]> git.baikalelectronics.ru Git - kernel.git/commitdiff
drm/panel: atna33xc20: Take advantage of wait_hpd_asserted() in struct drm_dp_aux
authorDouglas Anderson <dianders@chromium.org>
Tue, 14 Jun 2022 21:54:17 +0000 (14:54 -0700)
committerDouglas Anderson <dianders@chromium.org>
Tue, 21 Jun 2022 16:06:31 +0000 (09:06 -0700)
Let's add support for being able to read the HPD pin even if it's
hooked directly to the controller. This will let us take away the
waiting in the AUX transfer functions of the eDP controller drivers.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20220614145327.v4.3.I9ee239f6b95b944c8fa030f300ad222a7af9899d@changeid
drivers/gpu/drm/panel/panel-samsung-atna33xc20.c

index 3dd10412d147609c2422c6a2ff6ba1b33b7c149a..5a8b978c64158dab5e953239e0718131346cd885 100644 (file)
 #include <drm/drm_edid.h>
 #include <drm/drm_panel.h>
 
+/* T3 VCC to HPD high is max 200 ms */
+#define HPD_MAX_MS     200
+#define HPD_MAX_US     (HPD_MAX_MS * 1000)
+
 struct atana33xc20_panel {
        struct drm_panel base;
        bool prepared;
@@ -30,6 +34,7 @@ struct atana33xc20_panel {
 
        struct regulator *supply;
        struct gpio_desc *el_on3_gpio;
+       struct drm_dp_aux *aux;
 
        struct edid *edid;
 
@@ -79,7 +84,7 @@ static int atana33xc20_suspend(struct device *dev)
 static int atana33xc20_resume(struct device *dev)
 {
        struct atana33xc20_panel *p = dev_get_drvdata(dev);
-       bool hpd_asserted = false;
+       int hpd_asserted;
        int ret;
 
        /* T12 (Power off time) is min 500 ms */
@@ -90,23 +95,41 @@ static int atana33xc20_resume(struct device *dev)
                return ret;
        p->powered_on_time = ktime_get();
 
-       /*
-        * Handle HPD. Note: if HPD is hooked up to a dedicated pin on the
-        * eDP controller then "no_hpd" will be false _and_ "hpd_gpio" will be
-        * NULL. It's up to the controller driver to wait for HPD after
-        * preparing the panel in that case.
-        */
        if (p->no_hpd) {
-               /* T3 VCC to HPD high is max 200 ms */
-               msleep(200);
-       } else if (p->hpd_gpio) {
+               msleep(HPD_MAX_MS);
+               return 0;
+       }
+
+       if (p->hpd_gpio) {
                ret = readx_poll_timeout(gpiod_get_value_cansleep, p->hpd_gpio,
                                         hpd_asserted, hpd_asserted,
-                                        1000, 200000);
-               if (!hpd_asserted)
-                       dev_warn(dev, "Timeout waiting for HPD\n");
+                                        1000, HPD_MAX_US);
+               if (hpd_asserted < 0)
+                       ret = hpd_asserted;
+
+               if (ret)
+                       dev_warn(dev, "Error waiting for HPD GPIO: %d\n", ret);
+
+               return ret;
        }
 
+       if (p->aux->wait_hpd_asserted) {
+               ret = p->aux->wait_hpd_asserted(p->aux, HPD_MAX_US);
+
+               if (ret)
+                       dev_warn(dev, "Controller error waiting for HPD: %d\n", ret);
+
+               return ret;
+       }
+
+       /*
+        * Note that it's possible that no_hpd is false, hpd_gpio is
+        * NULL, and wait_hpd_asserted is NULL. This is because
+        * wait_hpd_asserted() is optional even if HPD is hooked up to
+        * a dedicated pin on the eDP controller. In this case we just
+        * assume that the controller driver will wait for HPD at the
+        * right times.
+        */
        return 0;
 }
 
@@ -263,6 +286,8 @@ static int atana33xc20_probe(struct dp_aux_ep_device *aux_ep)
                return -ENOMEM;
        dev_set_drvdata(dev, panel);
 
+       panel->aux = aux_ep->aux;
+
        panel->supply = devm_regulator_get(dev, "power");
        if (IS_ERR(panel->supply))
                return dev_err_probe(dev, PTR_ERR(panel->supply),