OSDN Git Service

android-x86/bionic.git
12 years agobionic: introduce libc.debug.malloc.program
Iliyan Malchev [Tue, 29 May 2012 23:46:17 +0000 (16:46 -0700)]
bionic: introduce libc.debug.malloc.program

libc.debug.malloc.program  provides an additional level of control over which
processes to enable libc.debug.malloc functionality for.  The string value of
libc.debug.malloc.program is matched against the program name; if the value of
libc.debug.malloc.program is a substring of the program name, then malloc debug
is applied to that program at whatever level libc.debug.malloc specifies.

If lib.debug.malloc.program is not specified, then libc.debug.malloc has the
same effect as before.

For example, to enable libc.deubug.malloc = 10 only to the mediaserver, do the
following:

   adb root # necessary for setprop
   adb setprop libc.debug.malloc.program mediaserver
   adb setprop libc.debug.malloc 10
   adb kill -9 $(pid mediaserver)

Change-Id: I6f01c12f033c8e2e015d73025369d7f1685ba200
Signed-off-by: Iliyan Malchev <malchev@google.com>
12 years agoMerge "Ensure that the port number and TXID are properly randomized." into jb-dev
Geremy Condra [Wed, 30 May 2012 18:06:54 +0000 (11:06 -0700)]
Merge "Ensure that the port number and TXID are properly randomized." into jb-dev

12 years agoPrint the corrupted address passed to free().
Ben Cheng [Thu, 24 May 2012 23:56:53 +0000 (16:56 -0700)]
Print the corrupted address passed to free().

For example:

@@@ ABORTING: INVALID HEAP ADDRESS IN dlfree addr=0x5c3bfbd0
Fatal signal 11 (SIGSEGV) at 0xdeadbaad (code=1), thread 2942

The addr=0x5c3bfbd0 part is new.

Change-Id: I8670144b2b0a3a6182384150d762c97dfee5452f

12 years agoEnsure that the port number and TXID are properly randomized.
Geremy Condra [Mon, 21 May 2012 21:20:59 +0000 (14:20 -0700)]
Ensure that the port number and TXID are properly randomized.

This fix reads from /dev/urandom to get the required entropy.

Bug: 6535492
Change-Id: Ibc2fec3f71a67607b608ad9b767b0b6504993c1d

12 years agobionic: add support for non-NEON memcpy() on NEON SoCs
Prajakta Gudadhe [Mon, 7 May 2012 21:17:44 +0000 (14:17 -0700)]
bionic: add support for non-NEON memcpy() on NEON SoCs

Some SoCs that support NEON nevertheless perform better with a non-NEON than a
NEON memcpy().  This patch adds build variable ARCH_ARM_USE_NON_NEON_MEMCPY,
which can be set in BoardConfig.mk.  When ARCH_ARM_USE_NON_NEON_MEMCPY is
defined, we compile in the non-NEON optimized memcpy() even if the SoC supports
NEON.

Change-Id: Ia0e5bee6bad5880ffc5ff8f34a1382d567546cf9

12 years agoImplement the "abort" stub in assembly for ARM.
Ben Cheng [Tue, 8 May 2012 20:36:37 +0000 (13:36 -0700)]
Implement the "abort" stub in assembly for ARM.

So that we can always get the full stack trace regardless of gcc's handling
of the "noreturn" attribute associated with abort().

[cherry-picked from master]

BUG:6455193
Change-Id: I0102355f5bf20e636d3feab9d1424495f38e39e2

12 years agoMerge "Update f_accessory.h kernel header" into jb-dev
Mike Lockwood [Thu, 26 Apr 2012 20:10:25 +0000 (13:10 -0700)]
Merge "Update f_accessory.h kernel header" into jb-dev

12 years agolibc: continue to use Android's custom linker script
Nick Kralevich [Thu, 26 Apr 2012 16:54:48 +0000 (09:54 -0700)]
libc: continue to use Android's custom linker script

By default, Android no longer compiles code using it's custom
linker script /build/core/armelf.xsc. However, this causes
problems for libc. Certain programs linked using older versions
of GOLD expect libc.so to export __exidx_start and __exidx_end.
Removing the custom linker script causes libc.so not to export
those symbols.

For now, continue using the old linker script, until we can
figure out a better solution.

Change-Id: Iaf002afd63a58b848818da24e5a4525620dc4d74

12 years agoUpdate f_accessory.h kernel header
Mike Lockwood [Tue, 17 Apr 2012 17:41:34 +0000 (10:41 -0700)]
Update f_accessory.h kernel header

Change-Id: I29ec4aa4843b9308cbfa412df88e026e8475b715
Signed-off-by: Mike Lockwood <lockwood@google.com>
12 years agolinker: remove STB_LOCAL hack
Nick Kralevich [Mon, 23 Apr 2012 23:50:00 +0000 (16:50 -0700)]
linker: remove STB_LOCAL hack

The ARM static linker wasn't properly handling __exidx_start
and __exidx_end symbols. Now that the static linker has been fixed,
we don't need the dynamic linker to work around this problem.

Change-Id: I041b94903609fafab33663a7d441a5e70b7ffcdd

12 years agoadd linux/sw_sync.h
Erik Gilling [Wed, 18 Apr 2012 22:37:01 +0000 (15:37 -0700)]
add linux/sw_sync.h

Change-Id: I79de18d04b950c21b985d5ebc01cd3306a43d318
Signed-off-by: Erik Gilling <konkers@android.com>
12 years agoadd linux/sync.h
Erik Gilling [Wed, 18 Apr 2012 21:23:04 +0000 (14:23 -0700)]
add linux/sync.h

Change-Id: I38bb9498e18cb2b2e84a97487d4ad1e15fabd9d4
Signed-off-by: Erik Gilling <konkers@android.com>
12 years agoFix segv when unwinding stack past __libc_init.
Evgeniy Stepanov [Wed, 18 Apr 2012 08:57:40 +0000 (12:57 +0400)]
Fix segv when unwinding stack past __libc_init.

This change mirrors cd15bac for statically-linked binaries.

