1 |
Mansour Moufid wrote: |
2 |
> On Sun, Apr 19, 2009 at 9:44 AM, Thomas Sachau <tommy@g.o> wrote: |
3 |
> |
4 |
>> basile schrieb: |
5 |
>> |
6 |
>>> Hi, a have a couple of question is for Gordon and Nedd regarding |
7 |
>>> rebuilding an entire desktop system with emerge -e world, both amd64 and |
8 |
>>> i686. I'm mostly worried about the security implications of the |
9 |
>>> choices I'm making and I'm not 100% sure of my understanding. |
10 |
>>> |
11 |
>>> 1) Regarding choice of compiler. gcc-config -l gives |
12 |
>>> |
13 |
>>> [1] x86_64-pc-linux-gnu-3.4.6 |
14 |
>>> [2] x86_64-pc-linux-gnu-3.4.6-hardenednopie |
15 |
>>> [3] x86_64-pc-linux-gnu-3.4.6-hardenednopiessp |
16 |
>>> [4] x86_64-pc-linux-gnu-3.4.6-hardenednossp |
17 |
>>> [5] x86_64-pc-linux-gnu-3.4.6-vanilla |
18 |
>>> [6] x86_64-pc-linux-gnu-4.1.2 |
19 |
>>> |
20 |
>>> My understanding is that [1] is fully hardened and that [2]-[5] are |
21 |
>>> exactly what they say, respectively no pie, no pie nor ssp, no ssp and |
22 |
>>> fully vanilla. My confusion is about 4.1.2. What hardening is present |
23 |
>>> in it? (Did some hardening which wasn't present in gcc-3 make it to |
24 |
>>> gcc-4 vanilla?) What's the best practice here? |
25 |
>>> |
26 |
>> You are right with gcc-3.4.6-r2. How did you install gcc-4? It should be masked as that version does |
27 |
>> not have any builtin hardened features, so is only a normal, none-hardened gcc-4.1.2 |
28 |
>> |
29 |
> |
30 |
> This can happen when using a non-hardened stage3 tarball during the |
31 |
> install, then switching to the hardened profile later. |
32 |
> |
33 |
> I've noticed it's not immediately clear where to get hardened stages |
34 |
> in the documentation. For those wondering, the mirror URL can be found |
35 |
> in the topic on #gentoo-hardened, i.e.: |
36 |
> http://gentoo.osuosl.org/releases/${ARCH}/2008.0/stages/hardened/ |
37 |
> |
38 |
> |
39 |
|
40 |
I followed a variation of the upgrade process discussed here: |
41 |
|
42 |
http://www.gentoo.org/proj/en/hardened/toolchain-upgrade-guide.xml |
43 |
|
44 |
The differences are I used binutils-2.18 and glibc-2.8_p20080602-r1 |
45 |
|
46 |
I understand that its a VERY EARLY draft, but it proceeded without any |
47 |
problems on both i686 and amd64. I'm pretty sure I didn't loose PIE, |
48 |
but I'm not so sure about SSP. I'm playing around now with |
49 |
-fstack-protector-all in my CFLAGS. |
50 |
|
51 |
|
52 |
>>> 2) Regarding the choice of profiles on amd64. I have |
53 |
>>> |
54 |
>>> [6] hardened/amd64 |
55 |
>>> [7] hardened/amd64/multilib * |
56 |
>>> [10] hardened/linux/amd64 |
57 |
>>> |
58 |
>>> I'm using the multilib and I'm wondering what the security implications |
59 |
>>> of this decision. Also, should I be thinking about the newer [10] on |
60 |
>>> amd64? What about the similar choice on i686? |
61 |
>>> |
62 |
>>> Thanks guys. |
63 |
>>> |
64 |
>>> |
65 |
>> What security implications should be there? |
66 |
>> The newer [10] is still experimental and may change without warning. Use either [6] or [7] for now. |
67 |
>> |
68 |
>> -- |
69 |
>> Thomas Sachau |
70 |
>> |
71 |
>> Gentoo Linux Developer |
72 |
>> |
73 |
>> |
74 |
I remember reading about lots of security bugs with emulating |
75 |
libraries. I just googled for it to remind myself. So I'm wondering |
76 |
whether profile 6 is better than 7. |
77 |
|
78 |
-- |
79 |
|
80 |
Anthony G. Basile, Ph.D. |
81 |
Chair of Information Technology |
82 |
D'Youville College |
83 |
Buffalo, NY 14201 |
84 |
USA |
85 |
|
86 |
(716) 829-8197 |