[PATCH 0/2] sched/deadline: Fixes for constrained deadline tasks

[PATCH 0/2] sched/deadline: Fixes for constrained deadline tasks

[Daniel Bristot de Oliveira]

While reading sched deadline code, I find out that a constrained
deadline task could be replenished before the next period if
activated after the deadline, opening the window to run for more
than Q/P. The patch [2] explains and fixes this problem.

Furthermore, while fixing this issue, I found that the replenishment
timer was being fired at the deadline of the task. This works fine
for implicit deadline tasks (deadline == period) because the deadline
is at the same point in time of the next period. But that is not true
for constrained deadline tasks (deadline < period). This problem is
not as visible as the first because the runtime leakage takes
place only in the second activation. Next activations receive the
correct bandwidth. However, after the 2nd activation, tasks are
activated in the (period - dl_deadline) instant, which is before
the expected activation. This problem is explained in the fix
description as well.

Daniel Bristot de Oliveira (2):
  sched/deadline: Replenishment timer should fire in the next period
  sched/deadline: Throttle a constrained deadline task activated after
    the deadline

 kernel/sched/deadline.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 51 insertions(+), 2 deletions(-)

-- 
2.9.3

[PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[Daniel Bristot de Oliveira]

Currently, the replenishment timer is set to fire at the deadline
of a task. Although that works for implicit deadline tasks because the
deadline is equals to the begin of the next period, that is not correct
for constrained deadline tasks (deadline < period).

For instance:

f.c:
 --------------- %< ---------------
int main (void)
{
	for(;;);
}
 --------------- >% ---------------

  # gcc -o f f.c

  # trace-cmd record -e sched:sched_switch                              \
   				   -e syscalls:sys_exit_sched_setattr   \
   chrt -d --sched-runtime  490000000 					\
           --sched-deadline 500000000 					\
	   --sched-period  1000000000 0 ./f

  # trace-cmd report | grep "{pid of ./f}"

After setting parameters, the task is replenished and continue running
until being throttled.
         f-11295 [003] 13322.113776: sys_exit_sched_setattr: 0x0

The task is throttled after running 492318 ms, as expected.
         f-11295 [003] 13322.606094: sched_switch:   f:11295 [-1] R ==> \
						   watchdog/3:32 [0]

But then, the task is replenished 500719 ms after the first
replenishment:
    <idle>-0     [003] 13322.614495: sched_switch:   swapper/3:0 [120] R \
						   ==> f:11295 [-1]

Running for 490277 ms:
         f-11295 [003] 13323.104772: sched_switch:   f:11295 [-1] R ==>  \
						   swapper/3:0 [120]

Hence, in the first period, the task runs 2 * runtime, and that is a bug.

During the first replenishment, the next deadline is set one period away.
So the runtime/period starts to be respected. However, as the second
replenishment took place in the wrong instant, the next replenishment
will also be held in a wrong instant of time. Rather than occurring in
the nth period away from the first activation, it is taking place
in the (nth period - relative deadline).

Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
Cc: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-kernel@vger.kernel.org

---
 kernel/sched/deadline.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 70ef2b1..3c94d85 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -505,10 +505,15 @@ static void update_dl_entity(struct sched_dl_entity *dl_se,
 	}
 }
 
+static inline u64 dl_next_period(struct sched_dl_entity *dl_se)
+{
+	return dl_se->deadline - dl_se->dl_deadline + dl_se->dl_period;
+}
+
 /*
  * If the entity depleted all its runtime, and if we want it to sleep
  * while waiting for some new execution time to become available, we
- * set the bandwidth enforcement timer to the replenishment instant
+ * set the bandwidth replenishment timer to the replenishment instant
  * and try to activate it.
  *
  * Notice that it is important for the caller to know if the timer
@@ -530,7 +535,7 @@ static int start_dl_timer(struct task_struct *p)
 	 * that it is actually coming from rq->clock and not from
 	 * hrtimer's time base reading.
 	 */
-	act = ns_to_ktime(dl_se->deadline);
+	act = ns_to_ktime(dl_next_period(dl_se));
 	now = hrtimer_cb_get_time(timer);
 	delta = ktime_to_ns(now) - rq_clock(rq);
 	act = ktime_add_ns(act, delta);
-- 
2.9.3

Re: [PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[Steven Rostedt]

On Fri, 10 Feb 2017 20:48:10 +0100
Daniel Bristot de Oliveira <bristot@redhat.com> wrote:

