This reverts commit 6a93b17c8c.
The reverted commit added more complexity than it was worth at the
current stage. Handling delta CapVer changes requires extensive changes
to relayManager datastructures in order to also support delta updates of
relay servers.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
nodeBackend now publishes filter and node changes to eventbus topics
that are consumed by magicsock.Conn
Updates tailscale/corp#27502
Updates tailscale/corp#29543
Signed-off-by: Jordan Whited <jordan@tailscale.com>
This enables us to mark nodes as relay capable or not. We don't actually
do that yet, as we haven't established a relay CapVer.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
It might complete, interrupting it reduces the chances of establishing a
relay path.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
This is simply for consistency with relayManagerInputEvent(), which
should be the sole launcher of runLoop().
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
relayManager can now hand endpoint a relay epAddr for it to consider
as bestAddr.
endpoint and Conn disco ping/pong handling are now VNI-aware.
Updates tailscale/corp#27502
Updates tailscale/corp#29422
Signed-off-by: Jordan Whited <jordan@tailscale.com>
This commit adds a new type to magicsock, epAddr, which largely ends up
replacing netip.AddrPort in packet I/O paths throughout, enabling
Geneve encapsulation over UDP awareness.
The conn.ReceiveFunc for UDP has been revamped to fix and more clearly
distinguish the different classes of packets we expect to receive: naked
STUN binding messages, naked disco, naked WireGuard, Geneve-encapsulated
disco, and Geneve-encapsulated WireGuard.
Prior to this commit, STUN matching logic in the RX path could swallow
a naked WireGuard packet if the keypair index, which is randomly
generated, happened to overlap with a subset of the STUN magic cookie.
Updates tailscale/corp#27502
Updates tailscale/corp#29326
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Fragmented datagrams would be processed instead of being dumped right
away. In reality, thse datagrams would be dropped anyway later so there
should functionally not be any change. Additionally, the feature is off
by default.
Closes#16203
Signed-off-by: Claus Lensbøl <claus@tailscale.com>
Our conn.Bind implementation is updated to make Send() offset-aware for
future VXLAN/Geneve encapsulation support.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
relayManager is responsible for disco ping/pong probing of relay
endpoints once a handshake is complete.
Future work will enable relayManager to set a relay endpoint as the best
UDP path on an endpoint if appropriate.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
This type improves code clarity and reduces the chance of heap alloc as
we pass it as a non-pointer. VNI being a 3-byte value enables us to
track set vs unset via the reserved/unused byte.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
CallMeMaybeVia reception and endpoint allocation have been collapsed to
a single event channel. discoInfo caching for active relay handshakes
is now implemented.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Add new rules to update DNAT rules for Kubernetes operator's
HA ingress where it's expected that rules will be added/removed
frequently (so we don't want to keep old rules around or rewrite
existing rules unnecessarily):
- allow deleting DNAT rules using metadata lookup
- allow inserting DNAT rules if they don't already
exist (using metadata lookup)
Updates tailscale/tailscale#15895
Signed-off-by: Irbe Krumina <irbe@tailscale.com>
Co-authored-by: chaosinthecrd <tom@tmlabs.co.uk>
The event loop removes the need for growing locking complexities and
synchronization. Now we simply use channels. The event loop only runs
while there is active work to do.
relayManager remains no-op inside magicsock for the time being.
endpoints are never 'relayCapable' and therefore endpoint & Conn will
not feed CallMeMaybeVia or allocation events into it.
A number of relayManager events remain unimplemented, e.g.
CallMeMaybeVia reception and relay handshaking.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
relayManager will eventually be responsible for handling the allocation
and handshaking of UDP relay server endpoints.
relay servers are endpoint-independent, and Conn must already maintain
handshake state for all endpoints. This justifies a new data structure
to fill these roles.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Conn.handleDiscoMessage() now makes a distinction between relay
handshake disco messages and peer disco messages.
Updates tailscale/corp#27502
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Conn.sendDiscoMessage() now verifies if the destination disco key is
associated with any known peer(s) in a thread-safe manner.
