2733 Commits (24db1a3c9b5e26e6ba4677abf762f9ea7a26cc58)

 

All Branches   

Search

Author	SHA1	Message	Date
Josh Bleecher Snyder	4b51fbf48c	internal/deephash: increase scratch space size ... e66d4e4c81 added AppendTo methods to some key types. Their marshaled form is longer than 64 bytes. name old time/op new time/op delta Hash-8 15.5µs ± 1% 14.8µs ± 1% -4.17% (p=0.000 n=9+9) name old alloc/op new alloc/op delta Hash-8 1.18kB ± 0% 0.47kB ± 0% -59.87% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Hash-8 12.0 ± 0% 6.0 ± 0% -50.00% (p=0.000 n=10+10) This is still a bit worse than explicitly handling the types, but much nicer. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Brad Fitzpatrick	e66d4e4c81	tailcfg, types/wgkey: add AppendTo methods on some types ... Add MarshalText-like appending variants. Like: https://pkg.go.dev/inet.af/netaddr#IP.AppendTo To be used by @josharian's pending deephash optimizations. Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Josh Bleecher Snyder	b340beff8e	internal/deephash: reset scratch before appending to it ... Oops. In practice this doesn't matter, but it's still wrong. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	15a7ff83de	internal/deephash: remove netaddr special cases ... All netaddr types that we are concerned with now implement AppendTo. Use the AppendTo method if available, and remove all references to netaddr. This is slower but cleaner, and more readily re-usable by others. name old time/op new time/op delta Hash-8 12.6µs ± 0% 14.8µs ± 1% +18.05% (p=0.000 n=8+10) HashMapAcyclic-8 21.4µs ± 1% 21.9µs ± 1% +2.39% (p=0.000 n=10+9) name old alloc/op new alloc/op delta Hash-8 408B ± 0% 408B ± 0% ~ (p=1.000 n=10+10) HashMapAcyclic-8 1.00B ± 0% 1.00B ± 0% ~ (all equal) name old allocs/op new allocs/op delta Hash-8 6.00 ± 0% 6.00 ± 0% ~ (all equal) HashMapAcyclic-8 0.00 0.00 ~ (all equal) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	051d2f47e5	internal/deephash: re-use MapIter ... name old time/op new time/op delta Hash-8 12.4µs ± 0% 12.4µs ± 0% -0.33% (p=0.002 n=10+9) HashMapAcyclic-8 21.2µs ± 0% 21.3µs ± 0% +0.45% (p=0.000 n=8+8) name old alloc/op new alloc/op delta Hash-8 793B ± 0% 408B ± 0% -48.55% (p=0.000 n=10+10) HashMapAcyclic-8 128B ± 0% 0B -100.00% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Hash-8 9.00 ± 0% 6.00 ± 0% -33.33% (p=0.000 n=10+10) HashMapAcyclic-8 1.00 ± 0% 0.00 -100.00% (p=0.000 n=10+10) Depends on https://github.com/golang/go/issues/46293. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	c06ec45f09	internal/deephash: document MapIter shims ... These exist so we can use the optimized MapIter APIs while still working with released versions of Go. They're pretty simple, but some docs won't hurt. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Elias Naur	adfe8cf41d	paths: generalize IOSSharedDir to cover Android ... Also fix an error message while here. Signed-off-by: Elias Naur <mail@eliasnaur.com>	4 years ago
Josh Bleecher Snyder	73adbb7a78	wgengine: pass an addressable value to deephash.UpdateHash ... This makes deephash more efficient. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	ce7a87e5e4	internal/deephash: use hash.BlockSize instead of a constant ... Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	135b641332	internal/deephash: add re-usable scratch space ... name old time/op new time/op delta Hash-8 13.9µs ± 0% 12.5µs ± 0% -10.10% (p=0.008 n=5+5) name old alloc/op new alloc/op delta Hash-8 793B ± 0% 793B ± 0% ~ (all equal) name old allocs/op new allocs/op delta Hash-8 14.0 ± 0% 12.0 ± 0% -14.29% (p=0.008 n=5+5) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	988dfcabef	internal/deephash: simplify API ... Reduce to just a single external endpoint. Convert from a variadic number of interfaces to a slice there. name old time/op new time/op delta Hash-8 14.4µs ± 0% 14.0µs ± 1% -3.08% (p=0.000 n=9+9) name old