DERP writes go via TCP and the host OS will have plenty of buffer space.
We've observed in the wild with a backed up TCP socket kernel side
buffers of >2.4MB. The DERP internal queue being larger causes an
increase in the probability that the contents of the backbuffer are
"dead letters" - packets that were assumed to be lost.
A first step to improvement is to size this queue only large enough to
avoid some of the initial connect stall problem, but not large enough
that it is contributing in a substantial way to buffer bloat /
dead-letter retention.
Updates tailscale/corp#31762
Signed-off-by: James Tucker <james@tailscale.com>
I need a ringbuffer in the more traditional sense, one that has a notion
of item removal as well as tail loss on overrun. This implementation is
really a clearable log window, and is used as such where it is used.
Updates #cleanup
Updates tailscale/corp#31762
Signed-off-by: James Tucker <james@tailscale.com>
Bump Go 1.25 release to include a go/types patch and resolve govulncheck
CI exceptions.
Updates tailscale/corp#31755
Signed-off-by: Patrick O'Doherty <patrick@tailscale.com>
Extract field comments from AST and include them in generated view
methods. Comments are preserved from the original struct fields to
provide documentation for the view accessors.
Fixes#16958
Signed-off-by: Maisem Ali <3953239+maisem@users.noreply.github.com>
fixestailscale/corp#26369
The suggested exit node is currently only calculated during a localAPI request.
For older UIs, this wasn't a bad choice - we could just fetch it on-demand when a menu
presented itself. For newer incarnations however, this is an always-visible field
that needs to react to changes in the suggested exit node's value.
This change recalculates the suggested exit node ID on netmap updates and
broadcasts it on the IPN bus. The localAPI version of this remains intact for the
time being.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
updates tailscale/corp#29841
Adds a node cap macOS UIs can query to determine
whether then should enable the new windowed UI.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
Pull the lock-bearing code into a closure, and use a clone rather than a
shallow copy of the hostinfo record.
Updates #11649
Change-Id: I4f1d42c42ce45e493b204baae0d50b1cbf82b102
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
The early unlock on this branch was required because the "send" method goes on
to acquire the mutex itself. Rather than release the lock just to acquire it
again, call the underlying locked helper directly.
Updates #11649
Change-Id: I50d81864a00150fc41460b7486a9c65655f282f5
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
In places where we are locking the LocakBackend and immediately deferring an
unlock, and where there is no shortcut path in the control flow below the
deferral, we do not need the unlockOnce helper. Replace all these with use of
the lock directly.
Updates #11649
Change-Id: I3e6a7110dfc9ec6c1d38d2585c5367a0d4e76514
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Instead of referring to groups, which is a term of art for a different entity,
update the doc comments to more accurately describe what tags are in reference
to the policy document.
Updates #cleanup
Change-Id: Iefff6f84981985f834bae7c6a6c34044f53f2ea2
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
There are several methods within the LocalBackend that used an unusual and
error-prone lock discipline whereby they require the caller to hold the backend
mutex on entry, but release it on the way out.
In #11650 we added some support code to make this pattern more visible.
Now it is time to eliminate the pattern (at least within this package).
This is intended to produce no semantic changes, though I am relying on
integration tests and careful inspection to achieve that.
To the extent possible I preserved the existing control flow. In a few places,
however, I replaced this with an unlock/lock closure. This means we will
sometimes reacquire a lock only to release it again one frame up the stack, but
these operations are not performance sensitive and the legibility gain seems
worthwhile.
We can probably also pull some of these out into separate methods, but I did
not do that here so as to avoid other variable scope changes that might be hard
to see. I would like to do some more cleanup separately.
As a follow-up, we could also remove the unlockOnce helper, but I did not do
that here either.
Updates #11649
Change-Id: I4c92d4536eca629cfcd6187528381c33f4d64e20
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
The serve code leaves it up to the system's DNS resolver and netstack to
figure out how to reach the proxy destination. Combined with k8s-proxy
running in userspace mode, this means we can't rely on MagicDNS being
available or tailnet IPs being routable. I'd like to implement that as a
feature for serve in userspace mode, but for now the safer fix to get
kube-apiserver ProxyGroups consistently working in all environments is to
switch to using localhost as the proxy target instead.
