Files
freemkv-keysources/CHANGELOG.md
T
Matthew Jackson c820b5ab3f Release 1.6.3
Version bump and changelog for the 1.6.3 sync. All eight crates ship the
same version, so the crates with no functional change this cycle say so
rather than carrying an empty section.
2026-08-09 20:01:33 -07:00

142 lines
5.4 KiB
Markdown

# Changelog
## [1.6.3] — 2026-08-10
### Changed
- **The key service is reached over a newer HTTP client, with the same
protections.** Before contacting a key service this crate resolves the host,
rejects any private or link-local address, and then pins the connection to the
addresses it just checked, so a hostname cannot be re-pointed at an internal
machine in the moment between the check and the request. That behaviour is
unchanged; only the client underneath it moved.
### Security
- **The address pinning is now proven by a test rather than assumed.** The checks
that reject a bad address were well covered, but nothing verified that the
approved addresses were the ones actually connected to — and every existing
test still passed with that wiring removed entirely, because none of them
opened a connection. A connection made to an unreachable name can only
succeed if the pinned address is honoured, and that is now asserted. No
defect is being fixed here: the pinning was correct. It simply had no way of
telling anyone if it stopped being correct.
## [1.6.2] — 2026-08-08
Version sync with the workspace. No functional change in this crate.
## [1.6.1] — 2026-08-07
Version sync with the workspace. No functional change in this crate.
## [1.6.0] — 2026-08-03
### Fixed
- **Key-service failures are no longer indistinguishable from "no key for this
disc".** `query` returned an empty vector for a 502, a 401, a 429, a
transport error and an unparseable body alike — the same value a successful
lookup with no match returns. It now returns a typed error, and the HTTP
status is classified into the operator action it implies: 401/403 the token,
429 back off, any other non-2xx the service. A genuine miss is a 200 with no
key, so every non-2xx is a failure by definition. A DNS failure or timeout
now counts as unreachable rather than as a bad URL.
Version sync with the workspace (freemkv-engine split release). No source change
in this crate; it remains a pluggable AACS key-source provider consumed by the
`freemkv` CLI and libfreemkv's key resolver.
## [1.5.2] — 2026-07-22
Version sync with the workspace; inherits libfreemkv 1.5.2 (CSS DVD descramble
fix). No source change in this crate.
## [1.4.5] — 2026-07-18
Version sync; inherits libfreemkv 1.4.5. `KeySource` split into `get_unit_keys` +
`get_fmts_indexes`, and a keydb device-key parse bug on an uppercase `0X` hex
prefix was fixed (case-insensitive hex parsing across the toolchain).
## [1.4.4] — 2026-07-17
Version sync; inherits libfreemkv 1.4.4. The online `/decode` request is built from
a `DecodeSampleSet` proven sufficient by type rather than a runtime length check.
## [1.4.3] — 2026-07-17
Version sync; inherits libfreemkv 1.4.3. The online unit-key reply is parsed as a
list (one key for an ordinary disc, the ordered set for a forensic-variant disc),
and `MIN_SAMPLE_UNITS` is re-exported from libfreemkv.
## [1.4.2] — 2026-07-15
Version sync with the workspace; inherits libfreemkv 1.4.2. The keydb test that
feeds real key material into the AACS crypto was adapted to the segregated
`decrypt_unit` + `is_clean` primitives (behaviour unchanged).
## [1.4.1] — 2026-07-14
Version sync with the workspace; inherits libfreemkv 1.4.1.
## [1.4.0] — 2026-07-13
Version sync with the workspace; inherits libfreemkv 1.4.0.
## [1.3.2] — 2026-07-10
### Changed
- Unit keys carry libfreemkv's new `UnitKey.variant_number`; every source
(keydb, online, VUK-derived) emits `0` — ordinary, non-forensic content —
via the `UnitKey::new` constructor. No behaviour change. Inherits
**libfreemkv 1.3.2**.
## [1.3.1] — 2026-07-10
### Licensing
- **Relicensed to the MIT License, from 1.3.1 onwards** (releases up to and
including 1.3.0 remain under AGPL-3.0).
Version sync with the workspace; inherits libfreemkv 1.3.1.
## [1.3.0] — 2026-07-08
### Added
- **AACS 2.0 host certs round-trip through `keydb.cfg`.** `to_keydb_cfg` now
emits the sibling `| HC2 |` line — the inverse of the v2 host-cert parser — so
writing a keydb back out no longer silently drops AACS 2.0 host certs.
### Changed
- **Resolve runs directly on `libfreemkv::aacs` primitives.** After libfreemkv
dropped its `aacs::boil` veneer, the resolve path now calls
`derive_media_key_from_{pk,dk}`, `derive_vuk`, and `decrypt_unit_key` from
`aacs::derive` with the `aacs::types` newtypes. No behaviour change.
- Inherits **libfreemkv 1.3.0**.
### Fixed
- **keydb save-validation matches the parser exactly.** A `0x` line counts as a
disc entry only when it also contains ` = `, so validating and persisting
content that parses to zero usable entries (e.g. a stray `0xDEADBEEF` line) can
no longer succeed.
- **Disc-entry titles round-trip verbatim** (parentheses and all) — the parse
path now keeps the title exactly as the emit path writes it.
## [1.2.0] — 2026-06-29
### Changed
- **One hex parser across the toolchain.** Online and keydb hex inputs now parse
through `libfreemkv::hex`, the same parser the library uses — no separate
decoder with its own length/nibble rules.
- **`DiscInputs` carries the disc's AACS version**, and the tests derive the
`Unit_Key_RO` stride from `inputs.version` instead of hardcoding it, so an
AACS-1.0 (V10, 48-byte) and AACS-2.x (V20/V21, 64-byte) disc are each handled
at their own stride.
- **Online MKB read cap aligned with libfreemkv (64 MiB)**, and an over-cap MKB
is logged rather than silently truncated.