JEP-370 Implementation (Part 3) #8501

babsingh · 2020-02-04T22:30:24Z

Closes: #8292.

1. Use OpenJDK ViewHandles in Java14+ to support JEP-370

OpenJDK implementation of ByteArrayViewHandle and ByteBufferViewHandle
is used to support JEP-370 in Java14+.

OpenJ9 ViewHandles don't handle memory scopes properly. Example:

try (MemorySegment segment = MemorySegment.allocateNative(bytes)) {
   ByteBuffer byteBuffer = segment.asByteBuffer();
   MethodHandle handle = Derived from ByteBuffer & ByteBufferViewHandle;
   handle.invoke();
}

Invoking handle.invoke() outside the memory scope should result in
IllegalStateException, which is thrown from jdk.internal.foreign.MemoryScope.

With OpenJ9 ViewHandles, the IllegalStateException is not thrown. The
correct behavior is achieved by using OpenJDK ViewHandles.

The deficiency in OpenJ9 ViewHandles could not be identified.

2. Handle NULL VarHandle.handleTable entries in VarHandle.accessModeType

3. Handle NULL VarHandle.handleTable entries in (Bytecode|MH)Interpreter

4. Handle NULL VarHandle.handleTable entries in VarHandle.toMethodHandle

5. Support isAccessModeSupportedHelper method with VarForm constructor

OpenJ9 ViewVarHandles have their own implementation for the
isAccessModeSupported method. OpenJDK ViewVarHandles are used in OpenJ9
Java 14+ which rely upon VarHandle.isAccessModeSupported instead of
having their own isAccessModeSupported implementation.

VarHandle.isAccessModeSupported is updated to support OpenJDK
VarHandles, which rely upon OpenJ9's VarHandle(VarForm varForm)
constructor.

Signed-off-by: Babneet Singh sbabneet@ca.ibm.com

babsingh · 2020-02-04T22:33:42Z

Depends on #8330, #8414 and #8460. Will require a rebase after dependents are merged.

runtime/bcutil/ConstantPoolMap.cpp

babsingh · 2020-02-13T21:27:34Z

@DanHeidinga Rebased this pull request since #8414 is merged (#8414 (comment)).

DanHeidinga · 2020-02-14T03:40:44Z

Jenkins test sanity plinux jdk11,jdk14

pshipton · 2020-02-14T13:34:45Z

jenkins compile win jdk8,jdk14

pshipton · 2020-02-14T16:13:36Z

Pls rebase to pick up #8583

babsingh · 2020-02-14T16:37:22Z

Rebased. @AdamBrousseau do we apply the pull request changes on top of openj9/master? are rebases required?

AdamBrousseau · 2020-02-14T16:49:53Z

Rebases are not required. We use the merge ref of the PR which is a merge of the PR into master. That being said, GitHub has publicly stated these are an undocumented, unsupported feature of a PR and they encourage users to create the merge themselves. I believe it is because when something is merged to master, not every pr is updated automatically by default. Certain events need to happen on a PR in order for the ref to be refreshed.

pshipton · 2020-02-14T16:54:29Z

jenkins compile win jdk8

pshipton · 2020-02-14T16:55:41Z

plinux PR testing is https://ci.eclipse.org/openj9/job/PullRequest-OpenJ9/2729/

pshipton · 2020-02-14T16:58:12Z

Past PR testing has found that sometimes (often?) a rebase is required to pick up other changes. I could have tried it without the rebase, but I know the rebase will fix it.

babsingh · 2020-02-14T18:42:41Z

https://ci.eclipse.org/openj9/job/PullRequest-OpenJ9/2729/: the plinux failure is a segfault in Jep359Tests_0, which is reported in #8582. This failure happens even without the changes of this pull request.

