Review Request: 8011161 NMT: Memory leak when encountering out of memory error while initializing memory snapshot
Daniel D. Daugherty
daniel.daugherty at oracle.com
Thu Apr 4 12:14:52 PDT 2013
On 4/4/13 11:42 AM, Zhengyu Gu wrote:
> Updated webrev: http://cr.openjdk.java.net/~zgu/8011161/webrev.01/
> <http://cr.openjdk.java.net/%7Ezgu/8011161/webrev.01/>
src/share/vm/services/memTracker.cpp
Please consider this version:
_snapshot = new (std::nothrow)MemSnapshot();
if (_snapshot != NULL) {
if (!_snapshot->out_of_memory() && start_worker()) {
_state = NMT_started;
NMT_track_callsite = (_tracking_level == NMT_detail &&
can_walk_stack());
return;
}
delete _snapshot;
_snapshot = NULL;
}
// fail to start native memory tracking, shut it down
shutdown(NMT_initialization);
Dan
>
> Thanks,
>
> -Zhengyu
>
>
>
> On 4/3/2013 4:35 PM, Calvin Cheung wrote:
>> Not directly related to this fix, I'm seeing
>> shutdown(NMT_initialization) can be called twice if start_worker()
>> has an error before returning false.
>> 552 shutdown(NMT_initialization);
>>
>> again in MemTracker::start():
>> 144 shutdown(NMT_initialization);
>>
>> Perhaps the one at line 552 isn't necessary?
>>
>> Calvin
>>
>> On 4/3/2013 11:01 AM, Zhengyu Gu wrote:
>>> Resend. Could anyone review it?
>>>
>>> Thanks,
>>>
>>> -Zhengyu
>>>
>>> On 4/2/2013 10:09 AM, Zhengyu Gu wrote:
>>>> Fix a memory leak when initializing memory snapshot due to out of
>>>> memory.
>>>> In normal case, memory snapshot is released by worker thread. But
>>>> this case, the worker thread is never started.
>>>>
>>>>
>>>> bug: http://bugs.sun.com/view_bug.do?bug_id=8011161
>>>> <http://bugs.sun.com/view_bug.do?bug_id=8011161>
>>>> webrev: http://cr.openjdk.java.net/~zgu/8011161/webrev.00/
>>>> <http://cr.openjdk.java.net/%7Ezgu/8011161/webrev.00/>
>>>>
>>>>
>>>> Thanks,
>>>>
>>>> -Zhengyu
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/attachments/20130404/7ad8c987/attachment.html
More information about the hotspot-runtime-dev
mailing list