Live install on the user's Pixel surfaced a fourth bootloader-lock
prop the source plan never enumerated. The Chunqiu-style detector
card listed three indicators on its bootloader-unlock row and
flagged red because ro.boot.vbmeta.device_state remained unlocked
even after ro.boot.verifiedbootstate, ro.boot.flash.locked, and
ro.boot.veritymode were all spoofed.
Add ro.boot.vbmeta.device_state=locked to BootStateManager.targets.
Verified post-reboot on device: all four props now report the
spoofed values.
The source plan's bulletin-history schema declared a four-value
status enum: success, network_error, parse_error,
validation_rejected. The poller only ever wrote the first three;
when PatchLevelManager.updateTo silently rejected a date for bad
format, floor violation, past/future bounds, or atomicWrite IO
error, the history still recorded status=success and applied=true
because applied was set from isNewer before updateTo ran.
Make updateTo return Boolean. Wire the result through fetchAndParse
so a rejected apply lands as status=validation_rejected with
applied=false and an error string identifying the date that failed.
Closes the only spec gap from fancy-humming-firefly.md uncovered
during the source-plan cross-audit.
currentPatch returned the raw system= value unchanged. A
malformed value such as system=tomorrow flowed into the
lexicographic comparison `date > current`, where any well-formed
YYYY-MM-DD from the bulletin sorts before lowercase letters, so
the poller permanently judged its date "not newer" and never
applied an update. Validate the read value against the date
pattern; on mismatch, log a warning and treat as passive.
The 60-day window covers Pixel monthly bulletin cadence plus
pre-announcement slip but rejects far-future hostile inputs. The
prior bare constant left readers wondering whether the value was
arbitrary; the KDoc closes that loop.
applyToProps reaches Runtime.exec("resetprop", name, value) twice
per call, once for system and once for vendor. Boot-time
initialize, BulletinPoller's handler thread, and the PifObserver
inotify thread can all reach applyToProps independently. Two
concurrent invocations for different dates could interleave such
that system and vendor end up with mismatched values. Synchronize
on the singleton so each apply runs to completion before the
next begins.
A read failure (partial-write race during user edit, SELinux
denial, or any other IOException) used to fall through the
runCatching and rewrite the file with only the global block,
silently destroying every existing [pkg] override. Drop the
runCatching so the failure bubbles to atomicWrite, where the M3
guard in updateTo logs and returns without applyToProps, leaving
both file and props untouched.
isGlobalKeyAssignment treated any bare line whose first token
matched system/boot/vendor/all as a global assignment and
stripped it. A user line of literally "all" or "system" written
without a value (malformed config but reachable) thus got eaten
on the next atomicWrite. Require '=' in the trimmed line before
treating it as a key=value assignment.
parsePatchLevelValue silently synthesized day=01 when input was
6 chars (YYYY-MM) and caller wanted 8-digit YYYYMMDD. If
ro.vendor.build.security_patch ever returns YYYY-MM on a target
device (older Samsung firmware does), the synthesized day
disagrees with the real bulletin day -- a detection fingerprint.
Return null so getRealDevicePatchLevelInt falls through to
Build.VERSION.SECURITY_PATCH, which is always YYYY-MM-DD.
Revert 59dfb2e. AOSP 15 KeyMint reference TA at
system/keymint/common/src/tag/info.rs:61-89 lists both Tag::EcCurve
and Tag::KeySize in KEYMINT_ENFORCED_CHARACTERISTICS, and
check_ec_params at common/src/tag.rs:632 says "Key size is not
needed, but if present should match the curve" -- the TA passes
through whatever the caller supplies and keystore2 supplies both
for EC keys per KeyMintBenchmark.cpp:234,259. Omitting KEY_SIZE
made the simulator's characteristics list shorter than real
hardware, a detection fingerprint.
