[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-runtime-dev
Subject:    Re: How are stacks of non-JavaThreads guarded?
From:       Thomas_Stüfe <thomas.stuefe () gmail ! com>
Date:       2015-11-30 13:09:04
Message-ID: CAA-vtUyPF2z9JLuhxA1-MXTPk3cgq=-mhHTNW3mDWkR+Gi0qEw () mail ! gmail ! com
[Download RAW message or body]

Hi David,



On Mon, Nov 30, 2015 at 12:54 PM, David Holmes <david.holmes@oracle.com>
wrote:

> On 30/11/2015 5:01 PM, Thomas St=C3=BCfe wrote:
>
>> Hi Andrew,
>>
>> On Sat, Nov 28, 2015 at 10:57 AM, Andrew Haley <aph@redhat.com> wrote:
>>
>> On 28/11/15 07:55, Thomas St=C3=BCfe wrote:
>>>
>>>> So, in JavaThread we establish stack guards (red and yellow zones). Bu=
t
>>>> I
>>>> do not find anything similar for non JavaThread threads, e.g. VmThread=
?
>>>>
>>>
>>> There isn't, and I don't think it's possible because there is no
>>> reasonably portable way to recover from a stack overflow in C++.
>>>
>>>
>>> thanks for clarifying.
>>
>> But I don't even want to recover; a hard-and-fast crash is better than
>> running over whatever happens to be below the VmThread stack. Well,
>> something to mull over.
>>
>
> You will run into the libc guard pages as I understand it and get your
> hard-and-fast crash.
>
>
On AIX I disable the OS guard pages. That was the reason for my original
question. I guess I have
to revert this and analyze the original problem I tried to solve back when
I disabled the guard pages.

Kind Regards, Thomas



> The Java thread guard pages are for recovery purposes.
>
> David
> -----
>
> Regards, Thomas
>>
>>
>>
>> Andrew.
>>>
>>>
>>>
[prev in list] [next in list] [prev in thread] [next in thread]