X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-netdev.c;h=193ffe4336930e900e2ac61b11c94401270f71b4;hb=85b20fd6ee585f462e012fbcc7f966a81edab2ed;hp=a6c06fe9de91e4d7f0e61979a3f36c384708c2ed;hpb=e0edde6fee279cdbbf3c179f5f50adaf0c7c7f1e;p=sliver-openvswitch.git diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index a6c06fe9d..193ffe433 100644 --- a/lib/dpif-netdev.c +++ b/lib/dpif-netdev.c @@ -70,8 +70,13 @@ enum { MAX_QUEUE_LEN = 128 }; /* Maximum number of packets per queue. */ enum { QUEUE_MASK = MAX_QUEUE_LEN - 1 }; BUILD_ASSERT_DECL(IS_POW2(MAX_QUEUE_LEN)); +struct dp_netdev_upcall { + struct dpif_upcall upcall; /* Queued upcall information. */ + struct ofpbuf buf; /* ofpbuf instance for upcall.packet. */ +}; + struct dp_netdev_queue { - struct dpif_upcall *upcalls[MAX_QUEUE_LEN]; + struct dp_netdev_upcall upcalls[MAX_QUEUE_LEN]; unsigned int head, tail; }; @@ -133,15 +138,15 @@ static struct shash dp_netdevs = SHASH_INITIALIZER(&dp_netdevs); /* Maximum port MTU seen so far. */ static int max_mtu = ETH_PAYLOAD_MAX; -static int get_port_by_number(struct dp_netdev *, uint16_t port_no, +static int get_port_by_number(struct dp_netdev *, uint32_t port_no, struct dp_netdev_port **portp); static int get_port_by_name(struct dp_netdev *, const char *devname, struct dp_netdev_port **portp); static void dp_netdev_free(struct dp_netdev *); static void dp_netdev_flow_flush(struct dp_netdev *); static int do_add_port(struct dp_netdev *, const char *devname, - const char *type, uint16_t port_no); -static int do_del_port(struct dp_netdev *, uint16_t port_no); + const char *type, uint32_t port_no); +static int do_del_port(struct dp_netdev *, uint32_t port_no); 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 *, @@ -165,6 +170,33 @@ get_dp_netdev(const struct dpif *dpif) return dpif_netdev_cast(dpif)->dp; } +static int +dpif_netdev_enumerate(struct sset *all_dps) +{ + struct shash_node *node; + + SHASH_FOR_EACH(node, &dp_netdevs) { + sset_add(all_dps, node->name); + } + return 0; +} + +static const char * +dpif_netdev_port_open_type(const struct dpif_class *class, const char *type) +{ + return strcmp(type, "internal") ? type + : class != &dpif_netdev_class ? "dummy" + : "tap"; +} + +static const char * +dpif_planetlab_port_open_type(const struct dpif_class *class, const char *type) +{ + return strcmp(type, "internal") ? type + : class != &dpif_planetlab_class ? "dummy" + : "pltap"; +} + static struct dpif * create_dpif_netdev(struct dp_netdev *dp) { @@ -181,6 +213,46 @@ create_dpif_netdev(struct dp_netdev *dp) return &dpif->dpif; } +static int +choose_port(struct dp_netdev *dp, const char *name) +{ + int port_no; + + if (dp->class != &dpif_netdev_class && + dp->class != &dpif_planetlab_class) { + const char *p; + int start_no = 0; + + /* If the port name begins with "br", start the number search at + * 100 to make writing tests easier. */ + if (!strncmp(name, "br", 2)) { + start_no = 100; + } + + /* If the port name contains a number, try to assign that port number. + * This can make writing unit tests easier because port numbers are + * predictable. */ + for (p = name; *p != '\0'; p++) { + if (isdigit((unsigned char) *p)) { + port_no = start_no + strtol(p, NULL, 10); + if (port_no > 0 && port_no < MAX_PORTS + && !dp->ports[port_no]) { + return port_no; + } + break; + } + } + } + + for (port_no = 1; port_no < MAX_PORTS; port_no++) { + if (!dp->ports[port_no]) { + return port_no; + } + } + + return -1; +} + static int create_dp_netdev(const char *name, const struct dpif_class *class, struct dp_netdev **dpp) @@ -198,6 +270,7 @@ create_dp_netdev(const char *name, const struct dpif_class *class, } hmap_init(&dp->flow_table); list_init(&dp->port_list); + error = do_add_port(dp, name, "internal", OVSP_LOCAL); if (error) { dp_netdev_free(dp); @@ -248,10 +321,8 @@ dp_netdev_purge_queues(struct dp_netdev *dp) struct dp_netdev_queue *q = &dp->queues[i]; while (q->tail != q->head) { - struct dpif_upcall *upcall = q->upcalls[q->tail++ & QUEUE_MASK]; - - ofpbuf_delete(upcall->packet); - free(upcall); + struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK]; + ofpbuf_uninit(&u->buf); } } } @@ -304,7 +375,7 @@ dpif_netdev_get_stats(const struct dpif *dpif, struct dpif_dp_stats *stats) static int do_add_port(struct dp_netdev *dp, const char *devname, const char *type, - uint16_t port_no) + uint32_t port_no) { struct dp_netdev_port *port; struct netdev *netdev; @@ -315,9 +386,7 @@ 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; @@ -357,46 +426,22 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, } static int -choose_port(struct dpif *dpif, struct netdev *netdev) +dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, + uint32_t *port_nop) { struct dp_netdev *dp = get_dp_netdev(dpif); int port_no; - if (dpif->dpif_class != &dpif_netdev_class) { - /* If the port name contains a number, try to assign that port number. - * This can make writing unit tests easier because port numbers are - * predictable. */ - const char *p; - - for (p = netdev_get_name(netdev); *p != '\0'; p++) { - if (isdigit((unsigned char) *p)) { - port_no = strtol(p, NULL, 10); - if (port_no > 0 && port_no < MAX_PORTS - && !dp->ports[port_no]) { - return port_no; - } - break; - } - } - } - - for (port_no = 0; port_no < MAX_PORTS; port_no++) { - if (!dp->ports[port_no]) { - return port_no; + if (*port_nop != UINT32_MAX) { + if (*port_nop >= MAX_PORTS) { + return EFBIG; + } else if (dp->ports[*port_nop]) { + return EBUSY; } + port_no = *port_nop; + } else { + port_no = choose_port(dp, netdev_get_name(netdev)); } - - return -1; -} - -static int -dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, - uint16_t *port_nop) -{ - struct dp_netdev *dp = get_dp_netdev(dpif); - int port_no; - - port_no = choose_port(dpif, netdev); if (port_no >= 0) { *port_nop = port_no; return do_add_port(dp, netdev_get_name(netdev), @@ -406,21 +451,21 @@ dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, } static int -dpif_netdev_port_del(struct dpif *dpif, uint16_t port_no) +dpif_netdev_port_del(struct dpif *dpif, uint32_t port_no) { struct dp_netdev *dp = get_dp_netdev(dpif); return port_no == OVSP_LOCAL ? EINVAL : do_del_port(dp, port_no); } static bool -is_valid_port_number(uint16_t port_no) +is_valid_port_number(uint32_t port_no) { return port_no < MAX_PORTS; } static int get_port_by_number(struct dp_netdev *dp, - uint16_t port_no, struct dp_netdev_port **portp) + uint32_t port_no, struct dp_netdev_port **portp) { if (!is_valid_port_number(port_no)) { *portp = NULL; @@ -447,7 +492,7 @@ get_port_by_name(struct dp_netdev *dp, } static int -do_del_port(struct dp_netdev *dp, uint16_t port_no) +do_del_port(struct dp_netdev *dp, uint32_t port_no) { struct dp_netdev_port *port; char *name; @@ -482,7 +527,7 @@ answer_port_query(const struct dp_netdev_port *port, } static int -dpif_netdev_port_query_by_number(const struct dpif *dpif, uint16_t port_no, +dpif_netdev_port_query_by_number(const struct dpif *dpif, uint32_t port_no, struct dpif_port *dpif_port) { struct dp_netdev *dp = get_dp_netdev(dpif); @@ -490,7 +535,7 @@ dpif_netdev_port_query_by_number(const struct dpif *dpif, uint16_t port_no, int error; error = get_port_by_number(dp, port_no, &port); - if (!error) { + if (!error && dpif_port) { answer_port_query(port, dpif_port); } return error; @@ -505,7 +550,7 @@ dpif_netdev_port_query_by_name(const struct dpif *dpif, const char *devname, int error; error = get_port_by_name(dp, devname, &port); - if (!error) { + if (!error && dpif_port) { answer_port_query(port, dpif_port); } return error; @@ -634,7 +679,7 @@ 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 @@ -702,10 +747,9 @@ set_flow_actions(struct dp_netdev_flow *flow, } static int -add_flow(struct dpif *dpif, const struct flow *key, - const struct nlattr *actions, size_t actions_len) +dp_netdev_flow_add(struct dp_netdev *dp, const struct flow *key, + const struct nlattr *actions, size_t