dpif-netlink: Allow MRU packet attribute.

User space now may receive re-assembled IP fragments. The user space
netlink handler can now accept packets with the new OVS_PACKET_ATTR_MRU
attribute. This allows the kernel to assemble fragmented packets for the
duration of OpenFlow processing, then re-fragment at output time. Most
notably this occurs for packets that are sent through the connection
tracker.

Note that the MRU attribute is not exported at the OpenFlow layer. As
such, if packets are reassembled by conntrack and subsequently sent to
the controller, then OVS has no way to re-serialize the packets to their
original size.

Signed-off-by: Andy Zhou <[email protected]>
Acked-by: Jarno Rajahalme <[email protected]>
Acked-by: Ben Pfaff <[email protected]>

datapath/linux/compat/include/linux/openvswitch.h

@@ -187,6 +187,10 @@ enum ovs_packet_cmd {
  * %OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute, which is sent only if the
  * output port is actually a tunnel port. Contains the output tunnel key
  * extracted from the packet as nested %OVS_TUNNEL_KEY_ATTR_* attributes.
+ * @OVS_PACKET_ATTR_MRU: Present for an %OVS_PACKET_CMD_ACTION and
+ * %OVS_PACKET_ATTR_USERSPACE action specify the Maximum received fragment
+ * size.
+ *
  * These attributes follow the &struct ovs_header within the Generic Netlink
  * payload for %OVS_PACKET_* commands.
  */
@@ -202,6 +206,7 @@ enum ovs_packet_attr {
 	OVS_PACKET_ATTR_UNUSED2,
 	OVS_PACKET_ATTR_PROBE,      /* Packet operation is a feature probe,
 				       error logging should be suppressed. */
+	OVS_PACKET_ATTR_MRU,	    /* Maximum received IP fragment size. */
 	__OVS_PACKET_ATTR_MAX
 };
 

lib/dpif-netlink.c

@@ -1547,6 +1547,9 @@ dpif_netlink_encode_execute(int dp_ifindex, const struct dpif_execute *d_exec,
     if (d_exec->probe) {
         nl_msg_put_flag(buf, OVS_PACKET_ATTR_PROBE);
     }
+    if (d_exec->mtu) {
+        nl_msg_put_u16(buf, OVS_PACKET_ATTR_MRU, d_exec->mtu);
+    }
 }
 
 /* Executes, against 'dpif', up to the first 'n_ops' operations in 'ops'.
@@ -1965,6 +1968,7 @@ parse_odp_packet(const struct dpif_netlink *dpif, struct ofpbuf *buf,
         [OVS_PACKET_ATTR_USERDATA] = { .type = NL_A_UNSPEC, .optional = true },
         [OVS_PACKET_ATTR_EGRESS_TUN_KEY] = { .type = NL_A_NESTED, .optional = true },
         [OVS_PACKET_ATTR_ACTIONS] = { .type = NL_A_NESTED, .optional = true },
+        [OVS_PACKET_ATTR_MRU] = { .type = NL_A_U16, .optional = true }
     };
 
     struct ovs_header *ovs_header;
@@ -2002,6 +2006,7 @@ parse_odp_packet(const struct dpif_netlink *dpif, struct ofpbuf *buf,
     upcall->userdata = a[OVS_PACKET_ATTR_USERDATA];
     upcall->out_tun_key = a[OVS_PACKET_ATTR_EGRESS_TUN_KEY];
     upcall->actions = a[OVS_PACKET_ATTR_ACTIONS];
+    upcall->mru = a[OVS_PACKET_ATTR_MRU];
 
     /* Allow overwriting the netlink attribute header without reallocating. */
     dp_packet_use_stub(&upcall->packet,

lib/dpif.c

@@ -1126,6 +1126,7 @@ dpif_execute_helper_cb(void *aux_, struct dp_packet **packets, int cnt,
         execute.packet = packet;
         execute.needs_help = false;
         execute.probe = false;
+        execute.mtu = 0;
         aux->error = dpif_execute(aux->dpif, &execute);
         log_execute_message(aux->dpif, &execute, true, aux->error);
 
@@ -1693,6 +1694,7 @@ log_execute_message(struct dpif *dpif, const struct dpif_execute *execute,
             ds_put_format(&ds, " failed (%s)", ovs_strerror(error));
         }
         ds_put_format(&ds, " on packet %s", packet);
+        ds_put_format(&ds, " mtu %d", execute->mtu);
         vlog(THIS_MODULE, error ? VLL_WARN : VLL_DBG, "%s", ds_cstr(&ds));
         ds_destroy(&ds);
         free(packet);

lib/dpif.h

@@ -697,6 +697,8 @@ struct dpif_execute {
     size_t actions_len;             /* Length of 'actions' in bytes. */
     bool needs_help;
     bool probe;                     /* Suppress error messages. */
+    unsigned int mtu;               /* Maximum transmission unit to fragment.
+                                       0 if not a fragmented packet */
 
