Sinan Eldem
b93e7b2355
* feat(madmin): add account and two-factor wire contract Defines the self-service account and MFA API shapes in one place so the console and the `rc` CLI decode identical payloads instead of each carrying its own copy of the contract. `AccountMutability` is part of the contract on purpose: a client needs to know whether the server will accept a password change for this identity before offering the control, rather than discovering it from a rejected request. * feat(s3-types): add IAM identity audit events Adds `iam:Identity:CredentialChanged` and `iam:Identity:AuthChallenge` so account and authentication activity reaches the audit pipeline in its own namespace, the way the KMS events already do. Neither is reachable from a bucket notification config. Two variants for the whole surface rather than one per operation: `mask()` gives every variant its own bit in a `u64`, and the budget is nearly spent (63 of 64 used after this). The per-operation detail lives in `AuditEntry::api.name` and the `iamOperation` tag, which is what a SIEM filters on anyway. Splitting these further needs `mask()` widened first. * feat(iam): add two-factor authentication primitives Implements the state machine behind TOTP enrollment and verification in the IAM domain, so the admin handlers stay HTTP plumbing and the console and CLI drive identical logic. * `totp`: RFC 6238 over the workspace's existing hmac/sha1, pinned to the published Appendix B vectors. SHA-1, 6 digits, 30s: the parameters every mainstream authenticator app implements. Verification returns the matched time step so the caller can burn it. * `recovery`: ten single-use codes, 100 bits each, in a Crockford base32 alphabet without I/L/O/U. Stored as domain-separated SHA-256 digests — a password KDF would have to run once per stored code on every attempt, turning each guess into an attacker-controlled cost, and with uniform 100-bit input there is no dictionary for it to defend against. * `challenge`: stateless HMAC tokens. A TTL cache would be node-local, so a cluster without session affinity would issue on one node and verify on another; nothing here needs replicating. * `record`: two-phase enrollment, replay high-water mark, and lockout. Pending enrollment never gates a login, so a mis-scanned QR cannot lock an operator out, and re-configuring keeps the old factor working until the new one is confirmed. * `store`: one object per identity under `config/mfa/`, a sibling of `config/iam/` so the IAM cache loader's startup walk does not sweep it up. Optimistic `If-Match` writes; deliberately uncached, because a cache would need cluster-wide invalidation to keep the replay mark and the lockout counter honest. * `qr`: server-side rendering, so neither client needs a QR encoder. Enrollment is refused without `RUSTFS_IAM_MASTER_KEY`. A TOTP secret is credential-equivalent, and one written in plaintext could be lifted off a disk — worse than no second factor, because the user believes they have one. IAM identities tolerate a missing master key for backward compatibility; a new feature has no such history to honour. Also adds `IamSys::revoke_sts_sessions_for_parent`, so a credential rotation can invalidate the sessions minted under the old secret. * feat(admin): add self-service account endpoints and the two-factor login gate Adds the account surface (`/v3/account/*`), the second-factor endpoints, the administrative reset (`/v3/user/mfa`), and `PUT /v3/set-user-secret-key`, plus the gate on `AssumeRole`. What the gate covers, and what it deliberately does not: * `AssumeRole` is the only interactive login RustFS has, so it is where a second factor can be enforced. With one enrolled it requires `TokenCode`; without an enrollment the code path is unchanged, so existing deployments are untouched. * A request signed directly with a long-term access key stays ungated. Gating it would break every script and CLI the moment a human enabled 2FA on their own account, and would add no protection: whoever holds the secret key already has full access without presenting a code. This is the division AWS draws; making 2FA meaningful for API access needs an `aws:MultiFactorAuthPresent` policy condition, tracked separately. `SerialNumber`/`TokenCode` are STS's own parameters, so an SDK or script authenticates the same way the console does. `caller_identity` resolves who a request acts as. The console signs with a short-lived STS session, so "the caller" is almost never the key that signed. It reports two separate capabilities: root cannot rotate its secret (a process-wide `OnceLock` that also derives the internode RPC secret) but *can* enroll a second factor — conflating the two would leave the default deployment's console login