OSDN Git Service

android-x86/art.git
9 years agoMerge "Dump kernel/native stacks of pure native threads." into lmp-dev
Christopher Ferris [Sun, 21 Sep 2014 21:43:05 +0000 (21:43 +0000)]
Merge "Dump kernel/native stacks of pure native threads." into lmp-dev

9 years agoFix sput-wide verification flag.
Logan Chien [Thu, 11 Sep 2014 09:36:05 +0000 (17:36 +0800)]
Fix sput-wide verification flag.

The vA register for sput-wide is a wide register.  Thus, the
verification flag should be kVerifyRegAWide instead of
kVerifyRegA.

Without this change, the invalid Dex bytecode won't be
rejected by VerifyInstruction().  As the result, the
DCHECK_LT() in register_line-inl.h will abort the program.

Change-Id: I24b746d8a85ddaf811d9aebed12a3dd9f97d755f
(cherry picked from commit be8d99c228a76e70a19c1122aabe111c25a2121c)

9 years agoDump kernel/native stacks of pure native threads.
Christopher Ferris [Sat, 20 Sep 2014 00:23:59 +0000 (17:23 -0700)]
Dump kernel/native stacks of pure native threads.

Bug: 16463406
Change-Id: Ifa18d7ec6c60721ce4d3e73944851d1f07372043

9 years agoMove spammy logs to JDWP verbose mode
Sebastien Hertz [Thu, 18 Sep 2014 08:20:42 +0000 (10:20 +0200)]
Move spammy logs to JDWP verbose mode

We are spammed by warning messages when debugging, especially each time we
suspend/resume all threads (to update instrumentation or collect monitor info).
It's common to get into the cases where these warnings are logged so they
shouldn't be warning but debug messages.

This CL moves these LOG(WARNING) to VLOG(jdwp) to not disturb developers when
debugging their app (especially when looking for specific messages in logcat).
We keep them in JDWP verbose mode because they help knowing when we initiate
these sequences of "suspend/resume all threads".

Also adds debug suspend count in the log message for more context.

Bug: 17524544
Bug: 17170697

(cherry picked from commit f272af4b9dcd39cdd50fa6655601a26e837eaea9)

Change-Id: I61df70ace1475bf10d83202b6bb774f7036354cb

9 years agoART: Only allow the zygote to create the global boot image
Andreas Gampe [Fri, 19 Sep 2014 03:56:04 +0000 (20:56 -0700)]
ART: Only allow the zygote to create the global boot image

Do not allow arbitrary processes, even when root, to write the
boot image in /data/dalvik-cache.

Bug: 174787521751048917439961
Change-Id: Iba2b74be6d0752f4221f4ff5ee295b45a34cb2e1

9 years agoDO NOT MERGE. Only have a portable entrypoint in portable builds.
Ian Rogers [Thu, 18 Sep 2014 15:56:45 +0000 (08:56 -0700)]
DO NOT MERGE. Only have a portable entrypoint in portable builds.

Bug: 16214885

Change-Id: Iff7b7415efdbdabd7e6020e221a540f6a774c852

9 years agoMerge "Update JDWP event filtering to avoid useless ids" into lmp-dev
Sebastien Hertz [Thu, 18 Sep 2014 12:20:26 +0000 (12:20 +0000)]
Merge "Update JDWP event filtering to avoid useless ids" into lmp-dev

9 years agoART: skip compilation of huge methods.
buzbee [Wed, 17 Sep 2014 20:23:58 +0000 (13:23 -0700)]
ART: skip compilation of huge methods.

Besides not being worthwhile to compile, it huge methods could cause
overflow of some internal compiler structures.

internal b/17524219

Change-Id: I0e47ff4757eec265809598830a2616ac367af77d

9 years agoMerge "Avoid crash in StringReference.Value JDWP command" into lmp-dev
Sebastien Hertz [Wed, 17 Sep 2014 09:46:44 +0000 (09:46 +0000)]
Merge "Avoid crash in StringReference.Value JDWP command" into lmp-dev

9 years agoMerge "ART: Avoid recursive abort on thread dump" into lmp-dev
Andreas Gampe [Wed, 17 Sep 2014 03:31:25 +0000 (03:31 +0000)]
Merge "ART: Avoid recursive abort on thread dump" into lmp-dev

9 years agoART: Sync oat file to disk before patching
Andreas Gampe [Tue, 16 Sep 2014 23:40:09 +0000 (16:40 -0700)]
ART: Sync oat file to disk before patching

Bug: 15567083174399611751048917478752
Change-Id: I828dc6775044b5050c2520eb097abe6a920fd3ee

9 years agoART: Avoid recursive abort on thread dump
Andreas Gampe [Tue, 16 Sep 2014 22:15:13 +0000 (15:15 -0700)]
ART: Avoid recursive abort on thread dump

Also clean up unused variable.

