Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <7c686fecb11b4ec1f55cd7075dc7cfcdd9b445ba.1670697358.git.christophe.jaillet@wanadoo.fr>
In-Reply-To: <7c686fecb11b4ec1f55cd7075dc7cfcdd9b445ba.1670697358.git.christophe.jaillet@wanadoo.fr>
From:   Ulf Hansson <ulf.hansson@linaro.org>
Date:   Mon, 2 Jan 2023 13:49:08 +0100
Message-ID: <CAPDyKFo7W3EVCFjACB+zaMi8O-wTmCX2winydAB_XtxfnA0Qiw@mail.gmail.com>
Subject: Re: [PATCH 1/3] mmc: sunlpus: Fix an error handling path in spmmc_drv_probe()
To:     Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc:     Tony Huang <tonyhuang.sunplus@gmail.com>,
        Li-hao Kuo <lhjeff911@gmail.com>,
        Arnd Bergmann <arnd@arndb.de>, linux-kernel@vger.kernel.org,
        kernel-janitors@vger.kernel.org, linux-mmc@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sat, 10 Dec 2022 at 19:36, Christophe JAILLET
<christophe.jaillet@wanadoo.fr> wrote:
>
> If an error occurs after successful clk_prepare_enable() call in the probe,
> the clk is not clk_disable_unprepare()'ed.
>
> Use devm_clk_get_enabled() instead of devm_clk_get() to fix, and simplify
> the probe and the remove function accordingly.
>
> Fixes: 4e268fed8b18 ("mmc: Add mmc driver for Sunplus SP7021")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> This changes the order of resource releasing when the driver is removed,
> but it looks ok to me.
> ---
>  drivers/mmc/host/sunplus-mmc.c | 7 +------
>  1 file changed, 1 insertion(+), 6 deletions(-)
>
> diff --git a/drivers/mmc/host/sunplus-mmc.c b/drivers/mmc/host/sunplus-mmc.c
> index db5e0dcdfa7f..3e8856a82188 100644
> --- a/drivers/mmc/host/sunplus-mmc.c
> +++ b/drivers/mmc/host/sunplus-mmc.c
> @@ -878,7 +878,7 @@ static int spmmc_drv_probe(struct platform_device *pdev)
>         if (IS_ERR(host->base))
>                 return PTR_ERR(host->base);
>
> -       host->clk = devm_clk_get(&pdev->dev, NULL);
> +       host->clk = devm_clk_get_enabled(&pdev->dev, NULL);
>         if (IS_ERR(host->clk))
>                 return dev_err_probe(&pdev->dev, PTR_ERR(host->clk), "clk get fail\n");
>
> @@ -896,10 +896,6 @@ static int spmmc_drv_probe(struct platform_device *pdev)
>         if (ret)
>                 return ret;
>
> -       ret = clk_prepare_enable(host->clk);
> -       if (ret)
> -               return dev_err_probe(&pdev->dev, ret, "failed to enable clk\n");
> -
>         ret = mmc_of_parse(mmc);
>         if (ret)
>                 goto probe_free_host;
> @@ -944,7 +940,6 @@ static int spmmc_drv_remove(struct platform_device *dev)
>
>         mmc_remove_host(host->mmc);
>         pm_runtime_get_sync(&dev->dev);
> -       clk_disable_unprepare(host->clk);

I don't think this improves code quality, but rather the opposite.

Removing the call to clk_disable_unprepare() here, makes the untrained
eye, to believe that the calls to pm_runtime_get_sync() and
pm_runtime_put_noidle() are redundant, while they aren't.

Instead, I suggest we add a call to clk_disable_unprepare() in the
error path in ->probe(), rather than using devm_clk_get_enabled().

>         pm_runtime_put_noidle(&dev->dev);
>         pm_runtime_disable(&dev->dev);
>         platform_set_drvdata(dev, NULL);

Kind regards
Uffe