Commit Graph

1457 Commits (ab637ea4970cf08e0bdb09bc05b06b2b776bc9c7)
 

Author SHA1 Message Date
Josh Bleecher Snyder ab637ea497 try a stupider way to have one goroutine per udp read (doesn't work)
name                  old time/op    new time/op    delta
ReceiveFrom-8           16.8µs ±29%    22.8µs ±11%  +35.22%  (p=0.000 n=10+10)
ReceiveFrom_Native-8    19.1µs ±11%    20.6µs ± 9%   +8.03%  (p=0.009 n=10+10)

name                  old alloc/op   new alloc/op   delta
ReceiveFrom-8             112B ± 0%      112B ± 0%     ~     (all equal)
ReceiveFrom_Native-8      112B ± 0%      112B ± 0%     ~     (all equal)

name                  old allocs/op  new allocs/op  delta
ReceiveFrom-8             3.00 ± 0%      3.00 ± 0%     ~     (all equal)
ReceiveFrom_Native-8      3.00 ± 0%      3.00 ± 0%     ~     (all equal)
4 years ago
Brad Fitzpatrick 9503be083d tailcfg: update comments a bit 4 years ago
Brad Fitzpatrick 88179121e3 version: bump date 4 years ago
Brad Fitzpatrick 7b92f8e718 wgengine/magicsock: add start of magicsock benchmarks (Conn.ReceiveIPv4 for now)
And only single-threaded for now. Will get fancier later.

Updates #414
4 years ago
Brad Fitzpatrick 713cbe84c1 wgengine/magicsock: use net.JoinHostPort when host might have colons (udp6)
Only affected tests. (where it just generated log spam)
4 years ago
David Anderson be6fe393c5 wgengine: don't try pinging IPv6 addresses in legacy pinger.
Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson dfbde3d3aa ipn: pass through the prefix length from control.
Control sets this to /32 for IPv4 and /128 for IPv6.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson 4c8ccd6dd6 tailcfg: document new debug flag. 4 years ago
Brad Fitzpatrick c0af7deb86 tailcfg, cmd/tailscale: add Hostinfo.ShareeNode, hide in "tailscale status" 4 years ago
Brad Fitzpatrick ab482118ad tailcfg: add some missing json omitempty
Noticed these in MapResponses to clients.

MachineAuthorized was set true, but once we fix the coordination server
to zero out that field, then it can be omittted.
4 years ago
Dmytro Tananayskiy c431382720 Fix receiver in order to be consistent: syncs.WaitGroupChan
Signed-off-by: Dmytro Tananayskiy <dmitriyminer@gmail.com>
4 years ago
Josh Bleecher Snyder 3a7402aa2d logtail: help the server be more efficient
Add content length hints to headers.
The server can use these hints to more efficiently select buffers.

Stop attempting to compress tiny requests.
The bandwidth savings are negligible (and sometimes negative!),
and it makes extra work for the server.

Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Brad Fitzpatrick cd6099113f ipn: add a comment about skipping files with null bytes
Updates #954
4 years ago
Alex Brainman 72e082aaf5 ipn: make LoadPrefs return os.ErrNotExist when reading corrupted files
It appears some users have corrupted pref.conf files. Have LoadPrefs
treat these files as non-existent. This way tailscale will make user
login, and not crash.

Fixes #954

Signed-off-by: Alex Brainman <alex.brainman@gmail.com>
4 years ago
David Crawshaw 2c48b4ee14 tailcfg: remove outdated comments about Clone methods
The cloner tool adds static checks that the Clone methods are up to
date, so failing to update Clone causes a compiler error.

Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
4 years ago
Sonia Appasamy 0710fca0cd
tailcfg: include ShieldsUp in HostInfo 4 years ago
Josh Bleecher Snyder aa9d7f4665 tstime: add Parse3339B, for byte slices
Use go4.org/mem for memory safety.
A slight performance hit, but a huge performance win
for clients who start with a []byte.
The perf hit is due largely to the MapHash call, which adds ~25ns.
That is necessary to keep the fast path allocation-free.

name                     old time/op    new time/op    delta
GoParse3339/Z-8             281ns ± 1%     283ns ± 2%     ~     (p=0.366 n=9+9)
GoParse3339/TZ-8            509ns ± 0%     510ns ± 1%     ~     (p=0.059 n=9+9)
GoParse3339InLocation-8     330ns ± 1%     330ns ± 0%     ~     (p=0.802 n=10+6)
Parse3339/Z-8              69.3ns ± 1%    74.4ns ± 1%   +7.45%  (p=0.000 n=9+10)
Parse3339/TZ-8              110ns ± 1%     140ns ± 3%  +27.42%  (p=0.000 n=9+10)
ParseInt-8                 8.20ns ± 1%    8.17ns ± 1%     ~     (p=0.452 n=9+9)

name                     old alloc/op   new alloc/op   delta
GoParse3339/Z-8             0.00B          0.00B          ~     (all equal)
GoParse3339/TZ-8             160B ± 0%      160B ± 0%     ~     (all equal)
GoParse3339InLocation-8     0.00B          0.00B          ~     (all equal)
Parse3339/Z-8               0.00B          0.00B          ~     (all equal)
Parse3339/TZ-8              0.00B          0.00B          ~     (all equal)

name                     old allocs/op  new allocs/op  delta
GoParse3339/Z-8              0.00           0.00          ~     (all equal)
GoParse3339/TZ-8             3.00 ± 0%      3.00 ± 0%     ~     (all equal)
GoParse3339InLocation-8      0.00           0.00          ~     (all equal)
Parse3339/Z-8                0.00           0.00          ~     (all equal)
Parse3339/TZ-8               0.00           0.00          ~     (all equal)


Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Josh Bleecher Snyder a5dd0bcb09 util/jsonutil: new package
The cornerstone API is a more memory-efficient Unmarshal.
The savings come from re-using a json.Decoder.

BenchmarkUnmarshal-8      	 4016418	       288 ns/op	       8 B/op	       1 allocs/op
BenchmarkStdUnmarshal-8   	 4189261	       283 ns/op	     184 B/op	       2 allocs/op

It also includes a Bytes type to reduce allocations
when unmarshalling a non-hex-encoded JSON string into a []byte.

Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Josh Bleecher Snyder b65eee0745 util/lineread: add docs to Reader
In particular, point out how to stop reading
and detect it on the other side.

Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Josh Bleecher Snyder 1ebbaaaebb net/interfaces: make syscall and netstat agree when multiple gateways are present
likelyHomeRouterIPDarwinSyscall iterates through the list of routes,
looking for a private gateway, returning the first one it finds.

likelyHomeRouterIPDarwinExec does the same thing,
except that it returns the last one it finds.

As a result, when there are multiple gateways,
TestLikelyHomeRouterIPSyscallExec fails.
(At least, I think that that is what is happening;
I am going inferring from observed behavior.)

Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Brad Fitzpatrick eccc167733 wgengine/monitor: fix memory corruption in Windows implementation
I used the Windows APIs wrong previously, but it had worked just
enough.

Updates #921

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
Brad Fitzpatrick 8f76548fd9 tempfork/osexec: remove old fork of os/exec
This package was a temporary fork of os/exec to fix an EINTR loop
bug that was fixed upstream for Go 1.15 in
8c1db77a92
(https://go-review.googlesource.com/c/go/+/232862), in
src/os/exec_unix.go:

8c1db77a92 (diff-72072cbd53a7240debad8aa506ff7ec795f9cfac7322e779f9bac29a4d0d0bd4)
4 years ago
Brad Fitzpatrick 5b338bf011 tempfork/registry: delete
It's unused.
4 years ago
Brad Fitzpatrick acade77c86 ipn/ipnserver: add knob to disable babysitter 4 years ago
Brad Fitzpatrick 5d96ecd5e6 net/netstat: remove a bit more unsafe
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
Brad Fitzpatrick c8939ab7c7 util/endian: add Native variable to get the platform's native binary.ByteOrder
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
Josh Bleecher Snyder 883a11f2a8 logtail: fix typo in comment
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
4 years ago
Brad Fitzpatrick d9e2edb5ae wgengine: reconfigure wireguard peer in two steps when its disco key changes
First remove the device (to clear its wireguard session key), and then
add it back.

