RFR: 8263476: NMT: Stack guard pages should not be marked as committed [v3]

David Holmes dholmes at openjdk.org
Tue Aug 5 12:04:03 UTC 2025


On Tue, 5 Aug 2025 09:13:28 GMT, Albert Mingkun Yang <ayang at openjdk.org> wrote:

> Since they will never use any physical memory, using committed-memory for them is unnecessary, 

Windows only allows you to protect committed pages! Linux seems more lax. I don't know about BSD, or macOS or AIX.

> I should update the ticket description to avoid confusions.

Yes please do so. I think changing the guard mechanism in this way needs a bit more obvious "socialising". There may be people who would be interested in this but have ignored it because it appeared to be about NMT labelling! And we need the AIX folk to confirm the changes on their platform.

>  I ran the usual tier1-3 testing.

Normally our VM never gets loaded on the primordial thread so your changes there are untested. We may have one test that loads the VM on the primordial thread, but even then I'm not sure that would adequately test your changes.

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

PR Comment: https://git.openjdk.org/jdk/pull/26571#issuecomment-3154923922


More information about the hotspot-runtime-dev mailing list