1 files changed, 286 insertions, 54 deletions

diff --git a/kernel/sched.c b/kernel/sched.c
index 5f889d0cbfcc..18b95520a2e2 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -1478,6 +1478,7 @@ static inline void prepare_task_switch(runqueue_t *rq, task_t *next)
 
 /**
  * finish_task_switch - clean up after a task-switch
+ * @rq: runqueue associated with task-switch
  * @prev: the thread we just switched away from.
  *
  * finish_task_switch must be called after the context switch, paired
@@ -4779,7 +4780,7 @@ static int sd_parent_degenerate(struct sched_domain *sd,
  * Attach the domain 'sd' to 'cpu' as its base domain.  Callers must
  * hold the hotplug lock.
  */
-void cpu_attach_domain(struct sched_domain *sd, int cpu)
+static void cpu_attach_domain(struct sched_domain *sd, int cpu)
 {
         runqueue_t *rq = cpu_rq(cpu);
         struct sched_domain *tmp;
@@ -4802,7 +4803,7 @@ void cpu_attach_domain(struct sched_domain *sd, int cpu)
 }
 
 /* cpus with isolated domains */
-cpumask_t __devinitdata cpu_isolated_map = CPU_MASK_NONE;
+static cpumask_t __devinitdata cpu_isolated_map = CPU_MASK_NONE;
 
 /* Setup the mask of cpus configured for isolated domains */
 static int __init isolated_cpu_setup(char *str)
@@ -4830,8 +4831,8 @@ __setup ("isolcpus=", isolated_cpu_setup);
  * covered by the given span, and will set each group's ->cpumask correctly,
  * and ->cpu_power to 0.
  */
-void init_sched_build_groups(struct sched_group groups[],
-                        cpumask_t span, int (*group_fn)(int cpu))
+static void init_sched_build_groups(struct sched_group groups[], cpumask_t span,
+                                    int (*group_fn)(int cpu))
 {
         struct sched_group *first = NULL, *last = NULL;
         cpumask_t covered = CPU_MASK_NONE;
@@ -4864,12 +4865,85 @@ void init_sched_build_groups(struct sched_group groups[],
         last->next = first;
 }
 
+#define SD_NODES_PER_DOMAIN 16
 
-#ifdef ARCH_HAS_SCHED_DOMAIN
-extern void build_sched_domains(const cpumask_t *cpu_map);
-extern void arch_init_sched_domains(const cpumask_t *cpu_map);
-extern void arch_destroy_sched_domains(const cpumask_t *cpu_map);
-#else
+#ifdef CONFIG_NUMA
+/**
+ * find_next_best_node - find the next node to include in a sched_domain
+ * @node: node whose sched_domain we're building
+ * @used_nodes: nodes already in the sched_domain
+ *
+ * Find the next node to include in a given scheduling domain.  Simply
+ * finds the closest node not already in the @used_nodes map.
+ *
+ * Should use nodemask_t.
+ */
+static int find_next_best_node(int node, unsigned long *used_nodes)
+{
+        int i, n, val, min_val, best_node = 0;
+
+        min_val = INT_MAX;
+
+        for (i = 0; i < MAX_NUMNODES; i++) {
+                /* Start at @node */
+                n = (node + i) % MAX_NUMNODES;
+
+                if (!nr_cpus_node(n))
+                        continue;
+
+                /* Skip already used nodes */
+                if (test_bit(n, used_nodes))
+                        continue;
+
+                /* Simple min distance search */
+                val = node_distance(node, n);
+
+                if (val < min_val) {
+                        min_val = val;
+                        best_node = n;
+                }
+        }
+
+        set_bit(best_node, used_nodes);
+        return best_node;
+}
+
+/**
+ * sched_domain_node_span - get a cpumask for a node's sched_domain
+ * @node: node whose cpumask we're constructing
+ * @size: number of nodes to include in this span
+ *
+ * Given a node, construct a good cpumask for its sched_domain to span.  It
+ * should be one that prevents unnecessary balancing, but also spreads tasks
+ * out optimally.
+ */
+static cpumask_t sched_domain_node_span(int node)
+{
+        int i;
+        cpumask_t span, nodemask;
+        DECLARE_BITMAP(used_nodes, MAX_NUMNODES);
+
+        cpus_clear(span);
+        bitmap_zero(used_nodes, MAX_NUMNODES);
+
+        nodemask = node_to_cpumask(node);
+        cpus_or(span, span, nodemask);
+        set_bit(node, used_nodes);
+
+        for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
+                int next_node = find_next_best_node(node, used_nodes);
+                nodemask = node_to_cpumask(next_node);
+                cpus_or(span, span, nodemask);
+        }
+
+        return span;
+}
+#endif
+
+/*
+ * At the moment, CONFIG_SCHED_SMT is never defined, but leave it in so we
+ * can switch it on easily if needed.
+ */
 #ifdef CONFIG_SCHED_SMT
 static DEFINE_PER_CPU(struct sched_domain, cpu_domains);
 static struct sched_group sched_group_cpus[NR_CPUS];
@@ -4891,36 +4965,20 @@ static int cpu_to_phys_group(int cpu)
 }
 
