We were inconsistent whether we checked if the feature was already
enabled which we could do cheaply using the locally available status.
We would do the checks fine if we were turning on funnel, but not serve.
This moves the cap checks down into enableFeatureInteractive so that
are always run.
Updates #9984
Co-authored-by: Tyler Smalley <tyler@tailscale.com>
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It would end up resetting whatever hostinfo we had constructed
and leave the backend statemachine in a broken state.
This fixes that by storing the PushDeviceToken on the LocalBackend
and populating it on Hostinfo before passing it to controlclient.
Updates tailscale/corp#8940
Updates tailscale/corp#15367
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The value being passed was the same as whats on b.hostinfo, so just
use that directly.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This adds a check to prevent changes to ServeConfig if tailscaled
is run with a Locked config.
Missed in 1fc3573446.
Updates #1412
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We were previously using the netlink API to see if there are chains/rules that
already exist. This works fine in environments where there is either full
nftable support or no support at all. However, we have identified certain
environments which have partial nftable support and the only feasible way of
detecting such an environment is to try to create some of the chains that we
need.
This adds a check to create a dummy postrouting chain which is immediately
deleted. The goal of the check is to ensure we are able to use nftables and
that it won't error out later. This check is only done in the path where we
detected that the system has no preexisting nftable rules.
Updates #5621
Updates #8555
Updates #8762
Signed-off-by: Maisem Ali <maisem@tailscale.com>
These tests were broken at HEAD. CI currently does not run these
as root, will figure out how to do that in a followup.
Updates #5621
Updates #8555
Updates #8762
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously we were just smushing together args and not trying
to parse the values at all. This resulted in the args to testwrapper
being limited and confusing.
This makes it so that testwrapper parses flags in the exact format as `go test`
command and passes them down in the provided order. It uses tesing.Init to
register flags that `go test` understands, however those are not the only
flags understood by `go test` (such as `-exec`) so we register these separately.
Updates tailscale/corp#14975
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This partially reverts commits a61a9ab087
and 7538f38671 and fully reverts
4823a7e591.
The goal of that commit was to reapply known config whenever the
container restarts. However, that already happens when TS_AUTH_ONCE was
false (the default back then). So we only had to selectively reapply the
config if TS_AUTH_ONCE is true, this does exactly that.
This is a little sad that we have to revert to `tailscale up`, but it
fixes the backwards incompatibility problem.
Updates tailscale/tailscale#9539
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We were eagerly doing a synchronous renewal of the cert while
trying to serve traffic. Instead of that, just do the cert
renewal in the background and continue serving traffic as long
as the cert is still valid.
This regressed in c1ecae13ab when
we introduced ARI support and were trying to make the experience
of `tailscale cert` better. However, that ended up regressing
the experience for tsnet as it would not always doing the renewal
synchronously.
Fixes#9783
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We were too strict and required the user not specify the host field at all
in the ingress rules, but that degrades compatibility with existing helm charts.
Relax the constraint so that rule.Host can either be empty, or match the tls.Host[0]
value exactly.
Fixes#9548
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Replace the deprecated var with the one in docs to avoid confusion.
Introduced in 335a5aaf9a.
Updates #8317Fixes#9764
Signed-off-by: Maisem Ali <maisem@tailscale.com>
These log paths were actually unexpected until the refactor in
fe95d81b43. This moves the logs
to the callsites where they are actually unexpected.
Fixes#9670
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The current structure meant that we were embedding netstack in
the tailscale CLI and in the GUIs. This removes that by isolating
the checksum munging to a different pkg which is only called from
`net/tstun`.
Fixes#9756
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This migrates containerboot to reuse the NetfilterRunner used
by tailscaled instead of manipulating iptables rule itself.
This has the added advantage of now working with nftables and
we can potentially drop the `iptables` command from the container
image in the future.
