Baby steps. This permits building without much of gvisor, but not all of it.
Updates #17283
Change-Id: I8433146e259918cc901fe86b4ea29be22075b32c
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This only saves ~32KB in the minimal linux/amd64 binary, but it's a
step towards permitting not depending on gvisor for small builds.
Updates #17283
Change-Id: Iae8da5e9465127de354dbcaf25e794a6832d891b
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
We can only register one key implementation per process. When running on
macOS or Android, trying to register a separate key implementation from
feature/tpm causes a panic.
Updates #15830
Signed-off-by: Andrew Lytvynov <awly@tailscale.com>
On platforms that are causing EPIPE at a high frequency this is
resulting in non-working connections, for example when Apple decides to
forcefully close UDP sockets due to an unsoliced packet rejection in the
firewall.
Too frequent rebinds cause a failure to solicit the endpoints triggering
the rebinds, that would normally happen via CallMeMaybe.
Updates #14551
Updates tailscale/corp#25648
Signed-off-by: James Tucker <james@tailscale.com>
This commit fixes a race condition where `tailscale up --force-reauth` would
exit prematurely on an already-logged in device.
Previously, the CLI would wait for IPN to report the "Running" state and then
exit. However, this could happen before the new auth URL was printed, leading
to two distinct issues:
* **Without seamless key renewal:** The CLI could exit immediately after
the `StartLoginInteractive` call, before IPN has time to switch into
the "Starting" state or send a new auth URL back to the CLI.
* **With seamless key renewal:** IPN stays in the "Running" state
throughout the process, so the CLI exits immediately without performing
any reauthentication.
The fix is to change the CLI's exit condition.
Instead of waiting for the "Running" state, if we're doing a `--force-reauth`
we now wait to see the node key change, which is a more reliable indicator
that a successful authentication has occurred.
Updates tailscale/corp#31476
Updates tailscale/tailscale#17108
Signed-off-by: Alex Chan <alexc@tailscale.com>
This partially reverts f3d2fd2.
When that patch was written, the goroutine that responds to IPN notifications
could call `StartLoginInteractive`, creating a race condition that led to
flaky integration tests. We no longer call `StartLoginInteractive` in that
goroutine, so the race is now impossible.
Moving the `WatchIPNBus` call earlier ensures the CLI gets all necessary
IPN notifications, preventing a reauth from hanging.
Updates tailscale/corp#31476
Signed-off-by: Alex Chan <alexc@tailscale.com>
A customer wants to allow their employees to restart tailscaled at will, when access rights and MDM policy allow it,
as a way to fully reset client state and re-create the tunnel in case of connectivity issues.
On Windows, the main tailscaled process runs as a child of a service process. The service restarts the child
when it exits (or crashes) until the service itself is stopped. Regular (non-admin) users can't stop the service,
and allowing them to do so isn't ideal, especially in managed or multi-user environments.
In this PR, we add a LocalAPI endpoint that instructs ipnserver.Server, and by extension the tailscaled process,
to shut down. The service then restarts the child tailscaled. Shutting down tailscaled requires LocalAPI write access
and an enabled policy setting.
Updates tailscale/corp#32674
Updates tailscale/corp#32675
Signed-off-by: Nick Khyl <nickk@tailscale.com>
And yay: tsnet (and thus k8s-operator etc) no longer depends on
portlist! And LocalBackend is smaller.
Removes 50 KB from the minimal binary.
Updates #12614
Change-Id: Iee04057053dc39305303e8bd1d9599db8368d926
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
We made changes to ipnext callback registration/unregistration/invocation in #15780
that made resetting b.exthost to a nil, no-op host in (*LocalBackend).Shutdown() unnecessary.
But resetting it is also racy: b.exthost must be safe for concurrent use with or without b.mu held,
so it shouldn't be written after NewLocalBackend returns. This PR removes it.
Fixes#17279
Signed-off-by: Nick Khyl <nickk@tailscale.com>
This change adds full IPv6 support to the Kubernetes operator's DNS functionality,
enabling dual-stack and IPv6-only cluster support.
Fixes#16633
Signed-off-by: Raj Singh <raj@tailscale.com>
Expand the integration tests to cover a wider range of scenarios, including:
* Before and after a successful initial login
* Auth URLs and auth keys
* With and without the `--force-reauth` flag
* With and without seamless key renewal
These tests expose a race condition when using `--force-reauth` on an
already-logged in device. The command completes too quickly, preventing
the auth URL from being displayed. This issue is identified and will be
fixed in a separate commit.
Updates #17108
Signed-off-by: Alex Chan <alexc@tailscale.com>
Ideally we would remove this warning entirely, as it is now possible to
reauthenticate without losing connectivty. However, it is still possible to
lose SSH connectivity if the user changes the ownership of the machine when
they do a force-reauth, and we have no way of knowing if they are going to
do that before they do it.
For now, let's just reduce the strength of the warning to warn them that
they "may" lose their connection, rather than they "will".
