RFR 8234270: [REDO] JDK-8204128 NMT might report incorrect numbers for Compiler area
Zhengyu Gu
zgu at redhat.com
Wed Nov 20 13:48:14 UTC 2019
JDK-8204128 did not fix the original bug. But new assertion helped to
catch the problem, as it consistently failed in Oracle internal tests.
The root cause is that, when NMT biases a resource area to compiler, it
did not adjust tracking data to reflect that. When the biased resource
area is released, there is a possibility that its size is greater than
total size recorded, and underflow a size_t counter.
JDK-8204128 patch also missed a long to ssize_t parameter type change,
that resulted new test failure on Windows, because long is 4-bytes on
Windows.
Many thanks to Leonid Mesnik, who helped to run this patch through
Oracle's internal stress tests.
Bug: https://bugs.openjdk.java.net/browse/JDK-8234270
Webrev: http://cr.openjdk.java.net/~zgu/JDK-8234270/webrev.00/index.html
Test:
hotspot_nmt
Submit test
Oracle internal stress tests.
Thanks,
-Zhengyu
More information about the hotspot-runtime-dev
mailing list