-
Bug
-
Resolution: Fixed
-
P2
-
17, 18, 19, 20
-
b21
Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
---|---|---|---|---|---|---|
JDK-8297512 | 19.0.2 | Tobias Hartmann | P2 | Closed | Fixed | b07 |
JDK-8296278 | 17.0.7-oracle | Tobias Hartmann | P2 | Closed | Fixed | b01 |
JDK-8299583 | 17.0.7 | Goetz Lindenmaier | P2 | Resolved | Fixed | b01 |
I also noticed that SharedRuntime::fixup_callers_callsite() patches callsites if the c2i adapter's Method's code is is_in_use(). It seems to miss an "&& !is_unloading()" in there. In practice that has not made a big difference until I removed the sweeper. Because nmethods would typically be is_unloading, because the Method is also unloading, in which case the c2i adapter entry barrier will take care of it. But now that an nmethod can become is_unloading also because it's "cold" or such, its Method is way more likely to be is_alive. And then we can get past the c2i adapter entry barrier, and end up calling fixup_callers_callsite, and miss the is_unloading check on the code, and break unlinking monotonicity of the inline cache cleaning. That could have pretty disasterous consequences when using ZGC.
- backported by
-
JDK-8299583 missing is_unloading() check in SharedRuntime::fixup_callers_callsite()
-
- Resolved
-
-
JDK-8296278 missing is_unloading() check in SharedRuntime::fixup_callers_callsite()
-
- Closed
-
-
JDK-8297512 missing is_unloading() check in SharedRuntime::fixup_callers_callsite()
-
- Closed
-
- relates to
-
JDK-8293648 Kitchensink.java failed with SIGSEGV in compiled frame with ZGC
-
- Closed
-
-
JDK-8295097 DirectNativeCallWrapper::verify_resolve_call fails with "assert(db != __null && !db->is_adapter_blob()) failed: must use stub!"
-
- Closed
-
-
JDK-8291832 RenaissanceStressTest24H.java fails with "fatal error: missing receiver type check"
-
- Closed
-
- links to
-
Commit openjdk/jdk17u-dev/013709f6
-
Commit openjdk/jdk19u/d1268e96
-
Commit openjdk/jdk/b5efa2af
-
Review openjdk/jdk17u-dev/1021
-
Review openjdk/jdk19u/81
-
Review openjdk/jdk/10747