Bjoern's Dissertation Code with Priority Donationwip-splitting-omlp-jerickso

1 files changed, 115 insertions, 12 deletions

diff --git a/kernel/sched.c b/kernel/sched.c
index dc85ceb90832..1f5327f8c012 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -79,6 +79,11 @@
 #include "sched_cpupri.h"
 #include "workqueue_sched.h"
 
+#include <litmus/sched_trace.h>
+#include <litmus/trace.h>
+
+static void litmus_tick(struct rq*, struct task_struct*);
+
 #define CREATE_TRACE_POINTS
 #include <trace/events/sched.h>
 
@@ -405,6 +410,12 @@ struct rt_rq {
 #endif
 };
 
+/* Litmus related fields in a runqueue */
+struct litmus_rq {
+        unsigned long nr_running;
+        struct task_struct *prev;
+};
+
 #ifdef CONFIG_SMP
 
 /*
@@ -471,6 +482,7 @@ struct rq {
 
         struct cfs_rq cfs;
         struct rt_rq rt;
+        struct litmus_rq litmus;
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
         /* list of leaf cfs_rq on this cpu: */
@@ -566,8 +578,14 @@ void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
          * A queue event has occurred, and we're going to schedule.  In
          * this case, we can save a useless back to back clock update.
          */
+        /* LITMUS^RT: turning off the clock update is buggy in Linux 2.6.36;
+         * the scheduler can "forget" to renable the runqueue clock in some
+         * cases. LITMUS^RT amplifies the effects of this problem. Hence, we
+         * turn it off to avoid stalling clocks. */
+        /*
         if (test_tsk_need_resched(p))
                 rq->skip_clock_update = 1;
+        */
 }
 
 static inline int cpu_of(struct rq *rq)
@@ -1042,6 +1060,7 @@ static enum hrtimer_restart hrtick(struct hrtimer *timer)
         raw_spin_lock(&rq->lock);
         update_rq_clock(rq);
         rq->curr->sched_class->task_tick(rq, rq->curr, 1);
+        litmus_tick(rq, rq->curr);
         raw_spin_unlock(&rq->lock);
 
         return HRTIMER_NORESTART;
@@ -1840,7 +1859,7 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
 
 static const struct sched_class rt_sched_class;
 
-#define sched_class_highest (&rt_sched_class)
+#define sched_class_highest (&litmus_sched_class)
 #define for_each_class(class) \
    for (class = sched_class_highest; class; class = class->next)
 
@@ -1920,6 +1939,7 @@ static void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
 #include "sched_idletask.c"
 #include "sched_fair.c"
 #include "sched_rt.c"
+#include "../litmus/sched_litmus.c"
 #ifdef CONFIG_SCHED_DEBUG
 # include "sched_debug.c"
 #endif
@@ -2352,6 +2372,9 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state,
         unsigned long en_flags = ENQUEUE_WAKEUP;
         struct rq *rq;
 
+        if (is_realtime(p))
+                TRACE_TASK(p, "try_to_wake_up() state:%d\n", p->state);
+
         this_cpu = get_cpu();
 
         smp_wmb();
@@ -2366,7 +2389,7 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state,
         orig_cpu = cpu;
 
 #ifdef CONFIG_SMP
-        if (unlikely(task_running(rq, p)))
+        if (unlikely(task_running(rq, p)) || is_realtime(p))
                 goto out_activate;
 
         /*
@@ -2428,6 +2451,8 @@ out_activate:
 out_running:
         ttwu_post_activation(p, rq, wake_flags, success);
 out:
+        if (is_realtime(p))
+                TRACE_TASK(p, "try_to_wake_up() done state:%d\n", p->state);
         task_rq_unlock(rq, &flags);
         put_cpu();
 
