1 |
PaX ---> |
2 |
Non-executable pages ---> |
3 |
[*] Enforce non-executable pages |
4 |
|
5 |
is the only option I see. I hope im blind :S |
6 |
|
7 |
On 5/20/07, Andrew Ross <aross@g.o> wrote: |
8 |
> > Hello all, I just finished installing hardened gentoo on an i686 dual |
9 |
> > p3 system and have some questions. I used the 2006.1 install CD and |
10 |
> > the stage3-hardened-2007.0 tarball. After configuring the kernel and |
11 |
> > recompiling the toolchain (binutils, gcc, virtual/libc) I did an |
12 |
> > 'emerge -e world'. This is my first hardened profile install but I |
13 |
> > have used the grsec patches on a kernel before (an amd64 system) and |
14 |
> > after rebuilding the toolchain then userland paxtest was killed on all |
15 |
> > but one or two types of stack execution. |
16 |
> >=20 |
17 |
> > However, this time (on the dual p3 system) paxtest is still able to do = |
18 |
> a |
19 |
> > lot.... |
20 |
> |
21 |
> I'm not a hardened dev, but AMD64 systems would have hardware support |
22 |
> for non-executable pages, and thus software NX (by Pax) wouldn't be |
23 |
> required. See http://en.wikipedia.org/wiki/NX_bit |
24 |
> |
25 |
> > Is this expected? Did I skip a step? Here is my kernel config. |
26 |
> |
27 |
> > # Non-executable pages |
28 |
> > # |
29 |
> > CONFIG_PAX_NOEXEC=3Dy |
30 |
> |
31 |
> I don't have a x86 system handy to compare this to, but don't you need |
32 |
> to select either page or segment-based NX? |
33 |
> |
34 |
> Cheers |
35 |
> |
36 |
> Andrew |
37 |
> |
38 |
> |
39 |
> |
40 |
|
41 |
|
42 |
-- |
43 |
Matthew Poletiek |
44 |
www.chill-fu.net |
45 |
-- |
46 |
gentoo-hardened@g.o mailing list |