On Thu, May 21, 2020 at 07:42:56PM +0800, dinghao.liu@xxxxxxxxxx wrote: > We need to make sure if pm_runtime_get_sync() is designed with > such behavior before modifying it. > > I received a response from Rafael when I commited a similar patch: > https://lkml.org/lkml/2020/5/20/1100 > It seems that this behavior is intentional and needs to be kept. Yes. This is why I have said twice or three times to not change pm_runtime_get_sync() but instead to write a replacement. A large percent of the callers are buggy. The pm_runtime_get_sync() is a -4 on Rusty's API scale. http://sweng.the-davies.net/Home/rustys-api-design-manifesto One could argue that anything above a -4 is really a 2 if you read the implementation thouroughly enough... regards, dan carpenter
