On Thu, Sep 7, 2017 at 12:21 PM, Marcel Holtmann [off-list ref] wrote:

Hi Fred,

quoted

Signed-off-by: Frédéric Danis <redacted>
---
drivers/tty/serdev/core.c | 99 ++++++++++++++++++++++++++++++++++++++++++++---
1 file changed, 94 insertions(+), 5 deletions(-)

[...]

quoted

@@ -404,9 +488,14 @@ int serdev_controller_add(struct serdev_controller *ctrl)
      if (ret)
              return ret;

-     ret = of_serdev_register_devices(ctrl);
-     if (ret)
+     ret_of = of_serdev_register_devices(ctrl);
+     ret_acpi = acpi_serdev_register_devices(ctrl);
+     if (ret_of && ret_acpi) {
+             dev_dbg(&ctrl->dev, "serdev%d no devices registered: of:%d acpi:%d\n",
+                     ctrl->nr, ret_of, ret_acpi);
+             ret = -ENODEV;
              goto out_dev_del;
+     }

      dev_dbg(&ctrl->dev, "serdev%d registered: dev:%p\n",
              ctrl->nr, &ctrl->dev);

Shouldn’t we just consider to always register the controller? Even if there are no devices attached to it.

You argued for the opposite at least in regards to a serdev ldisc. :)
The problem is we use the success or failure here to decide if we
create a tty char dev or not. I guess we could move that decision out
of the core and let the tty code check for devices and decide.

Rob