| Message ID | 20241215172133.178460-2-hdegoede@redhat.com (mailing list archive) |
|---|---|
| State | Handled Elsewhere, archived |
| Headers | show |
| Series | power: supply: Add adc-battery-helper lib and Intel Dollar Cove TI CC battery driver | expand |
diff --git a/drivers/power/supply/power_supply_core.c b/drivers/power/supply/power_supply_core.c index 76ebaef403ed..54ee4c83b32f 100644 --- a/drivers/power/supply/power_supply_core.c +++ b/drivers/power/supply/power_supply_core.c @@ -70,10 +70,8 @@ static int __power_supply_changed_work(struct power_supply *pst, void *data) { struct power_supply *psy = data; - if (__power_supply_is_supplied_by(psy, pst)) { - if (pst->desc->external_power_changed) - pst->desc->external_power_changed(pst); - } + if (__power_supply_is_supplied_by(psy, pst)) + power_supply_external_power_changed(pst); return 0; }
The power-supply core is designed so that power-supply driver callbacks such as get_property() and external_power_changed() will not be called until the power-supply's parent driver's probe() function has completed. There is a race where power_supply_changed() can be called for a supplier of a power-supply which is being probed after the device_add() in __power_supply_register() but before the parent driver's probe() function has completed. Hitting this race breaks the power-supply core's design to not call power-supply driver callbacks before probe() completion. This problem is caused by __power_supply_changed_work() calling the external_power_changed() directly rather then going through the power_supply_external_power_changed() helper which correcly checks psy->use_cnt . Switch to using power_supply_external_power_changed() to fix this race. Signed-off-by: Hans de Goede <hdegoede@redhat.com> --- drivers/power/supply/power_supply_core.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)