RFR 8234270: [REDO] JDK-8204128 NMT might report incorrect numbers for Compiler area

Zhengyu Gu zgu at redhat.com
Tue Nov 26 14:29:40 UTC 2019


>>
>> But I leave it up tp you if you change this. If you just reorder the 
>> calls, I do not need another Webrev,
> 
> I will re-order the calls, as you suggested.

Passed submit tests and pushed.

Thanks, Thomas and Yumin

-Zhengyu

> 
> Thanks,
> 
> -Zhengyu
> 
>>
>> ...
>>
>> Another unrelated question, what is the reason for the unusual 
>> creation of MallocMemorySummarySnapshot with placement new? Why not 
>> just put it as a member into MallocMemorySummary? I must be missing 
>> something.
>>
>>
>>     -Zhengyu
>>
>>
>>
>>      >
>>      >    66     wb.NMTFreeArena(arena1);
>>      >
>>      > On Mon, Nov 25, 2019 at 2:30 PM Zhengyu Gu <zgu at redhat.com
>>     <mailto:zgu at redhat.com>
>>      > <mailto:zgu at redhat.com <mailto:zgu at redhat.com>>> wrote:
>>      >
>>      >     Ping ... May I get a second review?
>>      >
>>      >     Thanks,
>>      >
>>      >     -Zhengyu
>>      >
>>      >     On 11/21/19 12:12 PM, yumin qi wrote:
>>      >      > Hi, Zhengyu
>>      >      >
>>      >      >    The fix looks good to me.
>>      >      >
>>      >      > Thanks
>>      >      > Yumin
>>      >      >
>>      >      >
>>      >      >
>>      >      > On Wed, Nov 20, 2019 at 5:49 AM Zhengyu Gu <zgu at redhat.com
>>     <mailto:zgu at redhat.com>
>>      >     <mailto:zgu at redhat.com <mailto:zgu at redhat.com>>
>>      >      > <mailto:zgu at redhat.com <mailto:zgu at redhat.com>
>>     <mailto:zgu at redhat.com <mailto:zgu at redhat.com>>>> wrote:
>>      >      >
>>      >      >     JDK-8204128 did not fix the original bug. But new
>>     assertion
>>      >     helped to
>>      >      >     catch the problem, as it consistently failed in Oracle
>>      >     internal tests.
>>      >      >
>>      >      >     The root cause is that, when NMT biases a resource 
>> area to
>>      >     compiler, it
>>      >      >     did not adjust tracking data to reflect that. When the
>>     biased
>>      >     resource
>>      >      >     area is released, there is a possibility that its 
>> size is
>>      >     greater than
>>      >      >     total size recorded, and underflow a size_t counter.
>>      >      >
>>      >      >     JDK-8204128 patch also missed a long to ssize_t 
>> parameter
>>      >     type change,
>>      >      >     that resulted new test failure on Windows, because 
>> long is
>>      >     4-bytes on
>>      >      >     Windows.
>>      >      >
>>      >      >     Many thanks to Leonid Mesnik, who helped to run this 
>> patch
>>      >     through
>>      >      >     Oracle's internal stress tests.
>>      >      >
>>      >      >     Bug: https://bugs.openjdk.java.net/browse/JDK-8234270
>>      >      >     Webrev:
>>      > http://cr.openjdk.java.net/~zgu/JDK-8234270/webrev.00/index.html
>>      >      >
>>      >      >
>>      >      >     Test:
>>      >      >         hotspot_nmt
>>      >      >         Submit test
>>      >      >         Oracle internal stress tests.
>>      >      >
>>      >      >
>>      >      >     Thanks,
>>      >      >
>>      >      >     -Zhengyu
>>      >      >
>>      >
>>



More information about the hotspot-runtime-dev mailing list