actions_len) { - struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; int error; @@ -751,7 +795,8 @@ dpif_netdev_flow_put(struct dpif *dpif, const struct dpif_flow_put *put) if (put->stats) { memset(put->stats, 0, sizeof *put->stats); } - return add_flow(dpif, &key, put->actions, put->actions_len); + return dp_netdev_flow_add(dp, &key, put->actions, + put->actions_len); } else { return EFBIG; } @@ -843,7 +888,7 @@ dpif_netdev_flow_dump_next(const struct dpif *dpif, void *state_, struct ofpbuf buf; ofpbuf_use_stack(&buf, &state->keybuf, sizeof state->keybuf); - odp_flow_key_from_flow(&buf, &flow->key); + odp_flow_key_from_flow(&buf, &flow->key, flow->key.in_port); *key = buf.data; *key_len = buf.size; @@ -893,7 +938,7 @@ 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, 0, -1, &key); + flow_extract(©, 0, 0, NULL, -1, &key); error = dpif_netdev_flow_from_nlattrs(execute->key, execute->key_len, &key); if (!error) { @@ -940,12 +985,13 @@ dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall, { struct dp_netdev_queue *q = find_nonempty_queue(dpif); if (q) { - struct dpif_upcall *u = q->upcalls[q->tail++ & QUEUE_MASK]; - *upcall = *u; - free(u); + struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK]; + + *upcall = u->upcall; + upcall->packet = buf; ofpbuf_uninit(buf); - *buf = *u->packet; + *buf = u->buf; return 0; } else { @@ -972,13 +1018,12 @@ dpif_netdev_recv_purge(struct dpif *dpif) } static void -dp_netdev_flow_used(struct dp_netdev_flow *flow, struct flow *key, - const struct ofpbuf *packet) +dp_netdev_flow_used(struct dp_netdev_flow *flow, const struct ofpbuf *packet) { flow->used = time_msec(); flow->packet_count++; flow->byte_count += packet->size; - flow->tcp_flags |= packet_get_tcp_flags(packet, key); + flow->tcp_flags |= packet_get_tcp_flags(packet, &flow->key); } static void @@ -991,10 +1036,10 @@ dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, if (packet->size < ETH_HEADER_LEN) { return; } - flow_extract(packet, 0, 0, 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, &key, packet); + dp_netdev_flow_used(flow, packet); dp_netdev_execute_actions(dp, packet, &key, flow->actions, flow->actions_len); dp->n_hit++; @@ -1054,7 +1099,7 @@ dp_netdev_set_dl(struct ofpbuf *packet, const struct ovs_key_ethernet *eth_key) static void dp_netdev_output_port(struct dp_netdev *dp, struct ofpbuf *packet, - uint16_t out_port) + uint32_t out_port) { struct dp_netdev_port *p = dp->ports[out_port]; if (p) { @@ -1067,6 +1112,7 @@ dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, int queue_no, const struct flow *flow, uint64_t arg) { 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; @@ -1076,22 +1122,23 @@ dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, return ENOBUFS; } - buf = ofpbuf_new(ODPUTIL_FLOW_KEY_BYTES + 2 + packet->size); - odp_flow_key_from_flow(buf, flow); + u = &q->upcalls[q->head++ & QUEUE_MASK]; + + 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); - upcall = xzalloc(sizeof *upcall); + upcall = &u->upcall; upcall->type = queue_no; upcall->packet = buf; upcall->key = buf->base; upcall->key_len = key_len; upcall->userdata = arg; - q->upcalls[q->head++ & QUEUE_MASK] = upcall; - return 0; } @@ -1147,13 +1194,15 @@ 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_SKB_MARK: + case OVS_KEY_ATTR_IPV4_TUNNEL: /* not implemented */ break; @@ -1168,6 +1217,13 @@ 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); @@ -1239,40 +1295,49 @@ dp_netdev_execute_actions(struct dp_netdev *dp, } } +#define DPIF_NETDEV_CLASS_FUNCTIONS(PORT_OPEN_TYPE) \ + dpif_netdev_enumerate, \ + 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", - NULL, /* 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(dpif_netdev_port_open_type) +}; + +const struct dpif_class dpif_planetlab_class = { + "planetlab", + DPIF_NETDEV_CLASS_FUNCTIONS(dpif_planetlab_port_open_type) }; static void @@ -1305,3 +1370,4 @@ dpif_dummy_register(bool override) dpif_dummy_register__("dummy"); } +