> Currently, the replenishment timer is set to fire at the deadline
> of a task. Although that works for implicit deadline tasks because the
> deadline is equals to the begin of the next period, that is not correct
> for constrained deadline tasks (deadline < period).
> 
> For instance:
> 
> f.c:
>  --------------- %< ---------------
> int main (void)
> {
> 	for(;;);
> }
>  --------------- >% ---------------  
> 
>   # gcc -o f f.c
> 
>   # trace-cmd record -e sched:sched_switch                              \
>    				   -e syscalls:sys_exit_sched_setattr   \
>    chrt -d --sched-runtime  490000000 					\
>            --sched-deadline 500000000 					\
> 	   --sched-period  1000000000 0 ./f
> 
>   # trace-cmd report | grep "{pid of ./f}"

 # trace-cmd report -F '.*:common_pid == 11295'

does the same ;-)

Or better yet, add "-F" (just -F, no extra paremeter. In report, -F
means "filter", in record -F means "follow"), and that will only record
the task that is created.

Also, it may be nice to use --ts-diff option to see results. I ran this
with my own spinning program, and here's the output:

 # trace-cmd report --ts-diff
[...]
        userspin-1540  [003]   298.725084: (+822)   sys_exit_sched_setattr: 0x0
        userspin-1540  [003]   299.218199: (+493115) sched_switch:         userspin:1540 [4294967295] R ==> rcuc/3:38 [98]
          <idle>-0     [003]   299.225111: (+6912)  sched_switch:         swapper/3:0 [120] R ==> userspin:1540 [4294967295]
        userspin-1540  [003]   299.718192: (+493081) sched_switch:         userspin:1540 [4294967295] R ==> ktimersoftd/3:39 [98]
          <idle>-0     [003]   300.225125: (+506933) sched_switch:         swapper/3:0 [120] R ==> userspin:1540 [4294967295]
        userspin-1540  [003]   300.717185: (+492060) sched_switch:         userspin:1540 [4294967295] R ==> rcuc/3:38 [98]
          <idle>-0     [003]   301.225116: (+507931) sched_switch:         swapper/3:0 [120] R ==> userspin:1540 [4294967295]
        userspin-1540  [003]   301.718177: (+493061) sched_switch:         userspin:1540 [4294967295] R ==> rcuc/3:38 [98]
          <idle>-0     [003]   302.225120: (+506943) sched_switch:         swapper/3:0 [120] R ==> userspin:1540 [4294967295]


> 
> After setting parameters, the task is replenished and continue running
> until being throttled.
>          f-11295 [003] 13322.113776: sys_exit_sched_setattr: 0x0
> 
> The task is throttled after running 492318 ms, as expected.
>          f-11295 [003] 13322.606094: sched_switch:   f:11295 [-1] R ==> \
> 						   watchdog/3:32 [0]
> 
> But then, the task is replenished 500719 ms after the first
> replenishment:
>     <idle>-0     [003] 13322.614495: sched_switch:   swapper/3:0 [120] R \
> 						   ==> f:11295 [-1]  
> 
> Running for 490277 ms:
>          f-11295 [003] 13323.104772: sched_switch:   f:11295 [-1] R ==>  \
> 						   swapper/3:0 [120]
> 
> Hence, in the first period, the task runs 2 * runtime, and that is a bug.
> 
> During the first replenishment, the next deadline is set one period away.
> So the runtime/period starts to be respected. However, as the second
> replenishment took place in the wrong instant, the next replenishment
> will also be held in a wrong instant of time. Rather than occurring in
> the nth period away from the first activation, it is taking place
> in the (nth period - relative deadline).

OK, after you patch, I get this:

        userspin-1581  [003]   103.450129: (+716)   sys_exit_sched_setattr: 0x0
        userspin-1581  [003]   103.943957: (+493828) sched_switch:         userspin:1581 [4294967295] R ==> rcuc/3:38 [98]
          <idle>-0     [003]   104.450172: (+506215) sched_switch:         swapper/3:0 [120] R ==> userspin:1581 [4294967295]
        userspin-1581  [003]   104.942349: (+492177) sched_switch:         userspin:1581 [4294967295] R ==> watchdog/3:36 [0]
          <idle>-0     [003]   105.450164: (+507815) sched_switch:         swapper/3:0 [120] R ==> userspin:1581 [4294967295]
        userspin-1581  [003]   105.942354: (+492190) sched_switch:         userspin:1581 [4294967295] R ==> ktimersoftd/3:39 [98]
          <idle>-0     [003]   106.450159: (+507805) sched_switch:         swapper/3:0 [120] R ==> userspin:1581 [4294967295]
        userspin-1581  [003]   106.942366: (+492207) sched_switch:         userspin:1581 [4294967295] R ==> ktimersoftd/3:39 [98]
          <idle>-0     [003]   107.450164: (+507798) sched_switch:         swapper/3:0 [120] R ==> userspin:1581 [4294967295]

