8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread #26287

dholmes-ora · 2025-07-14T05:34:57Z

The cv_internal_thread_to_JavaThread method will now check if the thread is a mounted virtual thread, and if so protect the carrier thread. User's of the API that need to deal with virtual threads must still check the carrier themselves as it could change asynchronously, but it is now guaranteed that the carrier JavaThread returned via this method cannot terminate whilst being used (the same as regular platform JavaThreads).

It was noticed whilst updating the documentation that the JvmtiExport variant cv_oop_to_JavaThread is unused and so can be removed.

Testing:

tiers 1-4

Thanks

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-8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread (Bug - P3)

Reviewers

Alex Menkov (@alexmenkov - Reviewer) Review applies to 820f26e4
Serguei Spitsyn (@sspitsyn - Reviewer) Review applies to 820f26e4
Patricio Chilano Mateo (@pchilano - Reviewer)
Daniel D. Daugherty (@dcubed-ojdk - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/26287/head:pull/26287
$ git checkout pull/26287

Update a local copy of the PR:
$ git checkout pull/26287
$ git pull https://git.openjdk.org/jdk.git pull/26287/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 26287

View PR using the GUI difftool:
$ git pr show -t 26287

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/26287.diff

Using Webrev

Link to Webrev Comment

for virtual threads, and add a new oop variant.

bridgekeeper · 2025-07-14T05:35:51Z

👋 Welcome back dholmes! 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 · 2025-07-14T05:36:02Z

@dholmes-ora 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:

8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread  does not deal with a virtual thread's carrier thread

Reviewed-by: pchilanomate, dcubed, amenkov, sspitsyn

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 201 new commits pushed to the master branch:

ea0b49c: 8362979: C2 fails with unexpected node in SuperWord truncation: CmpLTMask, RoundF
fe4d7f8: 8360507: JPKG002-006: SigningPackageTest: appOutput.txt cannot be found in user home directory and output doesn't contain: origin=Developer ID Application: jpackage.openjdk.java.net
f53d051: 8340312: sun.security.ssl.SSLLogger uses incorrect log level ALL for finest log events
... and 198 more: https://git.openjdk.org/jdk/compare/ee0d309bbd33302d8c6f35155e975db77aaea785...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.

openjdk · 2025-07-14T05:36:37Z

@dholmes-ora The following labels will be automatically applied to this pull request:

hotspot
serviceability

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

…ew API" This reverts commit 4ed9035.

mlbridge · 2025-07-21T08:32:16Z

Webrevs

src/hotspot/share/runtime/threadSMR.hpp

alexmenkov · 2025-07-22T21:55:56Z

I still think that it would be safer to explicitly specify that caller knows that jthread can be virtual thread (and maybe add assert when a caller works only with platform threads, but passed jthread is virtual), but ok, it's up to you.
Currently all callers except ThreadSnapshotFactory::get_thread_snapshot use cv_internal_thread_to_JavaThread for platform threads only. Do you want to update ThreadSnapshotFactory::get_thread_snapshot to utilize new functionality or prefer to do it by separate CR?

dholmes-ora · 2025-07-23T00:01:37Z

I still think that it would be safer to explicitly specify that caller knows that jthread can be virtual thread

I think that retains the risk of the caller not knowing they need to ensure they protect the carrier thread.

Do you want to update ThreadSnapshotFactory::get_thread_snapshot to utilize new functionality

No, we had some internal discussions and that code is incorrectly using the JVMTI transition disabler, so it needs a complete overhaul. That will be handled separately.

Thanks again for looking at this in depth Alex.

sspitsyn · 2025-07-24T09:44:20Z

src/hotspot/share/runtime/threadSMR.cpp

+         java_thread = java_lang_Thread::thread(carrier_thread);
+       }
+       if (java_thread == nullptr) {
+         // Virtual thread was unmounted, or else carrier has now terminated.


Nit: If java_thread for a virtual thread is null then this thread is unmounted. The words about the terminated carrier thread are confusing.