Updates tailscale/corp#32429
Signed-off-by: James Sanderson <jsanderson@tailscale.com>
I think this was originally a brain-o in 9380e2dfc6. It's
disabling the port _poller_, listing what open ports (i.e. services)
are open, not PMP/PCP/UPnP port mapping.
While there, drop in some more testenv.AssertInTest() in a few places.
Updates #cleanup
Change-Id: Ia6f755ad3544f855883b8a7bdcfc066e8649547b
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
PR #17258 extracted `derp.Server` into `derp/derpserver.Server`.
This followup patch adds the following cleanups:
1. Rename `derp_server*.go` files to `derpserver*.go` to match
the package name.
2. Rename the `derpserver.NewServer` constructor to `derpserver.New`
to reduce stuttering.
3. Remove the unnecessary `derpserver.Conn` type alias.
Updates #17257
Updates #cleanup
Signed-off-by: Simon Law <sfllaw@tailscale.com>
Sidestep cmd/viewer incompatibility hiccups with
HardwareAttestationPublic type due to its *ecdsa.PublicKey inner member
by serializing the key to a byte slice instead.
Updates tailscale/corp#31269
Signed-off-by: Patrick O'Doherty <patrick@tailscale.com>
This exports a number of things from the derp (generic + client) package
to be used by the new derpserver package, as now used by cmd/derper.
And then enough other misc changes to lock in that cmd/tailscaled can
be configured to not bring in tailscale.com/client/local. (The webclient
in particular, even when disabled, was bringing it in, so that's now fixed)
Fixes#17257
Change-Id: I88b6c7958643fb54f386dd900bddf73d2d4d96d5
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Some systems need to tell whether the monitored goroutine has finished
alongside other channel operations (notably in this case the relay server, but
there seem likely to be others similarly situated).
Updates #15160
Change-Id: I5f0f3fae827b07f9b7102a3b08f60cda9737fe28
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Help out the linker's dead code elimination.
Updates #12614
Change-Id: I6c13cb44d3250bf1e3a01ad393c637da4613affb
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This doesn't yet fully pull it out into a feature/captiveportal package.
This is the usual first step, moving the code to its own files within
the same packages.
Updates #17254
Change-Id: Idfaec839debf7c96f51ca6520ce36ccf2f8eec92
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
updates tailscale/corp#32600
A localAPI/cli call to reload-config can end up leaving magicsock's mutex
locked. We were missing an unlock for the early exit where there's no change in
the static endpoints when the disk-based config is loaded. This is not likely
the root cause of the linked issue - just noted during investigation.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
In MacOS GUI apps, users have to select folders to share via the GUI. This is both because
the GUI app keeps its own record of shares, and because the sandboxed version of the GUI
app needs to gain access to the shared folders by having the user pick them in a file
selector.
The new build tag `ts_mac_gui` allows the MacOS GUI app build to signal that this
is a MacOS GUI app, which causes the `drive` subcommand to be omitted so that people
do not mistakenly attempt to use it.
Updates tailscale/tailscale#17210
Signed-off-by: Percy Wegmann <percy@tailscale.com>
Also update to use the new DisplayNameOrDefault.
Updates tailscale/corp#30456
Change-Id: Ia101a4a3005adb9118051b3416f5a64a4a45987d
Signed-off-by: Will Norris <will@tailscale.com>
This commit does not change the order or meaning of any eventbus activity, it
only updates the way the plumbing is set up.
Updates #15160
Change-Id: I61b863f9c05459d530a4c34063a8bad9046c0e27
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Add a last seen time on the cli's status command, similar to the web
portal.
Before:
```
100.xxx.xxx.xxx tailscale-operator tagged-devices linux offline
```
After:
```
100.xxx.xxx.xxx tailscale-operator tagged-devices linux offline, last seen 20d ago
```
Fixes#16584
Signed-off-by: Mahyar Mirrashed <mah.mirr@gmail.com>
This commit does not change the order or meaning of any eventbus activity, it
only updates the way the plumbing is set up.
Updates #15160
Change-Id: I06860ac4e43952a9bb4d85366138c9d9a17fd9cd
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
It is a programming error to Publish or Subscribe on a closed Client, but now
the way you discover that is by getting a panic from down in the machinery of
the bus after the client state has been cleaned up.
To provide a more helpful error, let's panic explicitly when that happens and
say what went wrong ("the client is closed"), by preventing subscriptions from
interleaving with closure of the client. With this change, either an attachment
fails outright (because the client is already closed) or completes and then
shuts down in good order in the normal course.
This does not change the semantics of the client, publishers, or subscribers,
it's just making the failure more eager so we can attach explanatory text.
Updates #15160
Change-Id: Ia492f4c1dea7535aec2cdcc2e5ea5410ed5218d2
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Only changes how the go routine consuming the events starts and stops,
not what it does.
Updates #15160
Signed-off-by: Claus Lensbøl <claus@tailscale.com>
This commit modifies the k8s operator to wrap its logger using the logtail
logger provided via the tsnet server. This causes any logs written by
the operator to make their way to Tailscale in the same fashion as
wireguard logs to be used by support.
