[PATCH] sched: Resolve sd_idle and first_idle_cpu Catch-22

[Venkatesh Pallipadi]

Consider a system with { [ (A B) (C D) ] [ (E F) (G H) ] },
() denoting SMT siblings, [] cores on same socket and {} system wide
Further, A, C and D are idle, B is busy and one of EFGH has excess load.

With sd_idle logic, a check in rebalance_domains() converts tick
based load balance requests from CPU A to busy load balance for core
and above domains (lower rate of balance and higher load_idx).

With first_idle_cpu logic, when CPU C or D tries to balance across domains
the logic finds CPU A as first idle CPU in the group and nominates CPU A to
idle balance across sockets.

But, sd_idle above would not allow CPU A to do cross socket idle balance
as CPU A switches its higher level balancing to busy balance.

So, this can result is no cross socket balancing for extended periods.

The fix here adds additional check to detect sd_idle logic in
first_idle_cpu code path. We will now nominate (in order or preference):
* First fully idle CPU
* First semi-idle CPU
* First CPU

Note that this solution works fine for 2 SMT siblings case and won't be
perfect in picking proper semi-idle in case of more than 2 SMT threads.

The problem was found by looking at the code and schedstat output. I don't
yet have any data to show impact of this on any workload.

Signed-off-by: Venkatesh Pallipadi <venki@xxxxxxxxxx>
---
 kernel/sched_fair.c |   41 +++++++++++++++++++++++++++++++++++++++--
 1 files changed, 39 insertions(+), 2 deletions(-)

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 62723a4..1790cc2 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -2603,6 +2603,37 @@ fix_small_capacity(struct sched_domain *sd, struct sched_group *group)
 	return 0;
 }
 
+/*
+ * Find if there is any busy CPUs in SD_SHARE_CPUPOWER domain of
+ * requested CPU.
+ * Bypass the check in case of SD_POWERSAVINGS_BALANCE on
+ * parent domain. In that case requested CPU can still be nominated as
+ * balancer for higher domains.
+ */
+static int is_cpupower_sharing_domain_idle(int cpu)
+{
+	struct sched_domain *sd;
+	int i;
+
+	if (!(sysctl_sched_compat_yield & 0x4))
+		return 1;
+
+	for_each_domain(cpu, sd) {
+		if (!(sd->flags & SD_SHARE_CPUPOWER))
+			break;
+
+		if (test_sd_parent(sd, SD_POWERSAVINGS_BALANCE))
+			return 1;
+
+		for_each_cpu(i, sched_domain_span(sd)) {
+			if (!idle_cpu(i))
+				return 0;
+		}
+	}
+
+	return 1;
+}
+
 /**
  * update_sg_lb_stats - Update sched_group's statistics for load balancing.
  * @sd: The sched_domain whose statistics are to be updated.
@@ -2625,6 +2656,7 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
 	unsigned long load, max_cpu_load, min_cpu_load, max_nr_running;
 	int i;
 	unsigned int balance_cpu = -1, first_idle_cpu = 0;
+	unsigned int first_semiidle_cpu = 0;
 	unsigned long avg_load_per_task = 0;
 
 	if (local_group)
@@ -2644,8 +2676,13 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
 		/* Bias balancing toward cpus of our domain */
 		if (local_group) {
 			if (idle_cpu(i) && !first_idle_cpu) {
-				first_idle_cpu = 1;
-				balance_cpu = i;
+				if (is_cpupower_sharing_domain_idle(i)) {
+					first_idle_cpu = 1;
+					balance_cpu = i;
+				} else if (!first_semiidle_cpu) {
+					first_semiidle_cpu = 1;
+					balance_cpu = i;
+				}
 			}
 
 			load = target_load(i, load_idx);
-- 
1.7.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/