3[bionic](https://en.wikipedia.org/wiki/Bionic_(software)) is Android's
4C library, math library, and dynamic linker.
6# Using bionic as an app developer
8See the [user documentation](docs/).
10# Working on bionic itself
12This documentation is about making changes to bionic itself.
14## What are the big pieces of bionic?
16#### libc/ --- libc.so, libc.a
18The C library. Stuff like `fopen(3)` and `kill(2)`.
20#### libm/ --- libm.so, libm.a
22The math library. Traditionally Unix systems kept stuff like `sin(3)` and
23`cos(3)` in a separate library to save space in the days before shared
26#### libdl/ --- libdl.so
28The dynamic linker interface library. This is actually just a bunch of stubs
29that the dynamic linker replaces with pointers to its own implementation at
30runtime. This is where stuff like `dlopen(3)` lives.
32#### libstdc++/ --- libstdc++.so
34The C++ ABI support functions. The C++ compiler doesn't know how to implement
35thread-safe static initialization and the like, so it just calls functions that
36are supplied by the system. Stuff like `__cxa_guard_acquire` and
37`__cxa_pure_virtual` live here.
39#### linker/ --- /system/bin/linker and /system/bin/linker64
41The dynamic linker. When you run a dynamically-linked executable, its ELF file
42has a `DT_INTERP` entry that says "use the following program to start me". On
43Android, that's either `linker` or `linker64` (depending on whether it's a
4432-bit or 64-bit executable). It's responsible for loading the ELF executable
45into memory and resolving references to symbols (so that when your code tries to
46jump to `fopen(3)`, say, it lands in the right place).
48#### tests/ --- unit tests
50The `tests/` directory contains unit tests. Roughly arranged as one file per
51publicly-exported header file.
53#### benchmarks/ --- benchmarks
55The `benchmarks/` directory contains benchmarks, with its own [documentation](benchmarks/README.md).
58## What's in libc/?
69 # Each architecture has its own subdirectory for stuff that isn't shared
70 # because it's architecture-specific. There will be a .mk file in here that
71 # drags in all the architecture-specific files.
73 # Every architecture needs a handful of machine-specific assembler files.
74 # They live here.
76 # Most architectures have a handful of optional assembler files
77 # implementing optimized versions of various routines. The <string.h>
78 # functions are particular favorites.
80 # The syscalls directories contain script-generated assembler files.
81 # See 'Adding system calls' later.
84 # The public header files on everyone's include path. These are a mixture of
85 # files written by us and files taken from BSD.
88 # The kernel uapi header files. These are scrubbed copies of the originals
89 # in external/kernel-headers/. These files must not be edited directly. The
90 # generate_uapi_headers.sh script should be used to go from a kernel tree to
91 # external/kernel-headers/ --- this takes care of the architecture-specific
92 # details. The update_all.py script should be used to regenerate bionic's
93 # scrubbed headers from external/kernel-headers/.
96 # These are private header files meant for use within bionic itself.
99 # Contains the DNS resolver (originates from NetBSD code).
104 # These directories contain unmolested upstream source. Any time we can
105 # just use a BSD implementation of something unmodified, we should.
106 # The structure under these directories mimics the upstream tree,
107 # but there's also...
110 # This is where we keep the hacks necessary to build BSD source
111 # in our world. The *-compat.h files are automatically included
112 # using -include, but we also provide equivalents for missing
113 # header/source files needed by the BSD implementation.
116 # This is the biggest mess. The C++ files are files we own, typically
117 # because the Linux kernel interface is sufficiently different that we
118 # can't use any of the BSD implementations. The C files are usually
119 # legacy mess that needs to be sorted out, either by replacing it with
120 # current upstream source in one of the upstream directories or by
121 # switching the file to C++ and cleaning it up.
124 # The code that implements the functionality to enable debugging of
125 # native allocation problems.
128 # These are legacy files of dubious provenance. We're working to clean
129 # this mess up, and this directory should disappear.
132 # Various tools used to maintain bionic.
135 # A modified superset of the IANA tzcode. Most of the modifications relate
136 # to Android's use of a single file (with corresponding index) to contain
137 # time zone data.
139 # Android-format time zone data.
140 # See 'Updating tzdata' later.
144## Adding libc wrappers for system calls
146The first question you should ask is "should I add a libc wrapper for
147this system call?". The answer is usually "no".
149The answer is "yes" if the system call is part of the POSIX standard.
151The answer is probably "yes" if the system call has a wrapper in at
152least one other C library.
154The answer may be "yes" if the system call has three/four distinct
155users in different projects, and there isn't a more specific library
156that would make more sense as the place to add the wrapper.
158In all other cases, you should use
161Adding a system call usually involves:
163 1. Add entries to SYSCALLS.TXT.
164 See SYSCALLS.TXT itself for documentation on the format.
165 2. Run the gensyscalls.py script.
166 3. Add constants (and perhaps types) to the appropriate header file.
167 Note that you should check to see whether the constants are already in
168 kernel uapi header files, in which case you just need to make sure that
169 the appropriate POSIX header file in libc/include/ includes the
170 relevant file or files.
171 4. Add function declarations to the appropriate header file. Don't forget
172 to include the appropriate `__INTRODUCED_IN()`.
173 5. Add the function name to the correct section in libc/libc.map.txt and
174 run `./libc/tools/genversion-scripts.py`.
