RFR: 8267155: runtime/os/TestTracePageSizes times out [v4]

[Thomas Stuefe]

On Tue, 18 May 2021 12:32:23 GMT, Stefan Johansson <sjohanss at openjdk.org> wrote:

>> Thomas Stuefe has updated the pull request incrementally with two additional commits since the last revision:
>> 
>>  - Update test/hotspot/jtreg/runtime/os/TestTracePageSizes.java
>>    
>>    Co-authored-by: Aleksey Shipilëv <shade at redhat.com>
>>  - Update test/hotspot/jtreg/runtime/os/TestTracePageSizes.java
>>    
>>    oops
>>    
>>    Co-authored-by: Aleksey Shipilëv <shade at redhat.com>
>
> Did a quick test now with THP configured to `always` but running with `-XX:-UseLargePages -XX:-UseTransparentHugePages`. It gives me:
> 
> 80000000-ffe00000 rw-p 00000000 00:00 0 
> Size:            2095104 kB
> KernelPageSize:        4 kB
> MMUPageSize:           4 kB
> ...
> AnonHugePages:   2095104 kB
> THPeligible:            1
> VmFlags: rd wr mr mw me ac sd 
> 
> So no `hg` flag as expected.

Hi @kstefanj , @shipilev ,

I changed the test so that:
- we skip out if kernel < 3.8. This means we can rely on a sane set of tags from the smaps file.
- I skipped also for ppc because I found that the semantics of some tags in smaps differ and I honestly never used hugetlbfs on ppc; if it becomes an issue, we can spend more time on this.
- I reverted some of my earlier changes; now we again require all information to be there (so no null handling for missing smaps information in `RangeWithPageSize`.
- added some comments

Note that even though we skip out of the test we still at least start the VMs and go through VM initialization. That is totally fine by me since this is a good test that the VM survives these variations of large page flags (eg we found unrelated crashes on ppc with https://github.com/openjdk/jdk/pull/1153 that way).

Cheers, Thomas

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

PR: https://git.openjdk.java.net/jdk/pull/4064