Fixes #929

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
David Anderson 3c508a58cc wgengine/filter: don't filter GCP DNS.
Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
Brad Fitzpatrick 51c8fd1dfc logpolicy: add -race suffix to Go version when race detector in use 4 years ago
Brad Fitzpatrick ff50ddf1ee util/racebuild: add package to export a race-is-enabled const 4 years ago
Brad Fitzpatrick fc8bc76e58 wgengine/router: lock goroutine to OS thread before using OLE [windows]
See https://github.com/tailscale/tailscale/issues/921#issuecomment-727526807

Not yet sure whether this is our problem, but it can't hurt at least,
and seems like what we're supposed to do.

Updates #921
4 years ago
Brad Fitzpatrick 7a01cd27ca net/netstat: remove some unsafe
Just removing any unnecessary unsafe while auditing unsafe usage for #921.
4 years ago
Brad Fitzpatrick 45d96788b5 net/netns: remove use of unsafe on Windows
Found while auditing unsafe for #921 via the list at:

https://github.com/tailscale/tailscale/issues/921#issuecomment-727365383

No need for unsafe here, so remove it.
4 years ago
Brad Fitzpatrick 000347d4cf util/endian: add package with const for whether platform is big endian 4 years ago
Josh Bleecher Snyder b0526e8284 net/packet: remove unnecessary mark
There's no need to mask out the bottom four bits
of b[0] if we are about to shift them away.
4 years ago
Josh Bleecher Snyder efad55cf86 net/packet: speed up packet decoding
The compiler is failing to draw the connection between
slice cap and slice len, so is missing some obvious BCE opportunities.
Give it a hint by making the cap equal to the length.
The generated code is smaller and cleaner, and a bit faster.

name              old time/op    new time/op    delta
Decode/tcp4-8       12.2ns ± 1%    11.6ns ± 3%  -5.31%  (p=0.000 n=28+29)
Decode/tcp6-8       12.5ns ± 2%    11.9ns ± 2%  -4.84%  (p=0.000 n=30+30)
Decode/udp4-8       11.5ns ± 1%    11.1ns ± 1%  -3.11%  (p=0.000 n=25+24)
Decode/udp6-8       11.8ns ± 3%    11.4ns ± 1%  -3.08%  (p=0.000 n=30+26)
Decode/icmp4-8      11.0ns ± 3%    10.6ns ± 1%  -3.38%  (p=0.000 n=25+30)
Decode/icmp6-8      11.4ns ± 1%    11.1ns ± 2%  -2.29%  (p=0.000 n=27+30)
Decode/igmp-8       10.3ns ± 0%    10.0ns ± 1%  -3.26%  (p=0.000 n=19+23)
Decode/unknown-8    8.68ns ± 1%    8.38ns ± 1%  -3.55%  (p=0.000 n=28+29)
4 years ago
Brad Fitzpatrick cccdd81441 go.mod: update some deps to get past a wireguard-windows checkptr fix 4 years ago
David Anderson 2eb474dd8d wgengine/filter: add test cases for len(dsts) > 1.
While the code was correct, I broke it during a refactoring and
tests didn't detect it. This fixes that glitch.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson ce45f4f3ff wgengine/filter: inline ip6InList into match.
matchIPsOnly gets 5% slower when inlining, despite significantly reduced
memory ops and slightly tighter code.

Part of #19.