PatchLevelManager.initialize ran once at boot; PIF edits required
a reboot to take effect. Add a FileObserver on
/data/adb/modules/playintegrityfix for CLOSE_WRITE, MOVED_TO, and
DELETE on the four known PIF filenames, re-resolving and applying
the new date when any of them changes. Skip the watch when the
PIF dir is absent so the daemon does not start a stale inotify
node before the module is even installed.
resolvePifPatch selected the last existing file regardless of
size. A 0-byte file picked up by lastOrNull caused JSONObject("")
to throw, the catch silently fell back to SystemProperties, and a
preceding non-empty PIF was ignored. Filter out zero-length files
so the lookup walks past them to the next valid candidate.
writeText and Files.move can throw IOException, SecurityException,
or AtomicMoveNotSupportedException. The exception previously
propagated through updateTo into BulletinPoller.fetchAndParse's
broad catch, which mislabelled it as "network_error" in the
history. Wrap atomicWrite, log the real failure, and return
before resetprop so the on-disk file and live props stay
consistent on failure.
atomicWrite previously overwrote the entire security_patch.txt
with only the three global lines, destroying the per-package
[pkg] overrides supported by ConfigurationManager. Read the
existing file, strip only global system/boot/vendor/all key
assignments, prepend the refreshed global block, and append
everything else (comments, blanks, all [pkg] sections) verbatim.
PatchLevelManager.initialize previously called updateTo, which
overwrote security_patch.txt with explicit dates and destroyed
the Phase 1 default of system=prop. Split prop application into
a new private applyToProps so initialize only resetprops; never
writes the file. BulletinPoller now treats currentPatch() == null
(the signal for system=prop or missing/blank) as passive and
skips updateTo. The file becomes user-owned config; props track
PIF or the device default.
BootStateManager.apply and PatchLevelManager.initialize ran after
initializeInterceptors, so keystore2 cached ro.boot.* and
ro.build.version.security_patch from the un-spoofed values during
hook init. Move both before the interceptor so the hook sees the
spoofed snapshot. ConfigurationManager stays between them since
it only loads files and is independent of prop state.
BulletinPoller.start ran inside App.main's outer try{...} catch
that rethrows, so any HandlerThread or Looper init failure killed
the daemon including keystore interception. Wrap the start call in
its own try so a poller failure logs and falls through, leaving
the rest of the pipeline alive.
fetchAndParse and appendHistory each catch their own exceptions,
but scheduleNext can throw IllegalStateException if the Looper is
torn down or any helper raises an unanticipated error. Without an
outer catch, the reschedule chain broke and the poller stayed dead
until reboot. Wrap the entire body so a thrown exception still
attempts to schedule the next poll.
HttpsURLConnection resolves bulletin.source via getaddrinfo, which
uses UDP/53 first. Without UDP socket rules the resolver fails
before TCP even attempts, killing BulletinPoller silently on
enforcing SELinux kernels. Mirror the existing TCP rules onto UDP
for ksu and magisk.
PatchLevelManager only rejected dates more than ~1 year in the
past. A MITM serving <td>2099-12-31</td> from a spoofed bulletin
response slipped through validation and got written to
security_patch.txt plus resetprop'd. Add a 60-day upper bound past
today using LocalDate.plusDays so month boundaries are handled
correctly. The existing past bound stays.
Gradle's buildRustCertgen task uses commandLine("cargo", ...), which
ProcessBuilder resolves against the daemon's inherited PATH rather
than the env injected via Exec.environment(). Non-login shells (CI,
IDE-spawned terminals, fresh tmux panes) don't source the profile.d
hook that prepends ~/.cargo/bin, so the daemon dies with
"A problem occurred starting process 'command 'cargo''" even when
rustup is installed. Prepending ~/.cargo/bin at script entry makes
the script self-contained regardless of how the shell was launched.
