Re: [PATCH] ASoC: cs35l56: Fix pm_runtime imbalance if suspending before first attach

[Mark Brown]

On Mon, Aug 31, 2026 at 10:45:06AM +0100, Richard Fitzgerald wrote:
> Remove the check for init_done in cs35l56_sdw_system_suspend(). Instead,
> protect the call to cs35l56_mask_soundwire_interrupts() to only be done
> if the amp is currently enumerated.

> @@ -386,11 +386,8 @@ static int __maybe_unused cs35l56_sdw_system_suspend(struct device *dev)
>  {
>  	struct cs35l56_private *cs35l56 = dev_get_drvdata(dev);
>  
> -	if (!cs35l56->base.init_done)
> -		return 0;
> -
> -	/* runtime_resume unmasks the interrupt */
> -	cs35l56_mask_soundwire_interrupts(cs35l56);
> +	if (cs35l56->sdw_attached)
> +		cs35l56_mask_soundwire_interrupts(cs35l56);

There's a thin race here on init - we call cs35l56_sdw_init() and unmask
the interrupts prior to setting sdw_attached.

Attachment: signature.asc
Description: PGP signature