdtime);
-	put_lock_stats(stats);
 }
 #else
 static inline void lock_release_holdtime(struct held_lock *hlock)
@@ -4090,7 +4084,6 @@ __lock_contended(struct lockdep_map *lock, unsigned long ip)
 		stats->contending_point[contending_point]++;
 	if (lock->cpu != smp_processor_id())
 		stats->bounces[bounce_contended + !!hlock->read]++;
-	put_lock_stats(stats);
 }
 
 static void
@@ -4138,7 +4131,6 @@ __lock_acquired(struct lockdep_map *lock, unsigned long ip)
 	}
 	if (lock->cpu != cpu)
 		stats->bounces[bounce_acquired + !!hlock->read]++;
-	put_lock_stats(stats);
 
 	lock->cpu = cpu;
 	lock->ip = ip;

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-11 13:12   ` Peter Zijlstra
@ 2018-07-11 13:19     ` Steven Rostedt
  2018-07-11 13:22       ` Steven Rostedt
  2018-07-12  8:38       ` Joel Fernandes
  2018-07-12  0:44     ` Joel Fernandes
  1 sibling, 2 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 13:19 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 15:12:56 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Thu, Jun 28, 2018 at 11:21:47AM -0700, Joel Fernandes wrote:
> > One note, I have to check for lockdep recursion in the code that calls
> > the trace events API and bail out if we're in lockdep recursion  
> 
> I'm not seeing any new lockdep_recursion checks...

I believe he's talking about this part:

+void trace_hardirqs_on(void)
+{
+	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
+		return;
+

[etc]

> 
> > protection to prevent something like the following case: a spin_lock is
> > taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> > and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> > this function, a call to get_lock_stats happens which calls
> > preempt_disable, which calls trace IRQS off somewhere which enters my
> > tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> > This flag is then never cleared causing lockdep paths to never be
> > entered and thus causing splats and other bad things.  
> 
> Would it not be much easier to avoid that entirely, afaict all
> get/put_lock_stats() callers already have IRQs disabled, so that
> (traced) preempt fiddling is entirely superfluous.

Agreed. Looks like a good clean up.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-11 13:19     ` Steven Rostedt
@ 2018-07-11 13:22       ` Steven Rostedt
  2018-07-12  8:38       ` Joel Fernandes
  1 sibling, 0 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 13:22 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 09:19:44 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Wed, 11 Jul 2018 15:12:56 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Thu, Jun 28, 2018 at 11:21:47AM -0700, Joel Fernandes wrote:  
> > > One note, I have to check for lockdep recursion in the code that calls
> > > the trace events API and bail out if we're in lockdep recursion    
> > 
> > I'm not seeing any new lockdep_recursion checks...  
> 
> I believe he's talking about this part:
> 
> +void trace_hardirqs_on(void)
> +{
> +	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
> +		return;
> +

And the reason he said this is new, IIUC, is because the old way we
could still do irqsoff tracing even if lockdep_recursion is set. Now,
irqsoff tracing is disable within lockdep_recursion.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 13:00     ` Steven Rostedt
@ 2018-07-11 14:27       ` Paul E. McKenney
  2018-07-11 14:46         ` Steven Rostedt
  0 siblings, 1 reply; 52+ messages in thread
From: Paul E. McKenney @ 2018-07-11 14:27 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Peter Zijlstra, Joel Fernandes, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 09:00:03AM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 14:49:54 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> > > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\  
> > 
> > I would convert to rcu_dereference_raw() to appease sparse. The fancy
> > stuff below is pointless if you then turn off all checking.
> 
> The problem with doing this is if we use a trace event without the
> proper _idle() or whatever, we wont get a warning that it is used
> incorrectly with lockdep. Or does lockdep still check if "rcu is
> watching" with rcu_dereference_raw()?

No lockdep checking is done by rcu_dereference_raw().

							Thanx, Paul

> -- Steve
> 
> > 
> > > +									\
> > > +		/*							\
> > > +		 * For rcuidle callers, use srcu since sched-rcu	\
> > > +		 * doesn't work from the idle path.			\
> > > +		 */							\
> > > +		if (rcuidle) {						\
> > > +			if (in_nmi()) {					\
> > > +				WARN_ON_ONCE(1);			\
> > > +				return; /* no srcu from nmi */		\
> > > +			}						\
> > > +									\
> > > +			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
> > > +			it_func_ptr =					\
> > > +				srcu_dereference_notrace((tp)->funcs,	\
> > > +						&tracepoint_srcu);	\
> > > +			/* To keep it consistent with !rcuidle path */	\
> > > +			preempt_disable_notrace();			\
> > > +		} else {						\
> > > +			rcu_read_lock_sched_notrace();			\
> > > +			it_func_ptr =					\
> > > +				rcu_dereference_sched((tp)->funcs);	\
> > > +		}							\  
> 


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 14:27       ` Paul E. McKenney
@ 2018-07-11 14:46         ` Steven Rostedt
  2018-07-11 15:15           ` Paul E. McKenney
  2018-07-11 20:52           ` Joel Fernandes
  0 siblings, 2 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 14:46 UTC (permalink / raw)
  To: Paul E. McKenney
  Cc: Peter Zijlstra, Joel Fernandes, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, 11 Jul 2018 07:27:44 -0700
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com> wrote:

> On Wed, Jul 11, 2018 at 09:00:03AM -0400, Steven Rostedt wrote:
> > On Wed, 11 Jul 2018 14:49:54 +0200
> > Peter Zijlstra <peterz@infradead.org> wrote:
> >   
> > > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:  
> > > > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\    
> > > 
> > > I would convert to rcu_dereference_raw() to appease sparse. The fancy
> > > stuff below is pointless if you then turn off all checking.  
> > 
> > The problem with doing this is if we use a trace event without the
> > proper _idle() or whatever, we wont get a warning that it is used
> > incorrectly with lockdep. Or does lockdep still check if "rcu is
> > watching" with rcu_dereference_raw()?  
> 
> No lockdep checking is done by rcu_dereference_raw().

Correct, but I think we can do this regardless. So Joel please resend
with Peter's suggestion.

The reason being is because of this:


#define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
	extern struct tracepoint __tracepoint_##name;			\
	static inline void trace_##name(proto)				\
	{								\
		if (static_key_false(&__tracepoint_##name.key))		\
			__DO_TRACE(&__tracepoint_##name,		\
				TP_PROTO(data_proto),			\
				TP_ARGS(data_args),			\
				TP_CONDITION(cond), 0);			\
		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
			rcu_read_lock_sched_notrace();			\
			rcu_dereference_sched(__tracepoint_##name.funcs);\
			rcu_read_unlock_sched_notrace();		\
		}							\
	}

Because lockdep would only trigger warnings when the tracepoint was
enabled and used in a place it shouldn't be, we added the above
IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
tracepoint was enabled or not. Because we do this, we don't need to
have the test in the __DO_TRACE() code itself. That means we can clean
up the code as per Peter's suggestion.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 14:46         ` Steven Rostedt
@ 2018-07-11 15:15           ` Paul E. McKenney
  2018-07-11 20:56             ` Joel Fernandes
  2018-07-11 20:52           ` Joel Fernandes
  1 sibling, 1 reply; 52+ messages in thread
From: Paul E. McKenney @ 2018-07-11 15:15 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Peter Zijlstra, Joel Fernandes, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 10:46:18AM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 07:27:44 -0700
> "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> wrote:
> 
> > On Wed, Jul 11, 2018 at 09:00:03AM -0400, Steven Rostedt wrote:
> > > On Wed, 11 Jul 2018 14:49:54 +0200
> > > Peter Zijlstra <peterz@infradead.org> wrote:
> > >   
> > > > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:  
> > > > > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\    
> > > > 
> > > > I would convert to rcu_dereference_raw() to appease sparse. The fancy
> > > > stuff below is pointless if you then turn off all checking.  
> > > 
> > > The problem with doing this is if we use a trace event without the
> > > proper _idle() or whatever, we wont get a warning that it is used
> > > incorrectly with lockdep. Or does lockdep still check if "rcu is
> > > watching" with rcu_dereference_raw()?  
> > 
> > No lockdep checking is done by rcu_dereference_raw().
> 
> Correct, but I think we can do this regardless. So Joel please resend
> with Peter's suggestion.
> 
> The reason being is because of this:
> 
> 
> #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> 	extern struct tracepoint __tracepoint_##name;			\
> 	static inline void trace_##name(proto)				\
> 	{								\
> 		if (static_key_false(&__tracepoint_##name.key))		\
> 			__DO_TRACE(&__tracepoint_##name,		\
> 				TP_PROTO(data_proto),			\
> 				TP_ARGS(data_args),			\
> 				TP_CONDITION(cond), 0);			\
> 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> 			rcu_read_lock_sched_notrace();			\
> 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> 			rcu_read_unlock_sched_notrace();		\
> 		}							\
> 	}
> 
> Because lockdep would only trigger warnings when the tracepoint was
> enabled and used in a place it shouldn't be, we added the above
> IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> tracepoint was enabled or not. Because we do this, we don't need to
> have the test in the __DO_TRACE() code itself. That means we can clean
> up the code as per Peter's suggestion.

