[RFR] (XS) 8087153: EXCEPTION_ACCESS_VIOLATION when CDS RO section vanished on win32

[Chris Plummer]

Ok. I'm not certain of everything that was looked at before yesterday. I 
think there is still a race condition bug if we remap, and I think 
attempts to fix that failed (your protect_memory recollection). Someone 
else should open a new bug for that, because I don't fully understand 
the details. This change just greatly reduces the likelihood of the race 
condition bug being exposed by removing a common need to remap (one that 
was introduced by 8054386).

Chris

On 6/20/15 4:36 AM, David Holmes wrote:
> Chris,
>
> Could you please update the bug report with a conclusion because I 
> can't follow the twists in the plot with this one. We were going to 
> just use protect_memory then that was dropped. I'm not sure what we 
> are doing now.
>
> Thanks,
> David
>
> On 20/06/2015 7:35 AM, Chris Plummer wrote:
>> Please review the following fix for 8087153:
>>
>> http://spiro.us.oracle.com/~cplummer/webrevs/8087153/webrev.00/webrev.hotspot/ 
>>
>>
>> https://bugs.openjdk.java.net/browse/JDK-8087153
>>
>> This bug was introduced by the changes for 8054386, which mapped the
>> archive RW even if RO was requested. The state of the
>> space_info->read_only was not reflecting this, causing an unnecessary
>> attempt to remap to RW later on when a method's bytecodes were
>> rewritten. This remap exposed a race condition on win32.
>>
>> Since the changes are small, they will be committed once they are 
>> approved.
>>
>> thanks,
>>
>> Chris