Updates #9310
Co-authored-by: Irbe Krumina <irbe@tailscale.com>
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This allows using the fake runner in different packages
that need to manage filter rules.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Just a refactor to consolidate the firewall detection logic in a single
package so that it can be reused in a later commit by containerboot.
Updates #9310
Signed-off-by: Maisem Ali <maisem@tailscale.com>
These were missed when adding NodeCapMap and resulted
in tsnet binaries not being able to turn on funnel.
Fixes#9566
Signed-off-by: Maisem Ali <maisem@tailscale.com>
In typical k8s setups, the MTU configured on the eth0 interfaces is typically 1500 which
results in packets being dropped when they make it to proxy pods as the tailscale0 interface
has a 1280 MTU.
As the primary use of this functionality is TCP, add iptables based MSS clamping to allow
connectivity.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It was tailscale.com/ts-tailnet-target-ip, which was pretty
redundant. Change it to tailscale.com/tailnet-ip.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It might as well have been spewing out gibberish. This adds
a nicer output format for us to be able to read and identify
whats going on.
Sample output
```
natV4Config{nativeAddr: 100.83.114.95, listenAddrs: [10.32.80.33], dstMasqAddrs: [10.32.80.33: 407 peers]}
```
Fixestailscale/corp#14650
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Like PeerCapMap, add a field to `tailcfg.Node` which provides
a map of Capability to raw JSON messages which are deferred to be
parsed later by the application code which cares about the specific
capabilities. This effectively allows us to prototype new behavior
without having to commit to a schema in tailcfg, and it also opens up
the possibilities to develop custom behavior in tsnet applications w/o
having to plumb through application specific data in the MapResponse.
Updates #4217
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This adds a new RawMessage type backed by string instead of the
json.RawMessage which is backed by []byte. The byte slice makes
the generated views be a lot more defensive than the need to be
which we can get around by using a string instead.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The kube-apiserver proxy in the operator would only run in
auth proxy mode but thats not always desirable. There are
situations where the proxy should just be a transparent
proxy and not inject auth headers, so do that using a new
env var APISERVER_PROXY and deprecate the AUTH_PROXY env.
THe new env var has three options `false`, `true` and `noauth`.
Updates #8317
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It was only waiting for 0.5s (5ms * 100), but our CI
is too slow so make it wait up to 3s (10ms * 300).
Updates tailscale/corp#14515
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We would only check if the client was paused, but not
if the client was closed. This meant that a call to
Shutdown may block forever/leak goroutines
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It would acquire the lock, calculate `nextState`, relase
the lock, then call `enterState` which would acquire the lock
again. There were obvious races there which could lead to
nil panics as seen in a test in a different repo.
```
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x2 addr=0x70 pc=0x1050f2c7c]
goroutine 42240 [running]:
tailscale.com/ipn/ipnlocal.(*LocalBackend).enterStateLockedOnEntry(0x14002154e00, 0x6)
tailscale.com/ipn/ipnlocal/local.go:3715 +0x30c
tailscale.com/ipn/ipnlocal.(*LocalBackend).enterState(0x14002154e00?, 0x14002e3a140?)
tailscale.com/ipn/ipnlocal/local.go:3663 +0x8c
tailscale.com/ipn/ipnlocal.(*LocalBackend).stateMachine(0x14001f5e280?)
tailscale.com/ipn/ipnlocal/local.go:3836 +0x2c
tailscale.com/ipn/ipnlocal.(*LocalBackend).setWgengineStatus(0x14002154e00, 0x14002e3a190, {0x0?, 0x0?})
tailscale.com/ipn/ipnlocal/local.go:1193 +0x4d0
tailscale.com/wgengine.(*userspaceEngine).RequestStatus(0x14005d90300)
tailscale.com/wgengine/userspace.go:1051 +0x80
tailscale.com/wgengine.NewUserspaceEngine.func2({0x14002e3a0a0, 0x2, 0x140025cce40?})
tailscale.com/wgengine/userspace.go:318 +0x1a0
tailscale.com/wgengine/magicsock.(*Conn).updateEndpoints(0x14002154700, {0x105c13eaf, 0xf})
tailscale.com/wgengine/magicsock/magicsock.go:531 +0x424
created by tailscale.com/wgengine/magicsock.(*Conn).ReSTUN in goroutine 42077
tailscale.com/wgengine/magicsock/magicsock.go:2142 +0x3a4
```
Updates tailscale/corp#14480
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It would just fail the entire pkg, but would not print any
logs. It was already tracking all the logs, so have it emit
them when the pkg fails/times out.
