Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
authorLinus Torvalds <torvalds@linux-foundation.org>
Mon, 31 Dec 2018 17:54:17 +0000 (09:54 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 31 Dec 2018 17:54:17 +0000 (09:54 -0800)
Pull scheduler fix from Ingo Molnar:
 "This is a revert for a lockup in cgroups-intense workloads - the real
  fixes will come later"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/fair: Fix infinite loop in update_blocked_averages() by reverting a9e7f6544b9c

kernel/sched/fair.c

index d1907506318a4c7f22e71e9b8b63203fd20483e7..6483834f12788f80633c96e89d5cf874525a52e7 100644 (file)
@@ -352,10 +352,9 @@ static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
        }
 }
 
-/* Iterate thr' all leaf cfs_rq's on a runqueue */
-#define for_each_leaf_cfs_rq_safe(rq, cfs_rq, pos)                     \
-       list_for_each_entry_safe(cfs_rq, pos, &rq->leaf_cfs_rq_list,    \
-                                leaf_cfs_rq_list)
+/* Iterate through all leaf cfs_rq's on a runqueue: */
+#define for_each_leaf_cfs_rq(rq, cfs_rq) \
+       list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
 
 /* Do the two (enqueued) entities belong to the same group ? */
 static inline struct cfs_rq *
@@ -447,8 +446,8 @@ static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
 {
 }
 
-#define for_each_leaf_cfs_rq_safe(rq, cfs_rq, pos)     \
-               for (cfs_rq = &rq->cfs, pos = NULL; cfs_rq; cfs_rq = pos)
+#define for_each_leaf_cfs_rq(rq, cfs_rq)       \
+               for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
 
 static inline struct sched_entity *parent_entity(struct sched_entity *se)
 {
@@ -7647,27 +7646,10 @@ static inline bool others_have_blocked(struct rq *rq)
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 
-static inline bool cfs_rq_is_decayed(struct cfs_rq *cfs_rq)
-{
-       if (cfs_rq->load.weight)
-               return false;
-
-       if (cfs_rq->avg.load_sum)
-               return false;
-
-       if (cfs_rq->avg.util_sum)
-               return false;
-
-       if (cfs_rq->avg.runnable_load_sum)
-               return false;
-
-       return true;
-}
-
 static void update_blocked_averages(int cpu)
 {
        struct rq *rq = cpu_rq(cpu);
-       struct cfs_rq *cfs_rq, *pos;
+       struct cfs_rq *cfs_rq;
        const struct sched_class *curr_class;
        struct rq_flags rf;
        bool done = true;
@@ -7679,7 +7661,7 @@ static void update_blocked_averages(int cpu)
         * Iterates the task_group tree in a bottom up fashion, see
         * list_add_leaf_cfs_rq() for details.
         */
-       for_each_leaf_cfs_rq_safe(rq, cfs_rq, pos) {
+       for_each_leaf_cfs_rq(rq, cfs_rq) {
                struct sched_entity *se;
 
                /* throttled entities do not contribute to load */
@@ -7694,13 +7676,6 @@ static void update_blocked_averages(int cpu)
                if (se && !skip_blocked_update(se))
                        update_load_avg(cfs_rq_of(se), se, 0);
 
-               /*
-                * There can be a lot of idle CPU cgroups.  Don't let fully
-                * decayed cfs_rqs linger on the list.
-                */
-               if (cfs_rq_is_decayed(cfs_rq))
-                       list_del_leaf_cfs_rq(cfs_rq);
-
                /* Don't need periodic decay once load/util_avg are null */
                if (cfs_rq_has_blocked(cfs_rq))
                        done = false;
@@ -10570,10 +10545,10 @@ const struct sched_class fair_sched_class = {
 #ifdef CONFIG_SCHED_DEBUG
 void print_cfs_stats(struct seq_file *m, int cpu)
 {
-       struct cfs_rq *cfs_rq, *pos;
+       struct cfs_rq *cfs_rq;
 
        rcu_read_lock();
-       for_each_leaf_cfs_rq_safe(cpu_rq(cpu), cfs_rq, pos)
+       for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
                print_cfs_rq(m, cpu, cfs_rq);
        rcu_read_unlock();
 }