I noticed a deadlock in a test in a in-development PR where during a
shutdown storm of things (from a tsnet.Server.Close), LocalBackend was
trying to call magicsock.Conn.Synchronize but the magicsock and/or
eventbus was already shut down and no longer processing events.
Updates #16369
Change-Id: I58b1f86c8959303c3fb46e2e3b7f38f6385036f1
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Unfortunately I closed the tab and lost it in my sea of CI failures
I'm currently fighting.
Updates #cleanup
Change-Id: I4e3a652d57d52b75238f25d104fc1987add64191
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
So they're not all run N times on the sharded oss builders
and are only run one time each.
Updates tailscale/corp#28679
Change-Id: Ie21e84b06731fdc8ec3212eceb136c8fc26b0115
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
When systemd notification support was omitted from the build, or on
non-Linux systems, we were unnecessarily emitting code and generating
garbage stringifying addresses upon transition to the Running state.
Updates #12614
Change-Id: If713f47351c7922bb70e9da85bf92725b25954b9
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This removes one of the O(n=peers) allocs in getStatus, as
Engine.getStatus happens more often than Reconfig.
Updates #17814
Change-Id: I8a87fbebbecca3aedadba38e46cc418fd163c2b0
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Previously if `chains` was empty, it would be passed to `computeActiveAncestor()`,
which would fail with the misleading error "multiple distinct chains".
Updates tailscale/corp#33846
Signed-off-by: Alex Chan <alexc@tailscale.com>
Change-Id: Ib93a755dbdf4127f81cbf69f3eece5a388db31c8
* lock released early just to call `b.send` when it can call
`b.sendToLocked` instead
* `UnlockEarly` called to release the lock before trivially fast
operations, we can wait for a defer there
Updates #11649
Signed-off-by: Andrew Lytvynov <awly@tailscale.com>
It was disabled in May 2024 in #12205 (9eb72bb51).
This removes the unused symbols.
Updates #188
Updates tailscale/corp#19106
Updates tailscale/corp#19116
Change-Id: I5208b7b750b18226ed703532ed58c4ea17195a8e
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Use GetGlobalAddrs() to discover all STUN endpoints, handling bad NATs
that create multiple mappings. When MappingVariesByDestIP is true, also
add the first STUN IPv4 address with the relay's local port for static
port mapping scenarios.
Updates #17796
Signed-off-by: Raj Singh <raj@tailscale.com>
The feature is currently in private alpha, so requires a tailnet feature
flag. Initially focuses on supporting the operator's own auth, because the
operator is the only device we maintain that uses static long-lived
credentials. All other operator-created devices use single-use auth keys.
Testing steps:
* Create a cluster with an API server accessible over public internet
* kubectl get --raw /.well-known/openid-configuration | jq '.issuer'
* Create a federated OAuth client in the Tailscale admin console with:
* The issuer from the previous step
* Subject claim `system:serviceaccount:tailscale:operator`
* Write scopes services, devices:core, auth_keys
* Tag tag:k8s-operator
* Allow the Tailscale control plane to get the public portion of
the ServiceAccount token signing key without authentication:
* kubectl create clusterrolebinding oidc-discovery \
--clusterrole=system:service-account-issuer-discovery \
--group=system:unauthenticated
* helm install --set oauth.clientId=... --set oauth.audience=...
Updates #17457
Change-Id: Ib29c85ba97b093c70b002f4f41793ffc02e6c6e9
Signed-off-by: Tom Proctor <tomhjp@users.noreply.github.com>
Now that the feature is in beta, no one should encounter this error.
Updates #cleanup
Change-Id: I69ed3f460b7f28c44da43ce2f552042f980a0420
Signed-off-by: Tom Proctor <tomhjp@users.noreply.github.com>
This starts running the jsontags vet checker on the module.
All existing findings are adding to an allowlist.
Updates tailscale/corp#791
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
The cmd/jsontags is non-idiomatic since it is not a main binary.
Move it to a vet directory, which will eventually contain a vettool binary.
Update tailscale/corp#791
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
Include the node's OS with network flow log information.
Refactor the JSON-length computation to be a bit more precise.
Updates tailscale/corp#33352Fixestailscale/corp#34030
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
Prior to this change a SubscriberFunc treated the call to the subscriber's
function as the completion of delivery. But that means when we are closing the
subscriber, that callback could continue to execute for some time after the
close returns.
For channel-based subscribers that works OK because the close takes effect
before the subscriber ever sees the event. To make the two subscriber types
symmetric, we should also wait for the callback to finish before returning.
This ensures that a Close of the client means the same thing with both kinds of
subscriber.
Updates #17638
Change-Id: I82fd31bcaa4e92fab07981ac0e57e6e3a7d9d60b
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Add options to the eventbus.Bus to plumb in a logger.
Route that logger in to the subscriber machinery, and trigger a log message to
it when a subscriber fails to respond to its delivered events for 5s or more.
The log message includes the package, filename, and line number of the call
site that created the subscription.
Add tests that verify this works.
Updates #17680
Change-Id: I0546516476b1e13e6a9cf79f19db2fe55e56c698
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
In particular on Windows, the `transport.TPMCloser` we get is not safe
for concurrent use. This is especially noticeable because
`tpm.attestationKey.Clone` uses the same open handle as the original
key. So wrap the operations on ak.tpm with a mutex and make a deep copy
with a new connection in Clone.
