diff --git a/.config/nextest.toml b/.config/nextest.toml index 348c63799..6cb716248 100644 --- a/.config/nextest.toml +++ b/.config/nextest.toml @@ -144,10 +144,28 @@ test-group = 'e2e-reliability' # the guard now honours an off-by-default opt-in and this suite's source servers # set it (RUSTFS_REPLICATION_ALLOW_LOOPBACK_TARGET) — so the allowlist below is # restored. +# +# Security negative-auth subset (backlog#1151 sec-5): the three attacker-facing +# S3 auth-rejection suites join the first clause above by module name — +# presigned_negative (sec-2), negative_sigv4 (sec-1, header SigV4), and +# admin_auth (sec-4, admin gate + root-credential lifecycle). All three use +# RustFSTestEnvironment on a random port and are parallel-safe, so they meet the +# smoke admission criteria unchanged. This is the wiring step that makes those +# merged suites actually execute on every PR (they were dead until listed here). +# A rename that drops any of them out of this filter would silently thin the +# security gate with no CI signal, so scripts/check_security_smoke_count.sh owns +# a count-floor guard over exactly this subset (infra-12 mechanism, floor in +# .config/security-smoke-floor.txt), invoked from the e2e-tests job in ci.yml. +# NOT here by topology: the GHSA-3p3x FTPS/WebDAV constant-time e2e +# (protocols::test_protocol_core_suite) binds fixed ports and needs the +# ftps,webdav features, so it cannot join this random-port, default-feature +# profile; its GHSA-r5qv sibling is a unit test that already runs in the +# test-and-lint `--all --exclude e2e_test` pass. See +# docs/testing/security-regressions.md for the full CI-execution map. [profile.e2e-smoke] default-filter = """ package(e2e_test) & ( - test(/^(delete_marker_migration_semantics|version_id_regression|list_objects_v2_pagination|list_object_versions_regression|list_objects_duplicates|list_buckets_double_slash|leading_slash_key|special_chars|create_bucket_region|delete_objects_versioning|head_object_consistency|head_object_range|copy_object_metadata|copy_source_invalid_date|content_encoding|anonymous_access|bucket_policy_check|presigned_negative)_test::/) + test(/^(delete_marker_migration_semantics|version_id_regression|list_objects_v2_pagination|list_object_versions_regression|list_objects_duplicates|list_buckets_double_slash|leading_slash_key|special_chars|create_bucket_region|delete_objects_versioning|head_object_consistency|head_object_range|copy_object_metadata|copy_source_invalid_date|content_encoding|anonymous_access|bucket_policy_check|presigned_negative|negative_sigv4|admin_auth)_test::/) | test(/^replication_extension_test::(test_replication_check_succeeds_with_remote_target|test_replication_check_rejects_target_without_object_lock|test_set_remote_target_rejects_unversioned_source_bucket|test_replication_check_rejects_unversioned_source_bucket|test_replication_check_rejects_missing_replication_config|test_replication_check_rejects_invalid_bucket|test_set_remote_target_rejects_same_bucket_on_same_deployment|test_set_remote_target_rejects_unversioned_target_bucket|test_set_remote_target_update_requires_arn|test_set_remote_target_update_rejects_missing_target|test_set_remote_target_rejects_invalid_target_url|test_set_remote_target_rejects_self_signed_https_target_without_skip_tls_verify|test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_pem|test_list_remote_targets_rejects_empty_bucket|test_list_remote_targets_rejects_invalid_bucket|test_remove_remote_target_rejects_missing_target|test_remove_remote_target_rejects_missing_arn|test_remove_remote_target_rejects_invalid_bucket|test_remove_remote_target_rejects_target_used_by_replication|test_delete_bucket_replication_removes_remote_target)$/) | test(/^reliant::lifecycle::/) ) diff --git a/.config/security-smoke-floor.txt b/.config/security-smoke-floor.txt new file mode 100644 index 000000000..98a35fe08 --- /dev/null +++ b/.config/security-smoke-floor.txt @@ -0,0 +1,12 @@ +# Committed floor for the number of security negative-auth tests selected by the +# e2e-smoke PR profile (see scripts/check_security_smoke_count.sh, backlog#1151 +# sec-5). +# +# The floor equals the exact count of e2e_test cases whose name starts with a +# security module prefix (negative_sigv4_test, presigned_negative_test, +# admin_auth_test) that the [profile.e2e-smoke] default-filter in +# .config/nextest.toml selects, at the time this file was last updated. CI fails +# if the selected count drops below this number, so a rename or removal that +# thins the security smoke gate must update this file in the same PR. +# Adding tests does not require a bump, but bumping keeps the guard tight. +16 diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index 0e3b38fd8..b998266ab 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -435,6 +435,16 @@ jobs: - name: Make binary executable run: chmod +x ./target/debug/rustfs + # Guard the security negative-auth smoke subset (backlog#1151 sec-5) + # against a rename or deletion silently dropping it out of the e2e-smoke + # filter. The script lists what the profile selects and fails if the count + # of security auth-rejection tests falls below the committed floor in + # .config/security-smoke-floor.txt (infra-12 count-floor mechanism). Run + # before the smoke suite so a thinned gate fails fast; the `nextest list` + # here compiles the e2e_test binaries the run below reuses. + - name: Check security smoke subset count floor + run: ./scripts/check_security_smoke_count.sh check + # PR smoke subset of the in-repo e2e suite (backlog#1149 ci-4). The # profile.e2e-smoke default-filter in .config/nextest.toml is the single # wiring mechanism for e2e tests in CI — extend that filter instead of diff --git a/docs/testing/security-regressions.md b/docs/testing/security-regressions.md index 49dd1efc8..f0b5740d5 100644 --- a/docs/testing/security-regressions.md +++ b/docs/testing/security-regressions.md @@ -22,21 +22,45 @@ forced to update its pinned test (red -> green). | [GHSA-r5qv-rc46-hv8q](https://github.com/rustfs/rustfs/security/advisories/GHSA-r5qv-rc46-hv8q) | Internode RPC authentication must fail closed | rustfs/rustfs#4402 | `ghsa_r5qv_resolve_shared_secret_rejects_default_fallback`, `ghsa_r5qv_verify_rpc_signature_fails_closed_on_missing_or_invalid_auth` (`crates/ecstore/src/cluster/rpc/http_auth.rs`) | unit | | [GHSA-m77q-r63m-pj89](https://github.com/rustfs/rustfs/security/advisories/GHSA-m77q-r63m-pj89) | STS JWTs signed with shared root secret (intentionally unfixed) | n/a | `test_created_sts_credentials_authorize_with_session_token_claims` (`crates/iam/src/sys.rs`) — pins current behavior; sec-7 adds GHSA naming | unit | -## Where these run +## Where these run (CI-execution map) -- **Unit tests** (`ghsa_r5qv_*` in `crates/ecstore`) run automatically in the - default CI pass via `cargo nextest run` / `cargo test` — no special wiring. -- **Protocol e2e tests** (WebDAV/FTPS constant-time login rejection) live in the - `e2e_test` protocols suite, which binds **fixed ports** and requires - `--test-threads=1`. Because of the fixed ports they **cannot** join the - `e2e-smoke` nextest profile (parallel, dynamic ports). They run manually today: +Every security regression must land where CI actually runs it — a named test in +an unexecuted suite is theater. The suites split across three execution paths by +topology: + +- **Unit tests** — `ghsa_r5qv_*` (`crates/ecstore`) and the GHSA-m77q STS + pinning (`crates/iam`) run automatically in the default CI pass + (`cargo nextest run --profile ci --all --exclude e2e_test`) — no special + wiring. This is the CI-executed regression for the RPC fail-closed (r5qv) and + STS-signing (m77q) advisories. +- **S3-API negative-auth e2e (e2e-smoke, PR-gated)** — the attacker-facing S3 + auth-rejection suites run on every PR via the `e2e-smoke` nextest profile + (`.config/nextest.toml`), which each spawns its own server on a random port + and is parallel-safe: + - `negative_sigv4_test` — tampered/wrong-key/skewed header SigV4 (sec-1) + - `presigned_negative_test` — expired/tampered/wrong-key presigned URLs (sec-2) + - `admin_auth_test` — non-admin denial + root-credential lifecycle (sec-4) + + A count-floor guard (`scripts/check_security_smoke_count.sh`, floor in + `.config/security-smoke-floor.txt`) runs in the `e2e-tests` CI job and fails if + a rename drops any of these out of the smoke filter (infra-12 mechanism). This + is the sec-5 wiring: those merged suites were dead weight until listed here. +- **Protocol e2e (WebDAV/FTPS constant-time login, GHSA-3p3x)** — lives in the + `e2e_test` protocols suite (`test_protocol_core_suite`), which binds **fixed + ports**, needs the `ftps,webdav` build features, and is `#[serial]`. Those + three properties make it **structurally incompatible** with the random-port, + default-feature, parallel `e2e-smoke` profile: it cannot be a filterset change, + so sec-5 (scoped to a filterset change, global ruling G5) does not wire it. + It runs manually today: ```bash RUSTFS_BUILD_FEATURES=ftps,webdav cargo test --package e2e_test \ test_protocol_core_suite -- --test-threads=1 --nocapture ``` - Wiring the security e2e lane into CI is tracked separately (sec-5, backlog#1151). + **Open gap:** GHSA-3p3x's *e2e* layer has no PR-gated CI execution. A protocols + e2e CI lane is a ci-domain concern (a new profile/job, not a filterset change); + it is deliberately out of sec-5's boundary and left as a follow-up. ## Adding a new advisory guard @@ -44,5 +68,8 @@ forced to update its pinned test (red -> green). 