Re: [PATCH] gpio: tegra186: Support multi-socket devices

[Jon Hunter]

On 16/02/2026 11:29, Bartosz Golaszewski wrote:
> On Mon, 16 Feb 2026 05:58:42 +0100, Prathamesh Shete <pshete@xxxxxxxxxx> said:
> > On Tegra platforms, multiple SoC instances may be present with each
> > defining the same GPIO name. For such devices, this results in
> > duplicate GPIO names.
> >
> > When the device has a valid NUMA node, prepend the NUMA node ID
> > to the GPIO name prefix. The node ID identifies each socket,
> > ensuring GPIO line names remain distinct across multiple sockets.
> >
> > Signed-off-by: Prathamesh Shete <pshete@xxxxxxxxxx>
> > ---
> >   drivers/gpio/gpio-tegra186.c | 26 +++++++++++++++++---------
> >   1 file changed, 17 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/gpio/gpio-tegra186.c b/drivers/gpio/gpio-tegra186.c
> > index 9c874f07be75..daf5aaffa28a 100644
> > --- a/drivers/gpio/gpio-tegra186.c
> > +++ b/drivers/gpio/gpio-tegra186.c
> > @@ -857,7 +857,8 @@ static int tegra186_gpio_probe(struct platform_device *pdev)
> >   	struct device_node *np;
> >   	struct resource *res;
> >   	char **names;
> > -	int err;
> > +	char *instance = "";
> > +	int node, err;
> >
> >   	gpio = devm_kzalloc(&pdev->dev, sizeof(*gpio), GFP_KERNEL);
> >   	if (!gpio)
> > @@ -937,17 +938,21 @@ static int tegra186_gpio_probe(struct platform_device *pdev)
> >   	if (!names)
> >   		return -ENOMEM;
> >
> > +	node = dev_to_node(&pdev->dev);
> > +	if (node >= 0) {
> > +		instance = devm_kasprintf(gpio->gpio.parent, GFP_KERNEL, "%d-", node);
>
> I've never noticed it before (and it's not introduced by this patch) but I
> really dislike the child device registering devres nodes with its parent...

While it looks a bit odd, the parent is initialised as ...

 gpio->gpio.parent = &pdev->dev

> > +		if (!instance)
> > +			return -ENOMEM;
> > +	}
> > +
> >   	for (i = 0, offset = 0; i < gpio->soc->num_ports; i++) {
> >   		const struct tegra_gpio_port *port = &gpio->soc->ports[i];
> >   		char *name;
> >
> >   		for (j = 0; j < port->pins; j++) {
> > -			if (gpio->soc->prefix)
> > -				gpio->gpio.parent, GFP_KERNEL, "%s-P%s.%02x",
> > -						      gpio->soc->prefix, port->name, j);
> > -			else
> > -				name = devm_kasprintf(gpio->gpio.parent, GFP_KERNEL, "P%s.%02x",
> > -						      port->name, j);
> > +			name = devm_kasprintf(gpio->gpio.parent, GFP_KERNEL, "%s%sP%s.%02x",
> > +					      instance ?: "", gpio->soc->prefix ?: "",
>
> instance can't be NULL here, can it? Either print it unconditionally or
> initialize it to NULL.

Probably best to drop 'instance' altogether and just ...

 if (node >= 0)
       name = devm_kasprintf(gpio->gpio.parent, GFP_KERNEL,
                             "%d-%sP%s.%02x", node,
                             gpio->soc->prefix ?: "",
                             port->name, j);
 else
       name = devm_kasprintf(gpio->gpio.parent, GFP_KERNEL,
                             "%sP%s.%02x",
                             gpio->soc->prefix ?: "",
                             port->name, j);

> > +					      port->name, j);
> >   			if (!name)
> >   				return -ENOMEM;
> >
> > @@ -1373,6 +1378,9 @@ static const struct tegra_gpio_soc tegra256_main_soc = {
> >   	.has_vm_support = true,
> >   };
> >
> > +/* Macro to define GPIO name prefix with separator */
> > +#define TEGRA_GPIO_PREFIX(_x)	_x "-"
> > +
>
> This doesn't look like it belongs to this patch, why are you doing it

Yes we should split this into two patches; one to simplify the prefix 
handling and one to handle the socket.

Jon

--
nvpublic