[jdk11u-dev] RFR: 8234930: Use MAP_JIT when allocating pages for code cache on macOS [v2]

[Anton Kozlov]

On Wed, 1 Dec 2021 12:01:12 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

> I wish some minimal mac-os specific hack were possible to get MAP_JIT working, instead of downporting this chain of large patches. That would be way easier to review and would limit unintended side effects to just one platform.

I can relate to the last point. If something would crash, let it be the macos/aarch64 or at least */aarch64. :)

However, would not it be useful to have a queue of clean(er) backports, for ease of review as each patch is altered minimally?
And for the long-term maintenance, as the result will be less different from the mainline?
Of course, this depends on the nature of fixes, the queue of changes of formally speaking prerequistes for this patch:

* [JDK-8255254](https://bugs.openjdk.java.net/browse/JDK-8255254): Split os::reserve_memory and os::map_memory_to_file interfaces
* [JDK-8253638](https://bugs.openjdk.java.net/browse/JDK-8253638): Cleanup os::reserve_memory and remove MAP_FIXED
* [JDK-8253639](https://bugs.openjdk.java.net/browse/JDK-8253639): Change os::attempt_reserve_memory_at parameter order
* [JDK-8253650](https://bugs.openjdk.java.net/browse/JDK-8253650): Cleanup: remove alignment_hint parameter from os::reserve_memory

These are code base improvements with fairly low risk. (I have added two more from the description).
But they would help with codebase divergence between 11 and the mainline.

I'd wanted to discuss this approach for this and similar cases. Any comments?

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

PR: https://git.openjdk.java.net/jdk11u-dev/pull/580