 #ifdef CONFIG_NUMA
-
-static DEFINE_PER_CPU(struct sched_domain, node_domains);
-static struct sched_group sched_group_nodes[MAX_NUMNODES];
-static int cpu_to_node_group(int cpu)
-{
-        return cpu_to_node(cpu);
-}
-#endif
-
-#if defined(CONFIG_SCHED_SMT) && defined(CONFIG_NUMA)
 /*
- * The domains setup code relies on siblings not spanning
- * multiple nodes. Make sure the architecture has a proper
- * siblings map:
+ * The init_sched_build_groups can't handle what we want to do with node
+ * groups, so roll our own. Now each node has its own list of groups which
+ * gets dynamically allocated.
  */
-static void check_sibling_maps(void)
-{
-        int i, j;
+static DEFINE_PER_CPU(struct sched_domain, node_domains);
+static struct sched_group **sched_group_nodes_bycpu[NR_CPUS];
 
-        for_each_online_cpu(i) {
-                for_each_cpu_mask(j, cpu_sibling_map[i]) {
-                        if (cpu_to_node(i) != cpu_to_node(j)) {
-                                printk(KERN_INFO "warning: CPU %d siblings map "
-                                        "to different node - isolating "
-                                        "them.\n", i);
-                                cpu_sibling_map[i] = cpumask_of_cpu(i);
-                                break;
-                        }
-                }
-        }
+static DEFINE_PER_CPU(struct sched_domain, allnodes_domains);
+static struct sched_group *sched_group_allnodes_bycpu[NR_CPUS];
+
+static int cpu_to_allnodes_group(int cpu)
+{
+        return cpu_to_node(cpu);
 }
 #endif
 
@@ -4928,9 +4986,24 @@ static void check_sibling_maps(void)
  * Build sched domains for a given set of cpus and attach the sched domains
  * to the individual cpus
  */
-static void build_sched_domains(const cpumask_t *cpu_map)
+void build_sched_domains(const cpumask_t *cpu_map)
 {
         int i;
+#ifdef CONFIG_NUMA
+        struct sched_group **sched_group_nodes = NULL;
+        struct sched_group *sched_group_allnodes = NULL;
+
+        /*
+         * Allocate the per-node list of sched groups
+         */
+        sched_group_nodes = kmalloc(sizeof(struct sched_group*)*MAX_NUMNODES,
+                                           GFP_ATOMIC);
+        if (!sched_group_nodes) {
+                printk(KERN_WARNING "Can not alloc sched group node list\n");
+                return;
+        }
+        sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes;
+#endif
 
         /*
          * Set up domains for cpus specified by the cpu_map.
@@ -4943,11 +5016,35 @@ static void build_sched_domains(const cpumask_t *cpu_map)
                 cpus_and(nodemask, nodemask, *cpu_map);
 
