Without this, setting __FILE_OFFSET_BITS to 64 and targeting pre-L
made sendfile entirely unavailable.
Bug: https://github.com/android-ndk/ndk/issues/333
Test: builds
Change-Id: I82d326ff244473dd5443632ff6c19d726d5e29ee
This reverts commit 1daa18e737.
To be replaced with customized linker-namespaces for different kind of binaries.
Bug: http://b/34407260
Change-Id: Ib4e480669bbdf44ee4f59bf05f07d3e082c348c9
Test: m
Test: Used bionic tests available under bionic/tests folder.
Tested for mips32r1/mips32r2/mips64r6 on emulators.
Change-Id: I589415ddc496df3f6067ae34cb33ca58b3a1f276
Signed-off-by: Prashant Patil <prashant.patil@imgtec.com>
Prevent the compiler from being too smart and allocating a stack buffer
at the beginning of a function.
Bug: http://b/36206043
Test: 32/64-bit dynamic tests pass, static ones still don't
Change-Id: I90c575be43a9dd6c4fefc0d8b514f1ae0405b994
Test: Make sure arm, x86, x86_64, mips, mips64 emulators boot
Make sure sailfish still boots
Ran CTS test from
https://android-review.googlesource.com/#/c/348671/3 and it passed
The instructions for how to run mips emulators above worked, but
the CTS tests did not seem to actually run.
Change-Id: Iddee5acdb19ed32c7bd4657573313ca439cf6a49
Specifically, this change would translate /system/lib/hw/foo.so to /data/lib/hw/foo.so.
Bug: 30320104
Test: in SANITIZE_TARGET=address build, /proc/$PID/maps for
system_server show libraries from /data/lib64/hw and not from
/system/lib64/hw.
Change-Id: Ia1fa22fa1ca088fe76e41308762efe037541fc74
Due to b/33681361, libraries are now installed to different paths than
before. In order for linker to load libraries from the new paths, the
default lib paths are updated.
However, this patch is temporary; different paths must be used depending
on whether the process is executing vendor executable or framework
executable. For example, vendor executable should not be able to access
/system/lib/, which is for framework-internal libs. Similarly,
framework executables should not be able to access /vendor/lib.
We have several solutions (linker namespace and DT_RUNPATH) to support
the scenario, but haven't concluded which is better. Furthermore, if we
restrict the search path right now, it will break many parts of Android
that are not yet prepared for the restriction. So, until then, we
temporarily use the single search path for both framework/vendor
executables.
Test: build & run
Change-Id: I806e9b4379bce96653b1dd3354b9f4801abb0411
When I rewrote malloc debug, I accidentally thought that each
value returned in the info buffer contained the number of backtrace
frames in the backtrace. This was incorrect, it should have been
the total number of allocations with the same backtrace.
This is a temporary fix that sets that value to 1. The better fix is
to de-dupe backtraces and then return all allocations of the same size
with the same backtrace.
I updated the documents to describe this.
Bug: 31854476
Test: Unit tests pass.
Change-Id: Idf9efaa3d363923b5d7543d90dc7c65a0ed553d9
Let the fallback allocator be used on multiple threads (as long as only
one thread is using it at once).
Bug: http://b/35858739
Change-Id: Id3e2fc6b7c093c6e56870524ffda28946de09e29
Remove dependencies that are now linked directly into
libdebuggerd_handler_fallback via whole_static_libraries.
Test: m
Change-Id: I00e2edb2778c2986c533f97708252aa0996d0498
snprintf to a buffer of length PATH_MAX consumes about 7kB of stack.
Bug: http://b/35858739
Test: bionic-unit-tests --gtest_filter="*big_enough*"
Change-Id: I34a7f42c1fd2582ca0d0a9b7e7a5290bc1cc19b1
This changes system_properties' initialize_properties to prefer
loading property_contexts from /system/etc/selinux &
/vendor/etc/selinux, while falling back to the pre-existing behavior
of loading from /.
Test: Device with *_property_contexts in / boots up fine, no denials
to do with properties, getprop -Z lists correct labels.
Test: Device with *_property_contexts in /system & /vendor, but not
in /, boots up fine, no denials to do with properties,
getprop -Z lists correct labels.
Test: Device with *_property_contexts in /system & vendor and with
empty *_property_contexts in / boots up fine, no denials to do
with properties, getprop -Z lists correct labels.
Bug: 36002573
Change-Id: I15174acdf89ee8f5a96acf1e38a54d4214df51ef
__pthread_internal_free doesn't happen on threads that are detached,
causing the bionic TLS allocation (and guard pages) to be leaked.
Fix the leak, and name the allocations to make things apparent if this
ever happens again.
Bug: http://b/36045112
Test: manually ran a program that detached empty threads
Change-Id: Id1c7852b7384474244f7bf5a0f7da54ff962e0a1