Looks promising. But unrelated, I noticed this with and without your
patch:

        userspin-1567  [002]    81.455345: (+1196)  sys_exit_sched_setattr: 0x0
        userspin-1567  [002]    81.457599: (+2254)  sched_switch:         chrt:1567 [4294967295] D ==> rcuc/2:29 [98]
          <idle>-0     [002]    81.466030: (+8431)  sched_switch:         swapper/2:0 [120] R ==> chrt:1567 [4294967295]
        userspin-1567  [002]    81.466123: (+93)    sched_switch:         chrt:1567 [4294967295] D ==> rcuc/2:29 [98]
          <idle>-0     [002]    81.471862: (+5739)  sched_switch:         swapper/2:0 [120] R ==> chrt:1567 [4294967295]
        userspin-1567  [002]    81.471959: (+97)    sched_switch:         chrt:1567 [4294967295] D ==> swapper/2:0 [120]
          <idle>-0     [002]    81.478285: (+6326)  sched_switch:         swapper/2:0 [120] R ==> chrt:1567 [4294967295]
        userspin-1567  [002]    81.478369: (+84)    sched_switch:         chrt:1567 [4294967295] D ==> swapper/2:0 [120]
          <idle>-0     [002]    81.486664: (+8295)  sched_switch:         swapper/2:0 [120] R ==> chrt:1567 [4294967295]
        userspin-1567  [002]    81.487291: (+627)   sched_switch:         chrt:1567 [4294967295] D|W ==> swapper/2:0 [120]
          <idle>-0     [002]    81.494884: (+7593)  sched_switch:         swapper/2:0 [120] R ==> chrt:1567 [4294967295]
        userspin-1567  [002]    81.988095: (+493211) sched_switch:         userspin:1567 [4294967295] R ==> rcuc/2:29 [98]
          <idle>-0     [002]    82.496009: (+507914) sched_switch:         swapper/2:0 [120] R ==> userspin:1567 [4294967295]
        userspin-1567  [002]    82.988158: (+492149) sched_switch:         userspin:1567 [4294967295] R ==> rcuc/2:29 [98]
          <idle>-0     [002]    83.494897: (+506739) sched_switch:         swapper/2:0 [120] R ==> userspin:1567 [4294967295]
        userspin-1567  [002]    83.986181: (+491284) sched_switch:         userspin:1567 [4294967295] R ==> rcuc/2:29 [98]
          <idle>-0     [002]    84.494895: (+508714) sched_switch:         swapper/2:0 [120] R ==> userspin:1567 [4294967295]

A bit of choppiness to get started. I'll have to look more into this.

I'll take a look at this patch deeper to try to understand everything
before applying a reviewed-by.

-- Steve

> 
> Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Juri Lelli <juri.lelli@arm.com>
> Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
> Cc: Luca Abeni <luca.abeni@santannapisa.it>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: linux-kernel@vger.kernel.org
> 
> ---
>  kernel/sched/deadline.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index 70ef2b1..3c94d85 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -505,10 +505,15 @@ static void update_dl_entity(struct sched_dl_entity *dl_se,
>  	}
>  }
>  
> +static inline u64 dl_next_period(struct sched_dl_entity *dl_se)
> +{
> +	return dl_se->deadline - dl_se->dl_deadline + dl_se->dl_period;
> +}
> +
>  /*
>   * If the entity depleted all its runtime, and if we want it to sleep
>   * while waiting for some new execution time to become available, we
> - * set the bandwidth enforcement timer to the replenishment instant
> + * set the bandwidth replenishment timer to the replenishment instant
>   * and try to activate it.
>   *
>   * Notice that it is important for the caller to know if the timer
> @@ -530,7 +535,7 @@ static int start_dl_timer(struct task_struct *p)
>  	 * that it is actually coming from rq->clock and not from
>  	 * hrtimer's time base reading.
>  	 */
> -	act = ns_to_ktime(dl_se->deadline);
> +	act = ns_to_ktime(dl_next_period(dl_se));
>  	now = hrtimer_cb_get_time(timer);
>  	delta = ktime_to_ns(now) - rq_clock(rq);
>  	act = ktime_add_ns(act, delta);

