[RFC] Only a.out QMAGIC format is working

[RFC] Only a.out QMAGIC format is working

[Geyslan Gregório Bem]

Sirs,

I was researching about old binary formats and did some tests.
Meantime, I was able to run sucessfully only the QMAGIC format.
Nonetheless, the OMAGIC, NMAGIC and ZMAGIC didn't work anymore.

The test occurred using old slackware binaries and some new, the
latter compiled by me, using cross-compiled as and ld. In any case,
the QMAGIC was the only functional.

After some debugging I identified (when loading a OMAGIC) that the
kernel sigkill the current after this checking:

http://lxr.linux.no/linux+v3.13.5/arch/x86/ia32/ia32_aout.c#L325

 325                error = vm_brk(text_addr & PAGE_MASK, map_size);
 326
 327                if (error != (text_addr & PAGE_MASK)) {
 328                        send_sig(SIGKILL, current, 0);
 329                        return error;
 330                }

I suppose this happens due to changes made, in the course of time, in
the memory mapping (vm_brk/do_brk), therefore the only one that still
works is the QMAGIC (the aligned one). Or maybe, it's purposely. [RFC]

Is important to note that when a ZMAGIC is loaded what happens is a
"Segmentation fault" and not "SIGKILL".

That was reported by others too:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/966472

-- 
Regards,

Geyslan G. Bem
hackingbits.com

Re: [RFC] Only a.out QMAGIC format is working

[Pavel Machek]

Hi!

> I was researching about old binary formats and did some tests.
> Meantime, I was able to run sucessfully only the QMAGIC format.
> Nonetheless, the OMAGIC, NMAGIC and ZMAGIC didn't work anymore.

Some time ago, I ran into similar problem, and turning off userspace
randomization was neccessary... Jiri Kosina might remember details.

									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [Kernel-BR] Re: [RFC] Only a.out QMAGIC format is working

[Geyslan Gregório Bem]

Pavel,

Thank you.

Maintainers, is there some chance to fix it or a.out is really doomed?


2014-04-19 13:15 GMT-03:00 Pavel Machek <pavel@ucw.cz>:
> Hi!
>
>> I was researching about old binary formats and did some tests.
>> Meantime, I was able to run sucessfully only the QMAGIC format.
>> Nonetheless, the OMAGIC, NMAGIC and ZMAGIC didn't work anymore.
>
> Some time ago, I ran into similar problem, and turning off userspace
> randomization was neccessary... Jiri Kosina might remember details.
>
>                                                                         Pavel
>
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
>
> --
> Você está recebendo esta mensagem porque se inscreveu no grupo "Kernel Brasil" dos Grupos do Google.
> Para cancelar inscrição nesse grupo e parar de receber e-mails dele, envie um e-mail para kernel-br+unsubscribe@googlegroups.com.
> Para postar neste grupo, envie um e-mail para kernel-br@googlegroups.com.
> Para ver esta discussão na web, acesse https://groups.google.com/d/msgid/kernel-br/20140419161552.GB27776%40amd.pavel.ucw.cz.
> Para obter mais opções, acesse https://groups.google.com/d/optout.



-- 
Regards,

Geyslan G. Bem
hackingbits.com

Re: [Kernel-BR] Re: [RFC] Only a.out QMAGIC format is working

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 314 bytes --]

On Sat, 19 Apr 2014 13:37:27 -0300, Geyslan Gregório Bem said:

> Maintainers, is there some chance to fix it or a.out is really doomed?

Is there an actual use case for a.out on a modern kernel?

In other wods, is there any reason to really care if it's doomed, since
it's been *years* since that worked?

[-- Attachment #2: Type: application/pgp-signature, Size: 848 bytes --]

Re: [Kernel-BR] Re: [RFC] Only a.out QMAGIC format is working

[Geyslan Gregório Bem]

Valdis,

2014-04-19 15:33 GMT-03:00  <Valdis.Kletnieks@vt.edu>:
> On Sat, 19 Apr 2014 13:37:27 -0300, Geyslan Gregório Bem said:
>
>> Maintainers, is there some chance to fix it or a.out is really doomed?
>
> Is there an actual use case for a.out on a modern kernel?

Maybe retrocompatibility.

>
> In other wods, is there any reason to really care if it's doomed, since
> it's been *years* since that worked?

Perhaps not, but why to continue patching the ia32_aout.c and
binfmt_aout.c if that format doesn't matter (and work) anymore? That's
was the reason that I asked if it is really doomed in linux.


-- 
Regards,

Geyslan G. Bem
hackingbits.com

Re: [Kernel-BR] Re: [RFC] Only a.out QMAGIC format is working

[Pavel Machek]

On Sat 2014-04-19 13:37:27, Geyslan Gregório Bem wrote:
> Pavel,
> 
> Thank you.
> 
> Maintainers, is there some chance to fix it or a.out is really
> doomed?

There should be chance to fix it.

Do you have some idea when it stopped working?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [Kernel-BR] Re: [RFC] Only a.out QMAGIC format is working

[Andi Kleen]

Geyslan Gregório Bem <geyslan@gmail.com> writes:
>     
>     There should be chance to fix it.
>     
> I think it too.

If it's randomization the following could help:

echo 0 > /proc/sys/kernel/randomize_va_space

Does it?

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only