[PATCH] sched_domain: handle kmalloc failure
authorSrivatsa Vaddagiri <vatsa@in.ibm.com>
Tue, 27 Jun 2006 09:54:38 +0000 (02:54 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Wed, 28 Jun 2006 00:32:45 +0000 (17:32 -0700)
Try to handle mem allocation failures in build_sched_domains by bailing out
and cleaning up thus-far allocated memory.  The patch has a direct consequence
that we disable load balancing completely (even at sibling level) upon *any*
memory allocation failure.

[Lee.Schermerhorn@hp.com: bugfix]
Signed-off-by: Srivatsa Vaddagir <vatsa@in.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
include/linux/sched.h
kernel/sched.c

index 74a1e39e0d3d30530c31ebd3da7a8256c6f0c136..ab8ffc54423a3ef55a620b6041b84476bac988ac 100644 (file)
@@ -639,7 +639,7 @@ struct sched_domain {
 #endif
 };
 
-extern void partition_sched_domains(cpumask_t *partition1,
+extern int partition_sched_domains(cpumask_t *partition1,
                                    cpumask_t *partition2);
 
 /*
index 0ec84f57695df9f0691f2f56cba8b4f6b9713227..77a2ec55ef7d140c87bb40f3e1fa68b6cb523906 100644 (file)
@@ -5820,11 +5820,56 @@ next_sg:
 }
 #endif
 
+/* Free memory allocated for various sched_group structures */
+static void free_sched_groups(const cpumask_t *cpu_map)
+{
+#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];
+
+                       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
+}
+
 /*
  * Build sched domains for a given set of cpus and attach the sched domains
  * to the individual cpus
  */
-void build_sched_domains(const cpumask_t *cpu_map)
+static int build_sched_domains(const cpumask_t *cpu_map)
 {
        int i;
 #ifdef CONFIG_NUMA
@@ -5834,11 +5879,11 @@ void build_sched_domains(const cpumask_t *cpu_map)
        /*
         * Allocate the per-node list of sched groups
         */
-       sched_group_nodes = kmalloc(sizeof(struct sched_group*)*MAX_NUMNODES,
+       sched_group_nodes = kzalloc(sizeof(struct sched_group*)*MAX_NUMNODES,
                                           GFP_ATOMIC);
        if (!sched_group_nodes) {
                printk(KERN_WARNING "Can not alloc sched group node list\n");
-               return;
+               return -ENOMEM;
        }
        sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes;
 #endif
@@ -5864,7 +5909,7 @@ void build_sched_domains(const cpumask_t *cpu_map)
                                if (!sched_group_allnodes) {
                                        printk(KERN_WARNING
                                        "Can not alloc allnodes sched group\n");
-                                       break;
+                                       goto error;
                                }
                                sched_group_allnodes_bycpu[i]
                                                = sched_group_allnodes;
@@ -5978,23 +6023,20 @@ void build_sched_domains(const cpumask_t *cpu_map)
                cpus_and(domainspan, domainspan, *cpu_map);
 
                sg = kmalloc(sizeof(struct sched_group), GFP_KERNEL);
+               if (!sg) {
+                       printk(KERN_WARNING "Can not alloc domain group for "
+                               "node %d\n", i);
+                       goto error;
+               }
                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;
+               sg->next = sg;
                cpus_or(covered, covered, nodemask);
                prev = sg;
 
@@ -6017,15 +6059,15 @@ void build_sched_domains(const cpumask_t *cpu_map)
                        if (!sg) {
                                printk(KERN_WARNING
                                "Can not alloc domain group for node %d\n", j);
-                               break;
+                               goto error;
                        }
                        sg->cpu_power = 0;
                        sg->cpumask = tmp;
+                       sg->next = prev->next;
                        cpus_or(covered, covered, tmp);
                        prev->next = sg;
                        prev = sg;
                }
-               prev->next = sched_group_nodes[i];
        }
 #endif
 
@@ -6088,13 +6130,22 @@ void build_sched_domains(const cpumask_t *cpu_map)
         * Tune cache-hot values:
         */
        calibrate_migration_costs(cpu_map);
+
+       return 0;
+
+#ifdef CONFIG_NUMA
+error:
+       free_sched_groups(cpu_map);
+       return -ENOMEM;
+#endif
 }
 /*
  * Set up scheduler domains and groups.  Callers must hold the hotplug lock.
  */
-static void arch_init_sched_domains(const cpumask_t *cpu_map)
+static int arch_init_sched_domains(const cpumask_t *cpu_map)
 {
        cpumask_t cpu_default_map;
+       int err;
 
        /*
         * Setup mask for cpus without special case scheduling requirements.
@@ -6103,51 +6154,14 @@ static void arch_init_sched_domains(const cpumask_t *cpu_map)
         */
        cpus_andnot(cpu_default_map, *cpu_map, cpu_isolated_map);
 
-       build_sched_domains(&cpu_default_map);
+       err = build_sched_domains(&cpu_default_map);
+
+       return err;
 }
 
 static void arch_destroy_sched_domains(const cpumask_t *cpu_map)
 {
-#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];
-
-                       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
+       free_sched_groups(cpu_map);
 }
 
 /*
@@ -6172,9 +6186,10 @@ static void detach_destroy_domains(const cpumask_t *cpu_map)
  * correct sched domains
  * Call with hotplug lock held
  */
-void partition_sched_domains(cpumask_t *partition1, cpumask_t *partition2)
+int partition_sched_domains(cpumask_t *partition1, cpumask_t *partition2)
 {
        cpumask_t change_map;
+       int err = 0;
 
        cpus_and(*partition1, *partition1, cpu_online_map);
        cpus_and(*partition2, *partition2, cpu_online_map);
@@ -6183,9 +6198,11 @@ void partition_sched_domains(cpumask_t *partition1, cpumask_t *partition2)
        /* Detach sched domains from all of the affected cpus */
        detach_destroy_domains(&change_map);
        if (!cpus_empty(*partition1))
-               build_sched_domains(partition1);
-       if (!cpus_empty(*partition2))
-               build_sched_domains(partition2);
+               err = build_sched_domains(partition1);
+       if (!err && !cpus_empty(*partition2))
+               err = build_sched_domains(partition2);
+
+       return err;
 }
 
 #ifdef CONFIG_HOTPLUG_CPU