Commit Graph

6659 Commits

Author SHA1 Message Date
David Anderson
1fdadf06f1 testy: don't give Clock.Step==0 magical behavior.
Turns out it's sometimes useful to stop time entirely.

Signed-off-by: David Anderson <dave@natulte.net>
2020-03-27 00:38:39 -07:00
Brad Fitzpatrick
8b3f6be008 control/controlclient: flip IPv6 to be on by default
The DEBUG_INCLUDE_IPV6 environment variable is now an opt-out.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-26 22:02:56 -07:00
Brad Fitzpatrick
322499473e cmd/tailscaled, wgengine, ipn: add /debug/ipn handler with world state
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-26 14:26:24 -07:00
David Anderson
dbca186a64 stunner: fix data race.
In very low-latency conditions, a STUN request can complete before
the startup loop has finished firing off goroutines, leading to
a concurrent map mutation.
2020-03-25 23:11:22 -07:00
Brad Fitzpatrick
7fb8d873ac logpolicy: don't log timestamp when running under systemd
It'll do it anyway.

Makes journalctl -fu cleaner.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-25 21:03:06 -07:00
Brad Fitzpatrick
2d48f92a82 wgengine/magicsock: re-stun every [20,27] sec, not 28
28 is cutting it close, and we think jitter will help some spikes
we're seeing.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-25 14:25:33 -07:00
Brad Fitzpatrick
2c7ddd0828 go.mod, go.sum: tidy 2020-03-25 14:05:33 -07:00
Brad Fitzpatrick
577f321c38 wgengine/magicsock: revise derp fallback logic
Revision to earlier 6284454ae5

Don't be sticky if we have no peers.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-25 13:09:18 -07:00
David Anderson
cbd8aceb95 go.mod: bump wireguard-go version.
Fixes #219.

Signed-off-by: David Anderson <dave@natulte.net>
2020-03-25 12:20:46 -07:00
Brad Fitzpatrick
2e420ad8b6 wgengine, wgengine/filter: minor doc, style, performance, locking changes
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-25 08:41:01 -07:00
Brad Fitzpatrick
6284454ae5 wgengine/magicsock: if UDP blocked, pick DERP where most peers are
Updates #207

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-25 08:00:44 -07:00
Avery Pennarun
f53e78e0d5 wgengine: don't lose filter state on filter reconfig.
We were abandoning the UDP port LRU every time we got a new packet
filter from tailcontrol, which caused return packets to suddenly stop
arriving.
2020-03-25 04:52:41 -04:00
Avery Pennarun
4336de0d98 ipn/local: don't print packet filter every single time.
It's extremely noisy right now for domains with complex ACLs.
2020-03-25 04:52:41 -04:00
Avery Pennarun
a56e853b72 cmd/relaynode: delete it!
tailscale+tailscaled now completely obsoletes relaynode, so let's let
it finally go away.
2020-03-25 04:48:15 -04:00
David Anderson
73fccd8b06 control/controlclient: remove old compat PacketFilter initialization.
Signed-off-by: David Anderson <dave@natulte.net>
2020-03-24 16:00:41 -07:00
Brad Fitzpatrick
d321190578 wgengine/magicsock: stringify [IPv6]:port normally in AddrSet.String 2020-03-24 13:40:43 -07:00
Brad Fitzpatrick
3c3ea8bc8a wgengine/magicsock: finish IPv6 transport support
DEBUG_INCLUDE_IPV6=1 is still required, but works now.

Updates #18 (fixes it, once env var gate is removed)
2020-03-24 10:56:22 -07:00
Brad Fitzpatrick
82ed7e527e wgengine/magicsock: remove log allocation
This was the whole point but I goofed at the last line.
2020-03-24 08:14:47 -07:00
Brad Fitzpatrick
8454bbbda5 wgengine/magicsock: more logging improvements
* remove endpoint discovery noise when results unchanged
* consistently spell derp nodes as "derp-N"
* replace "127.3.3.40:" with "derp-" in CreateEndpoint log output
* stop early DERP setup before SetPrivateKey is called;
  it just generates log nosie
* fix stringification of peer ShortStrings (it had an old %x on it,
  rendering it garbage)
* describe why derp routes are changing, with one of:
  shared home, their home, our home, alt
2020-03-24 08:12:55 -07:00
Brad Fitzpatrick
64aff0db7b control/controlclient: add TODO to delete the UAPI method 2020-03-24 07:25:14 -07:00
Brad Fitzpatrick
7740cbd8d9 ipn: call SetNetInfoCallback later, in Start
It was being called back into ultimately from magicsock before there
was a control client.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-23 22:19:40 -07:00
Brad Fitzpatrick
680311b3df wgengine/magicsock: fix few remaining logs without package prefix 2020-03-23 22:11:49 -07:00
Brad Fitzpatrick
c473927558 wgengine/magicsock: clean up, add, improve DERP logs 2020-03-23 21:57:58 -07:00
Brad Fitzpatrick
e749377a56 derp/derphttp: remove a redundant log message 2020-03-23 21:23:36 -07:00
Brad Fitzpatrick
1d3f45f9be derp/derphttp: ignore PeerGoneMessage in test 2020-03-23 14:13:49 -07:00
Brad Fitzpatrick
ea9310403d wgengine/magicsock: re-STUN on DERP connection death
Fixes #201
2020-03-23 13:19:33 -07:00
Brad Fitzpatrick
1ab5b31c4b derp, magicsock: send new "peer gone" frames when previous sender disconnects
Updates #150 (not yet enabled by default in magicsock)