Updates #15844
Signed-off-by: Jordan Whited <jordan@tailscale.com>
When an event bus is plumbed in, use it to subscribe and react to port mapping
updates instead of using the client's callback mechanism. For now, the callback
remains available as a fallback when an event bus is not provided.
Updates #15160
Change-Id: I026adca44bf6187692ee87ae8ec02641c12f7774
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
In preparation for adding more parameters (and later, moving some away), rework
the portmapper constructor to accept its arguments on a Config struct rather
than positionally.
This is a breaking change to the function signature, but one that is very easy
to update, and a search of GitHub reveals only six instances of usage outside
clones and forks of Tailscale itself, that are not direct copies of the code
fixed up here.
While we could stub in another constructor, I think it is safe to let those
folks do the update in-place, since their usage is already affected by other
changes we can't test for anyway.
Updates #15160
Change-Id: I9f8a5e12b38885074c98894b7376039261b43f43
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
Although, at the moment, we do not yet require an event bus to be present, as
we start to add more pieces we will want to ensure it is always available. Add
a new constructor and replace existing uses of new(tsd.System) throughout.
Update generated files for import changes.
Updates #15160
Change-Id: Ie5460985571ade87b8eac8b416948c7f49f0f64b
Signed-off-by: M. J. Fromberger <fromberger@tailscale.com>
This adds netx.DialFunc, unifying a type we have a bazillion other
places, giving it now a nice short name that's clickable in
editors, etc.
That highlighted that my earlier move (03b47a55c7) of stuff from
nettest into netx moved too much: it also dragged along the memnet
impl, meaning all users of netx.DialFunc who just wanted netx for the
type definition were instead also pulling in all of memnet.
So move the memnet implementation netx.Network into memnet, a package
we already had.
Then use netx.DialFunc in a bunch of places. I'm sure I missed some.
And plenty remain in other repos, to be updated later.
Updates tailscale/corp#27636
Change-Id: I7296cd4591218e8624e214f8c70dab05fb884e95
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
The regular android app constructs its own wgengine with
additional FFI shims, so this default codepath only affects
other handcrafted buids like tsnet, which do not let the
caller customize the innards of wgengine.
Android >=14 forbids the use of netlink sockets, which makes
the standard linux router fail to initialize.
Fixes#9836
Signed-off-by: David Anderson <dave@tailscale.com>
fixestailscale/corp#27506
The source address link selection on sandboxed macOS doesn't deal
with loopback addresses correctly. This adds an explicit check to ensure
we return the loopback interface for loopback addresses instead of the
default empty interface.
Specifically, this allows the dns resolver to route queries to a loopback
IP which is a common tactic for local DNS proxies.
Tested on both macos, macsys and tailscaled. Forwarded requests to
127/8 all bound to lo0.
Signed-off-by: Jonathan Nobels <jonathan@tailscale.com>
This adds a new helper to the netmon package that allows us to
rate-limit log messages, so that they only print once per (major)
LinkChange event. We then use this when constructing the portmapper, so
that we don't keep spamming logs forever on the same network.
Updates #13145
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: I6e7162509148abea674f96efd76be9dffb373ae4
Ensure that the src address for a connection is one of the primary
addresses assigned by Tailscale. Not, for example, a virtual IP address.
Updates #14667
Signed-off-by: Fran Bull <fran@tailscale.com>
If we get a packet in over some DERP and don't otherwise know how to
reply (no known DERP home or UDP endpoint), this makes us use the
DERP connection on which we received the packet to reply. This will
almost always be our own home DERP region.
This is particularly useful for large one-way nodes (such as
hello.ts.net) that don't actively reach out to other nodes, so don't
need to be told the DERP home of peers. They can instead learn the
DERP home upon getting the first connection.
This can also help nodes from a slow or misbehaving control plane.
Updates tailscale/corp#26438
Change-Id: I6241ec92828bf45982e0eb83ad5c7404df5968bc
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
When in tun mode on Linux, AllowedIPs are not automatically added to
netstack because the kernel is responsible for handling subnet routes.
