tailscale: gvisor: panic: Incrementing non-positive count 0xc000533180 on stack.PacketBuffer

What is the issue?

From email report:

we compiled 1.22.1 and are experiencing problems on two of our macOS machines. tailscaled crashes directly after startup.

The ports 6883, 6885 are used by a local program which is running in the background before tailscaled is started.

We do see this error on our 2 Intel macOS systems. Arm macOS does not show the error.

We had a similar problem with 1.22.0. Downgrading on one of the intel machines to 1.20.4 solved the problem.

2022/03/10 15:08:59 logtail started
2022/03/10 15:08:59 Program starting: v1.22.1-tc8fb4f8c7, Go 1.17.8-tsdce70b6: []string{"/usr/local/bin/tailscaled", "-state=/var/db/tailscale/tailscaled.state", "-tun", "utun9"}
2022/03/10 15:08:59 LogID: 459c4a8ac464c58028859fb0d0bb79ee88adf8db6a27e083c7a7d63242103f9e
2022/03/10 15:08:59 logpolicy: using system state directory "/Library/Tailscale"
2022/03/10 15:08:59 wgengine.NewUserspaceEngine(tun "utun9") ...
2022/03/10 15:08:59 dns: using dns.noopManager
2022/03/10 15:08:59 link state: interfaces.State{defaultRoute=en13 ifs={en13:[10.0.5.6/24]} v4=true v6=false}
2022/03/10 15:08:59 magicsock: disco key = d:f376519d5ab5ccd3
2022/03/10 15:08:59 Creating wireguard device...
2022/03/10 15:08:59 Bringing wireguard device up...
2022/03/10 15:08:59 Bringing router up...
2022/03/10 15:08:59 external route: up
2022/03/10 15:08:59 Clearing router settings...
2022/03/10 15:08:59 Starting link monitor...
2022/03/10 15:08:59 Engine created.
2022/03/10 15:09:01 Start
2022/03/10 15:09:01 using backend prefs for "_daemon": Prefs{ra=true dns=true want=true routes=[0.0.0.0/0 ::/0 10.0.5.0/24] snat=true Persist{lm=, o=, n=[jR0DB] u="REDACTED"}}
2022/03/10 15:09:01 Backend: logs: be:REDACTED fe:
2022/03/10 15:09:01 control: client.Login(false, 0)
2022/03/10 15:09:01 control: doLogin(regen=false, hasUrl=false)
2022/03/10 15:09:01 control: control server key [nlFWp] from https://controlplane.tailscale.com/
2022/03/10 15:09:01 control: RegisterReq: onode= node=[jR0DB] fup=false
2022/03/10 15:09:02 control: RegisterReq: got response; nodeKeyExpired=false, machineAuthorized=true; authURL=false
2022/03/10 15:09:02 active login: REDACTED
2022/03/10 15:09:02 Switching ipn state NoState -> Starting (WantRunning=true, nm=true)
2022/03/10 15:09:02 health("overall"): error: state=Starting, wantRunning=true
2022/03/10 15:09:02 magicsock: SetPrivateKey called (init)
2022/03/10 15:09:02 wgengine: Reconfig: configuring userspace wireguard config (with 1/10 peers)
2022/03/10 15:09:02 wgengine: Reconfig: configuring router
2022/03/10 15:09:02 wgengine: Reconfig: configuring DNS
2022/03/10 15:09:02 dns: Set: {DefaultResolvers:[] Routes:{} SearchDomains:[] Hosts:11}
2022/03/10 15:09:02 dns: Resolvercfg: {Routes:{} Hosts:11 LocalDomains:[]}
2022/03/10 15:09:02 dns: OScfg: {Nameservers:[] SearchDomains:[] MatchDomains:[]}
2022/03/10 15:09:02 peerapi: serving on http://100.x.y.z:35528/
2022/03/10 15:09:02 peerapi: serving on http://[fd7a:115c:a1e0:ab12:q:r:s:t]:35528/
2022/03/10 15:09:02 magicsock: home is now derp-1 (nyc)
2022/03/10 15:09:02 magicsock: adding connection to derp-1 for home-keep-alive
2022/03/10 15:09:02 control: NetInfo: NetInfo{varies=false hairpin=false ipv6=false udp=true derp=#1 portmap= link=""}
2022/03/10 15:09:02 magicsock: 1 active derp conns: derp-1=cr0s,wr0s
2022/03/10 15:09:02 derphttp.Client.Connect: connecting to derp-1 (nyc)
2022/03/10 15:09:02 Switching ipn state Starting -> Running (WantRunning=true, nm=true)
2022/03/10 15:09:02 magicsock: endpoints changed: a.b.c.d:57147 (stun), 10.0.5.6:57147 (local)
2022/03/10 15:09:02 magicsock: adding connection to derp-4 for [/+7hj]
2022/03/10 15:09:02 magicsock: 2 active derp conns: derp-1=cr83ms,wr83ms derp-4=cr0s,wr0s
2022/03/10 15:09:02 derphttp.Client.Recv: connecting to derp-4 (fra)
2022/03/10 15:09:03 magicsock: derp-1 connected; connGen=1
2022/03/10 15:09:03 health("overall"): ok
2022/03/10 15:09:03 magicsock: derp-4 connected; connGen=1
2022/03/10 15:09:03 wgengine: idle peer [HOh7S] now active, reconfiguring wireguard
2022/03/10 15:09:03 wgengine: Reconfig: configuring userspace wireguard config (with 2/10 peers)
2022/03/10 15:09:03 magicsock: disco: node [HOh7S] d:c2b10c247333c6cc now using 10.0.5.8:62659
2022/03/10 15:09:03 Accept: UDP{100.x.y.1:6883 > 10.0.5.6:6885} 48 ok
2022/03/10 15:09:03 Accept: UDP{100.x.y.1:6883 > 10.0.5.6:6885} 48 ok
2022/03/10 15:09:03 Accept: UDP{100.x.y.1:6883 > 10.0.5.6:6885} 48 ok
2022/03/10 15:09:03 netstack: could not bind local port 6883: listen udp 0.0.0.0:6883: bind: address already in use, trying again with random port
panic: Incrementing non-positive count 0xc000533180 on stack.PacketBuffer

Steps to reproduce

No response

Are there any recent changes that introduced the issue?

Panic first appeared in 1.22.0.

OS

macOS

OS version

No response

Tailscale version

1.22

Bug report

No response

Front conversations