Pull in the latest version of wireguard-windows.
Switch to upstream wireguard-go.
This requires reverting all of our import paths.
Unfortunately, this has to happen at the same time.
The wireguard-go change is very low risk,
as that commit matches our fork almost exactly.
(The only changes are import paths, CI files, and a go.mod entry.)
So if there are issues as a result of this commit,
the first place to look is wireguard-windows changes.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
We repeat many peers each time we call SetPeers.
Instead of constructing strings for them from scratch every time,
keep strings alive across iterations.
name old time/op new time/op delta
SetPeers-8 3.58µs ± 1% 2.41µs ± 1% -32.60% (p=0.000 n=9+10)
name old alloc/op new alloc/op delta
SetPeers-8 2.53kB ± 0% 1.30kB ± 0% -48.73% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
SetPeers-8 99.0 ± 0% 16.0 ± 0% -83.84% (p=0.000 n=10+10)
We could reduce alloc/op 12% and allocs/op 23% if strs had
type map[string]strCache instead of map[string]*strCache,
but that wipes out the execution time impact.
Given that re-use is the most common scenario, let's optimize for it.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
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>
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>
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>
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>
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>
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>
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
https://github.com/tailscale/go/commit/728ecc58fd5c010c04bb7c8a56d238167e7d888f.
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
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>
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>
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>
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>
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>
Not that it matters, but we were missing a close parens.
It's cheap, so add it.
name old time/op new time/op delta
Hash-8 6.64µs ± 0% 6.67µs ± 1% +0.42% (p=0.008 n=9+10)
name old alloc/op new alloc/op delta
Hash-8 1.54kB ± 0% 1.54kB ± 0% ~ (all equal)
name old allocs/op new allocs/op delta
Hash-8 37.0 ± 0% 37.0 ± 0% ~ (all equal)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The struct field names don't change within a single run,
so they are irrelevant. Use the field index instead.
name old time/op new time/op delta
Hash-8 6.52µs ± 0% 6.64µs ± 0% +1.91% (p=0.000 n=6+9)
name old alloc/op new alloc/op delta
Hash-8 1.67kB ± 0% 1.54kB ± 0% -7.66% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Hash-8 53.0 ± 0% 37.0 ± 0% -30.19% (p=0.000 n=10+10)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
These show up a lot in our data structures.
name old time/op new time/op delta
Hash-8 11.5µs ± 1% 7.8µs ± 1% -32.17% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
Hash-8 1.98kB ± 0% 1.67kB ± 0% -15.73% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Hash-8 82.0 ± 0% 53.0 ± 0% -35.37% (p=0.000 n=10+10)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The sha256 hash writer doesn't implement WriteString.
(See https://github.com/golang/go/issues/38776.)
As a consequence, we end up converting many strings to []byte.
Wrapping a bufio.Writer around the hash writer lets us
avoid these conversions by using WriteString.
Using a bufio.Writer is, perhaps surprisingly, almost as cheap as using unsafe.
The reason is that the sha256 writer does internal buffering,
but doesn't do any when handed larger writers.
Using a bufio.Writer merely shifts the data copying from one buffer
to a different one.
Using a concrete type for Print and print cuts 10% off of the execution time.
name old time/op new time/op delta
Hash-8 15.3µs ± 0% 11.5µs ± 0% -24.84% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
Hash-8 2.82kB ± 0% 1.98kB ± 0% -29.57% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Hash-8 140 ± 0% 82 ± 0% -41.43% (p=0.000 n=10+10)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
deepprint currently accounts for 15% of allocs in tailscaled.
This is a useful benchmark to have.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
On benchmark completion, we shut down the wgengine.
If we happen to poll for status during shutdown,
we get an "engine closing" error.
It doesn't hurt anything; ignore it.
Fixestailscale/corp#1776
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Without any synchronization here, the "first packet" callback can
be delayed indefinitely, while other work continues.
Since the callback starts the benchmark timer, this could skew results.
Worse, if the benchmark manages to complete before the benchmark timer begins,
it'll cause a data race with the benchmark shutdown performed by package testing.
That is what is reported in #1881.
This is a bit unfortunate, in that it means that users of TrafficGen have
to be careful to keep this callback speedy and lightweight and to avoid deadlocks.
Fixes#1881
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It is possible to get multiple status callbacks from an Engine.
We need to wait for at least one from each Engine.
Without limiting to one per Engine,
wait.Wait can exit early or can panic due to a negative counter.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This reduces the speed with which these benchmarks exhaust their supply fds.
Not to zero unfortunately, but it's still helpful when doing long runs.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This function accounted for ~1% of all allocs by tailscaled.
