8310264: In PhaseChaitin::Split defs and phis are leaked #14530

jdksjolen · 2023-06-17T16:08:53Z

Hi,

defs and phis are leaked as they are resource allocated but not protected by a ResourceMark. The intention might have been for these to also live in the split_arena.. This change is the most conservative one, however, and does fix the memory leak.

Please consider, thanks.

Johan

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8310264: In PhaseChaitin::Split defs and phis are leaked (Bug - P4)

Reviewers

Tobias Hartmann (@TobiHartmann - Reviewer)
Christian Hagedorn (@chhagedorn - Reviewer)
Vladimir Kozlov (@vnkozlov - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/14530/head:pull/14530
$ git checkout pull/14530

Update a local copy of the PR:
$ git checkout pull/14530
$ git pull https://git.openjdk.org/jdk.git pull/14530/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 14530

View PR using the GUI difftool:
$ git pr show -t 14530

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/14530.diff

Webrev

Link to Webrev Comment

bridgekeeper · 2023-06-17T16:10:26Z

👋 Welcome back jsjolen! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2023-06-17T16:12:59Z

@jdksjolen The following label will be automatically applied to this pull request:

hotspot-compiler

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2023-06-17T16:17:21Z

Webrevs

TobiHartmann

Looks good to me.

openjdk · 2023-06-19T05:28:12Z

@jdksjolen This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8310264: In PhaseChaitin::Split defs and phis are leaked

Reviewed-by: thartmann, chagedorn, kvn

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 186 new commits pushed to the master branch:

7fffdb5: 8310405: Linker.Option.firstVariadicArg should specify which index values are valid
b6c789f: 8309140: ResourceHashtable failed "assert(~(_allocation_t[0] | allocation_mask) == (uintptr_t)this) failed: lost resource object"
9f98136: 6956385: URLConnection.getLastModified() leaks file handles for jar:file and file: URLs
46e4ee1: 8310974: NMT: Arena diffs miss the scale
f17bfee: 8311034: Fix typo in javac man page
2ccdd29: 8299825: Move StdoutLog and StderrLog to LogConfiguration
e3f18af: 8311007: jdk/jfr/tool/TestView.java can't find event
08c51f2: 8310920: Fix -Wconversion warnings in command line flags
c2e9485: 8310921: Fix -Wconversion warnings from GenerateOopMap
ef71c32: 8310110: Shenandoah: Trace page sizes
... and 176 more: https://git.openjdk.org/jdk/compare/955fc2fabafc4c4d79d137bf86194228d5cf8ad8...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

chhagedorn

Looks good! I'm wondering, why we don't stack allocate both Node_List instead of new Node_List(). But regardless of that, we should indeed add a ResourceMark.

shipilev · 2023-06-19T07:29:10Z

Looking at f0d08c0 -- it would seem that def and phi just missed the allocation in split_arena? I think it would a cleaner/safer to change def/phi to be allocated in split_arena, rather than doing the blank ResourceMark here. Are we sure nothing RA-allocated is used after we leave Split?

jdksjolen · 2023-06-20T10:30:20Z

Hi,

In pursuit of answering your questions I saw some more opportunities for improvement. I did end up removing the ResourceMark as the size of Split makes it difficult to see whether it changes something that's used later on or not.

First of all, the VectorSets allocated their elements on the arena, but themselves on the resource area, this was another unhandled memory leak. I fixed that by allocating the VSets themselves on the arena, also. This removes a leak of spill_cnt*56 bytes.

Second of all @shipilev asked a good question, and I think that we can move defs and phis into the arena. I was worried about the sizes of these arrays being quite large and thus triggering a lot of reallocations, but they're typically quite small (I measured) while sometimes reaching to defs being 256 and phis 512 elements. I picked a conservative lower bound of 8/16.

For what it's worth, the number of VectorSets seems to strongly correlate with the size of the sets. a future RFE might make these larger by default. All of these measurements are done by running a Spring Hello World-app as a substitute for a 'real'/'typical' workload.

TobiHartmann

Good catch! Looks good to me otherwise.

TobiHartmann · 2023-06-23T07:03:49Z

src/hotspot/share/opto/reg_split.cpp

+  Node_List defs{split_arena, 8};
+  Node_List phis{split_arena, 16};


Why do you use aggregate initialization instead of constructor invocation here?

This will call the constructor, and has since C++11 I believe. However, I'm clearly being inconsistent here and with the VectorSet change above. There's no reason that I picked brace initializer other than it being the 'modern' way: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Res-list

I'll revert this change and use regular parens to be more stylistically consistent.

Looking at https://github.com/openjdk/jdk/blob/master/doc/hotspot-style.md "uniform initialization", it's OK to use.

Thanks for the background, I wasn't aware of that. I don't have a strong opinion but consistency in the same area would be nice.

Please, change to normal (). Using '{}' is very confusing for not modern C++ experts and affects maintainability of this code.

Thanks for the input, reverted to ().

jdksjolen · 2023-06-28T13:31:23Z

@chhagedorn , @TobiHartmann , would you mind re-approving this PR? Thank you.

chhagedorn

Update looks good!

