I broke the mips build yesterday because it doesn't use
<private/bionic_asm.h> like the other architectures, including mips64.
I want to move mips closer to mips64 to try to avoid this kind of thing
in future.
Change-Id: Idb985587ff355b9e5e765c1f5671dc0144cd2488
This gives us:
* <dirent.h>
struct dirent64
readdir64, readdir64_r, alphasort64, scandir64
* <fcntl.h>
creat64, openat64, open64.
* <sys/stat.h>
struct stat64
fstat64, fstatat64, lstat64, stat64.
* <sys/statvfs.h>
struct statvfs64
statvfs64, fstatvfs64.
* <sys/vfs.h>
struct statfs64
statfs64, fstatfs64.
This also removes some of the incorrect #define hacks we've had in the
past (for stat64, for example, which we promised to clean up way back
in bug 8472078).
Bug: 11865851
Bug: 8472078
Change-Id: Ia46443521918519f2dfa64d4621027dfd13ac566
bionic/libc/arch-arm64/syscalls/read.S ends with:
b.hi __set_errno
ret
END(read)
If __set_errno returns int, it will set w0 to 0xFFFFFFFF, which means
x0 is 0x00000000FFFFFFFF. When interpreted as a ssize_t that is
INT_MAX, not -1.
Change __set_errno to return long, which will cause x0 to be set instead
of w0.
Change-Id: I9f9ea0f2995928d2ea240eb2ff7758ecdf0ff412
These were needed when bionic's header files were missing these macros (though
it would have made a lot more sense to just fix the header files!) but cause
warnings now.
Change-Id: I65a677122f4f6bd07dffc3f37a0c4c0e823d1bb0
We don't need quite so much duplication because we already have a way
to get the signal number from its name, and that already copes with the
fact that the mips/mips64 numbers are different from everyone else's.
Also remove sys_signame from LP64. glibc doesn't have this BSD-ism.
Change-Id: I6dc411a3d73589383c85d3b07d9d648311492a10
1. Moved arch-specific setup to their own files:
- <arch>/<arch>.mk, arch-specific configs. Variables in those config
end with the arch name.
- removed the extra complexity introduced by function libc-add-cpu-variant-src,
which seems to be not very useful these days.
2. Separated out the crt object files generation rules and set up the
rules for both TARGET_ARCH and TARGET_2ND_ARCH.
3. Build all the libraries for both TARGET_ARCH and TARGET_2ND_ARCH,
with the arch-specific LOCAL_ variables.
Bug: 11654773
Change-Id: I9c2d85db0affa49199d182236d2210060a321421
Our sigset_t definition hasn't been tied to our NSIG definition since we
switched to uapi headers, so we can now fix it without breaking the LP32 ABI.
The kernel uapi headers define and use _NSIG, so we need to have our scripts
rename the kernel's definitions out of the way, then we can define _NSIG
and NSIG in terms of the kernel's off-by-one value.
Bug: 12938442
Change-Id: Ic7c86fd5be5ad1d822f7b2b1d88c8a0d70a1ac0f
No cacheflush for LP64; use the GCC builtin instead. Clean up the
32-bit MIPS implementation now we no longer need to worry about
old versions of GCC.
Bug: 12924756
Change-Id: Ie23955b3ec194e226c4b2bce35b11d5e061f4753
Remove the linker's reliance on BSD cruft and use the glibc-style
ElfW macro. (Other code too, but the linker contains the majority
of the code that needs to work for Elf32 and Elf64.)
All platforms need dl_iterate_phdr_static, so it doesn't make sense
to have that part of the per-architecture configuration.
Bug: 12476126
Change-Id: I1d7f918f1303a392794a6cd8b3512ff56bd6e487
Also move some of the stuff that should be in <link.h> out of the
private "linker.h", to make it clearer that these are public API
known to gdb that we can't change.
Bug: 12554197
Change-Id: I830e1260d3d8b833ed99bc1518f1c6b6102be8af
libc/libm support for MIPS64 targets
Change-Id: I8271941d418612a286be55495f0e95822f90004f
Signed-off-by: Chris Dearman <chris.dearman@imgtec.com>
Signed-off-by: Raghu Gandham <raghu.gandham@imgtec.com>
Most of <machine/_types.h> was either unused, wrong, or identical across
all 32-/64-bit architectures.
I'm not a huge fan of <sys/_types.h> either, but moving the bits we need
up into there is a step forward.