unprotectable. The self-service routes carry no admin action. Giving them one would be wrong in both directions: it would stop an ordinary user from changing their own password, and let any holder of that action change someone else's. They gate on possession of the credential plus, for the mutations, knowledge of the current secret — a signature only proves a credential was used, so without that a hijacked tab could rewrite the account's credentials or strip its second factor. `set-user-secret-key` exists because the only prior way to change a password was to re-POST the whole user through `add-user`, which rewrote `status` and dropped the policy field — a password reset that silently re-enabled a disabled account. Wrong, replayed and malformed codes are indistinguishable on the wire; the distinction survives only in the audit trail, where no submitted value, secret or code is ever recorded. * test(e2e): cover the two-factor lifecycle and its regressions Unit tests cover the state machine at its edges; only an end-to-end test proves the pieces are wired together and that the existing authentication paths still behave. Asserts, against a real server: enrollment is refused without a master key; the full enroll/activate flow works with a genuine RFC 6238 code; `AssumeRole` refuses without a factor and accepts a valid one; a recovery code works exactly once; a direct SigV4 admin request keeps working with a factor enrolled; `AssumeRole` for an unenrolled identity is unchanged; and a password rotation invalidates the old secret. The test computes TOTP codes itself rather than calling the server's implementation — a shared helper could agree with a bug on both sides. This suite caught a real defect during development: enrollment was refused for root because its *password* is immutable, which would have left the default deployment — an administrator signing into the console as root — unable to protect the one login the feature exists for. * docs(operations): document the two-factor authentication model Records what the second factor protects and what it deliberately does not, because several of the boundaries look like gaps until the alternative is spelled out: why direct SigV4 access stays ungated, why root credentials cannot be rotated at runtime, why secret keys cannot be hashed in an S3 server, and why at-rest protection is mandatory for a TOTP secret but optional for an IAM identity. Also states the limitations plainly, including that GHSA-m77q-r63m-pj89 is unaffected: a holder of the root secret can still forge a session token, 2FA claim included. Placed alongside the other authentication and KMS security documents rather than under a new `docs/security/`, which `.gitignore` excludes. * fix(admin): route the new account handlers through the admin s3 facade Two of the guardrails in the CI "Quick Checks" job rejected the previous commits, so the required check would have gone red as soon as a maintainer approved the workflow run. `check_architecture_migration_rules.sh` requires everything under `rustfs/src/admin` to reach `ECStore` through a domain module rather than the root of `storage_api`. The MFA handler and the two `AssumeRole` signatures now use `storage_api::runtime::ECStore`, which is where the other ten admin handlers already take it from. `check_s3s_footprint.sh` ratchets two counters that new code may not grow: files referencing `s3s` and error-macro invocation lines. This branch added four files and thirty-two lines to them. The ratchet is lower-only and its header forbids raising a baseline to get green, so the construction moves behind the facade instead: `storage_api::s3` now re-exports the request and body types these handlers need and gains an `error` constructor over `S3Error::with_message`. That is the same constructor the macro expands to and the one `handlers/mod.rs`, `rebalance_internal_error` and `invalid_object_lock_configuration` already call, so this is the existing practice rather than a new one, and it keeps the `s3s` dependency in the boundary file the s3gate migration replaces. Every error code and message is carried over unchanged. In `sts.rs` only the call site this branch added is converted; the sixteen that predate it are left alone, because rewriting them would put unrelated churn in a feature PR and push the counter below the baseline it is meant to hold.
14 KiB
Two-Factor Authentication
Scope: the self-service account surface (
/rustfs/admin/v3/account/*), the login gate onAssumeRole, and the administrative reset (/rustfs/admin/v3/user/mfa).
This document records what the second factor does and does not protect, and why. The boundaries are deliberate; several of them look like gaps until the alternative is spelled out.