@@ -2532,7 +2557,8 @@ void sched_fork(struct task_struct *p, int clone_flags)
          * Revert to default priority/policy on fork if requested.
          */
         if (unlikely(p->sched_reset_on_fork)) {
-                if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) {
+                if (p->policy == SCHED_FIFO || p->policy == SCHED_RR ||
+                    p->policy == SCHED_LITMUS) {
                         p->policy = SCHED_NORMAL;
                         p->normal_prio = p->static_prio;
                 }
@@ -2748,6 +2774,8 @@ static void finish_task_switch(struct rq *rq, struct task_struct *prev)
          */
         prev_state = prev->state;
         finish_arch_switch(prev);
+        litmus->finish_switch(prev);
+        prev->rt_param.stack_in_use = NO_CPU;
 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
         local_irq_disable();
 #endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
@@ -2777,6 +2805,15 @@ static inline void pre_schedule(struct rq *rq, struct task_struct *prev)
 {
         if (prev->sched_class->pre_schedule)
                 prev->sched_class->pre_schedule(rq, prev);
+
+        /* LITMUS^RT not very clean hack: we need to save the prev task
+         * as our scheduling decision rely on it (as we drop the rq lock
+         * something in prev can change...); there is no way to escape
+         * this ack apart from modifying pick_nex_task(rq, _prev_) or
+         * falling back on the previous solution of decoupling
+         * scheduling decisions
+         */
+        rq->litmus.prev = prev;
 }
 
 /* rq->lock is NOT held, but preemption is disabled */
@@ -3578,18 +3615,26 @@ void scheduler_tick(void)
 
         sched_clock_tick();
 
+        TS_TICK_START(current);
+
         raw_spin_lock(&rq->lock);
         update_rq_clock(rq);
         update_cpu_load_active(rq);
         curr->sched_class->task_tick(rq, curr, 0);
+
+        /* litmus_tick may force current to resched */
+        litmus_tick(rq, curr);
+
         raw_spin_unlock(&rq->lock);
 
         perf_event_task_tick(curr);
 
 #ifdef CONFIG_SMP
         rq->idle_at_tick = idle_cpu(cpu);
-        trigger_load_balance(rq, cpu);
+        if (!is_realtime(current))
+                trigger_load_balance(rq, cpu);
 #endif
+        TS_TICK_END(current);
 }
 
 notrace unsigned long get_parent_ip(unsigned long addr)
@@ -3716,12 +3761,20 @@ pick_next_task(struct rq *rq)
         /*
          * Optimization: we know that if all tasks are in
          * the fair class we can call that function directly:
-         */
-        if (likely(rq->nr_running == rq->cfs.nr_running)) {
+
+         * NOT IN LITMUS^RT!
+
+         * This breaks many assumptions in the plugins.
+         * Do not uncomment without thinking long and hard
+         * about how this affects global plugins such as GSN-EDF.
+
+        if (rq->nr_running == rq->cfs.nr_running) {
+                TRACE("taking shortcut in pick_next_task()\n");
                 p = fair_sched_class.pick_next_task(rq);
                 if (likely(p))
                         return p;
         }
+        */
 
         class = sched_class_highest;
         for ( ; ; ) {
@@ -3748,6 +3801,7 @@ asmlinkage void __sched schedule(void)
 
 need_resched:
         preempt_disable();
+        sched_state_entered_schedule();
         cpu = smp_processor_id();
         rq = cpu_rq(cpu);
         rcu_note_context_switch(cpu);
@@ -3755,6 +3809,8 @@ need_resched:
 
         release_kernel_lock(prev);
 need_resched_nonpreemptible:
+        TS_SCHED_START;
+        sched_trace_task_switch_away(prev);
 
         schedule_debug(prev);
 
@@ -3803,7 +3859,10 @@ need_resched_nonpreemptible:
                 rq->curr = next;
                 ++*switch_count;
 
+                TS_SCHED_END(next);
+                TS_CXS_START(next);
                 context_switch(rq, prev, next); /* unlocks the rq */
+                TS_CXS_END(current);
                 /*
                  * The context switch have flipped the stack from under us
                  * and restored the local variables which were saved when
@@ -3812,17 +3871,23 @@ need_resched_nonpreemptible:
                  */
                 cpu = smp_processor_id();
                 rq = cpu_rq(cpu);
-        } else
+        } else {
+                TS_SCHED_END(prev);
                 raw_spin_unlock_irq(&rq->lock);
+        }
+
+        sched_trace_task_switch_to(current);
 
         post_schedule(rq);
 
