RFR: 8170355 - [BACKOUT] 8k class metaspace chunks misallocated from 4k chunk freelist

[Thomas Stüfe]

Hi Stefan,



On Fri, Nov 25, 2016 at 4:58 PM, Stefan Karlsson <stefan.karlsson at oracle.com
> wrote:

> Hi Jesper,
>
> On 2016-11-25 16:18, Jesper Wilhelmsson wrote:
>
>> Hi,
>>
>> Please review this backout of JDK-8169931 that is causing problems in
>> the nightlies that has proven difficult to fix.
>>
>
> The assert introduced in JDK-8169931 found a similar bug but this time in
> the metaspace initialization code. Using command line flags, there are ways
> to request a non-humongous chunk size (smaller than the Medium chunk size)
> that doesn't match any of the fixed chunk sizes.
>
>
I am currently working on a metaspace change and use your non-static
ChunkManager->list_index() a lot. I see that this backout has not yet found
its way into the repo, but that will happen shortly I guess.

The change looked reasonable, I am curious which issues exactly were you
guys encountered in the nightly tests?

Thanks, Thomas


>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8170355
>> Webrev: http://cr.openjdk.java.net/~jwilhelm/8170355/webrev.00/
>>
>
> The backout looks good.
>
> Thanks,
> StefanK
>
>
>> Thanks,
>> /Jesper
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20161214/b2785621/attachment.htm>