This ensures that virtual IPs are always added to netstack.
When in tun mode, pings were also not being handled, so this adds
explicit support for ping as well.
Fixestailscale/corp#26387
Change-Id: I6af02848bf2572701288125f247d1eaa6f661107
Signed-off-by: Adrian Dewhurst <adrian@tailscale.com>
Cubic performs better than Reno in higher BDP scenarios, and enables the
use of the hystart++ implementation contributed by Coder. This improves
throughput on higher BDP links with a much faster ramp.
gVisor is bumped as well for some fixes related to send queue processing
and RTT tracking.
Updates #9707
Updates #10408
Updates #12393
Updates tailscale/corp#24483
Updates tailscale/corp#25169
Signed-off-by: James Tucker <james@tailscale.com>
Originally identified by Coder and documented in their blog post, this
implementation differs slightly as our link endpoint was introduced for
a different purpose, but the behavior is the same: apply backpressure
rather than dropping packets. This reduces the negative impact of large
packet count bursts substantially. An alternative would be to swell the
size of the channel buffer substantially, however that's largely just
moving where buffering occurs and may lead to reduced signalling back to
lower layer or upstream congestion controls.
Updates #9707
Updates #10408
Updates #12393
Updates tailscale/corp#24483
Updates tailscale/corp#25169
Signed-off-by: James Tucker <james@tailscale.com>
The gVisor RACK implementation appears to perfom badly, particularly in
scenarios with higher BDP. This may have gone poorly noticed as a result
of it being gated on SACK, which is not enabled by default in upstream
gVisor, but itself has a higher positive impact on performance. Both the
RACK and DACK implementations (which are now one) have overlapping
non-completion of tasks in their work streams on the public tracker.
Updates #9707
Signed-off-by: James Tucker <james@tailscale.com>
We had the debug packet capture code + Lua dissector in the CLI + the
iOS app. Now we don't, with tests to lock it in.
As a bonus, tailscale.com/net/packet and tailscale.com/net/flowtrack
no longer appear in the CLI's binary either.
A new build tag ts_omit_capture disables the packet capture code and
was added to build_dist.sh's --extra-small mode.
Updates #12614
Change-Id: I79b0628c0d59911bd4d510c732284d97b0160f10
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Some natc instances have been observed with excessive memory growth,
dominant in gvisor buffers. It is likely that the connection buffers are
sticking around for too long due to the default long segment time, and
uptuned buffer size applied by default in wgengine/netstack. Apply
configurations in natc specifically which are a better match for the
natc use case, most notably a 5s maximum segment lifetime.
Updates tailscale/corp#25169
Signed-off-by: James Tucker <james@tailscale.com>
Remove "unexpected" labelling of PeerGoneReasonNotHere.
A peer being no longer connected to a DERP server
is not an unexpected case and causes confusion in looking at logs.
Fixestailscale/corp#25609
Signed-off-by: Mike O'Driscoll <mikeo@tailscale.com>
This commit intend to provide support for TCP and Web VIP services and also allow user to use Tun
for VIP services if they want to.
The commit includes:
1.Setting TCP intercept function for VIP Services.
2.Update netstack to send packet written from WG to netStack handler for VIP service.
3.Return correct TCP hander for VIP services when netstack acceptTCP.
This commit also includes unit tests for if the local backend setServeConfig would set correct TCP intercept
function and test if a hander gets returned when getting TCPHandlerForDst. The shouldProcessInbound
check is not unit tested since the test result just depends on mocked functions. There should be an integration
test to cover shouldProcessInbound and if the returned TCP handler actually does what the serveConfig says.
Updates tailscale/corp#24604
Signed-off-by: KevinLiang10 <37811973+KevinLiang10@users.noreply.github.com>
This deprecates the old "DERP string" packing a DERP region ID into an
IP:port of 127.3.3.40:$REGION_ID and just uses an integer, like
PeerChange.DERPRegion does.
