| 1 |
On Monday 25 March 2013 03:01:51 Kfir Lavi wrote: |
| 2 |
> I'm looking for a way to reduce glibc code size. |
| 3 |
> It can be a way to make system smaller and minimize the impact |
| 4 |
> of attack vectors in glibc, as in return-to-libc attack. |
| 5 |
> |
| 6 |
> Lets say I'm deleting the program 'mkdir', and mkdir uses a function |
| 7 |
> in glibc that non of the other parts of the system uses. |
| 8 |
> Then I want to eliminate this function from glibc. This leads to smaller |
| 9 |
> code and if this function is used in some attack scenario, maybe prevent |
| 10 |
> it. |
| 11 |
> |
| 12 |
> Is there a way to do it? |
| 13 |
> Can you help me think how to build a tool like this? or, integrate |
| 14 |
> with existing tools. |
| 15 |
|
| 16 |
the only thing i've seen in the past was a hacky script that utilize the |
| 17 |
uClibc build system to cull objects until things stopped linking. it had very |
| 18 |
constrained use where i'd safely work, and was never generalized. i don't |
| 19 |
remember the name of it now (was a few years ago), but having read the |
| 20 |
[limited] source, i wouldn't bother using it as a base. |
| 21 |
|
| 22 |
otherwise, i haven't heard of any tools that do what you want, but i've seen |
| 23 |
many people request it. unfortunately, it's a tough nut to crack, and the |
| 24 |
vast majority of people requesting it didn't have the technical skills to even |
| 25 |
think about a solution let alone implement it. |
| 26 |
|
| 27 |
i [pessimistically] suspect we're talking many man months here, and the skill |
| 28 |
set involves knowledge of the ELF format at a fairly low level. |
| 29 |
-mike |
Archives