| 1 |
Peter S. Mazinger wrote: |
| 2 |
|
| 3 |
> On 23 Jun 2004, Ned Ludd wrote: |
| 4 |
> |
| 5 |
> |
| 6 |
>>Side Note Off Topic: |
| 7 |
>> |
| 8 |
>>Both arm & mips fail a simple |
| 9 |
>>echo 'int main() { return 0; }' > main.c |
| 10 |
>>gcc -o main main.c -pie |
| 11 |
>> |
| 12 |
>>/usr/lib/libc_nonshared.a(elf-init.oS)(.text+0x38): In function |
| 13 |
>>`__libc_csu_init': |
| 14 |
>>: undefined reference to `__init_array_start' |
| 15 |
>>/usr/lib/libc_nonshared.a(elf-init.oS)(.text+0x3c): In function |
| 16 |
>>`__libc_csu_init': |
| 17 |
>>: undefined reference to `__init_array_end' |
| 18 |
>> |
| 19 |
>>I think these symbols are provided by linker scripts. |
| 20 |
>>Seen anything floating around where others have run into the same thing? |
| 21 |
>>If so recall the solution? |
| 22 |
> |
| 23 |
> |
| 24 |
> Is the used gcc surely patched for pie? |
| 25 |
> gcc (3.3.3+branch,3.4.0) does not support pie for arm (see my upstream |
| 26 |
> pie patches for arm). at least gcc-3.3.3-piepatches-8.7.5, |
| 27 |
> gcc-3.4.0-piepatches-8.7.6.1 have to be used. |
| 28 |
> check if the startfile/endfile sections mention pie. |
| 29 |
> |
| 30 |
> mips could be another problem, because it defaults already to PIC. |
| 31 |
> |
| 32 |
> could you run gcc -v and send the result for both cases? |
| 33 |
> |
| 34 |
> Peter |
| 35 |
> |
| 36 |
> |
| 37 |
> -- |
| 38 |
> gentoo-hardened@g.o mailing list |
| 39 |
> |
| 40 |
we already picked that out: |
| 41 |
ld does not have a linker script for -pie on the affected arches: |
| 42 |
the linker script normally contains a PROVIDE line for that. |
| 43 |
simply copy the linker script for shared executables to the extension of |
| 44 |
pie executables, this should be the "hack sluggish way" of doing things. |
| 45 |
|
| 46 |
|
| 47 |
bye, |
| 48 |
|
| 49 |
Alex (sunburned and happy pappy) |
| 50 |
-- |
| 51 |
Chaos is just a perceptional disorder of reality. |
| 52 |
|
| 53 |
-- |
| 54 |
gentoo-hardened@g.o mailing list |
Archives