I am attempting to write an init replacement that is capability-smart.
Though I'm pleased that prctl() lets me keep capabilities across a
setreuid(), maintaining caps over execve() seems impossible to do right.
I currently see a few options:
- use the CLOEXEC-pipe hack that execcap uses (parent notices
when pipe closes then rushes to set caps on child before
child notices they're gone). This looks like a race to me.
- tweak linux/fs/exec.c (prepare_binprm) to pretend that all
files have cap_inheritable and cap_effective fully set.
This seems a more elegant solution, but requires a kernel
patch.
- exec the child in a stopped state, mess with caps, then
send it SIGCONT. AFAIK, there is no way to do
execve_and_stop.
Is there a better solution available, or one in the works?
I think capabilites may be a key to achieving Pretty Good (tm) security
- but then again, so is running bind as non-root, and nobody even
bothers to do that...
-Eric
I personnaly use this simple patch which allows me
to keep caps over execve(). It allows me to give a
few more rights to some trusted users, such as
kill, insmod... without risking unlink, chown or
so. I couldn't find any other way to achieve this.
If needed, I can send you the complete prog which
sets the requested capabilities upon login,
eventually asking for a password and limited in
time of day.
Regards,
Willy
--- linux-2.2.18-wt11/fs/exec.c Fri Feb 16 23:11:52
2001
+++ linux-2.2.18-wt11+caps/fs/exec.c Thu Feb 22
20:45:33 2001
@@ -702,7 +702,10 @@
cap_clear(bprm->cap_inheritable);
cap_clear(bprm->cap_permitted);
cap_clear(bprm->cap_effective);
-
+/*** FIXME: just a test : keep permitted and
effective ******/
+bprm->cap_permitted =
cap_intersect(current->cap_inheritable,current->cap_permitted);
+bprm->cap_effective =
cap_intersect(current->cap_inheritable,current->cap_effective);
+/*** /FIXME ****/
/* To support inheritance of root-permissions
and suid-root
* executables under compatibility mode, we
raise all three
* capability sets for the file.
___________________________________________________________
Do You Yahoo!? -- Pour faire vos courses sur le Net,
Yahoo! Shopping : http://fr.shopping.yahoo.com
Hi!
> I am attempting to write an init replacement that is capability-smart.
> Though I'm pleased that prctl() lets me keep capabilities across a
> setreuid(), maintaining caps over execve() seems impossible to do right.
>
> I currently see a few options:
> - use the CLOEXEC-pipe hack that execcap uses (parent notices
> when pipe closes then rushes to set caps on child before
> child notices they're gone). This looks like a race to me.
> - tweak linux/fs/exec.c (prepare_binprm) to pretend that all
> files have cap_inheritable and cap_effective fully set.
> This seems a more elegant solution, but requires a kernel
> patch.
> - exec the child in a stopped state, mess with caps, then
> send it SIGCONT. AFAIK, there is no way to do
> execve_and_stop.
What about ptrace? It should be able to do this kind of stuff... but
it is going to be messy.
Pavel
--
I'm [email protected]. "In my country we have almost anarchy and I don't care."
Panos Katsaloulis describing me w.r.t. patents at [email protected]