Re: Problem with binutils 2.19.1 with Cortex-a9/Linux-2.6.28

[Nitin Garg <nitingarg98 at gmail dot com>]

Hi Ramana,

This is the section I can see in my LKM (.ko) with binutils 2.19.1:

RELOCATION RECORDS FOR [.eh_frame]:
OFFSET   TYPE              VALUE
0000001c R_ARM_REL32       .text
00000038 R_ARM_REL32       .text
00000058 R_ARM_REL32       .text

I tried the suggestion at this
link:http://forums.arm.com/index.php?s=da73b37532e83517674e1199b4166342&showtopic=8122&mode=linear

If -fno-dwarf2-cfi-asm is present in CFLAGS, R_ARM_REL32 section is
not generated.
This relocation needs to be handled in arch/arm/kernel/module.c, but
there is nothing in there till now.

Thanks,
Nitin


On Mon, Jul 27, 2009 at 6:19 AM, Ramana
Radhakrishnan<ramana.radhakrishnan@arm.com> wrote:
>
> On Thu, 2009-07-23 at 11:05 -0500, Nitin Garg wrote:
>> Hi All,
>>
>> With the binutils 2.19.1 and gcc 4.4.0+kernel Linux 2.6.28.10, I am
>> seeing an error "undefined relocation: 3" while doing insmod of
>> loadable kernel modules. The LKM are build using this toolchain.
>> If I change to binutils 2.19, the problem goes away.
>
> Unless you provide a testcase or an example of where this relocation is
> being produced, it wouldn't be possible to comment if this is incorrect
> generation of a relocation.
>
>
> Googling around seems to indicate that similar issues were encountered
> with the PPC Linux port which was to do with .eh_frame being generated.
> A reference thread is available here.
>
> http://linux.derkeiler.com/Mailing-Lists/Kernel/2009-02/msg10299.html
>
> I can only speculate that this is your issue.
>
>
> cheers
> Ramana
>
>
>