On 08/31/2015 10:08 AM, Paul Bolle wrote:

Hi Maxime,

On ma, 2015-08-31 at 09:49 +0200, Maxime Coquelin wrote:

quoted

On 07/09/2015 10:17 AM, Paul Bolle wrote:

quoted

quoted

quoted

+static int __exit st_fdma_remove(struct platform_device *pdev)
+{
+	struct st_fdma_dev *fdev = platform_get_drvdata(pdev);
+
+	wait_for_completion(&fdev->fw_ack);
+
+	st_fdma_clk_disable(fdev);
+
+	return 0;
+}

Since this driver is built-in only this means st_fdma_remove() can
never be used, right?

It's not because a driver is built-in only that it does not need a
remove callback.
An instance can be probed/removed any time via driver's bind/unbind
SysFS entries.
Am I missing something?

(This discussion is moot because Peter already stated that a new version
will be modular.)

It follows from the __exit tag that st_fdma_remove() should never be
part of the kernel image (in this version of the patch), doesn't it?

Yes, you are right.
The remove callback is relevant, but without the __exit tag.

(I don't know what happens in this situation if an unbind sysfs entry is
used to remove a driver. I've never tried that.)

Not checked whether the function is omitted when built-in, but in any 
case, I thnk remove callbacks should not be tagged with __exit.

Regards,
Maxime