X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-netdev.c;h=9f7bb6edad99713d42167de59292be19baf37e63;hb=7a97619ae4484604c15b846ce7268094e048729a;hp=82d7d74132e53b627e1764135dd65174ef587f24;hpb=db5ce51427906ed54aa3e8e1dffc0ba6d908aa8c;p=sliver-openvswitch.git diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index 82d7d7413..9f7bb6eda 100644 --- a/lib/dpif-netdev.c +++ b/lib/dpif-netdev.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010, 2011 Nicira Networks. + * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -36,6 +36,7 @@ #include "dpif.h" #include "dpif-provider.h" #include "dummy.h" +#include "dynamic-string.h" #include "flow.h" #include "hmap.h" #include "list.h" @@ -46,7 +47,9 @@ #include "ofpbuf.h" #include "packets.h" #include "poll-loop.h" +#include "random.h" #include "shash.h" +#include "sset.h" #include "timeval.h" #include "util.h" #include "vlog.h" @@ -54,8 +57,6 @@ VLOG_DEFINE_THIS_MODULE(dpif_netdev); /* Configuration parameters. */ -enum { N_QUEUES = 2 }; /* Number of queues for dpif_recv(). */ -enum { MAX_QUEUE_LEN = 100 }; /* Maximum number of packets per queue. */ enum { MAX_PORTS = 256 }; /* Maximum number of ports. */ enum { MAX_FLOWS = 65536 }; /* Maximum number of flows in flow table. */ @@ -63,6 +64,22 @@ enum { MAX_FLOWS = 65536 }; /* Maximum number of flows in flow table. */ * headers to be aligned on a 4-byte boundary. */ enum { DP_NETDEV_HEADROOM = 2 + VLAN_HEADER_LEN }; +/* Queues. */ +enum { N_QUEUES = 2 }; /* Number of queues for dpif_recv(). */ +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 dp_netdev_upcall upcalls[MAX_QUEUE_LEN]; + unsigned int head, tail; +}; + /* Datapath based on the network device interface from netdev.h. */ struct dp_netdev { const struct dpif_class *class; @@ -70,19 +87,15 @@ struct dp_netdev { int open_cnt; bool destroyed; - bool drop_frags; /* Drop all IP fragments, if true. */ - struct list queues[N_QUEUES]; /* Contain ofpbufs queued for dpif_recv(). */ - size_t queue_len[N_QUEUES]; /* Number of packets in each queue. */ + struct dp_netdev_queue queues[N_QUEUES]; struct hmap flow_table; /* Flow table. */ /* Statistics. */ - long long int n_frags; /* Number of dropped IP fragments. */ long long int n_hit; /* Number of flow table matches. */ long long int n_missed; /* Number of flow table misses. */ long long int n_lost; /* Number of misses not passed to client. */ /* Ports. */ - int n_ports; struct dp_netdev_port *ports[MAX_PORTS]; struct list port_list; unsigned int serial; @@ -93,7 +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; - bool internal; /* Internal port? */ + char *type; /* Port type as requested by user. */ }; /* A flow in dp_netdev's 'flow_table'. */ @@ -102,10 +115,10 @@ struct dp_netdev_flow { struct flow key; /* Statistics. */ - struct timespec used; /* Last used time. */ + long long int used; /* Last used time, in monotonic msecs. */ long long int packet_count; /* Number of packets matched. */ long long int byte_count; /* Number of bytes matched. */ - uint16_t tcp_ctl; /* Bitwise-OR of seen tcp_ctl values. */ + uint8_t tcp_flags; /* Bitwise-OR of seen tcp_flags values. */ /* Actions. */ struct nlattr *actions; @@ -116,7 +129,6 @@ struct dp_netdev_flow { struct dpif_netdev { struct dpif dpif; struct dp_netdev *dp; - int listen_mask; unsigned int dp_serial; }; @@ -137,14 +149,13 @@ static int do_add_port(struct dp_netdev *, const char *devname, static int do_del_port(struct dp_netdev *, uint16_t port_no); static int dpif_netdev_open(const struct dpif_class *, const char *name, bool create, struct dpif **); -static int dp_netdev_output_control(struct dp_netdev *, const struct ofpbuf *, - int queue_no, int port_no, uint64_t arg); -static int dp_netdev_execute_actions(struct dp_netdev *, - struct ofpbuf *, struct flow *, - const struct nlattr *actions, - size_t actions_len); - -static struct dpif_class dpif_dummy_class; +static int dp_netdev_output_userspace(struct dp_netdev *, const struct ofpbuf *, + int queue_no, const struct flow *, + uint64_t arg); +static void dp_netdev_execute_actions(struct dp_netdev *, + struct ofpbuf *, struct flow *, + const struct nlattr *actions, + size_t actions_len); static struct dpif_netdev * dpif_netdev_cast(const struct dpif *dpif) @@ -159,6 +170,17 @@ 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 struct dpif * create_dpif_netdev(struct dp_netdev *dp) { @@ -170,7 +192,6 @@ create_dpif_netdev(struct dp_netdev *dp) dpif = xmalloc(sizeof *dpif); dpif_init(&dpif->dpif, dp->class, dp->name, netflow_id >> 8, netflow_id); dpif->dp = dp; - dpif->listen_mask = 0; dpif->dp_serial = dp->serial; return &dpif->dpif; @@ -188,13 +209,12 @@ create_dp_netdev(const char *name, const struct dpif_class *class, dp->class = class; dp->name = xstrdup(name); dp->open_cnt = 0; - dp->drop_frags = false; for (i = 0; i < N_QUEUES; i++) { - list_init(&dp->queues[i]); + dp->queues[i].head = dp->queues[i].tail = 0; } hmap_init(&dp->flow_table); list_init(&dp->port_list); - error = do_add_port(dp, name, "internal", ODPP_LOCAL); + error = do_add_port(dp, name, "internal", OVSP_LOCAL); if (error) { dp_netdev_free(dp); return error; @@ -236,19 +256,30 @@ dpif_netdev_open(const struct dpif_class *class, const char *name, } static void -dp_netdev_free(struct dp_netdev *dp) +dp_netdev_purge_queues(struct dp_netdev *dp) { int i; + for (i = 0; i < N_QUEUES; i++) { + struct dp_netdev_queue *q = &dp->queues[i]; + + while (q->tail != q->head) { + struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK]; + ofpbuf_uninit(&u->buf); + } + } +} + +static void +dp_netdev_free(struct dp_netdev *dp) +{ + struct dp_netdev_port *port, *next; + dp_netdev_flow_flush(dp); - while (dp->n_ports > 0) { - struct dp_netdev_port *port = CONTAINER_OF( - dp->port_list.next, struct dp_netdev_port, node); + LIST_FOR_EACH_SAFE (port, next, node, &dp->port_list) { do_del_port(dp, port->port_no); } - for (i = 0; i < N_QUEUES; i++) { - ofpbuf_list_delete(&dp->queues[i]); - } + dp_netdev_purge_queues(dp); hmap_destroy(&dp->flow_table); free(dp->name); free(dp); @@ -275,38 +306,23 @@ dpif_netdev_destroy(struct dpif *dpif) } static int -dpif_netdev_get_stats(const struct dpif *dpif, struct odp_stats *stats) +dpif_netdev_get_stats(const struct dpif *dpif, struct dpif_dp_stats *stats) { struct dp_netdev *dp = get_dp_netdev(dpif); - memset(stats, 0, sizeof *stats); stats->n_flows = hmap_count(&dp->flow_table); - stats->cur_capacity = hmap_capacity(&dp->flow_table); - stats->max_capacity = MAX_FLOWS; - stats->n_ports = dp->n_ports; - stats->max_ports = MAX_PORTS; - stats->n_frags = dp->n_frags; stats->n_hit = dp->n_hit; stats->n_missed = dp->n_missed; stats->n_lost = dp->n_lost; - stats->max_miss_queue = MAX_QUEUE_LEN; - stats->max_action_queue = MAX_QUEUE_LEN; - return 0; -} - -static int -dpif_netdev_get_drop_frags(const struct dpif *dpif, bool *drop_fragsp) -{ - struct dp_netdev *dp = get_dp_netdev(dpif); - *drop_fragsp = dp->drop_frags; return 0; } -static int -dpif_netdev_set_drop_frags(struct dpif *dpif, bool drop_frags) +static const char* internal_port_type(const struct dp_netdev* dp) { - struct dp_netdev *dp = get_dp_netdev(dpif); - dp->drop_frags = drop_frags; - return 0; + if (dp->class == &dpif_netdev_class) + return "tap"; + if (dp->class == &dpif_planetlab_class) + return "pltap"; + return "dummy"; } static int @@ -314,39 +330,30 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, uint16_t port_no) { struct dp_netdev_port *port; - struct netdev_options netdev_options; struct netdev *netdev; - bool internal; + const char *open_type; int mtu; int error; /* XXX reject devices already in some dp_netdev. */ - if (type[0] == '\0' || !strcmp(type, "system")) { - internal = false; - } else if (!strcmp(type, "internal")) { - internal = true; - } else { - VLOG_WARN("%s: unsupported port type %s", devname, type); - return EINVAL; - } /* Open and validate network device. */ - memset(&netdev_options, 0, sizeof netdev_options); - netdev_options.name = devname; - netdev_options.ethertype = NETDEV_ETH_TYPE_ANY; - if (dp->class == &dpif_dummy_class) { - netdev_options.type = "dummy"; - } else if (internal) { - netdev_options.type = "tap"; - } - - error = netdev_open(&netdev_options, &netdev); + open_type = (strcmp(type, "internal") ? type : internal_port_type(dp)); + error = netdev_open(devname, open_type, &netdev); if (error) { return error; } /* XXX reject loopback devices */ /* XXX reject non-Ethernet devices */ + error = netdev_listen(netdev); + if (error) { + 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); if (error) { netdev_close(netdev); @@ -356,34 +363,76 @@ 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->internal = internal; + port->type = xstrdup(type); - netdev_get_mtu(netdev, &mtu); - if (mtu > max_mtu) { + error = netdev_get_mtu(netdev, &mtu); + if (!error) { max_mtu = mtu; } list_push_back(&dp->port_list, &port->node); dp->ports[port_no] = port; - dp->n_ports++; dp->serial++; return 0; } static int -dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, - uint16_t *port_nop) +choose_port(struct dpif *dpif, struct netdev *netdev) { struct dp_netdev *dp = get_dp_netdev(dpif); int port_no; + if (dpif->dpif_class != &dpif_netdev_class && + dpif->dpif_class != &dpif_planetlab_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]) { - *port_nop = port_no; - return do_add_port(dp, netdev_get_name(netdev), - netdev_get_type(netdev), port_no); + return port_no; + } + } + + 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; + + if (*port_nop != UINT16_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(dpif, netdev); + } + if (port_no >= 0) { + *port_nop = port_no; + return do_add_port(dp, netdev_get_name(netdev), + netdev_get_type(netdev), port_no); } return EFBIG; } @@ -392,7 +441,7 @@ static int dpif_netdev_port_del(struct dpif *dpif, uint16_t port_no) { struct dp_netdev *dp = get_dp_netdev(dpif); - return port_no == ODPP_LOCAL ? EINVAL : do_del_port(dp, port_no); + return port_no == OVSP_LOCAL ? EINVAL : do_del_port(dp, port_no); } static