ipv6,mcast: always hold idev->lock before mca_lock
[linux-3.10.git] / net / ipv6 / datagram.c
index 6157388..4b56cbb 100644 (file)
@@ -22,6 +22,7 @@
 #include <linux/ipv6.h>
 #include <linux/route.h>
 #include <linux/slab.h>
+#include <linux/export.h>
 
 #include <net/ipv6.h>
 #include <net/ndisc.h>
 #include <net/transp_v6.h>
 #include <net/ip6_route.h>
 #include <net/tcp_states.h>
+#include <net/dsfield.h>
 
 #include <linux/errqueue.h>
 #include <asm/uaccess.h>
 
+static bool ipv6_mapped_addr_any(const struct in6_addr *a)
+{
+       return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
+}
+
 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 {
        struct sockaddr_in6     *usin = (struct sockaddr_in6 *) uaddr;
        struct inet_sock        *inet = inet_sk(sk);
        struct ipv6_pinfo       *np = inet6_sk(sk);
-       struct in6_addr         *daddr, *final_p = NULL, final;
+       struct in6_addr         *daddr, *final_p, final;
        struct dst_entry        *dst;
-       struct flowi            fl;
+       struct flowi6           fl6;
        struct ip6_flowlabel    *flowlabel = NULL;
+       struct ipv6_txoptions   *opt;
        int                     addr_type;
        int                     err;
 
@@ -58,14 +66,14 @@ int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
        if (usin->sin6_family != AF_INET6)
                return -EAFNOSUPPORT;
 
-       memset(&fl, 0, sizeof(fl));
+       memset(&fl6, 0, sizeof(fl6));
        if (np->sndflow) {
-               fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
-               if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
-                       flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
+               fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
+               if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
+                       flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
                        if (flowlabel == NULL)
                                return -EINVAL;
-                       ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
+                       usin->sin6_addr = flowlabel->dst;
                }
        }
 
@@ -92,7 +100,7 @@ int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
                sin.sin_port = usin->sin6_port;
 
                err = ip4_datagram_connect(sk,
-                                          (struct sockaddr*) &sin,
+                                          (struct sockaddr *) &sin,
                                           sizeof(sin));
 
 ipv4_connected:
@@ -101,17 +109,22 @@ ipv4_connected:
 
                ipv6_addr_set_v4mapped(inet->inet_daddr, &np->daddr);
 
-               if (ipv6_addr_any(&np->saddr))
+               if (ipv6_addr_any(&np->saddr) ||
+                   ipv6_mapped_addr_any(&np->saddr))
                        ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
 
-               if (ipv6_addr_any(&np->rcv_saddr))
+               if (ipv6_addr_any(&np->rcv_saddr) ||
+                   ipv6_mapped_addr_any(&np->rcv_saddr)) {
                        ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
                                               &np->rcv_saddr);
+                       if (sk->sk_prot->rehash)
+                               sk->sk_prot->rehash(sk);
+               }
 
                goto out;
        }
 
-       if (addr_type&IPV6_ADDR_LINKLOCAL) {
+       if (__ipv6_addr_needs_scope_id(addr_type)) {
                if (addr_len >= sizeof(struct sockaddr_in6) &&
                    usin->sin6_scope_id) {
                        if (sk->sk_bound_dev_if &&
@@ -132,8 +145,8 @@ ipv4_connected:
                }
        }
 
-       ipv6_addr_copy(&np->daddr, daddr);
-       np->flow_label = fl.fl6_flowlabel;
+       np->daddr = *daddr;
+       np->flow_label = fl6.flowlabel;
 
        inet->inet_dport = usin->sin6_port;
 
@@ -142,62 +155,46 @@ ipv4_connected:
         *      destination cache for it.
         */
 
