On setting /proc/sys/kernel/sched_rt_runtime_us to -1, the intended
behavior is that rt-group throttling is disabled throughout. This
could be confirmed by trying to launch a real time task in a cgroup
which had zero rt runtime allotted to it.
This though was not happening when one tried to move a rt task into
a rt cgroup which had zero rt runtime. Ensure that this
tunable is honoured globally.
Signed-off-by: Dhaval Giani <[email protected]>
---
kernel/sched.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Index: master/kernel/sched.c
===================================================================
--- master.orig/kernel/sched.c
+++ master/kernel/sched.c
@@ -8542,7 +8542,8 @@ static int sched_rt_global_constraints(v
int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
{
/* Don't accept realtime tasks when there is no way for them to run */
- if (rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
+ if (rt_task(tsk) && rt_bandwidth_enabled() &&
+ tg->rt_bandwidth.rt_runtime == 0)
return 0;
return 1;