Change-Id: Id870832a50b37f0ef3e79e1ed03ed31390bfc9ef

12 years agoresolved conflicts for merge of 35d592cc to master
Elliott Hughes [Mon, 16 Apr 2012 21:40:26 +0000 (14:40 -0700)]
resolved conflicts for merge of 35d592cc to master

Change-Id: I8184302daf61814d26c837f9920b4e68d96d7f65

12 years agoresolved conflicts for merge of ef987656 to master
Elliott Hughes [Mon, 16 Apr 2012 21:26:43 +0000 (14:26 -0700)]
resolved conflicts for merge of ef987656 to master

Change-Id: I3854de8f4cddaf344444efa6f9da027642a237d9

12 years agoresolved conflicts for merge of 6b8fd054 to master
Elliott Hughes [Mon, 16 Apr 2012 21:16:42 +0000 (14:16 -0700)]
resolved conflicts for merge of 6b8fd054 to master

Change-Id: Ifc5a10d9c2f7764ad80d64cc552aad81d5fbf5eb

12 years agoam ff219e57: am 6435d27f: Merge "bionic: fix NULL parameter failure in getcwd()"
Elliott Hughes [Mon, 16 Apr 2012 20:15:13 +0000 (13:15 -0700)]
am ff219e57: am 6435d27f: Merge "bionic: fix NULL parameter failure in getcwd()"

* commit 'ff219e57c0ffe5ac2816f79677ce4f1afa677277':
  bionic: fix NULL parameter failure in getcwd()

12 years agoam 418e647a: Merge "libstdc++: Fix x86 thread-safe one-time-construction implementation."
Elliott Hughes [Mon, 16 Apr 2012 20:12:47 +0000 (13:12 -0700)]
am 418e647a: Merge "libstdc++: Fix x86 thread-safe one-time-construction implementation."

* commit '418e647a8300bb8e62b6b73814a5211152cbaacd':
  libstdc++: Fix x86 thread-safe one-time-construction implementation.

12 years agoam 4994deae: Merge "Bionic: Fix wrong prototype of system call clock_nanosleep"
Elliott Hughes [Mon, 16 Apr 2012 20:12:46 +0000 (13:12 -0700)]
am 4994deae: Merge "Bionic: Fix wrong prototype of system call clock_nanosleep"

* commit '4994deaef51d5fa1ac12e6160b47d9cbe3b2bc43':
  Bionic: Fix wrong prototype of system call clock_nanosleep

12 years agoam 7b8666e6: Merge "bionic: Fix wrong prototype of system call getresuid/getresgid"
Elliott Hughes [Mon, 16 Apr 2012 20:12:46 +0000 (13:12 -0700)]
am 7b8666e6: Merge "bionic: Fix wrong prototype of system call getresuid/getresgid"

* commit '7b8666e683e56549091b86fd7b9c421fd0124dbc':
  bionic: Fix wrong prototype of system call getresuid/getresgid

12 years agoam 6435d27f: Merge "bionic: fix NULL parameter failure in getcwd()"
Elliott Hughes [Mon, 16 Apr 2012 20:12:45 +0000 (13:12 -0700)]
am 6435d27f: Merge "bionic: fix NULL parameter failure in getcwd()"

* commit '6435d27f9d45d01fbd15bcc3dcd9617b86b825bb':
  bionic: fix NULL parameter failure in getcwd()

12 years agoMerge "libstdc++: Fix x86 thread-safe one-time-construction implementation."
Elliott Hughes [Mon, 16 Apr 2012 16:13:13 +0000 (09:13 -0700)]
Merge "libstdc++: Fix x86 thread-safe one-time-construction implementation."

12 years agoMerge "Bionic: Fix wrong prototype of system call clock_nanosleep"
Elliott Hughes [Mon, 16 Apr 2012 16:09:04 +0000 (09:09 -0700)]
Merge "Bionic: Fix wrong prototype of system call clock_nanosleep"

12 years agoMerge "bionic: Fix wrong prototype of system call getresuid/getresgid"
Elliott Hughes [Mon, 16 Apr 2012 16:06:22 +0000 (09:06 -0700)]
Merge "bionic: Fix wrong prototype of system call getresuid/getresgid"

12 years agoBionic: Fix wrong prototype of system call clock_nanosleep
Jack Ren [Tue, 17 Apr 2012 03:20:15 +0000 (11:20 +0800)]
Bionic: Fix wrong prototype of system call clock_nanosleep

In bionic/libc/SYSCALLS.TXT, the prototype of system call
clock_nanosleep is incorrect.

According to man page:
int clock_nanosleep(clockid_t clock_id, int flags,
                    const struct timespec *request,
                    struct timespec *remain);

Change-Id: Ic44c6db3d632293aa17998035554eacd664c2d57
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agoMerge "bionic: fix NULL parameter failure in getcwd()"
Elliott Hughes [Mon, 16 Apr 2012 15:48:19 +0000 (08:48 -0700)]
Merge "bionic: fix NULL parameter failure in getcwd()"

12 years agobionic: Fix wrong prototype of system call getresuid/getresgid
Jack Ren [Tue, 17 Apr 2012 03:16:18 +0000 (11:16 +0800)]
bionic: Fix wrong prototype of system call getresuid/getresgid

In bionic/libc/SYSCALLS.TXT, the prototypes of system call
getresuid/getresgid are incorrect.

According to man page, they should be:
    int getresuid(uid_t *ruid, uid_t *euid, uid_t *suid);
    int getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid);

Change-Id: I676098868bb05a9e1fe45419b234cf397626fdad
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agobionic: fix NULL parameter failure in getcwd()
Jack Ren [Tue, 13 Mar 2012 10:26:28 +0000 (18:26 +0800)]
bionic: fix NULL parameter failure in getcwd()

LTP: getcwd01 failed in LTP

Need to check getcwd parameters, otherwise it will lead to
posix test case to fail.

Change-Id: Ieb673b6dd4ca6481da81c5339dbf7ec0a463f263
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agolibstdc++: Fix x86 thread-safe one-time-construction implementation.
Fengwei Yin [Wed, 28 Mar 2012 09:25:17 +0000 (17:25 +0800)]
libstdc++: Fix x86 thread-safe one-time-construction implementation.

The root of the problem is that the existing implementation is based on the
ARM C++ ABI, which mandates a different guard variable layout than the
Itanium/x86 C++ one.

This patch modifies the implementation in a way that satisfies both ABIs (and
doesn't require changing the toolchains).

