In particular, tests showing that #3824 works. But that test doesn't
actually work yet; it only gets a DERP connection. (why?)
Updates #13038
Change-Id: Ie1fd1b6a38d4e90fae7e72a0b9a142a95f0b2e8f
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Troubleshooting DNS resolution issues often requires additional information.
This PR expands the effect of the TS_DEBUG_DNS_FORWARD_SEND envknob to forwarder.forwardWithDestChan,
and includes the request type, domain name length, and the first 3 bytes of the domain's SHA-256 hash in the output.
Fixes#13070
Signed-off-by: Nick Khyl <nickk@tailscale.com>
I noticed a few places with custom http.Transport where we are not
closing idle connections when transport is no longer used.
Updates tailscale/corp#21609
Signed-off-by: Anton Tolchanov <anton@tailscale.com>
We were copying 12 out of the 16 bytes which meant that
the 1:1 NAT required would only work if the last 4 bytes
happened to match between the new and old address, something
that our tests accidentally had. Fix it by copying the full
16 bytes and make the tests also verify the addr and use rand
addresses.
Updates #9511
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Updates tailscale/tailscale#1634
This PR optimizes captive portal detection on Android and iOS by excluding cellular data interfaces (`pdp*` and `rmnet`). As cellular networks do not present captive portals, frequent network switches between Wi-Fi and cellular would otherwise trigger captive detection unnecessarily, causing battery drain.
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
This commit implements TCP GRO for packets being written to gVisor on
Linux. Windows support will follow later. The wireguard-go dependency is
updated in order to make use of newly exported IP checksum functions.
gVisor is updated in order to make use of newly exported
stack.PacketBuffer GRO logic.
TCP throughput towards gVisor, i.e. TUN write direction, is dramatically
improved as a result of this commit. Benchmarks show substantial
improvement, sometimes as high as 2x. High bandwidth-delay product
paths remain receive window limited, bottlenecked by gVisor's default
TCP receive socket buffer size. This will be addressed in a follow-on
commit.
The iperf3 results below demonstrate the effect of this commit between
two Linux computers with i5-12400 CPUs. There is roughly ~13us of round
trip latency between them.
The first result is from commit 57856fc without TCP GRO.
Starting Test: protocol: TCP, 1 streams, 131072 byte blocks
- - - - - - - - - - - - - - - - - - - - - - - - -
Test Complete. Summary Results:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 4.77 GBytes 4.10 Gbits/sec 20 sender
[ 5] 0.00-10.00 sec 4.77 GBytes 4.10 Gbits/sec receiver
The second result is from this commit with TCP GRO.
Starting Test: protocol: TCP, 1 streams, 131072 byte blocks
- - - - - - - - - - - - - - - - - - - - - - - - -
Test Complete. Summary Results:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.6 GBytes 9.14 Gbits/sec 20 sender
[ 5] 0.00-10.00 sec 10.6 GBytes 9.14 Gbits/sec receiver
Updates #6816
Signed-off-by: Jordan Whited <jordan@tailscale.com>
I updated the address parsing stuff to return a specific error for
unspecified hosts passed as empty strings, and look for that
when logging errors. I explicitly did not make parseAddress return a
netip.Addr containing an unspecified address because at this layer,
in the absence of any host, we don't necessarily know the address
family we're dealing with.
For the purposes of this code I think this is fine, at least until
we implement #12588.
Fixes#12979
Signed-off-by: Aaron Klotz <aaron@tailscale.com>
It seems some security software or macOS itself might be MITMing TLS
(for ScreenTime?), so don't warn unless it fails x509 validation
against system roots.
Updates #3198
Change-Id: I6ea381b5bb6385b3d51da4a1468c0d803236b7bf
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This commit implements TCP GSO for packets being read from gVisor on
Linux. Windows support will follow later. The wireguard-go dependency is
updated in order to make use of newly exported GSO logic from its tun
package.
A new gVisor stack.LinkEndpoint implementation has been established
(linkEndpoint) that is loosely modeled after its predecessor
(channel.Endpoint). This new implementation supports GSO of monster TCP
segments up to 64K in size, whereas channel.Endpoint only supports up to
32K. linkEndpoint will also be required for GRO, which will be
implemented in a follow-on commit.
TCP throughput from gVisor, i.e. TUN read direction, is dramatically
improved as a result of this commit. Benchmarks show substantial
improvement through a wide range of RTT and loss conditions, sometimes
as high as 5x.
The iperf3 results below demonstrate the effect of this commit between
two Linux computers with i5-12400 CPUs. There is roughly ~13us of round
trip latency between them.
The first result is from commit 57856fc without TCP GSO.
