Hi Qiushi, On 17. 06. 20 16:30, David Laight wrote: > From: Wolfram Sang >> Sent: 14 June 2020 10:10 >> >> On Sat, Jun 13, 2020 at 04:59:23PM -0500, wu000273@xxxxxxx wrote: >>> From: Qiushi Wu <wu000273@xxxxxxx> >>> >>> pm_runtime_get_sync() increments the runtime PM usage counter even >>> when it returns an error code. Thus call pm_runtime_put_noidle() >>> if pm_runtime_get_sync() fails. >> >> Can you point me to a discussion where it was decided that this is a >> proper fix? I'd think we rather should fix pm_runtime_get_sync() but >> maybe there are technical reasons against it. > > Or, if there is one place that actually needs the reference split the > code so that unusual case keeps the reference. > > In one of the patches I also spotted: > ret = pm_runtime_get_sync(); > if (ret < 0 && ret != _EAGAIN) > ... > > (I think it was EAGAIN.) > I can't help feeling that is just wrong somewhere. Qiushi: Any update on this one? Thanks, Michal
