Re: Early crash (was: Re: module: show version information for built-in modules in sysfs)

[Dmitry Torokhov <dtor@vmware.com>]

On Wed, Feb 02, 2011 at 04:24:59PM -0800, Dmitry Torokhov wrote:
> On Wed, Feb 02, 2011 at 04:10:04PM -0800, Andreas Schwab wrote:
> > Dmitry Torokhov <dtor@vmware.com> writes:
> > 
> > > Even pointers? I'd expect pointers to be aligned on 4-bytes boundaries?
> > 
> > Pointers are not special in any way.  Why should they?  On the machine
> > level pointers are just numbers.
> 
> Are pointers (along with ints/longs) on m68k naturally aligned on word
> boundary even though they are 32 bit?
> 
> Anyway, here is the description that introduced alignment statement:
> 
> commit 02dba5c6439cff34936460b95cd1ba42b370f345
> Author: ak <ak>
> Date:   Sat Jun 21 16:18:16 2003 +0000
> 
>     [PATCH] Fix over-alignment problem on x86-64
> 
>     Thanks to Jan Hubicka who suggested this fix.
> 
>     The problem seems to be that gcc generates a 32byte alignment for static
>     objects > 32bytes.  This causes gas to set a high alignment on the
>     section, which causes the uneven (not multiple of sizeof(struct
>     kernel_param)) section size.  The pointer division with a base not being
>     a multiple of sizeof(*ptr) then causes the invalid result.
> 
>     This just forces a small alignment, which makes the section end come out
>     with the correct alignment.
> 
>     The only mystery left is why ld chose a 16 byte padding instead of
>     32byte.
> 
>     BKrev: 3ef485487jZN-h3PtASDeL2Vs55NIg
> 
> 
> I guess this does not directly apply to modversions since they are
> currently under 32 bytes, but I wonder what happen if we decide to
> extend one of the structures involved...
> 
> I guess explicitly setting alignment requirement for struct
> module_version_attribute is the best option.
> 

So here is the patch that explicitly specifies alignment for struct
module_version_attribute. I tested it on i386 and x86_64 and I believe
it will fix the issue with m68k but I do not have access to such a box.

Thanks,
Dmitry

>From f0e0e10b58b22047e36e21a022abf5e86b5819c2 Mon Sep 17 00:00:00 2001
From: Dmitry Torokhov <dtor@vmware.com>
Date: Fri, 4 Feb 2011 13:30:10 -0800
Subject: [PATCH] module: explicitly align module_version_attribute structure

We force particular alignment when we generate attribute structures
when generation MODULE_VERSION() data and we need to make sure that
this alignment is followed when we iterate over these structures,
otherwise we may crash on platforms whose natural alignment is not
sizeof(void *), such as m68k.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Dmitry Torokhov <dtor@vmware.com>
---
 include/linux/module.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/linux/module.h b/include/linux/module.h
index e7c6385..de5cd21 100644
--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -62,7 +62,7 @@ struct module_version_attribute {
 	struct module_attribute mattr;
 	const char *module_name;
 	const char *version;
-};
+} __attribute__ ((__aligned__(sizeof(void *))));
 
 struct module_kobject
 {
-- 
1.7.3.2