sched_credit: Raise bar for inter-socket migrations on mostly-idle systems

The credit scheduler ties to keep work balanced, even on a mostly idle
system.  Unfortunately, if you have one VM burning cpu and another VM
idle, the effect is that the busy VM will flip back and forth between
sockets.

This patch addresses this, by only migrating to a different socket if
the number of idle processors is twice that of the socket the vcpu is
currently on.

This will only affect mostly-idle systems; as the system becomes more
busy, other load-balancing code will come into effect.

Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>

diff --git a/xen/common/sched_credit.c b/xen/common/sched_credit.c
index 2a7e299f11f69041198001c78c7386df310387d7..64369c3462ba4675d49070d7f34aaf4960a9c6d9 100644 (file)
--- a/xen/common/sched_credit.c
+++ b/xen/common/sched_credit.c
@@ -499,26 +499,36 @@ _csched_cpu_pick(const struct scheduler *ops, struct vcpu *vc, bool_t commit)
         cpumask_t cpu_idlers;
         cpumask_t nxt_idlers;
         int nxt, weight_cpu, weight_nxt;
+        int migrate_factor;
 
         nxt = cycle_cpu(cpu, cpus);
 
         if ( cpu_isset(cpu, per_cpu(cpu_core_map, nxt)) )
         {
+            /* We're on the same socket, so check the busy-ness of threads.
+             * Migrate if # of idlers is less at all */
             ASSERT( cpu_isset(nxt, per_cpu(cpu_core_map, cpu)) );
+            migrate_factor = 1;
             cpus_and(cpu_idlers, idlers, per_cpu(cpu_sibling_map, cpu));
             cpus_and(nxt_idlers, idlers, per_cpu(cpu_sibling_map, nxt));
         }
         else
         {
+            /* We're on different sockets, so check the busy-ness of cores.
+             * Migrate only if the other core is twice as idle */
             ASSERT( !cpu_isset(nxt, per_cpu(cpu_core_map, cpu)) );
+            migrate_factor = 2;
             cpus_and(cpu_idlers, idlers, per_cpu(cpu_core_map, cpu));
             cpus_and(nxt_idlers, idlers, per_cpu(cpu_core_map, nxt));
         }
 
         weight_cpu = cpus_weight(cpu_idlers);
         weight_nxt = cpus_weight(nxt_idlers);
-        if ( ( (weight_cpu < weight_nxt) ^ sched_smt_power_savings )
-                && (weight_cpu != weight_nxt) )
+        /* smt_power_savings: consolidate work rather than spreading it */
+        if ( ( sched_smt_power_savings
+               && (weight_cpu > weight_nxt) )
+             || ( !sched_smt_power_savings
+                  && (weight_cpu * migrate_factor < weight_nxt) ) )
         {
             cpu = cycle_cpu(CSCHED_PCPU(nxt)->idle_bias, nxt_idlers);
             if ( commit )