X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-netdev.c;h=a37933d5a2b5c720fe2cb07994bca3df0d49d9d8;hb=c68cc3ef3fa0ee3ade3c86035e5c3d7eb066a605;hp=a80b1b062df4ed5168b8ab089670f75c54cd66ac;hpb=4afba28d559511a5ee03e005b65be8323bf26ad3;p=sliver-openvswitch.git diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index a80b1b062..a37933d5a 100644 --- a/lib/dpif-netdev.c +++ b/lib/dpif-netdev.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc. + * Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,7 +17,6 @@ #include #include "dpif.h" -#include #include #include #include @@ -41,6 +40,7 @@ #include "hmap.h" #include "list.h" #include "netdev.h" +#include "netdev-vport.h" #include "netlink.h" #include "odp-util.h" #include "ofp-print.h" @@ -106,6 +106,7 @@ struct dp_netdev_port { int port_no; /* Index into dp_netdev's 'ports'. */ struct list node; /* Element in dp_netdev's 'port_list'. */ struct netdev *netdev; + struct netdev_saved_flags *sf; char *type; /* Port type as requested by user. */ }; @@ -151,7 +152,7 @@ static int dpif_netdev_open(const struct dpif_class *, const char *name, bool create, struct dpif **); static int dp_netdev_output_userspace(struct dp_netdev *, const struct ofpbuf *, int queue_no, const struct flow *, - uint64_t arg); + const struct nlattr *userdata); static void dp_netdev_execute_actions(struct dp_netdev *, struct ofpbuf *, struct flow *, const struct nlattr *actions, @@ -160,7 +161,7 @@ static void dp_netdev_execute_actions(struct dp_netdev *, static struct dpif_netdev * dpif_netdev_cast(const struct dpif *dpif) { - assert(dpif->dpif_class->open == dpif_netdev_open); + ovs_assert(dpif->dpif_class->open == dpif_netdev_open); return CONTAINER_OF(dpif, struct dpif_netdev, dpif); } @@ -181,6 +182,27 @@ dpif_netdev_enumerate(struct sset *all_dps) return 0; } +static bool +dpif_netdev_class_is_dummy(const struct dpif_class *class) +{ + return class != &dpif_netdev_class; +} + +static bool +dpif_netdev_class_is_planetlab(const struct dpif_class *class) +{ + return class == &dpif_planetlab_class; +} + +static const char * +dpif_netdev_port_open_type(const struct dpif_class *class, const char *type) +{ + return strcmp(type, "internal") ? type + : dpif_netdev_class_is_planetlab(class) ? "pltap" + : dpif_netdev_class_is_dummy(class) ? "dummy" + : "tap"; +} + static struct dpif * create_dpif_netdev(struct dp_netdev *dp) { @@ -202,7 +224,8 @@ choose_port(struct dp_netdev *dp, const char *name) { int port_no; - if (dp->class != &dpif_netdev_class) { + if (dp->class != &dpif_netdev_class && + dp->class != &dpif_planetlab_class) { const char *p; int start_no = 0; @@ -241,7 +264,6 @@ create_dp_netdev(const char *name, const struct dpif_class *class, struct dp_netdev **dpp) { struct dp_netdev *dp; - int port_no; int error; int i; @@ -255,8 +277,7 @@ create_dp_netdev(const char *name, const struct dpif_class *class, hmap_init(&dp->flow_table); list_init(&dp->port_list); - port_no = !strncmp(name, "br", 2) ? choose_port(dp, name) : OVSP_LOCAL; - error = do_add_port(dp, name, "internal", port_no); + error = do_add_port(dp, name, "internal", OVSP_LOCAL); if (error) { dp_netdev_free(dp); return error; @@ -283,7 +304,7 @@ dpif_netdev_open(const struct dpif_class *class, const char *name, if (error) { return error; } - assert(dp != NULL); + ovs_assert(dp != NULL); } } else { if (dp->class != class) { @@ -331,7 +352,7 @@ static void dpif_netdev_close(struct dpif *dpif) { struct dp_netdev *dp = get_dp_netdev(dpif); - assert(dp->open_cnt > 0); + ovs_assert(dp->open_cnt > 0); if (--dp->open_cnt == 0 && dp->destroyed) { shash_find_and_delete(&dp_netdevs, dp->name); dp_netdev_free(dp); @@ -362,6 +383,7 @@ static int do_add_port(struct dp_netdev *dp, const char *devname, const char *type, uint32_t port_no) { + struct netdev_saved_flags *sf; struct dp_netdev_port *port; struct netdev *netdev; const char *open_type; @@ -371,9 +393,7 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, /* XXX reject devices already in some dp_netdev. */ /* Open and validate network device. */ - open_type = (strcmp(type, "internal") ? type - : dp->class != &dpif_netdev_class ? "dummy" - : "tap"); + open_type = dpif_netdev_port_open_type(dp->class, type); error = netdev_open(devname, open_type, &netdev); if (error) { return error; @@ -382,14 +402,15 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, /* XXX reject non-Ethernet devices */ error = netdev_listen(netdev); - if (error) { + if (error + && !(error == EOPNOTSUPP && dpif_netdev_class_is_dummy(dp->class))) { VLOG_ERR("%s: cannot receive packets on this network device (%s)", devname, strerror(errno)); netdev_close(netdev); return error; } - error = netdev_turn_flags_on(netdev, NETDEV_PROMISC, false); + error = netdev_turn_flags_on(netdev, NETDEV_PROMISC, &sf); if (error) { netdev_close(netdev); return error; @@ -398,10 +419,11 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, port = xmalloc(sizeof *port); port->port_no = port_no; port->netdev = netdev; + port->sf = sf; port->type = xstrdup(type); error = netdev_get_mtu(netdev, &mtu); - if (!error) { + if (!error && mtu > max_mtu) { max_mtu = mtu; } @@ -427,11 +449,11 @@ dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, } port_no = *port_nop; } else { - port_no = choose_port(dp, netdev_get_name(netdev)); + port_no = choose_port(dp, netdev_vport_get_dpif_port(netdev)); } if (port_no >= 0) { *port_nop = port_no; - return do_add_port(dp, netdev_get_name(netdev), + return do_add_port(dp, netdev_vport_get_dpif_port(netdev), netdev_get_type(netdev), port_no); } return EFBIG; @@ -470,7 +492,7 @@ get_port_by_name(struct dp_netdev *dp, struct dp_netdev_port *port; LIST_FOR_EACH (port, node, &dp->port_list) { - if (!strcmp(netdev_get_name(port->netdev), devname)) { + if (!strcmp(netdev_vport_get_dpif_port(port->netdev), devname)) { *portp = port; return 0; } @@ -482,7 +504,6 @@ static int do_del_port(struct dp_netdev *dp, uint32_t port_no) { struct dp_netdev_port *port; - char *name; int error; error = get_port_by_number(dp, port_no, &port); @@ -494,11 +515,9 @@ do_del_port(struct dp_netdev *dp, uint32_t port_no) dp->ports[port->port_no] = NULL; dp->serial++; - name = xstrdup(netdev_get_name(port->netdev)); netdev_close(port->netdev); + netdev_restore_flags(port->sf); free(port->type); - - free(name); free(port); return 0; @@ -508,7 +527,7 @@ static void