IPNExtension uses 400% cpu when connected

Hi

I’ve moved to Monterey recently, and I have TS 1.16.2.

If I stay connected (which I have connected most of the time), and sleep my laptop, when I open it up again, IPNExtension is hitting around 400% CPU (I have an 8 core i9). Disconnecting the VPN + tailscale drops it back down.

Any ideas?

Thanks

1 Like

Would you please capture tcpdump of the tailscale interface while the problem is occurring? Something like;

tcpdump -ni utun5

where utun5 is the interface from ifconfig that has a tailscale IP address.

1 Like

FINALLY managed to get it to do it again - mostly noticable cos the fans on my laptop are running :slight_smile:

Currently using 320% (8 core mac).

TCP dump is a flow of data - as fast as the console will show, but its all largely the same:

13:50:14.316281 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 384
13:50:14.316311 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 416
13:50:14.316343 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, bad length 1280 > 1248
13:50:14.316361 IP 100.999.99.12 > 192.168.1.29: ip-proto-17
13:50:14.316390 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 128
13:50:14.316401 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 192
13:50:14.316430 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 224
13:50:14.316460 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, bad length 1312 > 1248
13:50:14.316465 IP 100.999.99.12 > 192.168.1.29: ip-proto-17
13:50:14.316496 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 128
13:50:14.316536 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 192
13:50:14.316557 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 352
13:50:14.316590 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 224
13:50:14.316623 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 448
13:50:14.316654 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 480
13:50:14.316685 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 1088
13:50:14.316723 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 1152
13:50:14.316740 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 320
13:50:14.316777 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 384
13:50:14.316812 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, bad length 1312 > 1248
13:50:14.316835 IP 100.999.99.12 > 192.168.1.29: ip-proto-17
13:50:14.316839 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 128
13:50:14.316858 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 256

the 100.999.99.12 is my laptop. the 1.29 is a server here, which is also an exit node, tho it’s not selected as an exit node right now. It’s also a router for 192.168.1/24?

if I leave it, I can’t ssh into the server unless I kill Tailscale. Now I’ve disconnected and reconnected TS, I’m down to a few packets every 5 seconds

13:54:36.879276 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 110
13:54:36.879638 IP 100.999.99.12.41641 > 192.168.1.47.41641: UDP, length 124
13:54:36.879935 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 124
13:54:42.818897 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 110
13:54:42.845501 IP 100.999.99.12.41641 > 192.168.1.47.41641: UDP, length 124
13:54:42.845948 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 124
13:54:48.819544 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 110
13:54:48.846473 IP 100.999.99.12.41641 > 192.168.1.47.41641: UDP, length 124
13:54:48.846854 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 124
13:54:49.425296 IP 100.999.99.12.24204 > 192.168.1.33.53: 49724+ A? play.google.com. (33)
13:54:49.485131 IP 192.168.1.33.53 > 100.999.99.12.24204: 49724 1/0/0 A 142.250.66.206 (49)
13:54:54.821233 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 110
13:54:54.848500 IP 100.999.99.12.41641 > 192.168.1.29.41641: UDP, length 124
13:54:54.848726 IP 100.999.99.12.41641 > 192.168.1.47.41641: UDP, length 124

seams to be a change in the 1.16.x series - I’m on 1.16.2, and I’ve not had issues with this before that came out. My network here is 192.168.0/20, so likely should update that route.

We have logged an internal Github issue for the same. We will keep you posted with progress or any further information needed. Feel free to share your sensitive information using email on support@tailscale.com rather than sharing over the forum for security reasons.

@darshinimashar out of interest, which bits up there are sensitive? Not sure about to 100. addresses - the 192’s are pretty generic tho…

Your Tailscale IPs start with 100.x.x.x. just so that you know that if you have anything else to share in future you can email us.

I am also facing this issue — running macOS Ventura, 1.34.1 (App Store), also on an 8 core i9.

I’m on a 15" MBPR with Monterey 12.6.2 and Tailscale 1.34.2.

I’m hitting the “400%” problem with IPNExteension. Switching off Tailscale VPN immediately drops the CPU off. This occurred while the laptop was locked and not in use. Turning Tailscale back on, IPNExtension seems to be behaving again.

Me too, 15" MacBook Pro, macOS 11.6.8, tailscale 1.36.0
Turned tailscale off, issue went away.