This has a small knock-on in the code that does WhoIs lookups, which now
needs to check the X-Forwarded-For header that serve populates to get
the correct tailnet IP to look up, because the request's remote address
will be loopback.
Fixes#16920
Change-Id: I869ddcaf93102da50e66071bb00114cc1acc1288
Signed-off-by: Tom Proctor <tomhjp@users.noreply.github.com>
This increases throughput over long fat networks, and in the presence
of crypto/syscall-induced delay.
Updates tailscale/corp#31164
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Update odic-funnel-clients.json to take a path, this
allows setting the location of the file and prevents
it from landing in the root directory or users home directory.
Move setting of rootPath until after tsnet has started.
Previously this was added for the lazy creation of the
oidc-key.json. It's now needed earlier in the flow.
Updates #16734Fixes#16844
Signed-off-by: Mike O'Driscoll <mikeo@tailscale.com>
Add the ability for operators of natc in consensus mode to remove
servers from the raft cluster config, without losing other state.
Updates #14667
Signed-off-by: Fran Bull <fran@tailscale.com>
Currently consensus has a bootstrap routine where a tsnet node tries to
join each other node with the cluster tag, and if it is not able to join
any other node it starts its own cluster.
That algorithm is racy, and can result in split brain (more than one
leader/cluster) if all the nodes for a cluster are started at the same
time.
Add a FollowOnly argument to the bootstrap function. If provided this
tsnet node will never lead, it will try (and retry with exponential back
off) to follow any node it can contact.
Add a --follow-only flag to cmd/natc that uses this new tsconsensus
functionality.
Also slightly reorganize some arguments into opts structs.
Updates #14667
Signed-off-by: Fran Bull <fran@tailscale.com>
This significantly improves throughput of a peer relay server on Linux.
Server.packetReadLoop no longer passes sockets down the stack. Instead,
packet handling methods return a netip.AddrPort and []byte, which
packetReadLoop gathers together for eventual batched writes on the
appropriate socket(s).
Updates tailscale/corp#31164
Signed-off-by: Jordan Whited <jordan@tailscale.com>
We have been unintentionally ignoring errors from calling bootstrap.
bootstrap sometimes calls raft.BootstrapCluster which sometimes returns
a safe to ignore error, handle that case appropriately.
Updates #14667
Signed-off-by: Fran Bull <fran@tailscale.com>
This has come up in a few situations recently and adding these helpers
is much better than copying the slice (calling AsSlice()) in order to
use slices.Max and friends.
Updates #cleanup
Change-Id: Ib289a07d23c3687220c72c4ce341b9695cd875bf
Signed-off-by: Adrian Dewhurst <adrian@tailscale.com>
Update the runall handler to be more generic with an
exclude param to exclude multiple probes as the requesters
definition.
Updates tailscale/corp#27370
Signed-off-by: Mike O'Driscoll <mikeo@tailscale.com>
Cleanup nix support, make flake easier to read with nix-systems.
This also harmonizes with golinks flake setup and reduces an input
dependency by 1.
Update deps test to ensure the vendor hash stays harmonized
with go.mod.
Update make tidy to ensure vendor hash stays current.
Overlay the current version of golang, tailscale runs
recent releases faster than nixpkgs can update them into
the unstable branch.
Updates #16637
Signed-off-by: Mike O'Driscoll <mikeo@tailscale.com>
The -Environment argument to Start-Process is essentially being treated
as a delta; removing a particular variable from the argument's hash
table does not indicate to delete. Instead we must set the value of each
unwanted variable to $null.
Updates https://github.com/tailscale/corp/issues/29940
Signed-off-by: Aaron Klotz <aaron@tailscale.com>
Some of the operations of the local API need an event bus to correctly
instantiate other components (notably including the portmapper).
This commit adds that, and as the parameter list is starting to get a bit long
and hard to read, I took the opportunity to move the arguments to a config
type. Only a few call sites needed to be updated and this API is not intended
for general use, so I did not bother to stage the change.
Updates #15160
Updates #16842
Change-Id: I7b057d71161bd859f5acb96e2f878a34c85be0ef
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
gocross-wrapper.ps1 is a PowerShell core script that is essentially a
straight port of gocross-wrapper.sh. It requires PowerShell 7.4, which
is the latest LTS release of PSCore.
Why use PowerShell Core instead of Windows PowerShell? Essentially
because the former is much better to script with and is the edition
that is currently maintained.