-       fl.proto = sk->sk_protocol;
-       ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
-       ipv6_addr_copy(&fl.fl6_src, &np->saddr);
-       fl.oif = sk->sk_bound_dev_if;
-       fl.mark = sk->sk_mark;
-       fl.fl_ip_dport = inet->inet_dport;
-       fl.fl_ip_sport = inet->inet_sport;
-
-       if (!fl.oif && (addr_type&IPV6_ADDR_MULTICAST))
-               fl.oif = np->mcast_oif;
-
-       security_sk_classify_flow(sk, &fl);
-
-       if (flowlabel) {
-               if (flowlabel->opt && flowlabel->opt->srcrt) {
-                       struct rt0_hdr *rt0 = (struct rt0_hdr *) flowlabel->opt->srcrt;
-                       ipv6_addr_copy(&final, &fl.fl6_dst);
-                       ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
-                       final_p = &final;
-               }
-       } else if (np->opt && np->opt->srcrt) {
-               struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
-               ipv6_addr_copy(&final, &fl.fl6_dst);
-               ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
-               final_p = &final;
-       }
+       fl6.flowi6_proto = sk->sk_protocol;
+       fl6.daddr = np->daddr;
+       fl6.saddr = np->saddr;
+       fl6.flowi6_oif = sk->sk_bound_dev_if;
+       fl6.flowi6_mark = sk->sk_mark;
+       fl6.fl6_dport = inet->inet_dport;
+       fl6.fl6_sport = inet->inet_sport;
 
-       err = ip6_dst_lookup(sk, &dst, &fl);
-       if (err)
+       if (!fl6.flowi6_oif && (addr_type&IPV6_ADDR_MULTICAST))
+               fl6.flowi6_oif = np->mcast_oif;
+
+       security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
+
+       opt = flowlabel ? flowlabel->opt : np->opt;
+       final_p = fl6_update_dst(&fl6, opt, &final);
+
+       dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
+       err = 0;
+       if (IS_ERR(dst)) {
+               err = PTR_ERR(dst);
                goto out;
-       if (final_p)
-               ipv6_addr_copy(&fl.fl6_dst, final_p);
-
-       err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
-       if (err < 0) {
-               if (err == -EREMOTE)
-                       err = ip6_dst_blackhole(sk, &dst, &fl);
-               if (err < 0)
-                       goto out;
        }
 
        /* source address lookup done in ip6_dst_lookup */
 
        if (ipv6_addr_any(&np->saddr))
-               ipv6_addr_copy(&np->saddr, &fl.fl6_src);
+               np->saddr = fl6.saddr;
 
        if (ipv6_addr_any(&np->rcv_saddr)) {
-               ipv6_addr_copy(&np->rcv_saddr, &fl.fl6_src);
+               np->rcv_saddr = fl6.saddr;
                inet->inet_rcv_saddr = LOOPBACK4_IPV6;
+               if (sk->sk_prot->rehash)
+                       sk->sk_prot->rehash(sk);
        }
 
        ip6_dst_store(sk, dst,
-                     ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
+                     ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
                      &np->daddr : NULL,
 #ifdef CONFIG_IPV6_SUBTREES
-                     ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
+                     ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
                      &np->saddr :
 #endif
                      NULL);
@@ -207,6 +204,7 @@ out:
        fl6_sock_release(flowlabel);
        return err;
 }
+EXPORT_SYMBOL_GPL(ip6_datagram_connect);
 
 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
                     __be16 port, u32 info, u8 *payload)
@@ -243,7 +241,7 @@ void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
                kfree_skb(skb);
 }
 
-void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
+void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
 {
        struct ipv6_pinfo *np = inet6_sk(sk);
        struct sock_exterr_skb *serr;
@@ -262,7 +260,7 @@ void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
        skb_put(skb, sizeof(struct ipv6hdr));
        skb_reset_network_header(skb);
        iph = ipv6_hdr(skb);
-       ipv6_addr_copy(&iph->daddr, &fl->fl6_dst);
+       iph->daddr = fl6->daddr;
 
        serr = SKB_EXT_ERR(skb);
        serr->ee.ee_errno = err;
@@ -273,7 +271,7 @@ void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
        serr->ee.ee_info = info;
        serr->ee.ee_data = 0;
        serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
-       serr->port = fl->fl_ip_dport;
+       serr->port = fl6->fl6_dport;
 
        __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
        skb_reset_transport_header(skb);
@@ -282,6 +280,41 @@ void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
                kfree_skb(skb);
 }
 