Change-Id: I885e9adc7f088b9c0a78355bd752f1e6aeec9f07
Signed-off-by: Fengwei Yin <fengwei.yin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agoam cebc8fb0: am 6bc18fa5: Merge "libm: fix invalid result of function remquo/remquof"
Elliott Hughes [Sat, 14 Apr 2012 00:10:06 +0000 (17:10 -0700)]
am cebc8fb0: am 6bc18fa5: Merge "libm: fix invalid result of function remquo/remquof"

* commit 'cebc8fb0862a8e1ecd86f9d09b0a3fca0f0781fa':
  libm: fix invalid result of function remquo/remquof

12 years agoam 6bc18fa5: Merge "libm: fix invalid result of function remquo/remquof"
Elliott Hughes [Sat, 14 Apr 2012 00:07:43 +0000 (17:07 -0700)]
am 6bc18fa5: Merge "libm: fix invalid result of function remquo/remquof"

* commit '6bc18fa58849a4307cf6ddcfd526d9258e8175fc':
  libm: fix invalid result of function remquo/remquof

12 years agoMerge changes I427a1881,I959b6428
Kenny Root [Fri, 13 Apr 2012 23:03:51 +0000 (16:03 -0700)]
Merge changes I427a1881,I959b6428

* changes:
  Add faccessat to syscall list
  Update unistd.h for new syscalls

12 years agoAdd faccessat to syscall list
Kenny Root [Fri, 13 Apr 2012 21:56:25 +0000 (14:56 -0700)]
Add faccessat to syscall list

Change-Id: I427a18811089cb280769ac8da3ed8adc00a65a10

12 years agoMerge "libm: fix invalid result of function remquo/remquof"
Elliott Hughes [Fri, 13 Apr 2012 21:57:59 +0000 (14:57 -0700)]
Merge "libm: fix invalid result of function remquo/remquof"

12 years agoUpdate unistd.h for new syscalls
Kenny Root [Fri, 13 Apr 2012 21:50:14 +0000 (14:50 -0700)]
Update unistd.h for new syscalls

gensyscalls.py run from external/kernel-headers at commit
efab8f3e49f7f36ef0354b0996ecd5f3fa031e52

Change-Id: I959b64280e184655ef8c713aa79f9e23cb1f7df4

12 years agoam 0613dce0: regenerate linux/fb.h
Travis Geiselbrecht [Thu, 12 Apr 2012 23:09:29 +0000 (16:09 -0700)]
am 0613dce0: regenerate linux/fb.h

* commit '0613dce0a7b806d48758cabfb2d638d0ba4dd2bc':
  regenerate linux/fb.h

12 years agoregenerate linux/fb.h
Travis Geiselbrecht [Thu, 12 Apr 2012 21:51:43 +0000 (14:51 -0700)]
regenerate linux/fb.h

Change-Id: Icd8c0f53306a48ffd513378abdf387af21e886a3

12 years agoInitialize TLS before any application code is run.
Evgeniy Stepanov [Thu, 22 Mar 2012 14:01:53 +0000 (18:01 +0400)]
Initialize TLS before any application code is run.

Since e19d702b8e33, dlsym and friends use recursive mutexes that
require the current thread id, which is not available before the libc
constructor. This prevents us from using dlsym() in .preinit_array.

This change moves TLS initialization from libc constructor to the earliest
possible point - immediately after linker itself is relocated. As a result,
pthread_internal_t for the initial thread is available from the start.

As a bonus, values stored in TLS in .preinit_array are not lost when libc is
initialized.

Change-Id: Iee5a710ee000173bff63e924adeb4a4c600c1e2d

12 years agoMerge "Fix segv when unwinding stack past __libc_init."
Elliott Hughes [Thu, 12 Apr 2012 18:52:44 +0000 (11:52 -0700)]
Merge "Fix segv when unwinding stack past __libc_init."

12 years agoMove end of __on_dlclose up
Kenny Root [Wed, 11 Apr 2012 21:16:04 +0000 (14:16 -0700)]
Move end of __on_dlclose up

The END macro was put too far down which made the linker complain about
it. Move up to the end of the code.

Change-Id: Ica71a9c6083b437d2213c7cefe34b0083c78f16b

12 years ago__on_dlclose should be aligned
Kenny Root [Wed, 11 Apr 2012 00:24:24 +0000 (17:24 -0700)]
__on_dlclose should be aligned

Marking segments read-only was pushing the alignment of __on_dlclose by
2 bytes making it unaligned. This change makes sure the ARM code is
aligned to the 4 byte boundary.

Bug: 6313309
Change-Id: Ic2bf475e120dd61225ec19e5d8a9a8b1d0b7f081

12 years agoRevert "linker: remove STB_LOCAL hack"
Nick Kralevich [Tue, 10 Apr 2012 20:42:06 +0000 (13:42 -0700)]
Revert "linker: remove STB_LOCAL hack"

This reverts commit 61ff83475c6f3a3bb05a01ac89d668a331bfe9e9.

This code is harmless, and only applies to the linker, so
there's no harm in keeping it in the tree a little bit longer.
Let's roll this back while we try to figure out the root cause
of bug 6314858.

Bug: 6314858
Change-Id: I9f5ed81d23a7abe273baf792aa8a0a2839ef094c

12 years agoFix segv when unwinding stack past __libc_init.
Evgeniy Stepanov [Tue, 10 Apr 2012 10:10:11 +0000 (14:10 +0400)]
Fix segv when unwinding stack past __libc_init.

This change fixes a segmentation fault in the libc unwinder when it goes
past __libc_init.

Unwind instructions for __libc_init direct it to grab the return address from
the stack frame. Without this change, the unwinder gets a wild address and
looks up further unwind instructions for the routine at that address. If it's
unlucky enough to hit an existing function, it will try to unwind it. Bad
things happen then.

With this change, the return address always points to the _start function,
which does not have unwind instructions associated with it. This stop the
unwind process.

__libc_init never returns, so this does not affect program execution, other
than adding 4 bytes on the main thread stack.

Change-Id: Id58612172e8825c8729cccd081541a13bff96bd0

12 years agolibm: fix invalid result of function remquo/remquof
Jack Ren [Sat, 19 Nov 2011 07:52:08 +0000 (15:52 +0800)]
libm: fix invalid result of function remquo/remquof

Currently we will get the wrong result as follows:
remquof(0x7bb33336, 0x63000000) = -671088640, 0x00000000
remquo(0xbff0000000000003, 0x3ff0000000000003) = 1, 0x8000000000000000
remquo(0x9120000000000001, 0x0000000000000005) = -1288490188, 0x0000000000000004
while the correct one should be:
remquof(0x7bb33336, 0x63000000) = 1476395008, 0x00000000
remquo(0xbff0000000000003, 0x3ff0000000000003) = -1, 0x8000000000000000
remquo(0x9120000000000001, 0x0000000000000005) = -1288490189, 0x0000000000000001

Fixed in this patch.

