Throughout the network stack we have a few places of

	if (jailed(cred))
left.  If you are running with a vnet (virtual network stack) those will
return true and defer you to classic IP-jails handling and thus things
will be "denied" or returned with an error.

Work around this problem by introducing another "jailed()" function,
jailed_without_vnet(), that also takes vnets into account, and permits
the calls, should the jail from the given cred have its own virtual
network stack.

We cannot change the classic jailed() call to do that,  as it is used
outside the network stack as well.

Discussed with:	julian, zec, jamie, rwatson (back in Sept)
MFC after:	5 days

5 files changed, 30 insertions, 6 deletions

diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c
index 0cc330c..0900541 100644
--- a/sys/kern/kern_jail.c
+++ b/sys/kern/kern_jail.c
@@ -3161,7 +3161,7 @@ prison_check_af(struct ucred *cred, int af)
 	pr = cred->cr_prison;
 #ifdef VIMAGE
 	/* Prisons with their own network stack are not limited. */
-	if (pr->pr_flags & PR_VNET)
+	if (prison_owns_vnet(cred))
 		return (0);
 #endif
 
@@ -3222,6 +3222,11 @@ prison_if(struct ucred *cred, struct sockaddr *sa)
 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
 
+#ifdef VIMAGE
+	if (prison_owns_vnet(cred))
+		return (0);
+#endif
+
 	error = 0;
 	switch (sa->sa_family)
 	{
@@ -3279,6 +3284,24 @@ jailed(struct ucred *cred)
 }
 
 /*
+ * Return 1 if the passed credential is in a jail and that jail does not
+ * have its own virtual network stack, otherwise 0.
+ */
+int
+jailed_without_vnet(struct ucred *cred)
+{
+
+	if (!jailed(cred))
+		return (0);
+#ifdef VIMAGE
+	if (prison_owns_vnet(cred))
+		return (0);
+#endif
+
+	return (1);
+}
+
+/*
  * Return the correct hostname (domainname, et al) for the passed credential.
  */
 void
diff --git a/sys/net/rtsock.c b/sys/net/rtsock.c
index 4bbd6e3..a0677ec 100644
--- a/sys/net/rtsock.c
+++ b/sys/net/rtsock.c
@@ -651,7 +651,7 @@ route_output(struct mbuf *m, struct socket *so)
 		report:
 			RT_LOCK_ASSERT(rt);
 			if ((rt->rt_flags & RTF_HOST) == 0
-			    ? jailed(curthread->td_ucred)
+			    ? jailed_without_vnet(curthread->td_ucred)
 			    : prison_if(curthread->td_ucred,
 			    rt_key(rt)) != 0) {
 				RT_UNLOCK(rt);
@@ -1312,7 +1312,7 @@ sysctl_dumpentry(struct radix_node *rn, void *vw)
 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
 		return 0;
 	if ((rt->rt_flags & RTF_HOST) == 0
-	    ? jailed(w->w_req->td->td_ucred)
+	    ? jailed_without_vnet(w->w_req->td->td_ucred)
 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
 		return (0);
 	bzero((caddr_t)&info, sizeof(info));
diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c
index 02b51eb..3573472 100644
--- a/sys/netinet/raw_ip.c
+++ b/sys/netinet/raw_ip.c
@@ -291,7 +291,7 @@ rip_input(struct mbuf *m, int off)
 			continue;
 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
 			continue;
-		if (jailed(inp->inp_cred)) {
+		if (jailed_without_vnet(inp->inp_cred)) {
 			/*
 			 * XXX: If faddr was bound to multicast group,
 			 * jailed raw socket will drop datagram.
@@ -325,7 +325,7 @@ rip_input(struct mbuf *m, int off)
 		if (!in_nullhost(inp->inp_faddr) &&
 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
 			continue;
-		if (jailed(inp->inp_cred)) {
+		if (jailed_without_vnet(inp->inp_cred)) {
 			/*
 			 * Allow raw socket in jail to receive multicast;
 			 * assume process had PRIV_NETINET_RAW at attach,
diff --git a/sys/netinet6/raw_ip6.c b/sys/netinet6/raw_ip6.c
index 335eff5..9f1236a 100644
--- a/sys/netinet6/raw_ip6.c
+++ b/sys/netinet6/raw_ip6.c
@@ -184,7 +184,7 @@ rip6_input(struct mbuf **mp, int *offp, int proto)
 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
 			continue;
-		if (jailed(in6p->inp_cred)) {
+		if (jailed_without_vnet(in6p->inp_cred)) {
 			/*
 			 * Allow raw socket in jail to receive multicast;
 			 * assume process had PRIV_NETINET_RAW at attach,
diff --git a/sys/sys/jail.h b/sys/sys/jail.h
index cb26a64..2c5d178 100644
--- a/sys/sys/jail.h
+++ b/sys/sys/jail.h
@@ -335,6 +335,7 @@ struct mount;
 struct sockaddr;
 struct statfs;
 int jailed(struct ucred *cred);
+int jailed_without_vnet(struct ucred *);
 void getcredhostname(struct ucred *, char *, size_t);
 void getcreddomainname(struct ucred *, char *, size_t);
 void getcredhostuuid(struct ucred *, char *, size_t);