Give every member crate its own `[lints.clippy]` (`pedantic`, `todo`,
`unwrap_used`) and `[lints.rust]` block, add explicit `release` and
`production` profiles to the workspace manifest, and fix the code so that
`just clippy` -- `cargo clippy --workspace --all-targets --all-features
-- -D warnings` -- passes with zero warnings.
Declaring the lints is the easy half and on its own it was actively
harmful: `unsafe_code = "forbid"` cannot be overridden from source, so it
broke `lanparty-gateway`, `lanparty-client-tap`, and
`lanparty-client-route`, which all need OS FFI, and the workspace no
longer compiled on Linux at all. The rest of the workspace produced
roughly 120 denied lints. A lint policy nobody can satisfy is worse than
no policy, so this commit makes the tree actually clean.
Unsafe policy: the three FFI crates use `unsafe_code = "deny"`, and the
one module in each that holds the FFI (`gateway::packet`,
`client_tap::windows`, `client_route::windows`) opts back in with a
documented `#![allow(unsafe_code)]`. Every other crate keeps `"forbid"`.
`deny` was chosen over dropping the lint so that `unsafe` outside those
modules is still a hard error, and over per-block allows because the FFI
is dense enough that per-block attributes would drown the code.
Build profiles: `release` stays optimized but debuggable (debug info,
debug assertions, overflow checks, no LTO, incremental) so that a bug
reproduced at close-to-real speed still panics loudly and gives a usable
backtrace. `production` inherits from it and turns all of that off, adding
fat LTO and a single codegen unit, and is what ships. Note the
consequence: `cargo build --release` binaries are now slower than before
this commit and must not be shipped; use `just build-production`.
Code changes made to satisfy the lints, grouped by kind:
- `# Errors` sections on every public fallible function, and `# Panics`
on `RoomRegistry::new`. The Windows-only implementations and their
non-Windows `bail!` stubs are documented in parallel so the crates stay
clean when built for Windows too, which CI on Linux cannot check.
- Panicking paths removed rather than documented where the panic was only
an unreachable invariant: `OverlayHeader::decode` and
`declared_payload_len` take a fixed-size prefix via `first_chunk`
instead of `try_into().expect(...)`, and `RoomRegistry::join` returns an
`InternalError` reject instead of `expect`ing the room it just inserted.
- Tests keep using `unwrap`/`unwrap_err`: each crate root carries
`#![cfg_attr(test, allow(clippy::unwrap_used))]`. The lint is about
production code; an `expect` message per assertion buys nothing.
- Lossy casts replaced with `try_from` where a real conversion was
happening (datagram-size negotiation, control-frame length prefix,
MTU clamping). In `gateway::packet` the cast lints are allowed
module-wide instead: those casts move values between libc's C types
and ours after the value is already known to fit, so fallible
conversions would only add unreachable branches.
- Signature changes, all on private or crate-internal items except one:
`RoomRegistry::join` now takes `&EndpointHello` (public, it never
consumed the hello), `Room::join` and `reject_control_error` likewise.
`Room`'s `room_id` field is now `id`.
- `#[allow(clippy::too_many_lines)]` with a reason on three end-to-end
test scenarios and on `bridge_until_shutdown`, whose `select!` loop
mutates state shared by every arm; splitting it would hide that.
- Mechanical fixes from `cargo clippy --fix`: `map_or_else`, `let...else`,
inline format args, backticks in doc comments, `Duration::from_mins`.
No runtime behavior changes: the only observable differences are the
error text when a relay room lookup fails immediately after insertion
(previously a panic) and the profile rename described above.
Also documents the lint policy and the profile table in README.md, whose
build section now points at the `just` recipes it should have used all
along.
Test Plan:
- `just clippy` -- passes, zero warnings
- `just test` -- 186 tests pass, 0 failed
- `cargo check --profile production --workspace --all-features` -- passes
- `just fmt` -- clean
- Not verified: the Windows-only code paths and their doc comments, which
need a Windows target to compile.
Refs: https://doc.rust-lang.org/cargo/reference/profiles.html
Refs: https://rust-lang.github.io/rust-clippy/master/index.html#missing_errors_doc
Refs: https://doc.rust-lang.org/reference/attributes/diagnostics.html#lint-check-attributes
lanparty-client-core pulls getrandom from the workspace dependency set for
client identity generation. Bump the direct workspace dependency from 0.3.4 to
0.4.2 and refresh Cargo.lock so builds use the newer release for that direct
dependency.
