[PATCH] sched,nohz: fallback to housekeeping cpus before fallback to possible cpus

[PATCH] sched,nohz: fallback to housekeeping cpus before fallback to possible cpus

[Lai Jiangshan]

Let the processes which lose they original cpus run on housekeeping cpus
if they can to reduce disturbances on nohz_full cpus.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Christoph Lameter <cl@linux.com>
Cc: Kevin Hilman <khilman@linaro.org>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
---
 kernel/sched/core.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index f0f831e..f9ef41a 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1286,7 +1286,7 @@ static int select_fallback_rq(int cpu, struct task_struct *p)
 {
 	int nid = cpu_to_node(cpu);
 	const struct cpumask *nodemask = NULL;
-	enum { cpuset, possible, fail } state = cpuset;
+	enum { cpuset, housekeeping, possible, fail } state = cpuset;
 	int dest_cpu;
 
 	/*
@@ -1322,8 +1322,15 @@ static int select_fallback_rq(int cpu, struct task_struct *p)
 		case cpuset:
 			/* No more Mr. Nice Guy. */
 			cpuset_cpus_allowed_fallback(p);
+			state = housekeeping;
+			break;
+
+		case housekeeping:
+#ifdef CONFIG_NO_HZ_FULL
+			do_set_cpus_allowed(p, housekeeping_mask);
 			state = possible;
 			break;
+# endif /* CONFIG_NO_HZ_FULL. Fallthrough when !CONFIG_NO_HZ_FULL */
 
 		case possible:
 			do_set_cpus_allowed(p, cpu_possible_mask);
-- 
2.1.0

Re: [PATCH] sched,nohz: fallback to housekeeping cpus before fallback to possible cpus

[Peter Zijlstra]

On Wed, Apr 01, 2015 at 05:06:12PM +0800, Lai Jiangshan wrote:
> Let the processes which lose they original cpus run on housekeeping cpus
> if they can to reduce disturbances on nohz_full cpus.

Its a fail if you get here in the first place, why should we fix that?