Re: [microblaze-uclinux] PPPD problem with fork()

On Fri, Mar 13, 2009 at 9:58 AM, John Williams <john.williams@xxxxxxxxxxxxx> wrote:

Hi Wendy,

Can you please test the attached program on an petalinux MMU system
build from current SVN trunk? Any HW project is OK, s3aDSP or
whatever is easiest for you.

Carsten finds some error in fork() and pointer handling, with
different CPU optimisations. He is using original MMU test release,
not SVN head, so maybe it is already fixed.

Thanks,

John

---------- Forwarded message ----------
From: Bartsch Carsten <cbartsch@xxxxxxxxxx>
Date: Fri, Mar 13, 2009 at 1:29 AM
Subject: RE: [microblaze-uclinux] PPPD problem with fork()
To: microblaze-uclinux@xxxxxxxxxxxxxx

Attached you will find the test.
Make it with the default Makefile generated by petalinux-new-app.
Btw: I'm using a petalinux-mmu-v0.10 on a Sp3A-DSP-1800A-Starter.
It was only tested on my platform, not on a reference platform.

This following text is the output after a normal make:

# /ua/process-test/process-test
the main program process ID is 104
before forking: textptr=bf907e00 &textptr=bf907df8 *textptr=I am a
stupid text.
before forking: numpointer=bf907df4 &numptr=bf907dfc *numptr=9876
before forking: number=9876 &number=bf907df4

this is the parent process, with id 104 the child's process ID is 105
parent : textptr=bf907e00 &textptr=bf907df8 *textptr=I am a
stupid text.
parent : numptr=bf907df4 &numptr=bf907dfc *numptr=9876
parent : number=9876 &number=bf907df4

this is the child process, with id 105
child : textptr=bf907e00 &textptr=2 *textptr=I am a stupid text.
child : numptr=bf907df4 &numptr=0 *numptr=9876
child : number=9876 &number=0

The following text is the output after removing the -Os option:

# /ua/process-test/process-test
the main program process ID is 106
before forking: textptr=bfa67e04 &textptr=bfa67e1c *textptr=I am a
stupid text.
before forking: zahlpointer=bfa67e18 &numptr=bfa67e20 *numptr=9876
before forking: number=9876 &number=bfa67e18

this is the parent process, with id 106 the child's process ID is 107
parent : textptr=bfa67e04 &textptr=bfa67e1c *textptr=I am a
stupid text.
parent : numptr=bfa67e18 &numptr=bfa67e20 *numptr=9876
parent : number=9876 &number=bfa67e18

this is the child process, with id 107
child : textptr=bfa67e04 &textptr=bfa67e1c *textptr=I am a
stupid text.
child : numptr=bfa67e18 &numptr=bfa67e20 *numptr=9876
child : number=9876 &number=bfa67e18

The difference is the addresses. In the child process the content is
there but the addresses are lost.

-----Mensaje original-----
De: owner-microblaze-uclinux@xxxxxxxxxxxxxx
[mailto:owner-microblaze-uclinux@xxxxxxxxxxxxxx] En nombre de John
Williams
Enviado el: jueves, 12 de marzo de 2009 12:06
Para: microblaze-uclinux@xxxxxxxxxxxxxx
Asunto: Re: [microblaze-uclinux] PPPD problem with fork()

Hi Carsten,

On Wed, Mar 4, 2009 at 1:07 AM, Bartsch Carsten <cbartsch@xxxxxxxxxx>
wrote:
> with a small test-application I see the same problem (using MMU and
fork()).
>
> My first impression is that this happens if you want to share only a
> pointer. You could try sharing other "padding"- variables (use them in
the
> child). Some addresses look strange in the childs process. I didn't
dig
> deeper yet.

Can you please post this test app? The smaller the better :)

Thanks,

John
___________________________
microblaze-uclinux mailing list
microblaze-uclinux@xxxxxxxxxxxxxx
Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
Mailing List Archive :
http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/

--
John Williams, PhD, B.Eng, B.IT
PetaLogix - Linux Solutions for a Reconfigurable World
w: www.petalogix.com p: +61-7-30090663 f: +61-7-30090663