rcu: Control RCU_FANOUT_LEAF from boot-time parameter

Although making RCU_FANOUT_LEAF a kernel configuration parameter rather
than a fixed constant makes it easier for people to decrease cache-miss
overhead for large systems, it is of little help for people who must
run a single pre-built kernel binary.

This commit therefore allows the value of RCU_FANOUT_LEAF to be
increased (but not decreased!) via a boot-time parameter named
rcutree.rcu_fanout_leaf.

Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

Documentation/kernel-parameters.txt

@@ -2367,6 +2367,11 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 			Set maximum number of finished RCU callbacks to process
 			in one batch.
 
+	rcutree.fanout_leaf=	[KNL,BOOT]
+			Increase the number of CPUs assigned to each
+			leaf rcu_node structure.  Useful for very large
+			systems.
+
 	rcutree.qhimark=	[KNL,BOOT]
 			Set threshold of queued
 			RCU callbacks over which batch limiting is disabled.

kernel/rcutree.c

@@ -60,17 +60,10 @@
 
 /* Data structures. */
 
-static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];
+static struct lock_class_key rcu_node_class[RCU_NUM_LVLS];
 
 #define RCU_STATE_INITIALIZER(structname) { \
 	.level = { &structname##_state.node[0] }, \
-	.levelcnt = { \
-		NUM_RCU_LVL_0,  /* root of hierarchy. */ \
-		NUM_RCU_LVL_1, \
-		NUM_RCU_LVL_2, \
-		NUM_RCU_LVL_3, \
-		NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \
-	}, \
 	.fqs_state = RCU_GP_IDLE, \
 	.gpnum = -300, \
 	.completed = -300, \
@@ -91,6 +84,19 @@ DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
 
 static struct rcu_state *rcu_state;
 
+/* Increase (but not decrease) the CONFIG_RCU_FANOUT_LEAF at boot time. */
+static int rcu_fanout_leaf = CONFIG_RCU_FANOUT_LEAF;
+module_param(rcu_fanout_leaf, int, 0);
+int rcu_num_lvls __read_mostly = RCU_NUM_LVLS;
+static int num_rcu_lvl[] = {  /* Number of rcu_nodes at specified level. */
+	NUM_RCU_LVL_0,
+	NUM_RCU_LVL_1,
+	NUM_RCU_LVL_2,
+	NUM_RCU_LVL_3,
+	NUM_RCU_LVL_4,
+};
+int rcu_num_nodes __read_mostly = NUM_RCU_NODES; /* Total # rcu_nodes in use. */
+
 /*
  * The rcu_scheduler_active variable transitions from zero to one just
  * before the first task is spawned.  So when this variable is zero, RCU
@@ -2574,9 +2580,9 @@ static void __init rcu_init_levelspread(struct rcu_state *rsp)
 {
 	int i;
 
-	for (i = NUM_RCU_LVLS - 1; i > 0; i--)
+	for (i = rcu_num_lvls - 1; i > 0; i--)
 		rsp->levelspread[i] = CONFIG_RCU_FANOUT;
-	rsp->levelspread[0] = CONFIG_RCU_FANOUT_LEAF;
+	rsp->levelspread[0] = rcu_fanout_leaf;
 }
 #else /* #ifdef CONFIG_RCU_FANOUT_EXACT */
 static void __init rcu_init_levelspread(struct rcu_state *rsp)
@@ -2586,7 +2592,7 @@ static void __init rcu_init_levelspread(struct rcu_state *rsp)
 	int i;
 
 	cprv = NR_CPUS;
-	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
+	for (i = rcu_num_lvls - 1; i >= 0; i--) {
 		ccur = rsp->levelcnt[i];
 		rsp->levelspread[i] = (cprv + ccur - 1) / ccur;
 		cprv = ccur;
@@ -2613,13 +2619,15 @@ static void __init rcu_init_one(struct rcu_state *rsp,
 
 	/* Initialize the level-tracking arrays. */
 
-	for (i = 1; i < NUM_RCU_LVLS; i++)
+	for (i = 0; i < rcu_num_lvls; i++)
+		rsp->levelcnt[i] = num_rcu_lvl[i];
+	for (i = 1; i < rcu_num_lvls; i++)
 		rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];
 	rcu_init_levelspread(rsp);
 
