RFR: 8338383: Implement JEP 491: Synchronize Virtual Threads without Pinning [v25]

[Axel Boldt-Christmas]

On Thu, 31 Oct 2024 21:50:50 GMT, Patricio Chilano Mateo <pchilanomate at openjdk.org> wrote:

>> This is the implementation of JEP 491: Synchronize Virtual Threads without Pinning. See [JEP 491](https://bugs.openjdk.org/browse/JDK-8337395) for further details.
>> 
>> In order to make the code review easier the changes have been split into the following initial 4 commits:
>> 
>> - Changes to allow unmounting a virtual thread that is currently holding monitors.
>> - Changes to allow unmounting a virtual thread blocked on synchronized trying to acquire the monitor.
>> - Changes to allow unmounting a virtual thread blocked in `Object.wait()` and its timed-wait variants.
>> - Changes to tests, JFR pinned event, and other changes in the JDK libraries.
>> 
>> The changes fix pinning issues for all 4 ports that currently implement continuations: x64, aarch64, riscv and ppc. Note: ppc changes were added recently and stand in its own commit after the initial ones.
>> 
>> The changes fix pinning issues when using `LM_LIGHTWEIGHT`, i.e. the default locking mode, (and `LM_MONITOR` which comes for free), but not when using `LM_LEGACY` mode. Note that the `LockingMode` flag has already been deprecated ([JDK-8334299](https://bugs.openjdk.org/browse/JDK-8334299)), with the intention to remove `LM_LEGACY` code in future releases.
>> 
>> 
>> ## Summary of changes
>> 
>> ### Unmount virtual thread while holding monitors
>> 
>> As stated in the JEP, currently when a virtual thread enters a synchronized method or block, the JVM records the virtual thread's carrier platform thread as holding the monitor, not the virtual thread itself. This prevents the virtual thread from being unmounted from its carrier, as ownership information would otherwise go wrong. In order to fix this limitation we will do two things:
>> 
>> - We copy the oops stored in the LockStack of the carrier to the stackChunk when freezing (and clear the LockStack). We copy the oops back to the LockStack of the next carrier when thawing for the first time (and clear them from the stackChunk). Note that we currently assume carriers don't hold monitors while mounting virtual threads.
>> 
>> - For inflated monitors we now record the `java.lang.Thread.tid` of the owner in the ObjectMonitor's `_owner` field instead of a JavaThread*. This allows us to tie the owner of the monitor to a `java.lang.Thread` instance, rather than to a JavaThread which is only created per platform thread. The tid is already a 64 bit field so we can ignore issues of the counter wrapping around.
>> 
>> #### General notes about this part:
>> 
>> - Since virtual th...
>
> Patricio Chilano Mateo has updated the pull request incrementally with two additional commits since the last revision:
> 
>  - add comment to ThreadService::find_deadlocks_at_safepoint
>  - Remove assignments in preempt_kind enum

src/hotspot/share/oops/stackChunkOop.cpp line 445:

> 443: 
> 444: void stackChunkOopDesc::transfer_lockstack(oop* dst) {
> 445:   const bool requires_gc_barriers = is_gc_mode() || requires_barriers();

Given how careful we are in `Thaw` to not call `requires_barriers()` twice and use `_barriers` instead it would probably be nicer to pass in `_barriers` as a bool. 

There is only one other place we do the extra call and it is in `fix_thawed_frame`, but that only happens after we are committed to the slow path, so it might be nice for completeness, but should be negligible for performance. Here however we might still be in our new "medium" path where we could still do a fast thaw.

src/hotspot/share/oops/stackChunkOop.cpp line 460:

> 458:     } else {
> 459:       oop value = *reinterpret_cast<oop*>(at);
> 460:       HeapAccess<>::oop_store(reinterpret_cast<oop*>(at), nullptr);

Using HeapAccess when `!requires_gc_barriers` is wrong. This would crash with ZGC when/if we fix the flags race and changed `relativize_chunk_concurrently` to only be conditioned `requires_barriers() / _barriers` (and allowing the retry_fast_path "medium" path). 
So either use `*reinterpret_cast<oop*>(at) = nullptr;` or do what my initial suggestion with `clear_lockstack` did, just omit the clearing. Before we requires_barriers(), we are allowed to reuse the stackChuncks, so trying to clean them up seems fruitless.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/21565#discussion_r1825949756
PR Review Comment: https://git.openjdk.org/jdk/pull/21565#discussion_r1825942254