2. Name the test (or the helper/assertion) `ghsa__*`, or attach a `GHSA-` doc comment with the advisory URL and fix PR. 3. Add a row to the table above. -4. Land the unit-level guard where the default CI pass runs it; land any e2e - guard in the protocols suite (and register it with sec-5's filterset seam). +4. Land it where CI runs it (see the map above): a unit guard in the default CI + pass; an S3-API negative-auth e2e in the `e2e-smoke` filter (add the module to + `.config/nextest.toml` and bump `.config/security-smoke-floor.txt`); a + fixed-port protocol e2e in the protocols suite (still manual — see the open + gap above). diff --git a/scripts/check_security_smoke_count.sh b/scripts/check_security_smoke_count.sh new file mode 100755 index 000000000..aa5f6269d --- /dev/null +++ b/scripts/check_security_smoke_count.sh @@ -0,0 +1,85 @@ +#!/usr/bin/env bash +# Count-floor guard for the security negative-auth smoke subset (backlog#1151 +# sec-5), built on the infra-12 mechanism (scripts/check_migration_gate_count.sh, +# backlog#1153). +# +# The e2e-smoke profile in .config/nextest.toml selects tests BY MODULE NAME. +# The three attacker-facing S3 auth-rejection suites — negative_sigv4_test +# (header SigV4, sec-1), presigned_negative_test (presigned URL, sec-2), and +# admin_auth_test (admin gate + root-credential lifecycle, sec-4) — are wired +# into that filter so they run on every PR. A rename that drops a suite out of +# the filter (or a deletion) silently thins the security gate to nothing with no +# CI signal, exactly the failure mode the migration-gate guard was built for. +# +# This script lists what the e2e-smoke profile ACTUALLY selects and counts the +# security suites among them, then fails when that number drops below the +# committed floor in .config/security-smoke-floor.txt. Because it lists via the +# profile (not a standalone name filter), it verifies real smoke MEMBERSHIP: a +# suite renamed without updating the nextest.toml filter stops being selected, +# its tests vanish from the listing, and the count falls below the floor. To +# shrink the subset intentionally, update the floor file in the same PR so the +# reduction is explicit and reviewable (adding tests never requires a bump — +# the floor is a lower bound — but bumping keeps the guard tight). +# +# NAMING CONVENTION DEPENDENCY: the security suites are matched by these +# reserved module-name prefixes: +# +# negative_sigv4_test, presigned_negative_test, admin_auth_test +# +# Renaming a suite away from all of these prefixes removes it from the count. +# Keep this regex in sync with the module names in the e2e-smoke filter. +# +# Usage: +# scripts/check_security_smoke_count.sh # count check (alias of check) +# scripts/check_security_smoke_count.sh check # count check only + +set -euo pipefail + +cd "$(dirname "$0")/.." + +# Anchored module-name prefixes of the security negative-auth smoke subset. +# Kept deliberately in lockstep with the e2e-smoke filter in .config/nextest.toml. +SECURITY_SMOKE_REGEX='^(negative_sigv4|presigned_negative|admin_auth)_test::' +FLOOR_FILE=".config/security-smoke-floor.txt" + +mode="${1:-check}" +case "$mode" in + all | check) ;; + *) + echo "usage: $0 [check]" >&2 + exit 2 + ;; +esac + +floor="$(grep -Ev '^[[:space:]]*(#|$)' "$FLOOR_FILE" | head -n1 | tr -d '[:space:]')" +if ! [[ "$floor" =~ ^[0-9]+$ ]]; then + echo "error: $FLOOR_FILE does not contain a numeric floor (got: '$floor')" >&2 + exit 1 +fi + +# List via the e2e-smoke PROFILE so the default-filter is applied: only the +# tests the PR smoke suite would actually run appear in the output. Count the +# ones whose test name starts with a security module prefix. The structured JSON +# listing is used (not the human format) for the same reason infra-12 does: the +# human format's indentation varies across nextest versions; a JSON schema change +# makes jq fail loudly rather than silently collapsing the count to zero. +count="$(cargo nextest list --profile e2e-smoke -p e2e_test --message-format json \ + | jq --arg re "$SECURITY_SMOKE_REGEX" \ + '[."rust-suites"[].testcases | to_entries[] | select(.key | test($re))] | length')" +if ! [[ "$count" =~ ^[0-9]+$ ]]; then + echo "error: could not parse nextest JSON listing (got count: '$count')" >&2 + exit 1 +fi + +if ((count < floor)); then + echo "error: e2e-smoke selects $count security auth-rejection tests, below the committed floor of $floor." >&2 + echo "" >&2 + echo "The security smoke subset is matched by module name (negative_sigv4_test /" >&2 + echo "presigned_negative_test / admin_auth_test) in the e2e-smoke default-filter" >&2 + echo "(.config/nextest.toml). A rename or removal has dropped one out of the smoke" >&2 + echo "suite. Either restore the module names (and this script's regex), or — if the" >&2 + echo "reduction is intentional — update $FLOOR_FILE in this PR so the" >&2 + echo "security-gate change is explicit and reviewable." >&2 + exit 1 +fi +echo "security smoke count OK: $count selected auth-rejection tests (floor: $floor)"