xlpCodeCacheTests_1 failing on AIX XL

[pshipton]

https://ci.eclipse.org/openj9/job/Test_openjdk8_j9_extended.functional_ppc64_aix_xl_Personal/1
xlpCodeCacheTests_1

xlpCodeCacheTests_1 Start Time: Wed Nov  4 14:26:11 2020 Epoch Time (ms): 1604517971032
variation: Mode110
JVM_OPTIONS:  -Xjit -Xgcpolicy:gencon -Xnocompressedrefs 
test suite: read tests from "/home/jenkins/workspace/Test_openjdk8_j9_extended.functional_ppc64_aix_xl_Personal_testList_1/openjdk-tests/TKG/../../jvmtest/functional/VM_Test/VM_Test.jar"
+++ j9vm.test.xlpcodecache.XlpCodeCacheOptionsTest: +++
command: /home/jenkins/workspace/Test_openjdk8_j9_extended.functional_ppc64_aix_xl_Personal_testList_1/openjdkbinary/j2sdk-image/bin/java  -Xjit -Xgcpolicy:gencon -Xnocompressedrefs  -Xdump -Xms64m -Xmx64m -verbose:sizes    -classpath /home/jenkins/workspace/Test_openjdk8_j9_extended.functional_ppc64_aix_xl_Personal_testList_1/openjdk-tests/TKG/../../jvmtest/functional/VM_Test/VM_Test.jar  j9vm.test.xlpcodecache.XlpCodeCacheOptionsTest 

  -Xmca32K        RAM class segment increment
  -Xmco128K       ROM class segment increment
  -Xmcrs0K        compressed references metadata initial size
  -Xmns16M        initial new space size
  -Xmnx16M        maximum new space size
  -Xms64M         initial memory size
  -Xmos48M        initial old space size
  -Xmox48M        maximum old space size
  -Xmx64M         memory maximum
  -Xmr16K         remembered set size
  -Xlp:objectheap:pagesize=64K	 large page size
                  available large page sizes:
                  4K
                  64K
                  16M
  -Xlp:codecache:pagesize=64K	 large page size for JIT code cache
                  available large page sizes for JIT code cache:
                  4K
                  16M
  -Xmso256K       operating system thread stack size
  -Xiss2K         java thread stack initial size
  -Xssi16K        java thread stack increment
  -Xss1M          java thread stack maximum size
  -XX:SharedCacheHardLimit=300M shared class cache size
  -Xscmx64M       shared class cache soft max size
  -Xscdmx24544K   reserved shared class cache space for class debug attributes
  -Xscminaot0K    min reserved shared class cache space for AOT
  -Xscmaxaot282300K max allowed shared class cache space for AOT
  -Xscminjitdata0K min reserved shared class cache space for JIT data
  -Xscmaxjitdata282300K max allowed shared class cache space for JIT data
A message from XlpCodeCacheOptionsTest class
ERROR: Without -Xlp JIT should use default page size and type for executable pages

	 Default page size and type for executable pages: 4096 not used
	 Page size and type used by JIT for executable pages: 65536 not used
*** Test FAILED *** (j9vm.test.xlpcodecache.XlpCodeCacheOptionsTest)

Note this only fails on XL because it only runs on XL due to the test modes. FYI #11105

[pshipton]

@llxia I think this test was recently added? I expect it's incorrect somehow and should recognize 64K as the default for AIX. It's failing on IBM Java 8 as well (32-bit and 64-bit AIX) from Oct 29th (no previous history).

Also, let's run it in the default mode instead of 109, 110. There are likely many tests that need a mode change, but I can open another issue.

[pshipton]

I didn't see it at first, actually I think this is a bug in some JIT/port/verbose code since it's not listing 64K as a possible size although it's the default.

  -Xlp:codecache:pagesize=64K	 large page size for JIT code cache
                  available large page sizes for JIT code cache:
                  4K
                  16M

[llxia]

Just for the record, this test was ported recently (PR #10980)

xlpCodeCacheTests_0 passed with Mode109 -Xgcpolicy:gencon -Xint -Xnocompressedrefs
https://ci.eclipse.org/openj9/job/Test_openjdk8_j9_extended.functional_ppc64_aix_xl_Personal_testList_0/1/

[tajila]

@gita-omr can you please take a look at this?

see #11104 (comment)

[gita-omr]

If the test case was just added and seems to expose old issues which we are planning to rework anyway, maybe it should not be a blocker?

@AlenBadel fyi

[AlenBadel]

If the test case was just added and seems to expose old issues which we are planning to rework anyway, maybe it should not be a blocker?

I believe this is an issue I've already addressed on my re-work branch. This does not impact performance or functionality at all, rather it's just a bug in the verbose log which the test uses to verify the correct page configuration. In any matter, I'll be more than happy to create a PR to fix this verbose bug.

[gita-omr]

Sorry for sounding trivial but every change takes time and we don't know how it is going to affect other Xlp tests. I would rather us finish our other blockers.

[pshipton]

Test will be excluded via #11460.
Moving out to the next milestone.

[zl-wang]

with Alen leaving IBM, we will look into this in the next round.

[zl-wang]

AIX default pagesize is 4KB (whereas Linux default pagesize is 64KB). How does the test case expect AIX codecache default pagesize to be 64KB? through other setting(s)?

[zl-wang]

there are linker options to set these "default" pagesize(s) (on AIX only). i didn't see they are used in the makefile though.

[zl-wang]

as far as i can see, this is a preference testing. i.e. you can say 64KB page is the preference when 16MB page support is not available. let Amar chase this down ...

[dmitripivkine]

AIX default pagesize is 4KB (whereas Linux default pagesize is 64KB). How does the test case expect AIX codecache default pagesize to be 64KB? through other setting(s)?

Is it set to 64k here?

openj9/runtime/compiler/control/J9Options.cpp

Line 1821 in fc324bb

preferredPageSize = 65536;

[zl-wang]

yeah, it makes sense from -Xlp preference perspective (when 16MB is not available).