On Tue, May 13, 2014 at 5:51 PM, Stephen Hemminger
[off-list ref] wrote:

On Tue, 13 May 2014 13:44:24 -0700
Cong Wang [off-list ref] wrote:

quoted

On Tue, May 13, 2014 at 1:05 PM, Stephen Hemminger
[off-list ref] wrote:

quoted

That won't solve the problem.
 1. a pointer increase size of netns and requires refcount

Yes.

quoted

 2. iflink is still exposed in kernel user API and has to be set
    otherwise the device is falsely reported.

Then report 0 which is not a valid ifindex?

But that is what devices without a parent report.

Its parent of out of this netns, so this is expected. This is no way
to express "it's in another netns" with just ifindex. Maybe a new flag
would help.