 #ifdef CONFIG_NUMA
+                if (cpus_weight(*cpu_map)
+                                > SD_NODES_PER_DOMAIN*cpus_weight(nodemask)) {
+                        if (!sched_group_allnodes) {
+                                sched_group_allnodes
+                                        = kmalloc(sizeof(struct sched_group)
+                                                        * MAX_NUMNODES,
+                                                  GFP_KERNEL);
+                                if (!sched_group_allnodes) {
+                                        printk(KERN_WARNING
+                                        "Can not alloc allnodes sched group\n");
+                                        break;
+                                }
+                                sched_group_allnodes_bycpu[i]
+                                                = sched_group_allnodes;
+                        }
+                        sd = &per_cpu(allnodes_domains, i);
+                        *sd = SD_ALLNODES_INIT;
+                        sd->span = *cpu_map;
+                        group = cpu_to_allnodes_group(i);
+                        sd->groups = &sched_group_allnodes[group];
+                        p = sd;
+                } else
+                        p = NULL;
+
                 sd = &per_cpu(node_domains, i);
-                group = cpu_to_node_group(i);
                 *sd = SD_NODE_INIT;
-                sd->span = *cpu_map;
-                sd->groups = &sched_group_nodes[group];
+                sd->span = sched_domain_node_span(cpu_to_node(i));
+                sd->parent = p;
+                cpus_and(sd->span, sd->span, *cpu_map);
 #endif
 
                 p = sd;
@@ -4972,7 +5069,7 @@ static void build_sched_domains(const cpumask_t *cpu_map)
 
 #ifdef CONFIG_SCHED_SMT
         /* Set up CPU (sibling) groups */
-        for_each_online_cpu(i) {
+        for_each_cpu_mask(i, *cpu_map) {
                 cpumask_t this_sibling_map = cpu_sibling_map[i];
                 cpus_and(this_sibling_map, this_sibling_map, *cpu_map);
                 if (i != first_cpu(this_sibling_map))
@@ -4997,8 +5094,77 @@ static void build_sched_domains(const cpumask_t *cpu_map)
 
 #ifdef CONFIG_NUMA
         /* Set up node groups */
-        init_sched_build_groups(sched_group_nodes, *cpu_map,
-                                        &cpu_to_node_group);
+        if (sched_group_allnodes)
+                init_sched_build_groups(sched_group_allnodes, *cpu_map,
+                                        &cpu_to_allnodes_group);
+
+        for (i = 0; i < MAX_NUMNODES; i++) {
+                /* Set up node groups */
+                struct sched_group *sg, *prev;
+                cpumask_t nodemask = node_to_cpumask(i);
+                cpumask_t domainspan;
+                cpumask_t covered = CPU_MASK_NONE;
+                int j;
+
+                cpus_and(nodemask, nodemask, *cpu_map);
+                if (cpus_empty(nodemask)) {
+                        sched_group_nodes[i] = NULL;
+                        continue;
+                }
+
+                domainspan = sched_domain_node_span(i);
+                cpus_and(domainspan, domainspan, *cpu_map);
+
+                sg = kmalloc(sizeof(struct sched_group), GFP_KERNEL);
+                sched_group_nodes[i] = sg;
+                for_each_cpu_mask(j, nodemask) {
+                        struct sched_domain *sd;
+                        sd = &per_cpu(node_domains, j);
+                        sd->groups = sg;
+                        if (sd->groups == NULL) {
+                                /* Turn off balancing if we have no groups */
+                                sd->flags = 0;
+                        }
+                }
+                if (!sg) {
+                        printk(KERN_WARNING
+                        "Can not alloc domain group for node %d\n", i);
+                        continue;
+                }
+                sg->cpu_power = 0;
+                sg->cpumask = nodemask;
+                cpus_or(covered, covered, nodemask);
+                prev = sg;
+
+                for (j = 0; j < MAX_NUMNODES; j++) {
+                        cpumask_t tmp, notcovered;
+                        int n = (i + j) % MAX_NUMNODES;
+
+                        cpus_complement(notcovered, covered);
+                        cpus_and(tmp, notcovered, *cpu_map);
+                        cpus_and(tmp, tmp, domainspan);
+                        if (cpus_empty(tmp))
+                                break;
+
+                        nodemask = node_to_cpumask(n);
+                        cpus_and(tmp, tmp, nodemask);
+                        if (cpus_empty(tmp))
+                                continue;
+
+                        sg = kmalloc(sizeof(struct sched_group), GFP_KERNEL);
+                        if (!sg) {
+                                printk(KERN_WARNING
+                                "Can not alloc domain group for node %d\n", j);
+                                break;
+                        }
+                        sg->cpu_power = 0;
+                        sg->cpumask = tmp;
+                        cpus_or(covered, covered, tmp);
+                        prev->next = sg;
+                        prev = sg;
+                }
+                prev->next = sched_group_nodes[i];
+        }
 #endif
 