Change-Id: I540b348cd10a539f3b39b1753945c893c4c7ec46
Signed-off-by: Jingwei Zhang <jingwei.zhang@intel.com>
Signed-off-by: Jin Wei <wei.a.jin@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agoMerge "linker: remove STB_LOCAL hack"
Nick Kralevich [Mon, 9 Apr 2012 20:41:14 +0000 (13:41 -0700)]
Merge "linker: remove STB_LOCAL hack"

12 years agolinker: remove STB_LOCAL hack
Nick Kralevich [Mon, 9 Apr 2012 19:50:17 +0000 (12:50 -0700)]
linker: remove STB_LOCAL hack

The ARM static linker wasn't properly handling __exidx_start
and __exidx_end symbols. Now that the static linker has been fixed,
we don't need the dynamic linker to work around this problem.

Change-Id: Ic567122b6c7746cc016795e2befba2c6bd7c0478

12 years agoMerge "Update linux/input.h to version 3.4."
Jeff Brown [Mon, 9 Apr 2012 18:24:27 +0000 (11:24 -0700)]
Merge "Update linux/input.h to version 3.4."

12 years agoUpdate linux/input.h to version 3.4.
Jeff Brown [Sat, 7 Apr 2012 02:24:35 +0000 (19:24 -0700)]
Update linux/input.h to version 3.4.

Bug: 6292993
Change-Id: Ic7628068df6c8de9cb3711a4540d51365b96ebdb

12 years agolibc/kernel: update cleaned v4l2 header
Dima Zavin [Mon, 2 Apr 2012 19:57:03 +0000 (12:57 -0700)]
libc/kernel: update cleaned v4l2 header

Change-Id: Ib3747b45eb1e4095cca3de6d7692387953e8c4a7
Signed-off-by: Dima Zavin <dima@android.com>
12 years agoMerge "libc/kernel-headers: Auto generated files for netfilter/xt_IDLETIMER"
JP Abgrall [Fri, 6 Apr 2012 03:36:36 +0000 (20:36 -0700)]
Merge "libc/kernel-headers: Auto generated files for netfilter/xt_IDLETIMER"

12 years agolibc/kernel-headers: Auto generated files for netfilter/xt_IDLETIMER
Ashish Sharma [Fri, 6 Apr 2012 02:50:42 +0000 (19:50 -0700)]
libc/kernel-headers: Auto generated files for netfilter/xt_IDLETIMER

From Change I526b5fce: Add NETLINK_IDLETIMER msg type and include the corresponding header file.

Change-Id: I24bffc11394c8664e4d7d7f439b0600545f07536
Signed-off-by: Ashish Sharma <ashishsharma@google.com>
12 years agolinker: surround "debug_verbosity" by "#if LINKER_DEBUG"
Nick Kralevich [Wed, 4 Apr 2012 19:43:32 +0000 (12:43 -0700)]
linker: surround "debug_verbosity" by "#if LINKER_DEBUG"

Avoid executing useless code when we're not compiled
with debugging support.

Change-Id: Iea1821b241acacdf65d1a91d743356058dfef273

12 years agolinker: enable -fvisibility=hidden
Nick Kralevich [Wed, 4 Apr 2012 17:18:27 +0000 (10:18 -0700)]
linker: enable -fvisibility=hidden

Compile the linker with -fvisibility=hidden. This reduces the number
of symbols that show up in the .dynsym section of the linker.
These symbols are never exported to other applications.

In particular, this fixes a problem with setting -DLINKER_DEBUG=1
introduced in 468319ce4f3f7383d788b76c09cda2a405311f36.
Because the symbols "debug_verbosity" and "format_log" have not been
resolved before the linker links itself, any attempt to call
PRINT / INFO / TRACE / WARN / ERROR will result in a segfault.
This change allows the static linker to produce a relative reference
to these symbols rather than relying on relocation.

This also has a nice side effect of making the linker slightly smaller
and slightly more optimized.

The following symbols no longer in the .dynsym section of the linker
after this change:

-addr_to_name
-call_constructors_recursi
-calloc
-debugger_init
-debugger_signal_handler
-debug_verbosity
-dladdr
-dlclose
-dlerror
-dl_lock
-dlopen
-dlsym
-dl_unwind_find_exidx
-find_containing_library
-find_containing_symbol
-find_library
-format_buffer
-free
-libdl_info
-linker_env_get
-linker_env_init
-linker_env_secure
-linker_env_unset
-linker_get_error
-__linker_init
-lookup
-lookup_in_library
-malloc
-notify_gdb_of_libraries
-notify_gdb_of_load
-notify_gdb_of_unload
-realloc
-rtld_db_dlactivity
-unload_library
-vsnprintf

Bug: 5827809
Change-Id: I5e8cd7dcf48c1d6831a970a67f63f24916c5e437

12 years agoMerge remote-tracking branch 'goog/ics-aah'
Mike Lockwood [Tue, 3 Apr 2012 18:35:33 +0000 (11:35 -0700)]
Merge remote-tracking branch 'goog/ics-aah'

12 years agoam fdb11929: am b88f810d: Merge "Update to tzdata2012c."
Elliott Hughes [Mon, 2 Apr 2012 23:09:36 +0000 (16:09 -0700)]
am fdb11929: am b88f810d: Merge "Update to tzdata2012c."

* commit 'fdb119297ac421e2fc2ec096a6d5370b81938de8':
  Update to tzdata2012c.

12 years agoam b88f810d: Merge "Update to tzdata2012c."
Elliott Hughes [Mon, 2 Apr 2012 23:07:46 +0000 (16:07 -0700)]
am b88f810d: Merge "Update to tzdata2012c."

* commit 'b88f810d585a1736a1759c2c22e9f4999441ae00':
  Update to tzdata2012c.

12 years agoMerge "Update to tzdata2012c."
Elliott Hughes [Mon, 2 Apr 2012 17:25:43 +0000 (10:25 -0700)]
Merge "Update to tzdata2012c."