+void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
+{
+       struct ipv6_pinfo *np = inet6_sk(sk);
+       struct ipv6hdr *iph;
+       struct sk_buff *skb;
+       struct ip6_mtuinfo *mtu_info;
+
+       if (!np->rxopt.bits.rxpmtu)
+               return;
+
+       skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
+       if (!skb)
+               return;
+
+       skb_put(skb, sizeof(struct ipv6hdr));
+       skb_reset_network_header(skb);
+       iph = ipv6_hdr(skb);
+       iph->daddr = fl6->daddr;
+
+       mtu_info = IP6CBMTU(skb);
+
+       mtu_info->ip6m_mtu = mtu;
+       mtu_info->ip6m_addr.sin6_family = AF_INET6;
+       mtu_info->ip6m_addr.sin6_port = 0;
+       mtu_info->ip6m_addr.sin6_flowinfo = 0;
+       mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
+       mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
+
+       __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
+       skb_reset_transport_header(skb);
+
+       skb = xchg(&np->rxpmtu, skb);
+       kfree_skb(skb);
+}
+
 /*
  *     Handle MSG_ERRQUEUE
  */
@@ -322,19 +355,19 @@ int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
                sin->sin6_family = AF_INET6;
                sin->sin6_flowinfo = 0;
                sin->sin6_port = serr->port;
-               sin->sin6_scope_id = 0;
                if (skb->protocol == htons(ETH_P_IPV6)) {
-                       ipv6_addr_copy(&sin->sin6_addr,
-                                 (struct in6_addr *)(nh + serr->addr_offset));
+                       const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
+                                                                 struct ipv6hdr, daddr);
+                       sin->sin6_addr = ip6h->daddr;
                        if (np->sndflow)
-                               sin->sin6_flowinfo =
-                                       (*(__be32 *)(nh + serr->addr_offset - 24) &
-                                        IPV6_FLOWINFO_MASK);
-                       if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-                               sin->sin6_scope_id = IP6CB(skb)->iif;
+                               sin->sin6_flowinfo = ip6_flowinfo(ip6h);
+                       sin->sin6_scope_id =
+                               ipv6_iface_scope_id(&sin->sin6_addr,
+                                                   IP6CB(skb)->iif);
                } else {
                        ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
                                               &sin->sin6_addr);
+                       sin->sin6_scope_id = 0;
                }
        }
 
@@ -344,18 +377,19 @@ int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
        if (serr->ee.ee_origin != SO_EE_ORIGIN_LOCAL) {
                sin->sin6_family = AF_INET6;
                sin->sin6_flowinfo = 0;
-               sin->sin6_scope_id = 0;
                if (skb->protocol == htons(ETH_P_IPV6)) {
-                       ipv6_addr_copy(&sin->sin6_addr, &ipv6_hdr(skb)->saddr);
+                       sin->sin6_addr = ipv6_hdr(skb)->saddr;
                        if (np->rxopt.all)
-                               datagram_recv_ctl(sk, msg, skb);
-                       if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-                               sin->sin6_scope_id = IP6CB(skb)->iif;
+                               ip6_datagram_recv_ctl(sk, msg, skb);
+                       sin->sin6_scope_id =
+                               ipv6_iface_scope_id(&sin->sin6_addr,
+                                                   IP6CB(skb)->iif);
                } else {
                        struct inet_sock *inet = inet_sk(sk);
 
                        ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
                                               &sin->sin6_addr);
+                       sin->sin6_scope_id = 0;
                        if (inet->cmsg_flags)
                                ip_cmsg_recv(msg, skb);
                }
@@ -384,10 +418,60 @@ out_free_skb:
 out:
        return err;
 }
+EXPORT_SYMBOL_GPL(ipv6_recv_error);
 
+/*
+ *     Handle IPV6_RECVPATHMTU
+ */
+int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len)
+{
+       struct ipv6_pinfo *np = inet6_sk(sk);
+       struct sk_buff *skb;
+       struct sockaddr_in6 *sin;
+       struct ip6_mtuinfo mtu_info;
+       int err;
+       int copied;
 
+       err = -EAGAIN;
+       skb = xchg(&np->rxpmtu, NULL);
+       if (skb == NULL)
+               goto out;
 