Re: [PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[luca abeni]

Hi Daniel,

On Fri, 10 Feb 2017 20:48:10 +0100
Daniel Bristot de Oliveira <bristot@redhat.com> wrote:

[...]
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index 70ef2b1..3c94d85 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -505,10 +505,15 @@ static void update_dl_entity(struct
> sched_dl_entity *dl_se, }
>  }
>  
> +static inline u64 dl_next_period(struct sched_dl_entity *dl_se)
> +{
> +	return dl_se->deadline - dl_se->dl_deadline +
> dl_se->dl_period; +}
> +
>  /*
>   * If the entity depleted all its runtime, and if we want it to sleep
>   * while waiting for some new execution time to become available, we
> - * set the bandwidth enforcement timer to the replenishment instant
> + * set the bandwidth replenishment timer to the replenishment instant
>   * and try to activate it.
>   *
>   * Notice that it is important for the caller to know if the timer
> @@ -530,7 +535,7 @@ static int start_dl_timer(struct task_struct *p)
>  	 * that it is actually coming from rq->clock and not from
>  	 * hrtimer's time base reading.
>  	 */
> -	act = ns_to_ktime(dl_se->deadline);
> +	act = ns_to_ktime(dl_next_period(dl_se));

Looks like there is a real bug in the code, and your fix looks correct
to me. I think it should be committed.


			Thanks,
				Luca


>  	now = hrtimer_cb_get_time(timer);
>  	delta = ktime_to_ns(now) - rq_clock(rq);
>  	act = ktime_add_ns(act, delta);

Re: [PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[Peter Zijlstra]

On Sat, Feb 11, 2017 at 08:12:37AM +0100, luca abeni wrote:
> Hi Daniel,
> 
> On Fri, 10 Feb 2017 20:48:10 +0100
> Daniel Bristot de Oliveira <bristot@redhat.com> wrote:
> 
> [...]
> > diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> > index 70ef2b1..3c94d85 100644
> > --- a/kernel/sched/deadline.c
> > +++ b/kernel/sched/deadline.c
> > @@ -505,10 +505,15 @@ static void update_dl_entity(struct
> > sched_dl_entity *dl_se, }
> >  }
> >  
> > +static inline u64 dl_next_period(struct sched_dl_entity *dl_se)
> > +{
> > +	return dl_se->deadline - dl_se->dl_deadline +
> > dl_se->dl_period; +}
> > +
> >  /*
> >   * If the entity depleted all its runtime, and if we want it to sleep
> >   * while waiting for some new execution time to become available, we
> > - * set the bandwidth enforcement timer to the replenishment instant
> > + * set the bandwidth replenishment timer to the replenishment instant
> >   * and try to activate it.
> >   *
> >   * Notice that it is important for the caller to know if the timer
> > @@ -530,7 +535,7 @@ static int start_dl_timer(struct task_struct *p)
> >  	 * that it is actually coming from rq->clock and not from
> >  	 * hrtimer's time base reading.
> >  	 */
> > -	act = ns_to_ktime(dl_se->deadline);
> > +	act = ns_to_ktime(dl_next_period(dl_se));
> 
> Looks like there is a real bug in the code, and your fix looks correct
> to me. I think it should be committed.
> 

I've interpreted this as:

Reviewed-by: Luca Abeni <luca.abeni@santannapisa.it>

Holler if you disagree.

Re: [PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[Steven Rostedt]

On Mon, 13 Feb 2017 12:10:25 +0100
Peter Zijlstra <peterz@infradead.org> wrote:


> I've interpreted this as:
> 
> Reviewed-by: Luca Abeni <luca.abeni@santannapisa.it>
> 
> Holler if you disagree.

You can add mine too. I put in a lot of trace_printk()s and it all
appears to be exactly as Daniel describes.