BulletinPoller fetches the Pixel security bulletin index page on
its own HandlerThread with 5s/30s/2m/10m/30m bootstrap backoff,
then 24h steady cadence. The first <td>YYYY-MM-DD</td> match is
the latest published patch; newer-than-current dates flow through
PatchLevelManager.updateTo for validation + atomic write + resetprop.
Persists the last 10 attempts to last_bulletin_fetch.json (atomic
rename) with status, http_code, parsed_date, applied, and error
fields so operators can audit history without logcat.
Sepolicy rule appends TCP-socket allow rules for both ksu and
magisk source domains so HttpsURLConnection survives SELinux
enforcement on either root provider. Uninstall.sh cleans the
three new artifacts.
PatchLevelManager resolves the active security patch from
PlayIntegrityFix via the same six-path override chain as
Tricky-Addon's get_extra.sh (pif.json/pif.prop/custom.pif.*,
later entries override earlier ones). Falls back to live
ro.build.version.security_patch when no PIF source is present.
updateTo() validates YYYY-MM-DD format, rejects dates below
2020-01-01 or more than one year older than today, then atomically
stages security_patch.txt with explicit system/boot/vendor dates
and resetprops ro.build.version.security_patch plus
ro.vendor.build.security_patch. Cert tags 706/718/719 then encode
consistent dates via AndroidDeviceUtils.parsePatchLevelValue
(YYYYMM for OS, YYYYMMDD for VENDOR/BOOT per AOSP Tag.aidl).
Wired from App.main after BootStateManager.apply().
BootStateManager.apply() runs from App.main after ConfigurationManager
init and sets ro.boot.verifiedbootstate=green, ro.boot.flash.locked=1,
ro.boot.veritymode=enforcing via resetprop so the attestation
extension's hardcoded verifiedBootState=Verified agrees with what
detectors observe via getprop.
Adds an internal AndroidDeviceUtils.setProperty(name, value: String)
overload so the existing private ByteArray variant stays exclusive
to vbmeta digest persistence while config-package callers can set
plain string props without hex encoding.
Closes documented vulnerability D44 (countermeasure-matrix.md).
Out-of-box install seeds /data/adb/tricky_store/security_patch.txt
with system=prop so TEESimulator passively mirrors live device props.
ConfigurationManager auto-forces boot=prop+vendor=prop when system=prop
(ConfigurationManager.kt:253-256), giving full coverage with one line.
Eliminates Chunqiu code 26 (Tampered Attestation Key) on out-of-box
installs without requiring the Tricky-Addon-Update-Target-List
companion module.
Previously module/update.json had to be hand-bumped to keep
versionCode and zipUrl in lockstep with module.prop's expanded
$gitCommitCount. Wire a refreshUpdateJson task to the
prepareModuleFiles${variant} pipeline so every zipDebug/zipRelease
regenerates the file from current verName and gitCommitCount.
Gradle's exec environment does not inherit the user's interactive
shell PATH, so cargo-ndk could not find cargo even when it lived in
~/.cargo/bin. Prepend ~/.cargo/bin to PATH for buildRustCertgen so
the Rust toolchain resolves reliably from any shell.
Java sourceCompatibility/targetCompatibility were already 21, but
the Kotlin compiler defaulted to JVM 17 bytecode, producing a
toolchain skew warning on every build. Align the Kotlin target to
match the Java target.
AOSP keystore2 attestation lists KEY_SIZE only when there is no
authoritative key-shape tag. For EC keys the curve already pins the
key size, so emitting both KEY_SIZE and EC_CURVE is a forgery
fingerprint. Guard the createAuth call accordingly.
Regenerated keys were being filtered as deleted because the
deletion marker in Keystore2Interceptor.deletedSoftwareKeys
survived past the regen call. Clear the marker at both software
and TEE generation paths so the next getKeyEntry returns the
fresh key instead of NOT_FOUND.
Instrumentation-only. Logs entry (primitive class, callingUid, input size)
and throwable propagation (class, SSE error code, message, stack top) in
both SoftwareOperation.updateAad and SoftwareOperationBinder.updateAad.
