nohz/full, sched/rt: Fix missed tick-reenabling bug in sched_can_stop_tick()
authorPeter Zijlstra <peterz@infradead.org>
Thu, 21 Apr 2016 16:03:15 +0000 (18:03 +0200)
committerIngo Molnar <mingo@kernel.org>
Thu, 28 Apr 2016 08:28:55 +0000 (10:28 +0200)
Chris Metcalf reported a that sched_can_stop_tick() sometimes fails to
re-enable the tick.

His observed problem is that rq->cfs.nr_running can be 1 even though
there are multiple runnable CFS tasks. This happens in the cgroup
case, in which case cfs.nr_running is the number of runnable entities
for that level.

If there is a single runnable cgroup (which can have an arbitrary
number of runnable child entries itself) rq->cfs.nr_running will be 1.

However, looking at that function I think there's more problems with it.

It seems to assume that if there's FIFO tasks, those will run. This is
incorrect. The FIFO task can have a lower prio than an RR task, in which
case the RR task will run.

So the whole fifo_nr_running test seems misplaced, it should go after
the rr_nr_running tests. That is, only if !rr_nr_running, can we use
fifo_nr_running like this.

Reported-by: Chris Metcalf <cmetcalf@mellanox.com>
Tested-by: Chris Metcalf <cmetcalf@mellanox.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Wanpeng Li <kernellwp@gmail.com>
Fixes: 76d92ac305f2 ("sched: Migrate sched to use new tick dependency mask model")
Link: http://lkml.kernel.org/r/20160421160315.GK24771@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/core.c

index 8b489fcac37bd9d829439feb08d8bac1354c7e71..d1f7149f870439d65b9cfcfbc27d4160bbb1672f 100644 (file)
@@ -596,17 +596,8 @@ bool sched_can_stop_tick(struct rq *rq)
                return false;
 
        /*
-        * FIFO realtime policy runs the highest priority task (after DEADLINE).
-        * Other runnable tasks are of a lower priority. The scheduler tick
-        * isn't needed.
-        */
-       fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running;
-       if (fifo_nr_running)
-               return true;
-
-       /*
-        * Round-robin realtime tasks time slice with other tasks at the same
-        * realtime priority.
+        * If there are more than one RR tasks, we need the tick to effect the
+        * actual RR behaviour.
         */
        if (rq->rt.rr_nr_running) {
                if (rq->rt.rr_nr_running == 1)
@@ -615,8 +606,20 @@ bool sched_can_stop_tick(struct rq *rq)
                        return false;
        }
 
-       /* Normal multitasking need periodic preemption checks */
-       if (rq->cfs.nr_running > 1)
+       /*
+        * If there's no RR tasks, but FIFO tasks, we can skip the tick, no
+        * forced preemption between FIFO tasks.
+        */
+       fifo_nr_running = rq->rt.rt_nr_running - rq->rt.rr_nr_running;
+       if (fifo_nr_running)
+               return true;
+
+       /*
+        * If there are no DL,RR/FIFO tasks, there must only be CFS tasks left;
+        * if there's more than one we need the tick for involuntary
+        * preemption.
+        */
+       if (rq->nr_running > 1)
                return false;
 
        return true;