Files
marqdo-cli/docs/security-gate.md
T
cflmy 49835424ee docs(spec): adopt RFC-LANG-0002 unified language semantics
Replace the Strong/Emph/List/Quote mapping in place with the single
@exec/@knowledge heading-mode model, unified payload carriers and keyword
control flow.

- add the authoritative RFC, ADR-0052 and a historical ADR index
- rewrite spec/mapping.md and spec/expression.ebnf as the normative
  mapping/statement grammar; update evaluation, references, math,
  foreign, package-format, plugin-abi, security and schema notes
- rewrite README, AGENTS, COMPATIBILITY and CHANGELOG for the breaking
  language change; update the architecture index/overview and active docs
- mark superseded planning docs, compute RFCs, evidence and syntax-role
  ADRs as historical/superseded without rewriting their decisions
- add migration guide, per-file inventory and dev-only audit scripts

The frontend/stdlib migration is completed in the following commits; this
commit contains documentation only.
2026-11-21 12:24:19 +08:00

7.1 KiB

Linux strict-isolation security gate (G-C) — evidence

Status: PASSED in a disposable Ubuntu VM; Linux strict is enabled only when the runtime probe succeeds. Current convergence evidence is for source archive 8af7262 (see evidence/2026-11-20-strict-gc-8af7262.md); the earlier b6fc182 run is retained below as history. No remote self-hosted security-lane CI URL exists yet. Snapshot: workspace 0.1.0, 2026-11-20. See ADR-0021.

Language-syntax migration note. This evidence predates RFC-LANG-0002 (architecture/RFC-LANG-0002-unified-semantics.md): the .mq.md fixtures and the example paths used by the lane are being migrated to the unified syntax, and the lane must be re-run on the migrated tree before it is cited as evidencing the new syntax. The security facts here do not change — @exec is a compile-time projection, not a capability grant; @knowledge produces no IR and therefore no host effect; resource, file, network, stdin/stdout, memory and budget checks remain at the CapabilityHost boundary, and the strict sandbox is unaffected. See STATUS.md for the current red/green state.

This report separates three things that are often conflated:

  1. the mechanism (Bubblewrap namespaces + rlimits + supervisor),
  2. the runtime probe that decides whether strict is available, and
  3. the adversarial evidence that the boundary actually holds.

1. Why the earlier host could not pass

The first development host was a root-in-container environment:

virt: container-other
capsh: Current: = (no effective capabilities), no-new-privs=1
unshare --user/--mount/--pid/--net -> Operation not permitted
bwrap --unshare-user --unshare-net -> No permissions to create a new namespace
cgroup2 -> ro,nosuid,nodev,noexec

That environment cannot verify G-C. Strict stayed Unavailable there; no adapter was faked.

2. Disposable test VM

A QEMU/TCG Ubuntu 26.04 VM was created from the cloud image (no KVM available on the outer host). Inside the VM:

OS:               Ubuntu 26.04.1 LTS, kernel 7.0.0-43-generic x86_64
systemd-detect-virt --container: none
user:             dev (uid 1000), passwordless sudo for test setup only
cgroup2:          rw, nsdelegate, controllers: cpuset cpu io memory hugetlb pids rdma misc dmem
sysctl:           apparmor_restrict_unprivileged_userns = 1
bubblewrap:       0.11.1 with the packaged `bwrap-userns-restrict` profile
unprivileged:     bwrap --unshare-user --unshare-net --ro-bind / / true -> exit 0

Root can create namespaces, and the packaged Bubblewrap AppArmor profile allows the ordinary user to run bwrap without disabling AppArmor globally. The released strict mode runs as this ordinary user.