Bug: 15567083
Change-Id: Id83a3784bbe88520a4ffe654b2553d36a7b8e5bc

9 years agoMerge "Avoid suspending for alloc trace enabling when already suspended." into lmp-dev
Jeff Hao [Tue, 16 Sep 2014 21:53:22 +0000 (21:53 +0000)]
Merge "Avoid suspending for alloc trace enabling when already suspended." into lmp-dev

9 years agoAvoid suspending for alloc trace enabling when already suspended.
Jeff Hao [Tue, 16 Sep 2014 01:03:41 +0000 (18:03 -0700)]
Avoid suspending for alloc trace enabling when already suspended.

Bug: 17499772
Change-Id: Id98c10967b28e8859e5ac46f5878c304fb85c498

9 years agoMerge "ART: Check for exceptions from unresolved classes" into lmp-dev
Andreas Gampe [Tue, 16 Sep 2014 18:35:57 +0000 (18:35 +0000)]
Merge "ART: Check for exceptions from unresolved classes" into lmp-dev

9 years agoART: Check for exceptions from unresolved classes
Andreas Gampe [Tue, 16 Sep 2014 18:23:23 +0000 (11:23 -0700)]
ART: Check for exceptions from unresolved classes

In no-verify mode, classes may be unresolved because of missing
dependencies. Ignore and clear the exception.

Bug: 17506140
Change-Id: I70602b089e6631b1e177dbe8316c5fefdaf777a0

9 years agoMerge "Avoid printing absolute addresses in oatdump" into lmp-dev
Brian Carlstrom [Tue, 16 Sep 2014 18:25:29 +0000 (18:25 +0000)]
Merge "Avoid printing absolute addresses in oatdump" into lmp-dev

9 years agoAvoid printing absolute addresses in oatdump
Brian Carlstrom [Mon, 15 Sep 2014 03:34:17 +0000 (20:34 -0700)]
Avoid printing absolute addresses in oatdump

- Added printing of OatClass offsets.
- Added printing of OatMethod offsets.
- Added bounds checks for code size size, code size, mapping table, gc map, vmap table.
- Added sanity check of 100k for code size.
- Added partial disassembly of questionable code.
- Added --no-disassemble to disable disassembly.
- Added --no-dump:vmap to disable vmap dumping.
- Reordered OatMethod info to be in file order.

Bug: 15567083
Change-Id: Id86a21e06d4a28f29f16fd018cba7e55c57f849a

9 years agoMerge "ART: Do not unlink unreadable ELF files" into lmp-dev
Andreas Gampe [Tue, 16 Sep 2014 05:31:05 +0000 (05:31 +0000)]
Merge "ART: Do not unlink unreadable ELF files" into lmp-dev

9 years agoART: Do not unlink unreadable ELF files
Andreas Gampe [Tue, 16 Sep 2014 05:25:24 +0000 (22:25 -0700)]
ART: Do not unlink unreadable ELF files

Unlinking can be racy with respect to secondary oat files.

Bug: 17496483
Change-Id: I53f92a924b34dfcf50d7e02487c5f8f23de627ca

9 years agoMerge "Only log an error if an unattached thread is unregistered." into lmp-dev
Ian Rogers [Mon, 15 Sep 2014 23:08:15 +0000 (23:08 +0000)]
Merge "Only log an error if an unattached thread is unregistered." into lmp-dev

9 years agoOnly log an error if an unattached thread is unregistered.
Ian Rogers [Mon, 15 Sep 2014 22:17:07 +0000 (15:17 -0700)]
Only log an error if an unattached thread is unregistered.

Turn a fatal check into a diagnostic error. It looks like this issue arises
when a runtime shutsdown without properly being started.
Bug: 17011539

Change-Id: I2983c8332e83769e9480e8f30a46ca3b80a2e90e

9 years agoAvoid crash in StringReference.Value JDWP command
Sebastien Hertz [Mon, 15 Sep 2014 09:27:27 +0000 (11:27 +0200)]
Avoid crash in StringReference.Value JDWP command

Checks for null or invalid object id. Also checks whether the corresponding
object is a java.lang.String.

Bug: 17492221
Bug: 15005460

Change-Id: I2e9afe2ade69b130e59a5a0c00d02ded20b05b94

9 years agoAdd native support for FinalizerList.makeCircularListIfUnenqueued
Mathieu Chartier [Fri, 12 Sep 2014 19:58:05 +0000 (12:58 -0700)]
Add native support for FinalizerList.makeCircularListIfUnenqueued

Called from FinalizerReference.enqueueSentinelReference to prevent
a race where the GC updates pendingNext of the sentinel reference
before enqueueSentinelReference.

Bug: 17462553

Change-Id: I7ad2fd250c2715d1aeb919bd548ef9aab24f30a2

9 years agoUpdate JDWP event filtering to avoid useless ids
Sebastien Hertz [Tue, 9 Sep 2014 10:10:13 +0000 (12:10 +0200)]
Update JDWP event filtering to avoid useless ids

To reduce the number of JDWP ids in the debugger, we update the event filtering
support to work with runtime objects (Thread, Class, Object, ...) instead of
JDWP ids (ThreadId, RefTypeId, ObjectId, ...).

We used to create useless JDWP ids for events even if they were not reported
because of event filtering (thread only, class only, instance only, ...). Now
we only create JDWP ids when we know we're going to report an event.

