sched/numa: Use unsigned longs for numa group fault stats
authorMel Gorman <mgorman@suse.de>
Mon, 7 Oct 2013 10:29:40 +0000 (11:29 +0100)
committerIngo Molnar <mingo@kernel.org>
Wed, 9 Oct 2013 12:48:23 +0000 (14:48 +0200)
As Peter says "If you're going to hold locks you can also do away with all
that atomic_long_*() nonsense". Lock aquisition moved slightly to protect
the updates.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-63-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/fair.c

index e7884dc3416d27f5a77f67edb24a880574728bdc..5b2208e504a4d37317bfbeac2cb5c4f71dc7ebb0 100644 (file)
@@ -916,8 +916,8 @@ struct numa_group {
        struct list_head task_list;
 
        struct rcu_head rcu;
-       atomic_long_t total_faults;
-       atomic_long_t faults[0];
+       unsigned long total_faults;
+       unsigned long faults[0];
 };
 
 pid_t task_numa_group_id(struct task_struct *p)
@@ -944,8 +944,7 @@ static inline unsigned long group_faults(struct task_struct *p, int nid)
        if (!p->numa_group)
                return 0;
 
-       return atomic_long_read(&p->numa_group->faults[2*nid]) +
-              atomic_long_read(&p->numa_group->faults[2*nid+1]);
+       return p->numa_group->faults[2*nid] + p->numa_group->faults[2*nid+1];
 }
 
 /*
@@ -971,17 +970,10 @@ static inline unsigned long task_weight(struct task_struct *p, int nid)
 
 static inline unsigned long group_weight(struct task_struct *p, int nid)
 {
-       unsigned long total_faults;
-
-       if (!p->numa_group)
-               return 0;
-
-       total_faults = atomic_long_read(&p->numa_group->total_faults);
-
-       if (!total_faults)
+       if (!p->numa_group || !p->numa_group->total_faults)
                return 0;
 
-       return 1000 * group_faults(p, nid) / total_faults;
+       return 1000 * group_faults(p, nid) / p->numa_group->total_faults;
 }
 
 static unsigned long weighted_cpuload(const int cpu);
@@ -1397,9 +1389,9 @@ static void task_numa_placement(struct task_struct *p)
                        p->total_numa_faults += diff;
                        if (p->numa_group) {
                                /* safe because we can only change our own group */
-                               atomic_long_add(diff, &p->numa_group->faults[i]);
-                               atomic_long_add(diff, &p->numa_group->total_faults);
-                               group_faults += atomic_long_read(&p->numa_group->faults[i]);
+                               p->numa_group->faults[i] += diff;
+                               p->numa_group->total_faults += diff;
+                               group_faults += p->numa_group->faults[i];
                        }
                }
 
@@ -1475,7 +1467,7 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags,
 
        if (unlikely(!p->numa_group)) {
                unsigned int size = sizeof(struct numa_group) +
-                                   2*nr_node_ids*sizeof(atomic_long_t);
+                                   2*nr_node_ids*sizeof(unsigned long);
 
                grp = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
                if (!grp)
@@ -1487,9 +1479,9 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags,
                grp->gid = p->pid;
 
                for (i = 0; i < 2*nr_node_ids; i++)
-                       atomic_long_set(&grp->faults[i], p->numa_faults[i]);
+                       grp->faults[i] = p->numa_faults[i];
 
-               atomic_long_set(&grp->total_faults, p->total_numa_faults);
+               grp->total_faults = p->total_numa_faults;
 
                list_add(&p->numa_entry, &grp->task_list);
                grp->nr_tasks++;
@@ -1543,14 +1535,14 @@ unlock:
        if (!join)
                return;
 
+       double_lock(&my_grp->lock, &grp->lock);
+
        for (i = 0; i < 2*nr_node_ids; i++) {
-               atomic_long_sub(p->numa_faults[i], &my_grp->faults[i]);
-               atomic_long_add(p->numa_faults[i], &grp->faults[i]);
+               my_grp->faults[i] -= p->numa_faults[i];
+               grp->faults[i] += p->numa_faults[i];
        }
-       atomic_long_sub(p->total_numa_faults, &my_grp->total_faults);
-       atomic_long_add(p->total_numa_faults, &grp->total_faults);
-
-       double_lock(&my_grp->lock, &grp->lock);
+       my_grp->total_faults -= p->total_numa_faults;
+       grp->total_faults += p->total_numa_faults;
 
        list_move(&p->numa_entry, &grp->task_list);
        my_grp->nr_tasks--;
@@ -1571,12 +1563,11 @@ void task_numa_free(struct task_struct *p)
        void *numa_faults = p->numa_faults;
 
        if (grp) {
+               spin_lock(&grp->lock);
                for (i = 0; i < 2*nr_node_ids; i++)
-                       atomic_long_sub(p->numa_faults[i], &grp->faults[i]);
-
-               atomic_long_sub(p->total_numa_faults, &grp->total_faults);
+                       grp->faults[i] -= p->numa_faults[i];
+               grp->total_faults -= p->total_numa_faults;
 
-               spin_lock(&grp->lock);
                list_del(&p->numa_entry);
                grp->nr_tasks--;
                spin_unlock(&grp->lock);