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-22 21:00:47 -07:00
David Crawshaw
e60b433831 tailcfg: NetInfo.String: handle nil object without panicing
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
2020-03-23 13:34:33 +11:00
Brad Fitzpatrick
b6f77cc48d wgengine/magicsock: return early, outdent in derpWriteChanOfAddr 2020-03-22 14:08:59 -07:00
Brad Fitzpatrick
8c4cef60f8 derp: wait for send goroutine to finish before returning from run
I saw a test flake due to the sender goroutine logging (ultimately to
t.Logf) after the server was closed.

This makes sure the all goroutines are cleaned up before Server.Close
returns.
2020-03-22 13:10:07 -07:00
Brad Fitzpatrick
521ad7b0fc derp: only flush writes to clients when we're out of things to write
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-22 08:01:25 -07:00
Brad Fitzpatrick
0a25abcdaa derp: remove TODO comment
It was done in bf704a5218
2020-03-22 07:54:50 -07:00
Brad Fitzpatrick
8519a1b29f derp: revert rename of pkt to sendMsg
New plan, sorry.

Most reverts 77921a31b1, except some doc changes.
2020-03-21 22:17:22 -07:00
Brad Fitzpatrick
c34b350efa derp: remove the mutex around and closing of send channel
Makes it less complicated.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-21 22:00:04 -07:00
Brad Fitzpatrick
77921a31b1 derp: rename pkt to sendMsg, in prep for other types of writes
Updates #150
2020-03-21 20:34:49 -07:00
Brad Fitzpatrick
1453aecb44 derp: add sclient.done channel, simplify some context passing
This is mostly prep for a few future CLs, making sure we always have a
close-on-dead done channel available to select on when doing other
channel operations.
2020-03-21 18:28:34 -07:00
David Anderson
ea90780066 derp: specify type of the by-reason drop varz.
Signed-off-by: David Anderson <dave@natulte.net>
2020-03-20 15:53:04 -07:00
Brad Fitzpatrick
fd824df1fa derp: export metric for server's initial MemStats.Sys reading
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
2020-03-20 15:37:36 -07:00
David Anderson
64b5248929 derp: sprinkle comments and docstrings.
Signed-off-by: David Anderson <dave@natulte.net>
2020-03-20 15:14:01 -07:00
David Anderson
58b8ebd290 derp: add per-reason packet drop counter.
In parallel with the summed counter for now, for dashboard migration.

Signed-off-by: David Anderson <dave@natulte.net>
2020-03-20 15:14:01 -07:00
David Anderson
1e031001db derp: move write timeout to package constant.
Signed-off-by: David Anderson <dave@natulte.net>
2020-03-20 15:14:01 -07:00
David Anderson
dbfc916273 derp: add short queues between reader and writer, drop on overload.
This avoids the server blocking on misbehaving or heavily contended
clients. We attempt to drop from the head of the queue to keep
overall queueing time lower.

Also:
 - fixes server->client keepalives, which weren't happening.
 - removes read rate-limiter, deferring instead to kernel-level
   global limiter/fair queuer.

Signed-off-by: David Anderson <dave@natulte.net>
2020-03-20 15:14:01 -07:00
Brad Fitzpatrick
dd31285ad4 wgengine/magicsock: send IPv6 using pconn6, if available
In prep for IPv6 support. Nothing should make it this far yet.
2020-03-20 14:30:12 -07:00
Brad Fitzpatrick
af277a6762 controlclient, magicsock: add debug knob to request IPv6 endpoints
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>
2020-03-20 14:27:24 -07:00
Brad Fitzpatrick
221e7d7767 wgengine/magicsock: make log message include DERP port (node) 2020-03-20 13:51:20 -07:00
Brad Fitzpatrick
33bdcabf03 wgengine/magicsock: call stun callback w/ only valid part of STUN packet 2020-03-20 13:44:27 -07:00
David Crawshaw
13eed1e88b controlclient: remove common log on very quiet nodes
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
2020-03-21 07:11:10 +11:00
Brad Fitzpatrick
f9cc4e8f53 tailcfg: restore MapRequest.IncludeIPv6 bool
Partially restores 0be475ba46
2020-03-20 08:13:47 -07:00
David Anderson
0be475ba46 Revert "tailcfg, controlclient, magicsock: request IPv6 endpoints, but ignore them"
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>
2020-03-20 03:26:17 -07:00
Brad Fitzpatrick
f9184ad12b go.mod: update wireguard-go for https://github.com/tailscale/wireguard-go/pull/13 2020-03-19 22:46:36 -07:00