Indeed, the rcu_dereference_sched() would catch it in that case, so
agreed, Peter's suggestion isn't losing any debuggability.

						Thanx, Paul


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 13:06     ` Steven Rostedt
@ 2018-07-11 15:17       ` Peter Zijlstra
  2018-07-11 15:26         ` Steven Rostedt
  2018-07-11 16:40         ` Mathieu Desnoyers
  2018-07-12  0:31       ` Joel Fernandes
  1 sibling, 2 replies; 52+ messages in thread
From: Peter Zijlstra @ 2018-07-11 15:17 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, Jul 11, 2018 at 09:06:49AM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 14:56:47 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> > >  static inline void tracepoint_synchronize_unregister(void)
> > >  {
> > > +	synchronize_srcu(&tracepoint_srcu);
> > >  	synchronize_sched();
> > >  }  
> > 
> > Given you below do call_rcu_sched() and then call_srcu(), isn't the
> > above the wrong way around?
> 
> Good catch!
> 
> 	release_probes()
> 		call_rcu_sched()
> 			---> rcu_free_old_probes() queued
> 
> 	tracepoint_synchronize_unregister()
> 		synchronize_srcu(&tracepoint_srcu);
> 			< finishes right away >
> 		synchronize_sched()
> 			--> rcu_free_old_probes()
> 				--> srcu_free_old_probes() queued
> 	
> Here tracepoint_synchronize_unregister() returned before the srcu
> portion ran.

I just read the comment that goes with that function; the order doesn't
matter. All we want to ensure is that the unregistration is visible to
either sched or srcu tracepoint users.

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 15:17       ` Peter Zijlstra
@ 2018-07-11 15:26         ` Steven Rostedt
  2018-07-11 16:46           ` Mathieu Desnoyers
  2018-07-11 16:40         ` Mathieu Desnoyers
  1 sibling, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-11 15:26 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Joel Fernandes, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 17:17:51 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> I just read the comment that goes with that function; the order doesn't
> matter. All we want to ensure is that the unregistration is visible to
> either sched or srcu tracepoint users.

Yeah, but I think it is still good to change the order. It doesn't
hurt, and in my opinion makes the code a bit more robust.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 15:17       ` Peter Zijlstra
  2018-07-11 15:26         ` Steven Rostedt
@ 2018-07-11 16:40         ` Mathieu Desnoyers
  1 sibling, 0 replies; 52+ messages in thread
From: Mathieu Desnoyers @ 2018-07-11 16:40 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: rostedt, Joel Fernandes, Google, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Namhyung Kim, Paul E. McKenney,
	Thomas Gleixner, Tom Zanussi

----- On Jul 11, 2018, at 11:17 AM, Peter Zijlstra peterz@infradead.org wrote:

> On Wed, Jul 11, 2018 at 09:06:49AM -0400, Steven Rostedt wrote:
>> On Wed, 11 Jul 2018 14:56:47 +0200
>> Peter Zijlstra <peterz@infradead.org> wrote:
>> 
>> > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
>> > >  static inline void tracepoint_synchronize_unregister(void)
>> > >  {
>> > > +	synchronize_srcu(&tracepoint_srcu);
>> > >  	synchronize_sched();
>> > >  }
>> > 
>> > Given you below do call_rcu_sched() and then call_srcu(), isn't the
>> > above the wrong way around?
>> 
>> Good catch!
>> 
>> 	release_probes()
>> 		call_rcu_sched()
>> 			---> rcu_free_old_probes() queued
>> 
>> 	tracepoint_synchronize_unregister()
>> 		synchronize_srcu(&tracepoint_srcu);
>> 			< finishes right away >
>> 		synchronize_sched()
>> 			--> rcu_free_old_probes()
>> 				--> srcu_free_old_probes() queued
>> 	
>> Here tracepoint_synchronize_unregister() returned before the srcu
>> portion ran.
> 
> I just read the comment that goes with that function; the order doesn't
> matter. All we want to ensure is that the unregistration is visible to
> either sched or srcu tracepoint users.

Exactly, the order does not matter here.

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 15:26         ` Steven Rostedt
@ 2018-07-11 16:46           ` Mathieu Desnoyers
  0 siblings, 0 replies; 52+ messages in thread
From: Mathieu Desnoyers @ 2018-07-11 16:46 UTC (permalink / raw)
  To: rostedt
  Cc: Peter Zijlstra, Joel Fernandes, Google, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Namhyung Kim, Paul E. McKenney,
	Thomas Gleixner, Tom Zanussi

----- On Jul 11, 2018, at 11:26 AM, rostedt rostedt@goodmis.org wrote:

> On Wed, 11 Jul 2018 17:17:51 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
>> I just read the comment that goes with that function; the order doesn't
>> matter. All we want to ensure is that the unregistration is visible to
>> either sched or srcu tracepoint users.
> 
> Yeah, but I think it is still good to change the order. It doesn't
> hurt, and in my opinion makes the code a bit more robust.

I don't mind. It makes the code more regular. It does not change
anything wrt robustness here though.

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 14:46         ` Steven Rostedt
  2018-07-11 15:15           ` Paul E. McKenney
@ 2018-07-11 20:52           ` Joel Fernandes
  2018-07-12  3:21             ` Steven Rostedt
  1 sibling, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-11 20:52 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 10:46:18AM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 07:27:44 -0700
> "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> wrote:
> 
> > On Wed, Jul 11, 2018 at 09:00:03AM -0400, Steven Rostedt wrote:
> > > On Wed, 11 Jul 2018 14:49:54 +0200
> > > Peter Zijlstra <peterz@infradead.org> wrote:
> > >   
> > > > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:  
> > > > > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\    
> > > > 
> > > > I would convert to rcu_dereference_raw() to appease sparse. The fancy
> > > > stuff below is pointless if you then turn off all checking.  
> > > 
> > > The problem with doing this is if we use a trace event without the
> > > proper _idle() or whatever, we wont get a warning that it is used
> > > incorrectly with lockdep. Or does lockdep still check if "rcu is
> > > watching" with rcu_dereference_raw()?  
> > 
> > No lockdep checking is done by rcu_dereference_raw().
> 
> Correct, but I think we can do this regardless. So Joel please resend
> with Peter's suggestion.
> 
> The reason being is because of this:
> 
> 
> #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> 	extern struct tracepoint __tracepoint_##name;			\
> 	static inline void trace_##name(proto)				\
> 	{								\
> 		if (static_key_false(&__tracepoint_##name.key))		\
> 			__DO_TRACE(&__tracepoint_##name,		\
> 				TP_PROTO(data_proto),			\
> 				TP_ARGS(data_args),			\
> 				TP_CONDITION(cond), 0);			\
> 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> 			rcu_read_lock_sched_notrace();			\
> 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> 			rcu_read_unlock_sched_notrace();		\
> 		}							\
> 	}
> 
> Because lockdep would only trigger warnings when the tracepoint was
> enabled and used in a place it shouldn't be, we added the above
> IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> tracepoint was enabled or not. Because we do this, we don't need to
> have the test in the __DO_TRACE() code itself. That means we can clean
> up the code as per Peter's suggestion.

Sounds good, I'm Ok with making this change.

Just to clarify, are you proposing to change the rcu_dereference_sched to
rcu_dereference_raw in both __DECLARE_TRACE and __DO_TRACE?

thanks!

