Hi Romain,
On 2026-07-02T08:11:07, Romain Gantois <[email protected]> wrote:
> usb: gadget: Add Renesas RZ/N1 USBF controller support
>
> Add support for the Renesas USBF controller. This is an USB2.0 UDC
> controller available in the RZ/N1 SoC.
>
> This driver was originally written by Hervé Codina for the Linux kernel.
> Suspend/resume features have been removed in this U-Boot port.
>
> Linux revision this was ported from:
>
> 74851fbb6d64 usb: gadget: renesas_usbf: Handle devm_pm_runtime_...
>
> Signed-off-by: Herve Codina <[email protected]>
> Co-developed-by: Romain Gantois (Schneider Electric)
> <[email protected]>
> Signed-off-by: Romain Gantois (Schneider Electric)
> <[email protected]>
> Tested-by: Ralph Siemsen <[email protected]>
>
> MAINTAINERS | 5 +
> drivers/usb/gadget/Kconfig | 9 +
> drivers/usb/gadget/Makefile | 1 +
> drivers/usb/gadget/renesas_usbf.c | 3036
> +++++++++++++++++++++++++++++++++++++
> 4 files changed, 3051 insertions(+)
> This driver was originally written by Hervé Codina for the Linux kernel.
> Suspend/resume features have been removed in this U-Boot port.
Please also mention the two other removals noted in the cover letter -
handling of endpoints > 2 and dynamic detection of endpoint DMA
capability. Otherwise users will be surprised when a gadget needing
more than two endpoints fails to bind.
> + for (i = 0; i < ARRAY_SIZE(udc->ep); i++) {
> + ep = &udc->ep[i];
> +
> + if (!(usbf_reg_readl(udc, USBF_REG_USBSSCONF) &
> + USBF_SYS_EP_AVAILABLE(i))) {
> + continue;
> + }
> +
> + INIT_LIST_HEAD(&ep->queue);
If the availability check fails, INIT_LIST_HEAD() is never called and
ep->disabled stays 0. usbf_reset() and usbf_detach() then iterate over
all slots and only skip disabled == 1, so they will call
usbf_ep_nuke() on an ep whose queue list is uninitialised. On RZ/N1
all three endpoints are always present so this is latent, but please
set ep->disabled = 1 before the continue (or at least init the
list_head).
> diff --git a/drivers/usb/gadget/Kconfig b/drivers/usb/gadget/Kconfig
> @@ -159,6 +159,15 @@ config CI_UDC
> +config USB_GADGET_RENESAS_USBF
> + bool 'Renesas USB Function controller'
> + depends on RZN1
> + depends on DM_USB_GADGET
> + select USB_GADGET_DUALSPEED
> + help
> + Renesas USB Function controller is a USB peripheral controller
> + available on RZ/N1 Renesas SoCs.
Two Kconfig nits: prompts use double quotes throughout the file. Also,
the help text is indented one space more than the tab + two spaces
convention used by the neighbouring entries - compare with CI_UDC just
above.
> diff --git a/drivers/usb/gadget/renesas_usbf.c
> b/drivers/usb/gadget/renesas_usbf.c
> @@ -0,0 +1,3036 @@
> +static int usbf_remove(struct udevice *dev)
> +{
> + struct usbf_udc *udc = dev_get_priv(dev);
> +
> + usb_del_gadget_udc(&udc->gadget);
> +
> + return 0;
> +}
Just to check - if the device is removed while the pull-up is still
asserted, D+ stays driven and interrupts remain enabled at the EPC
level. Would it be worth calling usbf_detach() here so remove() always
leaves the hardware quiescent? What do you think?
Regards,
Simon
