Update the backtrace leak tests to share a map since this is the most
common way it will be used, and it runs much faster.
Remove the CreateNew functions, and references to them.
Remove benchmarks of CreateNew functions.
Test: Builds, ran new unit tests, ran art tests.
Change-Id: I4a25a412f1a74c6d43aebbebbf32ead20ead8f94
* changes:
Tests are run with proper namespace configs by their locations
Fix: make vndk namespace visible for 2016 Pixes
vndk: remove vendor/lib from the search path of the 'vndk' namespace
libz is back to LLNDK
Make the rs namespace visible
Add libft2.so to ll-ndk for rs namespace
Move libz vndk-sp from ll-ndk
Tests in /data/[nativetest|benchmarktest] run with namespace config for
system and tests in /data/[nativetest|benchmarktest]/vendor run with
namespace config for vendor.
They no longer run in the 'test' namespace config which didn't impose
any restriction for libraries.
Bug: 67028906
Test: sailfish/marlin builds and boots
Test: no VTS regression on system.img from GSI + vendor.img from
marlin/sailfish
Test: VtsKernelLibcutilsTest successful in above config
Merged-In: I28cdef960d087565c8a22dca0e9a154fb1c3bb94
Change-Id: I28cdef960d087565c8a22dca0e9a154fb1c3bb94
(cherry picked from commit 326b783ad9)
The nativeloader tries to find the 'vndk' namespace when there is a
vendor apk in the classloader paths. This can happen even for a
downloaded app if the app is using a vendor apk via <uses-library> tag.
In order to prevent the nativeloader from failing to find the vndk
namespace, the namespace is marked as visible.
Bug: 66482442
Test: download the app mentioned in the bug, it does not crash.
Merged-In: I82e2394eb6eedcb8645e1a5b3735bbfe2735b312
Change-Id: I82e2394eb6eedcb8645e1a5b3735bbfe2735b312
(cherry picked from commit 1cc09e749f)
'vndk' namespace is the place for VNDK-SP libs. So /vendor/lib must be
removed from its search paths. It was there just because libhardware
(which is VNDK-SP) is loading HAL libs in vendor partition from there.
However this problem has been solved by modifying the libhardware so
that HAL libs are loaded from the 'sphal' namespace and not from the
current (the 'vndk') namespace.
Bug: 37731053
Test: sailfish builds and boots
Merged-In: Ia88934a975aa8811e05b5ba408e42d132f20f4e9
Change-Id: Ia88934a975aa8811e05b5ba408e42d132f20f4e9
(cherry picked from commit f01b52895d)
For 2016 pixel devices, where VNDK is not fully enforced, move libz back
to LLNDK so that we can pass the CTS without requiring the default
namespace to be isolated.
If we have libz in vndk-sp directory, test_linker_namespaces fails
because /system/lib/vndk-sp/libz.so becomes accessible. In order to make
the lib inaccessible, we have to make the default namespace isolated,
but this can't be done for 2016 pixel devices where VNDK is not fully
enforced. So, we choose to remove /system/lib/vndk-sp/libz.so and keep
the single copy at /system/lib.
Bug: 65018779
Test: android.jni.cts.JniStaticTest#test_linker_namespaces successful on
2016 pixel devices
Merged-In: Ib6b6560b02be69d2015c0afb86acf07c02b30c2f
Change-Id: Ib6b6560b02be69d2015c0afb86acf07c02b30c2f
(cherry picked from commit 697eb2da0d)
This allows us to dlopen libRS_internal.so directly from the rs
namespace, not from the sphal namespace.
Bug: 64747884
Test: VtsHalRenderscriptV1_0TargetTest successful on the device built
with BOARD_VNDK_VERSION=current and [system] namespace config is applied
to /data/nativetest[64]/* processes.
Merged-In: I0b441791e395798e80a84592ca01e771bb1c201a
Change-Id: I0b441791e395798e80a84592ca01e771bb1c201a
(cherry picked from commit 421a5e46b7)
libft2.so is changed to ll-ndk that is available only for rs
namespace.
Bug: 64425518
Test: build and boot with BOARD_VNDK_VERSION=current
Merged-In: I991dc774ca9b92fb6e95a7656243a6a4ecdc0ab9
Change-Id: I991dc774ca9b92fb6e95a7656243a6a4ecdc0ab9
(cherry picked from commit 513f76394a)
The ABI of libz is not as stable as it is for ll-ndk.
Bug: 37617391
Test: build and boot
Merged-In: I883bc6fda268e98cc7cdd5888264170c58688794
Change-Id: I883bc6fda268e98cc7cdd5888264170c58688794
(cherry picked from commit dadd3a846f)
Nobody is looking at the mismatches, and it can cause problems
with tombstone parsers.
Also, fix the dump_header_info test and remove unused properties_fake.cpp.
Test: Ran unit tests, verified tombstones still work.
Change-Id: I4261646016b4e84b26a5aee72f3227f1ce48ec9a
Previously fastboot would carry on regardless if decompression failed:
fastboot: archive does not contain 'vbmeta.img'
fastboot: extracting vendor.img (260 MB)...
fastboot: W/ziparchive(56777): Zip: unable to allocate 272781472 bytes at offset 0 : No space left on device
fastboot: failed to extract 'vendor.img': I/O error
fastboot: archive does not contain 'vendor_other.img'
fastboot: wiping userdata...