Intended to distinguish F1 hypotheses H1 (binder swallows SSE) vs H3
(AIDL signature drift) once duck's probe key is routed through our
simulator instead of the real TEE.
The release job now uploads assets with stable names
(TEESimulator-RS-Release.zip) alongside versioned ones, so the
/latest/download/ URL in update.json always resolves. versionCode
in update.json is bumped to the commit count automatically after
each release, committed with [skip ci] to prevent loops.
v6.0 changed bare target.txt entries from AUTO to GENERATE, breaking
apps like BHIM that need TEE-backed attestation keys. Restore AUTO as
default and resolve it at config level (PATCH if TEE works, GENERATE
if not) to bypass the non-deterministic raceTeePatch path.
AOSP ta/src/keys.rs:451-478 requires self-signed leaf (depth 1) when
no attestation challenge is provided. Both Kotlin and Rust paths now
return subject==issuer, signed by generated key, no attestation
extension. Adds cert chain trace logging in debug builds.
createOperation was building effectiveParams from key-generation params
but dropping operation-time fields (nonce, blockMode, padding,
minMacLength). This caused GCM decrypt to fail with
"IV must be specified in GCM mode" since the nonce from the begin call
never reached CipherPrimitive.
Also adds nonce field to KeyMintAttestation and handles GCM/CBC/CTR IV
initialization in CipherPrimitive.
AOSP KeyMint only includes the attestation extension (OID
1.3.6.1.4.1.11129.2.1.17) when ATTESTATION_CHALLENGE is present.
Without a challenge, generateKey produces a plain self-signed cert.
Our code unconditionally added the extension, which behavioral
probes detect by generating a key without a challenge and checking
for the OID.
Fixes both the Rust native-certgen and BouncyCastle paths.
Also skips AAID computation when no challenge is provided,
matching keystore2 security_level.rs:457 behavior.
The retained cert chain was applied to response metadata but the
authorizations array was left unpatched, allowing a detector to compare
metadata patch levels against cert attestation values and spot the
divergence. Refs upstream JingMatrix #164.
PATCH and AUTO modes inherit the real TEE's attestation quirks (epoch 0
cert dates, version mismatch, missing USAGE_COUNT_LIMIT) which can't be
fixed in post-patch. GENERATE mode builds attestation from scratch with
full control over every field. Users who want real TEE key generation
can still use the ? suffix for explicit PATCH mode.
Under binder stress, debug builds hammered logd with 6-7 syscalls per
keygen, causing thread contention that spiked ping latency past G10b's
threshold. Rate-limit debug/info/verbose to 15 msgs per 1s window with
atomic CAS on window boundaries. Warnings and errors always pass.
Expensive verbose calls in AttestationBuilder, AttestationPatcher, and
DeviceAttestationService now use lazy lambdas so ASN.1 formatting only
runs when the message will actually be emitted.
Strip LOGV from the buffer parse loop and add a fast pre-check that
peeks at the first binder command before entering the full parser.
Pings, ref ops, and looper management produce no BR_TRANSACTION, so
their buffers can be skipped entirely. Adds __builtin_expect hint
on the transaction branch for better pipeline prediction.
Drops G2 binder ping ratio from 3.95x to 1.17x in debug builds.
The 256KB native size guard skipped interception entirely for oversized
transactions, causing them to reach the real TEE which returns different
attestation values. This inconsistency is exactly what G10 detects.
Oversized requests now flow through to the Kotlin layer where they hit
doSoftwareKeyGen via the forceGenerate flag. Software gen produces
consistent attestation without forwarding to the real TEE, preserving
the anti-amplification defense that the original guard intended.
B3: AttestationPatcher now accepts optional notBefore/notAfter overrides
so the PATCH path honors CERTIFICATE_NOT_BEFORE instead of inheriting
the real TEE's epoch 0.
