On 11/03/2019 22:56, Martin Blumenstingl wrote:

Hi Neil,

On Mon, Mar 4, 2019 at 11:41 AM Neil Armstrong [off-list ref] wrote:
[...]

quoted

+static void dwc3_meson_g12a_usb_init_mode(struct dwc3_meson_g12a *priv)
+{
+       if (priv->otg_phy_mode == PHY_MODE_USB_DEVICE) {
+               regmap_update_bits(priv->regmap, USB_R0,
+                               USB_R0_U2D_ACT, USB_R0_U2D_ACT);
+               regmap_update_bits(priv->regmap, USB_R0,
+                               USB_R0_U2D_SS_SCALEDOWN_MODE_MASK, 0);
+               regmap_update_bits(priv->regmap, USB_R4,
+                               USB_R4_P21_SLEEP_M0, USB_R4_P21_SLEEP_M0);
+       } else {
+               regmap_update_bits(priv->regmap, USB_R0,
+                               USB_R0_U2D_ACT, 0);
+               regmap_update_bits(priv->regmap, USB_R4,
+                               USB_R4_P21_SLEEP_M0, 0);
+       }
+}

I was already confused by the name of this function in v1.
do you think "dwc3_meson_g12a_usb_otg_apply_mode" is a suitable name?

Yep changed

[...]

quoted

+static int dwc3_meson_g12a_otg_mode_set(struct dwc3_meson_g12a *priv,
+                                       enum phy_mode mode)
+{
+       int ret;
+
+       if (!priv->phys[USB2_OTG_PHY])
+               return -EINVAL;
+
+       if (mode == PHY_MODE_USB_HOST)
+               dev_info(priv->dev, "switching to Host Mode\n");
+       else
+               dev_info(priv->dev, "switching to Device Mode\n");
+
+       if (priv->vbus) {
+               if (mode == PHY_MODE_USB_DEVICE)
+                       ret = regulator_disable(priv->vbus);
+               else
+                       ret = regulator_enable(priv->vbus);

do we need to track the regulator status (whether it's enabled or not)?
the regulator framework WARN()s if it detects "unbalanced disables for
<regulator>"
(I haven't tested this on one of my boards yet, so maybe it works
because the callers of dwc3_meson_g12a_otg_mode_set() protect against
this by not calling this function if the mode doesn't change)

I handled this case by enabling the regulator _before_, as we default in
Host/OTG mode in probe, then eventually switch to device.

In this case, we don't have to track the status, and further calls
to dwc3_meson_g12a_otg_mode_set() will only change the regulator
on mode change.

quoted

+               if (ret)
+                       return ret;
+               }
+
+               priv->otg_phy_mode = mode;
+
+               dwc3_meson_g12a_usb2_set_mode(priv, USB2_OTG_PHY, mode);
+
+               dwc3_meson_g12a_usb_init_mode(priv);
+
+               return phy_set_mode(priv->phys[USB2_OTG_PHY], mode);
+ }

this is the only place where phy_set_mode is called
I'm fine with keeping it but then it should be consistent at least for
all USB2 PHYs.
I suggest to either move the phy_set_mode call to
dwc3_meson_g12a_usb2_set_mode or to remove it.

I will remove it, it's useless for now.

[...]

quoted

+static int dwc3_meson_g12a_role_set(struct device *dev, enum usb_role role)
+{
+       struct dwc3_meson_g12a *priv = dev_get_drvdata(dev);
+       enum phy_mode mode;
+
+       if (role == USB_ROLE_NONE)
+               return 0;
+
+       mode = role == USB_ROLE_HOST ? PHY_MODE_USB_HOST : PHY_MODE_USB_DEVICE;

(without surrounding parens I find the "role == USB_ROLE_HOST" part
hard to distinguish from the "mode =" operation. if more people think
this way then please speak up - otherwise it's probably just my
personal taste)

Yep, I added parenthesis and changed indentation to make it clearer.

[...]

quoted

+static struct device *dwc3_meson_g12_find_child(struct device *dev,
+                                               const char *compatible)
+{
+       struct platform_device *pdev;
+       struct device_node *np;
+
+       np = of_find_compatible_node(dev->of_node, NULL, compatible);
+       if (!np)
+               return NULL;
+
+       pdev = of_find_device_by_node(np);

maybe switch to of_get_compatible_child() here? This was done for the
MMC driver in c483a5cc9d09f4 ("mmc: meson-mx-sdio: fix OF child-node
lookup"), but I'm not sure if the problem described there also applies
to dwc3_meson_g12_find_child

of_get_compatible_child() is new to me, switching to it !