Bug: 17343664
Change-Id: Ib83df6d82fbcb40895f5cd9201a1e9ecd0a07e17

9 years agoART: Make elf loading not abort
Andreas Gampe [Sat, 13 Sep 2014 01:38:24 +0000 (18:38 -0700)]
ART: Make elf loading not abort

Changes elf_file code to use less CHECKs and instead return error
values (usually nullptr). This avoids aborts.

In oat_file, when loading an oat file fails, try to unlink at. If
this succeeds, on the next run we may compile again.

Bug: 17491333
Change-Id: I50fdd2edacd86f25d4dacf2180ce2a6105eaf4af

9 years agoMerge "Fix stale root error in verifier" into lmp-dev
Mathieu Chartier [Sat, 13 Sep 2014 00:34:30 +0000 (00:34 +0000)]
Merge "Fix stale root error in verifier" into lmp-dev

9 years agoFix stale root error in verifier
Mathieu Chartier [Sat, 13 Sep 2014 00:01:24 +0000 (17:01 -0700)]
Fix stale root error in verifier

There was a stale root error caused by the static roots from
the reg types. These were visitied if there was an active verifier
in the method_verifiers_ but this is not always the case when a GC
is run. The fix is to always visit the static method verifier roots.

This only showed up as a bug without an image since these roots
were primitive classes and always in the image, and therefore
didn't ever need to be updated due to moving GC.

Bug: 17262039
Change-Id: I592f2770570de97b431671cfbd409f63697892f1

9 years agoMerge "ART: Allow quickening in the boot image" into lmp-dev
Andreas Gampe [Fri, 12 Sep 2014 22:31:51 +0000 (22:31 +0000)]
Merge "ART: Allow quickening in the boot image" into lmp-dev

9 years agoART: Allow quickening in the boot image
Andreas Gampe [Thu, 28 Aug 2014 21:41:02 +0000 (14:41 -0700)]
ART: Allow quickening in the boot image

Update the class linker to accept class status from the boot image
in compiler mode. Update compiler driver to allow quickening for
boot image classes. Update method verifier to accept quickened
instructions in compiler mode when we just want to dump. Update
oatdump to the new verifier API.

Bug: 17316928
Change-Id: I9ef1bfd78b0d93625b89b3d662131d7d6e5f2903

9 years agoMerge "ART: Change access flag behavior in verifier" into lmp-dev
Andreas Gampe [Fri, 12 Sep 2014 20:49:25 +0000 (20:49 +0000)]
Merge "ART: Change access flag behavior in verifier" into lmp-dev

9 years agoMerge "Quick compiler (arm64) Fix inline Math.round()" into lmp-dev
buzbee [Fri, 12 Sep 2014 19:49:26 +0000 (19:49 +0000)]
Merge "Quick compiler (arm64) Fix inline Math.round()" into lmp-dev

9 years agoMerge "Fix and re-enable FreeList large object space for 64 bit" into lmp-dev
Mathieu Chartier [Fri, 12 Sep 2014 19:42:41 +0000 (19:42 +0000)]
Merge "Fix and re-enable FreeList large object space for 64 bit" into lmp-dev

9 years agoQuick compiler (arm64) Fix inline Math.round()
buzbee [Fri, 12 Sep 2014 17:40:47 +0000 (10:40 -0700)]
Quick compiler (arm64) Fix inline Math.round()

Math.round is detected and inlined for arm64.  However, the
arm64 backend incorrectly modified a source operand in place
during the round sequence.  Depending on how registers are
allocated, that modification could persist.  Changed to use a
temp register for the intermediate result.

Internal b/17411468

Change-Id: I7c636f985e193f8ff838768fde3b741e443bb1bb

9 years agoART: Change access flag behavior in verifier
Andreas Gampe [Mon, 25 Aug 2014 22:05:04 +0000 (15:05 -0700)]
ART: Change access flag behavior in verifier

Note: this moves the miranda modifier to the upper 16 bit.

Bug: 16161620
Change-Id: I2f591d53b7d1559171e70aaaf22225d94b4882f5

9 years agoMerge "Fix host 64-bit ISA string" into lmp-dev
Andreas Gampe [Fri, 12 Sep 2014 18:25:28 +0000 (18:25 +0000)]
Merge "Fix host 64-bit ISA string" into lmp-dev

9 years agoFix and re-enable FreeList large object space for 64 bit
Mathieu Chartier [Tue, 9 Sep 2014 00:42:48 +0000 (17:42 -0700)]
Fix and re-enable FreeList large object space for 64 bit

Not enabled on 32 bit due to virtual memory fragmentation concerns.
The new free list large object space ensures that allocations are
page aligned by using a side table for accounting data.