C2: getKeymasterVersion delegates to getAttestVersion directly, ensuring
attestationVersion == keymasterVersion regardless of cache source.
F1: Remove incorrect EC+DECRYPT guard in AuthorizeCreate that returned
UNSUPPORTED_PURPOSE instead of INCOMPATIBLE_PURPOSE.
AUTO mode: Replace volatile teeFunctional boolean with AtomicReference
tri-state (null/true/false) so the first race winner locks the path for
all subsequent requests, preventing mixed attestation under concurrency.
AUTO mode now races TEE hardware against software generation via
CompletableFuture. If TEE succeeds, the cert chain is patched and
cached in teeResponses before returning, making attestation
stress-resilient. If TEE fails, software fallback is used.
ConfigurationManager no longer resolves AUTO at config time; it
passes Mode.AUTO through to KeyMintSecurityLevelInterceptor for
runtime dispatch. shouldPatch() returns true for both PATCH and
AUTO modes. TEE status file persistence removed entirely.
Aligns handleGenerateKey with upstream PR #157 three-way dispatch:
forceGenerate, raceTeePatch, or hardware forwarding with post-patch.
Hardware keygen rate limiting removed (replaced by raceTeePatch for
AUTO, plain Continue for PATCH). Attest key override in
Keystore2Interceptor now patches authorizations and uses null-safe
nspace assignment.
BinderInterceptor.onTransact now catches Throwable, preventing any
exception on a binder thread from killing the daemon. Adds a global
uncaught exception handler as defense in depth.
Replace Thread.sleep with TeeLatencySimulator (LockSupport.parkNanos +
statistical delay model) for keygen latency, reducing binder thread
blocking. Move GeneratedKeyPersistence.save to a background executor
to avoid disk I/O on binder threads.
Convert force-unwrap parcel reads to safe calls with early returns in
onPreTransact/onPostTransact hot paths. Add -DNDEBUG to native release
builds to compile out verbose logging from the ioctl hook.
Targets G2 (ping overhead) and G10 (stress attestation consistency).
Extend Rust native cert gen with software-enforced attestation tags
(CALLER_NONCE, ACTIVE_DATETIME, ORIGINATION_EXPIRE_DATETIME,
USAGE_EXPIRE_DATETIME, USAGE_COUNT_LIMIT, UNLOCKED_DEVICE_REQUIRED)
and make NO_AUTH_REQUIRED conditional in teeEnforced. Fixes F5/F6
test failures where these tags were missing from NativeCertGen path.
Add teeResponses cache so PATCH mode keys patched in onPostTransact
return consistent attestation via getKeyEntry. Without this, getKeyEntry
fell through to real keystore2, returning unpatched metadata.
Remove dead Rust enums (KeyPurpose, SecurityLevel, VerifiedBootState)
that were never referenced by the DER encoder.
Upstream removed the unconditional NO_AUTH_REQUIRED from toAuthorizations.
A key generated with auth requirements would incorrectly report
NO_AUTH_REQUIRED in metadata, creating a detectable inconsistency
with the attestation extension.
ConfigurationManager gains checkSELinuxPermission (reads /proc/pid/attr)
and hasPermissionForUid (delegates to IPackageManager.checkPermission)
for AOSP-compliant access control. TeeLatencySimulator provides log-normal
distribution matching real QTEE/Trustonic hardware timing profiles.
Module customize.sh now generates a device-unique hardware-bound key seed
(32 bytes from /dev/random) and clears stale tee_status.txt on install.
SoftwareOperation now throws ServiceSpecificException for all error paths
instead of raw Java exceptions, matching AIDL wire format. updateAad on
non-AEAD operations returns INVALID_TAG (-76) per AOSP operation.rs.
SoftwareOperationBinder methods are @Synchronized to match AOSP Mutex
semantics. GCM encrypt operations return the generated IV in
CreateOperationResponse.parameters.
