LKML Archive on lore.kernel.org
 help / color / Atom feed
From: Joel Fernandes <joel@joelfernandes.org>
To: linux-kernel@vger.kernel.org
Cc: kernel-team@android.com,
	"Joel Fernandes (Google)" <joel@joelfernandes.org>,
	Boqun Feng <boqun.feng@gmail.com>,
	Byungchul Park <byungchul.park@lge.com>,
	Ingo Molnar <mingo@redhat.com>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	Namhyung Kim <namhyung@kernel.org>,
	Paul McKenney <paulmck@linux.vnet.ibm.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Steven Rostedt <rostedt@goodmis.org>,
	Thomas Glexiner <tglx@linutronix.de>,
	Tom Zanussi <tom.zanussi@linux.intel.com>
Subject: [PATCH v12 2/3] tracepoint: Make rcuidle tracepoint callers use SRCU
Date: Mon, 30 Jul 2018 15:24:22 -0700
Message-ID: <20180730222423.196630-3-joel@joelfernandes.org> (raw)
In-Reply-To: <20180730222423.196630-1-joel@joelfernandes.org>

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/

[remove rcu_read_lock_sched_notrace as its the equivalent of
preempt_disable_notrace and is unnecessary to call in tracepoint code]
Cleaned-up-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 include/linux/tracepoint.h | 41 ++++++++++++++++++++++++++++++--------
 kernel/tracepoint.c        | 16 ++++++++++++++-
 2 files changed, 48 insertions(+), 9 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 19a690b559ca..6e7bc6ebfcd8 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 +138,32 @@ 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);	\
+									\
+		/* srcu can't be used from NMI */			\
+		if (rcuidle && in_nmi())				\
+			WARN_ON_ONCE(1);				\
+									\
+		/* keep srcu and sched-rcu usage consistent */		\
+		preempt_disable_notrace();				\
+									\
+		/*							\
+		 * For rcuidle callers, use srcu since sched-rcu	\
+		 * doesn't work from the idle path.			\
+		 */							\
+		if (rcuidle)						\
+			idx = srcu_read_lock_notrace(&tracepoint_srcu);	\
+									\
+		it_func_ptr = rcu_dereference_raw((tp)->funcs);		\
+									\
 		if (it_func_ptr) {					\
 			do {						\
 				it_func = (it_func_ptr)->func;		\
@@ -148,9 +171,11 @@ 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)						\
+			srcu_read_unlock_notrace(&tracepoint_srcu, idx);\
+									\
+		preempt_enable_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);
 	}
 }
-- 
2.18.0.345.g5c9ce644c3-goog


  parent reply index

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-07-30 22:24 [PATCH v12 0/3] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
2018-07-30 22:24 ` [PATCH v12 1/3] lockdep: use this_cpu_ptr instead of get_cpu_var stats Joel Fernandes
2018-07-30 22:24 ` Joel Fernandes [this message]
2018-07-30 23:10   ` [PATCH v12 2/3] tracepoint: Make rcuidle tracepoint callers use SRCU Steven Rostedt
2018-08-10 15:35   ` Steven Rostedt
2018-08-10 16:32     ` Steven Rostedt
2018-07-30 22:24 ` [PATCH v12 3/3] tracing: Centralize preemptirq tracepoints and unify their usage Joel Fernandes
2018-08-06 19:50   ` Steven Rostedt
2018-08-07  0:43     ` Joel Fernandes
2018-08-07  1:43       ` Steven Rostedt
2018-08-07 13:33         ` Joel Fernandes
2018-08-07 13:49           ` Steven Rostedt
2018-08-07 14:10             ` Joel Fernandes
2018-08-07 14:34               ` Steven Rostedt
2018-08-07 14:48                 ` Joel Fernandes
2018-08-07 15:09                   ` Steven Rostedt
2018-08-07 15:24                     ` Joel Fernandes
2018-08-07 23:45                       ` Steven Rostedt
2018-08-07 23:54                         ` Joel Fernandes
2018-08-08  0:48                           ` Steven Rostedt
2018-08-08  1:17                             ` Joel Fernandes
2018-08-08  1:55                               ` Steven Rostedt
2018-08-08  2:13                                 ` Joel Fernandes
2018-08-08  2:28                                   ` Steven Rostedt
2018-08-08  3:44                                     ` Joel Fernandes
2018-08-08  3:53                                       ` Joel Fernandes
2018-08-08  5:06                                         ` Joel Fernandes
2018-08-08 12:46                                         ` Steven Rostedt
2018-08-08 13:03                                           ` Paul E. McKenney
2018-08-08 13:07                                             ` Steven Rostedt
2018-08-08 14:33                                               ` Paul E. McKenney
2018-08-08 14:49                                                 ` Steven Rostedt
2018-08-08 15:05                                                   ` Paul E. McKenney
2018-08-08 15:23                                                     ` Steven Rostedt
2018-08-08 16:02                                                       ` Paul E. McKenney
2018-08-08 16:24                                                         ` Steven Rostedt
2018-08-08 17:21                                                           ` Paul E. McKenney
2018-08-08 13:00                                         ` Paul E. McKenney
2018-08-08 14:10                                           ` Joel Fernandes
2018-08-08 14:49                                             ` Paul E. McKenney
2018-08-08 19:24                                               ` Joel Fernandes
2018-08-08 20:18                                                 ` Paul E. McKenney
2018-08-08 22:15                                                   ` Joel Fernandes
2018-08-08 22:47                                                     ` Paul E. McKenney
2018-08-09 12:18                                                       ` joel
2018-08-08 14:27                                           ` Steven Rostedt
2018-08-08 14:42                                             ` Paul E. McKenney
2018-08-08 15:27                                               ` Steven Rostedt
2018-08-08 16:03                                                 ` Paul E. McKenney
2018-08-02 14:55 ` [PATCH v12 0/3] " Masami Hiramatsu
2018-08-03  2:57   ` Joel Fernandes
2018-08-03  7:23     ` Masami Hiramatsu
2018-08-04  4:51       ` Joel Fernandes
2018-08-05 16:46         ` Joel Fernandes
2018-08-06  2:07           ` Masami Hiramatsu
2018-08-06 15:24             ` Joel Fernandes
2018-08-03  7:34     ` Masami Hiramatsu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20180730222423.196630-3-joel@joelfernandes.org \
    --to=joel@joelfernandes.org \
    --cc=boqun.feng@gmail.com \
    --cc=byungchul.park@lge.com \
    --cc=kernel-team@android.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mathieu.desnoyers@efficios.com \
    --cc=mhiramat@kernel.org \
    --cc=mingo@redhat.com \
    --cc=namhyung@kernel.org \
    --cc=paulmck@linux.vnet.ibm.com \
    --cc=peterz@infradead.org \
    --cc=rostedt@goodmis.org \
    --cc=tglx@linutronix.de \
    --cc=tom.zanussi@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

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
	git clone --mirror https://lore.kernel.org/lkml/7 lkml/git/7.git
	git clone --mirror https://lore.kernel.org/lkml/8 lkml/git/8.git
	git clone --mirror https://lore.kernel.org/lkml/9 lkml/git/9.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
	public-inbox-index lkml

Example config snippet for mirrors

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.git