     /* Input, but possibly modified as a side effect of execution. */
     struct dp_packet *packet;          /* Packet to execute. */
@@ -780,6 +782,7 @@ struct dpif_upcall {
     struct nlattr *key;         /* Flow key. */
     size_t key_len;             /* Length of 'key' in bytes. */
     ovs_u128 ufid;              /* Unique flow identifier for 'key'. */
+    struct nlattr *mru;         /* Maximum receive unit. */
 
     /* DPIF_UC_ACTION only. */
     struct nlattr *userdata;    /* Argument to OVS_ACTION_ATTR_USERSPACE. */

ofproto/ofproto-dpif-upcall.c

@@ -183,6 +183,8 @@ struct upcall {
     unsigned pmd_id;               /* Datapath poll mode driver id. */
     const struct dp_packet *packet;   /* Packet associated with this upcall. */
     ofp_port_t in_port;            /* OpenFlow in port, or OFPP_NONE. */
+    uint16_t mru;                  /* If !0, Maximum receive unit of
+                                      fragmented IP packet */
 
     enum dpif_upcall_type type;    /* Datapath type of the upcall. */
     const struct nlattr *userdata; /* Userdata for DPIF_UC_ACTION Upcalls. */
@@ -329,6 +331,7 @@ static enum upcall_type classify_upcall(enum dpif_upcall_type type,
 static int upcall_receive(struct upcall *, const struct dpif_backer *,
                           const struct dp_packet *packet, enum dpif_upcall_type,
                           const struct nlattr *userdata, const struct flow *,
+                          const unsigned int mru,
                           const ovs_u128 *ufid, const unsigned pmd_id);
 static void upcall_uninit(struct upcall *);
 
@@ -720,6 +723,7 @@ recv_upcalls(struct handler *handler)
         struct dpif_upcall *dupcall = &dupcalls[n_upcalls];
         struct upcall *upcall = &upcalls[n_upcalls];
         struct flow *flow = &flows[n_upcalls];
+        unsigned int mru;
         int error;
 
         ofpbuf_use_stub(recv_buf, recv_stubs[n_upcalls],
@@ -734,8 +738,14 @@ recv_upcalls(struct handler *handler)
             goto free_dupcall;
         }
 
+        if (dupcall->mru) {
+            mru = nl_attr_get_u16(dupcall->mru);
+        } else {
+            mru = 0;
+        }
+
         error = upcall_receive(upcall, udpif->backer, &dupcall->packet,
-                               dupcall->type, dupcall->userdata, flow,
+                               dupcall->type, dupcall->userdata, flow, mru,
                                &dupcall->ufid, PMD_ID_NULL);
         if (error) {
             if (error == ENODEV) {
@@ -981,6 +991,7 @@ static int
 upcall_receive(struct upcall *upcall, const struct dpif_backer *backer,
                const struct dp_packet *packet, enum dpif_upcall_type type,
                const struct nlattr *userdata, const struct flow *flow,
+               const unsigned int mru,
                const ovs_u128 *ufid, const unsigned pmd_id)
 {
     int error;
@@ -1010,6 +1021,7 @@ upcall_receive(struct upcall *upcall, const struct dpif_backer *backer,
     upcall->ukey = NULL;
     upcall->key = NULL;
     upcall->key_len = 0;
+    upcall->mru = mru;
 
     upcall->out_tun_key = NULL;
     upcall->actions = NULL;
@@ -1137,7 +1149,7 @@ upcall_cb(const struct dp_packet *packet, const struct flow *flow, ovs_u128 *ufi
     atomic_read_relaxed(&udpif->flow_limit, &flow_limit);
 
     error = upcall_receive(&upcall, udpif->backer, packet, type, userdata,
-                           flow, ufid, pmd_id);
+                           flow, 0, ufid, pmd_id);
     if (error) {
         return error;
     }
@@ -1359,6 +1371,7 @@ handle_upcalls(struct udpif *udpif, struct upcall *upcalls,
             op->dop.u.execute.actions_len = upcall->odp_actions.size;
             op->dop.u.execute.needs_help = (upcall->xout.slow & SLOW_ACTION) != 0;
             op->dop.u.execute.probe = false;
+            op->dop.u.execute.mtu = upcall->mru;
         }
     }
 

ofproto/ofproto-dpif.c

@@ -1122,6 +1122,7 @@ check_variable_length_userdata(struct dpif_backer *backer)
     execute.packet = &packet;
     execute.needs_help = false;
     execute.probe = true;
+    execute.mtu = 0;
 
     error = dpif_execute(backer->dpif, &execute);
 
@@ -1220,6 +1221,7 @@ check_masked_set_action(struct dpif_backer *backer)
     execute.packet = &packet;
     execute.needs_help = false;
     execute.probe = true;
+    execute.mtu = 0;
 
     error = dpif_execute(backer->dpif, &execute);
 
@@ -3737,6 +3739,7 @@ ofproto_dpif_execute_actions__(struct ofproto_dpif *ofproto,
     execute.packet = packet;
     execute.needs_help = (xout.slow & SLOW_ACTION) != 0;
     execute.probe = false;
+    execute.mtu = 0;
 
     /* Fix up in_port. */
     in_port = flow->in_port.ofp_port;