Updates #9231
Signed-off-by: Maisem Ali <maisem@tailscale.com>
If Start was called multiple times concurrently, it would
create a new client and shutdown the previous one. However
there was a race possible between shutting down the old one
and assigning a new one where the concurent goroutine may
have assigned another one already and it would leak.
Updates tailscale/corp#14471
Signed-off-by: Maisem Ali <maisem@tailscale.com>
resetControlClientLocked is called while b.mu was held and
would call cc.Shutdown which would wait for the observer queue
to drain.
However, there may be active callbacks from cc already waiting for
b.mu resulting in a deadlock.
This makes it so that resetControlClientLocked does not call
Shutdown, and instead just returns the value.
It also makes it so that any status received from previous cc
are ignored.
Updates tailscale/corp#12827
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We already removed the async API, make it more sync and remove
the FinishLogout state too.
This also makes the callback be synchronous again as the previous
attempt was trying to work around the logout callback resulting
in a client shutdown getting blocked forever.
Updates #3833
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We have cases where the SetControlClientStatus would result in
a Shutdown call back into the auto client that would block
forever. The right thing to do here is to fix the LocalBackend
state machine but thats a different dumpster fire that we
are slowly making progress towards.
This makes it so that the SetControlClientStatus happens in a
different goroutine so that calls back into the auto client
do not block.
Also add a few missing mu.Unlocks in LocalBackend.Start.
Updates #9181
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Instead of confusing users, emit an event that explicitly tells the
user that HTTPS is disabled on the tailnet and that ingress may not
work until they enable it.
Updates #9141
Signed-off-by: Maisem Ali <maisem@tailscale.com>
While investigating the fix in 7538f38671,
I was curious why the testwrapper didn't fail. Turns out if the test
times out and there was no explicit failure, the only message we get
is that the overall pkg failed and no failure information about the
individual test. This resulted in a 0 exit code.
This fixes that by failing the explicit case of the pkg failing when
there is nothing to retry for that pkg.
Updates #8493
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We would call Update on the secret, but that was racey and would occasionaly
fail. Instead use patch whenever we can.
Fixes errors like
```
boot: 2023/08/29 01:03:53 failed to set serve config: sending serve config: updating config: writing ServeConfig to StateStore: Operation cannot be fulfilled on secrets "ts-webdav-kfrzv-0": the object has been modified; please apply your changes to the latest version and try again
{"level":"error","ts":"2023-08-29T01:03:48Z","msg":"Reconciler error","controller":"ingress","controllerGroup":"networking.k8s.io","controllerKind":"Ingress","Ingress":{"name":"webdav","namespace":"default"},"namespace":"default","name":"webdav","reconcileID":"96f5cfed-7782-4834-9b75-b0950fd563ed","error":"failed to provision: failed to create or get API key secret: Operation cannot be fulfilled on secrets \"ts-webdav-kfrzv-0\": the object has been modified; please apply your changes to the latest version and try again","stacktrace":"sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).reconcileHandler\n\tsigs.k8s.io/controller-runtime@v0.15.0/pkg/internal/controller/controller.go:324\nsigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).processNextWorkItem\n\tsigs.k8s.io/controller-runtime@v0.15.0/pkg/internal/controller/controller.go:265\nsigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).Start.func2.2\n\tsigs.k8s.io/controller-runtime@v0.15.0/pkg/internal/controller/controller.go:226"}
```
Updates #502
Updates #7895
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This was added in 3451b89e5f, but
resulted in the v6 Tailscale address being added to status when
when the forwarding only happened on the v4 address.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The new ingress reconcile raises events on failure, but I forgot to
add the updated permission.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
On k8s the serve-config secret mount is symlinked so checking against
the Name makes us miss the events.