Bug: 12213562
Change-Id: Id13551c78966e324beee2dd90c5575e37d2a71e6
libunwind has #define inline /* empty */ which breaks our fortified headers.
glibc uses __inline but our BSD-derived headers often override that. __inline__
is the third alternative understood by GCC that -- as far as I know -- neither
the C library itself nor third-party code tries to mess with.
Bug: 12871594
Change-Id: I6677e70ea531bb7d4c46021b43af760d4ad8ecf7
As suggested here: https://android-review.googlesource.com/#/c/71267/
it may be used for x86_64 libunwind enabling.
Change-Id: I21623261a48ea7099e030d33932556e294d226ff
Signed-off-by: Pavel Chupin <pavel.v.chupin@intel.com>
We don't actually need to worry about sign extension if we reject
negative values ourselves. Previously it was possible to come up
with negative but aligned values that we would pass to the kernel;
in the case of mmap (as opposed to mmap64) we'd incorrectly turn
those into large positive offsets.
Change-Id: I2aa583e0f892d59bb77429aea8730b72db32dcb0
The various committees decided that everyone should get all these macros,
all the time.
Bug: 12708004
Change-Id: Ib56010dcba9b0656e5701546fefb7f78dc0bf916
This is required to make the Nexus 10 graphics driver work on a system
compiled with gcc 4.9.
Change-Id: If3f3d488652a736d9ea3e583548d74fae3ffa902
Signed-off-by: Bernhard Rosenkränzer <Bernhard.Rosenkranzer@linaro.org>
These functions should print assertion violation messages and then
call abort(). They do really not return control flow afterwards.
Consider the declaration of the similar __assert_fail from glibc:
extern void __assert_fail (const char *__assertion,
const char *__file,
unsigned int __line,
const char *__function)
__THROW __attribute__ ((__noreturn__));
Bionic has __noreturn defined in sys/cdefs.h to be that GNU
noreturn attribute.
This patch has a practical value. Consider the following function:
void check(void* ptr) {
assert(ptr != NULL);
}
Without this patch applied, gcc (and presumably clang) shows even in
debug mode:
warning: unused parameter 'ptr' [-Wunused-parameter]
In release mode, NDEBUG is defined and assert() becomes a no-op, as
one should expect. Thus, the warning is shown correctly then.
Another code sample:
float array[2];
int i = 3;
...
assert(i < 2);
array[i] = 0;
gcc says,
warning: array subscript is below array bounds [-Warray-bounds]
In other words, without noreturn attribute, assertions do not
allow a compiler's static analyzer to properly understand
the preconditions.
Change-Id: I3be92e99787c528899cf243ed448c4730c00c45b
Signed-off-by: Vadim Markovtsev <gmarkhor@gmail.com>
This patch adds trivial implementations of the missing sys headers
needed by strace. All strace needs are the constants and structures,
so this is enough for now. We can come back and add the functions
if/when we ever need them.
Change-Id: Idb87c1a8b6b1c62f6e16ae94f147e1169722b48e
The situation here is a bit confusing. On 64-bit, rlimit and rlimit64 are
the same, and so getrlimit/getrlimit64, setrlimit/setrlimit64,
and prlimit/prlimit64 are all the same. On 32-bit, rlimit and rlimit64 are
different. 32-bit architectures other than MIPS go one step further by having
an even more limited getrlimit system call, so arm and x86 need to use
ugetrlimit instead of getrlimit. Worse, the 32-bit architectures don't have
64-bit getrlimit- and setrlimit-equivalent system calls, and you have to use
prlimit64 instead. There's no 32-bit prlimit system call, so there's no
easy implementation of that --- what should we do if the result of prlimit64
won't fit in a struct rlimit? Since 32-bit survived without prlimit/prlimit64
for this long, I'm not going to bother implementing prlimit for 32-bit.
We need the rlimit64 functions to be able to build strace 4.8 out of the box.
Change-Id: I1903d913b23016a2fc3b9f452885ac730d71e001
glibc has no <sys/dirent.h>. If we do have to bring this back, we
should probably just have one file #include the other.
Change-Id: I5c0bf9c03769daf3b23f69778e9f01f81c3de9ec
If glibc hadn't already done things this way round, I'd have
called the field sched_priority and the macro __sched_priority
since that would seem less likely to cause trouble, but glibc
source compatibility is probably more important.
Change-Id: I8a8a477f2aa87cae641069c5c84b4fcab3152a82
Modify the syscalls script to generate the cfi directives for x86
syscalls.
