More PI issues with -rt
Hi,
I found that PI boosted SCHED_OTHER tasks behave like they have SCHED_FIFO policy, while PI boosted SCHED_RR tasks continue to behave like they have SCHED_RR policy. This didn't seem right
Does something like the following patch make sense?
-Dinakar
kernel/sched.c +2 -1
@@ -2986,8 +2986,9 @@* On PREEMPT_RT, boosted tasks will also get into this * branch and wont get their timeslice decreased until * they have done their work. +* Boosted SCHED_OTHER tasks round-robin as well */ -if ((p->policy ==SCHED_RR ) && !--p->time_slice) { +if ((p->policy !=SCHED_FIFO ) && !--p->time_slice) { p->time_slice =task_timeslice (p); p->first_time_slice =0 ;set_tsk_need_resched (p);
Ingo/David,
Resending this as I did not get any reply.
This time I am attaching the testcases where I am noticing this Any feedback is appreciated
-Dinakar
Quoted 27 lines · Dinakar Guniguntala
Hi,
I found that PI boosted
SCHED_OTHERtasks behave like they haveSCHED_FIFOpolicy, while PI boostedSCHED_RRtasks continue to behave like they haveSCHED_RRpolicy. This didn't seem rightDoes something like the following patch make sense?
-Dinakar
Index:
linux-2.6.14/kernel/sched.c
===================================================================kernel/sched.c +2
-1@@ -2986,8 +2986,9 @@* On PREEMPT_RT, boosted tasks will also get into this * branch and wont get their timeslice decreased until * they have done their work. +* Boosted SCHED_OTHER tasks round-robin as well */ -if ((p->policy ==SCHED_RR ) && !--p->time_slice) { +if ((p->policy !=SCHED_FIFO ) && !--p->time_slice) { p->time_slice =task_timeslice (p); p->first_time_slice =0 ;set_tsk_need_resched (p);