Updates #15830
Updates #17662
Updates #17644
Signed-off-by: Andrew Lytvynov <awly@tailscale.com>
Specify the app apability that failed the test, instead of the
entire comma-separated list.
Fixes #cleanup
Signed-off-by: Gesa Stupperich <gesa@tailscale.com>
In #17639 we moved the subscription into NewLogger to ensure we would not race
subscribing with shutdown of the eventbus client. Doing so fixed that problem,
but exposed another: As we were only servicing events occasionally when waiting
for the network to come up, we could leave the eventbus to stall in cases where
a number of network deltas arrived later and weren't processed.
To address that, let's separate the concerns: As before, we'll Subscribe early
to avoid conflicts with shutdown; but instead of using the subscriber directly
to determine readiness, we'll keep track of the last-known network state in a
selectable condition that the subscriber updates for us. When we want to wait,
we'll wait on that condition (or until our context ends), ensuring all the
events get processed in a timely manner.
Updates #17638
Updates #15160
Change-Id: I28339a372be4ab24be46e2834a218874c33a0d2d
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Adds a new Redirect field to HTTPHandler for serving HTTP redirects
from the Tailscale serve config. The redirect URL supports template
variables ${HOST} and ${REQUEST_URI} that are resolved per request.
By default, it redirects using HTTP Status 302 (Found). For another
redirect status, like 301 - Moved Permanently, pass the HTTP status
code followed by ':' on Redirect, like: "301:https://tailscale.com"
Updates #11252
Updates #11330
Signed-off-by: Fernando Serboncini <fserb@tailscale.com>
In 3f5c560fd4 I changed to use std net/http's HTTP/2 support,
instead of pulling in x/net/http2.
But I forgot to update DialTLSContext to DialContext, which meant it
was falling back to using the std net.Dialer for its dials, instead
of the passed-in one.
The tests only passed because they were using localhost addresses, so
the std net.Dialer worked. But in prod, where a tsnet Dialer would be
needed, it didn't work, and would time out for 10 seconds before
resorting to the old protocol.
So this fixes the tests to use an isolated in-memory network to prevent
that class of problem in the future. With the test change, the old code
fails and the new code passes.
Thanks to @jasonodonnell for debugging!
Updates #17304
Updates 3f5c560fd4
Change-Id: I3602bafd07dc6548e2c62985af9ac0afb3a0e967
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Single letter 'l' variables can eventually become confusing when
they're rendered in some fonts that make them similar to 1 or I.
Updates #cleanup
Signed-off-by: Fernando Serboncini <fserb@tailscale.com>
A follow-up to #17411. Put AppConnector events into a task queue, as they may
take some time to process. Ensure that the queue is stopped at shutdown so that
cleanup will remain orderly.
Because events are delivered on a separate goroutine, slow processing of an
event does not cause an immediate problem; however, a subscriber that blocks
for a long time will push back on the bus as a whole. See
https://godoc.org/tailscale.com/util/eventbus#hdr-Expected_subscriber_behavior
for more discussion.
Updates #17192
Updates #15160
Change-Id: Ib313cc68aec273daf2b1ad79538266c81ef063e3
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
This migrates an internal tool to open source
so that we can run it on the tailscale.com module as well.
This PR does not yet set up a CI to run this analyzer.
Updates tailscale/corp#791
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
This rewrites the netlog package to support embedding node information in network flow logs.
Some bit of complexity comes in trying to pre-compute the expected size of the log message
after JSON serialization to ensure that we can respect maximum body limits in log uploading.
We also fix a bug in tstun, where we were recording the IP address after SNAT,
which was resulting in non-sensible connection flows being logged.
Updates tailscale/corp#33352
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
This migrates an internal tool to open source
so that we can run it on the tailscale.com module as well.
We add the "util/safediff" also as a dependency of the tool.
This PR does not yet set up a CI to run this analyzer.
Updates tailscale/corp#791
Signed-off-by: Joe Tsai <joetsai@digital-static.net>
Found by staticcheck, the test was calling derphttp.NewClient but not checking
its error result before doing other things to it.
Updates #cleanup
Change-Id: I4ade35a7de7c473571f176e747866bc0ab5774db
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
This reverts commit 4346615d77.
We averted the shutdown race, but will need to service the subscriber even when
we are not waiting for a change so that we do not delay the bus as a whole.
Updates #17638
Change-Id: I5488466ed83f5ad1141c95267f5ae54878a24657
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Drop usage of the branches filter with a single asterisk as this matches
against zero or more characters but not a forward slash, resulting in
PRs to branch names with forwards slashes in them not having these
workflow run against them as expected.
Updates https://github.com/tailscale/corp/issues/33523
Signed-off-by: Mario Minardi <mario@tailscale.com>
Also consolidates variable and header naming and amends the
CLI behavior
* multiple app-caps have to be specified as comma-separated
list
* simple regex-based validation of app capability names is
carried out during flag parsing
Signed-off-by: Gesa Stupperich <gesa@tailscale.com>
Given that we filter based on the usercaps argument now, truncation
should not be necessary anymore.
Updates tailscale/corp/#28372
Signed-off-by: Gesa Stupperich <gesa@tailscale.com>
Brad Fitzpatrick
Jordan Whited
Alex Chan
Andrew Lytvynov
Raj Singh
Tom Proctor
Joe Tsai
M. J. Fromberger
James 'zofrex' Sanderson
Gesa Stupperich
Fernando Serboncini
Erisa A
Harry Harpham
Will Norris
Mario Minardi
srwareham
Max Coulombe