Re: [PATCH 2/7] phy: qcom-qmp-combo: Move phy_mutex out of com_init/exit

[Johan Hovold <johan@kernel.org>]

On Mon, Apr 24, 2023 at 08:40:05PM -0700, Bjorn Andersson wrote:
> With the upcoming introduction of USB Type-C orientation switching the
> region of mutual exclusion needs to be extended to cover both the common
> init/exit as well as the individual functions.
> 
> So move the phy_mutex one step up the stack.
> 
> Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
> ---
>  drivers/phy/qualcomm/phy-qcom-qmp-combo.c | 51 +++++++++++++----------
>  1 file changed, 30 insertions(+), 21 deletions(-)
> 
> diff --git a/drivers/phy/qualcomm/phy-qcom-qmp-combo.c b/drivers/phy/qualcomm/phy-qcom-qmp-combo.c
> index 6850e04c329b..7280f7141961 100644
> --- a/drivers/phy/qualcomm/phy-qcom-qmp-combo.c
> +++ b/drivers/phy/qualcomm/phy-qcom-qmp-combo.c
> @@ -2463,16 +2463,13 @@ static int qmp_combo_com_init(struct qmp_combo *qmp)
>  	void __iomem *com = qmp->com;
>  	int ret;
>  
> -	mutex_lock(&qmp->phy_mutex);
> -	if (qmp->init_count++) {
> -		mutex_unlock(&qmp->phy_mutex);
> +	if (qmp->init_count++)
>  		return 0;
> -	}
>  
>  	ret = regulator_bulk_enable(cfg->num_vregs, qmp->vregs);
>  	if (ret) {
>  		dev_err(qmp->dev, "failed to enable regulators, err=%d\n", ret);
> -		goto err_unlock;
> +		goto err;

I was going to say that you can just return ret directly but then
realised we have a counter imbalance here that should be fixed.

I've just sent a couple of fixes which you could rebase on:

	https://lore.kernel.org/r/20230502103810.12061-1-johan+linaro@kernel.org

>  	}
>  
>  	ret = reset_control_bulk_assert(cfg->num_resets, qmp->resets);
> @@ -2514,16 +2511,13 @@ static int qmp_combo_com_init(struct qmp_combo *qmp)
>  	qphy_setbits(qmp->pcs, cfg->regs[QPHY_PCS_POWER_DOWN_CONTROL],
>  			SW_PWRDN);
>  
> -	mutex_unlock(&qmp->phy_mutex);
> -
>  	return 0;
>  
>  err_assert_reset:
>  	reset_control_bulk_assert(cfg->num_resets, qmp->resets);
>  err_disable_regulators:
>  	regulator_bulk_disable(cfg->num_vregs, qmp->vregs);
> -err_unlock:
> -	mutex_unlock(&qmp->phy_mutex);
> +err:
>  
>  	return ret;
>  }
 
> @@ -2686,14 +2683,19 @@ static int qmp_combo_usb_init(struct phy *phy)
>  	struct qmp_combo *qmp = phy_get_drvdata(phy);
>  	int ret;
>  
> +	mutex_lock(&qmp->phy_mutex);

Nit: I think adding a newline here would improve readability.

>  	ret = qmp_combo_com_init(qmp);
>  	if (ret)
> -		return ret;
> +		goto out_unlock;
>  
>  	ret = qmp_combo_usb_power_on(phy);
> -	if (ret)
> +	if (ret) {
>  		qmp_combo_com_exit(qmp);
> +		goto out_unlock;
> +	}
>  
> +out_unlock:
> +	mutex_unlock(&qmp->phy_mutex);

Same here.

>  	return ret;
>  }
>  
> @@ -2702,11 +2704,18 @@ static int qmp_combo_usb_exit(struct phy *phy)
>  	struct qmp_combo *qmp = phy_get_drvdata(phy);
>  	int ret;
>  
> +	mutex_lock(&qmp->phy_mutex);

And here.

>  	ret = qmp_combo_usb_power_off(phy);
>  	if (ret)
> -		return ret;
> +		goto out_unlock;
>  
> -	return qmp_combo_com_exit(qmp);
> +	ret = qmp_combo_com_exit(qmp);
> +	if (ret)
> +		goto out_unlock;
> +
> +out_unlock:
> +	mutex_unlock(&qmp->phy_mutex);

And here.

> +	return ret;
>  }
>  
>  static int qmp_combo_usb_set_mode(struct phy *phy, enum phy_mode mode, int submode)

Looks good otherwise: 

Reviewed-by: Johan Hovold <johan+linaro@kernel.org>

Johan