Re: [PATCH 5/5] hwspinlock/core: call pm_runtime_put in pm_runtime_get_sync failed case

From: "Rafael J. Wysocki" <rjw@sisk.pl>
To: Ohad Ben-Cohen <ohad@wizery.com>,
	Chuansheng Liu <chuansheng.liu@intel.com>
Cc: Li Fei <fei.li@intel.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 5/5] hwspinlock/core: call pm_runtime_put in pm_runtime_get_sync failed case
Date: Fri, 05 Apr 2013 13:39:58 +0200	[thread overview]
Message-ID: <1988602.WZduyKA6Mt@vostro.rjw.lan> (raw)
In-Reply-To: <CAK=WgbbDhfLa1uQn3v0Lu5iOLL5QgGxKuOVrq9AKD9fLtb42FQ@mail.gmail.com>

On Friday, April 05, 2013 09:27:40 AM Ohad Ben-Cohen wrote:
> Hi Li,
> 
> On Thu, Feb 28, 2013 at 10:02 AM, Li Fei <fei.li@intel.com> wrote:
> >
> > Even in failed case of pm_runtime_get_sync, the usage_count
> > is incremented. In order to keep the usage_count with correct
> > value and runtime power management to behave correctly, call
> > pm_runtime_put(_sync) in such case.
> 
> Is it better then to call pm_runtime_put_noidle instead? This way
> we're sure to only take care of usage_count without ever calling any
> underlying pm handler.

Both would break code that does

 pm_runtime_get_sync(dev);

 <device access>

 pm_runtime_put(dev);

without checking the result of pm_runtime_get_sync() - which BTW is completely
unnecessary in the majority of cases.

So no, it's not a good idea at all.

Thanks,
Rafael

-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.