AuthorizeCreate enforces PURPOSE validation, algorithm-purpose
compatibility (EC rejects ENCRYPT/DECRYPT, RSA rejects AGREE_KEY),
temporal constraints (ACTIVE_DATETIME, ORIGINATION_EXPIRE, USAGE_EXPIRE),
and CALLER_NONCE prohibition. GeneratedKeyInfo carries keyParams for
authorize_create enforcement on software createOperation.
Native binder_interceptor now accepts a filtered_codes vector per
registration, skipping JNI round-trip for non-intercepted transaction
codes. Keystore2Interceptor adds getNumberOfEntries software key counting,
deleteKey KEY_ID domain resolution, patchAuthorizations for OS/VENDOR/BOOT
patch levels, importedKeys tracking to prevent stale attest-key overrides,
and nspace consistency fix in the attest-key override path.
InterceptorUtils gains createServiceSpecificErrorReply for AIDL-compliant
error serialization and patchAuthorizations for authorization array patching.
KeyMintAttestation now carries all 17 enforcement tags that AOSP's
authorize_create and buildKeyDescription paths expect. AttestationBuilder
populates BLOCK_MODE as SET OF INTEGER, gates version-guarded tags
(RSA_OAEP_MGF_DIGEST >=100, ROLLBACK_RESISTANCE >=3, EARLY_BOOT_ONLY >=4),
computes INCLUDE_UNIQUE_ID via HMAC-SHA256 per KeyMint HAL spec, and
gates AAID on challenge presence.
CertificateGenerator uses AOSP cert validity defaults (epoch notBefore,
9999-12-31 notAfter), returns ServiceSpecificException(-75) for missing
keybox, and adds RSA exponent null safety.
trackAndEnforceOpLimit was only called in the Domain.KEY_ID not-found
path, so software-generated keys (found via Domain.APP) bypassed the
STRONGBOX_MAX_CONCURRENT_OPS=4 limit entirely. DuckDetector's concurrent
signing handles test created 24+ operations that all succeeded via LRU
pruning instead of being rejected with TOO_MANY_OPERATIONS (-29).
DuckDetector flags several behavioral anomalies that real TEE/StrongBox
hardware exhibits but our software interceptor did not:
- LRU pruning: cap concurrent ops at 15 (TEE) / 4 (StrongBox) per UID,
aborting oldest when exceeded, matches AOSP keystore2 malus scoring
- StrongBox param guard: forward unsupported params (RSA>2048, non-P256)
to real HAL for proper rejection instead of generating in software
- StrongBox latency floors: 250ms keygen, 80ms sign to match real SE
timing characteristics
- Sliding-window op limit for hardware-generated StrongBox keys that
bypass the software pruning path
- Domain.APP lookup path for createOperation to find software-generated
keys that never reach keystore2's database
Some Android 10 devices (e.g. Sony H8296) report EC private key
algorithm as "ECDSA" instead of "EC", causing IllegalArgumentException
in certificate signing and a SIGSEGV crash in the keystore process.
Closes#4
Symmetric keys (AES/HMAC/3DES) don't have KeyPairs or attestation
certs, routing them through doSoftwareKeyGen crashes with
"Unsupported algorithm: 32". Skip the software path entirely and
let the real HAL handle them.
Also adds CTR block mode, RSA_PKCS1_1_5_SIGN cipher padding, and
RSA_PSS signature padding to JcaAlgorithmMapper.
KeyDetector's OperationErrorPathChecker (flag 0x400000) probes three
error-path behaviors that real keystore2 operations expose. Our
SoftwareOperationBinder was missing all three, plus had no updateAad
implementation which caused AbstractMethodError on Android 16 where
the runtime Stub declares it abstract.