TobiHartmann

Looks good to me too.

vnkozlov

Good.

jdksjolen · 2023-06-28T21:19:16Z

Thank you.

/integrate

openjdk · 2023-06-28T21:20:38Z

Going to push as commit 02b17d7.
Since your change was applied there have been 191 commits pushed to the master branch:

a63afa4: 8294427: Check boxes and radio buttons have rendering issues on Windows in High DPI env
3df36c4: 8310061: Note if implicit annotation processing is being used
da0f832: 8310606: Fix signed integer overflow, part 3
f0c2f09: 8296972: [macos13] java/awt/Frame/MaximizedToIconified/MaximizedToIconified.java: getExtendedState() != 6 as expected.
9f46fc2: 8310906: Fix -Wconversion warnings in runtime, oops and some code header files.
7fffdb5: 8310405: Linker.Option.firstVariadicArg should specify which index values are valid
b6c789f: 8309140: ResourceHashtable failed "assert(~(_allocation_t[0] | allocation_mask) == (uintptr_t)this) failed: lost resource object"
9f98136: 6956385: URLConnection.getLastModified() leaks file handles for jar:file and file: URLs
46e4ee1: 8310974: NMT: Arena diffs miss the scale
f17bfee: 8311034: Fix typo in javac man page
... and 181 more: https://git.openjdk.org/jdk/compare/955fc2fabafc4c4d79d137bf86194228d5cf8ad8...master

Your commit was automatically rebased without conflicts.

openjdk · 2023-06-28T21:20:48Z

@jdksjolen Pushed as commit 02b17d7.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

Add in missing ResourceMark

1b0c549

openjdk bot added the rfr Pull request is ready for review label Jun 17, 2023

openjdk bot added the hotspot-compiler hotspot-compiler-dev@openjdk.org label Jun 17, 2023

TobiHartmann approved these changes Jun 19, 2023

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Jun 19, 2023

chhagedorn approved these changes Jun 19, 2023

View reviewed changes

jdksjolen added 5 commits June 20, 2023 10:19

Stack allocate

01d8f9f

Allocate whole of VectorSet on arena

0fecd85

Skip the resource mark, add in separate temporary arenas for each.

820b084

Move these defs

e6ab09e

Be conservative in sizing to be close to original behavior

7035bb4

TobiHartmann approved these changes Jun 23, 2023

View reviewed changes

Revert to list init

52e3acf

chhagedorn approved these changes Jun 28, 2023

View reviewed changes

TobiHartmann approved these changes Jun 28, 2023

View reviewed changes

vnkozlov approved these changes Jun 28, 2023

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label Jun 28, 2023

openjdk bot closed this Jun 28, 2023

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Jun 28, 2023

		Node_List defs{split_arena, 8};
		Node_List phis{split_arena, 16};

8310264: In PhaseChaitin::Split defs and phis are leaked #14530

8310264: In PhaseChaitin::Split defs and phis are leaked #14530

Uh oh!

Conversation

jdksjolen commented Jun 17, 2023 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Webrev

Uh oh!

bridgekeeper bot commented Jun 17, 2023

Uh oh!

openjdk bot commented Jun 17, 2023

Uh oh!

mlbridge bot commented Jun 17, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

TobiHartmann left a comment

Choose a reason for hiding this comment

Uh oh!

openjdk bot commented Jun 19, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chhagedorn left a comment

Choose a reason for hiding this comment

Uh oh!

shipilev commented Jun 19, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jdksjolen commented Jun 20, 2023

Uh oh!

TobiHartmann left a comment

Choose a reason for hiding this comment

Uh oh!

TobiHartmann Jun 23, 2023

Choose a reason for hiding this comment

Uh oh!

jdksjolen Jun 26, 2023

Choose a reason for hiding this comment

Uh oh!

jdksjolen Jun 26, 2023

Choose a reason for hiding this comment

Uh oh!

TobiHartmann Jun 26, 2023

Choose a reason for hiding this comment

Uh oh!

vnkozlov Jun 26, 2023

Choose a reason for hiding this comment

Uh oh!

jdksjolen Jun 28, 2023

Choose a reason for hiding this comment

Uh oh!

jdksjolen commented Jun 28, 2023

Uh oh!

chhagedorn left a comment

Choose a reason for hiding this comment

Uh oh!

TobiHartmann left a comment

Choose a reason for hiding this comment

Uh oh!

vnkozlov left a comment

Choose a reason for hiding this comment

Uh oh!

jdksjolen commented Jun 28, 2023

Uh oh!

openjdk bot commented Jun 28, 2023

Uh oh!

openjdk bot commented Jun 28, 2023

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

5 participants

jdksjolen commented Jun 17, 2023 •

edited by openjdk bot

Loading

mlbridge bot commented Jun 17, 2023 •

edited

Loading

openjdk bot commented Jun 19, 2023 •

edited

Loading

shipilev commented Jun 19, 2023 •

edited

Loading