runtime/vm/MHInterpreter.cpp

runtime/oti/vmconstantpool.xml

OpenJDK implementation of ByteArrayViewHandle and ByteBufferViewHandle is used to support JEP-370 in Java14+. OpenJ9 ViewHandles don't handle memory scopes properly. Example: try (MemorySegment segment = MemorySegment.allocateNative(bytes)) { ByteBuffer byteBuffer = segment.asByteBuffer(); MethodHandle handle = Derived from ByteBuffer & ByteBufferViewHandle; handle.invoke(); } Invoking handle.invoke() outside the memory scope should result in IllegalStateException, which is thrown from jdk.internal.foreign.MemoryScope. With OpenJ9 ViewHandles, the IllegalStateException is not thrown. The aforementioned behavior is achieved by using OpenJDK ViewHandles. The deficiency in OpenJ9 ViewHandles could not be identified. Signed-off-by: Babneet Singh <sbabneet@ca.ibm.com>

In OpenJ9, derived VarHandle classes define all AccessMode operations, and UnsupportedOperationException is thrown from the operation methods that are not supported. All OpenJ9 VarHandle.handleTable entries are non-null when supporting derived OpenJ9 VarHandle classes. In OpenJDK, derived VarHandle classes do not define unsupported AccessMode operations. Thus, OpenJ9 VarHandle.handleTable entries are null for unsupported AccessMode operations when supporting derived OpenJDK VarHandle classes. VarHandle.accessModeType relies upon VarHandle.handleTable to derive the AccessMode MethodType. When supporting derived OpenJDK VarHandle classes, accessModeTypeUncached is used to retrieve the AccessMode MethodType in case VarHandle.handleTable entries are null. Signed-off-by: Babneet Singh <sbabneet@ca.ibm.com>

runtime/vm/MHInterpreter.cpp

babsingh · 2020-02-26T14:51:45Z

Currently, investigating the failing extended tests in Test_openjdk14_j9_extended.functional_x86-64_windows_Personal-2:

05:45:44  FAILED test targets:
05:45:44  	VarHandle_0
05:45:44  	View-LE-OnHeap_0
05:45:44  	View-LE-OffHeap_0
05:45:44  	View-BE-OnHeap_0
05:45:44  	View-BE-OffHeap_0

runtime/vm/BytecodeInterpreter.hpp

babsingh · 2020-02-26T16:44:39Z

The assertion failures in VarHandle_0 are addressed in the following commit:

Support isAccessModeSupportedHelper method with VarForm constructor

babsingh · 2020-02-26T16:53:52Z

All failures seen in the extended functional test suite have been addressed.

DanHeidinga

@gacholio Anything major I missed in the interpreter changes?

runtime/vm/BytecodeInterpreter.hpp

runtime/vm/MHInterpreter.cpp

runtime/vm/BytecodeInterpreter.hpp

gacholio · 2020-02-27T17:57:24Z

@DanHeidinga I suggested one trivial change.

It might be better to move all of this code out of the interpreter - just do the update/call/hasbeenupdated in the interpreter itself.

DanHeidinga · 2020-02-27T18:30:47Z

It might be better to move all of this code out of the interpreter - just do the update/call/hasbeenupdated in the interpreter itself.

@babsingh Can you update the code to move most of this out of the interpreter as GAC suggested?

gacholio · 2020-02-27T18:38:11Z

You'll need to use the J9AllocateObject API instead of the inline allocate.

babsingh · 2020-02-27T20:55:03Z

@gacholio @DanHeidinga any suggestions on where the new function should be added? VMHelpers.hpp?

gacholio · 2020-02-27T21:58:26Z

exceptionsupport.cpp - the whole idea is that it not be inlined into the interpreter.

babsingh · 2020-02-27T23:03:59Z

@DanHeidinga Moved code out of the interpreter as per @gacholio's suggestion. New code is added to the following commit:

Handle NULL VarHandle.handleTable entries in (Bytecode|MH)Interpreter

gacholio · 2020-02-28T03:46:29Z

Yeah, literal means the direct string. If it's a #define then it can be used.

runtime/vm/MHInterpreter.cpp

OpenJ9 VarHandle.handleTable entries are null for unsupported AccessMode operations when supporting derived OpenJDK VarHandle classes. VM_MHInterpreter::dispatchLoop and BytecodeInterpreter::invokevarhandle look up entries in VarHandle.handleTable to invoke the operation method. A null handleTable entry reflects a unsupported AccessMode operation. In this case, an UnsupportedOperationException should be thrown. Signed-off-by: Babneet Singh <sbabneet@ca.ibm.com>