12 years agoUpdate to tzdata2012c.
Elliott Hughes [Mon, 2 Apr 2012 14:35:13 +0000 (07:35 -0700)]
Update to tzdata2012c.

From the release notes:

       africa
               Summer time changes for Morocco (to start late April 2012)

       asia
               Changes for 2012 for Gaza & the West Bank (Hebron) and Syria

       northamerica
               Haiti following US/Canada rules for 2012 (and we're assuming,
               for now anyway, for the future).

Also include a change made internally to the 'generate' script as part of
the tzdata2011m update that apparently never made it to AOSP; the original
checkin comment for which was:

    Update to tzdata2011m.

    Fixes for Europe/Tiraspol (Moldova) and all four Ukrainian zones.

    Also show the MD5 of the downloaded data, for comparison against the MD5
    given in the announcement mails. (There's a plan to move to proper signing,
    but that's not implemented on their end yet.)

(I'm repeating the tzdata change for the convenience of anyone grepping the
log, since the 2012 tzdata releases also contain the 2011m changes; 2011m
is the only missing release I noticed.)

Change-Id: I9a2e530b3a8ea88e3375334a12376e3d8526f267

12 years agoam cd834618: am 63b14755: Merge "libc/x86: ensure the stack 16-byte aligned when...
Elliott Hughes [Sat, 31 Mar 2012 05:16:02 +0000 (22:16 -0700)]
am cd834618: am 63b14755: Merge "libc/x86: ensure the stack 16-byte aligned when tasks created"

* commit 'cd834618c4752b61d54ff4005a8baa8219b822e4':
  libc/x86: ensure the stack 16-byte aligned when tasks created

12 years agoam 63b14755: Merge "libc/x86: ensure the stack 16-byte aligned when tasks created"
Elliott Hughes [Fri, 30 Mar 2012 22:53:19 +0000 (15:53 -0700)]
am 63b14755: Merge "libc/x86: ensure the stack 16-byte aligned when tasks created"

* commit '63b14755512d515762cbe3ad8544c561a576f32e':
  libc/x86: ensure the stack 16-byte aligned when tasks created

12 years agoMerge "libc/x86: ensure the stack 16-byte aligned when tasks created"
Elliott Hughes [Fri, 30 Mar 2012 20:42:42 +0000 (13:42 -0700)]
Merge "libc/x86: ensure the stack 16-byte aligned when tasks created"

12 years agoRevert "libc/kernel: Add rules to autogenerate device specific kernel headers"
Dima Zavin [Fri, 30 Mar 2012 17:21:16 +0000 (10:21 -0700)]
Revert "libc/kernel: Add rules to autogenerate device specific kernel headers"

This reverts commit 884147c7d0b1b6571256feb11e12d893432625b2.

Change-Id: I09723858ac961f3e1155791aa5c54d5d3abfbd36
Signed-off-by: Dima Zavin <dima@android.com>
12 years agoam ea76f414: am d509f9cc: am 09ce7749: Merge "[MIPS] Clean Kernel headers are generat...
Elliott Hughes [Wed, 28 Mar 2012 00:58:57 +0000 (17:58 -0700)]
am ea76f414: am d509f9cc: am 09ce7749: Merge "[MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture."

* commit 'ea76f4147825cc39d9aa91230cd863ed29f28e27':
  [MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture.

12 years agoam d509f9cc: am 09ce7749: Merge "[MIPS] Clean Kernel headers are generated by running...
Elliott Hughes [Wed, 28 Mar 2012 00:57:41 +0000 (17:57 -0700)]
am d509f9cc: am 09ce7749: Merge "[MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture."

* commit 'd509f9ccbba6e1dd53acfd7425bfc06d6b3f8693':
  [MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture.

12 years agoam 09ce7749: Merge "[MIPS] Clean Kernel headers are generated by running libc/kernel...
Elliott Hughes [Wed, 28 Mar 2012 00:53:35 +0000 (17:53 -0700)]
am 09ce7749: Merge "[MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture."

* commit '09ce7749d74733b28d4fa7a1d36457cb366cc5da':
  [MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture.

12 years agoMerge "[MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_...
Elliott Hughes [Tue, 27 Mar 2012 23:49:28 +0000 (16:49 -0700)]
Merge "[MIPS] Clean Kernel headers are generated by running libc/kernel/tools/update_all.py script. This patch ignores any changes to libc/kernel directory not related to MIPS architecture."

12 years ago[MIPS] Clean Kernel headers are generated by running
Raghu Gandham [Tue, 27 Mar 2012 18:37:17 +0000 (11:37 -0700)]
[MIPS] Clean Kernel headers are generated by running
libc/kernel/tools/update_all.py script. This patch ignores
any changes to libc/kernel directory not related to MIPS
architecture.

Change-Id: I2c9e461dccb7c33eb4420be2db1a562f45137c8d
Signed-off-by: Raghu Gandham <raghu@mips.com>
Signed-off-by: Chris Dearman <chris@mips.com>
12 years agoam b52c941e: am d7c6147e: am 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"
Jean-Baptiste Queru [Tue, 27 Mar 2012 17:10:50 +0000 (10:10 -0700)]
am b52c941e: am d7c6147e: am 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"

* commit 'b52c941ed4f24134b78a73daee84bdc328f969da':
  bionic: fix atfork hanlder_mutex deadlock

12 years agoam d7c6147e: am 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"
Jean-Baptiste Queru [Tue, 27 Mar 2012 01:29:23 +0000 (18:29 -0700)]
am d7c6147e: am 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"

* commit 'd7c6147eff65572762d55950b965db949ccd0281':
  bionic: fix atfork hanlder_mutex deadlock

12 years agoam 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"
Jean-Baptiste Queru [Tue, 27 Mar 2012 01:25:43 +0000 (18:25 -0700)]
am 56731351: Merge "bionic: fix atfork hanlder_mutex deadlock"

* commit '56731351de7230180fc99a1a4b0afd12f881b0f7':
  bionic: fix atfork hanlder_mutex deadlock

12 years agoMerge "Add auto-generated headers for USB FunctionFS"
Benoit Goby [Tue, 27 Mar 2012 00:14:24 +0000 (17:14 -0700)]
Merge "Add auto-generated headers for USB FunctionFS"

12 years agoMerge "bionic: fix atfork hanlder_mutex deadlock"
Jean-Baptiste Queru [Mon, 26 Mar 2012 23:19:35 +0000 (16:19 -0700)]
Merge "bionic: fix atfork hanlder_mutex deadlock"

12 years agoHandle pthread-related changes (mutex/atfork)
Jean-Baptiste Queru [Mon, 26 Mar 2012 22:25:19 +0000 (15:25 -0700)]
Handle pthread-related changes (mutex/atfork)

First commit:

Revert "Revert "am be741d47: am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()"""

This reverts commit 06823da2f0c8b4a4ce4c45113032f03df85c94b8.