Starting Test: protocol: TCP, 1 streams, 131072 byte blocks
- - - - - - - - - - - - - - - - - - - - - - - - -
Test Complete. Summary Results:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 2.51 GBytes 2.15 Gbits/sec 154 sender
[ 5] 0.00-10.00 sec 2.49 GBytes 2.14 Gbits/sec receiver
The second result is from this commit with TCP GSO.
Starting Test: protocol: TCP, 1 streams, 131072 byte blocks
- - - - - - - - - - - - - - - - - - - - - - - - -
Test Complete. Summary Results:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 12.6 GBytes 10.8 Gbits/sec 6 sender
[ 5] 0.00-10.00 sec 12.6 GBytes 10.8 Gbits/sec receiver
Updates #6816
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Fixestailscale/tailscale#12973
Updates tailscale/tailscale#1634
There was a logic issue in the captive detection code we shipped in https://github.com/tailscale/tailscale/pull/12707.
Assume a captive portal has been detected, and the user notified. Upon switching to another Wi-Fi that does *not* have a captive portal, we were issuing a signal to interrupt any pending captive detection attempt. However, we were not also setting the `captive-portal-detected` warnable to healthy. The result was that any "captive portal detected" alert would not be cleared from the UI.
Also fixes a broken log statement value.
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
fixes tailscale#12968
The dns manager cleanup func was getting passed a nil
health tracker, which will panic. Fixed to pass it
the system health tracker.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
updates tailscale/corp#21823
Misconfigured, broken, or blocked DNS will often present as
"internet is broken'" to the end user. This plumbs the health tracker
into the dns manager and forwarder and adds a health warning
with a 5 second delay that is raised on failures in the forwarder and
lowered on successes.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
Updates tailscale/corp#21949
As discussed with @raggi, this PR updates the static DERPMap embedded in the client to reflect the availability of HTTP on the DERP servers run by Tailscale.
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
Updates tailscale/tailscale#1634
This PR introduces a new `captive-portal-detected` Warnable which is set to an unhealthy state whenever a captive portal is detected on the local network, preventing Tailscale from connecting.
ipn/ipnlocal: fix captive portal loop shutdown
Change-Id: I7cafdbce68463a16260091bcec1741501a070c95
net/captivedetection: fix mutex misuse
ipn/ipnlocal: ensure that we don't fail to start the timer
Change-Id: I3e43fb19264d793e8707c5031c0898e48e3e7465
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
This commit truncates any additional information (mainly hostnames) that's passed to controlD via DOH URL in DoHIPsOfBase.
This change is to make sure only resolverID is passed to controlDv6Gen but not the additional information.
Updates: #7946
Signed-off-by: KevinLiang10 <37811973+KevinLiang10@users.noreply.github.com>
This adds a package with GP-related functions and types to be used in the future PRs.
It also updates nrptRuleDatabase to use the new package instead of its own gpNotificationWatcher implementation.
Updates #12687
Signed-off-by: Nick Khyl <nickk@tailscale.com>
Updates tailscale/corp#20677
The recover function wasn't getting set in the benchmark
tests. Default changed to an empty func.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
And some misc doc tweaks for idiomatic Go style.
Updates #cleanup
Change-Id: I3ca45f78aaca037f433538b847fd6a9571a2d918
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This turns the checklocks workflow into a real check, and adds
annotations to a few basic packages as a starting point.
Updates #12625
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: I2b0185bae05a843b5257980fc6bde732b1bdd93f
Previously, if we had a umask set (e.g. 0027) that prevented creating a
world-readable file, /etc/resolv.conf would be created without the o+r
bit and thus other users may be unable to resolve DNS.
Since a umask only applies to file creation, chmod the file after
creation and before renaming it to ensure that it has the appropriate
permissions.
Updates #12609
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: I2a05d64f4f3a8ee8683a70be17a7da0e70933137
Fixestailscale/corp#20677
Replaces the original attempt to rectify this (by injecting a netMon
event) which was both heavy handed, and missed cases where the
netMon event was "minor".
On apple platforms, the fetching the interface's nameservers can
and does return an empty list in certain situations. Apple's API
in particular is very limiting here. The header hints at notifications
for dns changes which would let us react ahead of time, but it's all
private APIs.
To avoid remaining in the state where we end up with no
nameservers but we absolutely need them, we'll react
to a lack of upstream nameservers by attempting to re-query
the OS.
We'll rate limit this to space out the attempts. It seems relatively
harmless to attempt a reconfig every 5 seconds (triggered
by an incoming query) if the network is in this broken state.
Missing nameservers might possibly be a persistent condition
(vs a transient error), but that would also imply that something
out of our control is badly misconfigured.