175 6. Add at least basic tests. Even a test that deliberately supplies
176 an invalid argument helps check that we're generating the right symbol
177 and have the right declaration in the header file, and that you correctly
178 updated the maps in step 5. (You can use strace(1) to confirm that the
179 correct system call is being made.)
182## Updating kernel header files
184As mentioned above, this is currently a two-step process:
186 1. Use generate_uapi_headers.sh to go from a Linux source tree to appropriate
187 contents for external/kernel-headers/.
188 2. Run update_all.py to scrub those headers and import them into bionic.
190Note that if you're actually just trying to expose device-specific headers to
191build your device drivers, you shouldn't modify bionic. Instead use
192`TARGET_DEVICE_KERNEL_HEADERS` and friends described in [config.mk](https://android.googlesource.com/platform/build/+/master/core/config.mk#186).
195## Updating tzdata
197This is fully automated (and these days handled by the libcore team, because
198they own icu, and that needs to be updated in sync with bionic):
200 1. Run update-tzdata.py in external/icu/tools/.
203## Verifying changes
205If you make a change that is likely to have a wide effect on the tree (such as a
206libc header change), you should run `make checkbuild`. A regular `make` will
207_not_ build the entire tree; just the minimum number of projects that are
208required for the device. Tests, additional developer tools, and various other
209modules will not be built. Note that `make checkbuild` will not be complete
210either, as `make tests` covers a few additional modules, but generally speaking
211`make checkbuild` is enough.
214## Running the tests
216The tests are all built from the tests/ directory.
218### Device tests
220 $ mma # In $ANDROID_ROOT/bionic.
221 $ adb root && adb remount && adb sync
222 $ adb shell /data/nativetest/bionic-unit-tests/bionic-unit-tests
223 $ adb shell \
225 # Only for 64-bit targets
226 $ adb shell /data/nativetest64/bionic-unit-tests/bionic-unit-tests
227 $ adb shell \
230Note that we use our own custom gtest runner that offers a superset of the
231options documented at
233in particular for test isolation and parallelism (both on by default).
235### Device tests via CTS
237Most of the unit tests are executed by CTS. By default, CTS runs as
238a non-root user, so the unit tests must also pass when not run as root.
239Some tests cannot do any useful work unless run as root. In this case,
240the test should check `getuid() == 0` and do nothing otherwise (typically
241we log in this case to prevent accidents!). Obviously, if the test can be
242rewritten to not require root, that's an even better solution.
244Currently, the list of bionic CTS tests is generated at build time by
245running a host version of the test executable and dumping the list of
246all tests. In order for this to continue to work, all architectures must
247have the same number of tests, and the host version of the executable
248must also have the same number of tests.
250Running the gtests directly is orders of magnitude faster than using CTS,
251but in cases where you really have to run CTS:
253 $ make cts # In $ANDROID_ROOT.
254 $ adb unroot # Because real CTS doesn't run as root.
255 # This will sync any *test* changes, but not *code* changes:
256 $ cts-tradefed \
257 run singleCommand cts --skip-preconditions -m CtsBionicTestCases
259### Host tests
261The host tests require that you have `lunch`ed either an x86 or x86_64 target.
262Note that due to ABI limitations (specifically, the size of pthread_mutex_t),
26332-bit bionic requires PIDs less than 65536. To enforce this, set /proc/sys/kernel/pid_max
266 $ ./tests/run-on-host.sh 32
267 $ ./tests/run-on-host.sh 64 # For x86_64-bit *targets* only.
269You can supply gtest flags as extra arguments to this script.
271### Against glibc
273As a way to check that our tests do in fact test the correct behavior (and not
274just the behavior we think is correct), it is possible to run the tests against
275the host's glibc.
277 $ ./tests/run-on-host.sh glibc
280## Gathering test coverage
282For either host or target coverage, you must first:
284 * `$ export NATIVE_COVERAGE=true`
285 * Note that the build system is ignorant to this flag being toggled, i.e. if
286 you change this flag, you will have to manually rebuild bionic.
287 * Set `bionic_coverage=true` in `libc/Android.mk` and `libm/Android.mk`.
289### Coverage from device tests
291 $ mma
292 $ adb sync
293 $ adb shell \
294 GCOV_PREFIX=/data/local/tmp/gcov \
295 GCOV_PREFIX_STRIP=`echo $ANDROID_BUILD_TOP | grep -o / | wc -l` \
297 $ acov
299`acov` will pull all coverage information from the device, push it to the right
300directories, run `lcov`, and open the coverage report in your browser.
302### Coverage from host tests
304First, build and run the host tests as usual (see above).
306 $ croot
307 $ lcov -c -d $ANDROID_PRODUCT_OUT -o coverage.info
308 $ genhtml -o covreport coverage.info # or lcov --list coverage.info
310The coverage report is now available at `covreport/index.html`.
313## Attaching GDB to the tests
315Bionic's test runner will run each test in its own process by default to prevent
316tests failures from impacting other tests. This also has the added benefit of
317running them in parallel, so they are much faster.
319However, this also makes it difficult to run the tests under GDB. To prevent
320each test from being forked, run the tests with the flag `--no-isolate`.
323## 32-bit ABI bugs
325See [32-bit ABI bugs](docs/32-bit-abi.md).