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

-- Steve

Re: [PATCH 1/2] sched/deadline: Replenishment timer should fire in the next period

[Juri Lelli]

On 13/02/17 09:59, Steven Rostedt wrote:
> On Mon, 13 Feb 2017 12:10:25 +0100
> Peter Zijlstra <peterz@infradead.org> wrote:
> 
> 
> > I've interpreted this as:
> > 
> > Reviewed-by: Luca Abeni <luca.abeni@santannapisa.it>
> > 
> > Holler if you disagree.
> 
> You can add mine too. I put in a lot of trace_printk()s and it all
> appears to be exactly as Daniel describes.
> 
> Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> 

I'm a bit late, but I looked at it as well and it looks good.

If you want,

Reviewed-by: Juri Lelli <juri.lelli@arm.com>

Thanks!

- Juri

[PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Daniel Bristot de Oliveira]

During the activation, CBS checks if it can reuse the current task's
runtime and period. If the deadline of the task is in the past, CBS
cannot use the runtime, and so it replenishes the task. This rule
works fine for implicit deadline tasks (deadline == period), and the
CBS was designed for implicit deadline tasks. However, a task with
constrained deadline (deadine < period) might be awakened after the
deadline, but before the next period. In this case, replenishing the
task would allow it to run for runtime / deadline. As in this case
deadline < period, CBS enables a task to run for more than the
runtime/period. In a very load system, this can cause the domino
effect, making other tasks to miss their deadlines.

To avoid this problem, in the activation of a constrained deadline
task after the deadline but before the next period, throttle the
task and set the replenishing timer to the begin of the next period,
unless it is boosted.

Reproducer:

 --------------- %< ---------------
  int main (int argc, char **argv)
  {
	int ret;
	int flags = 0;
	unsigned long l = 0;
	struct timespec ts;
	struct sched_attr attr;

	memset(&attr, 0, sizeof(attr));
	attr.size = sizeof(attr);

	attr.sched_policy   = SCHED_DEADLINE;
	attr.sched_runtime  = 2 * 1000 * 1000;		/* 2 ms */
	attr.sched_deadline = 2 * 1000 * 1000;		/* 2 ms */
	attr.sched_period   = 2 * 1000 * 1000 * 1000;	/* 2 s */

	ts.tv_sec = 0;
	ts.tv_nsec = 2000 * 1000;			/* 2 ms */

	ret = sched_setattr(0, &attr, flags);

	if (ret < 0) {
		perror("sched_setattr");
		exit(-1);
	}

	for(;;) {
		/* XXX: you may need to adjust the loop */
		for (l = 0; l < 150000; l++);
		/*
		 * The ideia is to go to sleep right before the deadline
		 * and then wake up before the next period to receive
		 * a new replenishment.
		 */
		nanosleep(&ts, NULL);
	}

	exit(0);
  }
  --------------- >% ---------------

On my box, this reproducer uses almost 50% of the CPU time, which is
obviously wrong for a task with 2/2000 reservation.

Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
Cc: Luca Abeni <luca.abeni@santannapisa.it>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-kernel@vger.kernel.org
---
 kernel/sched/deadline.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 44 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 3c94d85..b74d40e 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -694,6 +694,36 @@ void init_dl_task_timer(struct sched_dl_entity *dl_se)
 	timer->function = dl_task_timer;
 }
 
+/* During the activation, CBS checks if it can reuse the current task's
+ * runtime and period. If the deadline of the task is in the past, CBS
+ * cannot use the runtime, and so it replenishes the task. This rule
+ * works fine for implicit deadline tasks (deadline == period), and the
+ * CBS was designed for implicit deadline tasks. However, a task with
+ * constrained deadline (deadine < period) might be awakened after the
+ * deadline, but before the next period. In this case, replenishing the
+ * task would allow it to run for runtime / deadline. As in this case
+ * deadline < period, CBS enables a task to run for more than the
+ * runtime / period. In a very load system, this can cause the domino
+ * effect, making other tasks to miss their deadlines.
+ *
+ * To avoid this problem, in the activation of a constrained deadline
+ * task after the deadline but before the next period, throttle the
+ * task and set the replenishing timer to the begin of the next period,
+ * unless it is boosted.
+ */
+static inline void dl_check_constrained_dl(struct sched_dl_entity *dl_se)
+{
+	struct task_struct *p = dl_task_of(dl_se);
+	struct rq *rq = rq_of_dl_rq(dl_rq_of_se(dl_se));
+
+	if (dl_time_before(dl_se->deadline, rq_clock(rq)) &&
+	    dl_time_before(rq_clock(rq), dl_next_period(dl_se))) {
+		if (unlikely(dl_se->dl_boosted || !start_dl_timer(p)))
+			return;
+		dl_se->dl_throttled = 1;
+	}
+}
+
 static
 int dl_runtime_exceeded(struct sched_dl_entity *dl_se)
 {
@@ -927,6 +957,11 @@ static void dequeue_dl_entity(struct sched_dl_entity *dl_se)
 	__dequeue_dl_entity(dl_se);
 }
 
