x86, x2apic: Minimize IPI register writes using cluster groups

In the case of x2apic cluster mode we can group IPI register
writes based on the cluster group instead of individual per-cpu
destination messages.

This reduces the apic register writes and reduces the amount of
IPI messages (in the best case we can reduce it by a factor of
16).

With this change, the cost of flush_tlb_others(), with the flush
tlb IPI being sent from a cpu in the socket-1 to all the logical
cpus in socket-2 (on a Westmere-EX system that has 20 logical
cpus in a socket) is 3x times better now (compared to the former
'send one-by-one' algorithm).

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: steiner@sgi.com
Cc: yinghai@kernel.org
Link: http://lkml.kernel.org/r/20110519234637.512271057@sbsiddha-MOBL3.sc.intel.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 44 insertions, 14 deletions

diff --git a/arch/x86/kernel/apic/x2apic_cluster.c b/arch/x86/kernel/apic/x2apic_cluster.c
index 4b2bb1381ffa..4dfe9363ff4e 100644
--- a/arch/x86/kernel/apic/x2apic_cluster.c
+++ b/arch/x86/kernel/apic/x2apic_cluster.c
@@ -5,6 +5,7 @@
 #include <linux/ctype.h>
 #include <linux/init.h>
 #include <linux/dmar.h>
+#include <linux/cpu.h>
 
 #include <asm/smp.h>
 #include <asm/apic.h>
@@ -12,6 +13,7 @@
 
 static DEFINE_PER_CPU(u32, x86_cpu_to_logical_apicid);
 static DEFINE_PER_CPU(cpumask_var_t, cpus_in_cluster);
+static DEFINE_PER_CPU(cpumask_var_t, ipi_mask);
 
 static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 {
@@ -54,30 +56,52 @@ static inline u32 x2apic_cluster(int cpu)
         return per_cpu(x86_cpu_to_logical_apicid, cpu) >> 16;
 }
 
-/*
- * for now, we send the IPI's one by one in the cpumask.
- * TBD: Based on the cpu mask, we can send the IPI's to the cluster group
- * at once. We have 16 cpu's in a cluster. This will minimize IPI register
- * writes.
- */
 static void
 __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 {
-        unsigned long query_cpu;
-        unsigned long this_cpu;
+        struct cpumask *cpus_in_cluster_ptr;
+        struct cpumask *ipi_mask_ptr;
+        unsigned int cpu, this_cpu;
         unsigned long flags;
+        u32 dest;
 
         x2apic_wrmsr_fence();
 
         local_irq_save(flags);
 
         this_cpu = smp_processor_id();
-        for_each_cpu(query_cpu, mask) {
-                if (apic_dest == APIC_DEST_ALLBUT && query_cpu == this_cpu)
+
+        /*
+         * We are to modify mask, so we need an own copy
+         * and be sure it's manipulated with irq off.
+         */
+        ipi_mask_ptr = __raw_get_cpu_var(ipi_mask);
+        cpumask_copy(ipi_mask_ptr, mask);
+
+        /*
+         * The idea is to send one IPI per cluster.
+         */
+        for_each_cpu(cpu, ipi_mask_ptr) {
+                unsigned long i;
+
+                cpus_in_cluster_ptr = per_cpu(cpus_in_cluster, cpu);
+                dest = 0;
+
+                /* Collect cpus in cluster. */
+                for_each_cpu_and(i, ipi_mask_ptr, cpus_in_cluster_ptr) {
+                        if (apic_dest == APIC_DEST_ALLINC || i != this_cpu)
+                                dest |= per_cpu(x86_cpu_to_logical_apicid, i);
+                }
+
+                if (!dest)
                         continue;
-                __x2apic_send_IPI_dest(
-                        per_cpu(x86_cpu_to_logical_apicid, query_cpu),
-                        vector, apic->dest_logical);
+
+                __x2apic_send_IPI_dest(dest, vector, apic->dest_logical);
+                /*
+                 * Cluster sibling cpus should be discared now so
+                 * we would not send IPI them second time.
+                 */
+                cpumask_andnot(ipi_mask_ptr, ipi_mask_ptr, cpus_in_cluster_ptr);
         }
 
         local_irq_restore(flags);
@@ -198,6 +222,10 @@ update_clusterinfo(struct notifier_block *nfb, unsigned long action, void *hcpu)
                 if (!zalloc_cpumask_var(&per_cpu(cpus_in_cluster, this_cpu),
                                         GFP_KERNEL)) {
                         err = -ENOMEM;
+                } else if (!zalloc_cpumask_var(&per_cpu(ipi_mask, this_cpu),
+                                               GFP_KERNEL)) {
+                        free_cpumask_var(per_cpu(cpus_in_cluster, this_cpu));
+                        err = -ENOMEM;
                 }
                 break;
         case CPU_UP_CANCELED:
@@ -210,6 +238,7 @@ update_clusterinfo(struct notifier_block *nfb, unsigned long action, void *hcpu)
                         __cpu_clear(cpu, per_cpu(cpus_in_cluster, this_cpu));
                 }
                 free_cpumask_var(per_cpu(cpus_in_cluster, this_cpu));
+                free_cpumask_var(per_cpu(ipi_mask, this_cpu));
                 break;
         }
 
@@ -225,8 +254,9 @@ static int x2apic_init_cpu_notifier(void)
         int cpu = smp_processor_id();
 
         zalloc_cpumask_var(&per_cpu(cpus_in_cluster, cpu), GFP_KERNEL);
+        zalloc_cpumask_var(&per_cpu(ipi_mask, cpu), GFP_KERNEL);
 
-        BUG_ON(!per_cpu(cpus_in_cluster, cpu));
+        BUG_ON(!per_cpu(cpus_in_cluster, cpu) || !per_cpu(ipi_mask, cpu));
 
         __cpu_set(cpu, per_cpu(cpus_in_cluster, cpu));
         register_hotcpu_notifier(&x2apic_cpu_notifier);