RFR: JDK-8313781: Add regression tests for large page logging and user-facing error messages [v3]

Thomas Stuefe stuefe at openjdk.org
Tue Aug 15 15:16:11 UTC 2023


On Tue, 15 Aug 2023 12:29:59 GMT, Stefan Johansson <sjohanss at openjdk.org> wrote:

>> Thomas Stuefe has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains six additional commits since the last revision:
>> 
>>  - Merge branch 'JDK-8313781-Add-regression-tests-for-large-page-logging-and-user-facing-error-messages' of github.com:tstuefe/jdk into JDK-8313781-Add-regression-tests-for-large-page-logging-and-user-facing-error-messages
>>  - skip empties
>>    
>>    Co-authored-by: Stefan Johansson <54407259+kstefanj at users.noreply.github.com>
>>  - Remove THPmode unknown
>>  - Several test fixes
>>  - erge branch 'master' into JDK-8313781-Add-regression-tests-for-large-page-logging-and-user-facing-error-messages
>>  - JDK-8313781-Add-regression-tests-for-large-page-logging-and-user-facing-error-messages
>
> Looks good, just one minor thing. 
> 
> I saw the comment about `LargePageSizeInBytes` and thought that it should be required away but realized that `ProcessTools.createJavaProcessBuilder()` doesn't add the passed in options.
> 
> Also did some manual runs on an aarch64 system where the default 512m pages are not configured but the 2m pages are. This seem to generate output which will be accepted by the test.

Thanks @kstefanj!

Since we just changed the error warning texts (https://github.com/openjdk/jdk/pull/15158), I needed to change the error text somewhat.

I then retested on x64 with the following configurations:

- No Hugepages configured (but supported by kernel) and:
  - THP "never"
  - THP "always"
  - THP "advise"

- THP "never" and 2M default hugepage size and:
  - nonzero Hugepages enabled (2M)
  - nonzero Hugepages enabled (1G)
  - nonzero Hugepages enabled (2M, 1G)

- THP "never" and 1G default hugepage size (via kernel parameter) and nonzero 1G Hugepages
 ( this uncovered a small test bug where "buildSizeString()" had been wrong)

- Hugepages completely absent on a kernel built without hugepages (aarch64 SBC)
	(this uncovered another test bug where HugePageConfiguration would fall back to THP mode "unknown" which then confused the test. I removed that enum value.)

All issues have been fixed.

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

PR Comment: https://git.openjdk.org/jdk/pull/15159#issuecomment-1679106346


More information about the hotspot-runtime-dev mailing list