On 7/14/26 4:04 PM, Peng Fan wrote:
On Sat, Jul 11, 2026 at 10:00:52PM +0200, Marek Vasut wrote:
On 6/21/26 4:06 AM, Peng Fan (OSS) wrote:
From: Ye Li <[email protected]>
Call board_usb_init() during dwc3_generic_probe() and
board_usb_cleanup() during dwc3_generic_remove() to allow
board-level USB configuration (e.g. Type-C CC role, SS mux,
VBUS control) before DWC3 core initialization.
Signed-off-by: Ye Li <[email protected]>
Signed-off-by: Peng Fan <[email protected]>
...
dwc3_shutdown_phy(dev, &priv->phys);
unmap_physmem(dwc3->regs, MAP_NOCACHE);
+ if (mode == USB_DR_MODE_HOST)
+ board_usb_cleanup(dev_seq(dev), USB_INIT_HOST);
+ else if (mode == USB_DR_MODE_PERIPHERAL)
+ board_usb_cleanup(dev_seq(dev), USB_INIT_DEVICE);
+
This is supposed to be generic DT based glue code, can we avoid the
board-specific callbacks entirely ?
We need to invoke TCPM configuration code, dropping board callbacks means
we need put some #ifdef in dwc3 drivers:
#if CONFIG_IS_ENABLED(TYPEC_TCPM)
static int dwc3_typec_setup(struct udevice *dev, struct dwc3_generic_priv *priv,
enum usb_dr_mode mode)
{
/* Handling tcpm */
}
#else
static int dwc3_typec_setup(struct udevice *dev, struct dwc3_generic_priv *priv,
enum usb_dr_mode mode)
{
return 0;
}
#endif
I could use above in V3.
The typec_sel and typec_en GPIOs look like something that could be
handled by usb-c-connector , which could be made generic , can it not ?
How does Linux handle those GPIOs ?
I wanted to say look at the GPIO SBU mux Linux kernel driver, but that
operates those GPIOs as OUTPUTs and I think you need INPUTs here?
