Also, bit of behavior change: on non-nil err but expired context,
don't reset the consecutive failure count. I don't think the old
behavior was intentional.
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
NetworkMap text diffs being empty were currently used to short-circuit
calling magicsock's SetNetworkMap (via Engine.SetNetworkMap), but that
went away in c7582dc2 (0.100.0-230)
Prior to c7582dc2 (notably, in 0.100.0-225 and below, down to
0.100.0), a change in only disco key (as when a node restarts) but
without endpoint changes (as would happen for a client not behind a
NAT with random ports) could result in a "netmap diff: (none)" being
printed, as well as Engine.SetNetworkMap being skipped, leading to
broken discovery endpoints.
c7582dc2 fixed the Engine.SetNetworkMap skippage.
This change fixes the "netmap diff: (none)" print so we'll actually see when a peer
restarts with identical endpoints but a new discovery key.
This code is currently racy due to an incorrect assumption
that goal is never modified in-place, so does not require extra locking.
This change makes the assumption correct.
Signed-off-by: Dmytro Shynkevych <dmytro@tailscale.com>
This adds a new magicsock endpoint type only used when both sides
support discovery (that is, are advertising a discovery
key). Otherwise the old code is used.
So far the new code only communicates over DERP as proof that the new
code paths are wired up. None of the actually discovery messaging is
implemented yet.
Support for discovery (generating and advertising a key) are still
behind an environment variable for now.
Updates #483
As part of disabling background STUN packets when idle, we want an
emergency override switch to turn it back on, in case it interacts
poorly in the wild. We'll send that via control, but we'll want to
plumb it down to magicsock via NetworkMap.
Updates tailscale/corp#320
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
It can only be built with corp deps anyway, and having it split
from the control code makes our lives harder.
Signed-off-by: David Anderson <danderson@tailscale.com>
Instead of hard-coding the DERP map (except for cmd/tailscale netcheck
for now), get it from the control server at runtime.
And make the DERP map support multiple nodes per region with clients
picking the first one that's available. (The server will balance the
order presented to clients for load balancing)
This deletes the stunner package, merging it into the netcheck package
instead, to minimize all the config hooks that would've been
required.
Also fix some test flakes & races.
Fixes#387 (Don't hard-code the DERP map)
Updates #388 (Add DERP region support)
Fixes#399 (wgengine: flaky tests)
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
By default, nothing differentiates errors or fatals from regular logs, so they just
blend into the rest of the logs.
As a bonus, if you run a test using t.Run(), the log messages printed
via the sub-t.Run() are printed at a different time from log messages
printed via the parent t.Run(), making debugging almost impossible.
This doesn't actually fix the test flake I'm looking for, but at least
I can find it in the logs now.
Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>
If a test calls log.Printf, 'go test' horrifyingly rearranges the
output to no longer be in chronological order, which makes debugging
virtually impossible. Let's stop that from happening by making
log.Printf panic if called from any module, no matter how deep, during
tests.
This required us to change the default error handler in at least one
http.Server, as well as plumbing a bunch of logf functions around,
especially in magicsock and wgengine, but also in logtail and backoff.
To add insult to injury, 'go test' also rearranges the output when a
parent test has multiple sub-tests (all the sub-test's t.Logf is always
printed after all the parent tests t.Logf), so we need to screw around
with a special Logf that can point at the "current" t (current_t.Logf)
in some places. Probably our entire way of using subtests is wrong,
since 'go test' would probably like to run them all in parallel if you
called t.Parallel(), but it definitely can't because the're all
manipulating the shared state created by the parent test. They should
probably all be separate toplevel tests instead, with common
setup/teardown logic. But that's a job for another time.
Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>
This depends on improved support from the control server, to send the
new subnet width (Bits) fields. If these are missing, we fall back to
assuming their value is /32.
Conversely, if the server sends Bits fields to an older client, it will
interpret them as /32 addresses. Since the only rules we allow are
"accept" rules, this will be narrower or equal to the intended rule, so
older clients will simply reject hosts on the wider subnet (fail
closed).
With this change, the internal filter.Matches format has diverged
from the wire format used by controlclient, so move the wire format
into tailcfg and convert it to filter.Matches in controlclient.
Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>
This strictly sequences things such that c1 is fully registered in
the control server before c2 creates its poll. Failure to do this
can cause an inversion where c2's poll finishes establishing
before c1's poll starts, which results in c2 getting disconnected
rather than c1, and the test times out waiting for c1 to get kicked.
Fixes#98.
Signed-off-by: David Anderson <danderson@tailscale.com>
The test is straightforward, but it's a little perplexing if you're
not overly familiar with controlclient.
