From: Jesse Barnes <jbarnes@engr.sgi.com>
This patch limits the cpu span of each node's scheduler domain to prevent
balancing across too many cpus. The cpus included in a node's domain are
determined by the SD_NODES_PER_DOMAIN define and the arch specific
sched_domain_node_span routine if ARCH_HAS_SCHED_DOMAIN is defined. If
ARCH_HAS_SCHED_DOMAIN is not defined, behavior is unchanged--all possible cpus
will be included in each node's scheduling domain. Currently, only ia64
provides an arch specific sched_domain_node_span routine.
Signed-off-by: Jesse Barnes <jbarnes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/arch/ia64/kernel/smpboot.c | 67 +++++++++++++++++++++++++++++++++++
25-akpm/include/asm-ia64/processor.h | 5 ++
25-akpm/kernel/sched.c | 17 +++++++-
3 files changed, 86 insertions(+), 3 deletions(-)
diff -puN arch/ia64/kernel/smpboot.c~sched-domain-node-span-4 arch/ia64/kernel/smpboot.c
--- 25/arch/ia64/kernel/smpboot.c~sched-domain-node-span-4 Tue Jul 27 14:21:37 2004
+++ 25-akpm/arch/ia64/kernel/smpboot.c Tue Jul 27 14:21:37 2004
@@ -716,3 +716,70 @@ init_smp_config(void)
printk(KERN_ERR "SMP: Can't set SAL AP Boot Rendezvous: %s\n",
ia64_sal_strerror(sal_ret));
}
+
+#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 __init 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 < numnodes; i++) {
+ /* Start at @node */
+ n = (node + i) % numnodes;
+
+ /* Skip already used nodes */
+ if (test_bit(n, used_nodes))
+ continue;
+
+ /* Simple min distance search */
+ val = node_distance(node, i);
+
+ 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.
+ */
+cpumask_t __init sched_domain_node_span(int node, int size)
+{
+ int i;
+ cpumask_t span;
+ DECLARE_BITMAP(used_nodes, MAX_NUMNODES);
+
+ cpus_clear(span);
+ bitmap_zero(used_nodes, MAX_NUMNODES);
+
+ for (i = 0; i < size; i++) {
+ int next_node = find_next_best_node(node, used_nodes);
+ cpus_or(span, span, node_to_cpumask(next_node));
+ }
+
+ return span;
+}
+#endif /* CONFIG_NUMA */
+
diff -puN include/asm-ia64/processor.h~sched-domain-node-span-4 include/asm-ia64/processor.h
--- 25/include/asm-ia64/processor.h~sched-domain-node-span-4 Tue Jul 27 14:21:37 2004
+++ 25-akpm/include/asm-ia64/processor.h Tue Jul 27 14:21:37 2004
@@ -334,6 +334,11 @@ struct task_struct;
/* Prepare to copy thread state - unlazy all lazy status */
#define prepare_to_copy(tsk) do { } while (0)
+#ifdef CONFIG_NUMA
+/* smpboot.c defines a numa specific scheduler domain routine */
+#define ARCH_HAS_SCHED_DOMAIN
+#endif
+
/*
* This is the mechanism for creating a new kernel thread.
*
diff -puN kernel/sched.c~sched-domain-node-span-4 kernel/sched.c
--- 25/kernel/sched.c~sched-domain-node-span-4 Tue Jul 27 14:21:37 2004
+++ 25-akpm/kernel/sched.c Tue Jul 27 14:34:44 2004
@@ -3667,8 +3667,13 @@ void cpu_attach_domain(struct sched_doma
}
#ifdef ARCH_HAS_SCHED_DOMAIN
-extern void __init arch_init_sched_domains(void);
+extern cpumask_t __init sched_domain_node_span(int node, int size);
#else
+static cpumask_t __init sched_domain_node_span(int node, int size)
+{
+ return cpu_possible_map;
+}
+#endif /* ARCH_HAS_SCHED_DOMAIN */
#ifdef CONFIG_SCHED_SMT
static DEFINE_PER_CPU(struct sched_domain, cpu_domains);
@@ -3691,6 +3696,10 @@ __init static int cpu_to_phys_group(int
}
#ifdef CONFIG_NUMA
+
+/* Number of nearby nodes in a node's scheduling domain */
+#define SD_NODES_PER_DOMAIN 4
+
static DEFINE_PER_CPU(struct sched_domain, node_domains);
static struct sched_group sched_group_nodes[MAX_NUMNODES];
__init static int cpu_to_node_group(int cpu)
@@ -3755,10 +3764,13 @@ __init static void arch_init_sched_domai
cpumask_t nodemask = node_to_cpumask(cpu_to_node(i));
#ifdef CONFIG_NUMA
+ if (i != first_cpu(sd->groups->cpumask))
+ continue;
sd = &per_cpu(node_domains, i);
group = cpu_to_node_group(i);
*sd = SD_NODE_INIT;
- sd->span = cpu_possible_map;
+ /* FIXME: should be multilevel, in arch code */
+ sd->span = sched_domain_node_span(i, SD_NODES_PER_DOMAIN);
sd->groups = &sched_group_nodes[group];
#endif
@@ -3845,7 +3857,6 @@ __init static void arch_init_sched_domai
cpu_attach_domain(sd, i);
}
}
-#endif /* ARCH_HAS_SCHED_DOMAIN */
#define SCHED_DOMAIN_DEBUG
#ifdef SCHED_DOMAIN_DEBUG
_