OpenJ9 VarHandle.handleTable entries are null for unsupported AccessMode operations when supporting derived OpenJDK VarHandle classes. When supporting derived OpenJDK VarHandle classes, accessModeTypeUncached is used to retrieve the AccessMode MethodType in case VarHandle.handleTable entries are null. Signed-off-by: Babneet Singh <sbabneet@ca.ibm.com>

OpenJ9 ViewVarHandles have their own implementation for the isAccessModeSupported method. OpenJDK ViewVarHandles are used in OpenJ9 Java 14+ which rely upon VarHandle.isAccessModeSupported instead of having their own isAccessModeSupported implementation. VarHandle.isAccessModeSupported is updated to support OpenJDK VarHandles, which rely upon OpenJ9's VarHandle(VarForm varForm) constructor. Signed-off-by: Babneet Singh <sbabneet@ca.ibm.com>

DanHeidinga

lgtm

DanHeidinga · 2020-03-01T04:57:33Z

Jenkins test sanity zlinux jdk14,jdk11

DanHeidinga · 2020-03-01T04:57:52Z

Jenkins test extended win jdk14

DanHeidinga · 2020-03-01T04:58:27Z

~~@babsingh can you open a PR for this against the 0.19 release branch, assuming the PR builds will pass?~~

I've ported it in #8703

runtime/vm/exceptionsupport.c

babsingh force-pushed the jep370_impl4 branch 2 times, most recently from 21a245c to d721d80 Compare February 5, 2020 00:56

babsingh mentioned this pull request Feb 5, 2020

JEP-370 Foreign-Memory Access API #8292

Closed

babsingh marked this pull request as ready for review February 5, 2020 01:08

babsingh requested a review from DanHeidinga February 5, 2020 01:08

babsingh force-pushed the jep370_impl4 branch from d721d80 to 1d24c19 Compare February 5, 2020 01:28

DanHeidinga reviewed Feb 5, 2020

View reviewed changes

runtime/bcutil/ConstantPoolMap.cpp Outdated Show resolved Hide resolved

babsingh mentioned this pull request Feb 6, 2020

JEP-370 Implementation (Part 2) #8414

Merged

DanHeidinga self-assigned this Feb 13, 2020

DanHeidinga added this to the Release 0.19 (Java 14) milestone Feb 13, 2020

babsingh force-pushed the jep370_impl4 branch from 1d24c19 to 5efc135 Compare February 13, 2020 21:23

babsingh force-pushed the jep370_impl4 branch from 5efc135 to 3a4672a Compare February 14, 2020 16:36

DanHeidinga reviewed Feb 15, 2020

View reviewed changes

runtime/vm/MHInterpreter.cpp Outdated Show resolved Hide resolved

runtime/oti/vmconstantpool.xml Outdated Show resolved Hide resolved

babsingh added 2 commits February 17, 2020 22:03

babsingh force-pushed the jep370_impl4 branch 2 times, most recently from a894bae to 1f80a9f Compare February 18, 2020 21:00

DanHeidinga reviewed Feb 25, 2020

View reviewed changes

runtime/vm/MHInterpreter.cpp Outdated Show resolved Hide resolved

babsingh force-pushed the jep370_impl4 branch from 1f80a9f to 8b055be Compare February 25, 2020 19:58

babsingh force-pushed the jep370_impl4 branch from 8b055be to 8caf911 Compare February 26, 2020 16:33

babsingh commented Feb 26, 2020

View reviewed changes

runtime/vm/BytecodeInterpreter.hpp Outdated Show resolved Hide resolved

babsingh force-pushed the jep370_impl4 branch from 8caf911 to 1802734 Compare February 26, 2020 16:50

