8373695: G1: Using a value near integer max for ActiveProcessorCount causes fatal crash

[JonasNorlinder]

G1ConcRefinementThreads default value depends on ParallelGCThreads which in turn may depend on ActiveProcessorCount. These have different ranges that are valid. Patch will ensure argument setup code never try to assign a value larger than what G1ConcRefinementThreads expects.

---

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-8373695: G1: Using a value near integer max for ActiveProcessorCount causes fatal crash (Bug - P5)

Reviewers

• Stefan Karlsson (@stefank - Reviewer)
• Thomas Schatzl (@tschatzl - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/28831/head:pull/28831
$ git checkout pull/28831

Update a local copy of the PR:
$ git checkout pull/28831
$ git pull https://git.openjdk.org/jdk.git pull/28831/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 28831

View PR using the GUI difftool:
$ git pr show -t 28831

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/28831.diff

Using Webrev

Link to Webrev Comment

[bridgekeeper]

👋 Welcome back jnorlinder! 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]

@JonasNorlinder 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:

8373695: G1: Using a value near integer max for ActiveProcessorCount causes fatal crash

Reviewed-by: stefank, tschatzl

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 238 new commits pushed to the master branch:

• 08ff16f: 8374576: Disable MemoryEaterMT for VirtualThread
• a71326a: 8374528: C2 SuperWord: TestAliasingFuzzer.java strengthen no-multiversioning IR rule
• 1a6da44: 8374467: Incorrect ranges in jdk.internal.util.ByteArray JavaDoc
• ... and 235 more: https://git.openjdk.org/jdk/compare/ae85d899d074c531371dece30319ace701517528...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.

As you do not have Committer status in this project an existing Committer must agree to sponsor your change. Possible candidates are the reviewers of this PR (@stefank, @tschatzl) but any other Committer may sponsor as well.

➡️ To flag this PR as ready for integration with the above commit message, type /integrate in a new comment. (Afterwards, your sponsor types /sponsor in a new comment to perform the integration).

[openjdk]

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

• hotspot-gc

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]

Webrevs

• 03: Full - Incremental (62f78a13)
• 02: Full - Incremental (aee11b28)
• 01: Full - Incremental (416b88ed)
• 00: Full (a7d72b06)

[stefank]

Could an alternative be to limit ParallelGCThreads?

[JonasNorlinder]

Thanks for the suggestion Stefan. Thomas had the same comment and I’m open to it. But are we OK with using the upper bound limited by G1ConcRefinementThreads?

[tschatzl]

Thanks for the suggestion Stefan. Thomas had the same comment and I’m open to it. But are we OK with using the upper bound limited by G1ConcRefinementThreads?

Not sure what the question is, but limiting ParallelGCThreads by max_jint / 8 is an extremely generous limit anyway which should not give issues. The constant needs to be moved somewhere else though.

[JonasNorlinder]

FYI; I noticed that applying a limit on ParallelGCThreads still makes it possible to sneak in a larger value with ActiveProcessorCount. Looking into the interaction with ActiveProcessorCount with the rest of the code

[stefank]

I propose that you just add (max_jint - 1) / wordSize to the ParallelGCThread line, and let that be the entire fix.

I'm always hesitant to adding globals, functions, and classes to CollectedHeap. It requires extra motivation and acceptance from devs for all GCs. Often the proposal isn't a perfect fit for all GCs and we'll have to clean it out later. Here it is not clear if "WorkerThreadLimit" applies to all GCs. For example, ZGC could theoretically use more worker threads via ConcGCThreads. Not that anyone would, but I'm bringing this up to show a point that messing with CollectedHeap will likely make your PR a bit controversial and delay its review.

Another concern about this is that we haven't figured out a good, and unified way to bring constants into our GC flag cross-macros. I'd prefer if we could figure that out in a context of a separate RFE/discussion instead of in a bug fix.

Then we can see if our prioritize to figure out a good way to specify constants for our flag ranges rise to the level where we can sit down and figure out a clean way to do that.

[JonasNorlinder]

Thanks for that heads up! FWIW; At least from my side there is no rushing on getting this one accepted so we can take our time.

I propose that you just add (max_jint - 1) / wordSize to the ParallelGCThread line, and let that be the entire fix.

Did you mean this?

  product(uint, ParallelGCThreads, 0,                                       \
          "Number of parallel threads parallel gc will use")                \
          range(0, (max_jint-1)/wordSize)                                   \

I don't think that will not solve this bug when ParallelGCThreads is heuristically determined and may (i.e. it is defined) depend on ActiveProcessorCount. Applying the above patch and running with java -XX:+UseG1GC -XX:ActiveProcessorCount=2147483647 results in a fatal crash on my machine. That being said, it do solve the fatal crash that can be triggered with java -XX:+UseG1GC -XX:ParallelGCThreads=2147483646.

[stefank]

Did you mean this?

  product(uint, ParallelGCThreads, 0,                                       \
          "Number of parallel threads parallel gc will use")                \
          range(0, (max_jint-1)/wordSize)                                   \

Yes.

I don't think that will not solve this bug when ParallelGCThreads is heuristically determined and may (i.e. it is defined) depend on ActiveProcessorCount. Applying the above patch and running with java -XX:+UseG1GC -XX:ActiveProcessorCount=2147483647 results in a fatal crash on my machine. That being said, it do solve the fatal crash that can be triggered with java -XX:+UseG1GC -XX:ParallelGCThreads=2147483646.

My comment was made in the context provided before you found that issue.

[tschatzl]

Maybe use the flags API to programmatically get the maximum instead of hardcoding the value, something like this:

    const JVMTypedFlagLimit<uint>* conc_refinement_threads_limits = JVMFlagLimit::get_range_at(FLAG_MEMBER_ENUM(G1ConcRefinementThreads))->cast<uint>();
    FLAG_SET_ERGO(G1ConcRefinementThreads, MIN2(ParallelGCThreads, conc_refinement_threads_limits->max()));

[JonasNorlinder]

/integrate

[openjdk]

@JonasNorlinder
Your change (at version 62f78a1) is now ready to be sponsored by a Committer.

[tschatzl]

/sponsor

[openjdk]

Going to push as commit a855224.
Since your change was applied there have been 260 commits pushed to the master branch:

• 4231328: 8374698: Stub names should look more like identifiers
• a4fb07e: 8374644: Regression in GZIPInputStream performance after JDK-7036144
• 775f48d: 8365570: C2 fails assert(false) failed: Unexpected node in SuperWord truncation: CastII
• ... and 257 more: https://git.openjdk.org/jdk/compare/ae85d899d074c531371dece30319ace701517528...master

Your commit was automatically rebased without conflicts.

[openjdk]

@tschatzl @JonasNorlinder Pushed as commit a855224.

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