Update the x86 syscalls.
Change-Id: Ia1993dc714a7e79f917087fff8200e9a02c52603
Adds the TCPOPT_* constants from NetBSD. Note that the BSDs also have
TCPOPT_SIGNATURE, but Linux calls that TCPOPT_MD5SIG and glibc doesn't
have any corresponding constant yet, so let's wait until we see which name
wins out.
Change-Id: If53cdada5595285d9a7e7248ef74cd7502d804c0
32-bit Android's dev_t was wrong too. We can't fix that without ABI breakage,
but we can at least fix 64-bit Android. And add tests.
Bug: https://code.google.com/p/android/issues/detail?id=54966
Change-Id: Ie2e42cc042b78b669a1a44e55f959dbd9c52c5c9
This patch switches to using the uapi constants. It also adds the missing
setns system call, fixes sched_getcpu's error behavior, and fixes the
gensyscalls script now ARM is uapi-only too.
Change-Id: I8e16b1693d6d32cd9b8499e46b5d8b0a50bc4f1d
Also make the other architectures more similar to one another,
use NULL instead of 0 in calling code, and remove an unused #define.
Change-Id: I52b874afb6a351c802f201a0625e484df6d093bb
The caller is only required to allocate 16 bytes on the
stack for a0-a3. syscall is handling up to 6 arguments so
additional space is needed on the stack to avoid corrupting the
callers frame.
Change-Id: I054b31696decc3e17d9c70af18cd278b852235d1
From the release notes:
Changes affecting near-future time stamps:
Jordan switches back to standard time at 00:00 on December 20, 2013.
The 2006-2011 transition schedule is planned to resume in 2014.
(Thanks to Steffen Thorsen.)
Changes affecting past time stamps:
In 2004, Cuba began DST on March 28, not April 4.
(Thanks to Steffen Thorsen.)
Change-Id: I8f26cc50f6b571804a18ff2113b4a47a22bc56dd
This is needed if we use Clang to compile Bionic, which won't include
__popcountsi2 anymore as Clang generates inline instructions. However
prebuilt binary blobs still depend on libc.so to resolve __popcountsi2.
Change-Id: I9001a3884c4be250c0ceebcd79922783fae1a0b7
Even though code built with clang won't be fully fortified
and won't contain calls to our various helpers, binaries built
with GCC will.
Change-Id: I389b2f1e22a3e89b22aadedc46397bf704f9ca79
This patch changes the domain that the memory barrier operates on. Assumes
that the scope of bionic_atomic_barrier() does not include device memory,
memory shared with the GPU or any other memory external to the processor
cluster.
Change-Id: I291e741c98a64c86f3a3cf99811bbf1e714ac9aa
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
The bionic_atomic_cmpxchg() API states that the cmpxchg() will be done without
explicit memory barriers. LDAXR/STLXR semantics involve half barriers for
load/store.
This patch optimises cmpxchg() by using LDXR/STXR and avoiding unnecessary half
bariers. It also fixes the clobber list for all the bionic_atomic_*() functions.
Change-Id: Iae9468965785cfeeec791d52f1e8cbc524adb682
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
Don't use FORTIFY_SOURCE on functions which implement
FORTIFY_SOURCE, to avoid infinite recursion problems.
The previous patch only addressed one of the problems.
Bug: 12216860
Change-Id: I6f30ae7cb5b481be9942add18182ea4839d348a6
Because there was no default := for the aarch64 libc_crt_target_cflags,
the += was causing libc_crt_target_cflags to be recursively-defined
variable, which meant that when we were compiling crtbegin.c LOCAL_PATH
would be bionic/tests/ and we'd have -Ibionic/tests/include/ and find
none of our include files.
Also fix linking of pthread_debug.cpp, at least in the disabled mode.
The enabled mode was already broken for all architectures, and continues
to be broken after this change. It's been broken for long enough that
we might want to just remove it...
(aarch64 is using the FSF linker where arm uses the gold linker.)
Change-Id: I7db2e386694f6933db043138e6e97e5ae54d4174
This is a better solution than the old __warn_references because it's
a compile-time rather than link-time warning, it doesn't rely on something
that doesn't appear to be supported by gold (which is why you only used
to see these warnings on mips builds), and the errors refer to the exact
call site(s) rather than just telling you which object file contains a
reference to the bad function.
This is primarily so we can build bionic for aarch64; building libc.so
caused these warnings to fire (because link time is the wrong time) and
warnings are errors.
Change-Id: I5df9281b2a9d98b164a9b11807ea9472c6faa9e3
Addition of support for AArch64 in the linker64 target.
Change-Id: I8dfd9711278f6706063e91f626b6007ea7a3dd6e
Signed-off-by: Marcus Oakland <marcus.oakland@arm.com>
Previously we were checking against a positive errno which
would not be returned from a system call.
Change-Id: I8e3a36f6fbf5ccc2191a152a1def37e2d6f93124
This is the first patch out of a series of patches that add support for
AArch64, the new 64bit execution state of the ARMv8 Architecture. The
patches add support for LP64 programming model.
The patch adds:
* "arch-aarch64" to the architecture directories.
* "arch-aarch64/include" - headers used by libc
* "arch-aarch64/bionic":
- crtbegin, crtend support;
- aarch64 specific syscall stubs;
- setjmp, clone, vfork assembly files.
Change-Id: If72b859f81928d03ad05d4ccfcb54c2f5dbf99a5
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
This patch adds support for AArch64 atomic operations. Some
of the stubs use the lightweight store/load exclusive.
Change-Id: Iaf704d048b2dc15bf08cf8e4f0c3ea9f2052fe13
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
It looks like we can probably just use the generic GCC stuff instead;
the generated code looks pretty similar. We should come back to that.
These routines are only used by the pthread implementation, and
__bionic_atomic_inc isn't used, so we can remove it.
Change-Id: I8b5b8cb30a1b159f0e85c3675aee06ddef39b429
The original structure included four reserved 32-bit values. This
change adds these back into the structure so that the
__system_property_find_compat function will (again) process the system
properties correctly.
I fixed this bug a while back, but didn't remove it from the list,
could have added a better test, and could have written clearer code
that didn't require a comment.
Change-Id: Iebdf0f9a54537a7d5cbca254a5967b1543061f3d
For 64bit Architectures mmap() is equivalent to mmap64(). This patch
maps mmap64() to mmap() in a similar way to other syscalls that differ
based on the size of off_t and off64_t
Change-Id: If21b21ef71120bad23d9a608d02d4a7de5220a87
Signed-off-by: Serban Constantinescu <serban.constantinescu@arm.com>
Most callers won't check for EINVAL, so it's best to fail early.
GCC takes the nonnull attribute as a guarantee that an argument
won't be NULL, so these hacks were already ineffective, which is
how we found that at least one commercial game was using NULL
as if it's a mutex, but actually getting no-op behavior.
Bug: 11971278
Change-Id: I89646e043d931778805a8b692e07a34d076ee6bf
Since the ENTRY/END macros now have .cfi_startproc/.cfi_endproc, most of the
custom arm assembly has no unwind information. Adding the proper cfi directives
for these and removing the arm directives.
Update the gensyscalls.py script to add these cfi directives for the generated
assembly. Also fix the references to non-uapi headers to the proper uapi
header.
In addition, remove the kill.S, tkill.S, tgkill.S for arm since they are not
needed at all. The unwinder (libunwind) is able to properly unwind using the
normal abort.
After this change, I can unwind through the system calls again.
Bug: 11559337
Bug: 11825869
Bug: 11321283
Change-Id: I18b48089ef2d000a67913ce6febc6544bbe934a3
Unlike other architectures, on x86 (but not x86-64), CLONE_SETTLS
takes a pointer to a struct user_desc instead of a pointer to the
TLS itself. Rather than have to deal with this here, let's just use
the old __set_tls mechanism we used to use (and still use for the
main thread on all architectures, so it's not going away any time
soon).
Bug: 11826724
Change-Id: I02a27939a73ae6cea1134a3f4c1dd7eafea479da
Update headers and delete non-uapi headers that have been removed from
external/kernel-headers project.
Change-Id: I9ed9632a941095fe1bf3b207bafe1151f00de033
There is no uapi user.h file for arm, it was included by accident.
Move the user struct definition into the file to follow the pattern
used by the other architectures.
Change-Id: Ib9cea0deca551c9268382ddd6de9202fd32ef941
There are files in generated/asm that simply include asm-generic files. The
script now copies any file in generated/asm that also exists in asm-generic.
Change-Id: I075161c68624e9e9e81797224831988ce02220eb
Also fix the signature of usleep, and the definition of useconds_t which
should be unsigned, as the 'u' in its name implies.