DanHeidinga reviewed Feb 27, 2020

View reviewed changes

runtime/vm/BytecodeInterpreter.hpp Show resolved Hide resolved

runtime/vm/BytecodeInterpreter.hpp Outdated Show resolved Hide resolved

runtime/vm/BytecodeInterpreter.hpp Outdated Show resolved Hide resolved

runtime/vm/MHInterpreter.cpp Outdated Show resolved Hide resolved

gacholio reviewed Feb 27, 2020

View reviewed changes

runtime/vm/BytecodeInterpreter.hpp Outdated Show resolved Hide resolved

babsingh force-pushed the jep370_impl4 branch from 1802734 to 1b56e1a Compare February 27, 2020 22:58

DanHeidinga reviewed Feb 28, 2020

View reviewed changes

runtime/vm/MHInterpreter.cpp Show resolved Hide resolved

babsingh added 3 commits February 28, 2020 15:32

babsingh force-pushed the jep370_impl4 branch from 1b56e1a to 0c32d79 Compare February 28, 2020 20:33

DanHeidinga approved these changes Mar 1, 2020

View reviewed changes

DanHeidinga merged commit dd6fe85 into eclipse-openj9:master Mar 1, 2020

DanHeidinga mentioned this pull request Mar 1, 2020

(v0.19.0) JEP-370 Implementation (Part 3) #8703

Merged

keithc-ca reviewed Mar 2, 2020

View reviewed changes

runtime/vm/exceptionsupport.c Show resolved Hide resolved

babsingh mentioned this pull request Mar 2, 2020

Add a null check to handle internalFindClassUTF8 failure #8713

Merged

babsingh deleted the jep370_impl4 branch August 17, 2020 02:24

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

JEP-370 Implementation (Part 3) #8501

JEP-370 Implementation (Part 3) #8501

babsingh commented Feb 4, 2020 •

edited

Loading

babsingh commented Feb 4, 2020

babsingh commented Feb 13, 2020 •

edited

Loading

DanHeidinga commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

babsingh commented Feb 14, 2020

AdamBrousseau commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

babsingh commented Feb 14, 2020 •

edited

Loading

babsingh commented Feb 26, 2020

babsingh commented Feb 26, 2020 •

edited

Loading

babsingh commented Feb 26, 2020 •

edited

Loading

DanHeidinga left a comment

gacholio commented Feb 27, 2020

DanHeidinga commented Feb 27, 2020

gacholio commented Feb 27, 2020

babsingh commented Feb 27, 2020

gacholio commented Feb 27, 2020

babsingh commented Feb 27, 2020

gacholio commented Feb 28, 2020

DanHeidinga left a comment

DanHeidinga commented Mar 1, 2020

DanHeidinga commented Mar 1, 2020

DanHeidinga commented Mar 1, 2020 •

edited

Loading

JEP-370 Implementation (Part 3) #8501

JEP-370 Implementation (Part 3) #8501

Conversation

babsingh commented Feb 4, 2020 • edited Loading

babsingh commented Feb 4, 2020

babsingh commented Feb 13, 2020 • edited Loading

DanHeidinga commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

babsingh commented Feb 14, 2020

AdamBrousseau commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

pshipton commented Feb 14, 2020

babsingh commented Feb 14, 2020 • edited Loading

babsingh commented Feb 26, 2020

babsingh commented Feb 26, 2020 • edited Loading

babsingh commented Feb 26, 2020 • edited Loading

DanHeidinga left a comment

Choose a reason for hiding this comment

gacholio commented Feb 27, 2020

DanHeidinga commented Feb 27, 2020

gacholio commented Feb 27, 2020

babsingh commented Feb 27, 2020

gacholio commented Feb 27, 2020

babsingh commented Feb 27, 2020

gacholio commented Feb 28, 2020

DanHeidinga left a comment

Choose a reason for hiding this comment

DanHeidinga commented Mar 1, 2020

DanHeidinga commented Mar 1, 2020

DanHeidinga commented Mar 1, 2020 • edited Loading

babsingh commented Feb 4, 2020 •

edited

Loading

babsingh commented Feb 13, 2020 •

edited

Loading

babsingh commented Feb 14, 2020 •

edited

Loading

babsingh commented Feb 26, 2020 •

edited

Loading

babsingh commented Feb 26, 2020 •

edited

Loading

DanHeidinga commented Mar 1, 2020 •

edited

Loading