What is protected
TOTP gates session minting: the AssumeRole call that turns a long-term
credential into a short-lived STS session. That is the only interactive login
RustFS has — the Console holds nothing but an STS session, and obtains it by
signing an AssumeRole request with the access key the user typed.
When an identity has an active enrollment:
access key + secret key
│
▼
GET /v3/mfa/challenge (SigV4-signed; answers "is a factor needed?")
│ required: true
▼
POST / Action=AssumeRole
SerialNumber=<challenge>
TokenCode=<6-digit TOTP | recovery code>
│
▼
STS credentials, with the claim x-rustfs-mfa-verified: true
Without a TokenCode, AssumeRole fails with AccessDenied and a message
carrying the MultiFactorAuthRequired marker. Clients match on that marker to
prompt for a code rather than reporting a failed login — the password was
accepted.
SerialNumber and TokenCode are AssumeRole's own parameters, so an SDK or a
script authenticates the same way the Console does, with no RustFS-specific
protocol.
What is deliberately not protected
A request signed directly with a long-term access key is not gated. This is the most important boundary in the design, and it is intentional:
- Gating it would break every script, SDK client and
rcinvocation the moment a human enabled 2FA on their own account. An operator who turned on a security feature would discover it by way of a production outage. - It would add no protection. Whoever holds the secret key already has full access to everything that identity can reach; they never need to present a code, because they never need to mint a session.
This is the same division AWS draws: MFA gates AssumeRole and is enforced for
API calls through the aws:MultiFactorAuthPresent policy condition, not by
refusing signed requests.
Consequence to state plainly: 2FA raises the cost of a stolen password. It does not contain a stolen secret key. Because in RustFS the password is the S3 secret key (see below), those are the same string — so 2FA protects the console login path against credential reuse and phishing, and nothing more, until the policy-condition work lands.
The tracked follow-up is an aws:MultiFactorAuthPresent condition key populated
from the x-rustfs-mfa-verified session claim, which would let an operator write
a policy that denies administrative actions to a session that presented no
second factor. That is the mechanism that makes 2FA meaningful for API access.
OIDC and Keystone sessions are not gated either. Those identities are
authenticated by their provider; a RustFS-side TOTP enrollment would not be
consulted at login and would give a false impression of protection. MFA for a
federated identity belongs to its IdP. CallerIdentity reports such sessions as
FederatedIdentity and refuses enrollment.
Service-account credentials cannot manage their parent's factor. A machine credential must not be able to take over the human identity it was minted from.
Password reality: there is no password hash
RustFS is an S3 server. SigV4 requires the server to know the secret key itself in order to recompute a request signature, so secret keys cannot be hashed — not here, and not in any S3-compatible implementation. The "password" a user types into the Console is their S3 secret key.
What protects it instead:
| Protection | Mechanism |
|---|---|
| At rest | RUSTFS_IAM_MASTER_KEY + encrypt_stream_io (Argon2id → AES-GCM / ChaCha20-Poly1305) |
| Length floor | is_secret_key_valid (SECRET_KEY_MIN_LEN) |
| Rotation | POST /v3/account/password, requiring the current secret |
| Session cleanup | Every STS session minted from the identity is revoked on rotation |
There is deliberately no maximum length. The previous Console capped passwords at 40 characters, which was a client-side invention with no server constraint behind it; capping password length is an anti-pattern.
At-rest protection is mandatory for TOTP secrets
A TOTP secret is credential-equivalent: anyone holding it can mint valid codes
forever. So enrollment is refused when RUSTFS_IAM_MASTER_KEY is not
configured, rather than writing the secret in plaintext:
POST /v3/account/mfa/enroll → 501 NotImplemented
"two-factor authentication requires RUSTFS_IAM_MASTER_KEY to be configured
so the shared secret can be encrypted at rest"
IAM identities tolerate a missing master key for backward compatibility with existing deployments. A new feature has no such history to honour, and a second factor that can be lifted off a disk is worse than none, because the user believes they have one.
GET /v3/account/mfa reports enrollment_available: false with the reason, so
the Console and rc explain the remedy instead of offering a control that fails.
