Non-A/B devices need to include the DTBO image
within the recovery partition to be self-sufficient
and prevent OTA failures.
Test: Ran 'm dist' and verified that the DTBO image
was included in recovery.img using unpack_bootimg.
Also ran 'make' and verified that the DTBO image was
included in recovery.img using unpack_bootimg.
Also tested that aosp_angler-userdebug could
boot into recovery(including a fake dtbo.img
in recovery image).
Bug: 74763691
Change-Id: I38c9c395c95d21f4da42cfa646063bd4416f6bd8
The following is a buggy pattern that won't capture anything into err.
The issue is benign, since a failed run would be eventually captured by
a subsequent check.
p = Run(["aapt", ...], stdout=subprocess.PIPE)
output, err = p.communicate()
if err:
raise ...
This CL changes the error detection to be based on the return code from
aapt. It also adds some sanity test to ensure the call to aapt works.
The test app is built from AOSP com.android.cts.ctsshim (chosen mostly
because of its small size).
Test: python -m unittest test_common
Change-Id: I337f141bd0fc5f0801dfc628c601b88b7640789c
For the functions in common.py, capture the stdout/stderr outputs when
shelling out to external tools. Dump the outputs on errors.
Bug: 76123422
Test: Inject errors to signapk.jar / brotli / unzip. Check the outputs.
Change-Id: Ib2d4272528b83a50bb727788cf8a5211e2ddade8
build_image.py used to invoke build_verity_metadata.py with a hard-coded
path of 'system/extras/verity/build_verity_metadata.py', which makes it
hard to run unittests from non-$(ANDROID_BUILD_TOP) directory.
This CL adds the dependency on the tool, so that it gets installed to
$(HOST_OUT_EXECUTABLES), then removes the hard-coded path.
Bug: 74544459
Bug: 76015688
Test: `m dist`
Test: python -m unittest test_validate_target_files
Change-Id: I0dcf4eb067a0db6f099cb589eb99a151a05c7f2b
This CL handles a path that uses system-root-image on non-A/B device.
For this path, we can't generate recovery-from-boot patch with imgdiff,
because boot/recovery images contain different number of entries (only
recovery iamge has ramdisk image now).
Using BOARD_USES_FULL_RECOVERY_IMAGE can work around the issue, at the
cost of extra size. Compared to carrying full recovery image, this CL
saves the cost of the kernel size, by putting a patch that's roughly the
size of the recovery ramdisk.
The applypatch executable already detects and handles a bsdiff patch
automatically. No change required to that end.
Note that it won't further reduce the patch size by handling that
ramdisk entry specially, because (a) that's the only difference between
the two images; and (b) there's no corresponding data in boot image to
be diff'd against.
Bug: 72731506
Test: `m dist` with aosp_angler-userdebug. Check the device can install
recovery image successfully (after intentionally corrupting the
recovery image).
Test: Build aosp_angler-userdebug with BOARD_BUILD_SYSTEM_ROOT_IMAGE set.
Verify the generated patch files.
Test: Run validate_target_files.py with the target_files.zips.
Change-Id: I69c06f51ba8c39ae059c5e9a6872a9f10600cf17
In FinalizeMetadata and PropertyFiles, we need to reserve space between
the calls to Compute() and Finalize(). We used to put a 10-byte
placeholder, in the hope of covering the 'offset:length' space for the
metadata entry, as well as the possible value changes in other entries.
However, this could fail in two possible cases: (a) metadata entry
itself has a large offset (e.g. staying near the end of a 1-GiB package,
where the offset itself has 10-digit); or (b) the offsets for other
entries change substantially due to entry reordering. Note that for case
(b), it's space inefficient to always reserve 15-byte for _each_ token
in the property-files.
This CL handles both of these two cases. For (a), we bump up the 10-byte
to 15-byte, which is large enough to cover a package size up to 10-digit
number (i.e. ~9GiB) with a metadata entry size of 4-digit. All these
15-byte will be used for the metadata token alone.