Bug: 17414549

Change-Id: Idbcbe75cb86b6d9b3d8b20f3048631a48c511458

9 years agoMerge "Add fast path to VMClassLoader.findLoadedClass" into lmp-dev
Mathieu Chartier [Thu, 11 Sep 2014 23:51:03 +0000 (23:51 +0000)]
Merge "Add fast path to VMClassLoader.findLoadedClass" into lmp-dev

9 years agoMerge "Delete pin table" into lmp-dev
Mathieu Chartier [Thu, 11 Sep 2014 23:27:11 +0000 (23:27 +0000)]
Merge "Delete pin table" into lmp-dev

9 years agoMerge "Print the command line and ABI in stack dumps like debuggerd." into lmp-dev
Brian Carlstrom [Thu, 11 Sep 2014 23:13:17 +0000 (23:13 +0000)]
Merge "Print the command line and ABI in stack dumps like debuggerd." into lmp-dev

9 years agoPrint the command line and ABI in stack dumps like debuggerd.
Jeff Brown [Thu, 11 Sep 2014 01:41:18 +0000 (18:41 -0700)]
Print the command line and ABI in stack dumps like debuggerd.

Ensure the heading is "Cmd line:" just like debuggerd and
as it used to be in Dalvik.
Fix a missing newline.
Trim all training nulls from the command line.
Don't bother printing the original command line if unset.
Add the ABI to the dump to help the native stack symbol tool.

Bug: 17474152
Change-Id: Ie1da1bd4f38c92b13ea08c7122f4573d8ec8fc73

9 years agoMerge "Revert "Print the command line and ABI in stack dumps like debuggerd."" into...
Jeff Brown [Thu, 11 Sep 2014 23:01:55 +0000 (23:01 +0000)]
Merge "Revert "Print the command line and ABI in stack dumps like debuggerd."" into lmp-dev

9 years agoRevert "Print the command line and ABI in stack dumps like debuggerd."
Jeff Brown [Thu, 11 Sep 2014 23:01:44 +0000 (23:01 +0000)]
Revert "Print the command line and ABI in stack dumps like debuggerd."

This reverts commit 38ac77b6fb1fc626299022fe0f16d5b6c365cb9b.

Change-Id: I8e4952b80d03ad61c3ff2afaf8a96b8c2e36f0c7

9 years agoMerge "Print the command line and ABI in stack dumps like debuggerd." into lmp-dev
Jeff Brown [Thu, 11 Sep 2014 22:57:55 +0000 (22:57 +0000)]
Merge "Print the command line and ABI in stack dumps like debuggerd." into lmp-dev

9 years agoAdd fast path to VMClassLoader.findLoadedClass
Mathieu Chartier [Thu, 11 Sep 2014 21:21:41 +0000 (14:21 -0700)]
Add fast path to VMClassLoader.findLoadedClass

VMClassLoader.findLoadedClass now calls FindClassInPathClassLoader
as a fast path. Exclusive time results (trace view maps launch):

Before:
nativeFillInStackTrace 1.4%
defineClassNative 1.2%
findLoadedClass 0.2%

After:
nativeFillInStackTrace 0.5%
defineClassNative 0.0%
findLoadedClass 0.9%

Bug: 16828525

Change-Id: I1bde48effcd28529778c00ec0fa0dda4e32026a3

9 years agoPrint the command line and ABI in stack dumps like debuggerd.
Jeff Brown [Thu, 11 Sep 2014 01:41:18 +0000 (18:41 -0700)]
Print the command line and ABI in stack dumps like debuggerd.

Ensure the heading is "Cmd line:" just like debuggerd and
as it used to be in Dalvik.
Fix a missing newline.
Trim all training nulls from the command line.
Don't bother printing the original command line if unset.
Add the ABI to the dump to help the native stack symbol tool.

Bug: 17474152
Change-Id: I19c585dbd7d31ef6c98270e469445cae40cbd518

9 years agoDelete pin table
Mathieu Chartier [Thu, 11 Sep 2014 20:14:31 +0000 (13:14 -0700)]
Delete pin table

The pin table was brought over from dalvik but not really needed
since ART doesn't support pinning in movable spaces. The only
thing it did was hold objects live for JNI functions.
This shouldn't be necessary since people keep jni references to
these objects or else they could never release the elements.

Bug: 17456946

Change-Id: Ibed0d029157ffb9e75ecd80d4d544d690986c090

9 years agoART: Fix preverified setting in VerifyClass
Andreas Gampe [Thu, 11 Sep 2014 17:59:33 +0000 (10:59 -0700)]
ART: Fix preverified setting in VerifyClass

Make sure soft-failed classes cannot set methods to pre-verified.

Bug: 1682852517465185
Change-Id: I09c0a68ca722978459741311148eae7614f9ca49

9 years agoMerge "Improve dex location canonicalization-related performance." into lmp-dev
Vladimir Marko [Thu, 11 Sep 2014 16:18:49 +0000 (16:18 +0000)]
Merge "Improve dex location canonicalization-related performance." into lmp-dev

9 years agoMerge "Fix GetDexCanonicalLocation test for relative paths." into lmp-dev
Vladimir Marko [Thu, 11 Sep 2014 16:18:16 +0000 (16:18 +0000)]
Merge "Fix GetDexCanonicalLocation test for relative paths." into lmp-dev

9 years agoFix crash when connected to DDMS
Sebastien Hertz [Thu, 11 Sep 2014 14:23:44 +0000 (16:23 +0200)]
Fix crash when connected to DDMS

Restore init of WellKnownClasses::org_apache_harmony_dalvik_ddmc_Chunk_data and
removed duplicated line.