answer_port_query(const struct dp_netdev_port *port, struct dpif_port *dpif_port) { - dpif_port->name = xstrdup(netdev_get_name(port->netdev)); + dpif_port->name = xstrdup(netdev_vport_get_dpif_port(port->netdev)); dpif_port->type = xstrdup(port->type); dpif_port->port_no = port->port_no; } @@ -599,7 +618,7 @@ dpif_netdev_port_dump_next(const struct dpif *dpif, void *state_, struct dp_netdev_port *port = dp->ports[port_no]; if (port) { free(state->name); - state->name = xstrdup(netdev_get_name(port->netdev)); + state->name = xstrdup(netdev_vport_get_dpif_port(port->netdev)); dpif_port->name = state->name; dpif_port->type = port->type; dpif_port->port_no = port->port_no; @@ -666,7 +685,7 @@ static int dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len, struct flow *flow) { - if (odp_flow_key_to_flow(key, key_len, flow)) { + if (odp_flow_key_to_flow(key, key_len, flow) != ODP_FIT_PERFECT) { /* This should not happen: it indicates that odp_flow_key_from_flow() * and odp_flow_key_to_flow() disagree on the acceptable form of a * flow. Log the problem as an error, with enough details to enable @@ -925,7 +944,7 @@ dpif_netdev_execute(struct dpif *dpif, const struct dpif_execute *execute) ofpbuf_reserve(©, DP_NETDEV_HEADROOM); ofpbuf_put(©, execute->packet->data, execute->packet->size); - flow_extract(©, 0, NULL, -1, &key); + flow_extract(©, 0, 0, NULL, -1, &key); error = dpif_netdev_flow_from_nlattrs(execute->key, execute->key_len, &key); if (!error) { @@ -1023,7 +1042,7 @@ dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, if (packet->size < ETH_HEADER_LEN) { return; } - flow_extract(packet, 0, NULL, port->port_no, &key); + flow_extract(packet, 0, 0, NULL, port->port_no, &key); flow = dp_netdev_lookup_flow(dp, &key); if (flow) { dp_netdev_flow_used(flow, packet); @@ -1032,7 +1051,7 @@ dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, dp->n_hit++; } else { dp->n_missed++; - dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, 0); + dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, NULL); } } @@ -1058,7 +1077,8 @@ dpif_netdev_run(struct dpif *dpif) } else if (error != EAGAIN && error != EOPNOTSUPP) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); VLOG_ERR_RL(&rl, "error receiving data from %s: %s", - netdev_get_name(port->netdev), strerror(error)); + netdev_vport_get_dpif_port(port->netdev), + strerror(error)); } } ofpbuf_uninit(&packet); @@ -1096,37 +1116,51 @@ dp_netdev_output_port(struct dp_netdev *dp, struct ofpbuf *packet, static int dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, - int queue_no, const struct flow *flow, uint64_t arg) + int queue_no, const struct flow *flow, + const struct nlattr *userdata) { struct dp_netdev_queue *q = &dp->queues[queue_no]; - struct dp_netdev_upcall *u; - struct dpif_upcall *upcall; - struct ofpbuf *buf; - size_t key_len; - - if (q->head - q->tail >= MAX_QUEUE_LEN) { - dp->n_lost++; - return ENOBUFS; - } + if (q->head - q->tail < MAX_QUEUE_LEN) { + struct dp_netdev_upcall *u = &q->upcalls[q->head++ & QUEUE_MASK]; + struct dpif_upcall *upcall = &u->upcall; + struct ofpbuf *buf = &u->buf; + size_t buf_size; + + upcall->type = queue_no; + + /* Allocate buffer