SoftwareOperation changes:
- Add finalized state tracking; post-abort calls now throw
INVALID_OPERATION_HANDLE (-28) matching AOSP operation.rs
- Add input length guard (0x8000) throwing TOO_MUCH_DATA (29)
matching AOSP operation.rs MAX_RECEIVE_DATA
- Add updateAad to CryptoPrimitive interface and SoftwareOperationBinder
- Add KeystoreErrorCodes with runtime reflection + AOSP fallback values
KeyMintSecurityLevelInterceptor changes:
- Infer algorithm from stored key pair when operation params omit
ALGORITHM tag, matching AOSP behavior where createOperation uses
the key's stored algorithm rather than requiring it in op params
Stub addition:
- ServiceSpecificException compile stub (framework-internal class
resolved at runtime on device)
Tested on OnePlus Android 16 (SDK 36), KeyDetector passes all three
probes: updateAad succeeds, TOO_MUCH_DATA returns code=21,
INVALID_OPERATION_HANDLE returns after abort.
PADDING (tag 6) is ENUM_REP in Tag.aidl, meaning SET OF INTEGER in the
attestation extension ASN.1, same as PURPOSE and DIGEST. Commit 1fa6f5a
added it as individual [6] INTEGER entries, causing parsers to fail with
CertificateParsingException on any RSA key attestation.
Fork identity: rename across module metadata, CI pipeline, and build
scripts. Version scheme changed from v4.5-115-f388529 to v4.6-117
format, commit count auto-increments, git hash dropped from filenames.
The old Gaussian sleep (mean=55ms, stddev=12ms) triggered detection on
Chunqiu Native Check 2.8. A flat 15ms floor satisfies the minimum RTT
threshold without creating a detectable delay pattern, both attested
and non-attested paths get identical treatment, keeping the D50 ratio
at ~1.0 while staying above the >=15ms requirement.
signerAlgorithm was derived from params.algorithm (the generated key)
instead of the signing key, causing BouncyCastle to throw when signing
RSA keys with an EC attestation key. Now reads signingKeyPair.private.algorithm.
Device ID tags (serial/imei/meid/secondImei) were blanket-rejected
instead of flowing through to software cert gen like AOSP does.
Narrowed rejection to DEVICE_UNIQUE_ATTESTATION only.
toAuthorizations() was missing OS_VERSION, OS_PATCHLEVEL, VENDOR_PATCHLEVEL,
BOOT_PATCHLEVEL, CREATION_DATETIME, USER_ID, PADDING, and RSA_PUBLIC_EXPONENT
tags that real TEE-generated KeyMetadata always includes. EC_CURVE was also
hardcoded unconditionally, producing invalid authorizations for RSA keys.
Additionally, live-patched certificate chains in getKeyEntry weren't cached,
causing re-patching on every call with potentially different signatures.
Ports upstream JingMatrix/TEESimulator#148 and #150.
The deletion guard must not shadow re-generated keys. If an app
deletes a key then re-creates it, getKeyEntry was still returning
KEY_NOT_FOUND because deletedSoftwareKeys was checked first.
After deleting a software-generated key, getKeyEntry was falling
through to the real keystore2 service which could return a stale
hardware key with the same alias. The post-transact live-patch
fallback would then resurrect the key with a patched chain,
detectors flag this as binder inconsistency.
Track deleted software key aliases and return KEY_NOT_FOUND (7) for
subsequent getKeyEntry calls. Also always invoke cleanupKeyData on
delete to clear stale patchedChains entries for hardware keys.
Software-generated keys complete in ~4ms, real TEE averages 55-65ms
with a floor around 15ms. Detectors measure this RTT to distinguish
software from hardware paths. Gaussian delay sampling (mean=55ms,
σ=12ms, floor=15ms) brings total RTT into the expected range.
Cherry-pick three upstream fixes: Parcel position reset in hasException()
so the method doesn't consume reply data (bab7093), list_past_alias
enumeration filter inversion (71f75de), and KeyMetadata alignment with
AOSP semantics, modificationTimeMs, Tag.ORIGIN, KeyDescriptor
normalization (4e3dcc5).