Updates #7895
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously, the operator would only monitor Services and create
a Tailscale StatefulSet which acted as a L3 proxy which proxied
traffic inbound to the Tailscale IP onto the services ClusterIP.
This extends that functionality to also monitor Ingress resources
where the `ingressClassName=tailscale` and similarly creates a
Tailscale StatefulSet, acting as a L7 proxy instead.
Users can override the desired hostname by setting:
```
- tls
hosts:
- "foo"
```
Hostnames specified under `rules` are ignored as we only create a single
host. This is emitted as an event for users to see.
Fixes#7895
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This watches the provided path for a JSON encoded ipn.ServeConfig.
Everytime the file changes, or the nodes FQDN changes it reapplies
the ServeConfig.
At boot time, it nils out any previous ServeConfig just like tsnet does.
As the ServeConfig requires pre-existing knowledge of the nodes FQDN to do
SNI matching, it introduces a special `${TS_CERT_DOMAIN}` value in the JSON
file which is replaced with the known CertDomain before it is applied.
Updates #502
Updates #7895
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously users would have to unexpose/expose the service in order to
change Hostname/TargetIP. This now applies those changes by causing a
StatefulSet rollout now that a61a9ab087 is in.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously we would not reapply changes to TS_HOSTNAME etc when
then the container restarted and TS_AUTH_ONCE was enabled.
This splits those into two steps login and set, allowing us to
only rerun the set step on restarts.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
I forgot to move the defer out of the func, so the tsnet.Server
immediately closed after starting.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It was jumbled doing a lot of things, this breaks it up into
the svc reconciliation and the tailscale sts reconciliation.
Prep for future commit.
Updates #502
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Theory is that our long lived http2 connection to control would
get tainted by _something_ (unclear what) and would get closed.
This picks up the fix for golang/go#60818.
Updates tailscale/corp#5761
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It had a custom Clone func with a TODO to replace with cloner, resolve
that todo. Had to pull out the embedded Auth struct into a named struct.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously we would use the Impersonate-Group header to pass through
tags to the k8s api server. However, we would do nothing for non-tagged
nodes. Now that we have a way to specify these via peerCaps respect those
and send down groups for non-tagged nodes as well.
For tagged nodes, it defaults to sending down the tags as groups to retain
legacy behavior if there are no caps set. Otherwise, the tags are omitted.
Updates #5055
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Instead of having updates replace the map polls, create
a third goroutine which is solely responsible for making
sure that control is aware of the latest client state.
This also makes it so that the streaming map polls are only
broken when there are auth changes, or the client is paused.
Updates tailscale/corp#5761
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We would only look for duplicate profiles when a new login
occurred but when using `--force-reauth` we could switch
users which would end up with duplicate profiles.
Updates #7726
Signed-off-by: Maisem Ali <maisem@tailscale.com>
There are a few situations where we end up with duplicate profiles.
Add tests to identify those situations, fix in followup.
Updates #7726
Signed-off-by: Maisem Ali <maisem@tailscale.com>
It was being modified in two places in Direct for the auth routine
and then in LocalBackend when a new NetMap was received. This was
confusing, so make Direct also own changes to Persist when a new
NetMap is received.
Updates #7726
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Define PeerCapabilty and PeerCapMap as the new way of sending down
inter-peer capability information.
Previously, this was unstructured and you could only send down strings
which got too limiting for certain usecases. Instead add the ability
to send down raw JSON messages that are opaque to Tailscale but provide
the applications to define them however they wish.