Second commit:

bionic: fix atfork hanlder_mutex deadlock

This cherry-picks commit 34e89c232dd5645fe3b5f9b40856d8e3e4cae57a

After applying the kernel_id fix, the system refused to boot up and we
got following crash log:
I/DEBUG   (  113): pid: 618, tid: 618  >>> org.simalliance.openmobileapi.service:remote <<<
I/DEBUG   (  113): signal 16 (SIGSTKFLT), code -6 (?), fault addr --------
I/DEBUG   (  113):  eax fffffe00  ebx b77de994  ecx 00000080  edx 00724002
I/DEBUG   (  113):  esi 00000000  edi 00004000
I/DEBUG   (  113):  xcs 00000073  xds 0000007b  xes 0000007b  xfs 00000000 xss 0000007b
I/DEBUG   (  113):  eip b7761351  ebp bfdf3de8  esp bfdf3dc4  flags 00000202
I/DEBUG   (  113):     #00  eip: 00015351  /system/lib/libc.so
I/DEBUG   (  113):     #01  eip: 0000d13c  /system/lib/libc.so (pthread_mutex_lock)
I/DEBUG   (  113):     #02  eip: 00077b48  /system/lib/libc.so (__bionic_atfork_run_prepare)
I/DEBUG   (  113):     #03  eip: 00052cdb  /system/lib/libc.so (fork)
I/DEBUG   (  113):     #04  eip: 0009ae91  /system/lib/libdvm.so (_Z18dvmOptimizeDexFileillPKcjjb)
I/DEBUG   (  113):     #05  eip: 000819d6  /system/lib/libdvm.so (_Z14dvmJarFileOpenPKcS0_PP7JarFileb)
I/DEBUG   (  113):     #06  eip: 000b175e  /system/lib/libdvm.so (_ZL40Dalvik_dalvik_system_DexFile_openDexFilePKjP6JValue)
I/DEBUG   (  113):     #07  eip: 0011fb94  /system/lib/libdvm.so

Root cause:
The atfork uses the mutex handler_mutex to protect the atfork_head. The
parent will call __bionic_atfork_run_prepare() to lock the handler_mutex,
and need both the parent and child to unlock their own copy of handler_mutex
after fork. At that time, the owner of hanlder_mutex is set as the parent.
If we apply the kernel_id fix, then the child's kernel_id will be set as
child's tid.
The handler_mutex is a recursive lock, and pthread_mutex_unlock(&hander_mutex)
will fail because the mutex owner is the parent, while the current tid
(__get_thread()->kernel_id) is child, not matched with the mutex owner.
At that time, the handler_mutex is left in lock state.If the child wants to
fork other process after than, then it will try to lock handler_mutex, and
then be deadlocked.

Fix:
Since the child has its own copy of vm space from the the parent, the
child space's handler_mutex should be reset to the initialized state.

Change-Id: I3907dd9a153418fb78862f2aa6d0302c375d9e27
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Chenyang Du <chenyang.du@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
Change-Id: Ic8072f366a877443a60fe215f3c00b3df5a259c8

12 years agoAdd auto-generated headers for USB FunctionFS
Benoit Goby [Tue, 6 Mar 2012 00:10:41 +0000 (16:10 -0800)]
Add auto-generated headers for USB FunctionFS

linux/usb/ch9.h and linux/usb/functionfs.h

Change-Id: I2907081aba63b32740eb9916315759692a96ab42

12 years agoFixed to #include correct 32-bit headers; Refreshed libc/kernel headers
Andrew Hsieh [Fri, 23 Mar 2012 15:07:36 +0000 (23:07 +0800)]
Fixed to #include correct 32-bit headers; Refreshed libc/kernel headers

This patch fixes an issue where 64-bit hreaders are incorrectly included
in kernel headers.  For example, file "libc/kernel/arch-x86/asm/io.h"
incorreclty includes "io_64.h" (missing, BTW) instead of "io_32.h".

The reason is because CONFIG_X86_32 isn't considered pre-defined in
"kernel_default_arch_macros" for x86, and clean_header.py doesn't
look at it at all anyway (ie. __i386__ is also ignored, but it's
okay since x86 cross compiler defines it back)

Fixed 2 tools/*py, README.TXT, and refreshed libc/kernel headers

Change-Id: Iac834cc8b3548f055d3f2a214af36072dd679fe8

12 years agolibc/x86: ensure the stack 16-byte aligned when tasks created
Jack Ren [Wed, 21 Mar 2012 09:48:13 +0000 (17:48 +0800)]
libc/x86: ensure the stack 16-byte aligned when tasks created

Currently Renderscript sample code RsBalls crashed on x86 when SSE2
enabled. The root cause is that the stack was not 16-byte aligned
from the beginning when the processes/threads were created, so the
RsBalls crashed when SSE2 instructions tried to access the variables
on the stack.

- For the thread created by fork():
Its stack alignment is determined by crtbegin_{dynamic, static}.S

- For the thread created by pthread_create():
Its stack alignment is determined by clone.S. __thread_entry( ) is
a standard C function. In order to have its stack be aligned with
16 byte properly, __thread_entry() needs the stack with following
layout when it is called:
layout #1 (correct)
--------------
|            |
-------------- <--ESP (ECX - 20)
| ret EIP    |
-------------- <--ECX - 16
| arg0       |
-------------- <--ECX - 12
| arg1       |
-------------- <--ECX - 8
| arg2       |
-------------- <--ECX - 4
| unused     |
-------------- <--ECX (16-byte boundary)

But it has following layout for now:
layout #2: (incorrect)
--------------
|            |
-------------- <--ESP (ECX - 16)
| unused     |
-------------- <--ECX - 12
| arg0       |
-------------- <--ECX - 8
| arg1       |
-------------- <--ECX - 4
| arg2       |
-------------- <--ECX (16-byte boundary)

Fixed in this patch.

