1 |
>The question to ask yourself is, "Why would this process need |
2 |
>to get random numbers from /dev/urandom?" |
3 |
>Same for sshd below (though the answer for it is |
4 |
>probably a bit more obvious). |
5 |
|
6 |
That's fairly obvious in both cases; gradm was called as gradm -R and thus |
7 |
needs random numbers for handling of the password input; sshd needing random |
8 |
data for its operation is also understandable (new session keys etc.) |
9 |
|
10 |
>> /dev/random r |
11 |
|
12 |
>I would expect the same from these lines. However, I would |
13 |
>NEVER do this. You do not want all processes to have |
14 |
>access to your limited pool of high |
15 |
>grade random numbers (or any pool of random numbers). |
16 |
|
17 |
Definitely true, but I don't think I'm that far wrt my configuration. First I |
18 |
need to really understand how the system works and what's going on -only then |
19 |
does it make sense to start tuning. |
20 |
|
21 |
I think the default grsecurity policies could use some updates to make them a |
22 |
really useful starting point instead of a source of pitfalls for the unwary |
23 |
:-) |
24 |
|
25 |
Back to the log entries: the sshd entry is obvious now, I had overlooked the |
26 |
"o" subject in the sshd acl and /dev/urandom is not explicitely listet in the |
27 |
sshd acl. |
28 |
|
29 |
Understanding WHY it's not a good idea to let sshd inherit the default acls |
30 |
is another matter. |
31 |
|
32 |
Could the /dev/urandom message for gradm -R be an artefact of reloading the |
33 |
ruleset? |
34 |
|
35 |
Bye, Martin |
36 |
|
37 |
-- |
38 |
gentoo-hardened@g.o mailing list |