+static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
+{
+	return dl_se->dl_runtime < dl_se->dl_period;
+}
+
 static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
 	struct task_struct *pi_task = rt_mutex_get_top_task(p);
@@ -953,6 +988,15 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 	}
 
 	/*
+	 * Check if a constrained deadline task was activated
+	 * after the deadline but before the next period.
+	 * If that is the case, the task will be throttled and
+	 * the replenishment timer will be set to the next period.
+	 */
+	if (!pi_se->dl_throttled && dl_is_constrained(pi_se))
+		dl_check_constrained_dl(pi_se);
+
+	/*
 	 * If p is throttled, we do nothing. In fact, if it exhausted
 	 * its budget it needs a replenishment and, since it now is on
 	 * its rq, the bandwidth timer callback (which clearly has not
-- 
2.9.3

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[luca abeni]

Hi Daniel,

On Fri, 10 Feb 2017 20:48:11 +0100
Daniel Bristot de Oliveira <bristot@redhat.com> wrote:

> During the activation, CBS checks if it can reuse the current task's
> runtime and period. If the deadline of the task is in the past, CBS
> cannot use the runtime, and so it replenishes the task. This rule
> works fine for implicit deadline tasks (deadline == period), and the
> CBS was designed for implicit deadline tasks. However, a task with
> constrained deadline (deadine < period) might be awakened after the
> deadline, but before the next period. In this case, replenishing the
> task would allow it to run for runtime / deadline. As in this case
> deadline < period, CBS enables a task to run for more than the
> runtime/period. In a very load system, this can cause the domino
> effect, making other tasks to miss their deadlines.

I think you are right: SCHED_DEADLINE implements the original CBS
algorithm here, but uses relative deadlines different from periods in
other places (while the original algorithm only considered relative
deadlines equal to periods).
An this mix is dangerous... I think your fix is correct, and cures a
real problem.