Change-Id: Ibe01f64db14be14033c505d854c73033556ddaa8
Signed-off-by: Michael Liao <michael.liao@intel.com>
Signed-off-by: H.J. Lu <hongjiu.lu@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agoInitialize mspace->least_addr properly in the mmap path.
Ben Cheng [Wed, 21 Mar 2012 22:47:12 +0000 (15:47 -0700)]
Initialize mspace->least_addr properly in the mmap path.

BUG: 6206963
Change-Id: Id2ab580246de50a4511b56a734a7bece98fb945c

12 years agoMerge "string.h: add __attribute__ ((pure)) to string functions"
Nick Kralevich [Wed, 21 Mar 2012 21:40:41 +0000 (14:40 -0700)]
Merge "string.h: add __attribute__ ((pure)) to string functions"

12 years agoMerge "bionic: pass MADV_MERGEABLE on private & anonymous mmaps"
Iliyan Malchev [Wed, 21 Mar 2012 20:51:42 +0000 (13:51 -0700)]
Merge "bionic: pass MADV_MERGEABLE on private & anonymous mmaps"

12 years agoMerge "Update the libc kernel includes to support the newer mman for KSM support"
JP Abgrall [Wed, 21 Mar 2012 20:43:10 +0000 (13:43 -0700)]
Merge "Update the libc kernel includes to support the newer mman for KSM support"

12 years agobionic: pass MADV_MERGEABLE on private & anonymous mmaps
JP Abgrall [Wed, 21 Mar 2012 20:19:51 +0000 (13:19 -0700)]
bionic: pass MADV_MERGEABLE on private & anonymous mmaps

Change-Id: I8bc167bb33dec6417fe772172697ea6ff97da2f6
Signed-off-by: Iliyan Malchev <malchev@google.com>
12 years agoUpdate the libc kernel includes to support the newer mman for KSM support
JP Abgrall [Wed, 21 Mar 2012 20:21:08 +0000 (13:21 -0700)]
Update the libc kernel includes to support the newer mman for KSM support

These are generated from the matching external/kernel-headers/original
files (from kernel/samsung android-samsung-3.0-wip).

Change-Id: I982ff6a0d522ea250c3a437f5756766fcc6c5c91

12 years agoMerge "fnmatch.c: Update to version in OpenBSD HEAD"
Nick Kralevich [Wed, 21 Mar 2012 17:07:55 +0000 (10:07 -0700)]
Merge "fnmatch.c: Update to version in OpenBSD HEAD"

12 years agofnmatch.c: Update to version in OpenBSD HEAD
Nick Kralevich [Wed, 21 Mar 2012 16:46:46 +0000 (09:46 -0700)]
fnmatch.c: Update to version in OpenBSD HEAD

Upgrade fnmatch.c from OpenBSD version 1.13 to 1.16.
This is needed primarily to address CVE-2011-0419.

This is a straight copy from upstream's version at
http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/gen/fnmatch.c and
incorporates the following changes:

Revision 1.16:
New fnmatch(3) implementation which is not recursive.
Written and provided under BSD licence by William A. Rowe Jr.
Originally released in Apache APR-1.4.5.
Merged class matching code from r1.14 and PATH_MAX check from r1.15.
ok miod millert

Revision 1.15:
Put a limit on recursion during matching, and reject input of size greater
or equal PATH_MAX. Based on similar fix made in NetBSD.
ok miod@ millert@

Revision 1.14:
POSIX character class support for fnmatch(3) and glob(3).  OK deraadt@

Version 1.14 introduced charclasses.h, which we copy unmodified
from upstream version 1.1.
http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/gen/charclass.h

Bug: 3435120
Change-Id: I45133468f0c3d439fd10eb087a1c647799f9d25b

12 years agoMerge "New additions/bug fixes required/found when porting perf."
Ben Cheng [Wed, 21 Mar 2012 16:32:24 +0000 (09:32 -0700)]
Merge "New additions/bug fixes required/found when porting perf."

12 years agostring.h: add __attribute__ ((pure)) to string functions
Nick Kralevich [Wed, 21 Mar 2012 15:48:18 +0000 (08:48 -0700)]
string.h: add __attribute__ ((pure)) to string functions

cdefs.h: Introduce the __purefunc attribute, which allows us to mark
certain functions as being "pure".

http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html

  Many functions have no effects except the return value and their
  return value depends only on the parameters and/or global variables.
  Such a function can be subject to common subexpression elimination
  and loop optimization just as an arithmetic operator would be.

string.h: Mark many commently used string functions as "pure", to
allow for additional compiler optimizations.

Change-Id: I42961f90f822b6dbcbc3fd72cdbe774a7adc8785

12 years agolibc/kernel: Add rules to autogenerate device specific kernel headers
Dima Zavin [Fri, 16 Mar 2012 23:18:15 +0000 (16:18 -0700)]
libc/kernel: Add rules to autogenerate device specific kernel headers

This change will automatically post-process kernel headers
specified by device, board, and product. This will allow us
to not check in each kernel header twice, at least for the
device specific headers for now.

Change-Id: I3bb144b6535504b7c26b807daa75de495554356d
Signed-off-by: Dima Zavin <dima@android.com>
12 years agoNew additions/bug fixes required/found when porting perf.
Ben Cheng [Wed, 14 Mar 2012 06:04:57 +0000 (23:04 -0700)]
New additions/bug fixes required/found when porting perf.

New functions:
tfind
tsearch
tdelete
twalk
tdestroy (GNU extension)

Bug fix: the current implementation for realpath would crash
if the second argument (resolved_path) is NULL.

New headers:
ar.h
search.h

Change-Id: Ib6c1e42fc186a6d597a6e5a9692b16acaa155804

12 years agoMerge "linker: fix off-by-one error in GNU_RELRO handling"
Nick Kralevich [Tue, 20 Mar 2012 17:52:42 +0000 (10:52 -0700)]
Merge "linker: fix off-by-one error in GNU_RELRO handling"

12 years agoMerge "linker: initially reserved memory as PROT_NONE"
Nick Kralevich [Tue, 20 Mar 2012 17:52:23 +0000 (10:52 -0700)]
Merge "linker: initially reserved memory as PROT_NONE"

12 years agolinker: fix off-by-one error in GNU_RELRO handling
Nick Kralevich [Fri, 16 Mar 2012 18:38:58 +0000 (11:38 -0700)]
linker: fix off-by-one error in GNU_RELRO handling

Fix a bug where the GNU_RELRO end address could be exactly
the end of the loadable segment.

