RFR: 8338383: Implement JEP 491: Synchronize Virtual Threads without Pinning [v12]

[Dean Long]

On Thu, 31 Oct 2024 16:27:05 GMT, Patricio Chilano Mateo <pchilanomate at openjdk.org> wrote:

>> OK, so you're saying it's the stack adjustment that's the problem.  It sounds like there is code that is using rsp instead of last_Java_sp to compute the frame boundary.  Isn't that a bug that should be fixed?  I also think we should fix the aarch64 c2 stub to just store last_Java_pc like you suggest.  Adjusting the stack like this has in the past caused other problems, in particular making it hard to obtain safe stack traces during asynchronous profiling.
>> 
>> It's still unclear to me exactly how we resume after preemption.  It looks like we resume at last_Java_pc with rsp set based on last_Java_sp, which is why it needs to be adjusted.  If that's the case, an alternative simplification for aarch64 is to set a different last_Java_pc that is preemption-friendly that skips the stack adjustment.  In your example, last_Java_pc would be set to 0xffffdfdba5e4.  I think it is a reasonable requirement that preemption can return to last_Java_pc/last_Java_sp without adjustments.
>
>> OK, so you're saying it's the stack adjustment that's the problem. It sounds like there is code that is using rsp instead of last_Java_sp to compute the frame boundary. Isn't that a bug that should be fixed?
>>
> It's not a bug, it's just that the code from the runtime stub only cares about the actual rsp, not last_Java_sp. We are returning to the pc right after the call so we need to adjust rsp to what the runtime stub expects. Both alternatives will work, either changing the runtime stub to set last pc and not push those two extra words, or your suggestion of just setting the last pc to the instruction after the adjustment. Either way it requires to change the c2 code though which I'm not familiar with. But if you can provide a patch I'm happy to apply it and we can remove this `possibly_adjust_frame()` method.

It turns out if we try to set last pc to the instruction after the adjustment, then we need an oopmap there, and that would require more C2 changes.  Then I thought about restoring SP from FP or last_Java_fp, but I don't think we can rely on either of those being valid after resume from preemption, so I'll try the other alternative.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/21565#discussion_r1825498409