Signed-off-by: David Anderson <danderson@tailscale.com>
I removed the HTTPC field in b6fa5a69be but it was apparently still
used in [oss-skipped] tests.
Restore it, but name it so it's more obvious that it's only for
tests. (It currently is, and I'd like to keep it like that for now.)
The tests cheat at filling out web forms by directly POSTing to
the target. The target for authURLs has changed slightly, the base
authURL now redirects the user to the login page.
Additionally, the authURL cycle now checks the cookie is set
correctly, so we add cookie jars where necessary to pass the
cookie through.
Add opt-in method to request IPv6 endpoints from the control plane.
For now they should just be skipped. A previous version of this CL was
unconditional and reportedly had problems that I can't reproduce. So
make it a knob until the mystery is solved.
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Breaks something deep in wireguard or magicsock's brainstem, no packets at all
can flow. All received packets fail decryption with "invalid mac1".
This reverts commit 94024355ed.
Signed-off-by: David Anderson <dave@natulte.net>
This removes the need for go-cmp, which is extremely bloaty so we had
to leave it out of iOS. As a result, we had also left it out of macOS,
and so we didn't print netmap diffs at all on darwin-based platforms.
Oops.
As a bonus, the output format of the new function is way better.
Minor oddity: because I used the dumbest possible diff algorithm, the
sort order is a bit dumb. We print all "removed" lines and then print
all "added" lines, rather than doing the usual diff-like thing of
interspersing them. This probably doesn't matter (maybe it's an
improvement).
The .Concise() view had grown hard to read over time. Originally, we
assumed a peer almost always had just one endpoint and one-or-more
allowedips. With magicsock, we now almost always have multiple
endpoints per peer. And empirically, almost every peer has only one
allowedip.
Change their order so we can line up allowedips vertically. Also do
some tweaking to make multiple endpoints easier to read.
While we're here, add a column to show the home DERP server of each
peer, if any.
We log it once upon receiving the first copy of the map, then
subsequently when a new one appears, but only if we haven't logged one
less than 5 minutes ago.
This avoids overly cluttering the log (as we did before, logging the
netmap every time one appeared, which could be hundreds of lines every
few seconds), but still gives the log enough context to help in
diagnosing problems retroactively.
We still include them directly in the controlclient network map
just where we have been. Client plumbing we can do later.
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
This avoids a non-obvious data race, where the JSON decoder ends
up creating do-nothing writes into global variables.
==================
WARNING: DATA RACE
Write at 0x0000011e1860 by goroutine 201:
tailscale.com/wgengine/packet.(*IP).UnmarshalJSON()
/home/crawshaw/repo/corp/oss/wgengine/packet/packet.go:83 +0x2d9
encoding/json.(*decodeState).literalStore()
/home/crawshaw/go/go/src/encoding/json/decode.go:877 +0x445e
...
encoding/json.Unmarshal()
/home/crawshaw/go/go/src/encoding/json/decode.go:107 +0x1de
tailscale.com/control/controlclient.(*Direct).decodeMsg()
/home/crawshaw/repo/corp/oss/control/controlclient/direct.go:615 +0x1ab
tailscale.com/control/controlclient.(*Direct).PollNetMap()
/home/crawshaw/repo/corp/oss/control/controlclient/direct.go:525 +0x1053
tailscale.com/control/controlclient.(*Client).mapRoutine()
/home/crawshaw/repo/corp/oss/control/controlclient/auto.go:428 +0x3a6
Previous read at 0x0000011e1860 by goroutine 86:
tailscale.com/wgengine/filter.matchIPWithoutPorts()
/home/crawshaw/repo/corp/oss/wgengine/filter/match.go:108 +0x91
tailscale.com/wgengine/filter.(*Filter).runIn()
/home/crawshaw/repo/corp/oss/wgengine/filter/filter.go:147 +0x3c6
tailscale.com/wgengine/filter.(*Filter).RunIn()
/home/crawshaw/repo/corp/oss/wgengine/filter/filter.go:127 +0xb0
tailscale.com/wgengine.(*userspaceEngine).SetFilter.func1()
/home/crawshaw/repo/corp/oss/wgengine/userspace.go:390 +0xfc
github.com/tailscale/wireguard-go/device.(*Device).RoutineDecryption()
/home/crawshaw/repo/corp/wireguard-go/device/receive.go:295 +0xa1f
For #112
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
relaynode itself is not long for this world, deprecated in favour of
tailscale/tailscaled. But now that the control server supports central
distribution of packet filters, let's actually take advantage of it in
a final, backward compatible release of relaynode.
