RFR: 8313306: More sensible memory reservation logging [v15]

[Thomas Stuefe]

On Thu, 14 Mar 2024 20:08:06 GMT, Sonia Zaldana Calles <szaldana at openjdk.org> wrote:

>> This PR implements more specialized logs for virtual memory APIs in the “os” namespace. It uses “os” and “map” as log tags using unified JVM logging as introduced in JEP 158 (https://openjdk.org/jeps/158). 
>> 
>> As far as testing is concerned, I have added a regression test to verify the logging is working accordingly.
>
> Sonia Zaldana Calles has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Disabling tests on AIX due to VMA accounting and exchanging some test logic set up to trigger more subtle failures

Hi Sonia,

- testReserveFailed

I would either crank the reservation size up significantly or scratch this all together. Reservations of any size really are perfectly fine if tests run without virtual size limit. There is a limit somewhere dictated by the address space size and fragmentation, usually around 128TB or somewhere in the PB range.

So I would increase the size to > 100PB or so, but then be prepared for follow-up problems on some platforms. Either that, or outright scratch it. You already test reservation failures for reserve-at, that is sufficient.

- testCommitFailed
- testUncommitFailed
- testReleaseFailed

All of these I would only run in release VMs, where we will not assert anything like page-size granularity (neither now nor in the future).

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

Changes requested by stuefe (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/17356#pullrequestreview-1938181426