PM / Domains: Do not measure start time for "irq safe" devices
The genpd_start_dev() routine used by pm_genpd_runtime_resume()
to put "irq safe" devices into the full power state measures the
time necessary to "start" the device and updates its PM QoS timing
data if necessary. This may lead to a deadlock if the given device
is a clock source and genpd_start_dev() is invoked from within the
clock source's .enable() routine, which will happen if that routine
uses pm_runtime_get_sync(), for example, to ensure that the device
is operational.
For this reason, introduce a special routine analogous to
genpd_start_dev(), called genpd_start_dev_no_timing(), that doesn't
carry out the time measurement, and make pm_genpd_runtime_resume()
use it instead of genpd_start_dev() to power up "irq safe" devices.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
index 5217275..d7e71b5 100644
--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c
@@ -75,6 +75,12 @@
start_latency_ns, "start");
}
+static int genpd_start_dev_no_timing(struct generic_pm_domain *genpd,
+ struct device *dev)
+{
+ return GENPD_DEV_CALLBACK(genpd, int, start, dev);
+}
+
static bool genpd_sd_counter_dec(struct generic_pm_domain *genpd)
{
bool ret = false;
@@ -626,7 +632,7 @@
/* If power.irq_safe, the PM domain is never powered off. */
if (dev->power.irq_safe)
- return genpd_start_dev(genpd, dev);
+ return genpd_start_dev_no_timing(genpd, dev);
mutex_lock(&genpd->lock);
ret = __pm_genpd_poweron(genpd);