[PATCH] lsm: fix default return value of the socket_getpeersec_* hooks

[Ondrej Mosnacek]

On Tue, Jan 30, 2024 at 4:01 AM Paul Moore <paul at paul-moore.com> wrote:
>
> On Mon, Jan 29, 2024 at 6:25 PM Casey Schaufler <casey at schaufler-ca.com> wrote:
> >
> > On 1/29/2024 3:02 PM, Paul Moore wrote:
> > > On Fri, Jan 26, 2024 at 1:45 PM Ondrej Mosnacek <omosnace at redhat.com> wrote:
> > >> For these hooks the true "neutral" value is -EOPNOTSUPP, which is
> > >> currently what is returned when no LSM provides this hook and what LSMs
> > >> return when there is no security context set on the socket. Correct the
> > >> value in <linux/lsm_hooks.h> and adjust the dispatch functions in
> > >> security/security.c to avoid issues when the BPF LSM is enabled.
> > >>
> > >> Fixes: 98e828a0650f ("security: Refactor declaration of LSM hooks")
> > >> Signed-off-by: Ondrej Mosnacek <omosnace at redhat.com>
> > >> ---
> > >>  include/linux/lsm_hook_defs.h |  4 ++--
> > >>  security/security.c           | 31 +++++++++++++++++++++++++++----
> > >>  2 files changed, 29 insertions(+), 6 deletions(-)
> > >>
> > >> diff --git a/include/linux/lsm_hook_defs.h b/include/linux/lsm_hook_defs.h
> > >> index 185924c56378..76458b6d53da 100644
> > >> --- a/include/linux/lsm_hook_defs.h
> > >> +++ b/include/linux/lsm_hook_defs.h
> > >> @@ -315,9 +315,9 @@ LSM_HOOK(int, 0, socket_getsockopt, struct socket *sock, int level, int optname)
> > >>  LSM_HOOK(int, 0, socket_setsockopt, struct socket *sock, int level, int optname)
> > >>  LSM_HOOK(int, 0, socket_shutdown, struct socket *sock, int how)
> > >>  LSM_HOOK(int, 0, socket_sock_rcv_skb, struct sock *sk, struct sk_buff *skb)
> > >> -LSM_HOOK(int, 0, socket_getpeersec_stream, struct socket *sock,
> > >> +LSM_HOOK(int, -ENOPROTOOPT, socket_getpeersec_stream, struct socket *sock,
> > >>          sockptr_t optval, sockptr_t optlen, unsigned int len)
> > >> -LSM_HOOK(int, 0, socket_getpeersec_dgram, struct socket *sock,
> > >> +LSM_HOOK(int, -ENOPROTOOPT, socket_getpeersec_dgram, struct socket *sock,
> > >>          struct sk_buff *skb, u32 *secid)
> > >>  LSM_HOOK(int, 0, sk_alloc_security, struct sock *sk, int family, gfp_t priority)
> > >>  LSM_HOOK(void, LSM_RET_VOID, sk_free_security, struct sock *sk)
> > >> diff --git a/security/security.c b/security/security.c
> > >> index 6196ccaba433..3aaad75c9ce8 100644
> > >> --- a/security/security.c
> > >> +++ b/security/security.c
> > >> @@ -4624,8 +4624,20 @@ EXPORT_SYMBOL(security_sock_rcv_skb);
> > >>  int security_socket_getpeersec_stream(struct socket *sock, sockptr_t optval,
> > >>                                       sockptr_t optlen, unsigned int len)
> > >>  {
> > >> -       return call_int_hook(socket_getpeersec_stream, -ENOPROTOOPT, sock,
> > >> -                            optval, optlen, len);
> > >> +       struct security_hook_list *hp;
> > >> +       int rc;
> > >> +
> > >> +       /*
> > >> +        * Only one module will provide a security context.
> > >> +        */
> > >> +       hlist_for_each_entry(hp, &security_hook_heads.socket_getpeersec_stream,
> > >> +                            list) {
> > >> +               rc = hp->hook.socket_getpeersec_stream(sock, optval, optlen,
> > >> +                                                      len);
> > >> +               if (rc != LSM_RET_DEFAULT(socket_getpeersec_stream))
> > >> +                       return rc;
> > >> +       }
> > >> +       return LSM_RET_DEFAULT(socket_getpeersec_stream);
> > >>  }
> > >
> > > I'm beginning to wonder if we shouldn't update call_int_hook() so that
> > > it works for LSM_RET_DEFAULT() instead of assuming a zero/0 return
> > > value.  Thoughts?
> >
> > call_int_hook() was intended to address the "normal" case, where all
> > hooks registered would be called and the first error, if any, would
> > result in an immediate failure return. Hooks that behaved in any other
> > manner were expected to be open coded. The point of using the macros
> > was to reduce so much code duplication. I really don't want to see
> > call_int_hook() evolve into something hard to work with, or that has
> > non-obvious side effects. I think we could probably integrate
> > LSM_RET_DEFAULT() safely, but I'm wary of hiding these abnormal cases
> > in the macro.
>
> Yes, I'm not talking about modifying call_int_hook() to handle
> something like security_vm_enough_memory_mm(), I'm just talking about
> updating it use LSM_RET_DEFAULT() instead of zero.
>
> While we are at it, we should probably get rid of the second parameter
> too, @IRC, and just use the assigned LSM_RET_DEFAULT().  That always
> struck me as a bug waiting to happen if/when those two fell out of
> sync.

You're reading my mind :) I already started writing a patch that does
exactly that after I posted the security_inode_getsecctx() patch.
While working on it I gradually found two more pre-existing issues, so
I wanted to post fixes for them before the refactor for better
backportability. I should be able to post the patch today.

BTW, the IRC param removal means that a few of the existing
call_int_hook() calls have to be open-coded, but even then the patch
removes more LoC than it adds, so I think it's worth it.

-- 
Ondrej Mosnacek
Senior Software Engineer, Linux Security - SELinux kernel
Red Hat, Inc.