I have one concern here. Nothing protects this code from facing some unexpected conditions as this function is racy with mounting and unmounting of target virtual thread

this function can observe a JavaThread in mount or unmount transition where the thread identity of a java.lang.Thread associated with the JavaThread is not consistent (in a VTMS transition the thread identity might not match the stack trace)

nothing guaranties that the result returned by this function remains the same upon return as mount/unmount of the target virtual thread can be executed concurrently: a mounted virtual thread can be quickly unmounted or an unmounted virtual thread can be quickly mounted

So, it seems that this function is going to work correctly if used for target platform threads only or if the JavaThread* pointer is not really needed. Otherwise, the association with the JavaThread needs to be rechecked and its stability somehow guaranteed with any form of scheduling suspension or a VTMS transition disabler.

The java_thread can be null because the carrier of a mounted vthread was not contained in the ThreadsListHandle.

The only thing this function guarantees is that if the vthread appeared to have a carrier and that carrier is protected by the TLH, then the caller can safely interact with the carrier knowing it can't terminate - same as for regular threads. The caller has to account for the potential async mounting/unmounting of vthread - e.g. by handshaking the reported carrier and then confirming it is still the carrier.

Note that java_thread may already be null so we don't get to execute line 836.

Thanks, David. The potential issue I'm still concerned about is that a subsequent handshaking can also observe the JavaThread (and virtual thread as well) in a VTMS transition when there is no protection with a VTMSTransitionDisabler.

Note that java_thread may already be null so we don't get to execute line 836.

Right. I guess, this answers my comment about the line 839.

Okay, thanks.

Once this change is in main line then we can sync'up the loom repo and work on the follow-up changes. The loom repo has changes (that are not in main line) for "suspending" an unmounted thread. For the mounted case then we'll need checks in the handshake to ensure that the expected virtual threads is mounted. We'll need stress tests of course and we can collaborate there in advance of proposing changes for main line.

Note that java_thread may already be null so we don't get to execute line 836.

I'm having some trouble with this statement. Working backwards from L836, I see
the nullptr check on L826 and the check for non-virtual thread and bail on L827
and L831. However, if we saw nullptr on L826 and we have a virtual thread, then
we can get to L836 even when we started off with a null java_thread.

So I guess I'm not sure what point you're trying to make with the statement.

then we can get to L836 even when we started off with a null java_thread.

Sure. My point was more the other way round - we could have a null JavaThread without the null coming from line 836.

The comment was just trying to expand on my previous comment.

Thanks for the clarification.

sspitsyn

This looks okay to me.

dholmes-ora · 2025-07-25T06:39:08Z

Thanks for the review @sspitsyn !

pchilano

LGTM.

src/hotspot/share/runtime/threadSMR.hpp

dcubed-ojdk

Thumbs up on the changes. I think I only have a few nits.

src/hotspot/share/runtime/threadSMR.cpp

dcubed-ojdk · 2025-07-25T17:48:01Z

src/hotspot/share/runtime/threadSMR.cpp

+         java_thread = java_lang_Thread::thread(carrier_thread);
+       }
+       if (java_thread == nullptr) {
+         // Virtual thread was unmounted, or else carrier has now terminated.


Note that java_thread may already be null so we don't get to execute line 836.

I'm having some trouble with this statement. Working backwards from L836, I see
the nullptr check on L826 and the check for non-virtual thread and bail on L827
and L831. However, if we saw nullptr on L826 and we have a virtual thread, then
we can get to L836 even when we started off with a null java_thread.

So I guess I'm not sure what point you're trying to make with the statement.

src/hotspot/share/runtime/threadSMR.hpp

src/hotspot/share/prims/jvmtiExport.cpp

dholmes-ora · 2025-07-27T21:33:20Z

Thanks for the review @dcubed-ojdk !

Just need one Reviewer to re-review the nit fix-ups. Thanks

dcubed-ojdk

Thanks for the updates. Thumbs up.

dholmes-ora · 2025-07-28T23:46:10Z

Thanks Dan!

