There were two problems here:
* This would fail when run with unlimited stack, because it didn't know
that bionic reports unlimited stacks as 8MiB.
* This would leave RLIMIT_STACK small, causing failures to exec (so the
popen and system tests would fail).
(cherry-pick of 27a9aed819 plus the
new ScopeGuard.h from a3ad450a2e3fb6b3fe359683b247eba20896f646.)
Bug: 17394276
Change-Id: I5b92dc64ca089400223b2d9a3743e9b9d57c1bc2
This is needed to make L work correctly, and bionic tests pass
again, after applying the equivalent of
commit 00aaea3645 there.
It makes the preexisting code that uses __sync implementations
much more useful, although we should no longer be exercising that
code in AOSP.
Specifically fixes:
We were invoking __has_extension and __has_builtin for GCC compilations.
They're clang specific. Restructured the tests.
The __sync implementation was not defining the LOCK_FREE macros.
ATOMIC_VAR_INIT was using named field initializations. These are a
C, not C++, feature, that is not supported by g++ 4.6.
The stdatomic bionic test still failed with 4.6 and glibc with our
questionable LOCK_FREE macro implementation. Don't run that piece
with 4.6.
In L, this is a prerequisite for fixing:
Bug:16880454
Bug:16513433
Change-Id: I9b61e42307f96a114dce7552b6ead4ad1c544eab
(cherry picked from commit 32429606bf)
We seem to use this stdatomic.h sometimes, and slightly different prebuilts
at other times, making them all difficult to test, and making it unclear
which one we're testing. This generalizes the bionic header so that it
can be used directly as the prebuilt header as well. So long as they
don't diverge again, that should somewhat improve test coverage.
Use the correct builtin for atomic_is_lock_free.
Fix atomic_flag_init.
Turn on atomic tests even with __GLIBC__, since they now appear to pass.
Include uchar.h in stdatomic.h where needed.
Add a basic memory ordering test.
Fix bit-rotted comments in bionic tests makefile.
Prerequisite for fixing b/16880454 and
Bug:16513433
Change-Id: If6a14c1075b379395ba5d93357d56025c0ffab68
(cherry picked from commit 00aaea3645)
...rather than just what's already mapped in. This seems somewhat
contrary to POSIX's "All pages within the stack described by stackaddr
and stacksize shall be both readable and writable by the thread", but
it's what glibc does.
Bug: 17111575
(cherry picked from commit 9e4ffa7032)
Change-Id: I73f219a569917b2e4546c09436d7ef5231facc07
For tests that call uselocale(), the locale is stored in the
g_userlocale_key thread-specific key. If freelocale() is called later,
then g_uselocal_key points to a deleted pointer. CTS eventually calls
vfprintf to print the result, which calls MB_CUR_MAX and MB_CUR_MAX
accesses the deleted locale stored in g_uselocale_key, causing unpredictable
errors.
Fixed the tests by calling uselocale() with the old locale before
calling freelocale.
Bug: 17299565
Change-Id: I87efa2a9b16999a11d587f68d3aeedcbe6ac8a2c
On most architectures the kernel subtracts a random offset to the stack
pointer in create_elf_tables by calling arch_align_stack before writing
the auxval table and so on. On all but x86 this doesn't cause a problem
because the random offset is less than a page, but on x86 it's up to two
pages. This means that our old technique of rounding the stack pointer
doesn't work. (Our old implementation of that technique was wrong too.)
It's also incorrect to assume that the main thread's stack base and size
are constant. Likewise to assume that the main thread has a guard page.
The main thread is not like other threads.
This patch switches to reading /proc/self/maps (and checking RLIMIT_STACK)
whenever we're asked.
Bug: 17111575
Signed-off-by: Fengwei Yin <fengwei.yin@intel.com>
(cherry picked from commit 57b7a6110e)
Change-Id: I87e679ee1c0db8092f2d1221c8e7c1461545c5a4
This test only works if you're root (strictly: if you have permission to
CLONE_NEWNS), so it's useful to us when we're doing ad hoc testing (since
that's usually done as root), but it's not useful as part of CTS or when
running the tests on the host.
Bug: 16705621
Bug: 17170200
Change-Id: Ia92c871b15f7e45fc174bb59bc95540fd00ae745
During pthread_exit, the keys are cleaned. Unfortunately, a call to
free occurs after the cleanup and the memory for some of the keys
is recreated when using jemalloc. The solution is to do the key
cleanup twice.
Also, modify the pthread_detach__leak test to be less flaky
when run on a jemalloc system.
Bug: 16513133
Change-Id: Ic17e8344bdc1ba053c4f5b6d827a4c19c57860c1
Previously this was hard coded to 4. This is only the case for UTF-8
locales.
As a side effect, this properly reports C.UTF-8 as the default locale
instead of C.
Change-Id: I7c73cc8fe6ffac61d211cd5f75287e36de06f4fc
(cherry picked from commit 1aec7c1a35)
A mistake I made while cleaning this up the first time through.
mbstrtowcs(3) sets the src param to null if it finishes the string.
Change-Id: I6263646e25d9537043b7025fd1dd6ae195f365e2
(cherry picked from commit b6cc8e00cd)
The memchr implementation for 64 bit fails if these conditions occur:
- The buffer is 32 byte aligned.
- The buffer contains the character in the first byte.
- The count sent in is zero.
The function should return NULL, but it's not.
