On Wed, Feb 24, 2016 at 08:19:58AM +0100, Jiri Pirko wrote:

Tue, Feb 23, 2016 at 10:19:48PM CET, jarod@redhat.com wrote:

quoted

On Tue, Feb 23, 2016 at 04:51:26PM +0100, Jiri Pirko wrote:

quoted

From: Jiri Pirko <redacted>

Introduce devlink infrastructure for drivers to register and expose to
userspace via generic Netlink interface.

There are two basic objects defined:
devlink - one instance for every "parent device", for example switch ASIC
devlink port - one instance for every physical port of the device.

This initial portion implements basic get/dump of objects to userspace.
Also, port splitter and port type setting is implemented.

...

quoted

+/**
+ *	devlink_alloc - Allocate new devlink instance resources
+ *
+ *	@ops: ops
+ *	@priv_size: size of user private data
+ *
+ *	Allocate new devlink instance resources, including devlink index
+ *	and name.
+ */
+struct devlink *devlink_alloc(const struct devlink_ops *ops, size_t priv_size)
+{
+	static atomic_t dev_counter = ATOMIC_INIT(0);
+	struct devlink *devlink;
+
+	devlink = kzalloc(sizeof(*devlink) + priv_size, GFP_KERNEL);
+	if (!devlink)
+		return NULL;
+	devlink->ops = ops;
+	devlink_net_set(devlink, &init_net);
+
+different_name:
+	devlink->index = atomic_inc_return(&dev_counter);
+	if (devlink->index < 0) {
+		/* wrapped */
+		atomic_dec(&dev_counter);
+		kfree(devlink);
+		return NULL;
+	}
+	/* atomic_inc_return makes it start at 1, make it start at 0 */
+	devlink->index--;
+
+	dev_set_name(&devlink->dev, DEVLINK_GENL_NAME "%d", devlink->index);
+	if (devlink_name_exists(devlink_net(devlink), devlink_name(devlink)))
+		goto different_name;
+
+	INIT_LIST_HEAD(&devlink->port_list);
+	device_initialize(&devlink->dev);
+	devlink->dev.class = &devlink_class;
+	devlink->dev.platform_data = devlink;
+	return devlink;
+}
+EXPORT_SYMBOL_GPL(devlink_alloc);

The increment then decrement bit feels... Clunky and a little confusing.
Why not just atomic_read(&dev_counter) before different_name:, then an
atomic_inc_return(&dev_counter) under the if clause before the goto, and
eliminate the devlink->index-- bit?

If I split it it would not be atomic op anymore. But the
devlink->index-- is just to adjust the start, nothing else.
This is copied from nl80211 code.

Argh, I'm a moron. You need to increment dev_counter, can't just read it's
current value. Still fugly, but okay, does actually serve a purpose this
way.

-- 
Jarod Wilson
jarod@redhat.com