RFR: 8298992: runtime/NMT/SummarySanityCheck.java failed with "Total commi…tted (MMMMMM) did not match the summarized committed (NNNNNN)
Afshin Zafari
azafari at openjdk.org
Wed Aug 23 10:07:10 UTC 2023
On Wed, 23 Aug 2023 08:57:48 GMT, Johan Sjölen <jsjolen at openjdk.org> wrote:
> I don't understand what kind of concurrency problem we're reaching here. If the `ThreadCritical` lock is held during the summation and it attempts to be taken by the deallocations, then surely this would prevent any concurrent tampering of the data?
If we remove the while loop and keep the assert, then in some rare tests in tiers 4 and 7 ( 3-4 cases in 700+ tests) that the assertion raised. I had used other asserts/logs in the code to find the roots of mismatching Total and Sum.
So, theoretically the `ThreadCritical` should keep other threads from alloc/dealloc. But in real, it does not.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/15306#issuecomment-1689672091
More information about the hotspot-runtime-dev
mailing list