This functionality can also be opted-out of entirely using the
"TS_NO_LOGS_NO_SUPPORT" environment variable.
Updates https://github.com/tailscale/corp/issues/32037
Signed-off-by: David Bond <davidsbond93@gmail.com>
We never implemented the peercred package on OpenBSD (and I just tried
again and failed), but we've always documented that the creds pointer
can be nil for operating systems where we can't map the unix socket
back to its UID. On those platforms, we set the default unix socket
permissions such that only the admin can open it anyway and we don't
have a read-only vs read-write distinction. OpenBSD was always in that
camp, where any access to Tailscale's unix socket meant full access.
But during some refactoring, we broke OpenBSD in that we started
assuming during one logging path (during login) that Creds was non-nil
when looking up an ipnauth.Actor's username, which wasn't relevant (it
was called from a function "maybeUsernameOf" anyway, which threw away
errors).
Verified on an OpenBSD VM. We don't have any OpenBSD integration tests yet.
Fixes#17209
Updates #17221
Change-Id: I473c5903dfaa645694bcc75e7f5d484f3dd6044d
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
controlhttp has the responsibility of dialing a set of candidate control
endpoints in a way that minimizes user facing latency. If one control
endpoint is unavailable we promptly dial another, racing across the
dimensions of: IPv6, IPv4, port 80, and port 443, over multiple server
endpoints.
In the case that the top priority endpoint was not available, the prior
implementation would hang waiting for other results, so as to try to
return the highest priority successful connection to the rest of the
client code. This hang would take too long with a large dialplan and
sufficient client to endpoint latency as to cause the server to timeout
the connection due to inactivity in the intermediate state.
Instead of trying to prioritize non-ideal candidate connections, the
first successful connection is now used unconditionally, improving user
facing latency and avoiding any delays that would encroach on the
server-side timeout.
The tests are converted to memnet and synctest, running on all
platforms.
Fixes#8442Fixestailscale/corp#32534
Co-authored-by: James Tucker <james@tailscale.com>
Change-Id: I4eb57f046d8b40403220e40eb67a31c41adb3a38
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Signed-off-by: James Tucker <james@tailscale.com>
The controlhttp dialer with a ControlDialPlan IPv6 entry was hitting a
case where the dnscache Resolver was returning an netip.Addr zero
value, where it should've been returning the IPv6 address.
We then tried to dial "invalid IP:80", which would immediately fail,
at least locally.
Mostly this was causing spammy logs when debugging other stuff.
Updates tailscale/corp#32534
Change-Id: If8b9a20f10c1a6aa8a662c324151d987fe9bd2f8
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
tsnet apps in particular never use the Linux DNS OSManagers, so they don't need
DBus, etc. I started to pull that all out into separate features so tsnet doesn't
need to bring in DBus, but hit this first.
Here you can see that tsnet (and the k8s-operator) no longer pulls in inotify.
Updates #17206
Change-Id: I7af0f391f60c5e7dbeed7a080346f83262346591
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This commit does not change the order or meaning of any eventbus activity, it
only updates the way the plumbing is set up.
Updates #15160
Change-Id: I0a175e67e867459daaedba0731bf68bd331e5ebc
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
This commit does not change the order or meaning of any eventbus activity, it
only updates the way the plumbing is set up.
Updates #15160
Change-Id: I40c23b183c2a6a6ea3feec7767c8e5417019fc07
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
A common pattern in event bus usage is to run a goroutine to service a
collection of subscribers on a single bus client. To have an orderly shutdown,
however, we need a way to wait for such a goroutine to be finished.
This commit adds a Monitor type that makes this pattern easier to wire up:
rather than having to track all the subscribers and an extra channel, the
component need only track the client and the monitor. For example:
cli := bus.Client("example")
m := cli.Monitor(func(c *eventbus.Client) {
s1 := eventbus.Subscribe[T](cli)
s2 := eventbus.Subscribe[U](cli)
for {
select {
case <-c.Done():
return
case t := <-s1.Events():
processT(t)
case u := <-s2.Events():
processU(u)
}
}
})
To shut down the client and wait for the goroutine, the caller can write:
m.Close()
which closes cli and waits for the goroutine to finish. Or, separately:
cli.Close()
// do other stuff
m.Wait()
While the goroutine management is not explicitly tied to subscriptions, it is a
common enough pattern that this seems like a useful simplification in use.
Updates #15160
Change-Id: I657afda1cfaf03465a9dce1336e9fd518a968bca
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Brad Fitzpatrick
Nick Khyl
Andrew Lytvynov
Alex Chan
James Tucker
Raj Singh
James Sanderson
Simon Law
Patrick O'Doherty
M. J. Fromberger
Jordan Whited
Jonathan Nobels
License Updater
Percy Wegmann
Will Norris
Mahyar Mirrashed
Claus Lensbøl
David Bond
Kristoffer Dalby