Also update accessors to use the new values.
Updates #4217
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We were never resetting the backoff in streaming mapResponses.
The call to `PollNetMap` always returns with an error. Changing that contract
is harder, so manually reset backoff when a netmap is received.
Updates tailscale/corp#12894
Signed-off-by: Maisem Ali <maisem@tailscale.com>
`go test -json` outputs invalid JSON when a build fails.
Handle that case by reseting the json.Decode and continuing to read.
Updates #8493
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Without this, the client would just get stuck dialing even if the
context was canceled.
Updates tailscale/corp#12590
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously it would wait for all tests to run before printing anything,
instead stream the results over a channel so that they can be emitted
immediately.
Updates #8493
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Previously it would only print the failures without providing
more information on which package the failures from.
This commit makes it so that it prints out the package information
as well as the attempt numbers.
```
➜ tailscale.com git:(main) ✗ go run ./cmd/testwrapper ./cmd/...
ok tailscale.com/cmd/derper
ok tailscale.com/cmd/k8s-operator
ok tailscale.com/cmd/tailscale/cli
ok tailscale.com/cmd/tailscaled
=== RUN TestFlakeRun
flakytest.go:38: flakytest: issue tracking this flaky test: https://github.com/tailscale/tailscale/issues/0
flakytest_test.go:41: First run in testwrapper, failing so that test is retried. This is expected.
--- FAIL: TestFlakeRun (0.00s)
FAIL tailscale.com/cmd/testwrapper/flakytest
Attempt #2: Retrying flaky tests:
ok tailscale.com/cmd/testwrapper/flakytest
```
Updates #8493
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The panicLogWriter is too strict, and any panics that occur
get wrapped up in quotes. This makes it so that it will allow
panics to continue writing to Stderr without going through
logger.Logf.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Redo the testwrapper to track and only retry flaky tests instead
of retrying the entire pkg. It also fails early if a non-flaky test fails.
This also makes it so that the go test caches are used.
Fixes#7975
Signed-off-by: Maisem Ali <maisem@tailscale.com>
Trying to SSH when SELinux is enforced results in errors like:
```
➜ ~ ssh ec2-user@<ip>
Last login: Thu Jun 1 22:51:44 from <ip2>
ec2-user: no shell: Permission denied
Connection to <ip> closed.
```
while the `/var/log/audit/audit.log` has
```
type=AVC msg=audit(1685661291.067:465): avc: denied { transition } for pid=5296 comm="login" path="/usr/bin/bash" dev="nvme0n1p1" ino=2564 scontext=system_u:system_r:unconfined_service_t:s0 tcontext=unconfined_u:unconfined_r:unconfined_t:s0 tclass=process permissive=0
```
The right fix here would be to somehow install the appropriate context when
tailscale is installed on host, but until we figure out a way to do that
stop using the `login` cmd in these situations.
Updates #4908
Signed-off-by: Maisem Ali <maisem@tailscale.com>
The netstack code had a bunch of logic to figure out if the LocalBackend should handle an
incoming connection and then would call the function directly on LocalBackend. Move that
logic to LocalBackend and refactor the methods to return conn handlers.
Updates #cleanup
Signed-off-by: Maisem Ali <maisem@tailscale.com>
This basically allows running services on the SSH client and reaching
them from the SSH server during the session.
Updates #6575
Signed-off-by: Maisem Ali <maisem@tailscale.com>
We were only closing on side of the pty/tty pair.
Close the other side too.
Thanks to @fritterhoff for reporting and debugging the issue!
Fixes#8119
Signed-off-by: Maisem Ali <maisem@tailscale.com>
getSingleObject can return `nil, nil`, getDeviceInfo was not handling
that case which resulted in panics.
Fixes#7303
Signed-off-by: Maisem Ali <maisem@tailscale.com>