/integrate

openjdk · 2025-07-28T23:47:54Z

Going to push as commit 4669005.
Since your change was applied there have been 202 commits pushed to the master branch:

3e4e5dd: 8364081: Shenandoah & GenShen logging improvement
ea0b49c: 8362979: C2 fails with unexpected node in SuperWord truncation: CmpLTMask, RoundF
fe4d7f8: 8360507: JPKG002-006: SigningPackageTest: appOutput.txt cannot be found in user home directory and output doesn't contain: origin=Developer ID Application: jpackage.openjdk.java.net
... and 199 more: https://git.openjdk.org/jdk/compare/ee0d309bbd33302d8c6f35155e975db77aaea785...master

Your commit was automatically rebased without conflicts.

openjdk · 2025-07-28T23:48:04Z

@dholmes-ora Pushed as commit 4669005.

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

dholmes-ora added 3 commits July 14, 2025 01:30

Adapt thread conversion routines to return (and protect) the carrier

8099035

for virtual threads, and add a new oop variant.

Adapt ThreadSnapshotFactory::get_thread_snapshot to use the new API

4ed9035

Remove unused JvmtiExport::cv_oop_to_JavaThread API

9e13596

openjdk bot added serviceability [email protected] hotspot [email protected] labels Jul 14, 2025

dholmes-ora changed the title ~~8361912 thread smr~~ 8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread Jul 14, 2025

openjdk bot changed the title ~~8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread~~ 8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread Jul 14, 2025

dholmes-ora added 2 commits July 21, 2025 03:48

Revert "Adapt ThreadSnapshotFactory::get_thread_snapshot to use the n…

5d7eb80

…ew API" This reverts commit 4ed9035.

Revert to simpler fix without new API

a195fe2

dholmes-ora marked this pull request as ready for review July 21, 2025 08:27

openjdk bot added the rfr Pull request is ready for review label Jul 21, 2025

alexmenkov reviewed Jul 21, 2025

View reviewed changes

src/hotspot/share/runtime/threadSMR.hpp Outdated Show resolved Hide resolved

Move comment

820f26e

alexmenkov approved these changes Jul 23, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Jul 23, 2025

sspitsyn reviewed Jul 24, 2025

View reviewed changes

sspitsyn approved these changes Jul 25, 2025

View reviewed changes

pchilano approved these changes Jul 25, 2025

View reviewed changes

src/hotspot/share/runtime/threadSMR.hpp Outdated Show resolved Hide resolved

dcubed-ojdk approved these changes Jul 25, 2025

View reviewed changes

Fix comments

d863add

openjdk bot removed the ready Pull request is ready to be integrated label Jul 27, 2025

pchilano approved these changes Jul 28, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Jul 28, 2025

dcubed-ojdk approved these changes Jul 28, 2025

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label Jul 28, 2025

openjdk bot closed this Jul 28, 2025

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Jul 28, 2025

dholmes-ora deleted the 8361912-ThreadSMR branch July 28, 2025 23:53

8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread #26287

8361912: ThreadsListHandle::cv_internal_thread_to_JavaThread does not deal with a virtual thread's carrier thread #26287

Uh oh!

Conversation

dholmes-ora commented Jul 14, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Uh oh!

bridgekeeper bot commented Jul 14, 2025

Uh oh!

openjdk bot commented Jul 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk bot commented Jul 14, 2025

Uh oh!

mlbridge bot commented Jul 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

Uh oh!

alexmenkov commented Jul 22, 2025

Uh oh!

dholmes-ora commented Jul 23, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sspitsyn Jul 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sspitsyn left a comment

Choose a reason for hiding this comment

Uh oh!

dholmes-ora commented Jul 25, 2025

Uh oh!

pchilano left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

dcubed-ojdk left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dholmes-ora commented Jul 27, 2025

Uh oh!

dcubed-ojdk left a comment

Choose a reason for hiding this comment

Uh oh!

dholmes-ora commented Jul 28, 2025

Uh oh!

openjdk bot commented Jul 28, 2025

Uh oh!

openjdk bot commented Jul 28, 2025

Uh oh!

Uh oh!

dholmes-ora commented Jul 14, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Jul 14, 2025 •

edited

Loading

mlbridge bot commented Jul 21, 2025 •

edited

Loading

sspitsyn Jul 24, 2025 •

edited

Loading