[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20190227101544.581505206@infradead.org>
Date: Wed, 27 Feb 2019 11:12:57 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: torvalds@...ux-foundation.org, mingo@...nel.org, bp@...en8.de,
tglx@...utronix.de, luto@...nel.org, namit@...are.com,
peterz@...radead.org
Cc: linux-kernel@...r.kernel.org, Nadav Amit <nadav.amit@...il.com>
Subject: [PATCH 5/5] x86/percpu, sched/fair: Avoid local_clock()
Nadav reported that code-gen changed because of the this_cpu_*()
constraints, avoid this for select_idle_cpu() because that runs with
preemption (and IRQs) disabled anyway.
Reported-by: Nadav Amit <nadav.amit@...il.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
---
kernel/sched/fair.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6160,6 +6160,7 @@ static int select_idle_cpu(struct task_s
u64 time, cost;
s64 delta;
int cpu, nr = INT_MAX;
+ int this = smp_processor_id();
this_sd = rcu_dereference(*this_cpu_ptr(&sd_llc));
if (!this_sd)
@@ -6183,7 +6184,7 @@ static int select_idle_cpu(struct task_s
nr = 4;
}
- time = local_clock();
+ time = cpu_clock(this);
for_each_cpu_wrap(cpu, sched_domain_span(sd), target) {
if (!--nr)
@@ -6194,7 +6195,7 @@ static int select_idle_cpu(struct task_s
break;
}
- time = local_clock() - time;
+ time = cpu_clock(this) - time;
cost = this_sd->avg_scan_cost;
delta = (s64)(time - cost) / 8;
this_sd->avg_scan_cost += delta;
Powered by blists - more mailing lists