We still support servers sending the old form; they're converted to
the new form internally right when they're read off the network.
Updates #14636
Change-Id: I9427ec071f02a2c6d75ccb0fcbf0ecff9f19f26f
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
In this PR, we add a generic views.ValuePointer type that can be used as a view for pointers
to basic types and struct types that do not require deep cloning and do not have corresponding
view types. Its Get/GetOk methods return stack-allocated shallow copies of the underlying value.
We then update the cmd/viewer codegen to produce getters that return either concrete views
when available or ValuePointer views when not, for pointer fields in generated view types.
This allows us to avoid unnecessary allocations compared to returning pointers to newly
allocated shallow copies.
Updates #14570
Signed-off-by: Nick Khyl <nickk@tailscale.com>
sync.OnceValue and slices.Compact were both added in Go 1.21.
cmp.Or was added in Go 1.22.
Updates #8632
Updates #11058
Change-Id: I89ba4c404f40188e1f8a9566c8aaa049be377754
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Remove the platform specificity, it is unnecessary complexity.
Deduplicate repeated code as a result of reduced complexity.
Split out error identification code.
Update call-sites and tests.
Updates #14551
Updates tailscale/corp#25648
Signed-off-by: James Tucker <james@tailscale.com>
Observed in the wild some macOS machines gain broken sockets coming out
of sleep (we observe "time jumped", followed by EPIPE on sendto). The
cause of this in the platform is unclear, but the fix is clear: always
rebind if the socket is broken. This can also be created artificially on
Linux via `ss -K`, and other conditions or software on a system could
also lead to the same outcomes.
Updates tailscale/corp#25648
Signed-off-by: James Tucker <james@tailscale.com>
Fixes#14492
-----
Developer Certificate of Origin
Version 1.1
Copyright (C) 2004, 2006 The Linux Foundation and its contributors.
Everyone is permitted to copy and distribute verbatim copies of this
license document, but changing it is not allowed.
Developer's Certificate of Origin 1.1
By making a contribution to this project, I certify that:
(a) The contribution was created in whole or in part by me and I
have the right to submit it under the open source license
indicated in the file; or
(b) The contribution is based upon previous work that, to the best
of my knowledge, is covered under an appropriate open source
license and I have the right under that license to submit that
work with modifications, whether created in whole or in part
by me, under the same open source license (unless I am
permitted to submit under a different license), as indicated
in the file; or
(c) The contribution was provided directly to me by some other
person who certified (a), (b) or (c) and I have not modified
it.
(d) I understand and agree that this project and the contribution
are public and that a record of the contribution (including all
personal information I submit with it, including my sign-off) is
maintained indefinitely and may be redistributed consistent with
this project or the open source license(s) involved.
Change-Id: I6dc1068d34bbfa7477e7b7a56a4325b3868c92e1
Signed-off-by: Marc Paquette <marcphilippaquette@gmail.com>
Importing the ~deprecated golang.org/x/exp/maps as "xmaps" to not
shadow the std "maps" was getting ugly.
And using slices.Collect on an iterator is verbose & allocates more.
So copy (x)maps.Keys+Values into our slicesx package instead.
Updates #cleanup
Updates #12912
Updates #14514 (pulled out of that change)
Change-Id: I5e68d12729934de93cf4a9cd87c367645f86123a
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This provides an interface for a user to force a preferred DERP outcome
for all future netchecks that will take precedence unless the forced
region is unreachable.
The option does not persist and will be lost when the daemon restarts.
Updates tailscale/corp#18997
Updates tailscale/corp#24755
Signed-off-by: James Tucker <james@tailscale.com>
We have several places where LocalBackend instances are created for testing, but they are rarely shut down
when the tests that created them exit.
In this PR, we update newTestLocalBackend and similar functions to use testing.TB.Cleanup(lb.Shutdown)
to ensure LocalBackend instances are properly shut down during test cleanup.