Bug: 17467328
Bug: 17463118
Change-Id: I64638e23615b7c20061b49c41a9c0e9f35a4d537

9 years agoMerge "Fix deadlock in VirtualMachine.AllThreads" into lmp-dev
Sebastien Hertz [Thu, 11 Sep 2014 09:33:33 +0000 (09:33 +0000)]
Merge "Fix deadlock in VirtualMachine.AllThreads" into lmp-dev

9 years agoMerge "Don't hold any lock when visiting classes from JDWP" into lmp-dev
Sebastien Hertz [Thu, 11 Sep 2014 08:04:30 +0000 (08:04 +0000)]
Merge "Don't hold any lock when visiting classes from JDWP" into lmp-dev

9 years agoMerge "Use WellKnownClasses in ThrowStackOverflowError to reduce changes of further...
Brian Carlstrom [Thu, 11 Sep 2014 06:54:05 +0000 (06:54 +0000)]
Merge "Use WellKnownClasses in ThrowStackOverflowError to reduce changes of further stack overflow" into lmp-dev

9 years agoUse WellKnownClasses in ThrowStackOverflowError to reduce changes of further stack...
Brian Carlstrom [Thu, 11 Sep 2014 06:10:47 +0000 (23:10 -0700)]
Use WellKnownClasses in ThrowStackOverflowError to reduce changes of further stack overflow

Bug: 17463118
Change-Id: Ic05b07803d2d3112e0b67064f380cd46ba9e293f

9 years agoART: Correctly make methods preverified
Andreas Gampe [Thu, 11 Sep 2014 02:48:05 +0000 (19:48 -0700)]
ART: Correctly make methods preverified

Bug: 16828525
Change-Id: I66756348b2aa50e41dacca59769b6810a91c73b0

9 years agoImprove dex location canonicalization-related performance.
Vladimir Marko [Fri, 5 Sep 2014 13:01:17 +0000 (14:01 +0100)]
Improve dex location canonicalization-related performance.

Eagerly add canonical dex file locations to the OatFile's
primary lookup map in Setup(). This moves the boot.oat work
from every app startup to the zygote initialization. Since
we always ended up initializing the canonical location map
anyway due to the way that we're loading dex files, the lazy
initialization didn't save anything.

Clean up dex file name canonicalization to make sure we
free() the memory returned by realpath() rather than using
std::unique_ptr<> with the default deleter.

Avoid some unnecessary duplicate OatDexFile lookups.

Bug: 16828525
Bug: 17346103

(cherry picked from commit aa4497db59f1eeec954f2ba5da6d458fcdf9b3a4)

Change-Id: Icc4b14ebe903282ca91ce24e33a6d7c75dff991c

9 years agoFix GetDexCanonicalLocation test for relative paths.
Calin Juravle [Thu, 7 Aug 2014 10:54:36 +0000 (11:54 +0100)]
Fix GetDexCanonicalLocation test for relative paths.

If the reference file is created with a relative path the test will
fail.

(cherry picked from commit 61281dc0aaa22730fd4fee268cac40776b193042)

Change-Id: Ie82b9d381bd884111dfc91cc235b46ad9f3225be

9 years agoMerge "Ensure class is linked before resolution" into lmp-dev
Brian Carlstrom [Wed, 10 Sep 2014 18:02:55 +0000 (18:02 +0000)]
Merge "Ensure class is linked before resolution" into lmp-dev

9 years agoEnsure class is linked before resolution
Sebastien Hertz [Wed, 10 Sep 2014 17:44:39 +0000 (19:44 +0200)]
Ensure class is linked before resolution

Adds a missing call to ClassLinker::EnsureResolved to ensure we did link the
class and retired the temp class (placeholder) before doing the resolution.

Bug: 17435441

Change-Id: Ib1a7181d6e5e814ca9299d0504e739a2b69475ef

9 years agoMerge "Add proper cfi directives." into lmp-dev
Christopher Ferris [Wed, 10 Sep 2014 17:49:51 +0000 (17:49 +0000)]
Merge "Add proper cfi directives." into lmp-dev

9 years agoMerge "VisitClassesWithoutClassesLock isn't safe if classes move." into lmp-dev
Ian Rogers [Wed, 10 Sep 2014 14:31:36 +0000 (14:31 +0000)]
Merge "VisitClassesWithoutClassesLock isn't safe if classes move." into lmp-dev

9 years agoAdd proper cfi directives.
Christopher Ferris [Tue, 9 Sep 2014 20:55:13 +0000 (13:55 -0700)]
Add proper cfi directives.

This function has a mix of arm and cfi directives. Since our unwinders
can work with both, remove the arm directives and fix the cfi directives.

Bug: 17392751
Change-Id: Ibd4859a4f0b8f2da751a0928a2f5f4afa3df7df8

9 years agoIncrease dex2oat timeouts for debug builds
Mathieu Chartier [Wed, 10 Sep 2014 00:26:58 +0000 (17:26 -0700)]
Increase dex2oat timeouts for debug builds

Should fix the 056-const-string-jumbo test.

