On Fri, Sep 2, 2016 at 7:09 AM, Heikki Krogerus
[off-list ref] wrote:

Hi,

On Wed, Aug 31, 2016 at 05:40:17PM -0700, Stephen Boyd wrote:

quoted

@@ -174,14 +219,37 @@ static int ulpi_register(struct device *dev, struct ulpi *ulpi)
      ulpi->id.product = ulpi_read(ulpi, ULPI_PRODUCT_ID_LOW);
      ulpi->id.product |= ulpi_read(ulpi, ULPI_PRODUCT_ID_HIGH) << 8;

+     return 0;
+}
+
+static int ulpi_register(struct device *dev, struct ulpi *ulpi)
+{
+     int ret;
+
      ulpi->dev.parent = dev;
      ulpi->dev.bus = &ulpi_bus;
      ulpi->dev.type = &ulpi_dev_type;
      dev_set_name(&ulpi->dev, "%s.ulpi", dev_name(dev));

+     if (IS_ENABLED(CONFIG_OF)) {

I don't think you need to check that in this case.

quoted

+             ret = ulpi_of_register(ulpi);
+             if (ret)
+                     return ret;
+     }
+
      ACPI_COMPANION_SET(&ulpi->dev, ACPI_COMPANION(dev));

ACPI_COMPANION_SET will overwrite the primary fwnode unconditionally,
so just to play it safe, do this before you call ulpi_of_register().

Ok.

quoted

-     request_module("ulpi:v%04xp%04x", ulpi->id.vendor, ulpi->id.product);
+     ret = ulpi_read_id(ulpi);
+     /*
+      * Ignore failure in case of DT node because the device may
+      * not be powered up yet but we can still match by compatible
+      */
+     if (ret && !ulpi->dev.of_node)
+             return ret;
+
+     if (of_device_request_module(&ulpi->dev))
+             request_module("ulpi:v%04xp%04x", ulpi->id.vendor,
+                            ulpi->id.product);

I don't think this works in all cases. If of_device_request_module()
fails and we don't have the id.vendor/product set, we should not
register the device. It also looks a bit messy.

How about just using of_device_request_module() call as fallback in
ulpi_read_id() and moving also request_module() call there:

Sure I'll fold it in and test. Should we "goto err" if we can't read
the scratch register though? I would think that's a "real" failure and
we shouldn't try to support DT in that case.