Updates #12687
Signed-off-by: Nick Khyl <nickk@tailscale.com>
Initial support for SrcCaps was added in 5ec01bf but it was not actually
working without this.
Updates #12542
Signed-off-by: Anton Tolchanov <anton@tailscale.com>
This gets close to all of the remaining ones.
Updates #12912
Change-Id: I9c672bbed2654a6c5cab31e0cbece6c107d8c6fa
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
A filesystem was plumbed into netstack in 993acf4475
but hasn't been used since 2d5d6f5403. Remove it.
Noticed while rebasing a Tailscale fork elsewhere.
Updates tailscale/corp#16827
Change-Id: Ib76deeda205ffe912b77a59b9d22853ebff42813
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This required sharing the dropped packet metric between two packages
(tstun and magicsock), so I've moved its definition to util/usermetric.
Updates tailscale/corp#22075
Signed-off-by: Anton Tolchanov <anton@tailscale.com>
The user-facing metrics are intended to track data transmitted at
the overlay network level.
Updates tailscale/corp#22075
Signed-off-by: Anton Tolchanov <anton@tailscale.com>
This adds additional logging on DERP home changes to allow
better troubleshooting.
Updates tailscale/corp#18095
Signed-off-by: Tim Walters <tim@tailscale.com>
While looking at deflaking TestTwoDevicePing/ping_1.0.0.2_via_SendPacket,
there were a bunch of distracting:
WARNING: (non-fatal) nil health.Tracker (being strict in CI): ...
This pacifies those so it's easier to work on actually deflaking the test.
Updates #11762
Updates #11874
Change-Id: I08dcb44511d4996b68d5f1ce5a2619b555a2a773
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
connstats currently increments the packet counter whenever it is called
to store a length of data, however when udp batch sending was introduced
we pass the length for a series of packages, and it is only incremented
ones, making it count wrongly if we are on a platform supporting udp
batches.
Updates tailscale/corp#22075
Signed-off-by: Kristoffer Dalby <kristoffer@tailscale.com>
Like we do for the ones on iOS.
As a bonus, this removes a caller of tsaddr.IsTailscaleIP which we
want to revamp/remove soonish.
Updates #13687
Change-Id: Iab576a0c48e9005c7844ab52a0aba5ba343b750e
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
The AddSNATRuleForDst rule was adding a new rule each time it was called including:
- if a rule already existed
- if a rule matching the destination, but with different desired source already existed
This was causing issues especially for the in-progress egress HA proxies work,
where the rules are now refreshed more frequently, so more redundant rules
were being created.
This change:
- only creates the rule if it doesn't already exist
- if a rule for the same dst, but different source is found, delete it
- also ensures that egress proxies refresh firewall rules
if the node's tailnet IP changes
Updates tailscale/tailscale#13406
Signed-off-by: Irbe Krumina <irbe@tailscale.com>
The new logging in 2dd71e64ac is spammy at shutdown:
Receive func ReceiveIPv6 exiting with error: *net.OpError, read udp [::]:38869: raw-read udp6 [::]:38869: use of closed network connection
Receive func ReceiveIPv4 exiting with error: *net.OpError, read udp 0.0.0.0:36123: raw-read udp4 0.0.0.0:36123: use of closed network connection
Skip it if we're in the process of shutting down.
Updates #10976
Change-Id: I4f6d1c68465557eb9ffe335d43d740e499ba9786
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
* cmd/containerboot,kube,util/linuxfw: configure kube egress proxies to route to 1+ tailnet targets
This commit is first part of the work to allow running multiple
replicas of the Kubernetes operator egress proxies per tailnet service +
to allow exposing multiple tailnet services via each proxy replica.
This expands the existing iptables/nftables-based proxy configuration
mechanism.
A proxy can now be configured to route to one or more tailnet targets
via a (mounted) config file that, for each tailnet target, specifies:
- the target's tailnet IP or FQDN
- mappings of container ports to which cluster workloads will send traffic to
tailnet target ports where the traffic should be forwarded.
