RFR: 8159890: SIGSEGV with UseStringDeduplication and UseSharedSpaces/RequireSharedSpaces

Derek White derek.white at oracle.com
Mon Jun 27 20:29:10 UTC 2016


On 6/27/16 4:10 AM, Per Liden wrote:
> Thanks for reviewing Stefan/Derek/Thomas/Jiangli/Ioi.
>
> Thanks for the test Ioi, new webrev with test included:
>
> http://cr.openjdk.java.net/~pliden/8159890/webrev.1/
>
> cheers,
> Per
>

Hi Per,

Looks good.

  - Derek
> On 2016-06-22 18:20, Ioi Lam wrote:
>>
>>
>> On 6/22/16 9:11 AM, Thomas Schatzl wrote:
>>> Hi,
>>>
>>> On Wed, 2016-06-22 at 16:49 +0200, Per Liden wrote:
>>>> Hi,
>>>>
>>>> Summary: This was bug was introduced by JDK-8150607 - Clean up
>>>> CompactHashtable, where the rewrite of CompactHashtable::oops_do()
>>>> by
>>>> mistake now iterates over entry[0] instead of entry[1] for regular
>>>> bucket types. This causes garbage (non-narrowOop) values to be passed
>>>> to
>>>> the OopClosure passed to oops_do() and we crash. The bug is easy to
>>>> reproduce with:
>>>>
>>>> $ java -Xshare:dump
>>>> $ java -XX:+UseSharedSpaces -XX:+UseStringDeduplication -version
>>>>
>>>> Webrev: http://cr.openjdk.java.net/~pliden/8159890/webrev.0/
>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8159890
>>
>> Hi Per,
>>
>> Thanks for finding the bug. Sorry I checked code in without realizing
>> that it needed a test case :-(
>>
>>>    looks good. Not really sure if this needs a regression test. It 
>>> seems
>>> to be too trivial and obvious a fix.
>>
>> I wrote one anyway :-) I think it's good to have some minimal coverage
>> in the test suite to cover future issues.
>>
>> http://cr.openjdk.java.net/~iklam/jdk9/8159890_cds_strdedup.v02/SharedStringsDedup.java 
>>
>>
>>
>> It's a slight modification of
>>
>> hotspot/test/runtime/SharedArchiveFile/SharedStringsRunAuto.java
>>
>>
>> Thanks
>> - Ioi
>>
>>
>>> Thanks,
>>>    Thomas
>>



More information about the hotspot-runtime-dev mailing list