Filter/tcp6_syn_in-8     45.5ns ± 1%    42.4ns ± 2%   -6.86%  (p=0.000 n=10+10)
Filter/udp6_in-8          107ns ± 2%      94ns ± 2%  -11.50%  (p=0.000 n=9+10)

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson 3fdae12f0c wgengine/filter: eliminate unnecessary memory loads.
Doesn't materially affect benchmarks, but shrinks match6 by 30 instructions
and halves memory loads.

Part of #19.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
Josh Bleecher Snyder 47380ebcfb wgengine/filter: twiddle bits to optimize
Part of #19.

name            old time/op    new time/op    delta
Filter/icmp4-8    32.2ns ± 3%    32.5ns ± 2%     ~     (p=0.524 n=10+8)
Filter/icmp6-8    49.7ns ± 6%    43.1ns ± 4%  -13.12%  (p=0.000 n=9+10)

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson 5062131aad wgengine/filter: treat * as both a v4 and v6 wildcard.
Part of #19.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson 2d604b3791 net/packet: represent IP6 as two uint64s.
For the operations we perform on these types (mostly net6.Contains),
this encoding is much faster.

Part of #19.

name                   old time/op    new time/op    delta
Filter/icmp4-8           27.5ns ± 1%    28.0ns ± 2%   +1.89%  (p=0.016 n=5+5)
Filter/tcp4_syn_in-8     38.8ns ± 2%    38.3ns ± 1%   -1.24%  (p=0.024 n=5+5)
Filter/tcp4_syn_out-8    27.6ns ±12%    24.6ns ± 1%     ~     (p=0.063 n=5+5)
Filter/udp4_in-8         71.5ns ± 5%    65.9ns ± 1%   -7.94%  (p=0.008 n=5+5)
Filter/udp4_out-8         132ns ±13%     119ns ± 1%  -10.29%  (p=0.008 n=5+5)
Filter/icmp6-8            169ns ±10%      54ns ± 1%  -68.35%  (p=0.008 n=5+5)
Filter/tcp6_syn_in-8      149ns ± 6%      43ns ± 1%  -71.11%  (p=0.008 n=5+5)
Filter/tcp6_syn_out-8    37.7ns ± 4%    24.3ns ± 3%  -35.51%  (p=0.008 n=5+5)
Filter/udp6_in-8          179ns ± 5%     103ns ± 1%  -42.75%  (p=0.008 n=5+5)
Filter/udp6_out-8         156ns ± 3%     191ns ± 1%  +22.54%  (p=0.008 n=5+5)

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
David Anderson 04ff3c91ee wgengine/filter: add full IPv6 support.
Part of #19.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
Brad Fitzpatrick fac2b30eff control/controlclient: diagnose zero bytes from control
Updates #921

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
David Anderson a664aac877 wgengine/router: disable IPv6 if v6 policy routing is unavailable.
Fixes #895.

Signed-off-by: David Anderson <danderson@tailscale.com>
4 years ago
Brad Fitzpatrick a2d78b4d3e net/dnscache, control/controlclient: use DNS cache when dialing control
Cache DNS results of earlier login.tailscale.com control dials, and use
them for future dials if DNS is slow or broken.

Fixes various issues with trickier setups with the domain's DNS server
behind a subnet router.

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
4 years ago
Brad Fitzpatrick 97e82c6cc0 net/netcheck: remove unused DNSCache from netcheck
It's easy to add back later if/when the TODO is implemented.
4 years ago
Brad Fitzpatrick 19b0cfe89e all: prepare for GOOS=ios in Go 1.16
Work with either way for now on iOS (darwin/arm64 vs ios/arm64).

In February when Go 1.16 comes out we'll have a universal binary for
darwin/arm64 (macOS) and will drop support for Go 1.15 and its
darwin/amd64 meaning iOS. (it'll mean macOS).

Context:

* https://tip.golang.org/doc/go1.16#darwin
* https://github.com/golang/go/issues/38485
* https://github.com/golang/go/issues/42100
4 years ago