[PATCH] sched_domai: Allocate sched_group structures dynamically
authorSrivatsa Vaddagiri <vatsa@in.ibm.com>
Tue, 27 Jun 2006 09:54:41 +0000 (02:54 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Wed, 28 Jun 2006 00:32:45 +0000 (17:32 -0700)
As explained here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=114327539012323&w=2

there is a problem with sharing sched_group structures between two
separate sched_group structures for different sched_domains.

The patch has been tested and found to avoid the kernel lockup problem
described in above URL.

Signed-off-by: Srivatsa Vaddagiri <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: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/sched.c

index ee4211bd40c5568830a9c4b5ec3a32b675d8cc28..122b75584a138d9fd3bb063f30f85d226b8855b9 100644 (file)
@@ -5747,7 +5747,7 @@ static int cpu_to_cpu_group(int cpu)
 
 #ifdef CONFIG_SCHED_MC
 static DEFINE_PER_CPU(struct sched_domain, core_domains);
-static struct sched_group sched_group_core[NR_CPUS];
+static struct sched_group *sched_group_core_bycpu[NR_CPUS];
 #endif
 
 #if defined(CONFIG_SCHED_MC) && defined(CONFIG_SCHED_SMT)
@@ -5763,7 +5763,7 @@ static int cpu_to_core_group(int cpu)
 #endif
 
 static DEFINE_PER_CPU(struct sched_domain, phys_domains);
-static struct sched_group sched_group_phys[NR_CPUS];
+static struct sched_group *sched_group_phys_bycpu[NR_CPUS];
 static int cpu_to_phys_group(int cpu)
 {
 #if defined(CONFIG_SCHED_MC)
@@ -5823,9 +5823,9 @@ next_sg:
 /* Free memory allocated for various sched_group structures */
 static void free_sched_groups(const cpumask_t *cpu_map)
 {
+       int cpu;
 #ifdef CONFIG_NUMA
        int i;
-       int cpu;
 
        for_each_cpu_mask(cpu, *cpu_map) {
                struct sched_group *sched_group_allnodes
@@ -5863,6 +5863,18 @@ next_sg:
                sched_group_nodes_bycpu[cpu] = NULL;
        }
 #endif
+       for_each_cpu_mask(cpu, *cpu_map) {
+               if (sched_group_phys_bycpu[cpu]) {
+                       kfree(sched_group_phys_bycpu[cpu]);
+                       sched_group_phys_bycpu[cpu] = NULL;
+               }
+#ifdef CONFIG_SCHED_MC
+               if (sched_group_core_bycpu[cpu]) {
+                       kfree(sched_group_core_bycpu[cpu]);
+                       sched_group_core_bycpu[cpu] = NULL;
+               }
+#endif
+       }
 }
 
 /*
@@ -5872,6 +5884,10 @@ next_sg:
 static int build_sched_domains(const cpumask_t *cpu_map)
 {
        int i;
+       struct sched_group *sched_group_phys = NULL;
+#ifdef CONFIG_SCHED_MC
+       struct sched_group *sched_group_core = NULL;
+#endif
 #ifdef CONFIG_NUMA
        struct sched_group **sched_group_nodes = NULL;
        struct sched_group *sched_group_allnodes = NULL;
@@ -5930,6 +5946,18 @@ static int build_sched_domains(const cpumask_t *cpu_map)
                cpus_and(sd->span, sd->span, *cpu_map);
 #endif
 
+               if (!sched_group_phys) {
+                       sched_group_phys
+                               = kmalloc(sizeof(struct sched_group) * NR_CPUS,
+                                         GFP_KERNEL);
+                       if (!sched_group_phys) {
+                               printk (KERN_WARNING "Can not alloc phys sched"
+                                                    "group\n");
+                               goto error;
+                       }
+                       sched_group_phys_bycpu[i] = sched_group_phys;
+               }
+
                p = sd;
                sd = &per_cpu(phys_domains, i);
                group = cpu_to_phys_group(i);
@@ -5939,6 +5967,18 @@ static int build_sched_domains(const cpumask_t *cpu_map)
                sd->groups = &sched_group_phys[group];
 
 #ifdef CONFIG_SCHED_MC
+               if (!sched_group_core) {
+                       sched_group_core
+                               = kmalloc(sizeof(struct sched_group) * NR_CPUS,
+                                         GFP_KERNEL);
+                       if (!sched_group_core) {
+                               printk (KERN_WARNING "Can not alloc core sched"
+                                                    "group\n");
+                               goto error;
+                       }
+                       sched_group_core_bycpu[i] = sched_group_core;
+               }
+
                p = sd;
                sd = &per_cpu(core_domains, i);
                group = cpu_to_core_group(i);
@@ -6134,11 +6174,9 @@ static int build_sched_domains(const cpumask_t *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.