Motorola Shared BSPs not Linking

[Joel Sherrill]

I do not think ld can remove any symbols. Otherwise you would never have symbols like RAM_BASE.

If you notice closely, __rtems_start is from assembly and __rtems_end is from a linker script. Nm sees them both and has for a decade. Ld saw them both for a decade also. Something simply broke.

It is not valid to assume __rtems_start is 0. This is a unzipping loader which writes the unzipped data at __rtems_start. That could be any address that does not conflict with the loader.

--joel

Thomas Doerfler <Thomas.Doerfler at embedded-brains.de> wrote:

>Ralf,
>
>when I look at Joels further bug report on
>
>http://sourceware.org/bugzilla/show_bug.cgi?id=13991
>
>and the test case he provided, I can see that the symbol "_rtems_start"
>seems to be available in "hello.exe", otherwise the final line
>
>-bash-4.2$ powerpc-rtems4.10-nm -g hello.exe | grep rtems_start
>00000000 T __rtems_start
>
>would not report the symbol as content of "hello.exe".
>
>So I doubt that the reported problem is really due to a non-provided
>symbol, it looks rather like powerpc-rtems4.11-ld would somehow chew up
>the symbol "__rtems_start" and then complain about it being undefined.
>
>Or is ld allowed to remove text symbols located at 0x0000 ?
>
>wkr,
>
>Thomas.
>
>
>Am 05.04.2012 10:16, schrieb Ralf Corsepius:
>> On 04/04/2012 03:34 PM, Ralf Corsepius wrote:
>>> On 04/04/2012 02:49 PM, Ralf Corsepius wrote:
>>>> On 02/23/2012 05:15 PM, Till Straumann wrote:
>>>>>
>>>>> My last build of 'master' still linked fine but that's been a while
>>>>> (7/2011, I believe).
>>>>
>>>> FYI: I have narrowed down this breakdown to binutils.
>>>>
>>>> At least for me, building the mcp750 with gcc-4.6.x or gcc-4.7.0, _both_
>>>> fail when using binutils-2.22, but both do _not fail_ when using
>>>> binutils-2.20.1.
>>>
>>> It also works with binutils-2.21.1.
>>>
>>> => this issue was introduced between 2.21.1 and 2.22.
>> FWIW: As far as I can gather is a behavioral change in ld since
>> binutils-2.22
>> 
>> It seems to error out when linkage encounters references to undefined
>> symbols in linker scripts. This is unlike binutils < 2.22, which seem to
>> have silently ignored them and "set them to 0".
>> 
>> That said, the patch below fixes this issue for me:
>> 
>> index a2567a9..3ea50bb 100644
>> --- a/c/src/lib/libbsp/powerpc/shared/bootloader/ppcboot.lds
>> +++ b/c/src/lib/libbsp/powerpc/shared/bootloader/ppcboot.lds
>> @@ -78,6 +78,7 @@ SECTIONS
>>    }
>>    PROVIDE(_binary_initrd_gz_start = 0);
>>    PROVIDE(_binary_initrd_gz_end = 0);
>> +  PROVIDE(__rtems_start = 0);
>>    _rtems_gz_size = _binary_rtems_gz_end - _binary_rtems_gz_start;
>>    _rtems_size = __rtems_end - __rtems_start;
>>    .bss :
>> 
>> Unless somebody objects, I am going to commit it.
>> 
>> Ralf
>> 
>> _______________________________________________
>> rtems-users mailing list
>> rtems-users at rtems.org
>> http://www.rtems.org/mailman/listinfo/rtems-users
>
>
>-- 
>--------------------------------------------
>Embedded Brains GmbH
>Thomas Doerfler           Obere Lagerstr. 30
>D-82178 Puchheim          Germany
>email: Thomas.Doerfler at embedded-brains.de
>Phone: +49-89-18908079-2
>Fax:   +49-89-18908079-9
>_______________________________________________
>rtems-users mailing list
>rtems-users at rtems.org
>http://www.rtems.org/mailman/listinfo/rtems-users