8305670: Performance regression in LockSupport.unpark with lots of idle threads

[dcubed-ojdk]

A trivial fix to only check the ThreadsList in a ThreadsListHandle::cv_internal_thread_to_JavaThread()
call in fastdebug or slowdebug bits by default. The -XX:+EnableThreadSMRExtraValidityChecks option
can enable the check in release bits and the -XX:-EnableThreadSMRExtraValidityChecks option can
disable the check in fastdebug or slowdebug bits.

fastdebug or slowdebug bits will now fail an assert() if the java_thread is NOT found on the
ThreadsList (which should never happen).

---

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-8305670: Performance regression in LockSupport.unpark with lots of idle threads

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/13393/head:pull/13393
$ git checkout pull/13393

Update a local copy of the PR:
$ git checkout pull/13393
$ git pull https://git.openjdk.org/jdk.git pull/13393/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 13393

View PR using the GUI difftool:
$ git pr show -t 13393

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/13393.diff

Webrev

Link to Webrev Comment

[dcubed-ojdk]

/label add hotspot-runtime

[bridgekeeper]

👋 Welcome back dcubed! 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]

@dcubed-ojdk
The hotspot-runtime label was successfully added.

[mlbridge]

Webrevs

• 00: Full (dbc011d6)

[dcubed-ojdk]

Did some manual testing with this fix by adding code that would set the
is_on_list variable to false when the -XX:+UseNewCode option was
specified. That allowed me to verify the right output for the assertion failure:

#  Internal Error (/System/Volumes/Data/work/shared/bug_hunt/XXXXXXX_for_jdk21.git/open/src/hotspot/share/runtime/threadSMR.cpp:841), pid=26005, tid=2146563
#  assert(is_on_list) failed: java_thread=0x00007fac2ac29810 is not on ThreadsList(0x00007fac2ed51240)

I verified that the target thread (0x00007fac2ac29810) is listed in the
"Threads class SMR info:" section of the hs_err_pid file and I verified that
the ThreadsList (0x00007fac2ed51240) matches the current system
ThreadsList.

Also testing with Mach5 Tier[1-3].

[dcubed-ojdk]

It looks like a couple of tests are unhappy with the new code:

• runtime/handshake/HandshakeWalkOneExitTest.java
• runtime/handshake/HandshakeDirectTest.java

I'm going to switch back to draft while I sort this out...

[dcubed-ojdk]

To avoid confusion between the approach taken in this PR with new
approach that is being tested now, I'm closing this PR and will open
a new PR when I've done more testing of the merge of my latest
changes with @robehn's quick_mode idea.