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 15:15           ` Paul E. McKenney
@ 2018-07-11 20:56             ` Joel Fernandes
  2018-07-12  1:22               ` Steven Rostedt
  0 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-11 20:56 UTC (permalink / raw)
  To: Paul E. McKenney
  Cc: Steven Rostedt, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 08:15:59AM -0700, Paul E. McKenney wrote:
> On Wed, Jul 11, 2018 at 10:46:18AM -0400, Steven Rostedt wrote:
> > On Wed, 11 Jul 2018 07:27:44 -0700
> > "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> wrote:
> > 
> > > On Wed, Jul 11, 2018 at 09:00:03AM -0400, Steven Rostedt wrote:
> > > > On Wed, 11 Jul 2018 14:49:54 +0200
> > > > Peter Zijlstra <peterz@infradead.org> wrote:
> > > >   
> > > > > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:  
> > > > > > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\    
> > > > > 
> > > > > I would convert to rcu_dereference_raw() to appease sparse. The fancy
> > > > > stuff below is pointless if you then turn off all checking.  
> > > > 
> > > > The problem with doing this is if we use a trace event without the
> > > > proper _idle() or whatever, we wont get a warning that it is used
> > > > incorrectly with lockdep. Or does lockdep still check if "rcu is
> > > > watching" with rcu_dereference_raw()?  
> > > 
> > > No lockdep checking is done by rcu_dereference_raw().
> > 
> > Correct, but I think we can do this regardless. So Joel please resend
> > with Peter's suggestion.
> > 
> > The reason being is because of this:
> > 
> > 
> > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > 	extern struct tracepoint __tracepoint_##name;			\
> > 	static inline void trace_##name(proto)				\
> > 	{								\
> > 		if (static_key_false(&__tracepoint_##name.key))		\
> > 			__DO_TRACE(&__tracepoint_##name,		\
> > 				TP_PROTO(data_proto),			\
> > 				TP_ARGS(data_args),			\
> > 				TP_CONDITION(cond), 0);			\
> > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > 			rcu_read_lock_sched_notrace();			\
> > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > 			rcu_read_unlock_sched_notrace();		\
> > 		}							\
> > 	}
> > 
> > Because lockdep would only trigger warnings when the tracepoint was
> > enabled and used in a place it shouldn't be, we added the above
> > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > tracepoint was enabled or not. Because we do this, we don't need to
> > have the test in the __DO_TRACE() code itself. That means we can clean
> > up the code as per Peter's suggestion.
> 
> Indeed, the rcu_dereference_sched() would catch it in that case, so
> agreed, Peter's suggestion isn't losing any debuggability.

Hmm, but if we are doing the check later anyway, then why not do it in
__DO_TRACE itself?

Also I guess we are discussing about changing the rcu_dereference_sched which
I think should go into a separate patch since my patch isn't touching how the
rcuidle==0 paths use the RCU API. So I think this is an existing issue
independent of this series.

thanks!

- Joel






^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 13:06     ` Steven Rostedt
  2018-07-11 15:17       ` Peter Zijlstra
@ 2018-07-12  0:31       ` Joel Fernandes
  2018-07-12  1:26         ` Steven Rostedt
  1 sibling, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  0:31 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Peter Zijlstra, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, Jul 11, 2018 at 09:06:49AM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 14:56:47 +0200
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> > >  static inline void tracepoint_synchronize_unregister(void)
> > >  {
> > > +	synchronize_srcu(&tracepoint_srcu);
> > >  	synchronize_sched();
> > >  }  
> > 
> > Given you below do call_rcu_sched() and then call_srcu(), isn't the
> > above the wrong way around?
> 
> Good catch!
> 
> 	release_probes()
> 		call_rcu_sched()
> 			---> rcu_free_old_probes() queued
> 
> 	tracepoint_synchronize_unregister()
> 		synchronize_srcu(&tracepoint_srcu);
> 			< finishes right away >
> 		synchronize_sched()
> 			--> rcu_free_old_probes()
> 				--> srcu_free_old_probes() queued
> 	
> Here tracepoint_synchronize_unregister() returned before the srcu
> portion ran.

But isn't the point of synchronize_rcu to make sure that we're no longer in
an RCU read-side section, not that *all* queued callbacks already ran? So in that
case, I think it doesn't matter which order the 2 synchronize functions are
called in. Please let me know if if I missed something!

I believe what we're trying to guarantee here is that no tracepoints using
either flavor of RCU are active after tracepoint_synchronize_unregister
returns.

thanks!

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-11 13:12   ` Peter Zijlstra
  2018-07-11 13:19     ` Steven Rostedt
@ 2018-07-12  0:44     ` Joel Fernandes
  1 sibling, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  0:44 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 03:12:56PM +0200, Peter Zijlstra wrote:
> On Thu, Jun 28, 2018 at 11:21:47AM -0700, Joel Fernandes wrote:
> > One note, I have to check for lockdep recursion in the code that calls
> > the trace events API and bail out if we're in lockdep recursion
> 
> I'm not seeing any new lockdep_recursion checks...

I meant its this part:

void trace_hardirqs_on(void)
{
	if (lockdep_recursing(current) || !this_cpu_read(tracing_irq_cpu))
		return;

> > protection to prevent something like the following case: a spin_lock is
> > taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> > and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> > this function, a call to get_lock_stats happens which calls
> > preempt_disable, which calls trace IRQS off somewhere which enters my
> > tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> > This flag is then never cleared causing lockdep paths to never be
> > entered and thus causing splats and other bad things.
> 
> Would it not be much easier to avoid that entirely, afaict all
> get/put_lock_stats() callers already have IRQs disabled, so that
> (traced) preempt fiddling is entirely superfluous.

Let me try to apply Peter's diff and see if I still don't need lockdep
recursion checking. I believe it is still harmless to still check for lockdep
recursion just to be safe. But I'll give it a try and let you know how it
goes.

thanks!

- Joel


> ---
> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> index 5fa4d3138bf1..8f5ce0048d15 100644
> --- a/kernel/locking/lockdep.c
> +++ b/kernel/locking/lockdep.c
> @@ -248,12 +248,7 @@ void clear_lock_stats(struct lock_class *class)
>  
>  static struct lock_class_stats *get_lock_stats(struct lock_class *class)
>  {
> -	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
> -}
> -
> -static void put_lock_stats(struct lock_class_stats *stats)
> -{
> -	put_cpu_var(cpu_lock_stats);
> +	return &this_cpu_ptr(&cpu_lock_stats)[class - lock_classes];
>  }
>  
>  static void lock_release_holdtime(struct held_lock *hlock)
> @@ -271,7 +266,6 @@ static void lock_release_holdtime(struct held_lock *hlock)
>  		lock_time_inc(&stats->read_holdtime, holdtime);
>  	else
>  		lock_time_inc(&stats->write_holdtime, holdtime);
> -	put_lock_stats(stats);
>  }
>  #else
>  static inline void lock_release_holdtime(struct held_lock *hlock)
> @@ -4090,7 +4084,6 @@ __lock_contended(struct lockdep_map *lock, unsigned long ip)
>  		stats->contending_point[contending_point]++;
>  	if (lock->cpu != smp_processor_id())
>  		stats->bounces[bounce_contended + !!hlock->read]++;
> -	put_lock_stats(stats);
>  }
>  
>  static void
> @@ -4138,7 +4131,6 @@ __lock_acquired(struct lockdep_map *lock, unsigned long ip)
>  	}
>  	if (lock->cpu != cpu)
>  		stats->bounces[bounce_acquired + !!hlock->read]++;
> -	put_lock_stats(stats);
>  
>  	lock->cpu = cpu;
>

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 20:56             ` Joel Fernandes
@ 2018-07-12  1:22               ` Steven Rostedt
  2018-07-12  2:35                 ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12  1:22 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, 11 Jul 2018 13:56:39 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> > > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > > 	extern struct tracepoint __tracepoint_##name;			\
> > > 	static inline void trace_##name(proto)				\
> > > 	{								\
> > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > 				TP_PROTO(data_proto),			\
> > > 				TP_ARGS(data_args),			\
> > > 				TP_CONDITION(cond), 0);			\
> > > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > 			rcu_read_lock_sched_notrace();			\
> > > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > > 			rcu_read_unlock_sched_notrace();		\
> > > 		}							\
> > > 	}
> > > 
> > > Because lockdep would only trigger warnings when the tracepoint was
> > > enabled and used in a place it shouldn't be, we added the above
> > > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > > tracepoint was enabled or not. Because we do this, we don't need to
> > > have the test in the __DO_TRACE() code itself. That means we can clean
> > > up the code as per Peter's suggestion.  
> > 
> > Indeed, the rcu_dereference_sched() would catch it in that case, so
> > agreed, Peter's suggestion isn't losing any debuggability.  
> 
> Hmm, but if we are doing the check later anyway, then why not do it in
> __DO_TRACE itself?

Because __DO_TRACE is only called if the trace event is enabled. If we
never enable a trace event, we never know if it has a potential of
doing it wrong. The second part is to trigger the warning immediately
regardless if the trace event is enabled or not.

> 
> Also I guess we are discussing about changing the rcu_dereference_sched which
> I think should go into a separate patch since my patch isn't touching how the
> rcuidle==0 paths use the RCU API. So I think this is an existing issue
> independent of this series.

But the code you added made it much more complex to keep the checks as
is. If we remove the checks then this patch doesn't need to have all
the if statements, and we can do it the way Peter suggested.