Additionally, createErrorReply() was missing the empty remote stack
trace header int between the exception message and error code, per
AOSP Status.cpp:196. Binder readers expecting the standard
EX_SERVICE_SPECIFIC wire format would misparse our error replies.
paths-ignore for .github/** was preventing workflow-only pushes
from triggering the pipeline at all. Release job was gated to
push events only, so workflow_dispatch never published. Simplify
paths-ignore to just **.md and allow both push and dispatch to
trigger the release job.
Supervisor had zero-delay restart on crash loops, pins CPU core at
100% if daemon keeps dying. Add exponential backoff (500ms to 30s cap,
resets after 30s stable). Set nice=10 on daemon child to yield CPU
to foreground apps. Evict stale entries from fileLocks and rate limiter
ConcurrentHashMaps that grew unbounded. Upload pre-built flashable zips
in CI instead of unpacking and re-compressing loose files.
debug() was hitting Log.d() unconditionally in release builds,
every intercepted binder transaction triggered string formatting
and logcat syscalls. verbose() already had the guard; debug() was
just missing it. Also remove dead SERVICE_SLEEP_MS constant.
The workflow only uploaded unzipped contents as CI artifacts,
no GitHub release was ever created from CI. Restructured into
build + release jobs: build produces both debug and release ZIPs
(renamed to clean `TEESimulator-vX.Y-{Variant}.zip` format),
release extracts changelog from module/changelog.md and publishes
a GitHub release with both ZIPs attached.
Gradle's buildRustCertgen task requires cargo-ndk and Android NDK
targets to cross-compile libcertgen.so. Without these, CI fails on
any commit after 6aba82e which wired the Rust crate into the pipeline.
AIDL methods use codes 1..0x00ffffff. System transactions like
PING_TRANSACTION (0x5f4e4750) fall above that range. Intercepting
pings forces a full JNI round-trip to Java and back, adding enough
latency for timing detectors to flag the ratio (3.85x vs 3.0x
threshold). Early-return for codes above LAST_CALL_TRANSACTION
eliminates this overhead while preserving all AIDL interception.
Leaf cert Subject CN used "KeyStore" (capital S) but AOSP
KeyGenParameterSpec uses "Keystore" (lowercase s). Fixed in both
the Rust native certgen and BouncyCastle paths.
Replicate keystore2's security_level.rs parameter validation for
software-generated keys: reject CREATION_DATETIME (output-only tag,
ResponseCode 20) and device ID attestation tags (CANNOT_ATTEST_IDS
-66) that real keystore2 blocks before they reach the HAL.
Also fix createErrorReply parcel write order, AIDL protocol expects
exception_code, message, error_code but we had message and error_code
swapped, causing malformed replies for positive error codes.
resetprop overrides for ro.boot.* props don't survive reboots. On
devices where the kernel doesn't set ro.boot.vbmeta.public_key_digest,
the fallback chain hit random generation on every boot, producing a
different RootOfTrust hash each time.
Added file-based persistence (boot_hash.bin, boot_key.bin) as a
fallback layer between TEE cache and random generation. Once a value
is determined from any source, it's written to disk and reused on
subsequent boots.
Verified on Redmi 14C: second boot reads from persistent file instead
of regenerating random bytes.
DuckDetector flagged two issues:
1. Oversized challenge accepted, 256-byte attestation challenge should
return INVALID_INPUT_LENGTH (-21) like real KeyMint. Added early check
in handleGenerateKey before any path decision.
2. Issuer/subject chain mismatch, rcgen's HashMap loses DN attribute
ordering and converts PrintableString to UTF8String, producing
different DER bytes. Replaced rcgen with manual DER assembly that
injects raw keybox issuer_dn_der bytes directly.
Verified on device: TX_ID 315 rejects 256-byte challenge, TX_ID 501
generates valid 4-cert chain with correct issuer linkage.
Enginex0