octoturge ad55645ec9 Accept an AES-encrypted config bundle from the cloud
The bundle is authored by Continuum Cloud's configurator and downloaded by
the customer, so it travels over the internet and then on a USB stick while
holding the gateway token and possibly a wifi PSK. That makes it a secret at
rest, and confidentiality is the benefit encryption actually delivers here.

WinZip AES-256 ("AE-x"), via the zip crate's `aes-crypto` feature. That
pulls in aes, hmac, pbkdf2, sha1, constant_time_eq and zeroize - all
permissive (MIT/Apache-2.0, and CC0-1.0 OR MIT-0 OR Apache-2.0 for
constant_time_eq), each checked against docs/LICENSES.md, and
`cargo deny check licenses` passes.

**Legacy ZipCrypto is refused outright**, whatever key is configured: it
falls to a known-plaintext attack. The test is on metadata, not on a failed
decrypt - `encrypted()` is set for ZipCrypto and AES alike, but only AES
carries the AE-x extra field, so encrypted-without-it is ZipCrypto and the
whole bundle is rejected. Judged in the first pass, before any decrypt call,
so the crate's ZipCrypto path never runs on our input. Mixed bundles - one
entry encrypted, the other not - are rejected too.

The caps now bound *decrypted* output, so a small ciphertext that inflates
is capped exactly as a plaintext bomb is. Everything else holds unchanged
for an encrypted bundle: zip-slip, the inner config.toml.sha256, the
no-environment validation, atomic install, and a rejected bundle never
touching the installed config.

Failures are distinguished, because "wrong key" and "damaged" need different
answers: no key on the box and a wrong key both say "not for this box", while
tampered or damaged ciphertext says it failed its AES integrity check. A
wrong key passes AES's 16-bit password check 1 time in 65536 and then fails
the HMAC, so that message names a wrong key as a possibility.

**The limitation to be clear about: with plaintext bundles still accepted,
encryption does nothing against tampering.** An attacker supplies a
plaintext bundle and never touches the crypto. The tamper goal only begins to
hold once `accept_unencrypted` is false, which is one line in main.rs and is
deliberately not flipped yet - doing so breaks every stick made before the
configurator encrypts, and every hand-made development stick. That is a
product decision, not a code one.

More generally, encryption is not authenticity: the AES HMAC detects
modification only by someone without the password, and the password has to
live on the box, so extracting it from one image breaks every box sharing it.
It raises the bar from "anyone with a text editor" to "anyone who can read a
firmware image". The module docs say exactly that and do not call it
tamper-proof. `check_authenticity` is a named no-op where an Ed25519 check
over the config, verified against a *public* key in the box image, belongs -
it verifies nothing and says so, so the missing authenticity is a searchable
gap rather than an implicit one.

`load_bundle_key` is the single marked seam where provisioning lands. Its
current `$CONTINUUM_BUNDLE_KEY_FILE` / `/etc/continuum/bundle.key` lookup is
a placeholder, not a decision: shared-vs-per-box keys, where the key lives,
and rotation are open product questions and the mechanism depends on none of
them.

Encrypted fixtures are genuine AES ZIPs written by the zip crate itself, and
one test confirms the config text is absent from the ciphertext. The
ZipCrypto fixture sets the encrypted flag without an AE-x field, which is the
metadata shape this must refuse. Unproven: interop with whatever ZIP writer
the cloud actually uses (only archives written by this same crate version
were read back), AES-128/192, and `load_bundle_key` itself, which reads
process env and so has no test.
2026-10-19 07:56:25 +00:00

continuum-proxy

Edge gateway daemon for the Continuum print farm platform. Runs on a Linux SBC on-site, talks to printers over the LAN, and keeps a connection to the cloud control plane (continuum-backend).

What's real right now

  • Bambu MQTT - a real client for both protocol generations (MQTT v3.1.1+BambuMQTT v1 on older firmware, MQTT v5+BambuMQTT v2 on current firmware), auto-detected per printer on connect (a firmware update can move the same physical printer between them, so this isn't something config can hard-code once and forget - see printer/bambu.rs's BambuProtocolPreference::Auto). Request/reply correlation via sequence_id, an in-memory print.push_status cache that merges v1's partial updates and replaces on v2's full ones, and TLS trust that auto-pins a printer's certificate via trust-on-first-connect the first time the bundled CA doesn't verify (printer/bambu.rs's ensure_trusted) - see printer/bambu_mqtt.rs's doc comment for the full protocol write-up.
  • Marlin/RepRap serial - a real held-open serial connection that paces gcode lines by Marlin's own "ok" acknowledgment (printer/marlin_serial.rs).
  • Wire shapes from schemas, not hand duplication - Bambu's raw MQTT JSON is built straight from continuum-schemas' JSON Schema files via continuum-schema-instance, instead of a second hand-written copy of every request shape drifting out of sync with the schema.
  • Cloud uplink - a WebSocket client to continuum-backend that connects, sends a heartbeat on a timer, and reconnects on drop (src/uplink/).
  • Camera restreaming - generates go2rtc's configuration from the camera URLs the printers report, and restarts go2rtc when it changes (src/camera.rs + src/go2rtc.rs). Only the X1/H2 class publishes an RTSP URL: confirmed against two real machines, an H2C reports rtsps://<ip>:322/streaming/live/1 and a P1P reports a camera with no URL for it. A P1/A1 camera needs Bambu's proprietary port-6000 JPEG stream, which this crate does not speak, so those printers advertise no stream rather than one that will not play. Note the stream is reachable on the gateway's LAN only - there is no relay.
  • Bambu file upload - implicit FTPS (port 990) into the printer's own cache/ directory (src/printer/bambu_ftps.rs, via BambuPrinter::upload_file). Confirmed live against a P1P on 2026-10-03: login, binary mode, and the transfer all work, byte count matched exactly.
  • Print status - print.push_status parsed into PrintReport with real job progress, temperatures, fans and HMS entries. Confirmed live on both generations (idle P1P on v1, H2C 20% into a job on v2); regression fixtures from those captures are in tests/fixtures/.