But sure, go ahead and make a separate patch first that removes the
checks from __DO_TRACE() first if you want to.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12  0:31       ` Joel Fernandes
@ 2018-07-12  1:26         ` Steven Rostedt
  0 siblings, 0 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12  1:26 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Peter Zijlstra, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi

On Wed, 11 Jul 2018 17:31:00 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> On Wed, Jul 11, 2018 at 09:06:49AM -0400, Steven Rostedt wrote:
> > On Wed, 11 Jul 2018 14:56:47 +0200
> > Peter Zijlstra <peterz@infradead.org> wrote:
> >   
> > > On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:  
> > > >  static inline void tracepoint_synchronize_unregister(void)
> > > >  {
> > > > +	synchronize_srcu(&tracepoint_srcu);
> > > >  	synchronize_sched();
> > > >  }    
> > > 
> > > Given you below do call_rcu_sched() and then call_srcu(), isn't the
> > > above the wrong way around?  
> > 
> > Good catch!
> > 
> > 	release_probes()
> > 		call_rcu_sched()  
> > 			---> rcu_free_old_probes() queued  
> > 
> > 	tracepoint_synchronize_unregister()
> > 		synchronize_srcu(&tracepoint_srcu);
> > 			< finishes right away >
> > 		synchronize_sched()  
> > 			--> rcu_free_old_probes()
> > 				--> srcu_free_old_probes() queued  
> > 	
> > Here tracepoint_synchronize_unregister() returned before the srcu
> > portion ran.  
> 
> But isn't the point of synchronize_rcu to make sure that we're no longer in
> an RCU read-side section, not that *all* queued callbacks already ran? So in that
> case, I think it doesn't matter which order the 2 synchronize functions are
> called in. Please let me know if if I missed something!
> 
> I believe what we're trying to guarantee here is that no tracepoints using
> either flavor of RCU are active after tracepoint_synchronize_unregister
> returns.

Yes you are correct. If tracepoint_synchronize_unregister() is only to
make sure that there is no more trace events using the probes, then
this should work. I was focused on looking at it with release_probes()
too. So the patch is fine as is.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 12:53   ` Peter Zijlstra
@ 2018-07-12  2:32     ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  2:32 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: linux-kernel, Boqun Feng, Byungchul Park, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Steven Rostedt,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 02:53:22PM +0200, Peter Zijlstra wrote:
> On Thu, Jun 28, 2018 at 11:21:46AM -0700, Joel Fernandes wrote:
> > -		rcu_read_lock_sched_notrace();				\
> > -		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\
> > +									\
> > +		/*							\
> > +		 * For rcuidle callers, use srcu since sched-rcu	\
> > +		 * doesn't work from the idle path.			\
> > +		 */							\
> > +		if (rcuidle) {						\
> > +			if (in_nmi()) {					\
> > +				WARN_ON_ONCE(1);			\
> > +				return; /* no srcu from nmi */		\
> > +			}						\
> > +									\
> > +			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
> > +			it_func_ptr =					\
> > +				srcu_dereference_notrace((tp)->funcs,	\
> > +						&tracepoint_srcu);	\
> > +			/* To keep it consistent with !rcuidle path */	\
> > +			preempt_disable_notrace();			\
> > +		} else {						\
> > +			rcu_read_lock_sched_notrace();			\
> > +			it_func_ptr =					\
> > +				rcu_dereference_sched((tp)->funcs);	\
> > +		}							\
> > +									\
> >  		if (it_func_ptr) {					\
> >  			do {						\
> >  				it_func = (it_func_ptr)->func;		\
> > @@ -148,9 +177,13 @@ extern void syscall_unregfunc(void);
> >  				((void(*)(proto))(it_func))(args);	\
> >  			} while ((++it_func_ptr)->func);		\
> >  		}							\
> > -		rcu_read_unlock_sched_notrace();			\
> > -		if (rcucheck)						\
> > -			rcu_irq_exit_irqson();				\
> > +									\
> > +		if (rcuidle) {						\
> > +			preempt_enable_notrace();			\
> > +			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
> > +		} else {						\
> > +			rcu_read_unlock_sched_notrace();		\
> > +		}							\
> >  	} while (0)
> 
> In fact, I would write the thing like:
> 
> 		preempt_disable_notrace();
> 		if (rcuidle)
> 			idx = srcu_read_lock_notrace(&tracepoint_srcu);
> 
> 		it_func_ptr = rcu_dereference_raw((tp)->funcs);
> 
> 		/* ... */
> 
> 		if (rcu_idle)
> 			srcu_read_unlock_notrace(&tracepoint_srcu, idx);
> 		preempt_enable_notrace();
> 
> Much simpler and very much the same.

Cool, thanks! I will do it this way and resend.

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12  1:22               ` Steven Rostedt
@ 2018-07-12  2:35                 ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  2:35 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 09:22:37PM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 13:56:39 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > > > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > > > 	extern struct tracepoint __tracepoint_##name;			\
> > > > 	static inline void trace_##name(proto)				\
> > > > 	{								\
> > > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > > 				TP_PROTO(data_proto),			\
> > > > 				TP_ARGS(data_args),			\
> > > > 				TP_CONDITION(cond), 0);			\
> > > > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > > 			rcu_read_lock_sched_notrace();			\
> > > > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > > > 			rcu_read_unlock_sched_notrace();		\
> > > > 		}							\
> > > > 	}
> > > > 
> > > > Because lockdep would only trigger warnings when the tracepoint was
> > > > enabled and used in a place it shouldn't be, we added the above
> > > > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > > > tracepoint was enabled or not. Because we do this, we don't need to
> > > > have the test in the __DO_TRACE() code itself. That means we can clean
> > > > up the code as per Peter's suggestion.  
> > > 
> > > Indeed, the rcu_dereference_sched() would catch it in that case, so
> > > agreed, Peter's suggestion isn't losing any debuggability.  
> > 
> > Hmm, but if we are doing the check later anyway, then why not do it in
> > __DO_TRACE itself?
> 
> Because __DO_TRACE is only called if the trace event is enabled. If we
> never enable a trace event, we never know if it has a potential of
> doing it wrong. The second part is to trigger the warning immediately
> regardless if the trace event is enabled or not.

I see, thanks for the clarification.

> > 
> > Also I guess we are discussing about changing the rcu_dereference_sched which
> > I think should go into a separate patch since my patch isn't touching how the
> > rcuidle==0 paths use the RCU API. So I think this is an existing issue
> > independent of this series.
> 
> But the code you added made it much more complex to keep the checks as
> is. If we remove the checks then this patch doesn't need to have all
> the if statements, and we can do it the way Peter suggested.

Yes, I agree Peter's suggestion is very clean.

> But sure, go ahead and make a separate patch first that removes the
> checks from __DO_TRACE() first if you want to.

No its ok, no problem, I can just do it in the same patch now that I see the code is much simplified with what Peter is suggesting.

thanks!

- Joel



^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-11 20:52           ` Joel Fernandes
@ 2018-07-12  3:21             ` Steven Rostedt
  2018-07-12  4:28               ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12  3:21 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, 11 Jul 2018 13:52:49 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > 	extern struct tracepoint __tracepoint_##name;			\
> > 	static inline void trace_##name(proto)				\
> > 	{								\
> > 		if (static_key_false(&__tracepoint_##name.key))		\
> > 			__DO_TRACE(&__tracepoint_##name,		\
> > 				TP_PROTO(data_proto),			\
> > 				TP_ARGS(data_args),			\
> > 				TP_CONDITION(cond), 0);			\
> > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > 			rcu_read_lock_sched_notrace();			\
> > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > 			rcu_read_unlock_sched_notrace();		\
> > 		}							\
> > 	}
> > 
> > Because lockdep would only trigger warnings when the tracepoint was
> > enabled and used in a place it shouldn't be, we added the above
> > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > tracepoint was enabled or not. Because we do this, we don't need to
> > have the test in the __DO_TRACE() code itself. That means we can clean
> > up the code as per Peter's suggestion.  
> 
> Sounds good, I'm Ok with making this change.
> 
> Just to clarify, are you proposing to change the rcu_dereference_sched to
> rcu_dereference_raw in both __DECLARE_TRACE and __DO_TRACE?

No, just in __DO_TRACE(). The rcu_dereference_sched() above in
__DECLARE_TRACE() in the if (IS_ENABLED(CONFIG_LOCKDEP) block is
required to show the warnings if trace_##name() is used wrong, and is
the reason we can use rcu_dereference_raw() in __DO_TRACE() in the
first place ;-)

This brings up another point. We should probably add to
__DECLARE_TRACE_RCU() this:

#ifndef MODULE
#define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \
	static inline void trace_##name##_rcuidle(proto)		\
	{								\
		if (static_key_false(&__tracepoint_##name.key))		\
			__DO_TRACE(&__tracepoint_##name,		\
				TP_PROTO(data_proto),			\
				TP_ARGS(data_args),			\
				TP_CONDITION(cond), 1);			\
+		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
+			int idx;					\
+			idx = srcu_read_lock_notrace(&tracepoint_srcu); \
+			srcu_dereference_notrace(__tracepoint_##name.funcs, \
+						&tracepoint_srcu);	\
+			srcu_read_unlock_notrace(&tracepoint_srcu, idx); \
+		}							\
	}
#else


So that lockdep works with trace_##name##__rcuidle() when the trace
event is not enabled.