For (b), we add a fallback flow that would retry one more time, but
based on the already signed package that has entries in desired order.
Bug: 74210298
Test: python -m unittest test_ota_from_target_files
Test: Generate aosp-bullhead full OTA with '--no_signing' flag.
Change-Id: If20487602d2ad09b3797465c01972f2fa792a1f1
This CL exposes ota-property-files flag for non-A/B OTA packages.
Currently the line only contains the info for the METADATA entry, for
example "ota-property-files=metadata:69:286". This allows system updater
to just download the METADATA entry, as opposed to downloading the
entire package, to learn about the info regarding the OTA package (e.g.
post-OTA build fingerprint). Note that this requires the OTA server-side
support to pass down the flag along with the update URL.
Bug: 74210298
Test: python -m unittest test_ota_from_target_files
Test: Generate a non-A/B package and check the property-files string.
Change-Id: I1482c587e18ea7101c8328777ea988c2d8ca06ac
AbOtaPropertyFiles writes 'ota-property-files' into package metadata.
Comparing to StreamingPropertyFiles, AbOtaPropertyFiles additionally
writes a token of 'payload-metadata.bin' for a virtual entry that's head
of the 'payload.bin'. This entry can be used by the system updater to
verify if a payload is applicable to the underlying device.
AbOtaPropertyFiles intends to replace StreamingPropertyFiles, as it
covers all the info of the latter. We keep them both in P, and will
remove the latter in later release.
Bug: 74210298
Test: python -m unittest test_ota_from_target_files
Test: Generate full and incremental A/B OTAs. Check the property-files
in the METADATA entry.
Change-Id: If5872c8ee0a97a8a95b07d4c69b463aa3fe1b9b9
And move StreamingPropertyFiles as its subclass. We will need similar
PropertyFiles instance for non-A/B OTA as well (to expose the
offset/size for the METADATA entry).
Bug: 74210298
Test: python -m unittest test_ota_from_target_files
Test: Generate an A/B OTA. Check the generated property-files string.
Test: pylint --rcfile=pylintrc \
ota_from_target_files.py \
test_ota_from_target_files.py
Change-Id: If90d97f0b330749fd8a6cde2ed9d0d6cd6ea60a8
For a given (signed) target-files.zip, this CLs allows verifying the
Verified Boot related images. It works with both of VB 1.0 and VB 2.0
images.
As part of the CL, it also moves validate_target_files.py to argparse,
which is more flexible than the traditional getopt module.
Also add unittests for the VB 1.0 path. VB 2.0 tests will be added in
follow-up CL.
Example usage:
- Run the script on aosp_bullhead target-files.zip.
$ ./build/make/tools/releasetools/validate_target_files.py \
--verity_key build/target/product/security/verity.x509.pem \
--verity_key_mincrypt build/target/product/security/verity_key \
aosp_bullhead-target_files-4522605.zip
- Run the script on aosp_walleye target-files.zip.
$ ./build/make/tools/releasetools/validate_target_files.py \
--verity_key external/avb/test/data/testkey_rsa4096.pem \
aosp_walleye-target_files-4627254.zip
Bug: 63706333
Bug: 65486807
Test: Run validate_target_files.py on target_files.zip files.
Test: PYTHONPATH=build/make/tools/releasetools python -m unittest \
test_validate_target_files
Change-Id: I170f14d5828d15f3687d8af0a89a816968069057
This CL breaks down ComputeStreamingMetadata() into mutiple member
functions of StreamingPropertyFiles class, which correspond to the
two-pass logic when generating streaming property files (aka streaming
metadata).
StreamingPropertyFiles.Compute() does the work for the first pass, by
putting placeholders before doing initial signing. Finalize()
corresponds to the second pass, where the placeholders get replaced with
actual data. Verify() can be optionally called to assert the correctness
of the work.
The separation between Compute() and Finalize() is to allow having
multiple StreamingPropertyFiles instances (in coming up CLs). This way
we can call Compute() multiple times for each instance, followed by only
one call to SignOutput(). And similarly for Finalize().