bool @@ -443,11 +492,11 @@ do_del_port(struct dp_netdev *dp, uint16_t port_no) list_remove(&port->node); dp->ports[port->port_no] = NULL; - dp->n_ports--; dp->serial++; name = xstrdup(netdev_get_name(port->netdev)); netdev_close(port->netdev); + free(port->type); free(name); free(port); @@ -456,18 +505,17 @@ do_del_port(struct dp_netdev *dp, uint16_t port_no) } static void -answer_port_query(const struct dp_netdev_port *port, struct odp_port *odp_port) +answer_port_query(const struct dp_netdev_port *port, + struct dpif_port *dpif_port) { - memset(odp_port, 0, sizeof *odp_port); - ovs_strlcpy(odp_port->devname, netdev_get_name(port->netdev), - sizeof odp_port->devname); - odp_port->port = port->port_no; - strcpy(odp_port->type, port->internal ? "internal" : "system"); + dpif_port->name = xstrdup(netdev_get_name(port->netdev)); + dpif_port->type = xstrdup(port->type); + dpif_port->port_no = port->port_no; } static int dpif_netdev_port_query_by_number(const struct dpif *dpif, uint16_t port_no, - struct odp_port *odp_port) + struct dpif_port *dpif_port) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_port *port; @@ -475,14 +523,14 @@ dpif_netdev_port_query_by_number(const struct dpif *dpif, uint16_t port_no, error = get_port_by_number(dp, port_no, &port); if (!error) { - answer_port_query(port, odp_port); + answer_port_query(port, dpif_port); } return error; } static int dpif_netdev_port_query_by_name(const struct dpif *dpif, const char *devname, - struct odp_port *odp_port) + struct dpif_port *dpif_port) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_port *port; @@ -490,11 +538,17 @@ dpif_netdev_port_query_by_name(const struct dpif *dpif, const char *devname, error = get_port_by_name(dp, devname, &port); if (!error) { - answer_port_query(port, odp_port); + answer_port_query(port, dpif_port); } return error; } +static int +dpif_netdev_get_max_ports(const struct dpif *dpif OVS_UNUSED) +{ + return MAX_PORTS; +} + static void dp_netdev_free_flow(struct dp_netdev *dp, struct dp_netdev_flow *flow) { @@ -521,23 +575,48 @@ dpif_netdev_flow_flush(struct dpif *dpif) return 0; } +struct dp_netdev_port_state { + uint32_t port_no; + char *name; +}; + static int -dpif_netdev_port_list(const struct dpif *dpif, struct odp_port *ports, int n) +dpif_netdev_port_dump_start(const struct dpif *dpif OVS_UNUSED, void **statep) { - struct dp_netdev *dp = get_dp_netdev(dpif); - struct dp_netdev_port *port; - int i; + *statep = xzalloc(sizeof(struct dp_netdev_port_state)); + return 0; +} - i = 0; - LIST_FOR_EACH (port, node, &dp->port_list) { - struct odp_port *odp_port = &ports[i]; - if (i >= n) { - break; +static int +dpif_netdev_port_dump_next(const struct dpif *dpif, void *state_, + struct dpif_port *dpif_port) +{ + struct dp_netdev_port_state *state = state_; + struct dp_netdev *dp = get_dp_netdev(dpif); + uint32_t port_no; + + for (port_no = state->port_no; port_no < MAX_PORTS; port_no++) { + struct dp_netdev_port *port = dp->ports[port_no]; + if (port) { + free(state->name); + state->name = xstrdup(netdev_get_name(port->netdev)); + dpif_port->name = state->name; + dpif_port->type = port->type; + dpif_port->port_no = port->port_no; + state->port_no = port_no + 1; + return 0; } - answer_port_query(port, odp_port); - i++; } - return dp->n_ports; + return EOF; +} + +static int +dpif_netdev_port_dump_done(const struct dpif *dpif OVS_UNUSED, void *state_) +{ + struct dp_netdev_port_state *state = state_; + free(state->name); + free(state); + return 0; } static int @@ -574,141 +653,97 @@ dp_netdev_lookup_flow(const struct dp_netdev *dp, const struct flow *key) return NULL; } -/* The caller must fill in odp_flow->key itself. */ static void -answer_flow_query(struct dp_netdev_flow *flow, uint32_t query_flags, - struct odp_flow *odp_flow) +get_dpif_flow_stats(struct dp_netdev_flow *flow, struct dpif_flow_stats *stats) { - if (flow) { - odp_flow->stats.n_packets = flow->packet_count; - odp_flow->stats.n_bytes = flow->byte_count; - odp_flow->stats.used_sec = flow->used.tv_sec; - odp_flow->stats.used_nsec = flow->used.tv_nsec; - odp_flow->stats.tcp_flags = TCP_FLAGS(flow->tcp_ctl); - odp_flow->stats.reserved = 0; - odp_flow->stats.error = 0; - if (odp_flow->actions_len > 0) { - memcpy(odp_flow->actions, flow->actions, - MIN(odp_flow->actions_len, flow->actions_len)); - odp_flow->actions_len = flow->actions_len; - } - - if (query_flags & ODPFF_ZERO_TCP_FLAGS) { - flow->tcp_ctl = 0; - } - - } else { - odp_flow->stats.error = ENOENT; - } + stats->n_packets = flow->packet_count; + stats->n_bytes = flow->byte_count; + stats->used = flow->used; + stats->tcp_flags = flow->tcp_flags; } static int -dpif_netdev_flow_get(const struct dpif *dpif, struct odp_flow flows[], int n) +dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len, + struct flow *flow) { - struct dp_netdev *dp = get_dp_netdev(dpif); - int i; + if (odp_flow_key_to_flow(key, key_len, flow)) { + /* 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 + * debugging. */ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + if (!VLOG_DROP_ERR(&rl)) { + struct ds s; + + ds_init(&s); + odp_flow_key_format(key, key_len, &s); + VLOG_ERR("internal error parsing flow key %s", ds_cstr(&s)); + ds_destroy(&s); + } - for (i = 0; i < n; i++) { - struct odp_flow *odp_flow = &flows[i]; - struct flow key; + return EINVAL; + } - odp_flow_key_to_flow(&odp_flow->key, &key); - answer_flow_query(dp_netdev_lookup_flow(dp, &key), - odp_flow->flags, odp_flow); + if (flow->in_port < OFPP_MAX + ? flow->in_port >= MAX_PORTS + : flow->in_port != OFPP_LOCAL && flow->in_port != OFPP_NONE) { + return EINVAL; } + return 0; } static int -dpif_netdev_validate_actions(const struct nlattr *actions, - size_t actions_len, bool *mutates) +dpif_netdev_flow_get(const struct dpif *dpif, + const struct nlattr *nl_key, size_t nl_key_len, + struct ofpbuf **actionsp, struct dpif_flow_stats *stats) { - const struct nlattr *a; - unsigned int left; - - *mutates = false; - NL_ATTR_FOR_EACH (a, left, actions, actions_len) { - uint16_t type = nl_attr_type(a); - int len = odp_action_len(type); - - if (len != nl_attr_get_size(a)) { - return EINVAL; - } - - switch (type) { - case ODPAT_OUTPUT: - if (nl_attr_get_u32(a) >= MAX_PORTS) { - return EINVAL; - } - break; - - case ODPAT_CONTROLLER: - case ODPAT_DROP_SPOOFED_ARP: - break; - - case ODPAT_SET_DL_TCI: - *mutates = true; - if (nl_attr_get_be16(a) & htons(VLAN_CFI)) { - return EINVAL; - } - break; + struct dp_netdev *dp = get_dp_netdev(dpif); + struct dp_netdev_flow *flow; + struct flow key; + int error; - case ODPAT_SET_NW_TOS: - *mutates = true; - if (nl_attr_get_u8(a) & IP_ECN_MASK) { - return EINVAL; - } - break; + error = dpif_netdev_flow_from_nlattrs(nl_key, nl_key_len, &key); + if (error) { + return error; + } - case ODPAT_STRIP_VLAN: - case ODPAT_SET_DL_SRC: - case ODPAT_SET_DL_DST: - case ODPAT_SET_NW_SRC: - case ODPAT_SET_NW_DST: - case ODPAT_SET_TP_SRC: - case ODPAT_SET_TP_DST: - *mutates = true; - break; + flow = dp_netdev_lookup_flow(dp, &key); + if (!flow) { + return ENOENT; + } - case ODPAT_SET_TUNNEL: - case ODPAT_SET_PRIORITY: - case ODPAT_POP_PRIORITY: - default: - return EOPNOTSUPP; - } + if (stats) { + get_dpif_flow_stats(flow, stats); + } + if (actionsp) { + *actionsp = ofpbuf_clone_data(flow->actions, flow->actions_len); } return 0; } static int -set_flow_actions(struct dp_netdev_flow *flow, struct odp_flow *odp_flow) +set_flow_actions(struct dp_netdev_flow *flow, + const struct nlattr *actions, size_t actions_len) { - bool mutates; - int error; - - error = dpif_netdev_validate_actions(odp_flow->actions, - odp_flow->actions_len, &mutates); - if (error) { - return error; - } - - flow->actions = xrealloc(flow->actions, odp_flow->actions_len); - flow->actions_len = odp_flow->actions_len; - memcpy(flow->actions, odp_flow->actions, odp_flow->actions_len); + flow->actions = xrealloc(flow->actions, actions_len); + flow->actions_len = actions_len; + memcpy(flow->actions, actions, actions_len); return 0; } static int -add_flow(struct dpif *dpif, struct odp_flow *odp_flow) +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; flow = xzalloc(sizeof *flow); - odp_flow_key_to_flow(&odp_flow->key, &flow->key); + flow->key = *key; - error = set_flow_actions(flow, odp_flow); + error = set_flow_actions(flow, actions, actions_len); if (error) { free(flow); return error; @@ -721,26 +756,34 @@ add_flow(struct dpif *dpif, struct odp_flow *odp_flow) static void clear_stats(struct dp_netdev_flow *flow) { - flow->used.tv_sec = 0; - flow->used.tv_nsec = 0; + flow->used = 0; flow->packet_count = 0; flow->byte_count = 0; - flow->tcp_ctl = 0; + flow->tcp_flags = 0; } static int -dpif_netdev_flow_put(struct dpif *dpif, struct odp_flow_put *put) +dpif_netdev_flow_put(struct dpif *dpif, const struct dpif_flow_put *put) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; struct flow key; + int error; + + error = dpif_netdev_flow_from_nlattrs(put->key, put->key_len, &key); + if (error) { + return error; + } - odp_flow_key_to_flow(&put->flow.key, &key); flow = dp_netdev_lookup_flow(dp, &key); if (!flow) { - if (put->flags & ODPPF_CREATE) { + if (put->flags & DPIF_FP_CREATE) { if (hmap_count(&dp->flow_table) < MAX_FLOWS) { - return add_flow(dpif, &put->flow); + if (put->stats) { + memset(put->stats, 0, sizeof *put->stats); + } + return dp_netdev_flow_add(dp, &key, put->actions, + put->actions_len); } else { return EFBIG; } @@ -748,10 +791,15 @@ dpif_netdev_flow_put(struct dpif *dpif, struct odp_flow_put *put) return ENOENT; } } else { - if (put->flags & ODPPF_MODIFY) { - int error = set_flow_actions(flow, &put->flow); - if (!error && put->flags & ODPPF_ZERO_STATS) { - clear_stats(flow); + if (put->flags & DPIF_FP_MODIFY) { + int error = set_flow_actions(flow, put->actions, put->actions_len); + if (!error) { + if (put->stats) { + get_dpif_flow_stats(flow, put->stats); + } + if (put->flags & DPIF_FP_ZERO_STATS) { + clear_stats(flow); + } } return error; } else { @@ -760,18 +808,24 @@ dpif_netdev_flow_put(struct