But that should be a separate patch and not part of this series. I may
write that up tomorrow.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12  3:21             ` Steven Rostedt
@ 2018-07-12  4:28               ` Joel Fernandes
  2018-07-12 13:35                 ` Steven Rostedt
  0 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  4:28 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, Jul 11, 2018 at 11:21:20PM -0400, Steven Rostedt wrote:
> On Wed, 11 Jul 2018 13:52:49 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > > 	extern struct tracepoint __tracepoint_##name;			\
> > > 	static inline void trace_##name(proto)				\
> > > 	{								\
> > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > 				TP_PROTO(data_proto),			\
> > > 				TP_ARGS(data_args),			\
> > > 				TP_CONDITION(cond), 0);			\
> > > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > 			rcu_read_lock_sched_notrace();			\
> > > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > > 			rcu_read_unlock_sched_notrace();		\
> > > 		}							\
> > > 	}
> > > 
> > > Because lockdep would only trigger warnings when the tracepoint was
> > > enabled and used in a place it shouldn't be, we added the above
> > > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > > tracepoint was enabled or not. Because we do this, we don't need to
> > > have the test in the __DO_TRACE() code itself. That means we can clean
> > > up the code as per Peter's suggestion.  
> > 
> > Sounds good, I'm Ok with making this change.
> > 
> > Just to clarify, are you proposing to change the rcu_dereference_sched to
> > rcu_dereference_raw in both __DECLARE_TRACE and __DO_TRACE?
> 
> No, just in __DO_TRACE(). The rcu_dereference_sched() above in
> __DECLARE_TRACE() in the if (IS_ENABLED(CONFIG_LOCKDEP) block is
> required to show the warnings if trace_##name() is used wrong, and is
> the reason we can use rcu_dereference_raw() in __DO_TRACE() in the
> first place ;-)
> 
> This brings up another point. We should probably add to
> __DECLARE_TRACE_RCU() this:
> 
> #ifndef MODULE
> #define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \
> 	static inline void trace_##name##_rcuidle(proto)		\
> 	{								\
> 		if (static_key_false(&__tracepoint_##name.key))		\
> 			__DO_TRACE(&__tracepoint_##name,		\
> 				TP_PROTO(data_proto),			\
> 				TP_ARGS(data_args),			\
> 				TP_CONDITION(cond), 1);			\
> +		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> +			int idx;					\
> +			idx = srcu_read_lock_notrace(&tracepoint_srcu); \
> +			srcu_dereference_notrace(__tracepoint_##name.funcs, \
> +						&tracepoint_srcu);	\
> +			srcu_read_unlock_notrace(&tracepoint_srcu, idx); \
> +		}							\
> 	}
> #else
> 
> 
> So that lockdep works with trace_##name##__rcuidle() when the trace
> event is not enabled.
> 
> But that should be a separate patch and not part of this series. I may
> write that up tomorrow.

Yes, that sounds good to me and would be good to add the safe guard there.
But you meant srcu_dereference above, not srcu_dereference_notrace right?

Meanwhile I'll drop that lockdep_recursion tomorrow and run some tests and
see how it behaves with Peter's changes.

thanks!

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-11 13:19     ` Steven Rostedt
  2018-07-11 13:22       ` Steven Rostedt
@ 2018-07-12  8:38       ` Joel Fernandes
  2018-07-12 13:37         ` Steven Rostedt
  1 sibling, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12  8:38 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Peter Zijlstra, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi, kernel-team

On Wed, Jul 11, 2018 at 09:19:44AM -0400, Steven Rostedt wrote:
> > > protection to prevent something like the following case: a spin_lock is
> > > taken. Then lockdep_acquired is called.  That does a raw_local_irq_save
> > > and then sets lockdep_recursion, and then calls __lockdep_acquired. In
> > > this function, a call to get_lock_stats happens which calls
> > > preempt_disable, which calls trace IRQS off somewhere which enters my
> > > tracepoint code and sets the tracing_irq_cpu flag to prevent recursion.
> > > This flag is then never cleared causing lockdep paths to never be
> > > entered and thus causing splats and other bad things.  
> > 
> > Would it not be much easier to avoid that entirely, afaict all
> > get/put_lock_stats() callers already have IRQs disabled, so that
> > (traced) preempt fiddling is entirely superfluous.
> 
> Agreed. Looks like a good clean up.

So actually with or without the clean up, I don't see any issues with
dropping lockdep_recursing in my tests at the moment. I'm not sure something
else changed between then and now causing the issue to go away. I can include
Peter's clean up in my series though if he's Ok with it since you guys agree
its a good clean up anyway. Would you prefer I did that, and then also
dropped the lockdep_recursing checks? Or should I keep the
lockdep_recursing() checks just to be safe? Do you see cases where you want
irqsoff tracing while lockdep_recursing() is true?

thanks,

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12  4:28               ` Joel Fernandes
@ 2018-07-12 13:35                 ` Steven Rostedt
  2018-07-12 19:17                   ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12 13:35 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Wed, 11 Jul 2018 21:28:25 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> On Wed, Jul 11, 2018 at 11:21:20PM -0400, Steven Rostedt wrote:
> > On Wed, 11 Jul 2018 13:52:49 -0700
> > Joel Fernandes <joel@joelfernandes.org> wrote:
> >   
> > > > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > > > 	extern struct tracepoint __tracepoint_##name;			\
> > > > 	static inline void trace_##name(proto)				\
> > > > 	{								\
> > > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > > 				TP_PROTO(data_proto),			\
> > > > 				TP_ARGS(data_args),			\
> > > > 				TP_CONDITION(cond), 0);			\
> > > > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > > 			rcu_read_lock_sched_notrace();			\
> > > > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > > > 			rcu_read_unlock_sched_notrace();		\
> > > > 		}							\
> > > > 	}
> > > > 
> > > > Because lockdep would only trigger warnings when the tracepoint was
> > > > enabled and used in a place it shouldn't be, we added the above
> > > > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > > > tracepoint was enabled or not. Because we do this, we don't need to
> > > > have the test in the __DO_TRACE() code itself. That means we can clean
> > > > up the code as per Peter's suggestion.    
> > > 
> > > Sounds good, I'm Ok with making this change.
> > > 
> > > Just to clarify, are you proposing to change the rcu_dereference_sched to
> > > rcu_dereference_raw in both __DECLARE_TRACE and __DO_TRACE?  
> > 
> > No, just in __DO_TRACE(). The rcu_dereference_sched() above in
> > __DECLARE_TRACE() in the if (IS_ENABLED(CONFIG_LOCKDEP) block is
> > required to show the warnings if trace_##name() is used wrong, and is
> > the reason we can use rcu_dereference_raw() in __DO_TRACE() in the
> > first place ;-)
> > 
> > This brings up another point. We should probably add to
> > __DECLARE_TRACE_RCU() this:
> > 
> > #ifndef MODULE
> > #define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \
> > 	static inline void trace_##name##_rcuidle(proto)		\
> > 	{								\
> > 		if (static_key_false(&__tracepoint_##name.key))		\
> > 			__DO_TRACE(&__tracepoint_##name,		\
> > 				TP_PROTO(data_proto),			\
> > 				TP_ARGS(data_args),			\
> > 				TP_CONDITION(cond), 1);			\
> > +		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > +			int idx;					\
> > +			idx = srcu_read_lock_notrace(&tracepoint_srcu); \
> > +			srcu_dereference_notrace(__tracepoint_##name.funcs, \
> > +						&tracepoint_srcu);	\
> > +			srcu_read_unlock_notrace(&tracepoint_srcu, idx); \
> > +		}							\
> > 	}
> > #else
> > 
> > 
> > So that lockdep works with trace_##name##__rcuidle() when the trace
> > event is not enabled.
> > 
> > But that should be a separate patch and not part of this series. I may
> > write that up tomorrow.  
> 
> Yes, that sounds good to me and would be good to add the safe guard there.
> But you meant srcu_dereference above, not srcu_dereference_notrace right?

We don't need to trace them. I believe that the "srcu_*_notrace" still
performs the lockdep checks. That's what we want. If they don't then we
should not use notrace. But I believe they still do lockdep.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage
  2018-07-12  8:38       ` Joel Fernandes
@ 2018-07-12 13:37         ` Steven Rostedt
  0 siblings, 0 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12 13:37 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Peter Zijlstra, linux-kernel, Boqun Feng, Byungchul Park,
	Ingo Molnar, Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Thomas Glexiner,
	Tom Zanussi, kernel-team

On Thu, 12 Jul 2018 01:38:05 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> So actually with or without the clean up, I don't see any issues with
> dropping lockdep_recursing in my tests at the moment. I'm not sure something
> else changed between then and now causing the issue to go away. I can include
> Peter's clean up in my series though if he's Ok with it since you guys agree
> its a good clean up anyway. Would you prefer I did that, and then also
> dropped the lockdep_recursing checks? Or should I keep the
> lockdep_recursing() checks just to be safe? Do you see cases where you want
> irqsoff tracing while lockdep_recursing() is true?