Root credentials cannot be changed at runtime
The root identity comes from RUSTFS_ACCESS_KEY / RUSTFS_SECRET_KEY and lands
in a process-wide OnceLock (crates/credentials/src/credentials.rs). It cannot
be rotated while the server runs, and the account surface reports this as
credentials_source: "env" with mutable.password: false.
This is not merely a missing feature. The root secret key feeds three things:
- STS session token signing (
root_credentials::token_signing_key) — every live session in the cluster is HMAC-signed with it. - The internode RPC secret (
derive_rpc_secret), unlessRUSTFS_RPC_SECRETis set explicitly. - Legacy IAM at-rest decryption for blobs migrated from MinIO.
Rotating it at runtime would therefore invalidate every session cluster-wide and break node-to-node authentication. Making root mutable is a separate piece of work with those three couplings as prerequisites; it is not a side effect of adding a profile page.
Operational recommendation: treat root as a bootstrap identity. Create a
built-in IAM user with the consoleAdmin policy for day-to-day administration.
That identity has a working password change and full 2FA support.
Rate limiting, replay and expiry
| Control | Value | Where |
|---|---|---|
| Failed attempts before lockout | 5 | mfa/record.rs |
| First lockout | 15 minutes, doubling per further run | mfa/record.rs |
| Lockout ceiling | 1 hour | so a sustained attack cannot deny the owner indefinitely |
| TOTP clock skew | ±1 step (±30s) | three codes valid at once, no more |
| TOTP replay | Consumed time step is a high-water mark; step <= last_used is refused |
closes the ~90s window a captured code would otherwise have |
| Recovery code replay | used_at stamp, single use |
|
| Login challenge TTL | 5 minutes | |
| Pending enrollment TTL | 10 minutes | an abandoned enrollment leaves no usable secret |
A wrong code, a replayed code and a malformed code are indistinguishable on
the wire: all three return AccessDenied with the same message. The distinction
survives only in the audit trail, so an operator can tell a guessing attempt from
a replay without an attacker learning that a captured code was genuine.
The lockout is stored in the record and updated under an optimistic compare-and-set, so it holds across the cluster rather than per node.
Storage
.rustfs.sys/config/mfa/<access-key>/totp.json (encrypted with the IAM master key)
A sibling of config/iam/, not a child: the IAM cache loader walks the whole
config/iam/ tree on startup and buckets what it finds by first path segment, so
a new prefix under there would be swept into that walk for no benefit.
Records are not cached. Every verification reads from the store, because a cache would need cluster-wide invalidation to keep the replay mark and the lockout counter honest, and getting that wrong reopens exactly the holes this design closes. Verifications are rare enough that the read is not worth optimising.
Writes are read-modify-write under an If-Match precondition with bounded
retries — the same optimistic scheme the IAM lazy-rewrite path uses. It degrades
to a retry rather than to a distributed lock a crashed node would have to time
out.
Login challenges are stateless
A challenge is HMAC-SHA256(root_secret, "rustfs-mfa-challenge:v1" ‖ access_key ‖ issued_at), base64url-encoded with its payload.
The obvious alternative is a TTL cache, the way the OIDC flow stores its PKCE verifiers. That store is node-local, which is fine for OIDC because the whole authorization round trip returns to the node that started it. A second factor does not: a cluster behind a load balancer without session affinity would issue the challenge on one node and receive the code on another, and a node-local challenge would fail there for reasons no operator could debug.
Statelessness costs nothing, because the challenge is not what makes the exchange single-use — the consumed TOTP time step is.
