[PATCH] tracing: Have the trace_event benchmark thread call cond_resched_rcu_qs()

[PATCH] tracing: Have the trace_event benchmark thread call cond_resched_rcu_qs()

[Steven Rostedt]

From: "Steven Rostedt (VMware)" <rostedt@goodmis.org>

The trace_event benchmark thread runs in kernel space in an infinite loop
while also calling cond_resched() in case anything else wants to schedule
in. Unfortunately, on a PREEMPT kernel, that makes it a nop, in which case,
this will never voluntarily schedule. That will cause synchronize_rcu_tasks()
to forever block on this thread, while it is running.

This is exactly what cond_resched_rcu_qs() is for. Use that instead.

Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
---
 kernel/trace/trace_benchmark.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/kernel/trace/trace_benchmark.c b/kernel/trace/trace_benchmark.c
index e49fbe9..16a8cf0 100644
--- a/kernel/trace/trace_benchmark.c
+++ b/kernel/trace/trace_benchmark.c
@@ -153,10 +153,18 @@ static int benchmark_event_kthread(void *arg)
 		trace_do_benchmark();
 
 		/*
-		 * We don't go to sleep, but let others
-		 * run as well.
+		 * We don't go to sleep, but let others run as well.
+		 * This is bascially a "yield()" to let any task that
+		 * wants to run, schedule in, but if the CPU is idle,
+		 * we'll keep burning cycles.
+		 *
+		 * Note the _rcu_qs() version of cond_resched() will
+		 * notify synchronize_rcu_tasks() that this thread has
+		 * passed a quiescent state for rcu_tasks. Otherwise
+		 * this thread will never voluntarily schedule which would
+		 * block synchronize_rcu_tasks() indefinitely.
 		 */
-		cond_resched();
+		cond_resched_rcu_qs();
 	}
 
 	return 0;
-- 
2.9.3

Re: [PATCH] tracing: Have the trace_event benchmark thread call cond_resched_rcu_qs()

[Steven Rostedt]

On Tue, 11 Apr 2017 18:54:35 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:

> From: "Steven Rostedt (VMware)" <rostedt@goodmis.org>
> 
> The trace_event benchmark thread runs in kernel space in an infinite loop
> while also calling cond_resched() in case anything else wants to schedule
> in. Unfortunately, on a PREEMPT kernel, that makes it a nop, in which case,
> this will never voluntarily schedule. That will cause synchronize_rcu_tasks()
> to forever block on this thread, while it is running.
> 
> This is exactly what cond_resched_rcu_qs() is for. Use that instead.
> 
> Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

This actually should be:

Recommended-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

-- Steve

> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> ---

Re: [PATCH] tracing: Have the trace_event benchmark thread call cond_resched_rcu_qs()

[Paul E. McKenney]

On Tue, Apr 11, 2017 at 06:56:01PM -0400, Steven Rostedt wrote:
> On Tue, 11 Apr 2017 18:54:35 -0400
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > From: "Steven Rostedt (VMware)" <rostedt@goodmis.org>
> > 
> > The trace_event benchmark thread runs in kernel space in an infinite loop
> > while also calling cond_resched() in case anything else wants to schedule
> > in. Unfortunately, on a PREEMPT kernel, that makes it a nop, in which case,
> > this will never voluntarily schedule. That will cause synchronize_rcu_tasks()
> > to forever block on this thread, while it is running.
> > 
> > This is exactly what cond_resched_rcu_qs() is for. Use that instead.
> > 
> > Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
> 
> This actually should be:
> 
> Recommended-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

And:

Acked-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

> -- Steve
> 
> > Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> > ---
>