On Wed, Feb 21, 2018 at 01:33:11PM +0900, Katsuhiro Suzuki wrote:
+++ b/sound/soc/codecs/bd28623.c @@ -0,0 +1,258 @@ +// SPDX-License-Identifier: GPL-2.0 +/*
- ROHM BD28623MUV class D speaker amplifier codec driver.
Please make the entire comment C++ so this looks intentional.
dev_err(dev, "Failed to enable supplies: %d\n", ret);return ret;- }
- gpiod_set_value(bd->reset_gpio, 0);
Since this GPIO is not needed in atomic contexts you should use the _cansleep() versions of the GPIO functions - it doesn't cost you anything and means that if for some reason someone wired this up to a GPIO that can't be used in atomic context the driver will just work.
- bd->reset_gpio = devm_gpiod_get_optional(dev, "reset",
GPIOD_OUT_HIGH);
- bd->mute_gpio = devm_gpiod_get_optional(dev, "mute",
GPIOD_OUT_HIGH);
These properties were documented as mandatory in the binding but are optional here. It's fine that they're optional but I'd expect the binding to be consistent with this.
+static int bd28623_remove(struct platform_device *pdev) +{
- struct bd28623_priv *bd = platform_get_drvdata(pdev);
- regulator_bulk_disable(ARRAY_SIZE(bd->supplies), bd->supplies);
- return 0;
+}
We don't enable the supplies explicitly as part of the probe function so it feels wrong to disable on remove() - I'm sure it is fine in practice as-is but I'd have to think too hard to confirm that. I'd put this in a component level remove function instead so that it's consistent.