Authorization model
| Route | Gate |
|---|---|
GET /v3/account/info |
possession of the credential |
POST /v3/account/password |
credential + knowledge of the current secret |
GET /v3/account/mfa |
possession of the credential |
POST /v3/account/mfa/enroll |
credential, and the credential kind must be mutable |
POST /v3/account/mfa/activate |
credential + a valid code from the pending secret |
POST /v3/account/mfa/disable |
credential + a valid code and the account password |
POST /v3/account/mfa/recovery-codes |
credential + a valid code |
GET /v3/mfa/challenge |
possession of the credential |
GET /v3/user/mfa |
admin:GetUser |
DELETE /v3/user/mfa |
admin:EnableUser |
PUT /v3/set-user-secret-key |
admin:CreateUser |
The self-service routes carry no admin action. Giving them one would be wrong
in both directions: it would stop an ordinary user from changing their own
password, and it would let any holder of that action change somebody else's.
They are registered as CredentialOnly in the route-policy matrix.
POST /v3/account/password and POST /v3/account/mfa/disable require a
proof-of-knowledge step because a signature only proves a credential was used.
The Console signs with a short-lived session, so without it a hijacked browser
tab could rewrite the account's credentials or strip its second factor.
Why turning the factor off needs the password too
Requiring only a code would mean a single shoulder-surfed number, in a session someone walked away from, is enough to remove the protection. Requiring the password makes disabling the factor as hard as the thing the factor protects.
Break-glass
DELETE /v3/user/mfa clears another identity's factor, for a user who lost both
their authenticator and their recovery codes. It is gated on admin:EnableUser
rather than a bespoke action, because that is the same capability that can
already re-enable a disabled account — anyone who can do that can already take
the identity over, so a separate action would be a distinction without a security
difference.
The record is deleted outright rather than disabled, so no stale lockout counter survives to block the user's next enrollment. The acting administrator is recorded in the audit entry.
Recovery codes
Ten codes, XXXX-XXXX-XXXX-XXXX-XXXX, 100 bits of uniform randomness each, in a
Crockford base32 alphabet with I, L, O and U removed so a handwritten
code cannot be ambiguous.
Stored as domain-separated SHA-256 digests, not a password KDF. With 100 bits of uniform randomness there is no dictionary to try and no human-chosen pattern to exploit, so the attacks a slow KDF defends against do not apply — while a memory-hard KDF would have to run once per stored code on every verification attempt, turning each guess into an attacker-controlled multiple of that cost. This is the standard treatment for high-entropy bearer tokens, and the same reasoning is why there is no per-code salt.
Codes are returned in plaintext exactly once. Activation always replaces the set: reusing a previous one would leave codes valid for a secret they were never issued against. Disabling clears them, so no live bypass survives a factor the user believes is gone.
Audit
Two EventName variants carry the whole surface:
iam:Identity:CredentialChanged— password rotation, enrollment, activation, disable, recovery-code regeneration, administrative reset.iam:Identity:AuthChallenge— challenge issuance and second-factor verification.
The per-operation detail lives in api.name and the iamOperation tag, which is
what a SIEM filters on. The enum is coarse because EventName::mask() gives every
variant its own bit in a u64 and the budget is nearly spent — 63 of 64 used
after these two. Splitting these per-operation needs mask() widened first.
Redaction: no secret key, TOTP secret, provisioning URI, submitted code or
recovery code enters an audit entry — not even hashed, and not on the failure
paths where the submitted value would be the most tempting thing to record.
Failures are described by a closed set of static strings
(AccountAuditFailure), so no caller-supplied bytes can reach a log target
through this module.
Known limitations
- 2FA does not gate direct SigV4 access. By design; see above. The fix is
the
aws:MultiFactorAuthPresentpolicy condition. - Root cannot rotate its own credentials at runtime. By design; see above.
- GHSA-m77q-r63m-pj89 is unaffected. STS session tokens are signed with the
root secret key, so anyone holding it can still forge a session token —
including one carrying
x-rustfs-mfa-verified. 2FA does not close this; a dedicated STS signing key does, and that advisory is tracked separately. - Username changes are not supported for anyone. The access key is the
primary key for policy mappings, group membership, service-account parents and
bucket-policy principals. A rename is a migration that orphans service
accounts and silently breaks bucket-policy ARNs, not an edit;
mutable.usernameisfalsefor every identity.