big enough for everything. */ + buf_size = ODPUTIL_FLOW_KEY_BYTES + 2 + packet->size; + if (userdata) { + buf_size += NLA_ALIGN(userdata->nla_len); + } + ofpbuf_init(buf, buf_size); - u = &q->upcalls[q->head++ & QUEUE_MASK]; + /* Put ODP flow. */ + odp_flow_key_from_flow(buf, flow, flow->in_port); + upcall->key = buf->data; + upcall->key_len = buf->size; - buf = &u->buf; - ofpbuf_init(buf, ODPUTIL_FLOW_KEY_BYTES + 2 + packet->size); - odp_flow_key_from_flow(buf, flow, flow->in_port); - key_len = buf->size; - ofpbuf_pull(buf, key_len); - ofpbuf_reserve(buf, 2); - ofpbuf_put(buf, packet->data, packet->size); + /* Put userdata. */ + if (userdata) { + upcall->userdata = ofpbuf_put(buf, userdata, + NLA_ALIGN(userdata->nla_len)); + } - upcall = &u->upcall; - upcall->type = queue_no; - upcall->packet = buf; - upcall->key = buf->base; - upcall->key_len = key_len; - upcall->userdata = arg; + /* Put packet. + * + * We adjust 'data' and 'size' in 'buf' so that only the packet itself + * is visible in 'upcall->packet'. The ODP flow and (if present) + * userdata become part of the headroom. */ + ofpbuf_put_zeros(buf, 2); + buf->data = ofpbuf_put(buf, packet->data, packet->size); + buf->size = packet->size; + upcall->packet = buf; - return 0; + return 0; + } else { + dp->n_lost++; + return ENOBUFS; + } } static void @@ -1168,11 +1202,9 @@ dp_netdev_action_userspace(struct dp_netdev *dp, struct ofpbuf *packet, struct flow *key, const struct nlattr *a) { - const struct nlattr *userdata_attr; - uint64_t userdata; + const struct nlattr *userdata; - userdata_attr = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA); - userdata = userdata_attr ? nl_attr_get_u64(userdata_attr) : 0; + userdata = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA); dp_netdev_output_userspace(dp, packet, DPIF_UC_ACTION, key, userdata); } @@ -1181,14 +1213,14 @@ execute_set_action(struct ofpbuf *packet, const struct nlattr *a) { enum ovs_key_attr type = nl_attr_type(a); const struct ovs_key_ipv4 *ipv4_key; + const struct ovs_key_ipv6 *ipv6_key; const struct ovs_key_tcp *tcp_key; const struct ovs_key_udp *udp_key; switch (type) { - case OVS_KEY_ATTR_TUN_ID: case OVS_KEY_ATTR_PRIORITY: - case OVS_KEY_ATTR_IPV6: - case OVS_KEY_ATTR_IPV4_TUNNEL: + case OVS_KEY_ATTR_SKB_MARK: + case OVS_KEY_ATTR_TUNNEL: /* not implemented */ break; @@ -1203,6 +1235,13 @@ execute_set_action(struct ofpbuf *packet, const struct nlattr *a) ipv4_key->ipv4_tos, ipv4_key->ipv4_ttl); break; + case OVS_KEY_ATTR_IPV6: + ipv6_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv6)); + packet_set_ipv6(packet, ipv6_key->ipv6_proto, ipv6_key->ipv6_src, + ipv6_key->ipv6_dst, ipv6_key->ipv6_tclass, + ipv6_key->ipv6_label, ipv6_key->ipv6_hlimit); + break; + case OVS_KEY_ATTR_TCP: tcp_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_tcp)); packet_set_tcp_port(packet, tcp_key->tcp_src, tcp_key->tcp_dst); @@ -1213,6 +1252,10 @@ execute_set_action(struct ofpbuf *packet, const struct nlattr *a) packet_set_udp_port(packet, udp_key->udp_src, udp_key->udp_dst); break; + case OVS_KEY_ATTR_MPLS: + set_mpls_lse(packet, nl_attr_get_be32(a)); + break; + case OVS_KEY_ATTR_UNSPEC: case OVS_KEY_ATTR_ENCAP: case OVS_KEY_ATTR_ETHERTYPE: @@ -1238,7 +1281,6 @@ dp_netdev_execute_actions(struct dp_netdev *dp, unsigned int left; NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) { - const struct ovs_action_push_vlan *vlan; int type = nl_attr_type(a); switch ((enum ovs_action_attr) type) { @@ -1250,15 +1292,26 @@ dp_netdev_execute_actions(struct dp_netdev *dp, dp_netdev_action_userspace(dp, packet, key, a); break; - case OVS_ACTION_ATTR_PUSH_VLAN: - vlan = nl_attr_get(a); + case OVS_ACTION_ATTR_PUSH_VLAN: { + const struct ovs_action_push_vlan *vlan = nl_attr_get(a); eth_push_vlan(packet, vlan->vlan_tci); break; + } case OVS_ACTION_ATTR_POP_VLAN: eth_pop_vlan(packet); break; + case OVS_ACTION_ATTR_PUSH_MPLS: { + const struct ovs_action_push_mpls *mpls = nl_attr_get(a); + push_mpls(packet, mpls->mpls_ethertype, mpls->mpls_lse); + break; + } + + case OVS_ACTION_ATTR_POP_MPLS: + pop_mpls(packet, nl_attr_get_be16(a)); + break; + case OVS_ACTION_ATTR_SET: execute_set_action(packet, nl_attr_get(a)); break; @@ -1274,40 +1327,49 @@ dp_netdev_execute_actions(struct dp_netdev *dp, } } +#define DPIF_NETDEV_CLASS_FUNCTIONS \ + dpif_netdev_enumerate, \ + dpif_netdev_port_open_type, \ + dpif_netdev_open, \ + dpif_netdev_close, \ + dpif_netdev_destroy, \ + dpif_netdev_run, \ + dpif_netdev_wait, \ + dpif_netdev_get_stats, \ + dpif_netdev_port_add, \ + dpif_netdev_port_del, \ + dpif_netdev_port_query_by_number, \ + dpif_netdev_port_query_by_name, \ + dpif_netdev_get_max_ports, \ + NULL, /* port_get_pid */ \ + dpif_netdev_port_dump_start, \ + dpif_netdev_port_dump_next, \ + dpif_netdev_port_dump_done, \ + dpif_netdev_port_poll, \ + dpif_netdev_port_poll_wait, \ + dpif_netdev_flow_get, \ + dpif_netdev_flow_put, \ + dpif_netdev_flow_del, \ + dpif_netdev_flow_flush, \ + dpif_netdev_flow_dump_start, \ + dpif_netdev_flow_dump_next, \ + dpif_netdev_flow_dump_done, \ + dpif_netdev_execute, \ + NULL, /* operate */ \ + dpif_netdev_recv_set, \ + dpif_netdev_queue_to_priority, \ + dpif_netdev_recv, \ + dpif_netdev_recv_wait, \ + dpif_netdev_recv_purge, \ + const struct dpif_class dpif_netdev_class = { "netdev", - dpif_netdev_enumerate, - dpif_netdev_open, - dpif_netdev_close, - dpif_netdev_destroy, - dpif_netdev_run, - dpif_netdev_wait, - dpif_netdev_get_stats, - dpif_netdev_port_add, - dpif_netdev_port_del, - dpif_netdev_port_query_by_number, - dpif_netdev_port_query_by_name, - dpif_netdev_get_max_ports, - NULL, /* port_get_pid */ - dpif_netdev_port_dump_start, - dpif_netdev_port_dump_next, - dpif_netdev_port_dump_done, - dpif_netdev_port_poll, - dpif_netdev_port_poll_wait, - dpif_netdev_flow_get, - dpif_netdev_flow_put, - dpif_netdev_flow_del, - dpif_netdev_flow_flush, - dpif_netdev_flow_dump_start, - dpif_netdev_flow_dump_next, - dpif_netdev_flow_dump_done, - dpif_netdev_execute, - NULL, /* operate */ - dpif_netdev_recv_set, - dpif_netdev_queue_to_priority, - dpif_netdev_recv, - dpif_netdev_recv_wait, - dpif_netdev_recv_purge, + DPIF_NETDEV_CLASS_FUNCTIONS +}; + +const struct dpif_class dpif_planetlab_class = { + "planetlab", + DPIF_NETDEV_CLASS_FUNCTIONS }; static void @@ -1340,3 +1402,4 @@ dpif_dummy_register(bool override) dpif_dummy_register__("dummy"); } +