[...]

quoted

+static int dwc3_meson_g12a_probe(struct platform_device *pdev)
+{
+       struct dwc3_meson_g12a  *priv;
+       struct device           *dev = &pdev->dev;
+       struct device_node      *np = dev->of_node;
+       void __iomem *base;
+       struct resource *res;
+       enum phy_mode otg_id;
+       int ret, i;
+
+       priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+       if (!priv)
+               return -ENOMEM;
+
+       res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       base = devm_ioremap_resource(dev, res);
+       if (IS_ERR(base))
+               return PTR_ERR(base);
+
+       priv->regmap = devm_regmap_init_mmio(dev, base,
+                                            &phy_meson_g12a_usb3_regmap_conf);
+       if (IS_ERR(priv->regmap))
+               return PTR_ERR(priv->regmap);
+
+       priv->vbus = devm_regulator_get_optional(dev, "vbus");
+       if (IS_ERR(priv->vbus)) {
+               if (PTR_ERR(priv->vbus) == -EPROBE_DEFER)
+                       return PTR_ERR(priv->vbus);
+               priv->vbus = NULL;
+       }
+
+       priv->clk = devm_clk_get(dev, NULL);
+       if (IS_ERR(priv->clk))
+               return PTR_ERR(priv->clk);
+
+       ret = clk_prepare_enable(priv->clk);
+       if (ret)
+               return ret;
+
+       platform_set_drvdata(pdev, priv);
+       priv->dev = dev;
+
+       priv->reset = devm_reset_control_get(dev, NULL);
+       if (IS_ERR(priv->reset)) {
+               ret = PTR_ERR(priv->reset);
+               dev_err(dev, "failed to get device reset, err=%d\n", ret);
+               return ret;
+       }

clk_prepare_enable is called a few lines above but this (and a few
more) error-paths don't call clk_disable_unprepare.
Jerome suggested in the dwmac-meson8b driver that I use something
like, which will even allow you to drop the clk_disable_unprepare call
from dwc3_meson_g12a_remove and catch all error cases in
dwc3_meson_g12a_probe at the same time:
  devm_add_action_or_reset(dev, (void(*)(void
*))clk_disable_unprepare, priv->clk);
(if you go with this then you also need to remove the
clk_disable_unprepare after of_platform_populate)

Indeed I totally forgot this, thanks !

[...]

quoted

+static int dwc3_meson_g12a_remove(struct platform_device *pdev)
+{
+       struct dwc3_meson_g12a *priv = platform_get_drvdata(pdev);
+       struct device *dev = &pdev->dev;
+       int i;
+
+       usb_role_switch_unregister(priv->role_switch);
+
+       of_platform_depopulate(dev);
+
+       for (i = 0 ; i < PHY_COUNT ; ++i) {
+               phy_power_off(priv->phys[i]);
+               phy_exit(priv->phys[i]);
+               phy_put(priv->phys[i]);
+       }
+
+       clk_disable_unprepare(priv->clk);
+       clk_put(priv->clk);

priv->clk is obtained with devm_clk_get so the common clock framework
will call clk_put for us automatically

Right

[...]

quoted

+static int __maybe_unused dwc3_meson_g12a_suspend(struct device *dev)
+{
+       struct dwc3_meson_g12a *priv = dev_get_drvdata(dev);
+       int i;
+
+       for (i = 0 ; i < PHY_COUNT ; ++i)
+               if (priv->phys[i])
+                       phy_exit(priv->phys[i]);

phy_init is NULL-safe, so you can drop the NULL-check above

Forgot this one

quoted

+
+       reset_control_assert(priv->reset);
+
+       return 0;
+}
+
+static int __maybe_unused dwc3_meson_g12a_resume(struct device *dev)
+{
+       struct dwc3_meson_g12a *priv = dev_get_drvdata(dev);
+       int i, ret;
+
+       reset_control_deassert(priv->reset);
+
+       dwc3_meson_g12a_usb_init(priv);
+
+       /* Init PHYs */
+       for (i = 0 ; i < PHY_COUNT ; ++i) {
+               if (priv->phys[i]) {
+                       ret = phy_init(priv->phys[i]);

phy_init is NULL-safe, so you can drop the NULL-check above

Good catch...

Regards
Martin

Thanks !!

Neil

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