We removed it in #4806 in favor of the built-in functionality from the
nhooyr.io/websocket package. However, it has an issue with deadlines
that has not been fixed yet (see nhooyr/websocket#350). Temporarily
go back to using a custom wrapper (using the fix from our fork) so that
derpers will stop closing connections too aggressively.
Updates #5921
Signed-off-by: Mihai Parparita <mihai@tailscale.com>
Periodically poll the TCP RTT metric from all open TCP connections and
update a (bucketed) histogram metric.
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: I6214902196b05bf7829c9d0ea501ce0e13d984cf
NewNetcheckClient only initializes a subset of fields of derphttp.Client,
and the Close() call added by #5707 was result in a nil pointer dereference.
Make Close() safe to call when using NewNetcheckClient() too.
Fixes#5919
Signed-off-by: Mihai Parparita <mihai@tailscale.com>
The io/ioutil package has been deprecated as of Go 1.16 [1]. This commit
replaces the existing io/ioutil functions with their new definitions in
io and os packages.
Reference: https://golang.org/doc/go1.16#ioutil
Signed-off-by: Eng Zer Jun <engzerjun@gmail.com>
We can't do Noise-over-HTTP in Wasm/JS (because we don't have bidirectional
communication), but we should be able to do it over WebSockets. Reuses
derp WebSocket support that allows us to turn a WebSocket connection
into a net.Conn.
Updates #3157
Signed-off-by: Mihai Parparita <mihai@tailscale.com>
In cases where tailscale is operating behind a MITM proxy, we need to consider
that a lot more of the internals of our HTTP requests are visible and may be
used as part of authorization checks. As such, we need to 'behave' as closely
as possible to ideal.
- Some proxies do authorization or consistency checks based the on Host header
or HTTP URI, instead of just the IP/hostname/SNI. As such, we need to
construct a `*http.Request` with a valid URI everytime HTTP is going to be
used on the wire, even if its over TLS.
Aside from the singular instance in net/netcheck, I couldn't find anywhere
else a http.Request was constructed incorrectly.
- Some proxies may deny requests, typically by returning a 403 status code. We
should not consider these requests as a valid latency check, so netcheck
semantics have been updated to consider >299 status codes as a failed probe.
Signed-off-by: Tom DNetto <tom@tailscale.com>
This allows tests to verfiy that a DERP connection was actually
established.
Related to #4326
Updates tailscale/corp#2579
Signed-off-by: Maisem Ali <maisem@tailscale.com>
A new package can also later record/report which knobs are checked and
set. It also makes the code cleaner & easier to grep for env knobs.
Change-Id: Id8a123ab7539f1fadbd27e0cbeac79c2e4f09751
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Only if the source address isn't on the currently active interface or
a ping of the DERP server fails.
Updates #3619
Change-Id: I6bf06503cff4d781f518b437c8744ac29577acc8
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
So magicsock can later ask a DERP connection whether its source IP
would've changed if it reconnected.
Updates #3619
Change-Id: Ibc8810340c511d6786b60c78c1a61c09f5800e40
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Continuing work in 434af15a04, to make it possible for magicsock to
probe whether a DERP server is still there.
Updates #3619
Change-Id: I366a77c27e93b876734e64f445b85ef01eb590f2
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
In prep for a future change to have client ping derp connections
when their state is questionable, rather than aggressively tearing
them down and doing a heavy reconnect when their state is unknown.
We already support ping/pong in the other direction (servers probing
clients) so we already had the two frame types, but I'd never finished
this direction.
Updates #3619
Change-Id: I024b815d9db1bc57c20f82f80f95fb55fc9e2fcc
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Otherwise random browser requests to /derp cause log spam.
Change-Id: I7bdf991d2106f0323868e651156c788a877a90d5
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
We still try the host's x509 roots first, but if that fails (like if
the host is old), we fall back to using LetsEncrypt's root and
retrying with that.
tlsdial was used in the three main places: logs, control, DERP. But it
was missing in dnsfallback. So added it there too, so we can run fine
now on a machine with no DNS config and no root CAs configured.
Also, move SSLKEYLOGFILE support out of DERP. tlsdial is the logical place
for that support.
Fixes#1609
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
On about 1 out of 500 runs, TestSendFreeze failed:
derp_test.go:416: bob: unexpected message type derp.PeerGoneMessage
Closing alice before bob created a race.
If bob closed promptly, the test passed.
If bob closed slowly, and alice's disappearance caused
bob to receive a PeerGoneMessage before closing, the test failed.
Deflake the test by closing bob first.
With this fix, the test passed 12,000 times locally.
Fixes#2668
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
A public key should only have max one connection to a given
DERP node (or really: one connection to a node in a region).
But if people clone their machine keys (e.g. clone their VM, Raspbery
Pi SD card, etc), then we can get into a situation where a public key
is connected multiple times.
Originally, the DERP server handled this by just kicking out a prior
connections whenever a new one came. But this led to reconnect fights
where 2+ nodes were in hard loops trying to reconnect and kicking out
their peer.
Then a909d37a59 tried to add rate
limiting to how often that dup-kicking can happen, but empirically it
just doesn't work and ~leaks a bunch of goroutines and TCP
connections, tying them up for hour+ while more and more accumulate
and waste memory. Mostly because we were doing a time.Sleep forever
while not reading from their TCP connections.
Instead, just accept multiple connections per public key but track
which is the most recent. And if two both are writing back & forth,
then optionally disable them both. That last part is only enabled in
tests for now. The current default policy is just last-sender-wins
while we gather the next round of stats.
Updates #2751
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
If a peer is connected to multiple nodes in a region (so
multiForwarder is in use) and then a node restarts and re-sends all
its additions, this bug about whether an element is in the
multiForwarder could cause a one-time flip in the which peer node we
forward to. Note a huge deal, but not written as intended.
Thanks to @lewgun for the bug report in #2141.
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
