For A/B OTAs, by default it calls 'openssl pkeyutl' to sign the payload
and metadata with the package private key. If the private key cannot be
accessed directly, a payload signer that knows how to do that should be
supplied via "--payload_signer <signer>".
The signer will be called with "-inkey <path_to_private_key>",
"-in <input_file>" and "-out <output_file>" parameters.
Test: Use a dummy signer, call 'ota_from_target_files.py --payload_signer <signer> <target_files.zip> <ota.zip>' and verify the signatures in the generated package.
Bug: 28701652
Change-Id: I26cfdd3fdba6fc90799221741b75426988e46fd3
(cherry picked from commit dea0f8bfed)
Replace verity keyid with the keyid extracted from cert
passed through --replace_verity_keyid. The veritykeyid in the
BOOT/cmdline of input target files is replaced with keyid
extracted from --replace_verity_keyid and written to the
output target files.
BUG: 28384658
Change-Id: Ic683f36f543c4fcd94b6f95e40f01200fbf45ee1
(cherry picked from commit b58d23fe00)
It replaces the package verification key (change of path due to
system_root_image flag), as well as the payload verification key.
Bug: 29397395
Change-Id: I10435072aaf4356f2d8b5e1b6e82eb9cead7ad62
(cherry picked from commit 24a7206430)
This should make it easier to add parameters/options without breaking
existing clients.
Bug: 27461702
Change-Id: Ia4577f78d703a6b91828dd08492c78d5e9afb110
Now that the source trees all use NATIVE_TESTS for intermediate files
and generated sources, make it a requirement.
Change-Id: Id5718fabe63f6e8dde7981a6f0f5bd89e0ec7ee5
Standardize symlinking /system/vendor -> /vendor for aosp_* devices,
since some /vendor binaries still use /system/vendor/... paths.
Support using a prebuilt vendor image and including it into all the
normal packaging steps.
Bug: 28987532
Change-Id: I27040e8a8d1df0777e16cd1e3c3a9f1b28695e96
This surfaces relevant NoSuchAlgorithmExceptions to the caller instead
of rethrowing as other exception types. Some setups need to be able to
distringuish issues due to their own misconfiguration
(required crypto algorithm mising -- NoSuchAlgorithmException) from
issues with the APK being signed or verified.
Bug: 27461702
Change-Id: I993f73edb29b2cd4cc485734a89a924ec357ef19
This enables verification of APKs which are served to a specific
range of Android platform versions, or to replicate behavior of
particular platform versions.
Bug: 27461702
Change-Id: I44ab4c99419eb97d72c4ccd109137fe1efda577d
We're moving the platform/build repository down a level, then symlinking
the directories and necessary files back into build/. So if we're still
in build/, keep searching for Android.mk files, otherwise stop, since
they'll be found through the symlinks.
Bug: 28001743
Change-Id: Ieea6e3b1fca265b548395c6af148ebb4efa43b0f
There's only a single library, libhost, and it's only used by acp and
atree in build/tools, move it there.
Bug: 28001743
Change-Id: Ie404d2793710de4e265a6fa95d462c32d4042623
This addresses the TODO to mimic the behavior of Android when
verifying APK JAR signatures. Unfortunately, the behavior of Android
kept changing in interesting ways between different platform versions.
This is hard-coded as a big lookup.
Bug: 27461702
Change-Id: I49bc181ee05f774ef8ee041af870385b35212c23
update_engine now accepts POWERWASH=1 to schedule a factory reset in
the post-install phase. Hook up with the --wipe_user_data flag in the
OTA script.
Bug: 28700985
Change-Id: Ie73876a61db90d124d2af588d674757376e9aabc
(cherry picked from commit 38ca0be399)
Modules built against the NDK should only link against modules also
built against the NDK (or link to the NDK prebuilts). This patch
attempts to catch these cases, and prints a large warning when this is
violated. Once the tree is cleaned up, this will change to an error.
Change-Id: Ib6ffcc38d9161abdbe45a58af26ba429fb6f1876
Kati only supports a single use of --no_ignore_dirty, so we were
ignoring the SOONG_ANDROID_MK file, and only detecting changes to
SOONG_MAKEVARS_MK. Fix this by using a pattern that should apply to both
of those makefiles.
Change-Id: I46390f9887f95f1db0efe4e93339667f35ebc67c
Instead of specifying character encoding by name, the faster, cleaner,
and safer way is to use StandardCharsets.UTF_8.
Bug: 27461702
Change-Id: I897284d3ceeb44a21cc74de09a9b25f6aec8c205
For acp, we've been using an old prebuilt in prebuilts/sdk, but it's not
part of the SDK. Instead, we'll use a prebuilt in the build-tools
repository.
For ijar, we've been using the host libstdc++ to workaround the lack of
libc++ on some unbundled branches. Instead, use a prebuilt that can use
libc++.
For ziptime, we've been disabling it on unbundled branches, due to the
lack of libc++. Instead, use a prebuilt version of ziptime that can use
the prebuilt libc++.
Change-Id: If80f845ea06f76e3fe6765964e77c864eaf303d0
This switches PKCS#7 SignedData generation code from Bouncy Castle to
OpenJDK's proprietary internal API. This is to avoid depending on a
huge library that's not really needed. In the longer term, it's best
to add our own implementation of PKCS#7 SignedData building, parsing,
and verification. This will give the code more power to mimic what the
Android platform does.
Bug: 27461702
Change-Id: I29f1ceea1293e35424fde69c0f2969d551345100
