Re: [PATCH v2 07/10] power: reset: at91-reset: add reset_controller_dev support

[Philipp Zabel]

On Do, 2022-04-07 at 10:17 +0300, Claudiu Beznea wrote:
> SAMA7G5 reset controller has 5 extra lines that goes to different devices
> (3 lines to USB PHYs, 1 line to DDR controller, 1 line to DDR PHY
> controller). These reset lines could be requested by different controller
> drivers (e.g. USB PHY driver) and these controllers' drivers could
> assert/deassert these lines when necessary. Thus add support for
> reset_controller_dev which brings this functionality.
> 
> Signed-off-by: Claudiu Beznea <claudiu.beznea@xxxxxxxxxxxxx>
> ---
>  drivers/power/reset/at91-reset.c | 107 +++++++++++++++++++++++++++++--
>  1 file changed, 103 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/power/reset/at91-reset.c b/drivers/power/reset/at91-reset.c
> index 1b2aca3f490d..a6f65ac430cd 100644
> --- a/drivers/power/reset/at91-reset.c
> +++ b/drivers/power/reset/at91-reset.c
[...]
> +static int at91_reset_update(struct reset_controller_dev *rcdev,
> +			     unsigned long id, bool assert)
> +{
> +	struct at91_reset *reset = to_at91_reset(rcdev);
> +	u32 val;
> +
> +	spin_lock(&reset->lock);

Use spin_lock_irqsave. We don't know where we are called from and this
isn't a time critical path.

> +	val = readl_relaxed(reset->dev_base);
> +	if (assert)
> +		val |= BIT(id);
> +	else
> +		val &= ~BIT(id);
> +	writel_relaxed(val, reset->dev_base);
> +	spin_unlock(&reset->lock);
> +
> +	return 0;
> +}
> +
> +static int at91_reset_assert(struct reset_controller_dev *rcdev,
> +			     unsigned long id)
> +{
> +	return at91_reset_update(rcdev, id, true);
> +}
> +
> +static int at91_reset_deassert(struct reset_controller_dev *rcdev,
> +			       unsigned long id)
> +{
> +	return at91_reset_update(rcdev, id, false);
> +}
> +
> +static int at91_reset_dev_status(struct reset_controller_dev *rcdev,
> +				 unsigned long id)
> +{
> +	struct at91_reset *reset = to_at91_reset(rcdev);
> +	u32 val;
> +
> +	spin_lock(&reset->lock);

Locking is not necessary here. If the read is racing against an update,
it either returns the register value before or after the update, same
as without a lock.

With those changes,
Reviewed-by: Philipp Zabel <p.zabel@xxxxxxxxxxxxxx>

regards
Philipp