On 12 February 2018 at 11:38, Lucas Stach [off-list ref] wrote:

Am Freitag, den 09.02.2018, 14:58 +0100 schrieb Ulf Hansson:

quoted

On 26 January 2018 at 09:38, Jia-Ju Bai [off-list ref]
wrote:

quoted

After checking all possible call chains to genpd_dev_pm_detach()
and
genpd_dev_pm_attach() here,
my tool finds that these functions are never called in atomic
context,
namely never in an interrupt handler or holding a spinlock.
Thus mdelay can be replaced with msleep to avoid busy wait.

This is found by a static analysis tool named DCNS written by
myself.

Signed-off-by: Jia-Ju Bai <redacted>
---
 drivers/base/power/domain.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/base/power/domain.c

b/drivers/base/power/domain.c
index 0c80bea..f84ac72 100644

--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c

@@ -2144,7 +2144,7 @@ static void genpd_dev_pm_detach(struct device

*dev, bool power_off)
                if (ret != -EAGAIN)
                        break;

-               mdelay(i);
+               msleep(i);

This looks like a nice improvement, however moving to msleep() makes
the call to cond_resched() below a bit superfluous. Perhaps remove
that as well.

At least for small values of i, msleep also has a high chance to
overshoot the desired sleep by a lot. It would be better to convert
them to usleep_range with an acceptable slack.

Ack!

[...]

Kind regards
Uffe