			Thanks,
				Luca


> 
> To avoid this problem, in the activation of a constrained deadline
> task after the deadline but before the next period, throttle the
> task and set the replenishing timer to the begin of the next period,
> unless it is boosted.
> 
> Reproducer:
> 
>  --------------- %< ---------------
>   int main (int argc, char **argv)
>   {
> 	int ret;
> 	int flags = 0;
> 	unsigned long l = 0;
> 	struct timespec ts;
> 	struct sched_attr attr;
> 
> 	memset(&attr, 0, sizeof(attr));
> 	attr.size = sizeof(attr);
> 
> 	attr.sched_policy   = SCHED_DEADLINE;
> 	attr.sched_runtime  = 2 * 1000 * 1000;		/* 2 ms
> */ attr.sched_deadline = 2 * 1000 * 1000;		/* 2 ms */
> 	attr.sched_period   = 2 * 1000 * 1000 * 1000;	/* 2 s */
> 
> 	ts.tv_sec = 0;
> 	ts.tv_nsec = 2000 * 1000;			/* 2 ms */
> 
> 	ret = sched_setattr(0, &attr, flags);
> 
> 	if (ret < 0) {
> 		perror("sched_setattr");
> 		exit(-1);
> 	}
> 
> 	for(;;) {
> 		/* XXX: you may need to adjust the loop */
> 		for (l = 0; l < 150000; l++);
> 		/*
> 		 * The ideia is to go to sleep right before the
> deadline
> 		 * and then wake up before the next period to receive
> 		 * a new replenishment.
> 		 */
> 		nanosleep(&ts, NULL);
> 	}
> 
> 	exit(0);
>   }
>   --------------- >% ---------------  
> 
> On my box, this reproducer uses almost 50% of the CPU time, which is
> obviously wrong for a task with 2/2000 reservation.
> 
> Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Juri Lelli <juri.lelli@arm.com>
> Cc: Tommaso Cucinotta <tommaso.cucinotta@sssup.it>
> Cc: Luca Abeni <luca.abeni@santannapisa.it>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: linux-kernel@vger.kernel.org
> ---
>  kernel/sched/deadline.c | 44
> ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44
> insertions(+)
> 
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index 3c94d85..b74d40e 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -694,6 +694,36 @@ void init_dl_task_timer(struct sched_dl_entity
> *dl_se) timer->function = dl_task_timer;
>  }
>  
> +/* During the activation, CBS checks if it can reuse the current
> task's
> + * runtime and period. If the deadline of the task is in the past,
> CBS
> + * cannot use the runtime, and so it replenishes the task. This rule
> + * works fine for implicit deadline tasks (deadline == period), and
> the
> + * CBS was designed for implicit deadline tasks. However, a task with
> + * constrained deadline (deadine < period) might be awakened after
> the
> + * deadline, but before the next period. In this case, replenishing
> the
> + * task would allow it to run for runtime / deadline. As in this case
> + * deadline < period, CBS enables a task to run for more than the
> + * runtime / period. In a very load system, this can cause the domino
> + * effect, making other tasks to miss their deadlines.
> + *
> + * To avoid this problem, in the activation of a constrained deadline
> + * task after the deadline but before the next period, throttle the
> + * task and set the replenishing timer to the begin of the next
> period,
> + * unless it is boosted.
> + */
> +static inline void dl_check_constrained_dl(struct sched_dl_entity
> *dl_se) +{
> +	struct task_struct *p = dl_task_of(dl_se);
> +	struct rq *rq = rq_of_dl_rq(dl_rq_of_se(dl_se));
> +
> +	if (dl_time_before(dl_se->deadline, rq_clock(rq)) &&
> +	    dl_time_before(rq_clock(rq), dl_next_period(dl_se))) {
> +		if (unlikely(dl_se->dl_boosted
> || !start_dl_timer(p)))
> +			return;
> +		dl_se->dl_throttled = 1;
> +	}
> +}
> +
>  static
>  int dl_runtime_exceeded(struct sched_dl_entity *dl_se)
>  {
> @@ -927,6 +957,11 @@ static void dequeue_dl_entity(struct
> sched_dl_entity *dl_se) __dequeue_dl_entity(dl_se);
>  }
>  
> +static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
> +{
> +	return dl_se->dl_runtime < dl_se->dl_period;
> +}
> +
>  static void enqueue_task_dl(struct rq *rq, struct task_struct *p,
> int flags) {
>  	struct task_struct *pi_task = rt_mutex_get_top_task(p);
> @@ -953,6 +988,15 @@ static void enqueue_task_dl(struct rq *rq,
> struct task_struct *p, int flags) }
>  
>  	/*
> +	 * Check if a constrained deadline task was activated
> +	 * after the deadline but before the next period.
> +	 * If that is the case, the task will be throttled and
> +	 * the replenishment timer will be set to the next period.
> +	 */
> +	if (!pi_se->dl_throttled && dl_is_constrained(pi_se))
> +		dl_check_constrained_dl(pi_se);
> +
> +	/*
>  	 * If p is throttled, we do nothing. In fact, if it exhausted
>  	 * its budget it needs a replenishment and, since it now is
> on
>  	 * its rq, the bandwidth timer callback (which clearly has
> not

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Mike Galbraith]

On Sat, 2017-02-11 at 08:15 +0100, luca abeni wrote:
> Hi Daniel,
> 
> On Fri, 10 Feb 2017 20:48:11 +0100
> Daniel Bristot de Oliveira <bristot@redhat.com> wrote:
> 
> > During the activation, CBS checks if it can reuse the current
> > task's
> > runtime and period. If the deadline of the task is in the past, CBS
> > cannot use the runtime, and so it replenishes the task. This rule
> > works fine for implicit deadline tasks (deadline == period), and
> > the
> > CBS was designed for implicit deadline tasks. However, a task with
> > constrained deadline (deadine < period) might be awakened after the
> > deadline, but before the next period. In this case, replenishing
> > the
> > task would allow it to run for runtime / deadline. As in this case
> > deadline < period, CBS enables a task to run for more than the
> > runtime/period. In a very load system, this can cause the domino
> > effect, making other tasks to miss their deadlines.
> 
> I think you are right: SCHED_DEADLINE implements the original CBS
> algorithm here, but uses relative deadlines different from periods in
> other places (while the original algorithm only considered relative
> deadlines equal to periods).
> An this mix is dangerous... I think your fix is correct, and cures a
> real problem.

Both of these should be tagged for stable as well, or?

	-Mike

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Steven Rostedt]

On Sat, 11 Feb 2017 09:00:06 +0100
Mike Galbraith <efault@gmx.de> wrote:


> Both of these should be tagged for stable as well, or?
> 

Yes.