Cargo still keeps older getrandom versions in the lockfile through transitive
dependencies, so this change does not try to force the whole graph onto one
major line. The added WASI and WIT packages are getrandom 0.4's normal target
support dependencies.
Test Plan:
- cargo clippy, cargo clippy --benches, and cargo clippy --tests before and
after cargo +nightly fmt
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- cargo build --release -p lanparty-relay -p lanparty-gateway
- git diff --check
- git diff --cached --check
Refs: none
PLAN.md describes the first client flow as entering a relay domain and room
code, but the client and gateway CLIs only accepted socket-address literals.
Add a small shared RelayEndpoint parser so bare hosts default to UDP/443 while
IP literals and explicit host:port values stay supported.
The runtime configs still store resolved SocketAddr values. That keeps the
Windows route-pinning path on a concrete relay IP before TAP activation while
avoiding duplicated endpoint grammar between client and gateway. The relay
listen config reuses the same default port constant so UDP/443 has one source.
README examples now use lanparty-relay.local and document the shared endpoint
syntax.
Test Plan:
- cargo fmt --check
- cargo test -p lanparty-net
- cargo test -p lanparty-client-win \
accepts_relay_domain_with_default_port -- --nocapture
- cargo test -p lanparty-gateway \
accepts_iface_alias_for_gateway_interface -- --nocapture
- cargo test -p lanparty-net -p lanparty-client-win -p lanparty-gateway
- cargo clippy -p lanparty-net -p lanparty-client-win -p lanparty-gateway \
--all-targets -- -D warnings
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- git diff --check
Refs: PLAN.md
Add `lanparty-client-route` as the Win32 boundary for route-table work. The
first API is intentionally read-only: `best_route_to` wraps `GetBestRoute2`
and returns the selected source address, next hop, route prefix, interface
index/LUID, and route metric for a relay destination IP.
This keeps the route-protection work separate from the QUIC client binary, so
we can typecheck the Windows IP Helper calls on this Linux host without pulling
in the `ring` build path that currently blocks full Windows binary checks.
Actual route pinning and metric mutation remain later slices.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- cargo check -p lanparty-client-route --target x86_64-pc-windows-msvc
- cargo clippy -p lanparty-client-route --target x86_64-pc-windows-msvc -- -D warnings
- git diff --check
Refs: PLAN.md
The Windows client still needs a real Ethernet TAP boundary before it can pump
frames between the adapter and the relay session. Keep that OS-specific surface
separate from the QUIC client state by adding a `lanparty-client-tap` crate.
The new crate discovers TAP-Windows6 adapters through the Windows network
adapter registry class, validates `tap0901` component ids, constructs the
`\\.\Global\{NetCfgInstanceId}.tap` device path, opens the TAP device handle,
and exposes blocking Ethernet frame read/write helpers. It also wraps the
TAP-Windows IOCTLs for media status, driver MAC, and driver MTU.
This does not wire the TAP crate into `lanparty-client-win` yet and does not
attempt route protection. The value of this slice is the target-checkable OS
boundary that the next client pump can depend on.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- Windows-target cargo check for lanparty-client-tap with clang-cl/lld-link
- Windows-target cargo clippy for lanparty-client-tap with -D warnings
- git diff --check
Refs: PLAN.md Windows TAP client
Remote clients need a stable locally administered MAC address so the relay,
gateway, DHCP lease, and LAN peers keep seeing the same tunnel identity across
runs. Requiring users to pass `--virtual-mac` made that responsibility manual.
Add a platform-neutral client identity store that loads a JSON identity file or
generates a new valid virtual MAC with OS randomness and persists it. The file
stores the MAC in the same string form shown by the CLI. The Windows client now
uses `lanparty-client-identity.json` by default while keeping `--virtual-mac` as
a manual test override.
TAP binding still remains future work; this slice only owns the client identity
that will be assigned to the TAP adapter.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- git diff --check
Refs: PLAN.md MAC identity
The gateway now runs the actual frame bridge after relay admission. It registers
the AF_PACKET socket with Tokio using AsyncFd, reads valid LAN Ethernet frames
and forwards them as relay datagrams, and writes valid relay Ethernet datagrams
back to the LAN socket.
