sched/uclamp: Set max_spare_cap_cpu even if max_spare_cap is 0
authorQais Yousef <qyousef@layalina.io>
Sat, 16 Sep 2023 23:29:53 +0000 (00:29 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 20 Nov 2023 10:51:49 +0000 (11:51 +0100)
commitd77530a1d47fee1cb0c4142a79a26705ce66de01
tree78b3380a826897dad3ec2c1bccca8cf661c445fc
parentb7839197719f93b3f0454e3cceadce419e241a26
sched/uclamp: Set max_spare_cap_cpu even if max_spare_cap is 0

[ Upstream commit 6b00a40147653c8ea748e8f4396510f252763364 ]

When uclamp_max is being used, the util of the task could be higher than
the spare capacity of the CPU, but due to uclamp_max value we force-fit
it there.

The way the condition for checking for max_spare_cap in
find_energy_efficient_cpu() was constructed; it ignored any CPU that has
its spare_cap less than or _equal_ to max_spare_cap. Since we initialize
max_spare_cap to 0; this lead to never setting max_spare_cap_cpu and
hence ending up never performing compute_energy() for this cluster and
missing an opportunity for a better energy efficient placement to honour
uclamp_max setting.

max_spare_cap = 0;
cpu_cap = capacity_of(cpu) - cpu_util(p);  // 0 if cpu_util(p) is high

...

util_fits_cpu(...); // will return true if uclamp_max forces it to fit

...

// this logic will fail to update max_spare_cap_cpu if cpu_cap is 0
if (cpu_cap > max_spare_cap) {
max_spare_cap = cpu_cap;
max_spare_cap_cpu = cpu;
}

prev_spare_cap suffers from a similar problem.

Fix the logic by converting the variables into long and treating -1
value as 'not populated' instead of 0 which is a viable and correct
spare capacity value. We need to be careful signed comparison is used
when comparing with cpu_cap in one of the conditions.

Fixes: 1d42509e475c ("sched/fair: Make EAS wakeup placement consider uclamp restrictions")
Signed-off-by: Qais Yousef (Google) <qyousef@layalina.io>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230916232955.2099394-2-qyousef@layalina.io
Signed-off-by: Sasha Levin <sashal@kernel.org>
kernel/sched/fair.c