RFR: JDK-8311921: Inform about MaxExpectedDataSegmentSize in case of pthread_create failures on AIX

[Thomas Stuefe]

On Wed, 12 Jul 2023 07:52:06 GMT, Matthias Baesken <mbaesken at openjdk.org> wrote:

> In case pthread_create fails on AIX, MaxExpectedDataSegmentSize might be too small. We have already some UL logging when pthread_create fails, it would be helpful to mention there MaxExpectedDataSegmentSize too.
> 
> We observed the importance of the parameter MaxExpectedDataSegmentSize , when looking into failures of  https://github.com/openjdk/jdk/commits/master/test/hotspot/jtreg/vmTestbase/vm/mlvm/mixed/stress/java/findDeadlock/TestDescription.java  on AIX, this tests generates quite a few threads and might run on some machines into issues because of too low settings of MaxExpectedDataSegmentSize  .

Okay.

Annoying though  that we have to do this kind of thing.

@backwaterred @JoKern65 : Is there a way to prevent the AIX pthread library from allocating thread stacks from the break? To tell it to use mmap, perhaps? That's what the other libc implementations do, and then we don't have this problem.

Cheers, Thomas

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

Marked as reviewed by stuefe (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/14845#pullrequestreview-1526019148