[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-runtime-dev
Subject:    webrev review request for 6880029
From:       Thomas.Rodriguez () Sun ! COM (Tom Rodriguez)
Date:       2009-09-23 22:00:24
Message-ID: 898CC4EF-7DC0-412A-8C39-63DC76F92855 () sun ! com
[Download RAW message or body]

Doesn't that introduce a race in the initialization?   
inc_page_trap_count is called from the safepointing thread so multiple  
threads could perform the initialization.  Why can't being_statistics  
be moved up above the guarding of the polling page?  BTW, you used the  
wrong bug id in the comment.


tom

On Sep 23, 2009, at 2:35 PM, Xiaobin Lu wrote:

> Webrev: http://javaweb.sfbay/~xl116366/webrev/6880029/
>
> Details:
>
> The bug was uncovered by CBOE when their application crashed pretty  
> early during startup when -XX:+PrintSafepointStatistics was  
> specified. The root cause is that the polling page exception used by  
> JIT could occur without safepoint being involved and the array which  
> is used to record the page trap count hasn't been initialized at  
> that point. The fix is to do initialization at the point where we  
> try to record the page trap count.
>
> Verified by: CBOE
>
> Thanks,
>
> -Xiaobin


[prev in list] [next in list] [prev in thread] [next in thread]