-- Steve

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Peter Zijlstra]

On Sat, Feb 11, 2017 at 08:15:26AM +0100, luca abeni wrote:
> Hi Daniel,
> 
> On Fri, 10 Feb 2017 20:48:11 +0100
> Daniel Bristot de Oliveira <bristot@redhat.com> wrote:
> 
> > During the activation, CBS checks if it can reuse the current task's
> > runtime and period. If the deadline of the task is in the past, CBS
> > cannot use the runtime, and so it replenishes the task. This rule
> > works fine for implicit deadline tasks (deadline == period), and the
> > CBS was designed for implicit deadline tasks. However, a task with
> > constrained deadline (deadine < period) might be awakened after the
> > deadline, but before the next period. In this case, replenishing the
> > task would allow it to run for runtime / deadline. As in this case
> > deadline < period, CBS enables a task to run for more than the
> > runtime/period. In a very load system, this can cause the domino
> > effect, making other tasks to miss their deadlines.
> 
> I think you are right: SCHED_DEADLINE implements the original CBS
> algorithm here, but uses relative deadlines different from periods in
> other places (while the original algorithm only considered relative
> deadlines equal to periods).
> An this mix is dangerous... I think your fix is correct, and cures a
> real problem.
> 

Made that a "Reviewed-by:" tag in your name again.

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Peter Zijlstra]

On Fri, Feb 10, 2017 at 08:48:11PM +0100, Daniel Bristot de Oliveira wrote:
> +/* During the activation, CBS checks if it can reuse the current task's
> + * runtime and period. If the deadline of the task is in the past, CBS

Broken comment style, fixed that for you.

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Daniel Bristot de Oliveira]

On 02/13/2017 12:12 PM, Peter Zijlstra wrote:
> On Fri, Feb 10, 2017 at 08:48:11PM +0100, Daniel Bristot de Oliveira wrote:
>> +/* During the activation, CBS checks if it can reuse the current task's
>> + * runtime and period. If the deadline of the task is in the past, CBS
> Broken comment style, fixed that for you.

Thanks! I was thinking sending a v1 just because of this...

-- Daniel

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Steven Rostedt]

On Fri, 10 Feb 2017 20:48:11 +0100
Daniel Bristot de Oliveira <bristot@redhat.com> wrote:

>  
> +static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
> +{
> +	return dl_se->dl_runtime < dl_se->dl_period;
> +}
> +

Is it ever appropriate for a dl task to have runtime == period? What
purpose would that serve? Just run the task as FIFO higher than
everything else.

Or was this suppose to be dl_deadline < dl_period?

-- Steve

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Daniel Bristot de Oliveira]

On 02/13/2017 04:33 PM, Steven Rostedt wrote:
>> +static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
>> +{
>> +	return dl_se->dl_runtime < dl_se->dl_period;
>> +}
>> +
> Is it ever appropriate for a dl task to have runtime == period? What
> purpose would that serve? Just run the task as FIFO higher than
> everything else.
> 
> Or was this suppose to be dl_deadline < dl_period?

Oooooops, my bad :-(, this was supposed to be dl_deadline < dl_period.

I will send a v2 fixing it, along with the problem in the comment that
peterz mentioned before.

Thanks Steven!

-- Daniel

Re: [PATCH 2/2] sched/deadline: Throttle a constrained deadline task activated after the deadline

[Daniel Bristot de Oliveira]

On 02/13/2017 04:46 PM, Daniel Bristot de Oliveira wrote:
> On 02/13/2017 04:33 PM, Steven Rostedt wrote:
>>> +static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
>>> +{
>>> +	return dl_se->dl_runtime < dl_se->dl_period;
>>> +}
>>> +
>> Is it ever appropriate for a dl task to have runtime == period? What
>> purpose would that serve? Just run the task as FIFO higher than
>> everything else.
>>
>> Or was this suppose to be dl_deadline < dl_period?
> Oooooops, my bad :-(, this was supposed to be dl_deadline < dl_period.

As

+       if (dl_time_before(dl_se->deadline, rq_clock(rq)) &&
+           dl_time_before(rq_clock(rq), dl_next_period(dl_se))) {


is never true for implicit deadline tasks, the problem was not causing
further logical issues (apart from being conceptually obviously totally
wrong, and causes the check to be called without the need to... me--).

-- Daniel