Comment 52 for bug 1711407

> Here is the output of the kernel when the "unregister_netdevice" bug was hit

Thanks! And just to verify, that's the full output of all those lines before the next "waiting..." line, right? If so, it looks like the problem in your case is not a kernel socket holding the netns open, it looks like a good-old dst leak, or something else holding a dst, so I'll need to add debug tracking those now...

Also, when you reproduce the problem, if you leave it for a while (like, 5 minutes or so), does it resolve itself? Or stay hung and printing the "waiting..." messages forever? Make sure to check the netns addr in the "waiting..." message to verify it's stuck waiting for the *same* netns to close forever, not just reproducing the problem over and over with different netns.

I'll get another kernel ready with additional dst debug. I think there is at least one dst leak patch upstream too, I'll check for those.