[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170328115306.GR6986@localhost.localdomain>
Date: Tue, 28 Mar 2017 12:53:06 +0100
From: Charles Keepax <ckeepax@...nsource.wolfsonmicro.com>
To: Daniel Baluta <daniel.baluta@...il.com>
CC: Daniel Baluta <daniel.baluta@....com>,
Nicolin Chen <nicoleotsuka@...il.com>,
Mark Brown <broonie@...nel.org>, <alsa-devel@...a-project.org>,
Timur Tabi <timur@...i.org>, <Xiubo.Lee@...il.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Liam Girdwood <lgirdwood@...il.com>,
Takashi Iwai <tiwai@...e.com>, <fabio.estevam@....com>,
<linuxppc-dev@...ts.ozlabs.org>
Subject: Re: [alsa-devel] [PATCH v3 2/2] ASoC: imx-wm8962: Fix codec_clk
cleanup
On Tue, Mar 28, 2017 at 01:47:04PM +0300, Daniel Baluta wrote:
> On Tue, Mar 28, 2017 at 12:21 PM, Charles Keepax
> <ckeepax@...nsource.wolfsonmicro.com> wrote:
> > On Tue, Mar 28, 2017 at 11:58:52AM +0300, Daniel Baluta wrote:
> >> Resource managed devm_clk_get only works with platform's device dev.
> >>
> >
> > I feel like this could use an explaination of why not using devm
> > is the correct fix, rather than just using the platform device
> > for the call. Its not obvious to me, that using the platform
> > device would be an issue.
>
> Hi Charles,
>
> I see where the confusion comes from :) and I thought the explanation
> is obvious from the code, see inline comments.
>
> Would an explanation like the one below, work better?
>
> " We cannot use devm_clk_get with &codec_dev->dev device because
> the kernel uses pdev->dev to free the managed resources, so we will end
> up with a leaking codec_clk reference"
>
> >> @@ -231,7 +231,7 @@ static int imx_wm8962_probe(struct platform_device *pdev)
> >> goto fail;
> >> }
>
> pdev->dev from here:
>
> >>
> >> - codec_clk = devm_clk_get(&codec_dev->dev, NULL);
> >> + codec_clk = clk_get(&codec_dev->dev, NULL);
>
> is different from &codec_dev->dev.
>
I get that they are different, I just don't get why changing
from a devm_clk_get to a clk_get is a better fix than changing
&codec->dev to &pdev->dev.
Thanks,
Charles
Powered by blists - more mailing lists