-int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
+       copied = skb->len;
+       if (copied > len) {
+               msg->msg_flags |= MSG_TRUNC;
+               copied = len;
+       }
+       err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
+       if (err)
+               goto out_free_skb;
+
+       sock_recv_timestamp(msg, sk, skb);
+
+       memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
+
+       sin = (struct sockaddr_in6 *)msg->msg_name;
+       if (sin) {
+               sin->sin6_family = AF_INET6;
+               sin->sin6_flowinfo = 0;
+               sin->sin6_port = 0;
+               sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
+               sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
+       }
+
+       put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
+
+       err = copied;
+
+out_free_skb:
+       kfree_skb(skb);
+out:
+       return err;
+}
+
+
+int ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
+                         struct sk_buff *skb)
 {
        struct ipv6_pinfo *np = inet6_sk(sk);
        struct inet6_skb_parm *opt = IP6CB(skb);
@@ -397,7 +481,7 @@ int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
                struct in6_pktinfo src_info;
 
                src_info.ipi6_ifindex = opt->iif;
-               ipv6_addr_copy(&src_info.ipi6_addr, &ipv6_hdr(skb)->daddr);
+               src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
                put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
        }
 
@@ -407,13 +491,14 @@ int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
        }
 
        if (np->rxopt.bits.rxtclass) {
-               int tclass = (ntohl(*(__be32 *)ipv6_hdr(skb)) >> 20) & 0xff;
+               int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
                put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
        }
 
-       if (np->rxopt.bits.rxflow && (*(__be32 *)nh & IPV6_FLOWINFO_MASK)) {
-               __be32 flowinfo = *(__be32 *)nh & IPV6_FLOWINFO_MASK;
-               put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
+       if (np->rxopt.bits.rxflow) {
+               __be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
+               if (flowinfo)
+                       put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
        }
 
        /* HbH is allowed only once */
@@ -437,10 +522,10 @@ int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
                u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 
                while (off <= opt->lastopt) {
-                       unsigned len;
+                       unsigned int len;
                        u8 *ptr = nh + off;
 
-                       switch(nexthdr) {
+                       switch (nexthdr) {
                        case IPPROTO_DSTOPTS:
                                nexthdr = ptr[0];
                                len = (ptr[1] + 1) << 3;
@@ -472,7 +557,7 @@ int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
                struct in6_pktinfo src_info;
 
                src_info.ipi6_ifindex = opt->iif;
-               ipv6_addr_copy(&src_info.ipi6_addr, &ipv6_hdr(skb)->daddr);
+               src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
                put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
        }
        if (np->rxopt.bits.rxohlim) {
@@ -495,13 +580,35 @@ int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
                u8 *ptr = nh + opt->dst1;
                put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
        }
+       if (np->rxopt.bits.rxorigdstaddr) {
+               struct sockaddr_in6 sin6;
+               __be16 *ports = (__be16 *) skb_transport_header(skb);
+
+               if (skb_transport_offset(skb) + 4 <= skb->len) {
+                       /* All current transport protocols have the port numbers in the
+                        * first four bytes of the transport header and this function is
+                        * written with this assumption in mind.
+                        */
+
+                       sin6.sin6_family = AF_INET6;
+                       sin6.sin6_addr = ipv6_hdr(skb)->daddr;
+                       sin6.sin6_port = ports[1];
+                       sin6.sin6_flowinfo = 0;
+                       sin6.sin6_scope_id =
+                               ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
+                                                   opt->iif);
+
+                       put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
+               }
+       }
        return 0;
 }
+EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
 
-int datagram_send_ctl(struct net *net,
-                     struct msghdr *msg, struct flowi *fl,
-                     struct ipv6_txoptions *opt,
-                     int *hlimit, int *tclass)
+int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
+                         struct msghdr *msg, struct flowi6 *fl6,
+                         struct ipv6_txoptions *opt,
+                         int *hlimit, int *tclass, int *dontfrag)
 {
        struct in6_pktinfo *src_info;
        struct cmsghdr *cmsg;
@@ -535,16 +642,17 @@ int datagram_send_ctl(struct net *net,
                        src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
 