This patch also cleans up the existing FreeBSD hacks by moving the libm
stuff from <sys/cdefs.h> to a libm-private header, and adding comments
about the hacks we use to build FreeBSD source.
Change-Id: Ibe5067a380502df94a0a3a7901969b35411085b6
The kernel now maintains the pthread_internal_t::tid field for us,
and __clone was only used in one place so let's inline it so we don't
have to leave such a dangerous function lying around. Also rename
files to match their content and remove some useless #includes.
Change-Id: I24299fb4a940e394de75f864ee36fdabbd9438f9
Let the kernel keep pthread_internal_t::tid updated, including
across forks and for the main thread. This then lets us fix
pthread_join to only return after the thread has really exited.
Also fix the thread attributes of the main thread so we don't
unmap the main thread's stack (which is really owned by the
dynamic linker and contains things like environment variables),
which fixes crashes when joining with an exited main thread
and also fixes problems reported publicly with accessing environment
variables after the main thread exits (for which I've added a new
unit test).
In passing I also fixed a bug where if the clone(2) inside
pthread_create(3) fails, we'd unmap the child's stack and TLS (which
contains the mutex) and then try to unlock the mutex. Boom! It wasn't
until after I'd uploaded the fix for this that I came across a new
public bug reporting this exact failure.
Bug: 8206355
Bug: 11693195
Bug: https://code.google.com/p/android/issues/detail?id=57421
Bug: https://code.google.com/p/android/issues/detail?id=62392
Change-Id: I2af9cf6e8ae510a67256ad93cad891794ed0580b
As of 61e699a133, stdio clean up
functions are no longer registered in atexit and must be called
manually via __cleanup.
The issue this fixes is some static binaries linked against bionic
cannot output properly when piped or redirected because the buffer
is not flushed before closing.
This is done by pulling in exit.c (and other dependencies) from
netbsd.
Change-Id: I193e54a6d08900f291550029fe75ce76394d9e22
Some MIPS kernels do not correctly restart interrupted system calls that
have been invoked using the indirect syscall (NR_syscall).
The simplest workaround is to handle the indirection in userland and then
call the required system call directly.
Change-Id: I8385399621529db9a52b463c96925f6decaaca30
The kernel doesn't have an nlink_t; it just uses the equivalent of
uint32_t. We already had a usable __nlink_t in the C library, so
let's just define our nlink_t in terms of __nlink_t, which is what
__nlink_t was meant for anyway.
Note that our struct stat just follows the kernel, and doesn't refer
to nlink_t anyway.
Change-Id: I2a56e418e42404b1741b08c50554b03c11caebae
<time.h> didn't need to copy the cruft from <signal.h>, and
<signal.h> only needs the uid_t hack when it's not using
uapi headers.
pthread_exit.cpp should include what it uses.
Change-Id: I836c36abe0f0a781d41fc425b249d1c7686bb124
We only need it for MAX_ERRNO, and it's time we had somewhere to put
the little assembler utility macros we've been putting off writing.
Change-Id: I9354d2e0dc47c689296a34b5b229fc9ba75f1a83
In practice, thanks to all the registers the stubs don't actually change,
but it's confusing to have an incorrect declaration.
I suspect that fcntl remains broken for aarch64; it happens to work for
x86_64 because the first vararg argument gets placed in the right register
anyway, but I have no reason to believe that's true for aarch64.
This patch adds a unit test, though, so we'll be able to tell when we get
as far as running the unit tests.
Change-Id: I58dd0054fe99d7d51d04c22781d8965dff1afbf3
Unlike on 32-bit systems where off_t is 32-bit, we don't want to
throw away the top 32 bits of an LP64 system's 64-bit off_t.
Change-Id: Ib2e0daeb4fc0b8ab3d1b983d0b371d8f81033b50
The old code ignored operator precedence (!), despite having two tables
of operator precedence. The code's still pretty awful, but I've cleaned
it up enough to fix this, the most important bug.
This patch lets us correctly clean the uapi unistd.h, stat.h, and swab.h files,
and also fixes the mess we were already making of various old kernel
header files. I've added a bunch more tests, fixed the existing tests that
the existing script was already failing (!), and changed the script so that
the tests are run every time the script is run.
We can probably remove some of the old kernel header files that we were
parsing incorrectly, but we can worry about that later.
Bug: 11253477
Change-Id: Ie66c65b3a7ae13b4e98ed8038a6a534f06eae0e5
