On Thu, Jul 17, 2014 at 09:36:08AM -0700, Joe Perches wrote:

On Thu, 2014-07-17 at 18:25 +0200, Veaceslav Falico wrote:

quoted

On Thu, Jul 17, 2014 at 09:18:44AM -0700, Joe Perches wrote:

quoted

On Thu, 2014-07-17 at 16:09 +0200, Veaceslav Falico wrote:

quoted

netdev_name() returns dev->name only when the net_device is in
NETREG_REGISTERED state.

[]

quoted

quoted

Maybe this should not be inline and become something like:

It will miss the states then, when it's not NETREG_REGISTERED.

If it's registered, it has a valid name via
dev_get_valid_name() doesn't it?

Yes, I'm speaking about the NETREG_* states when it's not registered.

i.e.:

Jul 17 13:35:29 darkmag kernel: [  602.686489] bond2 (unregistering): Released all slaves

that's with my patchset, when the bond device is unregistering
(NETREG_UNREGISTERING). With your patch it would be only "bond2: Released
all slaves".

As the most time the device is in the NETREG_REGISTERED state, there will
be no differencies in the output (with either my or your approach), however
in less-common states/codepaths it will also output its state, which might
be quite valuable when analyzing the logs.

quoted

quoted

const char *netdev_name(const struct net_device *dev)
{
if (dev->reg_state == NETREG_REGISTERED)
	return dev->name;

if (!dev->name[0])
	return "(unnamed net_device)";

if (!strchr(dev->name, '%'))
	return "(unregistered net_device)";

return dev->name;
}
EXPORT_SYMBOL(netdev_name);