         /* Calculate CPU power for physical packages and nodes */
@@ -5017,14 +5183,46 @@ static void build_sched_domains(const cpumask_t *cpu_map)
                 sd->groups->cpu_power = power;
 
 #ifdef CONFIG_NUMA
-                if (i == first_cpu(sd->groups->cpumask)) {
-                        /* Only add "power" once for each physical package. */
-                        sd = &per_cpu(node_domains, i);
-                        sd->groups->cpu_power += power;
+                sd = &per_cpu(allnodes_domains, i);
+                if (sd->groups) {
+                        power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE *
+                                (cpus_weight(sd->groups->cpumask)-1) / 10;
+                        sd->groups->cpu_power = power;
                 }
 #endif
         }
 
+#ifdef CONFIG_NUMA
+        for (i = 0; i < MAX_NUMNODES; i++) {
+                struct sched_group *sg = sched_group_nodes[i];
+                int j;
+
+                if (sg == NULL)
+                        continue;
+next_sg:
+                for_each_cpu_mask(j, sg->cpumask) {
+                        struct sched_domain *sd;
+                        int power;
+
+                        sd = &per_cpu(phys_domains, j);
+                        if (j != first_cpu(sd->groups->cpumask)) {
+                                /*
+                                 * Only add "power" once for each
+                                 * physical package.
+                                 */
+                                continue;
+                        }
+                        power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE *
+                                (cpus_weight(sd->groups->cpumask)-1) / 10;
+
+                        sg->cpu_power += power;
+                }
+                sg = sg->next;
+                if (sg != sched_group_nodes[i])
+                        goto next_sg;
+        }
+#endif
+
         /* Attach the domains */
         for_each_cpu_mask(i, *cpu_map) {
                 struct sched_domain *sd;
@@ -5039,13 +5237,10 @@ static void build_sched_domains(const cpumask_t *cpu_map)
 /*
  * Set up scheduler domains and groups.  Callers must hold the hotplug lock.
  */
-static void arch_init_sched_domains(cpumask_t *cpu_map)
+static void arch_init_sched_domains(const cpumask_t *cpu_map)
 {
         cpumask_t cpu_default_map;
 
-#if defined(CONFIG_SCHED_SMT) && defined(CONFIG_NUMA)
-        check_sibling_maps();
-#endif
         /*
          * Setup mask for cpus without special case scheduling requirements.
          * For now this just excludes isolated cpus, but could be used to
@@ -5058,10 +5253,47 @@ static void arch_init_sched_domains(cpumask_t *cpu_map)
 
 static void arch_destroy_sched_domains(const cpumask_t *cpu_map)
 {
-        /* Do nothing: everything is statically allocated. */
-}
+#ifdef CONFIG_NUMA
+        int i;
+        int cpu;
+
+        for_each_cpu_mask(cpu, *cpu_map) {
+                struct sched_group *sched_group_allnodes
+                        = sched_group_allnodes_bycpu[cpu];
+                struct sched_group **sched_group_nodes
+                        = sched_group_nodes_bycpu[cpu];
+
+                if (sched_group_allnodes) {
+                        kfree(sched_group_allnodes);
+                        sched_group_allnodes_bycpu[cpu] = NULL;
+                }
+
+                if (!sched_group_nodes)
+                        continue;
+
+                for (i = 0; i < MAX_NUMNODES; i++) {
+                        cpumask_t nodemask = node_to_cpumask(i);
+                        struct sched_group *oldsg, *sg = sched_group_nodes[i];
 
-#endif /* ARCH_HAS_SCHED_DOMAIN */
+                        cpus_and(nodemask, nodemask, *cpu_map);
+                        if (cpus_empty(nodemask))
+                                continue;
+
+                        if (sg == NULL)
+                                continue;
+                        sg = sg->next;
+next_sg:
+                        oldsg = sg;
+                        sg = sg->next;
+                        kfree(oldsg);
+                        if (oldsg != sched_group_nodes[i])
+                                goto next_sg;
+                }
+                kfree(sched_group_nodes);
+                sched_group_nodes_bycpu[cpu] = NULL;
+        }
+#endif
+}
 
 /*
  * Detach sched domains from a group of cpus specified in cpu_map