[PATCH 5/5] UBUNTU: SAUCE: AppArmor: Fix oops there is no tracer and doing unsafe transition.

[Andy Whitcroft]

On Tue, Nov 10, 2009 at 10:29:13AM -0800, John Johansen wrote:
> BugLink: http://bugs.launchpad.net/bugs/480112
> 
> SRU Justification:  This bug can cause confined process to oops at address 0.
> This can occur when executing a process if the LSM_UNSAFE_PTRACE |
> LSM_UNSAFE_PTRACE_CAP flags are set.  The likely hood of if/how often this
> will occur depends on if ptrace is being used.
> 
> As reported by Tetsuo Handa on kernel-team mailing list:
> 
> In aa_may_change_ptraced_domain, if (!tracer) cred == NULL, and
> put_cred(cred) will oops.  This will only happen on exec if the task
> is marked as LSM_UNSAFE_PTRACE | LSM_UNSAFE_PTRACE_CAP, so should
> only happen to ptraced tasks that are confined.
> 
> Fix this by returning directly from aa_may_change_ptrace_domain if
> there is now tracer.
> 
> Signed-off-by: John Johansen <john.johansen at canonical.com>
> ---
>  ubuntu/apparmor/domain.c |    4 ++++
>  1 files changed, 4 insertions(+), 0 deletions(-)
> 
> diff --git a/ubuntu/apparmor/domain.c b/ubuntu/apparmor/domain.c
> index fe89ddc..12e45c6 100644
> --- a/ubuntu/apparmor/domain.c
> +++ b/ubuntu/apparmor/domain.c
> @@ -64,6 +64,10 @@ static int aa_may_change_ptraced_domain(struct task_struct *task,
>  		cred = aa_get_task_policy(tracer, &tracerp);
>  	rcu_read_unlock();
>  
> +	/* not ptraced */
> +	if (!tracer)
> +		return 0;
> +
>  	if (!tracerp)
>  		goto out;
>  
> -- 
> 1.6.3.3

Hrm, now perhaps this is fixing the concern I raised in the previous
patch.  Perhaps it would be safer to simply make the put_cred()
incantation instead?

  if (cred)
    put_cred(cred);

-apw