This is because all but "boot" and "system" are considered "optional",
and the implementation of "optional" was "ignore any failures". What it
_should_ have meant was "it's okay if these don't exist, but if they do,
failures matter".
Fix this logic, use die() more aggressively, and remove spurious "\n"s
from die() format strings.
Also fix spurious whitespace in the libziparchive format string. Before:
Zip: unable to allocate 272781472 bytes at offset 0 : No space left on device
After:
Zip: unable to allocate 272781472 bytes at offset 0: No space left on device
Bug: http://b/68383022
Test: `fastboot update` on marlin
Change-Id: I3cbf55f1a33ca125f293f873eafbcfb86c880ba8
If there is a restart follow a stop/reset immediately or vice versa,
clear previous flag bits.
Test: manual - trigger restart after stop immediately to check if
service get started.
Change-Id: I4503177d7cb5ed054dbcf50cd8e09728415404d4
* changes:
bootstat: Do not allow unknown boot reasons to land in first field.
bootstat: move boot reason validation transformation policy into subroutine.
bootstat: test: fix Its_Just_So_Hard_reboot
For a oneshot service, if start happens immediately after stop,
the service could be still in stopping status and then start
won't do anything. This fix this race condition.
Test: manual - see reproduce instructions in bug.
Bug: 68020256
Change-Id: I20202fa346f1949a8bda3d90deedc8b6a6d814d3
If libbacktrace is statically linked into a library that interposes
libc, exit-time destructors can lead to a segfault upon exit if static
variables are used after their destruction.
Remove the one use of a static variable (with inconsequential
performance impact of less than 1us per backtrace), and enable
-Wexit-time-destructors to ensure this stays fixed in the future.
Test: mma
Change-Id: Icdaf1cf1c1f166cf501d9ecd1983c44e4305f7e9
Correct for the number of skipped frames when unwinding with
libunwindstack.
Test: backtrace_test32 --gtest_filter="unwind_frame_skip_*"
Change-Id: I9528977104fde3c4ec792a6db1ada24ed571b867
Manually doing 1000 iterations of the benchmark doesn't seem to add any
significant amount of precision, and it makes the benchmark take
forever and obfuscates the results. Just let benchmark figure out the
time (with the option of using command line flags to increase the
number of iterations).
Test: backtrace_benchmarks64
Change-Id: I8de912c1b3c904755c8e2ac4175ff70176544ba3
Create a private rfind that allows a fuzzy match based on a bit error
rate (BER) of 1 every 8 bits. last kmsg is affected by pstore ramoops
backing that suffers from data corruption. Add some additional
validation based on possible data corruption scenarios, as a noisy
match means higher chance of noisy data.
Noisy data notably can affect the battery level detection, but do not
typically result in false positives. Battery level, or failure, is
the responsibility of the BatteryStats service, providing a positive
signal and strong device-independent algorithm. The checking done in
bootstat is likely to be deprecated in favour of an API request to
BatteryStats once their algorithms deal with surprise outages due to
aging.
The kernel logging heuristic and BER fixup handily deals with a
prevalent issue where some bootloaders failure to properly notify us
of panics. This is where the gains are noticed with this improvement.
Test: system/core/bootstat/boot_reason_test.sh
Bug: 63736262
Change-Id: I93b4210f12fb47c5c036f4d6eb4cafeee4896d35
Replace simple strtoull with loop that ensures no leading zeros.
Restrict size of value buffer being checked as allocation was
going to end of retrieved buffer, which can cause unnecessary
memory pressure during boot.
Test: system/core/bootstat/boot_reason_test.sh
Bug: 63736262
Change-Id: Ifdc1d4fd3a73794c001577024ce7cbfde9c25028
If we sniff an unknown boot reason from last kmsg, make sure it
has a "reboot," prefix.
Test: system/core/bootstat/boot_reason_test.sh
Bug: 63736262
Change-Id: Ia1c401b8899d1f0c56bd4f5d8d2d19b7fc889a30
Add the test injection to known list, and deal with an error
propagation issue.
Test: system/core/bootstat/boot_reason_test.sh Its_Just_So_Hard_reboot
Bug: 63736262
Change-Id: I4799956978a8884c69c830fcedd7febd143651fd
Remove dependencies on zip_archive specific data structures. In follow
up changes, this method will be promoted to a public API and used from
androidfw/ZipUtils as well as tools/zipalign, thereby allowing us to
remove a near-exact copy of this code and to simplify depdencies.
Test: zip_archive_test
Bug: 35246701
Change-Id: If24a9965fbd6fff308c8758859026684fd2af3b9
Fixed issues related to forking services into new PID + mount
namespaces.
Remounting rootfs recursively as slave when creating a service in new
PID + mount namespaces. This prevents the service from interfering with
mount points in the parent namespace.
Unmount then mount /proc instead of mounting it with MS_REMOUNT, since
MS_REMOUNT is not sufficient to update /proc to the state appropriate
for the new PID namespace. Note that the /proc mount options specified
here are not the same as those used in the default mount namespace. I
kept them consistent with those used in the code prior to this fix.
Test: Used custom sleepd service to test init 'namespace' keyword.
Tested on angler in oreo-dev - I had to add PID namespaces to the
kernel (commit ad82c662).
Change-Id: I859104525f82fef3400d5abbad465331fc3d732f
