Segmentation fault with NUMA aware ParallelScavange GC

[Igor Veresov]

On 7/28/11 10:39 PM, Lokesh Gidra wrote:
>
>
>
>>
>> Hi, Lokesh,
>>
>> Could you please check if 6u25 works?
>
>
> You mean kernel version 2.6.25 right?

No, the jdk.

>
> BTW, what kind of bug are you suspecting?
>

The only thing I can think of is that we started using 
madvise(MADV_FREE) to uncommit pages in lgrp chunks instead of doing 
mmap over mmap. I've tested it with earlier kernels and it works fine. 
It could be the semantics of MADV_FREE somehow changed. jdk6u25 has 
still the hotspot that would do mmap. So running it would confirm the 
theory.

Alternatively, if you're open to building hotspot, I can send you a 
patch to try.

igor

>
>
>>
>> Thanks you!
>> igor
>>
>
> Thank you!
> Lokesh
>
>> On 7/28/11 3:36 PM, Lokesh Gidra wrote:
>>> Hi Igor,
>>>
>>> Yes, its a Linux box with 2.6.35-22-server kernel version. Its running a
>>> 64-bit ubuntu server.
>>>
>>> Yes it does reproduce with -XX:-UseLargePages option.
>>>
>>> Please let me know if you need any other information.
>>>
>>> Regards,
>>> Lokesh
>>>
>>>       ------------------------------------------------------------------------
>>>       *From:* Igor Veresov<igor.veresov at oracle.com>
>>>       *To:* hotspot-gc-dev at openjdk.java.net
>>>       *Sent:* Thursday, 28 July 2011 11:14 PM
>>>       *Subject:* Re: Segmentation fault with NUMA aware ParallelScavange GC
>>>
>>>       Hi!
>>>
>>>       This is Linux, right?
>>>       What kernel version do you have?
>>>       Does it reproduce with -XX:-UseLargePages?
>>>
>>>       igor
>>>
>>>       On 7/28/11 2:23 AM, Lokesh Gidra wrote:
>>>        >  Hi,
>>>        >
>>>        >  I am trying to use the NUMA-aware implementation of
>>>       ParallelScavenge GC
>>>        >  of OpenJDK7 on a 8-node machine. Almost everytime I run it, I get a
>>>        >  segmentation fault trying to access a page aligned address. The
>>>       test is
>>>        >  with DaCapo benchmarks which works fine with default setup.
>>>       Although the
>>>        >  address lies in the range of one of the lgrp's but in the memory
>>>       maps,
>>>        >  its not mapped anywhere. Can anyone please explain me what could
>>>       be the
>>>        >  problem. Please let me know if some additional information is needed.
>>>        >
>>>        >
>>>        >  Best Regards,
>>>        >  Lokesh
>>>
>>>
>>>
>>
>>
>>
>>