Bug: 74210298
Test: Generate an A/B OTA package. Check the METADATA entry.
Test: python -m unittest test_ota_from_target_files
Change-Id: I45be0372a4863c4405e6d8e20bcb9ccdc29e7e11
The file was broken due to earlier touches:
- Missing 'import zipfile';
- Mismatching arguments when calling GetSparseImage().
Bug: 73996151
Test: Run validate_target_files.py with a walleye-target_files.zip.
Test: pylint --rcfile=pylintrc validate_target_files.py
Change-Id: I3692bd51fb27a3da698e06b75155e84502549f66
And add unittests for ComputeStreamingMetadata().
This prepares for the changes that add additional property-files (for
both of A/B and non-A/B).
Bug: 74210298
Bug: 72751683
Test: python -m unittest test_ota_from_target_files
Test: Generate A/B OTA package. Check the ota-streaming-property-files
in the METADATA entry.
Change-Id: Ib4b069f61c2c06c035c0cff73a55112f3936b969
There's a mismatch in WriteABOTAPackageWithBrilloScript().
temp_zip_file = tempfile.NamedTemporaryFile()
...
common.ZipClose(temp_zip_file)
It's benign since common.ZipClose() happens to be calling
"temp_zip_file.close()". This CL moves the use of tempfile to
common.MakeTempFile(), so that the tempfile will be cleaned up
automatically as part of the call to common.Cleanup(). (Not fixing the
close() directly, since the nearby lines will be refactored into another
function shortly.)
Also remove one assert in the same function, which trivially holds in
the current code.
Test: Generate an A/B OTA.
Change-Id: I53b375d1150820de718dec0ead55abf5f4951071
Also minor cleanups to make it pylint clean.
Test: Run check_target_files_signatures.py with a target-files.zip.
Test: pylint --rcfile=pylintrc check_target_files_signatures.py
Change-Id: Ife3b54c7805c2f2562e87e91ab4b4de355782012
In addition to the unzipping work, common.UnzipTemp() kindly bundles an
open ZipFile object as part of the return value. It doesn't look very
helpful to the callers though. It also looks less obvious that the
caller needs to properly close the handle (missing the close here is
benign though). This CL just removes the ZipFile object out of the
return value, and leaves the work to callers.
Test: `m dist` on both of A/B and non-A/B target.
Test: python -m unittest test_add_img_to_target_files
Test: python -m unittest test_common
Test: python -m unittest test_ota_from_target_files
Test: Check the callers to common.UnzipTemp() in code search.
Change-Id: Id47da3fd42a0e76d6ae8851f05780db319ee48cf
This should be the last case to be moved over.
Test: Generate an incremental BBOTA (which exercises the changed code).
Test: `rgrep mkdtemp` gives no more instance.
Change-Id: I76db069476201cdfaf3a2de9d9635dfe54507f7a
... from the following functions in add_img_to_target_files.py.
AddSystem
AddSystemOther
AddVendor
AddProduct
AddDtbo
AddUserdata
AddVBMeta
AddPartitionTable
AddCache
The last user of the parameter in img_from_target_files.py has been
removed in commit 2bb109709a (in O).
Test: pylint --rcfile=pylintrc add_img_to_target_files.py
Test: Check all the callers to the above functions.
Test: m dist
Change-Id: I551d1683def8f8535062fc90f68dafa0f4252822
The file has been removed from target-files.zip since commit
c19a8d5590 (Gingerbread), whose info has
been consolidated into META/misc_info.txt.
Test: `m dist`
Change-Id: Ic144457954f5742ea082dcd9ffbea71df4afe46e
With test_utils.construct_sparse_image(), creating valid sparse image is
no longer a blocker for testing these functions.
Test: python -m unittest test_add_img_to_target_files
Change-Id: Iff1f15837cc744bad52e2120f9a9ad94d7db50d5
This CL adds a new flag '--skip_postinstall' that allows skipping all
the postinstall hooks when generating an A/B OTA package (default:
False). Note that this discards ALL the hooks, including non-optional
ones. Should only be used if caller knows it's safe to do so (e.g. all
the postinstall work is to dexopt apps and a data wipe will happen
immediately after).
