[PATCH 09/13] syscon: Probe device first in syscon_get_regmap

[Jonas Karlman]

Hi Jiaxun,

On 2024-05-13 20:13, Jiaxun Yang wrote:
> When bootph-all is enabled for a syscon driver, the device
> may leave unprobed when syscon_get_regmap is called by another
> driver.
> 
> Perform device_probe in syscon_get_regmap, there is no side
> affect as device_probe will return 0 quickly for an activated
> device.
> 
> Signed-off-by: Jiaxun Yang <jiaxun.yang at flygoat.com>
> ---
>  drivers/core/syscon-uclass.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/core/syscon-uclass.c b/drivers/core/syscon-uclass.c
> index f0e69d7216b3..b09f7013194d 100644
> --- a/drivers/core/syscon-uclass.c
> +++ b/drivers/core/syscon-uclass.c
> @@ -32,10 +32,14 @@
>   */
>  struct regmap *syscon_get_regmap(struct udevice *dev)
>  {
> +	int ret;
>  	struct syscon_uc_info *priv;
>  
>  	if (device_get_uclass_id(dev) != UCLASS_SYSCON)
>  		return ERR_PTR(-ENOEXEC);
> +	ret = device_probe(dev);
> +	if (ret)
> +		return ERR_PTR(ret);

Please explain in more details what the issue this is trying to solve.

Typically syscon_get_regmap() is called on a udevice returned from a
uclass_get_device call, and that should trigger a probe for the device
and its parents.

Adding device_probe() here possible just hides an issue that exists
somewhere else. In what instance are you ending up with a call to
this function with a udevice that has not been probed?

Also, please add a new test to test/dm/regmap.c if this solves a real
issue.

Regards,
Jonas

>  	priv = dev_get_uclass_priv(dev);
>  	return priv->regmap;
>  }
>