Example configfile contents:
{
"some-svc": {"tailnetTarget":{"fqdn":"foo.tailnetxyz.ts.net","ports"{"tcp:4006:80":{"protocol":"tcp","matchPort":4006,"targetPort":80},"tcp:4007:443":{"protocol":"tcp","matchPort":4007,"targetPort":443}}}}
}
A proxy that is configured with this config file will configure firewall rules
to route cluster traffic to the tailnet targets. It will then watch the config file
for updates as well as monitor relevant netmap updates and reconfigure firewall
as needed.
This adds a bunch of new iptables/nftables functionality to make it easier to dynamically update
the firewall rules without needing to restart the proxy Pod as well as to make
it easier to debug/understand the rules:
- for iptables, each portmapping is a DNAT rule with a comment pointing
at the 'service',i.e:
-A PREROUTING ! -i tailscale0 -p tcp -m tcp --dport 4006 -m comment --comment "some-svc:tcp:4006 -> tcp:80" -j DNAT --to-destination 100.64.1.18:80
Additionally there is a SNAT rule for each tailnet target, to mask the source address.
- for nftables, a separate prerouting chain is created for each tailnet target
and all the portmapping rules are placed in that chain. This makes it easier
to look up rules and delete services when no longer needed.
(nftables allows hooking a custom chain to a prerouting hook, so no extra work
is needed to ensure that the rules in the service chains are evaluated).
The next steps will be to get the Kubernetes Operator to generate
the configfile and ensure it is mounted to the relevant proxy nodes.
Updates tailscale/tailscale#13406
Signed-off-by: Irbe Krumina <irbe@tailscale.com>
Like Linux, macOS will reply to sendto(2) with EPERM if the firewall is
currently blocking writes, though this behavior is like Linux
undocumented. This is often caused by a faulting network extension or
content filter from EDR software.
Updates #11710
Updates #12891
Updates #13511
Signed-off-by: James Tucker <james@tailscale.com>
When querying for an exit node suggestion, occasionally it triggers a
new report concurrently with an existing report in progress. Generally,
there should always be a recent report or one in progress, so it is
redundant to start one there, and it causes concurrency issues.
Fixes#12643
Change-Id: I66ab9003972f673e5d4416f40eccd7c6676272a5
Signed-off-by: Adrian Dewhurst <adrian@tailscale.com>
this commit changes usermetrics to be non-global, this is a building
block for correct metrics if a go process runs multiple tsnets or
in tests.
Updates #13420
Updates tailscale/corp#22075
Signed-off-by: Kristoffer Dalby <kristoffer@tailscale.com>
netcheck.Client.GetReport() applies its own deadlines. This 2s deadline
was causing GetReport() to never fall back to HTTPS/ICMP measurements
as it was shorter than netcheck.stunProbeTimeout, leaving no time
for fallbacks.
Updates #13394
Updates #6187
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Disable TCP & UDP GRO if the probe fails.
torvalds/linux@e269d79c7d broke virtio_net
TCP & UDP GRO causing GRO writes to return EINVAL. The bug was then
resolved later in
torvalds/linux@89add40066. The offending
commit was pulled into various LTS releases.
Updates #13041
Signed-off-by: Jordan Whited <jordan@tailscale.com>
Previously, despite what the commit said, we were using a raw IP socket
that was *not* an AF_PACKET socket, and thus was subject to the host
firewall rules. Switch to using a real AF_PACKET socket to actually get
the functionality we want.
Updates #13140
Signed-off-by: Andrew Dunham <andrew@du.nham.ca>
Change-Id: If657daeeda9ab8d967e75a4f049c66e2bca54b78
When the TS_DEBUG_NETSTACK_LOOPBACK_PORT environment variable is set,
netstack will loop back (dnat to addressFamilyLoopback:loopbackPort)
TCP & UDP flows originally destined to localServicesIP:loopbackPort.
localServicesIP is quad-100 or the IPv6 equivalent.
Updates tailscale/corp#22713
Signed-off-by: Jordan Whited <jordan@tailscale.com>
