Mailing-List: contact dev-help@harmony.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@harmony.apache.org
Received-SPF: pass (nike.apache.org: domain of ilya.berezhniuk@gmail.com
 designates 74.125.82.43 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=mime-version:in-reply-to:references:from:date:message-id:subject:to
         :content-type:content-transfer-encoding;
        b=qpuO9vvK8XvnjRBiHA3iOy6hC/yRBqwpcwT+2PaEt/AzH6pFuCUD4DBKxmHQbfu/sq
         YuaQvE2aAhgDXYq/e4fDZWAYpb9xOBRk8JVJnKhYen/6kDSYwFdkbuHMMUraG/TRUWZO
         g9e3saPQZzfhgBhwnWU6Nn1HriCg/+S+Ui0SI=
MIME-Version: 1.0
In-Reply-To: <BANLkTimsis4oLifRM38934UnYEVpTcJoGw@mail.gmail.com>
References: <3B18D33A-7F6D-4D2E-8A47-2D37B7561185@zenika.com>
 <BANLkTimsis4oLifRM38934UnYEVpTcJoGw@mail.gmail.com>
From: Ilya Berezhniuk <ilya.berezhniuk@gmail.com>
Date: Mon, 25 Apr 2011 19:43:05 -0700
Message-ID: <BANLkTikv_sMtu-aO-=8DHAXSLz+RiFjAPg@mail.gmail.com>
Subject: Re: [drlvm][thread] stacksize
To: dev@harmony.apache.org
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Hi Pierre,

1) The stack size you see is most probably a main thread stack size,
obtained once the global instance of "port shared data" object is
created, while attaching main thread to the port threading system.
You cannot change main thread stack size via Xss, it's set up in
compile time I believe.

To get size of all managed threads created by the VM, you can examine
results returned by the call to find_stack_addr_size() in init_stack()
function (vm/port/src/thread/linux/thread_os.c:395).

2) The meaning of the test is to check if our current stack position
(obtained through getting an address of local variable) with a fixed
spare space (mem_protect_size) overlaps with an area which is to be
protected by set_guard_page() later in this function. If they do
overlap, then we have way too small stack to work properly, even less
than 0x400 (0x100 on 32-bit) bytes between current stack and special
area supposed for catching stack overflow.

Thanks,
Ilya.


On Thu, Apr 21, 2011 at 1:45 PM, Alexei Fedotov
<alexei.fedotov@gmail.com> wrote:
> Pierre,
>
> Please try the following searches:
>
> The parameter is set:
> http://www.google.com/codesearch?q=3D"thread.stacksize"&exact_package=3Dh=
ttp%3A%2F%2Fsvn.apache.org%2Frepos%2Fasf%2Fharmony%2Fenhanced%2F<http://www=
.google.com/codesearch?q=3D%22thread.stacksize%22&exact_package=3Dhttp%3A%2=
F%2Fsvn.apache.org%2Frepos%2Fasf%2Fharmony%2Fenhanced%2F>
>
> The thread is created:
> http://www.google.com/codesearch?lr=3D&q=3D"port_thread_create"&sbtn=3D=
=D0=9F=D0=BE=D0=B8=D1=81=D0=BA&exact_package=3Dhttp%3A%2F%2Fsvn.apache.org%=
2Frepos%2Fasf%2Fharmony%2Fenhanced%2F<http://www.google.com/codesearch?lr=
=3D&q=3D%22port_thread_create%22&sbtn=3D%D0%9F%D0%BE%D0%B8%D1%81%D0%BA&exac=
t_package=3Dhttp%3A%2F%2Fsvn.apache.org%2Frepos%2Fasf%2Fharmony%2Fenhanced%=
2F>
>
> Maybe if you add more debugging there, this will help.
>
> --
> With best regards / =D1=81 =D0=BD=D0=B0=D0=B8=D0=BB=D1=83=D1=87=D1=88=D0=
=B8=D0=BC=D0=B8 =D0=BF=D0=BE=D0=B6=D0=B5=D0=BB=D0=B0=D0=BD=D0=B8=D1=8F=D0=
=BC=D0=B8,
> Alexei Fedotov / =D0=90=D0=BB=D0=B5=D0=BA=D1=81=D0=B5=D0=B9 =D0=A4=D0=B5=
=D0=B4=D0=BE=D1=82=D0=BE=D0=B2,
> http://dataved.ru/
> +7 916 562 8095
>
>
>
> On Fri, Apr 22, 2011 at 12:13 AM, Pierre Queinnec <
> pierre.queinnec@zenika.com> wrote:
>
>> Hi guys,
>>
>> I'm seeing some strange thread stack size behavior while porting to Mac,
>> and I've been trying to compare to a build on a 64bit linux (Debian 6).
>>
>> Here's what I see on Linux, whatever I can define for my Xss, the stack
>> size in init_psd_structure always calculates 8M :
>>
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>> queinnec@squeezy:~/Code/harmony-svn-java6/target/hdk/jdk$ ./bin/java
>> -Xmx1024m -Xss12m -version
>> init_psd_structure: stack_size =3D 8388608
>> init_psd_structure: guard_size =3D 4096
>> init_psd_structure: mem_protect_size =3D 1024
>> PMQ - setup_stack - (&res - mem_protect_size) =3D 140736255271036
>> PMQ - setup_stack - (guard_page_addr + guard_page_size =3D 1407362469601=
28
>> PMQ - setup_stack - (&res - mem_protect_size) =3D 140715543796204
>> PMQ - setup_stack - (guard_page_addr + guard_page_size =3D 1407155417784=
32
>> PMQ - setup_stack - (&res - mem_protect_size) =3D 140715528108524
>> PMQ - setup_stack - (guard_page_addr + guard_page_size =3D 1407155156049=
92
>> PMQ - setup_stack - (&res - mem_protect_size) =3D 140715515521516
>> PMQ - setup_stack - (guard_page_addr + guard_page_size =3D 1407155135037=
44
>> PMQ - setup_stack - (&res - mem_protect_size) =3D 140715513420268
>> PMQ - setup_stack - (guard_page_addr + guard_page_size =3D 1407155114024=
96
>> Apache Harmony Launcher : (c) Copyright 1991, 2010 The Apache Software
>> Foundation or its licensors, as applicable.
>> java version "1.5.0"
>> Apache Harmony (1.5.0)
>> DRLVM (1.5.0-r1092753)
>> pre-alpha : not complete or compatible
>> svn =3D r1092753, (Apr 15 2011), Linux/em64t/gcc 4.4.5, release build
>> http://harmony.apache.org
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>>
>> (patch attached, maybe I just missed something obvious... I've seen some
>> maybe related JIRA but they were supposed to be closed)
>>
>> As a second question, in drlvm/vm/port/src/thread/linux/thread_os.c ther=
e's
>> this test :
>>
>> =C2=A0 =C2=A0 =C2=A0if ((size_t)(&res) - PSD->mem_protect_size
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0< (size_t)tlsdata->guard=
_page_addr + tlsdata->guard_page_size)
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0return EINVAL;
>>
>> I'm not sure I get this... Anyone care to explain?
>>
>> As a side note, the port seems to start fine but crashes after starting
>> some threads.
>> Cheers,
>> -- Pierre Queinnec
>>
>>
>>
>>
>>
>>
>