Hi,
On Thu, 2019-10-17 at 09:34 -0700, Tony Lindgren wrote:

Hi,

Just few comments for future changes that might help below.

* min.guo@mediatek.com [off-list ref] [191017 09:42]:

quoted

--- /dev/null
+++ b/drivers/usb/musb/mediatek.c
+static int musb_usb_role_sx_set(struct device *dev, enum usb_role role)
+{
+	struct mtk_glue *glue = dev_get_drvdata(dev);
+	struct musb *musb = glue->musb;
+	u8 devctl = readb(musb->mregs + MUSB_DEVCTL);
+	enum usb_role new_role;
+
+	if (role == glue->role)
+		return 0;
+
+	switch (role) {
+	case USB_ROLE_HOST:
+		musb->xceiv->otg->state = OTG_STATE_A_WAIT_VRISE;
+		glue->phy_mode = PHY_MODE_USB_HOST;
+		new_role = USB_ROLE_HOST;
+		if (glue->role == USB_ROLE_NONE)
+			phy_power_on(glue->phy);
+
+		devctl |= MUSB_DEVCTL_SESSION;
+		musb_writeb(musb->mregs, MUSB_DEVCTL, devctl);
+		MUSB_HST_MODE(musb);
+		break;
+	case USB_ROLE_DEVICE:
+		musb->xceiv->otg->state = OTG_STATE_B_IDLE;
+		glue->phy_mode = PHY_MODE_USB_DEVICE;
+		new_role = USB_ROLE_DEVICE;
+		devctl &= ~MUSB_DEVCTL_SESSION;
+		musb_writeb(musb->mregs, MUSB_DEVCTL, devctl);
+		if (glue->role == USB_ROLE_NONE)
+			phy_power_on(glue->phy);
+
+		MUSB_DEV_MODE(musb);
+		break;
+	case USB_ROLE_NONE:
+		glue->phy_mode = PHY_MODE_USB_OTG;
+		new_role = USB_ROLE_NONE;
+		devctl &= ~MUSB_DEVCTL_SESSION;
+		musb_writeb(musb->mregs, MUSB_DEVCTL, devctl);
+		if (glue->role != USB_ROLE_NONE)
+			phy_power_off(glue->phy);
+
+		break;
+	default:
+		dev_err(glue->dev, "Invalid State\n");
+		return -EINVAL;
+	}
+
+	glue->role = new_role;
+	phy_set_mode(glue->phy, glue->phy_mode);
+
+	return 0;
+}

For the role change, I recently posted a patch "[PATCH 4/7] usb: musb:
Add musb_set_host and peripheral and use them for omap2430". That
should work for you looking at the code above, so later on you might
want to change to use that. Probably best done as a follow-up patch
to avoid adding extra dependencies to your series.

OK, thanks. I will try this patch.

Please also note that musb core attempts to do things automagically
on it's own. So trying to force mode in general does not work reliably.
This is because VBUS may not yet have risen for example.

VBUS control is through a GPIO on MediaTek musb controller. The host/device switching method is to use the driver of gpio-usb-b-connector, it sets the debounce of GPIO, the link of patch is as follows:
	Usb: roles: add USB Type-B GPIO connector driver
	Https://patchwork.kernel.org/patch/10966361/

The role change is best done based on the USB PHY as then usually
musb has already switched to the right mode automatically :)

Considering the use of type-c connector later, the foce phy method is used.

quoted

+static const struct musb_platform_ops mtk_musb_ops = {
+	.quirks = MUSB_DMA_INVENTRA,
+	.init = mtk_musb_init,
+	.get_toggle = mtk_musb_get_toggle,
+	.set_toggle = mtk_musb_set_toggle,
+	.exit = mtk_musb_exit,
+#ifdef CONFIG_USB_INVENTRA_DMA
+	.dma_init = musbhs_dma_controller_create_noirq,
+	.dma_exit = musbhs_dma_controller_destroy,
+#endif
+	.clearb = mtk_musb_clearb,
+	.clearw = mtk_musb_clearw,
+	.busctl_offset = mtk_musb_busctl_offset,
+	.set_mode = mtk_musb_set_mode,
+};

So you may want to consider getting rid of .set_mode completely
and rely on USB PHY calls instead.

In some cases you need to use struct phy_companion for set_vbus
depending how things are wired.

Regards,

Tony

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel