OSDN Git Service

android-x86/bionic.git
8 years agoMerge "Bionic: Add a TLS slot for ART for Thread self"
Andreas Gampe [Tue, 4 Aug 2015 22:35:34 +0000 (22:35 +0000)]
Merge "Bionic: Add a TLS slot for ART for Thread self"

8 years agoMerge "name the arc4random structure mappings"
Elliott Hughes [Tue, 4 Aug 2015 00:58:57 +0000 (00:58 +0000)]
Merge "name the arc4random structure mappings"

8 years agoMerge "Cherrypick upstream fnmatch.c fix."
Elliott Hughes [Mon, 3 Aug 2015 18:22:42 +0000 (18:22 +0000)]
Merge "Cherrypick upstream fnmatch.c fix."

8 years agoCherrypick upstream fnmatch.c fix.
Elliott Hughes [Mon, 3 Aug 2015 18:11:22 +0000 (11:11 -0700)]
Cherrypick upstream fnmatch.c fix.

Bug: http://b/22855370
Change-Id: I3865cff1e2ffeb6709cea092be3770c1bebf1623

8 years agoMerge "[MIPS] Link .dex or .oat code lacking .MIPS.abiflags segment"
Elliott Hughes [Mon, 3 Aug 2015 15:17:18 +0000 (15:17 +0000)]
Merge "[MIPS] Link .dex or .oat code lacking .MIPS.abiflags segment"

8 years agoname the arc4random structure mappings
Daniel Micay [Sun, 2 Aug 2015 02:25:42 +0000 (22:25 -0400)]
name the arc4random structure mappings

Change-Id: I6e150e526f6ceee31aa9bd219fa45317d5a729cb

8 years agoMerge "add a fortified implementation of getcwd"
Yabin Cui [Sat, 1 Aug 2015 05:14:20 +0000 (05:14 +0000)]
Merge "add a fortified implementation of getcwd"

8 years agoadd a fortified implementation of getcwd
Daniel Micay [Wed, 20 May 2015 19:31:26 +0000 (15:31 -0400)]
add a fortified implementation of getcwd

Change-Id: Ice3e6d3e1ff07788305dc85f8ee4059baad5fac4

8 years ago[MIPS] Link .dex or .oat code lacking .MIPS.abiflags segment
Duane Sand [Tue, 28 Jul 2015 21:04:29 +0000 (14:04 -0700)]
[MIPS] Link .dex or .oat code lacking .MIPS.abiflags segment

This corrects an issue with mips32 Art on mips64r6 Android, where
Java ran slowly due to unintended use of kernel-trap emulation of
single-precision floating point registers. This also regressed all
Art tests due to an extra logcat line
    WARNING: linker: Using FRE=1 mode to run "..."

When targeting mips32r6, Art generates modeless or FR=1 floating point
code, same as Android's own native mips32r6 modules. So the trapping was
unneeded. Linker was confusing Art-generated modules with those from
old NDK compilers, which do need that trapping mode.

This linker filename check may become unnecessary, if Art learns how to
generate .MIPS.abiflags segments in its generated elf-like codefiles.

Change-Id: I18069d1234960c680c5df739514da09015a7fdb6

8 years agoMerge "Sync current upstream fnmatch.c."
Elliott Hughes [Fri, 31 Jul 2015 18:04:38 +0000 (18:04 +0000)]
Merge "Sync current upstream fnmatch.c."

8 years agoSync current upstream fnmatch.c.
Elliott Hughes [Fri, 31 Jul 2015 17:57:27 +0000 (10:57 -0700)]
Sync current upstream fnmatch.c.

Bug: http://b/22855370
Change-Id: I589f849aaede3b69b6b3682989bab7d8649dcd97

8 years agoMerge "Revert "add a fortified implementation of getcwd""
Elliott Hughes [Fri, 31 Jul 2015 17:33:05 +0000 (17:33 +0000)]
Merge "Revert "add a fortified implementation of getcwd""

8 years agoRevert "add a fortified implementation of getcwd"
Elliott Hughes [Fri, 31 Jul 2015 17:32:50 +0000 (17:32 +0000)]
Revert "add a fortified implementation of getcwd"

This reverts commit 89a121d43a274131cf9f84b0e685bebf1e70cc1b.

Change-Id: Ib6f6fd424acea68a063777e9f324405450e23316

8 years agoMerge "add a fortified implementation of getcwd"
Yabin Cui [Fri, 31 Jul 2015 16:55:12 +0000 (16:55 +0000)]
Merge "add a fortified implementation of getcwd"

8 years agoadd a fortified implementation of getcwd
Daniel Micay [Wed, 20 May 2015 19:31:26 +0000 (15:31 -0400)]
add a fortified implementation of getcwd

Change-Id: I95001ae4fe8f206db83e5c44d129ba11310695ce

8 years agoMerge "Stop sending SIGPIPE to debuggerd."
Elliott Hughes [Thu, 30 Jul 2015 15:10:59 +0000 (15:10 +0000)]
Merge "Stop sending SIGPIPE to debuggerd."

8 years agoStop sending SIGPIPE to debuggerd.
Elliott Hughes [Thu, 30 Jul 2015 05:24:13 +0000 (22:24 -0700)]
Stop sending SIGPIPE to debuggerd.

SIGPIPE is a pretty normal way for command-line apps to die, but because
we catch it and report it via debuggerd, we get a lot of bogus bugs. We
could catch SIGPIPE in our tools, but that's not really legit and slightly
misleading.

"But", you say, "catching SIGPIPE is useful for app bugs!". Except a trawl
through buganizer suggests it's misleading there too. Not least because
it's usually an innocent victim that dies --- the problem is usually on the
other end of the pipe (which you learn nothing about because that process
already died, which is what closed the pipe).

We also don't catch SIGALRM, which is another signal that will terminate
your process if you don't catch it, but that one actually represents a
logic error in the crashing process, so there's a stronger argument for
catching that. (Except it too is not a real source of bugs.)

Bug: http://b/20659371
Change-Id: I79820b36573ddaa9a7bad0561a52f23e7a8d15ac

8 years agoMerge "Prevent buffer over-read in linker.cpp's parse_path."
Elliott Hughes [Thu, 30 Jul 2015 00:22:48 +0000 (00:22 +0000)]
Merge "Prevent buffer over-read in linker.cpp's parse_path."

8 years agoPrevent buffer over-read in linker.cpp's parse_path.
tony.ys_liu [Wed, 29 Jul 2015 10:00:22 +0000 (18:00 +0800)]
Prevent buffer over-read in linker.cpp's parse_path.

Also, the old behavior of skipping empty entries doesn't match glibc.

Change-Id: I497774377113ab6c5d962e0f20066e2192748f06

8 years agoMerge "Remove PAGE_SIZE from <limits.h>."
Elliott Hughes [Wed, 29 Jul 2015 15:18:06 +0000 (15:18 +0000)]
Merge "Remove PAGE_SIZE from <limits.h>."

8 years agoRemove PAGE_SIZE from <limits.h>.
Elliott Hughes [Wed, 29 Jul 2015 02:52:31 +0000 (19:52 -0700)]
Remove PAGE_SIZE from <limits.h>.

It turns out that everyone's still getting PAGE_SIZE from <sys/user.h> via
<sys/ucontext.h> via <signal.h> anyway.

glibc has PAGE_SIZE in <sys/user.h> rather than <limits.h> so this part is
good. The bad part is that we have such wide transitive inclusion of
<sys/user.h>!

Bug: http://b/22735893
Change-Id: I363adffe4a27b4ca1eedf695ea621f5dd2d5ca10

8 years agoBionic: Add a TLS slot for ART for Thread self
Andreas Gampe [Wed, 29 Jul 2015 02:53:54 +0000 (19:53 -0700)]
Bionic: Add a TLS slot for ART for Thread self

To improve the performance of Thread::Current(), add a tls slot so
ART can avoid a pthreads call.

Change-Id: Icc86a2b7590734637366f9d5e41a5c6d18cc5772

8 years agoMerge "Don't transitively include <limits.h> from <sys/user.h>."
Elliott Hughes [Wed, 29 Jul 2015 01:02:37 +0000 (01:02 +0000)]
Merge "Don't transitively include <limits.h> from <sys/user.h>."

8 years agoDon't transitively include <limits.h> from <sys/user.h>.
Elliott Hughes [Wed, 29 Jul 2015 00:44:32 +0000 (17:44 -0700)]
Don't transitively include <limits.h> from <sys/user.h>.

Bug: http://b/22735893
Change-Id: I27d0a1f7e2d477b8038e76bfc7fe05dd7e187344

8 years agoMerge "Remove PAGESIZE."
Elliott Hughes [Wed, 29 Jul 2015 00:09:20 +0000 (00:09 +0000)]
Merge "Remove PAGESIZE."

8 years agoRemove PAGESIZE.
Elliott Hughes [Tue, 28 Jul 2015 23:46:06 +0000 (16:46 -0700)]
Remove PAGESIZE.

This was pretty much unused, and isn't in glibc.

Bug: http://b/22735893
Change-Id: If17f0dcd931c90ef1ccb134a3950c3b0011a03f4

8 years agoMerge "Move PAGE_MASK into <sys/user.h>."
Elliott Hughes [Tue, 28 Jul 2015 23:11:18 +0000 (23:11 +0000)]
Merge "Move PAGE_MASK into <sys/user.h>."

8 years agoMove PAGE_MASK into <sys/user.h>.
Elliott Hughes [Tue, 28 Jul 2015 21:58:37 +0000 (14:58 -0700)]
Move PAGE_MASK into <sys/user.h>.

I'm removing the TODO on the assumption that being compatible with glibc
is more useful than BSD. The new internal "bionic_page.h" header factors
out some duplication between libc and the linker.

Bug: http://b/22735893
Change-Id: I4aec4dcba5886fb6f6b9290a8f85660643261321

8 years agoMerge "deprecate TARGET_USES_LOGD"
Mark Salyzyn [Tue, 28 Jul 2015 17:48:08 +0000 (17:48 +0000)]
Merge "deprecate TARGET_USES_LOGD"

8 years agodeprecate TARGET_USES_LOGD
Mark Salyzyn [Tue, 28 Jul 2015 15:51:17 +0000 (08:51 -0700)]
deprecate TARGET_USES_LOGD

This is not the kernel logger you are looking for

Bug: 22787659
Change-Id: I340d8bb5cdaa73be9565521681ee238b7033934b

8 years agoMerge "Add a regression test for a fixed strnlen bug."
Elliott Hughes [Tue, 28 Jul 2015 14:55:50 +0000 (14:55 +0000)]
Merge "Add a regression test for a fixed strnlen bug."

8 years agoAdd a regression test for a fixed strnlen bug.
Elliott Hughes [Tue, 28 Jul 2015 03:55:03 +0000 (20:55 -0700)]
Add a regression test for a fixed strnlen bug.

Bug: https://code.google.com/p/android/issues/detail?id=74741
Change-Id: I78b0114bdbe8e680b0e938af608a634e73c86eda

8 years agoMerge "Add one simple thread local storage test."
Chih-hung Hsieh [Tue, 28 Jul 2015 01:33:01 +0000 (01:33 +0000)]
Merge "Add one simple thread local storage test."

8 years agoAdd one simple thread local storage test.
Chih-Hung Hsieh [Mon, 27 Jul 2015 17:46:21 +0000 (10:46 -0700)]
Add one simple thread local storage test.

This test should pass with gcc and fail with clang,
until clang/llvm supports emutls or Android linker supports
ELF TLS models.

BUG: 21082792
Change-Id: Id8c97da52c68ec230c7d26af703f6ed32d53a4fe

8 years agoMerge "Remove pushes from memsets (krait/cortex-a9)."
Christopher Ferris [Mon, 27 Jul 2015 23:09:32 +0000 (23:09 +0000)]
Merge "Remove pushes from memsets (krait/cortex-a9)."

8 years agoMerge "add fortified implementations of fread/fwrite"
Yabin Cui [Mon, 27 Jul 2015 22:46:29 +0000 (22:46 +0000)]
Merge "add fortified implementations of fread/fwrite"

8 years agoMerge "Restore protection flags for ifunc during relocs."
Elliott Hughes [Mon, 27 Jul 2015 21:34:53 +0000 (21:34 +0000)]
Merge "Restore protection flags for ifunc during relocs."

8 years agoRemove pushes from memsets (krait/cortex-a9).
Christopher Ferris [Mon, 27 Jul 2015 20:51:31 +0000 (13:51 -0700)]
Remove pushes from memsets (krait/cortex-a9).

On the path that only uses r0 in both the krait and cortex-a9
memset, remove the push and use r3 instead.

In addition, for cortex-a9, remove the artificial function since
it's not needed since dwarf unwinding is now supported on arm.

Change-Id: Ia4ed1cc435b03627a7193215e76c8ea3335f949a

8 years agoMerge "name the atexit handler pages"
Elliott Hughes [Mon, 27 Jul 2015 18:06:35 +0000 (18:06 +0000)]
Merge "name the atexit handler pages"

8 years agoname the atexit handler pages
Daniel Micay [Sat, 25 Jul 2015 19:40:14 +0000 (15:40 -0400)]
name the atexit handler pages

Change-Id: I1718ddee7415a673b5818bc1455b8d90f1e8f9c6

8 years agoRestore protection flags for ifunc during relocs.
Dmitriy Ivanov [Sun, 26 Jul 2015 14:37:09 +0000 (07:37 -0700)]
Restore protection flags for ifunc during relocs.

IFUNC relocations require executable flag for the load
segment containing .text. When dso has text relocs linker
removes exec which causes crash during ifunc relocations.

This patch fixes this problem by restoring segments protection
for ifunc relocs.

Bug: http://b/22611399
Change-Id: Icbf3be0fec0e42bf805bcad7533e2032a2e11b9c
(cherry picked from commit de0fb393ae8136a5958fe17eee0c6285e2f7f91a)

8 years agoMerge "Use AT_PAGESZ for sysconf(_SC_PAGE_SIZE)."
Elliott Hughes [Sat, 25 Jul 2015 18:32:46 +0000 (18:32 +0000)]
Merge "Use AT_PAGESZ for sysconf(_SC_PAGE_SIZE)."

8 years agoUse AT_PAGESZ for sysconf(_SC_PAGE_SIZE).
Elliott Hughes [Sat, 25 Jul 2015 06:22:07 +0000 (23:22 -0700)]
Use AT_PAGESZ for sysconf(_SC_PAGE_SIZE).

Bug: http://b/18342333
Change-Id: Id12ed4e85a0f35d7d27202f7792df42a65a74b4e

8 years agoMerge "Make all labels local."
Christopher Ferris [Fri, 24 Jul 2015 00:47:06 +0000 (00:47 +0000)]
Merge "Make all labels local."

8 years agoadd fortified implementations of fread/fwrite
Daniel Micay [Sat, 18 Jul 2015 17:55:51 +0000 (13:55 -0400)]
add fortified implementations of fread/fwrite

A __size_mul_overflow utility is used to take advantage of the checked
overflow intrinsics in Clang and GCC (>= 5). The fallback for older
compilers is the optimized but less than ideal overflow checking pattern
used in OpenBSD.

Change-Id: Ibb0d4fd9b5acb67983e6a9f46844c2fd444f7e69

8 years agoMake all labels local.
Christopher Ferris [Thu, 23 Jul 2015 19:12:55 +0000 (12:12 -0700)]
Make all labels local.

Change the non-local labels to .L labels.

Change-Id: I720e894f2e311af8f4a0970303d8b86575fb69a5

8 years agoMerge "Don't abort when failed to write tracing message."
Yabin Cui [Thu, 23 Jul 2015 17:20:30 +0000 (17:20 +0000)]
Merge "Don't abort when failed to write tracing message."

8 years agoDon't abort when failed to write tracing message.
Yabin Cui [Wed, 22 Jul 2015 00:27:54 +0000 (17:27 -0700)]
Don't abort when failed to write tracing message.

Also make the code thread-safe with lock.

Bug: 20666100
Change-Id: I0f331a617b75280f36179c187418450230d713ef

8 years agoMerge "Add getgrgid_r/getgrnam_r."
Yabin Cui [Wed, 22 Jul 2015 03:22:26 +0000 (03:22 +0000)]
Merge "Add getgrgid_r/getgrnam_r."

8 years agoAdd getgrgid_r/getgrnam_r.
Yabin Cui [Tue, 21 Jul 2015 02:46:26 +0000 (19:46 -0700)]
Add getgrgid_r/getgrnam_r.

Bug: 22568551
Change-Id: I3c0772d119d6041063c6be53f5bcc5ea1768f0d5

8 years agoMerge "libc: arch-x86: implement kernel vdso time functions"
Elliott Hughes [Wed, 22 Jul 2015 01:51:00 +0000 (01:51 +0000)]
Merge "libc: arch-x86: implement kernel vdso time functions"

8 years agolibc: arch-x86: implement kernel vdso time functions
Robert Jarzmik [Wed, 15 Jul 2015 13:26:43 +0000 (15:26 +0200)]
libc: arch-x86: implement kernel vdso time functions

This patch give the possibility of time vdso support on 32bit kernel.
If the 32bit x86 kernel provides gettimeofday() and clock_gettime()
primitives in vdso. In this case make bionic use them. If the kernel
doesn't provide them, fallback to the legacy system call versions.

Change-Id: I87b772a9486fa356903e1f98f486ab9eb0b6f6f7
Signed-off-by: Robert Jarzmik <robert.jarzmik@intel.com>
Signed-off-by: Mingwei Shi <mingwei.shi@intel.com>
9 years agoMerge "We don't have CXX_BARE any more"
Ying Wang [Tue, 21 Jul 2015 21:10:22 +0000 (21:10 +0000)]
Merge "We don't have CXX_BARE any more"

9 years agoWe don't have CXX_BARE any more
Ying Wang [Tue, 21 Jul 2015 01:51:30 +0000 (18:51 -0700)]
We don't have CXX_BARE any more

Now we have split CXX_WRAPPER from TARGET_CXX and CXX_BARE isn't needed
any more.

Bug: 22612634
Change-Id: I52c78b0d1b325910e875a786d17f780731778b4b
(cherry-pick from commit 9d11a7087c860195bf23f2825a7fe70d43ff7b78)

9 years agoMerge "Use a less misleading name for the code that sets up the main thread."
Elliott Hughes [Tue, 21 Jul 2015 20:28:15 +0000 (20:28 +0000)]
Merge "Use a less misleading name for the code that sets up the main thread."

9 years agoUse a less misleading name for the code that sets up the main thread.
Elliott Hughes [Tue, 21 Jul 2015 18:57:09 +0000 (11:57 -0700)]
Use a less misleading name for the code that sets up the main thread.

Change-Id: I50c1b0a3b633cf8bc40a6bd86f12adb6b91e2888

9 years agoMerge "Add support for non-zero vaddr in maps."
Christopher Ferris [Tue, 21 Jul 2015 17:29:09 +0000 (17:29 +0000)]
Merge "Add support for non-zero vaddr in maps."

9 years agoMerge "Revert "Revert "make vdso function pointers read-only at runtime"""
Elliott Hughes [Tue, 21 Jul 2015 02:15:19 +0000 (02:15 +0000)]
Merge "Revert "Revert "make vdso function pointers read-only at runtime"""

9 years agoMerge "Revert "Stop libc from cross-referencing unwind symbols""
Dan Albert [Tue, 21 Jul 2015 00:22:20 +0000 (00:22 +0000)]
Merge "Revert "Stop libc from cross-referencing unwind symbols""

9 years agoRevert "Stop libc from cross-referencing unwind symbols"
Dan Albert [Mon, 20 Jul 2015 23:56:51 +0000 (16:56 -0700)]
Revert "Stop libc from cross-referencing unwind symbols"

This reverts commit cd13b14e98d4921af126667fae0cf6613a5615c5.

Bug: http://b/19958712
Change-Id: I7cc7f69728c42c37e129aee30d761c4cd7e30e94

9 years agoRevert "Revert "make vdso function pointers read-only at runtime""
Elliott Hughes [Mon, 20 Jul 2015 22:34:27 +0000 (22:34 +0000)]
Revert "Revert "make vdso function pointers read-only at runtime""

This reverts commit 1946856b1f18a27c51ba30bb9e304f25b722ee05.

This goes back to the original scheme of PROT_NONEing a page within
libc. Allocating a new page didn't fail safe for cases where these
functions are called from the dynamic linker.

Bug: http://b/22568628
Change-Id: I3e7241c8b54c27ea4a898bc952375c1e9ae38c80

9 years agoMerge "Fix dt_runpath test."
Dmitriy Ivanov [Mon, 20 Jul 2015 18:31:07 +0000 (18:31 +0000)]
Merge "Fix dt_runpath test."

9 years agoFix dt_runpath test.
Dmitriy Ivanov [Mon, 20 Jul 2015 17:47:49 +0000 (10:47 -0700)]
Fix dt_runpath test.

Bfd linker produces DT_RPATH instead of DT_RUNPATH
unless --enable-new-dtags option is specified.

Bug: http://b/22560945
Change-Id: I71bdc5137b899ab736e55c5a9574cfbba8e1983b

9 years agoMerge "<sys/time.h> should include [most of] <sys/select.h>."
Elliott Hughes [Mon, 20 Jul 2015 17:36:52 +0000 (17:36 +0000)]
Merge "<sys/time.h> should include [most of] <sys/select.h>."

9 years agoMerge "Revert "make vdso function pointers read-only at runtime""
Elliott Hughes [Mon, 20 Jul 2015 17:30:49 +0000 (17:30 +0000)]
Merge "Revert "make vdso function pointers read-only at runtime""

9 years agoRevert "make vdso function pointers read-only at runtime"
Elliott Hughes [Mon, 20 Jul 2015 17:30:33 +0000 (17:30 +0000)]
Revert "make vdso function pointers read-only at runtime"

This reverts commit df1a3c6d21702e3e96cfcddadee4a50bfac82110.

This change prevented N9 from booting (http://b/22568628).

Change-Id: I071d6d6a0ae7881d65641839e665acdcf58462b4

9 years ago<sys/time.h> should include [most of] <sys/select.h>.
Elliott Hughes [Sat, 18 Jul 2015 02:10:24 +0000 (19:10 -0700)]
<sys/time.h> should include [most of] <sys/select.h>.

Found by trying to build c-ares with bionic.

Change-Id: I38684abcb806ef1e8806c68b4a6dd9200f48849e

9 years agoMerge "Add all the glibc one-line "synonym" header files."
Elliott Hughes [Fri, 17 Jul 2015 21:42:03 +0000 (21:42 +0000)]
Merge "Add all the glibc one-line "synonym" header files."

9 years agoAdd all the glibc one-line "synonym" header files.
Elliott Hughes [Fri, 17 Jul 2015 20:59:36 +0000 (13:59 -0700)]
Add all the glibc one-line "synonym" header files.

Ported code keeps needing these one by one, so let's just add the whole
set...

Change-Id: I9662b256aeaa1b0ffb22354130e503f0c4aa472b

9 years agoMerge "make vdso function pointers read-only at runtime"
Elliott Hughes [Fri, 17 Jul 2015 18:13:31 +0000 (18:13 +0000)]
Merge "make vdso function pointers read-only at runtime"

9 years agomake vdso function pointers read-only at runtime
Daniel Micay [Fri, 17 Jul 2015 16:13:27 +0000 (12:13 -0400)]
make vdso function pointers read-only at runtime

Global, writable function pointers are low-hanging fruit for hijacking
control flow with an overflow from a global buffer or an arbitrary write
vulnerability. This moves the function pointer table into a dedicated
page and makes it read-only at runtime, similar to RELRO.

This increases the memory usage of the library by just under one page.
This could be avoided by having the linker load the vdso by replacing
weak symbols. It's not significant within the Zygote spawning model
though because it's read-only after early init.

Change-Id: Id7a49c96c1b15c2e1926528304b3c54a81e78caf

9 years agoMerge "Fix potential race condition on dlopen"
Dmitriy Ivanov [Fri, 17 Jul 2015 17:49:20 +0000 (17:49 +0000)]
Merge "Fix potential race condition on dlopen"

9 years agoFix potential race condition on dlopen
Dmitriy Ivanov [Fri, 17 Jul 2015 17:36:10 +0000 (10:36 -0700)]
Fix potential race condition on dlopen

  Call to find_containing_library should be guarded.

Change-Id: I985a903da48b83bcd35e957a979158eb9b80e70b

9 years agoAdd support for non-zero vaddr in maps.
Christopher Ferris [Thu, 16 Jul 2015 21:49:17 +0000 (14:49 -0700)]
Add support for non-zero vaddr in maps.

If a map has a non-zero vaddr then it needs to be added to the
computed relative offset.

Bug: 22532054
Change-Id: If55015165d25ecc4b530f674b481c8c7ef760a23

9 years agoMerge "Add timespec/timeval conversion helpers."
Bertrand SIMONNET [Thu, 16 Jul 2015 21:53:25 +0000 (21:53 +0000)]
Merge "Add timespec/timeval conversion helpers."

9 years agoAdd timespec/timeval conversion helpers.
Bertrand SIMONNET [Thu, 16 Jul 2015 18:50:39 +0000 (11:50 -0700)]
Add timespec/timeval conversion helpers.

TIMESPEC_TO_TIMEVAL and TIMEVAL_TO_TIMESPEC are usually provided by
sys/time.h.

BUG: 22532790

Change-Id: I8b9d7298e3250323faa20b1adf9a7514954c6096

9 years agoMerge "Add constant for zip separator"
Dmitriy Ivanov [Thu, 16 Jul 2015 15:59:15 +0000 (15:59 +0000)]
Merge "Add constant for zip separator"

9 years agoAdd constant for zip separator
Dmitriy Ivanov [Thu, 16 Jul 2015 11:52:06 +0000 (04:52 -0700)]
Add constant for zip separator

Change-Id: I1dcbb15ac612fc28e922eb3de8148fcaebbc0820

9 years agoMerge "Replace spaces with tabs in a Makefile."
Evgenii Stepanov [Wed, 15 Jul 2015 23:22:38 +0000 (23:22 +0000)]
Merge "Replace spaces with tabs in a Makefile."

9 years agoReplace spaces with tabs in a Makefile.
Evgenii Stepanov [Wed, 15 Jul 2015 23:19:59 +0000 (16:19 -0700)]
Replace spaces with tabs in a Makefile.

Sorry, I undertested my previous change and these guys somehow sneaked in.
This only affects SANITIZE_TARGET builds.

Change-Id: Id6986d89fa3d4965339f7b6fe973e9db73b380da

9 years agoMerge "A special linker for ASan executables."
Evgenii Stepanov [Wed, 15 Jul 2015 17:24:19 +0000 (17:24 +0000)]
Merge "A special linker for ASan executables."

9 years agoA special linker for ASan executables.
Evgenii Stepanov [Sat, 11 Jul 2015 00:54:01 +0000 (17:54 -0700)]
A special linker for ASan executables.

Setup a /system/bin/linker_asan as a symlink to "linker".
Read the linker name from PT_INTERP, and if it is linker_asan,
switch default library lookup paths to the ASan set, which
starts with the path to the instrumented libraries
(/data/lib), followed by /system/lib as a fallback.

This ensures that ASan binaries prefer ASan libraries, when
available. This approach is way better then RPATH/RUNPATH and even
better than LD_LIBRARY_PATH:
- RUNPATH is per-DSO, while default paths are global.
- LD_LIBRARY_PATH is overwritten by android_update_LD_LIBRARY_PATH.
- neither RUNPATH nor LD_LIBRARY_PATH appear in
  android_get_LD_LIBRARY_PATH which is used to build java.lang.path.
  Having ASan libraries in java.lang.path is a good thing.

Bug: 22355945
Change-Id: I1d2791fbf5740618f18f71a3ae3d873714669d3f

9 years agoMerge "Add test for an Android kernel bug about prctl."
Yabin Cui [Tue, 14 Jul 2015 02:48:44 +0000 (02:48 +0000)]
Merge "Add test for an Android kernel bug about prctl."

9 years agoAdd test for an Android kernel bug about prctl.
Yabin Cui [Mon, 13 Jul 2015 23:54:29 +0000 (16:54 -0700)]
Add test for an Android kernel bug about prctl.

Bug: 20017123

Change-Id: I8c109a06999a8bffcc8e8298b04ee637bb67d878

9 years agoMerge "Update to v3.18.18 kernel headers."
Christopher Ferris [Tue, 14 Jul 2015 01:29:46 +0000 (01:29 +0000)]
Merge "Update to v3.18.18 kernel headers."

9 years agoUpdate to v3.18.18 kernel headers.
Christopher Ferris [Tue, 14 Jul 2015 00:21:18 +0000 (17:21 -0700)]
Update to v3.18.18 kernel headers.

Change-Id: I0a0f423bda89500ad8e13a80582f05ec8bc756ea

9 years agoMerge "Do not hold hash table lock while backtracing."
Christopher Ferris [Mon, 13 Jul 2015 20:27:37 +0000 (20:27 +0000)]
Merge "Do not hold hash table lock while backtracing."

9 years agoDo not hold hash table lock while backtracing.
Christopher Ferris [Sun, 12 Jul 2015 05:47:06 +0000 (22:47 -0700)]
Do not hold hash table lock while backtracing.

There is a deadlock if holding the hash table lock while trying to do
a backtrace. Change the code so that the hash table lock is only held
while actually modifying either g_hash_table, or while modifying an
entry from g_hash_table.

Bug: 22423683
(cherry picked from commit 9fee99b06013787054a312449b94115038e2ad7c)

Change-Id: I72173bfe6f824ceaceea625c24e7851b87467135

9 years agoMerge "Prevent user-defined basename_r from breaking basename(3)."
Elliott Hughes [Sat, 11 Jul 2015 07:03:53 +0000 (07:03 +0000)]
Merge "Prevent user-defined basename_r from breaking basename(3)."

9 years agoPrevent user-defined basename_r from breaking basename(3).
Elliott Hughes [Sat, 11 Jul 2015 06:58:59 +0000 (23:58 -0700)]
Prevent user-defined basename_r from breaking basename(3).

LP64 is immune because basename_r is hidden there, but on LP32 a basename_r
defined in the executable breaks basename because its call to basename_r
will resolve to that one rather than the one in libc.

Bug: http://b/22415484
Change-Id: Ied3ca7ad3fb0e744eb705fc924743f893b4ad490

9 years agoMerge "Fix memory-leak on soinfo_free()"
Dmitriy Ivanov [Thu, 9 Jul 2015 23:32:18 +0000 (23:32 +0000)]
Merge "Fix memory-leak on soinfo_free()"

9 years agoMerge "Use delete[] to de-allocate pointers from new[]"
Pirama Arumuga Nainar [Thu, 9 Jul 2015 22:40:50 +0000 (22:40 +0000)]
Merge "Use delete[] to de-allocate pointers from new[]"

9 years agoUse delete[] to de-allocate pointers from new[]
Pirama Arumuga Nainar [Thu, 9 Jul 2015 17:15:15 +0000 (10:15 -0700)]
Use delete[] to de-allocate pointers from new[]

Upstream clang has a new diagnostic, '-Wmismatched-new-delete' to
generate a warning (that becomes an error with -Werror) when delete is
used to deallocate pointers created with 'new[]'.  This patch fixes
'delete's that trigger this warning/error.

Change-Id: Id7b6f6c2df3e860e576de55e2f61f3d2be3a3986

9 years agoFix memory-leak on soinfo_free()
Dmitriy Ivanov [Wed, 8 Jul 2015 22:26:46 +0000 (15:26 -0700)]
Fix memory-leak on soinfo_free()

Bug: http://b/22290728
Change-Id: I00747056db2cd3ff8b33936278464daed7236536

9 years agoMerge "libc: Add sys/fcntl.h for compatibility."
Daniel Erat [Wed, 8 Jul 2015 21:15:22 +0000 (21:15 +0000)]
Merge "libc: Add sys/fcntl.h for compatibility."

9 years agolibc: Add sys/fcntl.h for compatibility.
Daniel Erat [Wed, 8 Jul 2015 21:01:50 +0000 (15:01 -0600)]
libc: Add sys/fcntl.h for compatibility.

Add a /usr/include/sys/fcntl.h file that includes <fcntl.h>.

Change-Id: Ie34bf7519607c5aae1fbea7c7dca249c9b7801cc

9 years agoMerge "Respect caller DT_RUNPATH in dlopen()."
Evgenii Stepanov [Wed, 8 Jul 2015 19:50:02 +0000 (19:50 +0000)]
Merge "Respect caller DT_RUNPATH in dlopen()."

9 years agoMerge "Replace bx lr with update of pc from the stack."
Christopher Ferris [Wed, 8 Jul 2015 19:15:37 +0000 (19:15 +0000)]
Merge "Replace bx lr with update of pc from the stack."

9 years agoReplace bx lr with update of pc from the stack.
Christopher Ferris [Mon, 6 Jul 2015 19:03:40 +0000 (12:03 -0700)]
Replace bx lr with update of pc from the stack.

When there is arm assembler of this format:

ldmxx sp!, {..., lr} or pop {..., lr}
bx lr

It can be replaced with:

ldmxx sp!, {..., pc} or pop {..., pc}

Change-Id: Ic27048c52f90ac4360ad525daf0361a830dc22a3

9 years agoRespect caller DT_RUNPATH in dlopen().
Evgenii Stepanov [Tue, 7 Jul 2015 00:56:31 +0000 (17:56 -0700)]
Respect caller DT_RUNPATH in dlopen().

When dlopen-ing a library, add the caller's DT_RUNPATH to the directory search
list. This fixes dlfcn.dt_runpath in bionic-unit-tests-glibc(32|64).

Bug: 21899363
Change-Id: Ife6a7e192939292cf4dc291b7e6b95945761cde3