3. Implemented boundary (ADR-0021)

marqdo run --mode strict:

  • probes Bubblewrap at runtime; failure => Unavailable, run exits 1, no project code executes;
  • re-executes the CLI under bwrap --unshare-all --as-pid-1 --die-with-parent --new-session --cap-drop ALL --clearenv;
  • mounts only runtime libraries, /etc/ld.so.cache, the program and the project tree read-only at /project; private /proc, minimal /dev, tmpfs /tmp;
  • applies hard rlimits: CPU, 512 MiB address space, 64 fds, 64 processes, 8 MiB stack, 0 core, output-byte file size;
  • supervises wall-clock time and kills the process group;
  • the child verifies PID 1, NoNewPrivs: 1, empty CapEff, a non-initial uid map and no non-loopback interface before running project code;
  • rejects Foreign, native/worker plugins, kernels, read/write grants and trace files with exit 2 (strict L1 = pure language + native Math).

4. Adversarial evidence

The helper is tests/security/escape_helper.c, launched through the same marqdo-sandbox supervisor, not through a separate ad-hoc bwrap command. The lane is tests/security/run-lane.sh and the test is crates/marqdo-sandbox/tests/security_linux.rs.

At 8af7262 the same lane was re-run from a clean git archive in the disposable VM as dev uid 1000: 7/7 adversarial tests, positive CLI contract, external-surface refusals, and forged-marker rejection all PASS. The command now honours CARGO_TARGET_DIR; the parallel workspace run also passes all seven security tests.

7/7 tests PASS:
  strict_child_identity_is_unprivileged_and_namespaced
  strict_child_cannot_read_host_files_but_reads_the_project
  strict_child_cannot_write_outside_or_into_the_project
  strict_child_cannot_escape_via_proc_or_privileges
  strict_child_cannot_reach_the_network
  strict_supervisor_enforces_fd_and_process_limits
  strict_supervisor_enforces_memory_cpu_and_output_limits

Observed results inside the sandbox (ordinary user dev):

Case Observation
identity uid=1000, pid=1, NoNewPrivs: 1, CapEff: 0
/etc/shadow, /root, /home/dev DENIED
/project/allowed.txt OK
write /home/dev/..., /project/... DENIED (read-only)
write /tmp allowed (private tmpfs; host unaffected)
/proc/1/root/etc/shadow DENIED; no inherited fds above stdio
setuid(0), mount, unshare DENIED
IPv4 connect 1.1.1.1:80, raw socket DENIED
open /dev/null loop EMFILE at <= 64 fds
fork loop bounded below 200 (user process rlimit)
768 MiB allocation denied/killed by 512 MiB RLIMIT_AS
infinite CPU loop stopped by wall-clock/CPU budget
4 MiB stdout stopped at the 64 KiB output limit

CLI evidence (dev, VM):

run examples/hello.mq.md --mode strict            -> 你好,Marqdo (exit 0)
call factorial 阶乘 --arg n=5 --mode strict       -> 120 (exit 0)
run examples/multi-module/index.mq.md --mode strict -> 12.566... (exit 0)
run --mode strict --json                          -> success=true
run --mode strict --allow-foreign                 -> exit 2
run --mode strict --trust-native-plugin ...       -> exit 2
run --mode strict --kernel ...                    -> exit 2
run --mode strict --trace-file ...                -> exit 2
MARQDO_STRICT_ACTIVE=1 run --mode strict          -> exit 1 (not PID 1)
hide /usr/bin/bwrap; run --mode strict            -> exit 1 (not installed)

5. Explicit non-goals / remaining work

  • Strict L2 (Worker/Foreign inheriting the sandbox) is not implemented; those flags are refused in strict. Native ABI is never allowed in strict.
  • rlimits are used instead of cgroup-v2 accounting; the adapter does not claim cgroup enforcement.
  • Filesystem TOCTOU is reduced by read-only mounts but not formally proven against a racing host user.
  • Network is denied by a network namespace, not a seccomp allowlist.
  • The disposable VM is reproducible from security-vm.md; that document also explains how to register it as a dedicated self-hosted runner.
  • A remote security-lane CI URL is still required before a release calls G-C CI-verified. .github/workflows/security-linux.yml targets a dedicated self-hosted runner; MARQDO_REQUIRE_STRICT_SECURITY=1 makes missing mechanisms fail instead of skip.