Tested by randomly returning [] for the nameservers. When switching
between Wifi networks, or cell->wifi, this will randomly trigger
the bug, and we appear to reliably heal the DNS state.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
This is implemented via GetBestInterfaceEx. Should we encounter errors
or fail to resolve a valid, non-Tailscale interface, we fall back to
returning the index for the default interface instead.
Fixes#12551
Signed-off-by: Aaron Klotz <aaron@tailscale.com>
I meant to do this in the earlier change and had a git fail.
To atone, add a test too while I'm here.
Updates #12486
Updates #12507
Change-Id: I4943b454a2530cb5047636f37136aa2898d2ffc7
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
I noticed we were allocating these every time when they could just
share the same memory. Rather than document ownership, just lock it
down with a view.
I was considering doing all of the fields but decided to just do this
one first as test to see how infectious it became. Conclusion: not
very.
Updates #cleanup (while working towards tailscale/corp#20514)
Change-Id: I8ce08519de0c9a53f20292adfbecd970fe362de0
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
For pprof cosmetic/confusion reasons more than performance, but it
might have tiny speed benefit.
Updates #12486
Change-Id: I40e03714f3afa3a7e7f5e1fa99b81c7e889b91b6
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
So profiles show more useful names than just func1, func2, func3, etc.
There will still be func1 on them all, but the symbol before will say
what the lookup type is.
Updates #12486
Change-Id: I910b024a7861394eb83d07f5a899eae338cb1f22
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This moves NewContainsIPFunc from tsaddr to new ipset package.
And wgengine/filter types gets split into wgengine/filter/filtertype,
so netmap (and thus the CLI, etc) doesn't need to bring in ipset,
bart, etc.
Then add a test making sure the CLI deps don't regress.
Updates #1278
Change-Id: Ia246d6d9502bbefbdeacc4aef1bed9c8b24f54d5
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
NewContainsIPFunc was previously documented as performing poorly if
there were many netip.Prefixes to search over. As such, we never it used it
in such cases.
This updates it to use bart at a certain threshold (over 6 prefixes,
currently), at which point the bart lookup overhead pays off.
This is currently kinda useless because we're not using it. But now we
can and get wins elsewhere. And we can remove the caveat in the docs.
goos: darwin
goarch: arm64
pkg: tailscale.com/net/tsaddr
│ before │ after │
│ sec/op │ sec/op vs base │
NewContainsIPFunc/empty-8 2.215n ± 11% 2.239n ± 1% +1.08% (p=0.022 n=10)
NewContainsIPFunc/cidr-list-1-8 17.44n ± 0% 17.59n ± 6% +0.89% (p=0.000 n=10)
NewContainsIPFunc/cidr-list-2-8 27.85n ± 0% 28.13n ± 1% +1.01% (p=0.000 n=10)
NewContainsIPFunc/cidr-list-3-8 36.05n ± 0% 36.56n ± 13% +1.41% (p=0.000 n=10)
NewContainsIPFunc/cidr-list-4-8 43.73n ± 0% 44.38n ± 1% +1.50% (p=0.000 n=10)
NewContainsIPFunc/cidr-list-5-8 51.61n ± 2% 51.75n ± 0% ~ (p=0.101 n=10)
NewContainsIPFunc/cidr-list-10-8 95.65n ± 0% 68.92n ± 0% -27.94% (p=0.000 n=10)
NewContainsIPFunc/one-ip-8 4.466n ± 0% 4.469n ± 1% ~ (p=0.491 n=10)
NewContainsIPFunc/two-ip-8 8.002n ± 1% 7.997n ± 4% ~ (p=0.697 n=10)
NewContainsIPFunc/three-ip-8 27.98n ± 1% 27.75n ± 0% -0.82% (p=0.012 n=10)
geomean 19.60n 19.07n -2.71%
Updates #12486
Change-Id: I2e2320cc4384f875f41721374da536bab995c1ce
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Without this rule, Windows 8.1 and newer devices issue parallel DNS requests to DNS servers
associated with all network adapters, even when "Override local DNS" is enabled and/or
a Mullvad exit node is being used, resulting in DNS leaks.
This also adds "disable-local-dns-override-via-nrpt" nodeAttr that can be used to disable
the new behavior if needed.
Fixestailscale/corp#20718
Signed-off-by: Nick Khyl <nickk@tailscale.com>
Updates tailscale/tailscale#4136
This PR is the first round of work to move from encoding health warnings as strings and use structured data instead. The current health package revolves around the idea of Subsystems. Each subsystem can have (or not have) a Go error associated with it. The overall health of the backend is given by the concatenation of all these errors.