Change-Id: If6c43acabc06e9aff9217c0f6016e158b28bb41f

12 years agobionic: fix atfork hanlder_mutex deadlock
Jack Ren [Fri, 16 Mar 2012 08:37:35 +0000 (16:37 +0800)]
bionic: fix atfork hanlder_mutex deadlock

After applying the kernel_id fix, the system refused to boot up and we
got following crash log:
I/DEBUG   (  113): pid: 618, tid: 618  >>> org.simalliance.openmobileapi.service:remote <<<
I/DEBUG   (  113): signal 16 (SIGSTKFLT), code -6 (?), fault addr --------
I/DEBUG   (  113):  eax fffffe00  ebx b77de994  ecx 00000080  edx 00724002
I/DEBUG   (  113):  esi 00000000  edi 00004000
I/DEBUG   (  113):  xcs 00000073  xds 0000007b  xes 0000007b  xfs 00000000 xss 0000007b
I/DEBUG   (  113):  eip b7761351  ebp bfdf3de8  esp bfdf3dc4  flags 00000202
I/DEBUG   (  113):     #00  eip: 00015351  /system/lib/libc.so
I/DEBUG   (  113):     #01  eip: 0000d13c  /system/lib/libc.so (pthread_mutex_lock)
I/DEBUG   (  113):     #02  eip: 00077b48  /system/lib/libc.so (__bionic_atfork_run_prepare)
I/DEBUG   (  113):     #03  eip: 00052cdb  /system/lib/libc.so (fork)
I/DEBUG   (  113):     #04  eip: 0009ae91  /system/lib/libdvm.so (_Z18dvmOptimizeDexFileillPKcjjb)
I/DEBUG   (  113):     #05  eip: 000819d6  /system/lib/libdvm.so (_Z14dvmJarFileOpenPKcS0_PP7JarFileb)
I/DEBUG   (  113):     #06  eip: 000b175e  /system/lib/libdvm.so (_ZL40Dalvik_dalvik_system_DexFile_openDexFilePKjP6JValue)
I/DEBUG   (  113):     #07  eip: 0011fb94  /system/lib/libdvm.so

Root cause:
The atfork uses the mutex handler_mutex to protect the atfork_head. The
parent will call __bionic_atfork_run_prepare() to lock the handler_mutex,
and need both the parent and child to unlock their own copy of handler_mutex
after fork. At that time, the owner of hanlder_mutex is set as the parent.
If we apply the kernel_id fix, then the child's kernel_id will be set as
child's tid.
The handler_mutex is a recursive lock, and pthread_mutex_unlock(&hander_mutex)
will fail because the mutex owner is the parent, while the current tid
(__get_thread()->kernel_id) is child, not matched with the mutex owner.
At that time, the handler_mutex is left in lock state.If the child wants to
fork other process after than, then it will try to lock handler_mutex, and
then be deadlocked.

Fix:
Since the child has its own copy of vm space from the the parent, the
child space's handler_mutex should be reset to the initialized state.

Change-Id: I3907dd9a153418fb78862f2aa6d0302c375d9e27
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Chenyang Du <chenyang.du@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
12 years agolinker: initially reserved memory as PROT_NONE
Nick Kralevich [Fri, 16 Mar 2012 20:06:12 +0000 (13:06 -0700)]
linker: initially reserved memory as PROT_NONE

When the dynamic linker loads a shared library into memory, it
initially allocates a chunk of memory. The memory is then carved
into smaller chunks for each LOAD region, and appropriate memory
protections applied.

Modify the initial memory allocation so that the pages are mapped
as PROT_NONE, rather than PROT_READ / PROT_EXEC. This ensures that
gaps between LOAD regions are not inadvertantly readable / executable.

(Long term, we should munmap() these gaps entirely)

Change-Id: If128a203ccc6fe12dcbbd2bfe0cf13a2045675af

12 years agoAdd in auto-generated if_alg.h header
Kenny Root [Fri, 16 Mar 2012 04:59:01 +0000 (21:59 -0700)]
Add in auto-generated if_alg.h header

Change-Id: I5d0934069e74be2eafecdee43074590124db57a7

12 years agoMerge "Update kernel headers and add syscall "perf_event_open""
Ben Cheng [Tue, 13 Mar 2012 20:04:22 +0000 (13:04 -0700)]
Merge "Update kernel headers and add syscall "perf_event_open""

12 years agoUpdate kernel headers and add syscall "perf_event_open"
Ben Cheng [Tue, 13 Mar 2012 19:28:40 +0000 (12:28 -0700)]
Update kernel headers and add syscall "perf_event_open"

Change-Id: I43f12b727881df002a8524f2738586c043833bae

12 years agoRevert "am be741d47: am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id...
Guang Zhu [Tue, 13 Mar 2012 05:05:36 +0000 (22:05 -0700)]
Revert "am be741d47: am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()""

This reverts commit 76d56cf4a94e875a8b621025e2780775247adb6e, reversing
changes made to c59ba4595be25a1213955233fcf9bcd1afe6438e.

Bug: 6157577

12 years agoam be741d47: am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread...
Elliott Hughes [Tue, 13 Mar 2012 00:12:35 +0000 (17:12 -0700)]
am be741d47: am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()"

* commit 'be741d472868a8ffcb455588f18cda889b0f465c':
  bionic: Fix wrong kernel_id in pthread descriptor after fork()

12 years agoam 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor...
Elliott Hughes [Tue, 13 Mar 2012 00:10:46 +0000 (17:10 -0700)]
am 2f460fbe: am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()"

* commit '2f460fbee9abe2936175379c49c6618de7421233':
  bionic: Fix wrong kernel_id in pthread descriptor after fork()

12 years agoam 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()"
Elliott Hughes [Tue, 13 Mar 2012 00:06:09 +0000 (17:06 -0700)]
am 73b5cad9: Merge "bionic: Fix wrong kernel_id in pthread descriptor after fork()"

* commit '73b5cad989da317cc8089b57ee25f502b1cac71f':
  bionic: Fix wrong kernel_id in pthread descriptor after fork()