RFR: 8314225: SIGSEGV in JavaThread::is_lock_owned

[Kevin Walls]

On Thu, 25 Jan 2024 14:20:08 GMT, Kevin Walls <kevinw at openjdk.org> wrote:

>> src/hotspot/share/runtime/javaThread.cpp line 1043:
>> 
>>> 1041:     chunks = _monitor_chunks; // will be null, as deopt frees when finished
>>> 1042:   } else {
>>> 1043:     ReadMonitorChunksHandshake rmch;
>> 
>> How expensive is the Handshake versus the direct read?  Is it worth optimistically reading using `monitor_chunks()` and only attempt the handshake if it returns non-null?
>> 
>> Or is there an API to probe if a thread is in a deopt that we can wrap around the handshake?
>> 
>> If the handshake is cheap enough, then this isn't worth looking at further.
>
> Actually yes I was thinking about checking the value for non-null first, I will try that out.  No handshake has to be faster... 8-)

Updated - if _monitor_chunks is null, avoid the handshake, and the check if we are at a safepoint.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/17566#discussion_r1466983462