I say rewrite it as per Peter's suggestion. Perhaps even add credit to
Peter like:

Cleaned-up-code-by: Peter Zijlstra <peterz@infradead.org>

  ;-)

And yes, I would recommend dropping the lockdep_recursion() if you
can't trigger issues from within your tests. If it shows up later, we
can always add it back.

Thanks!

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12 13:35                 ` Steven Rostedt
@ 2018-07-12 19:17                   ` Joel Fernandes
  2018-07-12 20:15                     ` Steven Rostedt
  0 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12 19:17 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Thu, Jul 12, 2018 at 09:35:12AM -0400, Steven Rostedt wrote:

> On Wed, 11 Jul 2018 21:28:25 -0700
> Joel Fernandes <joel@joelfernandes.org> wrote:
> 
> > On Wed, Jul 11, 2018 at 11:21:20PM -0400, Steven Rostedt wrote:
> > > On Wed, 11 Jul 2018 13:52:49 -0700
> > > Joel Fernandes <joel@joelfernandes.org> wrote:
> > >   
> > > > > #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \
> > > > > 	extern struct tracepoint __tracepoint_##name;			\
> > > > > 	static inline void trace_##name(proto)				\
> > > > > 	{								\
> > > > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > > > 				TP_PROTO(data_proto),			\
> > > > > 				TP_ARGS(data_args),			\
> > > > > 				TP_CONDITION(cond), 0);			\
> > > > > 		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > > > 			rcu_read_lock_sched_notrace();			\
> > > > > 			rcu_dereference_sched(__tracepoint_##name.funcs);\
> > > > > 			rcu_read_unlock_sched_notrace();		\
> > > > > 		}							\
> > > > > 	}
> > > > > 
> > > > > Because lockdep would only trigger warnings when the tracepoint was
> > > > > enabled and used in a place it shouldn't be, we added the above
> > > > > IS_ENABLED(CONFIG_LOCKDEP) part to test regardless if the the
> > > > > tracepoint was enabled or not. Because we do this, we don't need to
> > > > > have the test in the __DO_TRACE() code itself. That means we can clean
> > > > > up the code as per Peter's suggestion.    
> > > > 
> > > > Sounds good, I'm Ok with making this change.
> > > > 
> > > > Just to clarify, are you proposing to change the rcu_dereference_sched to
> > > > rcu_dereference_raw in both __DECLARE_TRACE and __DO_TRACE?  
> > > 
> > > No, just in __DO_TRACE(). The rcu_dereference_sched() above in
> > > __DECLARE_TRACE() in the if (IS_ENABLED(CONFIG_LOCKDEP) block is
> > > required to show the warnings if trace_##name() is used wrong, and is
> > > the reason we can use rcu_dereference_raw() in __DO_TRACE() in the
> > > first place ;-)
> > > 
> > > This brings up another point. We should probably add to
> > > __DECLARE_TRACE_RCU() this:
> > > 
> > > #ifndef MODULE
> > > #define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \
> > > 	static inline void trace_##name##_rcuidle(proto)		\
> > > 	{								\
> > > 		if (static_key_false(&__tracepoint_##name.key))		\
> > > 			__DO_TRACE(&__tracepoint_##name,		\
> > > 				TP_PROTO(data_proto),			\
> > > 				TP_ARGS(data_args),			\
> > > 				TP_CONDITION(cond), 1);			\
> > > +		if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) {		\
> > > +			int idx;					\
> > > +			idx = srcu_read_lock_notrace(&tracepoint_srcu); \
> > > +			srcu_dereference_notrace(__tracepoint_##name.funcs, \
> > > +						&tracepoint_srcu);	\
> > > +			srcu_read_unlock_notrace(&tracepoint_srcu, idx); \
> > > +		}							\
> > > 	}
> > > #else
> > > 
> > > 
> > > So that lockdep works with trace_##name##__rcuidle() when the trace
> > > event is not enabled.
> > > 
> > > But that should be a separate patch and not part of this series. I may
> > > write that up tomorrow.  
> > 
> > Yes, that sounds good to me and would be good to add the safe guard there.
> > But you meant srcu_dereference above, not srcu_dereference_notrace right?
> 
> We don't need to trace them. I believe that the "srcu_*_notrace" still
> performs the lockdep checks. That's what we want. If they don't then we
> should not use notrace. But I believe they still do lockdep.

AFAICT, _notrace doesn't call into lockdep or tracing (there's also a comment
that says so):

/**
 * srcu_dereference_notrace - no tracing and no lockdep calls from here
 */

So then, we should use the regular variant for this additional check you're
suggesting.

thanks,

- Joel


^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12 19:17                   ` Joel Fernandes
@ 2018-07-12 20:15                     ` Steven Rostedt
  2018-07-12 20:29                       ` Joel Fernandes
  0 siblings, 1 reply; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12 20:15 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Paul E. McKenney, Peter Zijlstra, linux-kernel, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Thu, 12 Jul 2018 12:17:01 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> AFAICT, _notrace doesn't call into lockdep or tracing (there's also a comment
> that says so):
> 
> /**
>  * srcu_dereference_notrace - no tracing and no lockdep calls from here
>  */

Note, I had a different tree checked out, so I didn't have the source
available without digging through my email.

> 
> So then, we should use the regular variant for this additional check you're
> suggesting.

OK, I thought we had a rcu_dereference_notrace() that did checks and
thought that this followed suit, but it appears there is no such call.
That's where my confusion was.

Sure, I'll nuke the notrace() portion, thanks.

Also, I've applied 1-3, since 4 and 5 looks to be getting a remake, I'm
going to remove them from my queue. Please fold the SPDX patch into 5.

Thanks!

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12 20:15                     ` Steven Rostedt
@ 2018-07-12 20:29                       ` Joel Fernandes
  2018-07-12 20:31                         ` Steven Rostedt
  0 siblings, 1 reply; 52+ messages in thread
From: Joel Fernandes @ 2018-07-12 20:29 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Paul E. McKenney, Peter Zijlstra, LKML, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Thu, Jul 12, 2018 at 1:15 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
> On Thu, 12 Jul 2018 12:17:01 -0700
>> So then, we should use the regular variant for this additional check you're
>> suggesting.
>
> OK, I thought we had a rcu_dereference_notrace() that did checks and
> thought that this followed suit, but it appears there is no such call.
> That's where my confusion was.
>
> Sure, I'll nuke the notrace() portion, thanks.
>
> Also, I've applied 1-3, since 4 and 5 looks to be getting a remake, I'm
> going to remove them from my queue. Please fold the SPDX patch into 5.

Will do, and send out the 4 and 5 shortly with the SPDK folded.

Also the kselftest patches were acked and can be taken independently,
I had reposted them as a separate 2 patch series with some minor
changes based on your suggestions. Could you check them?

thanks!

- Joel

^ permalink raw reply	[flat|nested] 52+ messages in thread

* Re: [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-07-12 20:29                       ` Joel Fernandes
@ 2018-07-12 20:31                         ` Steven Rostedt
  0 siblings, 0 replies; 52+ messages in thread
From: Steven Rostedt @ 2018-07-12 20:31 UTC (permalink / raw)
  To: Joel Fernandes
  Cc: Paul E. McKenney, Peter Zijlstra, LKML, Boqun Feng,
	Byungchul Park, Ingo Molnar, Julia Cartwright, linux-kselftest,
	Masami Hiramatsu, Mathieu Desnoyers, Namhyung Kim,
	Thomas Glexiner, Tom Zanussi

On Thu, 12 Jul 2018 13:29:32 -0700
Joel Fernandes <joel@joelfernandes.org> wrote:

> Also the kselftest patches were acked and can be taken independently,
> I had reposted them as a separate 2 patch series with some minor
> changes based on your suggestions. Could you check them?
>

Yep, I saw them. I was going to wait till these patches were sent, but
since they are agnostic, I'll look at them now. Thanks for letting me
know.

-- Steve

^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-21 22:32 Joel Fernandes
@ 2018-06-21 22:32 ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-21 22:32 UTC (permalink / raw)
  To: linux-kernel
  Cc: kernel-team, Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Erick Reyes, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Shuah Khan, Steven Rostedt, Thomas Glexiner, Todd Kjos,
	Tom Zanussi, Will Deacon

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In recent tests with IRQ on/off tracepoints, a large performance
overhead ~10% is noticed when running hackbench. This is root caused to
calls to rcu_irq_enter_irqson and rcu_irq_exit_irqson from the
tracepoint code. Following a long discussion on the list [1] about this,
we concluded that srcu is a better alternative for use during rcu idle.
Although it does involve extra barriers, its lighter than the sched-rcu
version which has to do additional RCU calls to notify RCU idle about
entry into RCU sections.