Bug: 16676625
(cherry picked from commit e03e1eac0b)
Change-Id: Ie4cca2c445127a0936ee2b96651a8e7204fbaffd
The len parameter is a _maximum_ length. The previous code was treating
it as an exact length, causing the following typical call to fail:
mbsrtowcs(out, &in, sizeof(out), state); // sizeof(out) > strlen(in)
Change-Id: I48e474fd54ea5f122bc168a4d74bfe08704f28cc
(cherry picked from commit 6b55ba54ef)
Also clean up the implementation of all the pty functions, add tests,
and fix the stub implementations of ttyname(3) and ttyname_r(3).
Bug: https://code.google.com/p/android/issues/detail?id=58888
(cherry picked from commit 4916706cfe)
Change-Id: I5cb7a1c17b156456e4c4818e65f256eb8d045424
This doesn't require us to change any of the syscall implementations
because (a) the LP32 ones have sizeof(int) == sizeof(long) anyway,
which is how we never noticed this bug before and (b) the LP64 ones
all use a 64-bit register for the result (and for the syscall number
too).
Bug: https://code.google.com/p/android/issues/detail?id=73952
Bug: 16568314
(cherry picked from commit 21972b61ec)
Change-Id: Ifbc424be29e5650ec72a24df25dd35f24fdd5b3c
It seemed like a clever trick to use the internal log message formatting
code in syslog(3), but on reflection that means you can't (for example)
format floating point numbers. This patch switches us over to using good
old vsnprintf(3), even though that requires us to jump through a few hoops.
There's no obvious way to unit test this, so I wrote a little program and
ran that.
Bug: 14292866
Change-Id: I9c83500ba9cbb209b6f496067a91bf69434eeef5
dlsym(3) with handle != RTLD_DEFAULT|RTLD_NEXT performs
breadth first search through the dependency tree.
Bug: 16653281
(cherry picked from commit aa0f2bdbc2)
Change-Id: I0ba8c2034ab341f8a279cdb4e2e7e47f1aef7897
On 32-bit MIPS, 64-bit atomic ops are achieved through locks.
So allow the test to fail for atomic_intmax_t on 32-bit MIPS.
(cherry picked from commit f1837377d2)
Change-Id: I973d999c31c9ab89b5a7b709beff6486b93408f2
I've also added insque(3) and remque(3) (from NetBSD because the OpenBSD
ones are currently broken for non-circular lists).
I've not added the three hash table functions that should be in this header
because they operate on a single global hash table and thus aren't likely
to be useful.
Bug: https://code.google.com/p/android/issues/detail?id=73719
(cherry picked from commit 3e424d0a24)
Change-Id: I5882a6b48c80fea8ac6b9c27e7b9de10b202b4ff
PR_GET_DUMPABLE is used by an application to indicate whether or
not core dumps / PTRACE_ATTACH should work.
Security sensitive applications often set PR_SET_DUMPABLE to 0 to
disable core dumps, to avoid leaking sensitive memory to persistent
storage. Similarly, they also set PR_SET_DUMPABLE to zero to prevent
PTRACE_ATTACH from working, again to avoid leaking the contents
of sensitive memory.
Honor PR_GET_DUMPABLE when connecting to debuggerd. If an application
has said it doesn't want its memory dumped, then we shouldn't
ask debuggerd to dump memory on its behalf.
FORTIFY_SOURCE tests: Modify the fortify_source tests to set
PR_SET_DUMPABLE=0. This reduces the total runtime of
/data/nativetest/bionic-unit-tests/bionic-unit-tests32 from approx
53 seconds to 25 seconds. There's no need to connect to debuggerd
when running these tests.
Bug: 16513137
(cherry picked from commit be0e43b776)
Change-Id: I6e1a9bce564e94fc19893d639b15f38c549cabfa
Since we don't have syslogd on Android and you can't run one on a non-rooted
device, it's more useful if syslog output just goes to the regular Android
logging system.
Bug: 14292866
(cherry picked from commit 3ad8ecb64e)
Change-Id: I3038855ca4f22532bf6d2c45d3f8028b866975f9
Code developed for glibc or older versions of bionic might expect more
randomness than the BSD implementation provides.
Bug: 15829381
(cherry picked from commit 76c241b091)
Change-Id: If721b3f16efdb21cb67df5ec5034c0ba905bd029
* Static linker optimizes protected local symbol
out of existence, which leads to test failure.
Disabling it for now.
Change-Id: I8de327e5073f98b64639f7a0bba3a273aa419884
If you make clone, fork, or vfork system calls directly, you're still
on your own, but we now do the right thing for the clone wrapper.
With this implementation, children lose the getpid caching, but we've
no reason to think that that covers any significant use cases.
Bug: 15387103
Change-Id: Icfab6b63c708fea830960742ec92aeba8ce7680d
In practice, with this implementation we never need to make a system call.
We get the main thread's tid (which is the same as our pid) back from
the set_tid_address system call we have to make during initialization.
A new pthread will have the same pid as its parent, and a fork child's
main (and only) thread will have a pid equal to its tid, which we get for
free from the kernel before clone returns.
The only time we'd actually have to make a getpid system call now is if
we take a signal during fork and the signal handler calls getpid. (That,
or we call getpid in the dynamic linker while it's still dealing with its
own relocations and hasn't even set up the main thread yet.)
Bug: 15387103
Change-Id: I6d4718ed0a5c912fc75b5f738c49a023dbed5189
This allows an easier way to share config parameters between unit tests
and the bionic code.
It also fixes a problem where the 32 bit bionic tests based on glibc, or
the cts list executable did not have the pvalloc,valloc tests.
Change-Id: Ib47942cb8a278252faa7498a6ef23e9578db544f