Bug: 73547992
Test: python -m unittest test_ota_from_target_files
Test: Generate a full OTA package for walleye. Examine the generated
payload.
Change-Id: Ifc069e897b4019605051eabfd221230a6a37867c
Since we have been carrying test certificates in testdata/ for other
tests, do the same for test_sign_target_files_apks.py. Copy
verity.x509.pem from build/target/product/security/ to testdata/ for
that purpose.
Also capture the stderr output in ReplaceVerityKeyId().
Test: python -m unittest test_sign_target_files_apks
Change-Id: Ie11e042086952e8a4a5a63950cb0b16cc436b7e6
Also minor clean up to the function, to suppress the following lint
warning.
R:1813, 2: Redefinition of cert type from list to unicode (redefined-variable-type)
Test: python -m unittest -v test_common.CommonApkUtilsTest
Change-Id: Ib4b1dfa8b19e505fc70684b648efc36171c73bbf
Add construct_sparse_image() to test_utils.py, which is a util function
to create sparse images. The new tests also partially cover the recent
changes that add 'incomplete' and 'uses_shard_blocks' tags.
Test: python -m unittest test_common
Change-Id: Ia15f5c4ad12423691216ebbad2c28f95c8427d7e
When target defines 'BOARD_EXT4_SHARE_DUP_BLOCKS := true', the generated
system/vendor images may contain shared blocks (i.e. some blocks will
show up in multiple files' block list), which violates the current
assumptions in BBOTA script.
This CL allows generating BBOTAs by considering the first occurrence as
the "owner" of the shared blocks. All the later users of the shared
blocks will have an incomplete block list, whose RangeSet's will be
tagged with 'uses_shared_blocks'.
Files with 'uses_shared_blocks' tag will not be diff'd with imgdiff,
potentially with patch size penalty. Such files will be accounted for in
imgdiff stats report, where we can revisit for a better solution.
Bug: 64109868
Test: Generate BBOTA full and incremental package with targets defining
'BOARD_EXT4_SHARE_DUP_BLOCKS := true'.
Change-Id: I87fbc22eef7fafe2a470a03fdcfa1babf088ea8d
This CL uses the 'incomplete' tag to skip applying imgdiff to files with
incomplete block list. It's not the ideal fix to address the holes in
ext4 images, but would unhide other imgdiff issues covered by the
unconditional fallback.
Bug: 68016761
Test: Generate an incremental OTA package from images with incomplete
block list. Check the imgdiff stats report.
Test: `python -m unittest test_blockimgdiff`
Change-Id: Ice77686414e70f5e42de35c1757fb31cf02e4fd4
pylint complains about undefined `diff_done`:
W:754, 8: Global variable 'diff_done' undefined at the module level (global-variable-undefined)
W:820,14: Global variable 'diff_done' undefined at the module level (global-variable-undefined)
It would still warn about using global statement after adding the
definition.
W:859, 8: Using the global statement (global-statement)
W:925,14: Using the global statement (global-statement)
This CL computes 'diff_done' via 'len(diff_queue)' instead. It also
moves the progress reporting _before_ the diff work. This way it avoids
showing 100% progress with still changing filenames (because multiple
workers could see an empty queue simultaneously upon finishing their own
works).
There're possible alternatives, such as using the 'nonlocal' keyword in
Python 3 (which we're not there yet), or by using mutable object instead
(e.g. 'diff_done = [0]'). This CL looks cleaner, since it just kills the
var.
Test: Generate a BBOTA incremental. Check the on-screen progress
report.
Test: `pylint --rcfile=pylintrc blockimgdiff.py` no longer complains
about the global diff_done.
Change-Id: I339824735527e1f794b5b1dc99ff3fdb2da85744
The secondary payload should always be applied with
SWITCH_SLOT_ON_REBOOT=0. This CL moves the 'secondary' parameter from
Payload.WriteToZip() to Payload.__init__(). So it can append the flag to
secondary/payload_properties.txt.
