Request for approval: 6929067: Stack guard pages should be removed when thread is detached

[Coleen Phillimore - Sun Microsystems]

Andrew, Since I missed the contributed by in the last putback, this one 
will look like this:

6936168: Recent fix for unmapping stack guard pages doesn't close 
/proc/self/maps
Summary: Add close to returns (fix for 6929067 also contributed by aph)
Contributed-by: aph at redhat.com
Reviewed-by: coleenp, aph <other reviewers???>

It would be nice if we had one more reviewer....

Thanks,
Coleen

On 03/18/10 13:46, Andrew Haley wrote:
> Yes, thanks indeed.
>
> Andrew.
>
>
> On 03/18/2010 05:35 PM, Coleen Phillimore - Sun Microsystems wrote:
>   
>> I filed bug  6936168 and have a repository with 3 additional fcloses()
>> in them for the returns after the file is opened.  See:
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/6936168/
>> bug link at http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6936168
>>
>> It's running our tests that failed now.  Please review this change and
>> I'll push it back before it gets to the main hotspot repository (it's
>> still in hotspot_rt right now).
>>
>> Thanks for reporting this.  This would have taken us a while to
>> diagnose.  Sorry for the crummy code review the first time around.
>>
>> Coleen
>>
>> On 03/18/10 06:05, Andrew Haley wrote:
>>     
>>> On 03/18/2010 09:10 AM, Andreas Kohn wrote:
>>>  
>>>       
>>>> On Fri, 2010-03-12 at 09:44 +0000, Andrew Haley wrote:
>>>>    
>>>>         
>>>>> On 03/11/2010 09:06 PM, Coleen Phillimore wrote:
>>>>>      
>>>>>           
>>>>>> I've added the test to the changeset and a script to run in our
>>>>>> harness.
>>>>>>
>>>>>> Also in os_linux.cpp, I changed the SYS_gettid call to go through our
>>>>>> os::Linux::gettid() because on at least one linux, syscall() returns a
>>>>>> long int which gets a compilation warning with %d.
>>>>>>
>>>>>> open webrev at http://cr.openjdk.java.net/~coleenp/6929067/
>>>>>> bug link at http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6929067
>>>>>>
>>>>>> Andrew, please have a look since you're the contributor.
>>>>>>         
>>>>>>             
>>>>> That's OK, but you don't need SYS_gettid.
>>>>> Please look at
>>>>> http://cr.openjdk.java.net/~aph/6929067-jdk7-webrev-4/hotspot.patch
>>>>> I changed to "/proc/self/maps", as you requested.  I think this is
>>>>> better.
>>>>>
>>>>> The copy of my webrev to cr.openjdk.java.net failed for some reason
>>>>> I don't understand.
>>>>>       
>>>>>           
>>>> With this change I seem to hit the limit on the number of open files.
>>>> Looking through it, shouldn't get_stack_bounds() close the FILE* it
>>>> opened?
>>>>     
>>>>         
>>> Oh, how stupid of me!  If this were gcc I'd just push a fix
>>> immediately as obvious/trivial, but I think we need a bug opened to
>>> push the change.
>>>
>>> (BTW, this happened because of a mistake translating the patch I wrote
>>> from using the C++ library into C.  The original patch used an
>>> fstream, whose destructor closes the file.  When I did the translation
>>> I missed the fact that I had to close the file manually.)
>>>
>>> Andrew.
>>>   
>>>       
>
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-dev/attachments/20100318/37730dce/attachment.html