[PATCH usb-next v2 5/5] usb: dwc3: Add Apple Silicon DWC3 glue layer driver

[Philipp Zabel]

On Mi, 2025-10-15 at 15:40 +0000, Sven Peter wrote:
> The dwc3 controller present on Apple Silicon SoCs like the M1 requires
> a specific order of operations synchronized between its PHY and its
> Type-C controller. Specifically, the PHY first has to go through initial
> bringup (which requires knowledge of the lane mode and orientation)
> before dwc3 itself can be brought up and can then finalize the PHY
> configuration.
> Additionally, dwc3 has to be teared down and re-initialized whenever
> the cable is changed due to hardware quirks that prevent a new device
> from being recognized and due to the PHY being unable to switch lane
> mode or orientation while dwc3 is up and running.
> 
> These controllers also have a Apple-specific MMIO region after the
> common dwc3 region where some controls have to be updated. PHY bringup
> and shutdown also requires SUSPHY to be enabled for the ports to work
> correctly.
> 
> In the future, this driver will also gain support for USB3-via-USB4
> tunneling which will require additional tweaks.
> 
> Add a glue driver that takes of all of these constraints.
> 
> Reviewed-by: Neal Gompa <neal at gompa.dev>
> Acked-by: Thinh Nguyen <Thinh.Nguyen at synopsys.com>
> Signed-off-by: Sven Peter <sven at kernel.org>
> ---
>  MAINTAINERS                   |   1 +
>  drivers/usb/dwc3/Kconfig      |  11 +
>  drivers/usb/dwc3/Makefile     |   1 +
>  drivers/usb/dwc3/dwc3-apple.c | 489 ++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 502 insertions(+)
> 
[...]
> diff --git a/drivers/usb/dwc3/dwc3-apple.c b/drivers/usb/dwc3/dwc3-apple.c
> new file mode 100644
> index 0000000000000000000000000000000000000000..6e41bd0e34f461b0c3db9b8a646116458ff816b6
> --- /dev/null
> +++ b/drivers/usb/dwc3/dwc3-apple.c
> @@ -0,0 +1,489 @@
[...]
> +static int dwc3_apple_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct dwc3_apple *appledwc;
> +	int ret;
> +
> +	appledwc = devm_kzalloc(&pdev->dev, sizeof(*appledwc), GFP_KERNEL);
> +	if (!appledwc)
> +		return -ENOMEM;
> +
> +	appledwc->dev = &pdev->dev;
> +	mutex_init(&appledwc->lock);
> +
> +	appledwc->resets = devm_reset_control_array_get_exclusive(dev);

Why is this using an array? The bindings say there is only a single
reset.

regards
Philipp