 	/* Initialize the elements themselves, starting from the leaves. */
 
-	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
+	for (i = rcu_num_lvls - 1; i >= 0; i--) {
 		cpustride *= rsp->levelspread[i];
 		rnp = rsp->level[i];
 		for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {
@@ -2649,7 +2657,7 @@ static void __init rcu_init_one(struct rcu_state *rsp,
 	}
 
 	rsp->rda = rda;
-	rnp = rsp->level[NUM_RCU_LVLS - 1];
+	rnp = rsp->level[rcu_num_lvls - 1];
 	for_each_possible_cpu(i) {
 		while (i > rnp->grphi)
 			rnp++;
@@ -2658,11 +2666,72 @@ static void __init rcu_init_one(struct rcu_state *rsp,
 	}
 }
 
+/*
+ * Compute the rcu_node tree geometry from kernel parameters.  This cannot
+ * replace the definitions in rcutree.h because those are needed to size
+ * the ->node array in the rcu_state structure.
+ */
+static void __init rcu_init_geometry(void)
+{
+	int i;
+	int j;
+	int n = NR_CPUS;
+	int rcu_capacity[MAX_RCU_LVLS + 1];
+
+	/* If the compile-time values are accurate, just leave. */
+	if (rcu_fanout_leaf == CONFIG_RCU_FANOUT_LEAF)
+		return;
+
+	/*
+	 * Compute number of nodes that can be handled an rcu_node tree
+	 * with the given number of levels.  Setting rcu_capacity[0] makes
+	 * some of the arithmetic easier.
+	 */
+	rcu_capacity[0] = 1;
+	rcu_capacity[1] = rcu_fanout_leaf;
+	for (i = 2; i <= MAX_RCU_LVLS; i++)
+		rcu_capacity[i] = rcu_capacity[i - 1] * CONFIG_RCU_FANOUT;
+
+	/*
+	 * The boot-time rcu_fanout_leaf parameter is only permitted
+	 * to increase the leaf-level fanout, not decrease it.  Of course,
+	 * the leaf-level fanout cannot exceed the number of bits in
+	 * the rcu_node masks.  Finally, the tree must be able to accommodate
+	 * the configured number of CPUs.  Complain and fall back to the
+	 * compile-time values if these limits are exceeded.
+	 */
+	if (rcu_fanout_leaf < CONFIG_RCU_FANOUT_LEAF ||
+	    rcu_fanout_leaf > sizeof(unsigned long) * 8 ||
+	    n > rcu_capacity[MAX_RCU_LVLS]) {
+		WARN_ON(1);
+		return;
+	}
+
+	/* Calculate the number of rcu_nodes at each level of the tree. */
+	for (i = 1; i <= MAX_RCU_LVLS; i++)
+		if (n <= rcu_capacity[i]) {
+			for (j = 0; j <= i; j++)
+				num_rcu_lvl[j] =
+					DIV_ROUND_UP(n, rcu_capacity[i - j]);
+			rcu_num_lvls = i;
+			for (j = i + 1; j <= MAX_RCU_LVLS; j++)
+				num_rcu_lvl[j] = 0;
+			break;
+		}
+
+	/* Calculate the total number of rcu_node structures. */
+	rcu_num_nodes = 0;
+	for (i = 0; i <= MAX_RCU_LVLS; i++)
+		rcu_num_nodes += num_rcu_lvl[i];
+	rcu_num_nodes -= n;
+}
+
 void __init rcu_init(void)
 {
 	int cpu;
 
 	rcu_bootup_announce();
+	rcu_init_geometry();
 	rcu_init_one(&rcu_sched_state, &rcu_sched_data);
 	rcu_init_one(&rcu_bh_state, &rcu_bh_data);
 	__rcu_init_preempt();

kernel/rcutree.h

@@ -42,28 +42,28 @@
 #define RCU_FANOUT_4	      (RCU_FANOUT_3 * CONFIG_RCU_FANOUT)
 