                        if (src_info->ipi6_ifindex) {
-                               if (fl->oif && src_info->ipi6_ifindex != fl->oif)
+                               if (fl6->flowi6_oif &&
+                                   src_info->ipi6_ifindex != fl6->flowi6_oif)
                                        return -EINVAL;
-                               fl->oif = src_info->ipi6_ifindex;
+                               fl6->flowi6_oif = src_info->ipi6_ifindex;
                        }
 
                        addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
 
                        rcu_read_lock();
-                       if (fl->oif) {
-                               dev = dev_get_by_index_rcu(net, fl->oif);
+                       if (fl6->flowi6_oif) {
+                               dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
                                if (!dev) {
                                        rcu_read_unlock();
                                        return -ENODEV;
@@ -556,11 +664,12 @@ int datagram_send_ctl(struct net *net,
 
                        if (addr_type != IPV6_ADDR_ANY) {
                                int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
-                               if (!ipv6_chk_addr(net, &src_info->ipi6_addr,
+                               if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) &&
+                                   !ipv6_chk_addr(net, &src_info->ipi6_addr,
                                                   strict ? dev : NULL, 0))
                                        err = -EINVAL;
                                else
-                                       ipv6_addr_copy(&fl->fl6_src, &src_info->ipi6_addr);
+                                       fl6->saddr = src_info->ipi6_addr;
                        }
 
                        rcu_read_unlock();
@@ -577,13 +686,13 @@ int datagram_send_ctl(struct net *net,
                                goto exit_f;
                        }
 
-                       if (fl->fl6_flowlabel&IPV6_FLOWINFO_MASK) {
-                               if ((fl->fl6_flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
+                       if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
+                               if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
                                        err = -EINVAL;
                                        goto exit_f;
                                }
                        }
-                       fl->fl6_flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
+                       fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
                        break;
 
                case IPV6_2292HOPOPTS:
@@ -599,7 +708,7 @@ int datagram_send_ctl(struct net *net,
                                err = -EINVAL;
                                goto exit_f;
                        }
-                       if (!capable(CAP_NET_RAW)) {
+                       if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
                                err = -EPERM;
                                goto exit_f;
                        }
@@ -619,7 +728,7 @@ int datagram_send_ctl(struct net *net,
                                err = -EINVAL;
                                goto exit_f;
                        }
-                       if (!capable(CAP_NET_RAW)) {
+                       if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
                                err = -EPERM;
                                goto exit_f;
                        }
@@ -644,7 +753,7 @@ int datagram_send_ctl(struct net *net,
                                err = -EINVAL;
                                goto exit_f;
                        }
-                       if (!capable(CAP_NET_RAW)) {
+                       if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
                                err = -EPERM;
                                goto exit_f;
                        }
@@ -667,7 +776,7 @@ int datagram_send_ctl(struct net *net,
                        rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
 
                        switch (rthdr->type) {
-#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+#if IS_ENABLED(CONFIG_IPV6_MIP6)
                        case IPV6_SRCRT_TYPE_2:
                                if (rthdr->hdrlen != 2 ||
                                    rthdr->segments_left != 1) {
@@ -728,9 +837,8 @@ int datagram_send_ctl(struct net *net,
                        int tc;
 
                        err = -EINVAL;
-                       if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
+                       if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
                                goto exit_f;
-                       }
 
                        tc = *(int *)CMSG_DATA(cmsg);
                        if (tc < -1 || tc > 0xff)
@@ -741,6 +849,24 @@ int datagram_send_ctl(struct net *net,
 
                        break;
                    }
+
+               case IPV6_DONTFRAG:
+                   {
+                       int df;
+
+                       err = -EINVAL;
+                       if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
+                               goto exit_f;
+
+                       df = *(int *)CMSG_DATA(cmsg);
+                       if (df < 0 || df > 1)
+                               goto exit_f;
+
+                       err = 0;
+                       *dontfrag = df;
+
+                       break;
+                   }
                default:
                        LIMIT_NETDEBUG(KERN_DEBUG "invalid cmsg type: %d\n",
                                       cmsg->cmsg_type);
@@ -752,3 +878,4 @@ int datagram_send_ctl(struct net *net,
 exit_f:
        return err;
 }
+EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);