alloc/op new alloc/op delta Hash-8 873B ± 0% 793B ± 0% -9.16% (p=0.000 n=9+6) name old allocs/op new allocs/op delta Hash-8 18.0 ± 0% 14.0 ± 0% -22.22% (p=0.000 n=10+10) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	b371588ce6	internal/deephash: use netaddr AppendTo methods ... Slightly slower, but lots less garbage. We will recover the speed lost in a follow-up commit. name old time/op new time/op delta Hash-8 13.5µs ± 1% 14.3µs ± 0% +5.84% (p=0.000 n=10+9) name old alloc/op new alloc/op delta Hash-8 1.46kB ± 0% 0.87kB ± 0% -40.10% (p=0.000 n=7+10) name old allocs/op new allocs/op delta Hash-8 43.0 ± 0% 18.0 ± 0% -58.14% (p=0.000 n=10+10) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	09afb8e35b	internal/deephash: re-use map iteration values ... This requires changes to the Go toolchain. The changes are upstream at https://golang.org/cl/320929. They haven't been pulled into our fork yet. No need to allocate new iteration scratch values for every map. name old time/op new time/op delta Hash-8 13.6µs ± 0% 13.5µs ± 0% -1.01% (p=0.008 n=5+5) HashMapAcyclic-8 21.2µs ± 1% 21.1µs ± 2% ~ (p=0.310 n=5+5) name old alloc/op new alloc/op delta Hash-8 1.58kB ± 0% 1.46kB ± 0% -7.60% (p=0.008 n=5+5) HashMapAcyclic-8 152B ± 0% 128B ± 0% -15.79% (p=0.008 n=5+5) name old allocs/op new allocs/op delta Hash-8 49.0 ± 0% 43.0 ± 0% -12.24% (p=0.008 n=5+5) HashMapAcyclic-8 4.00 ± 0% 2.00 ± 0% -50.00% (p=0.008 n=5+5) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	a2d7a2aeb1	internal/deephash: use MapIter.Set{Key,Value} ... To get the benefit of this optimization requires help from the Go toolchain. The changes are upstream at https://golang.org/cl/320929, and have been pulled into the Tailscale fork at 728ecc58fd. It also requires building with the build tag tailscale_go. name old time/op new time/op delta Hash-8 14.0µs ± 0% 13.6µs ± 0% -2.88% (p=0.008 n=5+5) HashMapAcyclic-8 24.3µs ± 1% 21.2µs ± 1% -12.47% (p=0.008 n=5+5) name old alloc/op new alloc/op delta Hash-8 2.16kB ± 0% 1.58kB ± 0% -27.01% (p=0.008 n=5+5) HashMapAcyclic-8 2.53kB ± 0% 0.15kB ± 0% -93.99% (p=0.008 n=5+5) name old allocs/op new allocs/op delta Hash-8 77.0 ± 0% 49.0 ± 0% -36.36% (p=0.008 n=5+5) HashMapAcyclic-8 202 ± 0% 4 ± 0% -98.02% (p=0.008 n=5+5) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com> setkey	4 years ago
Josh Bleecher Snyder	020e904f4e	internal/deephash: add special handling for netaddr.IPPort ... The acyclic map code interacts badly with netaddr.IPs. One of the netaddr.IP fields is an *intern.Value, and we use a few sentinel values. Those sentinel values make many of the netaddr data structures appear cyclic. One option would be to replace the cycle-detection code with a Floyd-Warshall style algorithm. The downside is that this will take longer to detect cycles, particularly if the cycle is long. This problem is exacerbated by the fact that the acyclic cycle detection code shares a single visited map for the entire data structure, not just the subsection of the data structure localized to the map. Unfortunately, the extra allocations and work (and code) to use per-map visited maps make this option not viable. Instead, continue to special-case netaddr data types. name old time/op new time/op delta Hash-8 22.4µs ± 0% 14.0µs ± 0% -37.59% (p=0.008 n=5+5) HashMapAcyclic-8 23.8µs ± 0% 24.3µs ± 1% +1.75% (p=0.008 n=5+5) name old alloc/op new alloc/op delta Hash-8 2.49kB ± 0% 2.16kB ± 0% ~ (p=0.079 n=4+5) HashMapAcyclic-8 2.53kB ± 0% 2.53kB ± 0% ~ (all equal) name old allocs/op new allocs/op delta Hash-8 86.0 ± 0% 77.0 ± 0% -10.47% (p=0.008 n=5+5) HashMapAcyclic-8 202 ± 0% 202 ± 0% ~ (all equal) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	bbb79f2d6a	internal/deephash: fix accidental naked return ... name old time/op new time/op delta Hash-8 23.0µs ± 1% 22.4µs ± 0% -2.43% (p=0.008 n=5+5) HashMapAcyclic-8 24.0µs ± 0% 23.8µs ± 0% -0.56% (p=0.008 n=5+5) name