RFR: 8213713: Minor issues during MetaspaceShared::initialize_runtime_shared_and_meta_spaces
Jiangli Zhou
jiangli.zhou at oracle.com
Wed Nov 14 17:58:46 UTC 2018
Thanks, Ioi.
Jiangli
On 11/13/18 11:34 PM, Ioi Lam wrote:
> Hi Jiangli,
>
> The changes looks good.
>
> Thanks
>
> - Ioi
>
>
> On 11/13/18 5:56 PM, Jiangli Zhou wrote:
>> Please review the following change for fixing some non-fatal issues
>> during MetaspaceShared::initialize_runtime_shared_and_meta_spaces
>> process.
>>
>> webrev: http://cr.openjdk.java.net/~jiangli/8213713/webrev.00/
>> bug: https://bugs.openjdk.java.net/browse/JDK-8213713
>>
>> Issue 1)
>> MetaspaceShared::core_spaces_size() is not populated early enough
>> during runtime. As a result, the beginning (instead of the end)
>> address of the mapped shared spaces is used as the requested address
>> for reserving the compressed class space. So the first
>> ReservedSpace() always fails at runtime and we retry with a new
>> address that's 1G higher. That adds unwanted overhead and also causes
>> a gap between the mapped shared spaces and the compressed class space.
>>
>> Issue 2)
>> FileMapInfo::validate_shared_path_table returns true even when
>> shared_path(i)->validate() fails.
>>
>> Issue 3)
>> Dead code in FileMapInfo::validate_shared_path_table. The 'else if
>> (!PrintSharedArchiveAndExit)' is never taken and should be removed.
>>
>> As part of the change, I've also added some more asserts.
>>
>> Tested with appcds tests locally on linux-x64. Running tier1-tier3.
>>
>> Thanks,
>>
>> Jiangli
>>
>
More information about the hotspot-runtime-dev
mailing list