regulator: slg51000: use devm_gpiod_get_optional() in probe

The CS GPIO line is clearly optional GPIO (and marked as such in the
binding document) and we should handle it accordingly. The current code
treats all errors as meaning that there is no GPIO defined, which is
wrong, as it does not handle deferrals raised by the underlying code
properly, nor does it recognize non-existing GPIO from any other
initialization error.

As far as I can see the only reason the driver, unlike all others,
is using OF-specific devm_gpiod_get_from_of_node() so that it can
assign a custom label to the selected GPIO line. Given that noone else
needs that, it should not be doing that either.

Let's switch to using more appropriate devm_gpiod_get_optional().

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Link: https://lore.kernel.org/r/20190904214200.GA66118@dtor-ws
Signed-off-by: Mark Brown <broonie@kernel.org>

drivers/regulator/slg51000-regulator.c

@@ -447,19 +447,20 @@ static int slg51000_i2c_probe(struct i2c_client *client,
 {
 	struct device *dev = &client->dev;
 	struct slg51000 *chip;
-	struct gpio_desc *cs_gpiod = NULL;
+	struct gpio_desc *cs_gpiod;
 	int error, ret;
 
 	chip = devm_kzalloc(dev, sizeof(struct slg51000), GFP_KERNEL);
 	if (!chip)
 		return -ENOMEM;
 
-	cs_gpiod = devm_gpiod_get_from_of_node(dev, dev->of_node,
-					       "dlg,cs-gpios", 0,
-					       GPIOD_OUT_HIGH
-					       | GPIOD_FLAGS_BIT_NONEXCLUSIVE,
-					       "slg51000-cs");
-	if (!IS_ERR(cs_gpiod)) {
+	cs_gpiod = devm_gpiod_get_optional(dev, "dlg,cs",
+					   GPIOD_OUT_HIGH |
+						GPIOD_FLAGS_BIT_NONEXCLUSIVE);
+	if (IS_ERR(cs_gpiod))
+		return PTR_ERR(cs_gpiod);
+
+	if (cs_gpiod) {
 		dev_info(dev, "Found chip selector property\n");
 		chip->cs_gpiod = cs_gpiod;
 	}