RFR: 8342818: Implement JEP 509: JFR CPU-Time Profiling [v5]

[Markus Grönlund]

On Mon, 26 May 2025 06:32:03 GMT, Johannes Bechberger <jbechberger at openjdk.org> wrote:

>> src/hotspot/share/jfr/periodic/sampling/jfrCPUTimeThreadSampler.cpp line 200:
>> 
>>> 198:   void sample_thread(JfrSampleRequest& request, void* ucontext, JavaThread* jt, JfrThreadLocal* tl);
>>> 199: 
>>> 200:   // sample all marked threads out of safepoint
>> 
>> What is meant by "out of safepoint"?
>
> Stack walking that occurs not at a safepoint, but during longer periods of a thread being in native. We can't walk the stack directly when we're in native, as we can't properly walk stacks in signal handlers. So we defer this to a separate thread.

A thread in native is still stackwalked at a safe location, a safepoint code position.

It's guaranteed by the last java frame (ljf).

>> src/hotspot/share/jfr/periodic/sampling/jfrCPUTimeThreadSampler.cpp line 373:
>> 
>>> 371:     }
>>> 372:   }
>>> 373:   assert(first_index == 0, "invariant");
>> 
>> How is this possible?
>
> Because we have a safepoint before the thread goes into native (as far as I understand). I'll remove the code above, because it is therefore not needed.

There is no safepoint when a thread goes from Java to native.

It's done on the way back.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/25302#discussion_r2106808431
PR Review Comment: https://git.openjdk.org/jdk/pull/25302#discussion_r2106814829