 #if NR_CPUS <= RCU_FANOUT_1
-#  define NUM_RCU_LVLS	      1
+#  define RCU_NUM_LVLS	      1
 #  define NUM_RCU_LVL_0	      1
 #  define NUM_RCU_LVL_1	      (NR_CPUS)
 #  define NUM_RCU_LVL_2	      0
 #  define NUM_RCU_LVL_3	      0
 #  define NUM_RCU_LVL_4	      0
 #elif NR_CPUS <= RCU_FANOUT_2
-#  define NUM_RCU_LVLS	      2
+#  define RCU_NUM_LVLS	      2
 #  define NUM_RCU_LVL_0	      1
 #  define NUM_RCU_LVL_1	      DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
 #  define NUM_RCU_LVL_2	      (NR_CPUS)
 #  define NUM_RCU_LVL_3	      0
 #  define NUM_RCU_LVL_4	      0
 #elif NR_CPUS <= RCU_FANOUT_3
-#  define NUM_RCU_LVLS	      3
+#  define RCU_NUM_LVLS	      3
 #  define NUM_RCU_LVL_0	      1
 #  define NUM_RCU_LVL_1	      DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
 #  define NUM_RCU_LVL_2	      DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
 #  define NUM_RCU_LVL_3	      (NR_CPUS)
 #  define NUM_RCU_LVL_4	      0
 #elif NR_CPUS <= RCU_FANOUT_4
-#  define NUM_RCU_LVLS	      4
+#  define RCU_NUM_LVLS	      4
 #  define NUM_RCU_LVL_0	      1
 #  define NUM_RCU_LVL_1	      DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_3)
 #  define NUM_RCU_LVL_2	      DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
@@ -76,6 +76,9 @@
 #define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3 + NUM_RCU_LVL_4)
 #define NUM_RCU_NODES (RCU_SUM - NR_CPUS)
 
+extern int rcu_num_lvls;
+extern int rcu_num_nodes;
+
 /*
  * Dynticks per-CPU state.
  */
@@ -206,7 +209,7 @@ struct rcu_node {
  */
 #define rcu_for_each_node_breadth_first(rsp, rnp) \
 	for ((rnp) = &(rsp)->node[0]; \
-	     (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++)
+	     (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++)
 
 /*
  * Do a breadth-first scan of the non-leaf rcu_node structures for the
@@ -215,7 +218,7 @@ struct rcu_node {
  */
 #define rcu_for_each_nonleaf_node_breadth_first(rsp, rnp) \
 	for ((rnp) = &(rsp)->node[0]; \
-	     (rnp) < (rsp)->level[NUM_RCU_LVLS - 1]; (rnp)++)
+	     (rnp) < (rsp)->level[rcu_num_lvls - 1]; (rnp)++)
 
 /*
  * Scan the leaves of the rcu_node hierarchy for the specified rcu_state
@@ -224,8 +227,8 @@ struct rcu_node {
  * It is still a leaf node, even if it is also the root node.
  */
 #define rcu_for_each_leaf_node(rsp, rnp) \
-	for ((rnp) = (rsp)->level[NUM_RCU_LVLS - 1]; \
-	     (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++)
+	for ((rnp) = (rsp)->level[rcu_num_lvls - 1]; \
+	     (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++)
 
 /* Index values for nxttail array in struct rcu_data. */
 #define RCU_DONE_TAIL		0	/* Also RCU_WAIT head. */
@@ -357,9 +360,9 @@ do {									\
  */
 struct rcu_state {
 	struct rcu_node node[NUM_RCU_NODES];	/* Hierarchy. */
-	struct rcu_node *level[NUM_RCU_LVLS];	/* Hierarchy levels. */
+	struct rcu_node *level[RCU_NUM_LVLS];	/* Hierarchy levels. */
 	u32 levelcnt[MAX_RCU_LVLS + 1];		/* # nodes in each level. */
-	u8 levelspread[NUM_RCU_LVLS];		/* kids/node in each level. */
+	u8 levelspread[RCU_NUM_LVLS];		/* kids/node in each level. */
 	struct rcu_data __percpu *rda;		/* pointer of percu rcu_data. */
 
 	/* The following fields are guarded by the root rcu_node's lock. */

kernel/rcutree_plugin.h

@@ -70,6 +70,8 @@ static void __init rcu_bootup_announce_oddness(void)
 #if NUM_RCU_LVL_4 != 0
 	printk(KERN_INFO "\tExperimental four-level hierarchy is enabled.\n");
 #endif
+	if (rcu_fanout_leaf != CONFIG_RCU_FANOUT_LEAF)
+		printk(KERN_INFO "\tExperimental boot-time adjustment of leaf fanout to %d.\n", rcu_fanout_leaf);
 }
 
 #ifdef CONFIG_TREE_PREEMPT_RCU

kernel/rcutree_trace.c

@@ -278,7 +278,7 @@ static void print_one_rcu_state(struct seq_file *m, struct rcu_state *rsp)
 		   rsp->n_force_qs, rsp->n_force_qs_ngp,
 		   rsp->n_force_qs - rsp->n_force_qs_ngp,
 		   rsp->n_force_qs_lh, rsp->qlen_lazy, rsp->qlen);
-	for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < NUM_RCU_NODES; rnp++) {
+	for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < rcu_num_nodes; rnp++) {
 		if (rnp->level != level) {
 			seq_puts(m, "\n");
 			level = rnp->level;