Loading...

XML

Word

Printable

Type: Bug
Resolution: Fixed
Priority: P2
Fix Version/s: 21
Affects Version/s: 20, 21
Component/s: hotspot
Labels:

Subcomponent:
jfr
Resolved In Build:
b07

Issue	Fix Version	Assignee	Priority	Status	Resolution	Resolved In Build
JDK-8301978	20.0.2	Stefan Karlsson	P2	Resolved	Fixed	b01
JDK-8301974	20.0.1	Stefan Karlsson	P2	Resolved	Fixed	b05
JDK-8300660	20	Stefan Karlsson	P2	Resolved	Fixed	b34

The JfrVFrameStream is used while generating stack traces for events. One of the events are the ZPage allocation event. This event is sometimes sent when ZGC is relocating. The current implementation of JfrVFrameStream uses WalkContinuation::include, which causes JFR to walk the continuation and perform GC barriers. This is problematic, since ZGC has a requirement that we never perform load barriers while running the relocation code. If we do, we might end up performing other reloctions from the the relocation code, and in some cases that causes dead locks.

I propose that JFR doesn't walk the continuations when sending events. An alternative could be to limit this to ZGC, but I'd like to get some feedback around that from JFR / Loom devs.

backported by

JDK-8300660 JfrVframeStream causes deadlocks in ZGC

Resolved

JDK-8301974 JfrVframeStream causes deadlocks in ZGC

Resolved

JDK-8301978 JfrVframeStream causes deadlocks in ZGC

Resolved

relates to

JDK-8298058 jdk/internal/vm/Continuation/Fuzz.java#default fails "assert(oopDesc::is_oop_or_null(val)) failed: bad oop found at 0xNNNN in_cont: 0 compressed: 0"

Closed

links to

Commit openjdk/jdk20/1330d4ea

Commit openjdk/jdk/453dbd12

Review openjdk/jdk20/120

Review openjdk/jdk/11586

Review openjdk/jdk/11778

(4 links to)

Assignee:: Stefan Karlsson

Reporter:: Stefan Karlsson

Votes:: 0 Vote for this issue

Watchers:: 7 Start watching this issue

Created:: 2022-12-08 03:16

Updated:: 2023-06-29 14:43

Resolved:: 2023-01-19 05:50

Details

Backports

Description

Attachments

Issue Links

Activity

People

Dates