In this patch, we change the underlying implementation of the
trace_*_rcuidle API to use SRCU. This has shown to improve performance
alot for the high frequency irq enable/disable tracepoints.

Test: Tested idle and preempt/irq tracepoints.

Here are some performance numbers:

With a run of the following 30 times on a single core x86 Qemu instance
with 1GB memory:
hackbench -g 4 -f 2 -l 3000

Completion times in seconds. CONFIG_PROVE_LOCKING=y.

No patches (without this series)
Mean: 3.048
Median: 3.025
Std Dev: 0.064

With Lockdep using irq tracepoints with RCU implementation:
Mean: 3.451   (-11.66 %)
Median: 3.447 (-12.22%)
Std Dev: 0.049

With Lockdep using irq tracepoints with SRCU implementation (this series):
Mean: 3.020   (I would consider the improvement against the "without
	       this series" case as just noise).
Median: 3.013
Std Dev: 0.033

[1] https://patchwork.kernel.org/patch/10344297/

Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/tracepoint.h | 49 +++++++++++++++++++++++++++++++-------
 kernel/tracepoint.c        | 16 ++++++++++++-
 2 files changed, 56 insertions(+), 9 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 19a690b559ca..beeb01e147f8 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -15,6 +15,7 @@
  */
 
 #include <linux/smp.h>
+#include <linux/srcu.h>
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/cpumask.h>
@@ -33,6 +34,8 @@ struct trace_eval_map {
 
 #define TRACEPOINT_DEFAULT_PRIO	10
 
+extern struct srcu_struct tracepoint_srcu;
+
 extern int
 tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data);
 extern int
@@ -75,10 +78,16 @@ int unregister_tracepoint_module_notifier(struct notifier_block *nb)
  * probe unregistration and the end of module exit to make sure there is no
  * caller executing a probe when it is freed.
  */
+#ifdef CONFIG_TRACEPOINTS
 static inline void tracepoint_synchronize_unregister(void)
 {
+	synchronize_srcu(&tracepoint_srcu);
 	synchronize_sched();
 }
+#else
+static inline void tracepoint_synchronize_unregister(void)
+{ }
+#endif
 
 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
 extern int syscall_regfunc(void);
@@ -129,18 +138,38 @@ extern void syscall_unregfunc(void);
  * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just
  * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto".
  */
-#define __DO_TRACE(tp, proto, args, cond, rcucheck)			\
+#define __DO_TRACE(tp, proto, args, cond, rcuidle)			\
 	do {								\
 		struct tracepoint_func *it_func_ptr;			\
 		void *it_func;						\
 		void *__data;						\
+		int __maybe_unused idx = 0;				\
 									\
 		if (!(cond))						\
 			return;						\
-		if (rcucheck)						\
-			rcu_irq_enter_irqson();				\
-		rcu_read_lock_sched_notrace();				\
-		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\
+									\
+		/*							\
+		 * For rcuidle callers, use srcu since sched-rcu	\
+		 * doesn't work from the idle path.			\
+		 */							\
+		if (rcuidle) {						\
+			if (in_nmi()) {					\
+				WARN_ON_ONCE(1);			\
+				return; /* no srcu from nmi */		\
+			}						\
+									\
+			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
+			it_func_ptr =					\
+				srcu_dereference_notrace((tp)->funcs,	\
+						&tracepoint_srcu);	\
+			/* To keep it consistent with !rcuidle path */	\
+			preempt_disable_notrace();			\
+		} else {						\
+			rcu_read_lock_sched_notrace();			\
+			it_func_ptr =					\
+				rcu_dereference_sched((tp)->funcs);	\
+		}							\
+									\
 		if (it_func_ptr) {					\
 			do {						\
 				it_func = (it_func_ptr)->func;		\
@@ -148,9 +177,13 @@ extern void syscall_unregfunc(void);
 				((void(*)(proto))(it_func))(args);	\
 			} while ((++it_func_ptr)->func);		\
 		}							\
-		rcu_read_unlock_sched_notrace();			\
-		if (rcucheck)						\
-			rcu_irq_exit_irqson();				\
+									\
+		if (rcuidle) {						\
+			preempt_enable_notrace();			\
+			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
+		} else {						\
+			rcu_read_unlock_sched_notrace();		\
+		}							\
 	} while (0)
 
 #ifndef MODULE
diff --git a/kernel/tracepoint.c b/kernel/tracepoint.c
index 6dc6356c3327..955148d91b74 100644
--- a/kernel/tracepoint.c
+++ b/kernel/tracepoint.c
@@ -31,6 +31,9 @@
 extern struct tracepoint * const __start___tracepoints_ptrs[];
 extern struct tracepoint * const __stop___tracepoints_ptrs[];
 
+DEFINE_SRCU(tracepoint_srcu);
+EXPORT_SYMBOL_GPL(tracepoint_srcu);
+
 /* Set to 1 to enable tracepoint debug output */
 static const int tracepoint_debug;
 
@@ -67,16 +70,27 @@ static inline void *allocate_probes(int count)
 	return p == NULL ? NULL : p->probes;
 }
 
-static void rcu_free_old_probes(struct rcu_head *head)
+static void srcu_free_old_probes(struct rcu_head *head)
 {
 	kfree(container_of(head, struct tp_probes, rcu));
 }
 
+static void rcu_free_old_probes(struct rcu_head *head)
+{
+	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
+}
+
 static inline void release_probes(struct tracepoint_func *old)
 {
 	if (old) {
 		struct tp_probes *tp_probes = container_of(old,
 			struct tp_probes, probes[0]);
+		/*
+		 * Tracepoint probes are protected by both sched RCU and SRCU,
+		 * by calling the SRCU callback in the sched RCU callback we
+		 * cover both cases. So let us chain the SRCU and sched RCU
+		 * callbacks to wait for both grace periods.
+		 */
 		call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
 	}
 }
-- 
2.18.0.rc2.346.g013aa6912e-goog


^ permalink raw reply	[flat|nested] 52+ messages in thread

* [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU
  2018-06-07 20:38 [PATCH v9 0/7] Centralize and unify usage of preempt/irq Joel Fernandes
@ 2018-06-07 20:38 ` Joel Fernandes
  0 siblings, 0 replies; 52+ messages in thread
From: Joel Fernandes @ 2018-06-07 20:38 UTC (permalink / raw)
  To: linux-kernel
  Cc: kernel-team, Joel Fernandes (Google),
	Boqun Feng, Byungchul Park, Erick Reyes, Ingo Molnar,
	Julia Cartwright, linux-kselftest, Masami Hiramatsu,
	Mathieu Desnoyers, Namhyung Kim, Paul McKenney, Peter Zijlstra,
	Shuah Khan, Steven Rostedt, Thomas Glexiner, Todd Kjos,
	Tom Zanussi

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In recent tests with IRQ on/off tracepoints, a large performance
overhead ~10% is noticed when running hackbench. This is root caused to
calls to rcu_irq_enter_irqson and rcu_irq_exit_irqson from the
tracepoint code. Following a long discussion on the list [1] about this,
we concluded that srcu is a better alternative for use during rcu idle.
Although it does involve extra barriers, its lighter than the sched-rcu
version which has to do additional RCU calls to notify RCU idle about
entry into RCU sections.

In this patch, we change the underlying implementation of the
trace_*_rcuidle API to use SRCU. This has shown to improve performance
alot for the high frequency irq enable/disable tracepoints.

Test: Tested idle and preempt/irq tracepoints.

Here are some performance numbers:

With a run of the following 30 times on a single core x86 Qemu instance
with 1GB memory:
hackbench -g 4 -f 2 -l 3000

Completion times in seconds. CONFIG_PROVE_LOCKING=y.