Change-Id: I04d98dfd59003b964c1feee6eee31079f9f4a476

9 years agoChange Reference.get() intrinsic to Reference.getReferent().
Mathieu Chartier [Tue, 9 Sep 2014 20:51:09 +0000 (13:51 -0700)]
Change Reference.get() intrinsic to Reference.getReferent().

The reference intrinsic was incorrectly inlining
PhantomReference.get(). We now get around this by adding a
layer of indirection. Reference.get() now calls getReferent()
which is intrinsified and inlined.

Requires:
https://android-review.googlesource.com/#/c/107100/

Bug: 17429865

(cherry picked from commit cd48f2d86197d4fe87cc88077bc4af5ba66e5295)

Change-Id: Ie91e70abf43cedf3c707c7bb8a5059e19d2a2577

9 years agoMerge "Change 099-vmdebug test to use File.createTempFile." into lmp-dev
Jeff Hao [Tue, 9 Sep 2014 20:13:24 +0000 (20:13 +0000)]
Merge "Change 099-vmdebug test to use File.createTempFile." into lmp-dev

9 years agoChange 099-vmdebug test to use File.createTempFile.
Jeff Hao [Tue, 9 Sep 2014 20:07:59 +0000 (13:07 -0700)]
Change 099-vmdebug test to use File.createTempFile.

Avoids issues with concurrency.

Bug: 17439227
Change-Id: I4a5956d9f26fad416a244216c2c77d5e6f072c62

9 years agoMerge "Temporarily disable LOS on 64 bit" into lmp-dev
Mathieu Chartier [Tue, 9 Sep 2014 17:39:37 +0000 (17:39 +0000)]
Merge "Temporarily disable LOS on 64 bit" into lmp-dev

9 years agoFix deadlock in VirtualMachine.AllThreads
Sebastien Hertz [Tue, 9 Sep 2014 10:08:49 +0000 (12:08 +0200)]
Fix deadlock in VirtualMachine.AllThreads

We cannot add any object in the JDWP object registry while holding the
Locks::thread_list_lock. Indeed we may need to suspend a thread and take it,
causing a deadlock by waiting for ourself on this lock.

Bug: 17343664
Change-Id: I2335c63e418c0d485daf2c35f8debcd52f943140

9 years agoART: Better error reporting on broken oat file
Andreas Gampe [Tue, 9 Sep 2014 06:39:45 +0000 (23:39 -0700)]
ART: Better error reporting on broken oat file

Do not blow up when in GetHashBucket if GetHashBucketNum is zero.
Instead back-step to ElfFileOpen, which prints which symbol was
missing from which file.

Bug: 17422404
Change-Id: I3b890dd1c31c08b1ccc0f7f668afcceee95f8d00

9 years agoMerge "Workaround Darwin ContainedWithinExistingMap issue" into lmp-dev
Brian Carlstrom [Tue, 9 Sep 2014 06:02:19 +0000 (06:02 +0000)]
Merge "Workaround Darwin ContainedWithinExistingMap issue" into lmp-dev

9 years agoWorkaround Darwin ContainedWithinExistingMap issue
Brian Carlstrom [Tue, 9 Sep 2014 05:41:35 +0000 (22:41 -0700)]
Workaround Darwin ContainedWithinExistingMap issue

Bug: 16861075
Change-Id: I577a0c7ede919e610c356e7c40dd60d87914dd3a

9 years agoMerge "Fix monitor visit for class initializer" into lmp-dev
Sebastien Hertz [Tue, 9 Sep 2014 05:12:55 +0000 (05:12 +0000)]
Merge "Fix monitor visit for class initializer" into lmp-dev

9 years agoMerge "ART: Fix unchecked register index validity" into lmp-dev
Andreas Gampe [Tue, 9 Sep 2014 05:07:38 +0000 (05:07 +0000)]
Merge "ART: Fix unchecked register index validity" into lmp-dev

9 years agoFix darwin run-test problems
Brian Carlstrom [Tue, 9 Sep 2014 04:37:39 +0000 (21:37 -0700)]
Fix darwin run-test problems

Bug: 17428908
Change-Id: Id74a0383fdae6753558e27cd447c1ca349819c91

9 years agoTemporarily disable LOS on 64 bit
Mathieu Chartier [Tue, 9 Sep 2014 02:49:50 +0000 (19:49 -0700)]
Temporarily disable LOS on 64 bit

Problem is that MAP_32BIT simulation can take seconds on large object
allocations in the worst case.

Bug: 17414549

Change-Id: I463bb0b728eed14847a32a27a93640ec6080f6b3

9 years agoMerge "Add fast path for FindClass using the type dex file." into lmp-dev
Mathieu Chartier [Tue, 9 Sep 2014 01:14:22 +0000 (01:14 +0000)]
Merge "Add fast path for FindClass using the type dex file." into lmp-dev

9 years agoAdd fast path for FindClass using the type dex file.
Mathieu Chartier [Fri, 5 Sep 2014 01:33:17 +0000 (18:33 -0700)]
Add fast path for FindClass using the type dex file.

