[RFC PATCH v2] security, capability: pass object information to security_capable

[Paul Moore]

On Thu, Aug 1, 2019 at 10:43 AM Aaron Goidel <acgoide at tycho.nsa.gov> wrote:
> From: Nicholas Franck <nhfran2 at tycho.nsa.gov>
>
> At present security_capable does not pass any object information
> and therefore can neither audit the particular object nor take it
> into account. Augment the security_capable interface to support
> passing supplementary data. Use this facility initially to convey
> the inode for capability checks relevant to inodes. This only
> addresses capable_wrt_inode_uidgid calls; other capability checks
> relevant to inodes will be addressed in subsequent changes. In the
> future, this will be further extended to pass object information for
> other capability checks such as the target task for CAP_KILL.
>
> In SELinux this new information is leveraged here to include the inode
> in the audit message. In the future, it could also be used to perform
> a per inode capability checks.
>
> It would be possible to fold the existing opts argument into this new
> supplementary data structure. This was omitted from this change to
> minimize changes.
>
> Signed-off-by: Nicholas Franck <nhfran2 at tycho.nsa.gov>
> Signed-off-by: Aaron Goidel <acgoide at tycho.nsa.gov>
> ---
> v2:
> - Changed order of audit prints so optional information comes second
> ---
>  include/linux/capability.h             |  7 ++++++
>  include/linux/lsm_audit.h              |  5 +++-
>  include/linux/lsm_hooks.h              |  3 ++-
>  include/linux/security.h               | 23 +++++++++++++-----
>  kernel/capability.c                    | 33 ++++++++++++++++++--------
>  kernel/seccomp.c                       |  2 +-
>  security/apparmor/capability.c         |  8 ++++---
>  security/apparmor/include/capability.h |  4 +++-
>  security/apparmor/ipc.c                |  2 +-
>  security/apparmor/lsm.c                |  5 ++--
>  security/apparmor/resource.c           |  2 +-
>  security/commoncap.c                   | 11 +++++----
>  security/lsm_audit.c                   | 21 ++++++++++++++--
>  security/safesetid/lsm.c               |  3 ++-
>  security/security.c                    |  5 ++--
>  security/selinux/hooks.c               | 20 +++++++++-------
>  security/smack/smack_access.c          |  2 +-
>  17 files changed, 110 insertions(+), 46 deletions(-)

You should CC the linux-audit list, I've added them on this mail.

I had hoped to see some thought put into the idea of dynamically
emitting the proper audit records as I mentioned in the previous patch
set, but regardless there are some comments on this code as written
...

> diff --git a/security/lsm_audit.c b/security/lsm_audit.c
> index 33028c098ef3..18cc7c956b69 100644
> --- a/security/lsm_audit.c
> +++ b/security/lsm_audit.c
> @@ -229,9 +229,26 @@ static void dump_common_audit_data(struct audit_buffer *ab,
>         case LSM_AUDIT_DATA_IPC:
>                 audit_log_format(ab, " key=%d ", a->u.ipc_id);
>                 break;
> -       case LSM_AUDIT_DATA_CAP:
> -               audit_log_format(ab, " capability=%d ", a->u.cap);
> +       case LSM_AUDIT_DATA_CAP: {
> +               const struct inode *inode;
> +
> +               audit_log_format(ab, " capability=%d ", a->u.cap_struct.cap);
> +               if (a->u.cap_struct.cad) {
> +                       switch (a->u.cap_struct.cad->type) {
> +                       case CAP_AUX_DATA_INODE: {
> +                               inode = a->u.cap_struct.cad->u.inode;
> +
> +                               audit_log_format(ab, " dev=");
> +                               audit_log_untrustedstring(ab,
> +                                       inode->i_sb->s_id);
> +                               audit_log_format(ab, " ino=%lu",
> +                                       inode->i_ino);
> +                               break;
> +                       }

Since you are declaring "inode" further up, there doesn't appear to be
any need for the CAP_AUX_DATA_INODE braces, please remove them.

The general recommended practice when it comes to "sometimes" fields
in an audit record, is to always record them in the record, but use a
value of "?" when there is nothing relevant to record.  For example,
when *not* recording inode information you would do something like the
following:

  audit_log_format(ab, " dev=? ino=?");

> +                       }
> +               }
>                 break;
> +       }
>         case LSM_AUDIT_DATA_PATH: {
>                 struct inode *inode;
>

-- 
paul moore
www.paul-moore.com