Because we're using PowerShell Core, but many people will be running
scripts from a machine that only has Windows PowerShell, go.cmd has
been updated to prompt the user for PowerShell core installation if
necessary.
gocross-wrapper.sh has also been updated to utilize the PSCore script
when running under cygwin or msys.
gocross itself required a couple of updates:
We update gocross to output the PowerShell Core wrapper alongside the
bash wrapper, which will propagate the revised scripts to other repos
as necessary.
We also fix a couple of things in gocross that didn't work on Windows:
we change the toolchain resolution code to use os.UserHomeDir instead
of directly referencing the HOME environment variable, and we fix a
bug in the way arguments were being passed into exec.Command on
non-Unix systems.
Updates https://github.com/tailscale/corp/issues/29940
Signed-off-by: Aaron Klotz <aaron@tailscale.com>
Add a Run all probes handler that executes all
probes except those that are continuous or the derpmap
probe.
This is leveraged by other tooling to confirm DERP
stability after a deploy.
Updates tailscale/corp#27370
Signed-off-by: Mike O'Driscoll <mikeo@tailscale.com>
fixestailscale/corp#31299
Fixes two issues:
getInterfaceIndex would occasionally race with netmon's state, returning
the cached default interface index after it had be changed by NWNetworkMonitor.
This had the potential to cause connections to bind to the prior default. The fix
here is to preferentially use the interface index provided by NWNetworkMonitor
preferentially.
When no interfaces are available, macOS will set the tunnel as the default
interface when an exit node is enabled, potentially causing getInterfaceIndex
to return utun's index. We now guard against this when taking the
defaultIdx path.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
This pulls in a change from github.com/tailscale/QDK to verify code signing
when using QNAP_SIGNING_SCRIPT.
It also upgrades to the latest Google Cloud PKCS#11 library, and reorders
the Dockerfile to allow for more efficient future upgrades to the included QDK.
Updates tailscale/corp#23528
Signed-off-by: Percy Wegmann <percy@tailscale.com>
Define the HardwareAttestionKey interface describing a platform-specific
hardware backed node identity attestation key. Clients will register the
key type implementations for their platform.
Updates tailscale/corp#31269
Signed-off-by: Patrick O'Doherty <patrick@tailscale.com>
dnstype.Resolver adds a boolean UseWithExitNode that controls
whether the resolver should be used in tailscale exit node contexts
(not wireguard exit nodes). If UseWithExitNode resolvers are found,
they are installed as the global resolvers. If no UseWithExitNode resolvers
are found, the exit node resolver continues to be installed as the global
resolver. Split DNS Routes referencing UseWithExitNode resolvers are also
installed.
Updates #8237Fixestailscale/corp#30906Fixestailscale/corp#30907
Signed-off-by: Michael Ben-Ami <mzb@tailscale.com>
We already show a message in the menu itself, this just adds it to the
CLI output as well.
Updates #1708
Change-Id: Ia101a4a3005adb9118051b3416f5a64a4a45987d
Signed-off-by: Will Norris <will@tailscale.com>
This adds support for having every viewer type implement
jsonv2.MarshalerTo and jsonv2.UnmarshalerFrom.
This provides a significant boost in performance
as the json package no longer needs to validate
the entirety of the JSON value outputted by MarshalJSON,
nor does it need to identify the boundaries of a JSON value
in order to call UnmarshalJSON.
For deeply nested and recursive MarshalJSON or UnmarshalJSON calls,
this can improve runtime from O(N²) to O(N).
This still references "github.com/go-json-experiment/json"
instead of the experimental "encoding/json/v2" package
now available in Go 1.25 under goexperiment.jsonv2
so that code still builds without the experiment tag.
Of note, the "github.com/go-json-experiment/json" package
aliases the standard library under the right build conditions.
Updates tailscale/corp#791
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
James Tucker
License Updater
Patrick O'Doherty
Maisem Ali
Jonathan Nobels
Jordan Whited
M. J. Fromberger
Aaron Klotz
Claus Lensbøl
Kot C
Mohammed Naser
Need-an-AwP
Tom Proctor
Mike O'Driscoll
dependabot[bot]
Fran Bull
Adrian Dewhurst
Will Norris
Percy Wegmann
Will Hannah
Michael Ben-Ami
Joe Tsai