[11u] RFR(M): 8219586 backport: CodeHeap State Analytics processes dead nmethods

[Hohensee, Paul]

The maintainers prefer to see individual backports rather than bulk ones.

You're bulk backporting codeHeapState.cpp. Instead of that, you could backport these commits individually.

8219586: CodeHeap State Analytics processes dead nmethods
8217465: [REDO] - Optimize CodeHeap Analytics
8216314: SIGILL in CodeHeapState::print_names()
8214526: Change CodeHeap State Analytics control from UL to Print*

8214526: applies cleanly.
8216314: needs a copyright change in codeHeapState.cpp.
8217465:
Needs an implicit bufferedStream copy constructor removed in codeHeapState.cpp. This change was in 8224487, which was previously backported to 11u, but the change was omitted in that backport.
8219586:
Context adjustment in codeHeapState.cpp because 8183574 has not been backported.
Context adjustment in compiledMethod.cpp. Also, your webrev has global_lock_1 and function_lock_1 used with _no_safepoint_check_flag, but the tip version uses _safepoint_check_flag.
Context adjustment in mutexLocker.cpp. 

I made incremental webrevs of the above. They apply in sequence.

https://cr.openjdk.java.net/~phh/8214526/webrev.11u.01/
https://cr.openjdk.java.net/~phh/8216314/webrev.11u.01/
https://cr.openjdk.java.net/~phh/8217465/webrev.11u.01/
https://cr.openjdk.java.net/~phh/8219586/webrev.11u.02/

I'm running tier1 on linux-x64, but don't have access to other platforms. Would you be able to apply these patches and test them thoroughly? Are there any specific tests I can run beyond tier1?

Thanks,
Paul

-----Original Message-----
From: jdk-updates-dev <jdk-updates-dev-retn at openjdk.java.net> on behalf of "Schmidt, Lutz" <lutz.schmidt at sap.com>
Date: Wednesday, April 14, 2021 at 9:09 AM
To: "jdk-updates-dev at openjdk.java.net" <jdk-updates-dev at openjdk.java.net>
Subject: [11u] RFR(M): 8219586 backport: CodeHeap State Analytics processes dead nmethods

Dear Community,

I would happily accept reviews for this downport change. The change puts an end to discussions on the "accessing dead storage" topic, at least to the best of Erik Österlund's and my knowledge.

Most merge conflicts (and the complicated ones) were encountered codeHeapState.{c|h}pp. To resolve them, I copied the files from OpenJDK/jdk. That was the desired state for these files anyway.

To resolve compileBroker.cpp, I copied the OpenJDK/jdk version of CompileBroker::print_heapinfo() and retrofitted the Monitor and Mutex objects and calls to the jdk11 interfaces.

The remaining conflicts were trivial to resolve.

Original bug:          https://bugs.openjdk.java.net/browse/JDK-8219586
Downport webrev:       https://cr.openjdk.java.net/~lucy/webrevs/8219586.11u.01/
Merge conflicts:       https://cr.openjdk.java.net/~lucy/webrevs/8219586-jdk11u.conflicts
Conflict resolve diff: https://cr.openjdk.java.net/~lucy/webrevs/8219586-jdk11u.conflictresolve

Tests:
SAP's internal build and test farm (all OpenJDK platforms (no 32-bit), and more). Tests include JCK, jtreg (hotspot and jdk), and SAP-private tests. Results pending (expected by April 15th, start of business).

Thanks,
Lutz