1 |
On April 20, 2004 12:13 pm, Ned Ludd wrote: |
2 |
> Robert, |
3 |
> |
4 |
> I must apologize the previous example code was printing the address |
5 |
> of __guard and not value of the __guard itself. |
6 |
> |
7 |
> Can we try something like? |
8 |
> ------------------------- |
9 |
> extern unsigned long __guard[]; |
10 |
> |
11 |
> int main () { |
12 |
> printf("__guard\t=\t0x%08x;\n", __guard[0]); |
13 |
> return 0; |
14 |
> } |
15 |
> ----------------------- |
16 |
|
17 |
This output changes for each runtime using sysctl erandom. |
18 |
|
19 |
> I've also been giving some more thought into using the gettimeofday() as |
20 |
> we had discussed one day.. It could work across the board and would and |
21 |
> no kernel patching would be required which is probably more ideal for |
22 |
> Gentoo users as we have a butt ton of kernels. Disadvantage of course |
23 |
> would be potential timing attacks by local users. |
24 |
|
25 |
Using gettimeofday as a seed is less secure than we have now. I went to the |
26 |
trouble of getting sysctl-erandom working because it is equaly secure as |
27 |
urandom. I also have plans to patch mkstemp to use sysctl-erandom, and |
28 |
anywhere else that isn't crypto. Sysctl-erandom will not be only for SSP. |
29 |
Gettimeofday could be added to the canary fallback however. Infact I have an |
30 |
arcfour lib to hash gettimeofday that could be added to the ssp.c file, but |
31 |
it is a lot of bloat for a fallback feature, and again erandom does a better |
32 |
job. |
33 |
|
34 |
|
35 |
|
36 |
-- |
37 |
gentoo-hardened@g.o mailing list |