This PR polishes the concept of Warnable introduced by @bradfitz a few weeks ago. Each Warnable is a component of the backend (for instance, things like 'dns' or 'magicsock' are Warnables). Each Warnable has a unique identifying code. A Warnable is an entity we can warn the user about, by setting (or unsetting) a WarningState for it. Warnables have:
- an identifying Code, so that the GUI can track them as their WarningStates come and go
- a Title, which the GUIs can use to tell the user what component of the backend is broken
- a Text, which is a function that is called with a set of Args to generate a more detailed error message to explain the unhappy state
Additionally, this PR also begins to send Warnables and their WarningStates through LocalAPI to the clients, using ipn.Notify messages. An ipn.Notify is only issued when a warning is added or removed from the Tracker.
In a next PR, we'll get rid of subsystems entirely, and we'll start using structured warnings for all errors affecting the backend functionality.
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
This commit introduces a userspace program for managing an experimental
eBPF XDP STUN server program. derp/xdp contains the eBPF pseudo-C along
with a Go pkg for loading it and exporting its metrics.
cmd/xdpderper is a package main user of derp/xdp.
Updates tailscale/corp#20689
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Fixestailscale/corp#20677
On macOS sleep/wake, we're encountering a condition where reconfigure the network
a little bit too quickly - before apple has set the nameservers for our interface.
This results in a persistent condition where we have no upstream resolver and
fail all forwarded DNS queries.
No upstream nameservers is a legitimate configuration, and we have no (good) way
of determining when Apple is ready - but if we need to forward a query, and we
have no nameservers, then something has gone badly wrong and the network is
very broken.
A simple fix here is to simply inject a netMon event, which will go through the
configuration dance again when we hit the SERVFAIL condition.
Tested by artificially/randomly returning [] for the list of nameservers in the bespoke
ipn-bridge code responsible for getting the nameservers.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
As an alterative to #11935 using #12003.
Updates #11935
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: I05f643fe812ceeaec5f266e78e3e529cab3a1ac3
When we're starting child processes on Windows that are CLI programs that
don't need to output to a console, we should pass in DETACHED_PROCESS as a
CreationFlag on SysProcAttr. This prevents the OS from even creating a console
for the child (and paying the associated time/space penalty for new conhost
processes). This is more efficient than letting the OS create the console
window and then subsequently trying to hide it, which we were doing at a few
callsites.
Fixes#12270
Signed-off-by: Aaron Klotz <aaron@tailscale.com>
As quad-100 is an authoritative server for 4via6 domains, it should always return responses
with a response code of 0 (indicating no error) when resolving records for these domains.
If there's no resource record of the specified type (e.g. A), it should return a response
with an empty answer section rather than NXDomain. Such a response indicates that there
is at least one RR of a different type (e.g., AAAA), suggesting the Windows stub resolver
to look for it.
Fixestailscale/corp#20767
Signed-off-by: Nick Khyl <nickk@tailscale.com>
Add a new TS_EXPERIMENTAL_ENABLE_FORWARDING_OPTIMIZATIONS env var
that can be set for tailscale/tailscale container running as
a subnet router or exit node to enable UDP GRO forwarding
for improved performance.
See https://tailscale.com/kb/1320/performance-best-practices#linux-optimizations-for-subnet-routers-and-exit-nodes
This is currently considered an experimental approach;
the configuration support is partially to allow further experimentation
with containerized environments to evaluate the performance
improvements.
Updates tailscale/tailscale#12295
Signed-off-by: Irbe Krumina <irbe@tailscale.com>
Updates corp#15802.
Adds the ability for control to disable the recently added change that uses split DNS in more cases on iOS. This will allow us to disable the feature if it leads to regression in production. We plan to remove this knob once we've verified that the feature works properly.
Signed-off-by: Andrea Gottardo <andrea@gottardo.me>
This bug was introduced in e6b84f215 (May 2020) but was only used in
tests when stringifying probeProto values on failure so it wasn't
noticed for a long time.
But then it was moved into non-test code in 8450a18aa (Jun 2024) and I
didn't notice during the code movement that it was wrong. It's still
only used in failure paths in logs, but having wrong/ambiguous
debugging information isn't the best.
Whoops.
Updates tailscale/corp#20654
Change-Id: I296c727ed1c292a04db7b46ecc05c07fc1abc774
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Brad Fitzpatrick
Nick Khyl
Anton Tolchanov
VimT
Jordan Whited
Maisem Ali
Andrea Gottardo
Aaron Klotz
Jonathan Nobels
KevinLiang10
James Tucker
Andrew Dunham
Irbe Krumina