It is trivial to improve, so may as well.
name old time/op new time/op delta
KeyMarshalText-8 197ns ± 0% 47ns ± 0% -76.12% (p=0.016 n=4+5)
name old alloc/op new alloc/op delta
KeyMarshalText-8 200B ± 0% 80B ± 0% -60.00% (p=0.008 n=5+5)
name old allocs/op new allocs/op delta
KeyMarshalText-8 5.00 ± 0% 1.00 ± 0% -80.00% (p=0.008 n=5+5)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It is unused, and has been since early Feb 2021 (Tailscale 1.6).
We can't get delete the DeviceOptions entirely yet;
first #1831 and #1839 need to go in, along with some wireguard-go changes.
Deleting this chunk of code now will make the later commits more clearly correct.
Pingers can now go too.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It existed to work around the frequent opening and closing
of the conn.Bind done by wireguard-go.
The preceding commit removed that behavior,
so we can simply close the connections
when we are done with them.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
We don't use the port that wireguard-go passes to us (via magicsock.connBind.Open).
We ignore it entirely and use the port we selected.
When we tell wireguard-go that we're changing the listen_port,
it calls connBind.Close and then connBind.Open.
And in the meantime, it stops calling the receive functions,
which means that we stop receiving and processing UDP and DERP packets.
And that is Very Bad.
That was never a problem prior to b3ceca1dd7,
because we passed the SkipBindUpdate flag to our wireguard-go fork,
which told wireguard-go not to re-bind on listen_port changes.
That commit eliminated the SkipBindUpdate flag.
We could write a bunch of code to work around the gap.
We could add background readers that process UDP and DERP packets when wireguard-go isn't.
But it's simpler to never create the conditions in which wireguard-go rebinds.
The other scenario in which wireguard-go re-binds is device.Down.
Conveniently, we never call device.Down. We go from device.Up to device.Close,
and the latter only when we're shutting down a magicsock.Conn completely.
Rubber-ducked-by: Avery Pennarun <apenwarr@tailscale.com>
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Upstream wireguard-go has changed its receive model.
NewDevice now accepts a conn.Bind interface.
The conn.Bind is stateless; magicsock.Conns are stateful.
To work around this, we add a connBind type that supports
cheap teardown and bring-up, backed by a Conn.
The new conn.Bind allows us to specify a set of receive functions,
rather than having to shoehorn everything into ReceiveIPv4 and ReceiveIPv6.
This lets us plumbing DERP messages directly into wireguard-go,
instead of having to mux them via ReceiveIPv4.
One consequence of the new conn.Bind layer is that
closing the wireguard-go device is now indistinguishable
from the routine bring-up and tear-down normally experienced
by a conn.Bind. We thus have to explicitly close the magicsock.Conn
when the close the wireguard-go device.
One downside of this change is that we are reliant on wireguard-go
to call receiveDERP to process DERP messages. This is fine for now,
but is perhaps something we should fix in the future.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The code is not obviously better or worse, but this makes the little warning
triangle in my editor go away, and the distraction removal is worth it.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The call to appendEndpoint updates cpeer.Endpoints.
Then it is overwritten in the next line.
The only errors from appendEndpoint occur when
the host/port pair is malformed, but that cannot happen.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It can end up executing an a new goroutine,
at which point instead of immediately stopping test execution, it hangs.
Since this is unexpected anyway, panic instead.
As a bonus, it makes call sites nicer and removes a kludge comment.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The Windows CI machine experiences significant random execution delays.
For example, in this code from watchdog.go:
done := make(chan bool)
go func() {
start := time.Now()
mu.Lock()
There was a 500ms delay from initializing done to locking mu.
This test checks that we receive a sufficient number of events quickly enough.
In the face of random 500ms delays, unsurprisingly, the test fails.
There's not much principled we can do about it.
We could build a system of retries or attempt to detect these random delays,
but that game isn't worth the candle.
Skip the test.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This works around the close syscall being slow.
We can revert this if we find a fix or if Apple makes close fast again.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
When a handshake race occurs, a queued data packet can get lost.
TestTwoDevicePing expected that the very first data packet would arrive.
This caused occasional flakes.
Change TestTwoDevicePing to repeatedly re-send packets
and succeed when one of them makes it through.
This is acceptable (vs making WireGuard not drop the packets)
because this only affects communication with extremely old clients.
And those extremely old clients will eventually connect,
because the kernel will retry sends on timeout.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
We modified the standard net package to not allocate a *net.UDPAddr
during a call to (*net.UDPConn).ReadFromUDP if the caller's use
of the *net.UDPAddr does not cause it to escape.
That is https://golang.org/cl/291390.
This is the companion change to magicsock.
There are two changes required.
First, call ReadFromUDP instead of ReadFrom, if possible.
ReadFrom returns a net.Addr, which is an interface, which always allocates.
