On Thu, May 31, 2018 at 02:50:41AM -0400, Mathieu Desnoyers wrote:
> ----- On May 30, 2018, at 2:04 AM, Joel Fernandes [email protected] wrote:
>
> > From: "Joel Fernandes (Google)" <[email protected]>
> >
> > 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/
> >
> > Signed-off-by: Joel Fernandes (Google) <[email protected]>
> > ---
> > include/linux/tracepoint.h | 48 +++++++++++++++++++++++++++++++-------
> > kernel/tracepoint.c | 15 +++++++++++-
> > 2 files changed, 54 insertions(+), 9 deletions(-)
> >
> > diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
> > index c94f466d57ef..880794207921 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 +78,15 @@ 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) { }
>
> In order to add some consistency to the style in this header file,
> this empty function should look like:
>
> static inline void tracepoint_synchronize_unregister(void)
> { }
>
> (with newline before "{")
Ok, I changed it to the style you're proposing.
> > +#endif
> >
> > #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
> > extern int syscall_regfunc(void);
> > @@ -129,18 +137,38 @@ 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 */ \
>
> I find it odd to have a "return" in a macro that consists of a
> do { } while (0). I'm tempted to replace "return" by "break" here,
> to break the macro do/while (0) loop.
"return;" is also used from "if (!(cond))" above so I prefer be consistent
and just use return than break as done above, but please let me know if you
still object.
> > + } \
> > + \
> > + 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 +176,13 @@ 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..54157792f5ab 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 +70,26 @@ 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 lets chain the SRCU and RCU callbacks.
>
> lets -> let's
>
> But I would use a different wording for that sentence that does not include
> "let's" which is short for "let us", e.g.
>
> "Chain the SRCU and sched RCU callbacks to wait for both grace periods."
Fixed, thanks.
> With the comments above taken care of, please add my:
>
> Reviewed-by: Mathieu Desnoyers <[email protected]>
Will do, and thanks for the review!
Below is the updated patch.
- Joel
---8<-----------------------
From: "Joel Fernandes (Google)" <[email protected]>
Date: Thu, 26 Apr 2018 20:44:07 -0700
Subject: [PATCH v8.1] tracepoint: Make rcuidle tracepoint callers use SRCU
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 <[email protected]>
Signed-off-by: Joel Fernandes (Google) <[email protected]>
---
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);
}
}
--
2.17.0.921.gf22659ad46-goog