No patches (without this series)
Mean: 3.048
Median: 3.025
Std Dev: 0.064

With Lockdep using irq tracepoints with RCU implementation:
Mean: 3.451   (-11.66 %)
Median: 3.447 (-12.22%)
Std Dev: 0.049

With Lockdep using irq tracepoints with SRCU implementation (this series):
Mean: 3.020   (I would consider the improvement against the "without
	       this series" case as just noise).
Median: 3.013
Std Dev: 0.033

[1] https://patchwork.kernel.org/patch/10344297/

Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/tracepoint.h | 49 +++++++++++++++++++++++++++++++-------
 kernel/tracepoint.c        | 16 ++++++++++++-
 2 files changed, 56 insertions(+), 9 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index c94f466d57ef..9cf2882d2ef4 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -15,6 +15,7 @@
  */
 
 #include <linux/smp.h>
+#include <linux/srcu.h>
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/cpumask.h>
@@ -33,6 +34,8 @@ struct trace_eval_map {
 
 #define TRACEPOINT_DEFAULT_PRIO	10
 
+extern struct srcu_struct tracepoint_srcu;
+
 extern int
 tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data);
 extern int
@@ -75,10 +78,16 @@ int unregister_tracepoint_module_notifier(struct notifier_block *nb)
  * probe unregistration and the end of module exit to make sure there is no
  * caller executing a probe when it is freed.
  */
+#ifdef CONFIG_TRACEPOINTS
 static inline void tracepoint_synchronize_unregister(void)
 {
+	synchronize_srcu(&tracepoint_srcu);
 	synchronize_sched();
 }
+#else
+static inline void tracepoint_synchronize_unregister(void)
+{ }
+#endif
 
 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
 extern int syscall_regfunc(void);
@@ -129,18 +138,38 @@ extern void syscall_unregfunc(void);
  * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just
  * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto".
  */
-#define __DO_TRACE(tp, proto, args, cond, rcucheck)			\
+#define __DO_TRACE(tp, proto, args, cond, rcuidle)			\
 	do {								\
 		struct tracepoint_func *it_func_ptr;			\
 		void *it_func;						\
 		void *__data;						\
+		int __maybe_unused idx = 0;				\
 									\
 		if (!(cond))						\
 			return;						\
-		if (rcucheck)						\
-			rcu_irq_enter_irqson();				\
-		rcu_read_lock_sched_notrace();				\
-		it_func_ptr = rcu_dereference_sched((tp)->funcs);	\
+									\
+		/*							\
+		 * For rcuidle callers, use srcu since sched-rcu	\
+		 * doesn't work from the idle path.			\
+		 */							\
+		if (rcuidle) {						\
+			if (in_nmi()) {					\
+				WARN_ON_ONCE(1);			\
+				return; /* no srcu from nmi */		\
+			}						\
+									\
+			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
+			it_func_ptr =					\
+				srcu_dereference_notrace((tp)->funcs,	\
+						&tracepoint_srcu);	\
+			/* To keep it consistent with !rcuidle path */	\
+			preempt_disable_notrace();			\
+		} else {						\
+			rcu_read_lock_sched_notrace();			\
+			it_func_ptr =					\
+				rcu_dereference_sched((tp)->funcs);	\
+		}							\
+									\
 		if (it_func_ptr) {					\
 			do {						\
 				it_func = (it_func_ptr)->func;		\
@@ -148,9 +177,13 @@ extern void syscall_unregfunc(void);
 				((void(*)(proto))(it_func))(args);	\
 			} while ((++it_func_ptr)->func);		\
 		}							\
-		rcu_read_unlock_sched_notrace();			\
-		if (rcucheck)						\
-			rcu_irq_exit_irqson();				\
+									\
+		if (rcuidle) {						\
+			preempt_enable_notrace();			\
+			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
+		} else {						\
+			rcu_read_unlock_sched_notrace();		\
+		}							\
 	} while (0)
 
 #ifndef MODULE
diff --git a/kernel/tracepoint.c b/kernel/tracepoint.c
index 1e37da2e0c25..b2e3be589aff 100644
--- a/kernel/tracepoint.c
+++ b/kernel/tracepoint.c
@@ -31,6 +31,9 @@
 extern struct tracepoint * const __start___tracepoints_ptrs[];
 extern struct tracepoint * const __stop___tracepoints_ptrs[];
 
+DEFINE_SRCU(tracepoint_srcu);
+EXPORT_SYMBOL_GPL(tracepoint_srcu);
+
 /* Set to 1 to enable tracepoint debug output */
 static const int tracepoint_debug;
 
@@ -67,16 +70,27 @@ static inline void *allocate_probes(int count)
 	return p == NULL ? NULL : p->probes;
 }
 
-static void rcu_free_old_probes(struct rcu_head *head)
+static void srcu_free_old_probes(struct rcu_head *head)
 {
 	kfree(container_of(head, struct tp_probes, rcu));
 }
 
+static void rcu_free_old_probes(struct rcu_head *head)
+{
+	call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
+}
+
 static inline void release_probes(struct tracepoint_func *old)
 {
 	if (old) {
 		struct tp_probes *tp_probes = container_of(old,
 			struct tp_probes, probes[0]);
+		/*
+		 * Tracepoint probes are protected by both sched RCU and SRCU,
+		 * by calling the SRCU callback in the sched RCU callback we
+		 * cover both cases. So let us chain the SRCU and sched RCU
+		 * callbacks to wait for both grace periods.
+		 */
 		call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
 	}
 }
-- 
2.17.1.1185.g55be947832-goog

^ permalink raw reply	[flat|nested] 52+ messages in thread

end of thread, back to index

Thread overview: 52+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-06-28 18:21 [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 1/7] srcu: Add notrace variants of srcu_read_{lock,unlock} Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 2/7] srcu: Add notrace variant of srcu_dereference Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 3/7] trace/irqsoff: Split reset into separate functions Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
2018-07-11 12:49   ` Peter Zijlstra
2018-07-11 13:00     ` Steven Rostedt
2018-07-11 14:27       ` Paul E. McKenney
2018-07-11 14:46         ` Steven Rostedt
2018-07-11 15:15           ` Paul E. McKenney
2018-07-11 20:56             ` Joel Fernandes
2018-07-12  1:22               ` Steven Rostedt
2018-07-12  2:35                 ` Joel Fernandes
2018-07-11 20:52           ` Joel Fernandes
2018-07-12  3:21             ` Steven Rostedt
2018-07-12  4:28               ` Joel Fernandes
2018-07-12 13:35                 ` Steven Rostedt
2018-07-12 19:17                   ` Joel Fernandes
2018-07-12 20:15                     ` Steven Rostedt
2018-07-12 20:29                       ` Joel Fernandes
2018-07-12 20:31                         ` Steven Rostedt
2018-07-11 12:53   ` Peter Zijlstra
2018-07-12  2:32     ` Joel Fernandes
2018-07-11 12:56   ` Peter Zijlstra
2018-07-11 13:06     ` Steven Rostedt
2018-07-11 15:17       ` Peter Zijlstra
2018-07-11 15:26         ` Steven Rostedt
2018-07-11 16:46           ` Mathieu Desnoyers
2018-07-11 16:40         ` Mathieu Desnoyers
2018-07-12  0:31       ` Joel Fernandes
2018-07-12  1:26         ` Steven Rostedt
2018-06-28 18:21 ` [PATCH v9 5/7] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
2018-07-06 22:06   ` Steven Rostedt
2018-07-07  4:20     ` Joel Fernandes
2018-07-10 14:20   ` Steven Rostedt
2018-07-10 17:33     ` Joel Fernandes
2018-07-11 13:12   ` Peter Zijlstra
2018-07-11 13:19     ` Steven Rostedt
2018-07-11 13:22       ` Steven Rostedt
2018-07-12  8:38       ` Joel Fernandes
2018-07-12 13:37         ` Steven Rostedt
2018-07-12  0:44     ` Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 6/7] lib: Add module to simulate atomic sections for testing preemptoff tracers Joel Fernandes
2018-07-11  0:47   ` Steven Rostedt
2018-07-11  5:26     ` Joel Fernandes
2018-06-28 18:21 ` [PATCH v9 7/7] kselftests: Add tests for the preemptoff and irqsoff tracers Joel Fernandes
2018-07-11  0:49   ` Steven Rostedt
2018-07-11  5:27     ` Joel Fernandes
2018-07-03 14:15 ` [PATCH v9 0/7] Centralize and unify usage of preempt/irq tracepoints Joel Fernandes
2018-07-03 14:23   ` Steven Rostedt
  -- strict thread matches above, loose matches on Subject: below --
2018-06-21 22:32 Joel Fernandes
2018-06-21 22:32 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes
2018-06-07 20:38 [PATCH v9 0/7] Centralize and unify usage of preempt/irq Joel Fernandes
2018-06-07 20:38 ` [PATCH v9 4/7] tracepoint: Make rcuidle tracepoint callers use SRCU Joel Fernandes

LKML Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/lkml/0 lkml/git/0.git
	git clone --mirror https://lore.kernel.org/lkml/1 lkml/git/1.git
	git clone --mirror https://lore.kernel.org/lkml/2 lkml/git/2.git
	git clone --mirror https://lore.kernel.org/lkml/3 lkml/git/3.git
	git clone --mirror https://lore.kernel.org/lkml/4 lkml/git/4.git
	git clone --mirror https://lore.kernel.org/lkml/5 lkml/git/5.git
	git clone --mirror https://lore.kernel.org/lkml/6 lkml/git/6.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 lkml lkml/ https://lore.kernel.org/lkml \
		linux-kernel@vger.kernel.org linux-kernel@archiver.kernel.org
	public-inbox-index lkml


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-kernel


AGPL code for this site: git clone https://public-inbox.org/ public-inbox