If we are using the PathClassLoader with a BootClassLoader
parent, we can handle the common case in the FindClass function
without needing to go back to java code.

Around 10% speedup measured of maps launch, could be noise due to
large variation of app launch times. Eliminates defineClassNative
from being anywhere near the top of sampling profiles.

Bug: 17397179
Bug: 16828525
Change-Id: Ide0db2b5f6cf5b96fc46e89178e0799de667cb88

9 years agoMerge "ART: Change ART for new native bridge sequence" into lmp-dev
Andreas Gampe [Tue, 9 Sep 2014 00:41:35 +0000 (00:41 +0000)]
Merge "ART: Change ART for new native bridge sequence" into lmp-dev

9 years agoMerge "Add test for hashCode and System.identityHashCode." into lmp-dev
Mathieu Chartier [Mon, 8 Sep 2014 23:01:52 +0000 (23:01 +0000)]
Merge "Add test for hashCode and System.identityHashCode." into lmp-dev

9 years agoDisable tracing if enabled by default for new tracing test.
Jeff Hao [Mon, 8 Sep 2014 22:32:39 +0000 (15:32 -0700)]
Disable tracing if enabled by default for new tracing test.

Bug: 17412385
Change-Id: If963a9ba2214c72efbef673a840bbeb6fdbb1795

9 years agoMerge "Change when sampling thread is reset during shutdown to prevent races." into...
Jeff Hao [Mon, 8 Sep 2014 22:22:08 +0000 (22:22 +0000)]
Merge "Change when sampling thread is reset during shutdown to prevent races." into lmp-dev

9 years agoChange when sampling thread is reset during shutdown to prevent races.
Jeff Hao [Mon, 8 Sep 2014 16:38:18 +0000 (09:38 -0700)]
Change when sampling thread is reset during shutdown to prevent races.

Also adds some extra argument checking and testing for tracing.

Bug: 17412385
Change-Id: I876378d5bd17d75ab346584713b04dc0656d2120

9 years agoAdd test for hashCode and System.identityHashCode.
Mathieu Chartier [Mon, 8 Sep 2014 21:48:25 +0000 (14:48 -0700)]
Add test for hashCode and System.identityHashCode.

(cherry picked from commit bda2836af470c45e30c567b14d6747a7450c4741)

Change-Id: Ieba8f8eb09441ec145a176df00c293695eae0f47

9 years agoIgnore heap trim requests if we are the zygote
Mathieu Chartier [Mon, 8 Sep 2014 20:35:11 +0000 (13:35 -0700)]
Ignore heap trim requests if we are the zygote

Done to prevent app launching lag due to sleep in heap trimmer
daemon.

Bug: 17310019

Change-Id: Ia593e7eced1c1583771985ec9e7b60ee0c0e7912

9 years agoRevert "Fix heap trimmer daemon sleeping."
Mathieu Chartier [Mon, 8 Sep 2014 20:08:41 +0000 (20:08 +0000)]
Revert "Fix heap trimmer daemon sleeping."

This reverts commit 698b44634396e8111e96e1116586fde89a838347.

Change-Id: Ifa8c2d554b82289cc31803ad6027735d2c3c9be8

9 years agoART: Fix unchecked register index validity
Andreas Gampe [Sun, 7 Sep 2014 20:06:12 +0000 (13:06 -0700)]
ART: Fix unchecked register index validity

The static check of index validity is against the type given by
the instruction, e.g., boolean for SPUT_BOOLEAN, but the target_type
is the resolved field type and can differ. An additional check is
necessary to avoid a read out of bounds.

Bug: 17411109
Change-Id: Ie4ed8bbda79f3f6403a24e727450a943447aa71d

9 years agoART: Change ART for new native bridge sequence
Andreas Gampe [Wed, 3 Sep 2014 04:22:18 +0000 (21:22 -0700)]
ART: Change ART for new native bridge sequence

Initialize or unload after a fork.

(cherry picked from commit 6be67eeedbe60afce42300ae3e7f0e7180a96efa)

Change-Id: I5a20de1cb68dd1802937b369b14c50c9c1031c67

9 years agoVisitClassesWithoutClassesLock isn't safe if classes move.
Ian Rogers [Fri, 29 Aug 2014 22:40:08 +0000 (15:40 -0700)]
VisitClassesWithoutClassesLock isn't safe if classes move.

Which they do, so avoid by doing an array allocation.
Also, tidy member variables to the end of ClassLinker.
Remove unnecessary mutable. Tidy and fix a locks required/excluded.

Change-Id: I2404a9e7a1ea997d68ab1206f97d2a20dffbda06
(cherry picked from commit dbf3be0f133c0bdf454f637fee2452dbb5f7c027)

9 years agoFix monitor visit for class initializer
Sebastien Hertz [Mon, 8 Sep 2014 08:15:19 +0000 (10:15 +0200)]
Fix monitor visit for class initializer

We do not hold class object's lock when invoking class initializer. Therefore
it is wrong to visit it when looking for monitors owned by a thread.