old alloc/op new alloc/op delta Hash-8 2.92kB ± 0% 2.49kB ± 0% -14.80% (p=0.000 n=5+4) HashMapAcyclic-8 2.53kB ± 0% 2.53kB ± 0% ~ (all equal) name old allocs/op new allocs/op delta Hash-8 93.0 ± 0% 86.0 ± 0% -7.53% (p=0.008 n=5+5) HashMapAcyclic-8 202 ± 0% 202 ± 0% ~ (all equal) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Brad Fitzpatrick	79b7fa9ac3	internal/deephash: hash maps without sorting in the acyclic common case ... Hash and xor each entry instead, then write final xor'ed result. name old time/op new time/op delta Hash-4 33.6µs ± 4% 34.6µs ± 3% +3.03% (p=0.013 n=10+9) name old alloc/op new alloc/op delta Hash-4 1.86kB ± 0% 1.77kB ± 0% -5.10% (p=0.000 n=10+9) name old allocs/op new allocs/op delta Hash-4 51.0 ± 0% 49.0 ± 0% -3.92% (p=0.000 n=10+10) Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Brad Fitzpatrick	a86a0361a7	go.mod: upgrade all deps ... At the start of a dev cycle we'll upgrade all dependencies. Done with: $ for Dep in $(cat go.mod | perl -ne '/(\S+) v/ and print "$1\n"'); do go get $Dep@upgrade; done Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Josh Bleecher Snyder	8bf2a38f29	go.mod: update wireguard-go, taking control over iOS memory usage from our fork ... Our wireguard-go fork used different values from upstream for package device's memory limits on iOS. This was the last blocker to removing our fork. These values are now vars rather than consts for iOS. c27ff9b9f6 Adjust them on startup to our preferred values. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	5666663370	net/packet: use netaddr AppendTo methods ... This lets us remote the types/strbuilder package, which had only a single user. And it's faster. name old time/op new time/op delta String/tcp4-8 175ns ± 0% 58ns ± 1% -66.95% (p=0.000 n=10+9) String/tcp6-8 226ns ± 1% 136ns ± 1% -39.85% (p=0.000 n=10+10) String/udp4-8 175ns ± 1% 58ns ± 1% -67.01% (p=0.000 n=10+9) String/udp6-8 230ns ± 1% 140ns ± 0% -39.32% (p=0.000 n=10+9) String/icmp4-8 164ns ± 0% 50ns ± 1% -69.89% (p=0.000 n=10+10) String/icmp6-8 217ns ± 1% 129ns ± 0% -40.46% (p=0.000 n=10+10) String/igmp-8 196ns ± 0% 56ns ± 1% -71.32% (p=0.000 n=10+10) String/unknown-8 2.06ns ± 1% 2.06ns ± 2% ~ (p=0.985 n=10+10) name old alloc/op new alloc/op delta String/tcp4-8 32.0B ± 0% 32.0B ± 0% ~ (all equal) String/tcp6-8 168B ± 0% 96B ± 0% -42.86% (p=0.000 n=10+10) String/udp4-8 32.0B ± 0% 32.0B ± 0% ~ (all equal) String/udp6-8 168B ± 0% 96B ± 0% -42.86% (p=0.000 n=10+10) String/icmp4-8 32.0B ± 0% 32.0B ± 0% ~ (all equal) String/icmp6-8 104B ± 0% 64B ± 0% -38.46% (p=0.000 n=10+10) String/igmp-8 48.0B ± 0% 48.0B ± 0% ~ (all equal) String/unknown-8 0.00B 0.00B ~ (all equal) name old allocs/op new allocs/op delta String/tcp4-8 1.00 ± 0% 1.00 ± 0% ~ (all equal) String/tcp6-8 3.00 ± 0% 1.00 ± 0% -66.67% (p=0.000 n=10+10) String/udp4-8 1.00 ± 0% 1.00 ± 0% ~ (all equal) String/udp6-8 3.00 ± 0% 1.00 ± 0% -66.67% (p=0.000 n=10+10) String/icmp4-8 1.00 ± 0% 1.00 ± 0% ~ (all equal) String/icmp6-8 3.00 ± 0% 1.00 ± 0% -66.67% (p=0.000 n=10+10) String/igmp-8 1.00 ± 0% 1.00 ± 0% ~ (all equal) String/unknown-8 0.00 0.00 ~ (all equal) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	d6d1951897	net/packet: add BenchmarkString ... Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
David Anderson	df350e2069	ipn/ipnlocal: initialize DNS config maps unconditionally. ... Fixes #1963. Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
Josh Bleecher Snyder	eb9757a290	go.mod: upgrade netaddr to get AppendTo methods ... Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	cd54792fe9	internal/deephash: add a few more benchmarking map entries ... Typical maps in production are considerably longer. This helps benchmarks more accurately reflect the costs per key vs the costs per map in deephash. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
