linux-kernel - Re: [PATCH -mm] ALSA portman2x4 section mismatch

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Thu, 28 Jun 2007 20:15:21 +0200
From:	Sam Ravnborg <sam@...nborg.org>
To:	Randy Dunlap <randy.dunlap@...cle.com>
Cc:	Takashi Iwai <tiwai@...e.de>, lkml <linux-kernel@...r.kernel.org>,
	akpm <akpm@...ux-foundation.org>, perex@...e.cz
Subject: Re: [PATCH -mm] ALSA portman2x4 section mismatch

On Thu, Jun 28, 2007 at 10:24:38AM -0700, Randy Dunlap wrote:
> On Wed, 27 Jun 2007 12:02:32 +0200 Takashi Iwai wrote:
> 
> > At Tue, 26 Jun 2007 15:30:13 -0700,
> > Randy Dunlap wrote:
> > > 
> > > From: Randy Dunlap <randy.dunlap@...cle.com>
> > > 
> > > FIx section mismatch when CONFIG_HOTPLUG=n:
> > > 
> > > WARNING: sound/built-in.o(.exit.text+0x271): Section mismatch: reference to .init.text:snd_p
> > > ortman_unregister_all (between 'snd_portman_module_exit' and 'alsa_mpu401_uart_exit')
> > > 
> > > Signed-off-by: Randy Dunlap <randy.dunlap@...cle.com>
> > 
> > I applied it to ALSA tree now because it's the simplest way to fix
> > this warning.  But, I'm still wondering whether it was really wrong.
> > 
> > The function snd_portmap_unregister_all() is called both from the
> > module init and exit functions.  Would the module_exit function be
> > called even CONFIG_MODULE=n ?
> 
> No, it's not called when the driver is built-in.
> 
> > If not, __init_or_module should be
> > correct since the function is called only from module_init.
> 
> [adding Sam to cc:]
> 
> I see.  So to try to summarize:
> 
> When CONFIG_MODULES=n, __exit functions can be allowed to reference
> __init functions since the __init functions won't be called because
> the __exit functions are not invoked.
> 
> Does that sound correct?
In the CONFIG_MODULES=n case the __exit section is discarded so yes they
can reference anything without causing breakage.
But that said this is by no means a logical thing to allow from the principle
of minimum suprise.

>  Can & should modpost add this knowledge?
So modpost should not have this knowledge. A better approach is either to fix
the relations or drom the __init/__exit marker for a few sections.
The __init/__exit parts are mainly of benefit in two cases:
1) When someone has a lot of built-in drivers
2) In the embedded world with scarce memory

As for 1) this is a non-typical case not worth optimizing too much against.
As for 2) this is where the biggest benefit are.

Considering this alsa drivers are likely not typical embedded so here dropping a few
__init/__exit markings should be safe.
But it also depends on what other code is invoked from these functions.

	Sam
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/