Second, reduce the lifetime of the returned *net.UDPAddr.
We do this by immediately converting it into a netaddr.IPPort.
We left the existing RebindingUDPConn.ReadFrom method in place,
as it is required to satisfy the net.PacketConn interface.
With the upstream change and both of these fixes in place,
we have removed one large allocation per packet received.
name old time/op new time/op delta
ReceiveFrom-8 16.7µs ± 5% 16.4µs ± 8% ~ (p=0.310 n=5+5)
name old alloc/op new alloc/op delta
ReceiveFrom-8 112B ± 0% 64B ± 0% -42.86% (p=0.008 n=5+5)
name old allocs/op new allocs/op delta
ReceiveFrom-8 3.00 ± 0% 2.00 ± 0% -33.33% (p=0.008 n=5+5)
Co-authored-by: Sonia Appasamy <sonia@tailscale.com>
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
addrSet maintained duplicate lists of netaddr.IPPorts and net.UDPAddrs.
Unify to use the netaddr type only.
This makes (*Conn).ReceiveIPvN a bit uglier,
but that'll be cleaned up in a subsequent commit.
This is preparatory work to remove an allocation from ReceiveIPv4.
Co-authored-by: Sonia Appasamy <sonia@tailscale.com>
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
I based my estimation of the required timeout based on locally
observed behavior. But CI machines are worse than my local machine.
16s was enough to reduce flakiness but not eliminate it. Bump it up again.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This was in place because retrieved allowed_ips was very expensive.
Upstream changed the data structure to make them cheaper to compute.
This commit is an experiment to find out whether they're now cheap enough.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
We removed the "fast retry" code from our wireguard-go fork.
As a result, pings can take longer to transit when retries are required.
Allow that.
Fixes#1277
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Upstream wireguard-go decided to use errors.Is(err, net.ErrClosed)
instead of checking the error string.
It also provided an unsafe linknamed version of net.ErrClosed
for clients running Go 1.15. Switch to that.
This reduces the time required for the wgengine/magicsock tests
on my machine from ~35s back to the ~13s it was before
456cf8a3765948d6f1992162993eaf3844371592.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Use tb.Cleanup to simplify both the API and the implementation.
One behavior change: When the number of goroutines shrinks, don't log.
I've never found these logs to be useful, and they frequently add noise.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This is mostly code movement from the wireguard-go repo.
Most of the new wgcfg package corresponds to the wireguard-go wgcfg package.
wgengine/wgcfg/device{_test}.go was device/config{_test}.go.
There were substantive but simple changes to device_test.go to remove
internal package device references.
The API of device.Config (now wgcfg.DeviceConfig) grew an error return;
we previously logged the error and threw it away.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
We log lines like this:
c.logf("[v1] magicsock: disco: %v->%v (%v, %v) sent %v", c.discoShort, dstDisco.ShortString(), dstKey.ShortString(), derpStr(dst.String()), disco.MessageSummary(m))
The leading [v1] causes it to get unintentionally rate limited.
Until we have a proper fix, work around it.
Fixes#1216
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Rewrite log lines on the fly, based on the set of known peers.
This enables us to use upstream wireguard-go logging,
but maintain the Tailscale-style peer public key identifiers
that the rest of our systems (and people) expect.
Fixes#1183
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
DstToString is used in two places in wireguard-go: Logging and uapi.
We are switching to use uapi for wireguard-go config.
To preserve existing behavior, we need the full set of addrs.
And for logging, having the full set of addrs seems useful.
(The Addrs method itself is slated for removal. When that happens,
the implementation will move to DstToString.)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This reverts commit 08baa17d9a.
It caused deadlocks due to lock ordering violations.
It was not the right fix, and thus should simply be reverted
while we look for the right fix (if we haven't already found it
in the interim; we've fixed other logging-after-test issues).
Fixes#1161
Windows has a low resolution timer.
Some of the tests assumed that unblock takes effect immediately.
Consider:
t := time.Now()
elapsed := time.Now().After(t)
It seems plausible that elapsed should always be true.
However, with a low resolution timer, that might fail.
Change time.Now().After to !time.Now().Before,
so that unblocking always takes effect immediately.
Fixes#873.
22507adf54 stopped relying on
our fork of wireguard-go's UpdateDst callback.
As a result, we can unwind that code,
and the extra return value of ReceiveIPv{4,6}.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This prevents us from continuing to do unnecessary work
(including logging) after the connection has closed.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This eliminates a dependency on wgcfg.Endpoint,
as part of the effort to eliminate our wireguard-go fork.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
If any goroutine continues to use the logger in TestLocalLogLines
after the test finishes, the test panics.
The culprit for this was wireguard-go; the previous commit fixed that.
This commit adds suspenders: When the test is done, make logging calls
into no-ops.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