David Crawshaw	293a2b11cd	ipn: allow b to be nil in NewBackendServer ... A couple of code paths in ipnserver use a NewBackendServer with a nil backend just to call the callback with an encapsulated error message. This covers a panic case seen in logs. For #1920 Signed-off-by: David Crawshaw <crawshaw@tailscale.com>	4 years ago
David Anderson	e2dcf63420	net/dns: replace AuthoritativeSuffixes with nil Route entries. ... This leads to a cleaner separation of intent vs. implementation (Routes is now the only place specifying who handles DNS requests), and allows for cleaner expression of a configuration that creates MagicDNS records without serving them to the OS. Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
David Anderson	6690f86ef4	net/dns: always offer MagicDNS records at 100.100.100.100. ... Fixes #1886. Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
simenghe	dd0b690e7b	Added new Addresses / AllowedIPs fields to testcontrol when creating tailcfg.Node (#1948) ... * Added new Addresses / AllowedIPs fields to testcontrol when creating new &tailcfg.Node Signed-off-by: Simeng He <simeng@tailscale.com> * Added single node test to check Addresses and AllowedIPs Signed-off-by: Simeng He <simeng@tailscale.com> Co-authored-by: Simeng He <simeng@tailscale.com>	4 years ago
David Anderson	85df1b0fa7	go.mod: bump wireguard-go. ... Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
Maisem Ali	234cc87f48	cmd/tailscaled: use the wf package instead of wireguard-windows/firewall ... Signed-off-by: Maisem Ali <maisem@tailscale.com>	4 years ago
Josh Bleecher Snyder	25df067dd0	all: adapt to opaque netaddr types ... This commit is a mishmash of automated edits using gofmt: gofmt -r 'netaddr.IPPort{IP: a, Port: b} -> netaddr.IPPortFrom(a, b)' -w . gofmt -r 'netaddr.IPPrefix{IP: a, Port: b} -> netaddr.IPPrefixFrom(a, b)' -w . gofmt -r 'a.IP.Is4 -> a.IP().Is4' -w . gofmt -r 'a.IP.As16 -> a.IP().As16' -w . gofmt -r 'a.IP.Is6 -> a.IP().Is6' -w . gofmt -r 'a.IP.As4 -> a.IP().As4' -w . gofmt -r 'a.IP.String -> a.IP().String' -w . And regexps: \w*(.*)\.Port = (.*) -> $1 = $1.WithPort($2) \w*(.*)\.IP = (.*) -> $1 = $1.WithIP($2) And lots of manual fixups. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
David Anderson	4f92f405ee	scripts: fix up installer script comments. ... Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
David Anderson	0e9ea9f779	scripts: detect curl vs. wget and use the right one. ... Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
David Anderson	783f125003	scripts: use codenames for ubuntu, since that's what our repo uses. ... Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
David Anderson	01a359cec9	scripts: add an install script. ... The script detects one of the supported OS/version combos, and issues the right install instructions for it. Co-authored-by: Christine Dodrill <xe@tailscale.com> Signed-off-by: David Anderson <danderson@tailscale.com>	4 years ago
Brad Fitzpatrick	5b52b64094	tsnet: add Tailscale-as-a-library package ... Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Josh Bleecher Snyder	6f62bbae79	cmd/tailscale: make ping --until-direct require direct connection to exit 0 ... If --until-direct is set, the goal is to make a direct connection. If we failed at that, say so, and exit with an error. RELNOTE=tailscale ping --until-direct (the default) now exits with a non-zero exit code if no direct connection was established. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Avery Pennarun	6fd4e8d244	ipnlocal: fix switching users while logged in + Stopped. ... This code path is very tricky since it was originally designed for the "re-authenticate to refresh my keys" use case, which didn't want to lose the original session even if the refresh cycle failed. This is why it acts differently from the Logout(); Login(); case. Maybe that's too fancy, considering that it probably never quite worked at all, for switching between users without logging out first. But it works now. This was more invasive than I hoped, but the necessary fixes actually removed several other suspicious BUG: lines from state_test.go, so I'm pretty confident this is a significant net improvement. Fixes tailscale/corp#1756. Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>	4 years ago