The packet socket is opened nonblocking so the bridge can shut down cleanly on
Ctrl-C without leaving a blocking recv thread behind. Existing send_ethernet and
recv_ethernet helpers now share the same validation and encoding helpers used by
the bridge.
This still needs a privileged LAN-host smoke test with a real wired interface,
but the compile-time and loopback coverage now include the gateway relay side of
the bridge and the non-root-safe packet-socket validation.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
Refs: PLAN.md gateway AF_PACKET to relay bridge loop
The gateway now has a small Linux PacketSocket wrapper for raw Ethernet frame
I/O. It resolves the configured interface with if_nametoindex, opens an
AF_PACKET/SOCK_RAW socket for ETH_P_ALL, binds it to the interface, and exposes
thin send_frame and recv_frame helpers around the owned file descriptor.
The gateway binary opens this socket after completing the relay control
handshake. The frame bridge loop is still intentionally left for a later slice,
but the process now proves the two required resources are available: relay
admission and raw L2 access on the LAN interface.
Tests cover interface-name validation and missing-interface lookup without
requiring root or CAP_NET_RAW.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
Refs: PLAN.md Linux AF_PACKET gateway socket
The relay now keeps a shared room registry behind the QUIC endpoint and
runs an accept loop instead of only binding the socket. Each accepted
connection must open its first bidirectional control stream with a hello
frame; the relay joins the room registry and replies with welcome or reject.
Admission clamps the hello datagram budget to Quinn's negotiated peer
datagram size before choosing the effective room MTU, so room state is based
on what the connection can actually carry. Accepted peers remain present
until the QUIC connection closes, then the relay removes them through the
existing leave cleanup path.
The development self-signed certificate helper now exposes the certificate
to tests so a loopback Quinn client can trust the relay and exercise the real
stream codec path.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
Refs: PLAN.md relay QUIC control-stream startup flow
Make the relay binary bind a real Quinn endpoint instead of only printing its
configuration. This is the next runtime step toward the public relay while still
keeping connection handling out of this commit.
The relay now builds a self-signed development TLS configuration, advertises the
lanparty ALPN, enables QUIC datagram buffers, binds the configured UDP address,
prints the actual local address, and waits for Ctrl-C before closing the
endpoint. The generated certificate is explicitly a development placeholder;
production certificate and client trust handling remain future work.
The rustls dependency is pinned to the ring provider to match Quinn's selected
crypto backend and avoid process-level provider ambiguity at runtime.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- timeout 2s cargo run -p lanparty-relay -- --listen 127.0.0.1:0 || test $? -eq 124
Refs: PLAN.md public relay QUIC data path
Refs: https://docs.rs/quinn/0.11.9
Replace the placeholder relay binary with a typed command-line configuration
entry point. This gives the future QUIC server loop the listen endpoint and room
limit configuration it needs without mixing command parsing into networking or
room-state code.
The relay now accepts --listen as either a socket address or an explicit UDP
shorthand such as 443/udp, defaults to 0.0.0.0:443/udp, and validates that the
per-room client limit is positive. The binary currently reports the parsed
configuration and clearly states that the QUIC server loop is not wired yet, so
this commit does not pretend to provide a working relay.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
- cargo run -p lanparty-relay -- --listen 443/udp
Refs: PLAN.md public relay --listen requirement
Add the framing layer that turns typed control messages into bytes for reliable
QUIC streams. This keeps the codec next to the control schema while leaving the
actual QUIC read/write loops for a later relay/client/gateway slice.
The codec uses a four-byte big-endian length prefix followed by JSON. JSON is a
phase-1 choice for inspectability during manual tunnel bring-up; the explicit
length prefix keeps stream parsing deterministic and the 64 KiB cap prevents a
peer from announcing unbounded control payloads. Decoding validates the message
after deserialization so forged stream bytes cannot bypass constructor checks.
The next networking slice can use complete_control_frame_len to split a stream
buffer and decode_control_frame once a complete frame is available.
Test Plan:
- cargo fmt --check
- cargo test --workspace
- cargo clippy --workspace --all-targets -- -D warnings
Refs: PLAN.md reliable QUIC control stream requirements
ddidderr