Re: [PATCH v5 1/1] input: add driver for Bosch Sensortec's BMA150 accelerometer

From: Dmitry Torokhov <dmitry.torokhov@gmail.com>
To: Eric Andersson <eric.andersson@unixphere.com>
Cc: linux-input@vger.kernel.org, linux-kernel@vger.kernel.org,
	zhengguang.guo@bosch-sensortec.com, stefan.nilsson@unixphere.com,
	alan@lxorguk.ukuu.org.uk,
	Albert Zhang <xu.zhang@bosch-sensortec.com>
Subject: Re: [PATCH v5 1/1] input: add driver for Bosch Sensortec's BMA150 accelerometer
Date: Mon, 25 Jul 2011 18:51:21 -0700	[thread overview]
Message-ID: <20110726015121.GA2076@core.coreip.homeip.net> (raw)
In-Reply-To: <20110722094521.GB710@scully.xfiles.lan>

On Fri, Jul 22, 2011 at 11:45:21AM +0200, Eric Andersson wrote:
> Hi Dmitry,
> 
> Thanks for reviewing!
> 
> > > +static int bma150_open(struct bma150_data *bma150)
> > > +{
> > > +#ifdef CONFIG_PM_RUNTIME
> > > +	return pm_runtime_get_sync(&bma150->client->dev);
> > > +#else
> > > +	return bma150_set_mode(bma150->client, BMA150_MODE_NORMAL);
> > 
> > Hmm, this is kind of weird. I'd expect you want to try waking up the
> > parent in both cases (if no runtime pm then call to pm_runtime_get_sync
> > is basically a noop) and then wake up your device.
> 
> I am not sure what you mean? Are you suggesting something like:
> 	bma150_set_mode(bma150->client, BMA150_MODE_NORMAL);
> 	return pm_runtime_get_sync(&bma150->client->dev);
> 
> That would cause multiple bma150_set_mode() calls since
> pm_runtime_get_sync() will trigger one if needed.
> 

Right... It looks like instead of calling pm_runtime_get_sync() we need
to do something like this:

static int bma150_open(struct bma150_data *bma150)
{
	int error;

	error = pm_runtime_set_active(&bma150->client->dev);
	if (error)
		return error;

	return bma150_set_mode(bma150->client, BMA150_MODE_NORMAL);
}

and opposite in bma150_close().

Thanks.

-- 
Dmitry