Bug: 17377924

Change-Id: I7fddcdb9a554301e040c571971eec9be6cba5eda

9 years agoMerge "Fix numerous issues with DdmVmInternal allocation tracking" into lmp-dev
Brian Carlstrom [Fri, 5 Sep 2014 18:54:44 +0000 (18:54 +0000)]
Merge "Fix numerous issues with DdmVmInternal allocation tracking" into lmp-dev

9 years agoFix numerous issues with DdmVmInternal allocation tracking
Brian Carlstrom [Fri, 5 Sep 2014 05:15:18 +0000 (22:15 -0700)]
Fix numerous issues with DdmVmInternal allocation tracking

Issues addressed:
- Using without JDWP attached caused native crash.
- When buffer is full (64k entries), number of entries reported was 0.
- Disabling tracking after disabling tracking caused native crash.
- Asking for allocations after disabled caused native crash.
- Lock ordering issues between mutator lock and alloc tracker lock.

Adding 098-ddmc test to cover these cases.

Bug: 17392248
Change-Id: I45c0abbfb3dd7a7421aea3bbba2ea8d05225e48b

9 years agoMerge "Avoid recomputing the dex checksum during class loading" into lmp-dev
Calin Juravle [Fri, 5 Sep 2014 15:33:21 +0000 (15:33 +0000)]
Merge "Avoid recomputing the dex checksum during class loading" into lmp-dev

9 years agoAvoid recomputing the dex checksum during class loading
Calin Juravle [Tue, 2 Sep 2014 14:53:55 +0000 (15:53 +0100)]
Avoid recomputing the dex checksum during class loading

Thread the already computed checksum to VerifyOatAndDexFileChecksums and
LoadMultiDexFilesFromOatFile to avoid recomputing it.

Bug:17346103
Change-Id: Ifa0c1cad952853751e98cbb3c999631b9909a9f9

9 years agoEnsure alternate signal stack is minimum size.
Ian Rogers [Wed, 13 Aug 2014 21:37:26 +0000 (14:37 -0700)]
Ensure alternate signal stack is minimum size.

On Mac a sigaltstack may need to be 32KB, setting it to 16KB causes
sigaltstack to fail.

(cherry picked from commit c24a1e00b13a7ebf10dd896f6a18b1ad09617ec7)

Change-Id: Ic06d80f1acf5c4eb3d2dd1156bfa2805ab0df492

9 years agoMerge "Fix heap trimmer daemon sleeping." into lmp-dev
Mathieu Chartier [Thu, 4 Sep 2014 17:38:39 +0000 (17:38 +0000)]
Merge "Fix heap trimmer daemon sleeping." into lmp-dev

9 years agoFix heap trimmer daemon sleeping.
Mathieu Chartier [Thu, 4 Sep 2014 04:26:36 +0000 (21:26 -0700)]
Fix heap trimmer daemon sleeping.

Problem:
Heap trimmer daemon sleeping caused app launch occasionally to take a
while due stopping the heap trimming thread taking a long time.

The heap trimmer thread now never sleeps, we prevent issues caused
by frequent back and forth process state changes by only
transitioning to background if it has been kCollectorTransitionWait
time since the last transition to background. Similar logic for heap
trimming.

Bug: 17310019

Change-Id: I23980421cc388f36b66a4c03ed15dd11d43f59b5

9 years agoART: Fix x86_64 GenSelect case when destination is Ref
Jean Christophe Beyler [Thu, 4 Sep 2014 15:34:28 +0000 (08:34 -0700)]
ART: Fix x86_64 GenSelect case when destination is Ref

Reference in x86_64 is a 64-bit solo register. As a result, the invocation
  of OpRegImm results in an error when Select opcode of the kind:
    ref = boolean ? null : null;
because opRegImm does not support 64-bit destination for OpMov.

The case above is only possible for ref because no one other constant except
  null is possible.

Bug: 17327895
Change-Id: I7541e744ec1c8619711712fd17be72764efcf3a8
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
(cherry picked from commit 3f51e7d942c22edaab3a7e703a1e6a2dd6a26f77)

9 years agoMerge "ART: Fix computation of frame size for direct proxy methods" into lmp-dev
Ian Rogers [Thu, 4 Sep 2014 15:53:16 +0000 (15:53 +0000)]
Merge "ART: Fix computation of frame size for direct proxy methods" into lmp-dev

9 years agoART: Fix computation of frame size for direct proxy methods
Serguei Katkov [Fri, 29 Aug 2014 11:20:15 +0000 (18:20 +0700)]
ART: Fix computation of frame size for direct proxy methods

Proxy method has only one direct method and it is a constructor which
is cloned from java.lang.reflect.Proxy class together with code.
As a result its body is a compiled quick code and frame size should be
computed accordingly in contrast with other virtual methods which are
invoked through stub.

Change-Id: I0ac99b2f567b281c9342152bad1149dd6cd39480
Signed-off-by: Serguei Katkov <serguei.i.katkov@intel.com>
(cherry picked from commit 805bab1738549b2477b3ad4d9d57fd7c681451b9)