Bug: 35724498
Test: Generate an A/B OTA with --include_secondary. Check
secondary/payload_properties.txt entry in the generated ZIP.
Test: `python -m unittest test_ota_from_target_files`
Change-Id: I816c07ab57a1c8a52eff785801634b8b1cb134d4
We have a couple of active imgdiff workarounds (and likely with one more
inbounding that allows having shared blocks in ext4 image). Most of
these workarounds need extending imgdiff's capability. While us not
getting there anytime soon, collect the stats to better understand the
impact of each kind so we can prioritize accordingly.
A sample report is as follows.
Imgdiff Stats Report
========================
APK files diff'd with imgdiff (count: 88)
-------------------------------------------
/system/priv-app/Shell/Shell.apk
...
Large APK files split and diff'd with imgdiff (count: 4)
----------------------------------------------------------
/system/priv-app/Settings/Settings.apk
...
Bug: 68016761
Test: Generate an incremental BBOTA package. Check the stats report.
Test: python -m unittest test_blockimgdiff
Change-Id: I27ad862cde472ab2806db877632ce5a0607420f2
In Transfer class, unbundle 'intact' with the monotonicity of the input
ranges. Negate the logic of 'intact', and thus rename it to 'trimmed'.
Move this property from an attribute of Transfer class as the one in
RangeSet.extra. 'trimmed' indicates whether the source / target ranges
have been modified after creating the Transfer() instance.
The logic that determines whether we can apply imgdiff has been
refactored and consolidated into BlockImageDiff.CanUseImgdiff(). Now
both of the two paths call this single copy, i.e. the one that detects
large APKs (before creating Transfer()'s), and the other that's about to
generate the patch for a given Transfer instance.
Bug: 68016761
Test: python -m unittest test_blockimgdiff
Test: Generate an incremental BBOTA package.
Change-Id: Id07195f63f1fa6c3af6e9091940d251cf09fa104
'monotonic' has been non-optional since [1] (L-MR1). Fix the comment in
RangeSet.parse(), as well as the use in blockimgdiff.py.
[1] commit 8b72aefb5a.
Test: Generate an incremental BBOTA package.
Change-Id: I7f95231683473b4f0f07f9c83fccc0e36a1340cb
Prior to this CL, the call to delta_generator in
check_ota_package_signature.VerifyAbOtaPayload() didn't redirect stderr.
The logs (mostly INFO) on successful verification added noise to the
normal output, which also upset the unittest result parser.
This CL captures stderr outputs from delta_generator, and will only dump
them on error.
Bug: 72884343
Test: `python -m unittest -v test_ota_from_target_files > /dev/null`
gives clean output.
Test: Inject error into delta_generator. The call to
check_ota_package_signature correctly dumps both of stdout and
stderr outputs.
Change-Id: I014a4b21bf758dcf0a4b9963259d6019851935ee
By default, an A/B OTA package doesn't contain the images for the
secondary slot (e.g. system_other.img). Specifying
"--include_secondary" that's introduced in this CL allows generating
a separate payload that will install secondary slot images. Both
payloads will be added to the generated A/B OTA package.
An example A/B OTA package with secondary payload
|
+-- payload.bin
|
+-- payload_properties.txt
|
+-- secondary/payload.bin
|
+-- secondary/payload_properties.txt
|
+-- ...
Such a package needs to be applied in a two-stage manner. During the
first stage, the updater applies the primary payload only. Upon
finishing, it reboots the device into the newly updated slot. It then
continues to install the secondary payload to the inactive slot, but
without switching the active slot at the end (needs the matching support
in update_engine, i.e. SWITCH_SLOT_ON_REBOOT flag).
Due to the special install procedure, the secondary payload will be
always generated as a full payload.
Bug: 35724498
Test: Generate full and incremental OTAs with --include_secondary. Check
the generated OTAs.
Test: python -m unittest test_ota_from_target_files
Change-Id: I975e826bec492e86eb400f99de0c355a32420127
