Tue, Aug 14, 2012 at 03:14:00PM CEST, fbl@redhat.com wrote:

On Tue, 14 Aug 2012 14:24:33 +0200
Jiri Pirko [off-list ref] wrote:

quoted

Mon, Aug 13, 2012 at 07:52:17PM CEST, fbl@redhat.com wrote:

quoted

On Mon, 13 Aug 2012 17:27:00 +0200
Jiri Pirko [off-list ref] wrote:

quoted

+	/*
+	 * To prevent loops, check if dev is not upper device to upper_dev.
+	 */
+	if (__netdev_has_upper_dev(upper_dev, dev, true))
+		return -EBUSY;
+
+	if (__netdev_find_upper(dev, upper_dev))
+		return -EEXIST;

__netdev_has_upper_dev() can go all the way up finding the device and
the __netdev_find_upper() just check the first level.

I do not think this ordering is somewhat inportant.

it's not the order, see below:

quoted

quoted

I think it would be better to use:
__netdev_find_upper_dev(,,deep=true/false)
__netdev_has_upper(,)

It's their names.  Currently, the function ..._find_... look at
one level only, while the function ..._has_... does one or more
levels.  I think it's better to swap 'has' and 'find' in their names:

__netdev_find_upper_dev(,,deep=true/false) <-- find in all levels
__netdev_has_upper(,)  <-- check only the one level.

Oh, now I think I see your point. But realise this:

The main reason for __netdev_find_upper() is to find "struct upper" for
netdev_upper_dev_unlink(). Therefore the name is not
"__netdev_find_upper_dev" and there's no need to go deep here.

On the orher hand, __netdev_has_upper_dev() only says whether device is lower
to specified upper device. In this case I think the name is quite
convenient as well.

fbl