Examples worth knowing

  • cargo run --example live_reports [printer-id] - read-only, safe against a printer that is mid-print. Connects, reports firmware, and watches the status stream. Redirect stderr to capture raw payloads.
  • cargo run --example bambu_idle_checks -- <printer-id> - writes (FTPS upload, fan speed). Requires an explicit printer id and refuses to run unless that printer reports gcode_state: "IDLE".

What routes between the cloud and the printers

The uplink and the printer fleet are wired together, in both directions (uplink/mod.rs):

  • Up: every 5s, one PrinterTelemetry snapshot per printer that has reported, batched into a single Telemetry message. Verified live against a P1P and an H2C.
  • Down: a Command from the cloud is looked up by our own printer id, executed, and answered with a CommandResult. The set of commands is deliberately small - a fan speed and a single gcode line - and deliberately has no "start a print" variant; see PrinterCommand.

One quirk worth knowing: a Bambu on protocol v1 only sends a complete status when asked, so each telemetry tick also fires a pushing.pushall for the next one. A v1 printer's telemetry is therefore one tick behind, which is why it shows up on the second frame rather than the first.

What's still a stub

start_uploaded_file - the MQTT command that actually starts a print from an uploaded file - is written but has deliberately never been fired, since that commits a machine to hours of work. Its doc comment lists the three details to confirm on the first real attempt. upload_file is proven, but only with a small text file, not a multi-megabyte sliced .3mf.

PrusaLink (REST) and Klipper/Moonraker (WebSocket) - printer/prusa.rs and printer/klipper.rs just println! what they'd send. Also not built yet: local buffering for telemetry across connectivity gaps (so telemetry produced while the uplink is down is simply lost), LAN printer discovery (SSDP/mDNS), and the mechanical plate-changer interface.

Getting started

cp .env.example .env
cargo run                                  # the daemon: connects every configured printer, then runs the cloud uplink + go2rtc watchdog

cp printers.example.toml printers.toml     # fill in your real printers (gitignored)
cargo run --example auto_connect_bambu     # checks/auto-pins each Bambu printer's certificate, no MQTT needed
cargo run --example test_bambu_certs       # narrower: just the raw TLS handshake, no auto-pinning
cargo run --example printer_polymorphism   # standalone trait+enum demo, fake data, no network/env needed
cargo run --example bambu_commands_demo    # round-trips hand-written structs against real captured Bambu JSON

Structure

src/
  main.rs               Loads+connects the printer fleet, then runs the uplink, camera config, watchdog and completion watcher side by side
  config.rs             Loads settings from environment variables
  fleet.rs              Loads printers.toml into ready-to-use PrinterHandles
  uplink/                WebSocket client to continuum-backend: connect, heartbeat, reconnect on drop
  camera.rs             Writes go2rtc's config from the camera URLs the printers report
  go2rtc.rs             Restarts the go2rtc camera-restreaming process if it dies, or when the config changes
  printer/
    mod.rs               GenericPrinter trait + PrinterBase + PrinterHandle enum - the trait+enum
                         "inheritance" pattern this project uses instead of class inheritance
    bambu.rs             The real Bambu printer type: cert trust, protocol version auto-detect, commands
    bambu_mqtt.rs        The real MQTT wiring underneath bambu.rs - see its doc comment for the protocol details
    bambu_commands/      Hand-written serde structs mirroring continuum-schemas' bambulab/ shapes
    marlin_serial.rs     A real held-open Marlin/RepRap serial connection
    prusa.rs, klipper.rs Still stubs - see "What's still a stub" above
examples/
  auto_connect_bambu.rs   Checks/auto-pins every Bambu printer's certificate (no MQTT)
  test_bambu_certs.rs     Narrower: just a raw TLS handshake per Bambu printer
  printer_polymorphism.rs Standalone trait+enum demo, five fake printers, no network needed
  bambu_commands_demo.rs  Round-trips bambu_commands' structs against real captured Bambu JSON

printers.toml (gitignored - copy from printers.example.toml) holds real per-printer connection details: host, access code/API key/serial port depending on vendor, and - for Bambu - an optional per-printer CA certificate override (see certs/README.md; most printers don't need one, since connect() auto-pins automatically).

Learning Rust through this repo

src/printer/mod.rs's doc comment is where the "how do I do inheritance without classes" question lives - trait (shared behavior) + struct (shared fields) + enum (the closed set of concrete types), instead of GenericPrinter -> BambuPrinter/PrusaLinkPrinter/... class inheritance. Run printer_polymorphism to see it work end to end without needing any real printer on the network.

S
Description
Continuum platform: continuum-proxy
Readme
914 KiB
Languages
Rust 100%