RFR: 8053885: test/runtime/memory/RunUnitTestsConcurrently.java fails on solaris with largepage options

Jon Masamitsu jon.masamitsu at oracle.com
Thu Apr 14 17:15:39 UTC 2016



On 04/14/2016 02:00 AM, Stefan Karlsson wrote:
> Hi Jon,
>
> On 2016-04-13 20:40, Jon Masamitsu wrote:
>> Stefan,
>>
>> Is the cause of this problem that there should be better argument 
>> checking on
>> LargePageSizeInBytes?
>
> I think the cause is that I made some incorrect assumptions when I 
> wrote the unit test.
>
> But yes, restricting the LargePageSizeInBytes on solaris would also 
> solve the problem. I could make that change, but it probably requires 
> a CCC approval, since it changes the behavior of a product flag.
>

I understand.  I'll create a CR to improve the argument checking on
LargePageSizeInBytes.  Your change is correct.

Reviewed.

Jon

> StefanK
>
>>
>> Jon
>>
>> On 04/13/2016 02:21 AM, Stefan Karlsson wrote:
>>> Hi all,
>>>
>>> Please review this small patch to fix a unit test problem when 
>>> running tests with -XX:LargePageSizeInBytes=<small page size>.
>>>
>>> http://cr.openjdk.java.net/~stefank/8053885/webrev.01/
>>>
>>> The problem was easily reproducible with the following command line 
>>> on solaris sparc:
>>>  java -XX:LargePageSizeInBytes=8192 -XX:+ExecuteInternalVMTests 
>>> -version
>>>
>>> # Internal Error 
>>> (/opt/jprt/T/P1/215954.amurillo/s/hotspot/src/share/vm/memory/metaspace.cpp:1204), 
>>> pid=5652, tid=5668
>>> # assert(is_size_aligned(vs_byte_size, 
>>> Metaspace::reserve_alignment())) failed: 32768 is not aligned to 
>>> 2097152
>>>
>>> Thanks,
>>> StefanK
>>>
>>>
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20160414/20b42cf7/attachment.html>


More information about the hotspot-gc-dev mailing list