-        if (unlikely(reacquire_kernel_lock(prev)))
+        if (sched_state_validate_switch() || unlikely(reacquire_kernel_lock(prev)))
                 goto need_resched_nonpreemptible;
 
         preempt_enable_no_resched();
         if (need_resched())
                 goto need_resched;
+
+        srp_ceiling_block();
 }
 EXPORT_SYMBOL(schedule);
 
@@ -4108,6 +4173,17 @@ void complete_all(struct completion *x)
 }
 EXPORT_SYMBOL(complete_all);
 
+void complete_n(struct completion *x, int n)
+{
+        unsigned long flags;
+
+        spin_lock_irqsave(&x->wait.lock, flags);
+        x->done += n;
+        __wake_up_common(&x->wait, TASK_NORMAL, n, 0, NULL);
+        spin_unlock_irqrestore(&x->wait.lock, flags);
+}
+EXPORT_SYMBOL(complete_n);
+
 static inline long __sched
 do_wait_for_common(struct completion *x, long timeout, int state)
 {
@@ -4550,7 +4626,9 @@ __setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
         p->normal_prio = normal_prio(p);
         /* we are holding p->pi_lock already */
         p->prio = rt_mutex_getprio(p);
-        if (rt_prio(p->prio))
+        if (p->policy == SCHED_LITMUS)
+                p->sched_class = &litmus_sched_class;
+        else if (rt_prio(p->prio))
                 p->sched_class = &rt_sched_class;
         else
                 p->sched_class = &fair_sched_class;
@@ -4595,7 +4673,7 @@ recheck:
 
                 if (policy != SCHED_FIFO && policy != SCHED_RR &&
                                 policy != SCHED_NORMAL && policy != SCHED_BATCH &&
-                                policy != SCHED_IDLE)
+                                policy != SCHED_IDLE && policy != SCHED_LITMUS)
                         return -EINVAL;
         }
 
@@ -4610,6 +4688,8 @@ recheck:
                 return -EINVAL;
         if (rt_policy(policy) != (param->sched_priority != 0))
                 return -EINVAL;
+        if (policy == SCHED_LITMUS && policy == p->policy)
+                return -EINVAL;
 
         /*
          * Allow unprivileged RT tasks to decrease priority:
@@ -4650,6 +4730,12 @@ recheck:
                         return retval;
         }
 
+        if (policy == SCHED_LITMUS) {
+                retval = litmus_admit_task(p);
+                if (retval)
+                        return retval;
+        }
+
         /*
          * make sure no PI-waiters arrive (or leave) while we are
          * changing the priority of the task:
@@ -4692,10 +4778,19 @@ recheck:
 
         p->sched_reset_on_fork = reset_on_fork;
 
+        if (p->policy == SCHED_LITMUS)
+                litmus_exit_task(p);
+
         oldprio = p->prio;
         prev_class = p->sched_class;
         __setscheduler(rq, p, policy, param->sched_priority);
 
+        if (policy == SCHED_LITMUS) {
+                p->rt_param.stack_in_use = running ? rq->cpu : NO_CPU;
+                p->rt_param.present = running;
+                litmus->task_new(p, on_rq, running);
+        }
+
         if (running)
                 p->sched_class->set_curr_task(rq);
         if (on_rq) {
@@ -4755,6 +4850,13 @@ do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
         if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
                 return -EFAULT;
 
+#ifdef CONFIG_LITMUS_LOCKING
+        /* Hack to allow plugin to call into schedule
+         * prio to a setscheduler() call. */
+        if (is_realtime(current))
+                litmus->pre_setsched(current, policy);
+#endif
+
         rcu_read_lock();
         retval = -ESRCH;
         p = find_process_by_pid(pid);
@@ -4865,10 +4967,11 @@ long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
         rcu_read_lock();
 
         p = find_process_by_pid(pid);
-        if (!p) {
+        /* Don't set affinity if task not found and for LITMUS tasks */
+        if (!p || is_realtime(p)) {
                 rcu_read_unlock();
                 put_online_cpus();
-                return -ESRCH;
+                return p ? -EPERM : -ESRCH;
         }
 
         /* Prevent p going away */