Avery Pennarun	6307a9285d	controlclient: update Persist.LoginName when it changes. ... Well, that was anticlimactic. Fixes tailscale/corp#461. Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>	4 years ago
Avery Pennarun	285d0e3b4d	ipnlocal: fix deadlock in RequestEngineStatusAndWait() error path. ... If the engine was shutting down from a previous session (e.closing=true), it would return an error code when trying to get status. In that case, ipnlocal would never unblock any callers that were waiting on the status. Not sure if this ever happened in real life, but I accidentally triggered it while writing a test. Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>	4 years ago
Brad Fitzpatrick	5a7c6f1678	tstest/integration{,/testcontrol}: add node update support, two node test ... Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Brad Fitzpatrick	d32667011d	tstest/integration: build test binaries with -race if test itself is ... Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Brad Fitzpatrick	314d15b3fb	version: add func IsRace to report whether race detector enabled ... Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Brad Fitzpatrick	ed9d825552	tstest/integration: fix integration test on linux/386 ... Apparently can't use GOBIN with GOARCH. Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Brad Fitzpatrick	c0158bcd0b	tstest/integration{,/testcontrol}: add testcontrol.RequireAuth mode, new test ... Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago
Josh Bleecher Snyder	ebcd7ab890	wgengine: remove wireguard-go DeviceOptions ... We no longer need them. This also removes the 32 bytes of prefix junk before endpoints. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	aacb2107ae	all: add extra information to serialized endpoints ... magicsock.Conn.ParseEndpoint requires a peer's public key, disco key, and legacy ip/ports in order to do its job. We currently accomplish that by: * adding the public key in our wireguard-go fork * encoding the disco key as magic hostname * using a bespoke comma-separated encoding It's a bit messy. Instead, switch to something simpler: use a json-encoded struct containing exactly the information we need, in the form we use it. Our wireguard-go fork still adds the public key to the address when it passes it to ParseEndpoint, but now the code compensating for that is just a couple of simple, well-commented lines. Once this commit is in, we can remove that part of the fork and remove the compensating code. Signed-off-by: Josh Bleecher Snyder <josharian@gmail.com>	4 years ago
Josh Bleecher Snyder	98cae48e70	wgengine/wglog: optimize wireguardGoString ... The new code is ugly, but much faster and leaner. name old time/op new time/op delta SetPeers-8 7.81µs ± 1% 3.59µs ± 1% -54.04% (p=0.000 n=9+10) name old alloc/op new alloc/op delta SetPeers-8 7.68kB ± 0% 2.53kB ± 0% -67.08% (p=0.000 n=10+10) name old allocs/op new allocs/op delta SetPeers-8 237 ± 0% 99 ± 0% -58.23% (p=0.000 n=10+10) Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Josh Bleecher Snyder	9356912053	wgengine/wglog: add BenchmarkSetPeer ... Because it showed up on hello profiles. Cycle through some moderate-sized sets of peers. This should cover the "small tweaks to netmap" and the "up/down cycle" cases. Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>	4 years ago
Brad Fitzpatrick	36a26e6a71	internal/deephash: rename from deepprint ... Yes, it printed, but that was an implementation detail for hashing. And coming optimization will make it print even less. Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>	4 years ago