dpif *dpif, struct odp_flow_put *put) } } - static int -dpif_netdev_flow_del(struct dpif *dpif, struct odp_flow *odp_flow) +dpif_netdev_flow_del(struct dpif *dpif, const struct dpif_flow_del *del) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; struct flow key; + int error; + + error = dpif_netdev_flow_from_nlattrs(del->key, del->key_len, &key); + if (error) { + return error; + } - odp_flow_key_to_flow(&odp_flow->key, &key); flow = dp_netdev_lookup_flow(dp, &key); if (flow) { - answer_flow_query(flow, 0, odp_flow); + if (del->stats) { + get_dpif_flow_stats(flow, del->stats); + } dp_netdev_free_flow(dp, flow); return 0; } else { @@ -779,113 +833,152 @@ dpif_netdev_flow_del(struct dpif *dpif, struct odp_flow *odp_flow) } } +struct dp_netdev_flow_state { + uint32_t bucket; + uint32_t offset; + struct nlattr *actions; + struct odputil_keybuf keybuf; + struct dpif_flow_stats stats; +}; + static int -dpif_netdev_flow_list(const struct dpif *dpif, struct odp_flow flows[], int n) +dpif_netdev_flow_dump_start(const struct dpif *dpif OVS_UNUSED, void **statep) { + struct dp_netdev_flow_state *state; + + *statep = state = xmalloc(sizeof *state); + state->bucket = 0; + state->offset = 0; + state->actions = NULL; + return 0; +} + +static int +dpif_netdev_flow_dump_next(const struct dpif *dpif, void *state_, + const struct nlattr **key, size_t *key_len, + const struct nlattr **actions, size_t *actions_len, + const struct dpif_flow_stats **stats) +{ + struct dp_netdev_flow_state *state = state_; struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; - int i; + struct hmap_node *node; - i = 0; - HMAP_FOR_EACH (flow, node, &dp->flow_table) { - if (i >= n) { - break; - } + node = hmap_at_position(&dp->flow_table, &state->bucket, &state->offset); + if (!node) { + return EOF; + } + + flow = CONTAINER_OF(node, struct dp_netdev_flow, node); + + if (key) { + struct ofpbuf buf; + + ofpbuf_use_stack(&buf, &state->keybuf, sizeof state->keybuf); + odp_flow_key_from_flow(&buf, &flow->key); + + *key = buf.data; + *key_len = buf.size; + } + + if (actions) { + free(state->actions); + state->actions = xmemdup(flow->actions, flow->actions_len); + + *actions = state->actions; + *actions_len = flow->actions_len; + } - odp_flow_key_from_flow(&flows[i].key, &flow->key); - answer_flow_query(flow, 0, &flows[i]); - i++; + if (stats) { + get_dpif_flow_stats(flow, &state->stats); + *stats = &state->stats; } - return hmap_count(&dp->flow_table); + + return 0; +} + +static int +dpif_netdev_flow_dump_done(const struct dpif *dpif OVS_UNUSED, void *state_) +{ + struct dp_netdev_flow_state *state = state_; + + free(state->actions); + free(state); + return 0; } static int -dpif_netdev_execute(struct dpif *dpif, - const struct nlattr *actions, size_t actions_len, - const struct ofpbuf *packet) +dpif_netdev_execute(struct dpif *dpif, const struct dpif_execute *execute) { struct dp_netdev *dp = get_dp_netdev(dpif); struct ofpbuf copy; - bool mutates; struct flow key; int error; - if (packet->size < ETH_HEADER_LEN || packet->size > UINT16_MAX) { + if (execute->packet->size < ETH_HEADER_LEN || + execute->packet->size > UINT16_MAX) { return EINVAL; } - error = dpif_netdev_validate_actions(actions, actions_len, &mutates); - if (error) { - return error; - } + /* Make a deep copy of 'packet', because we might modify its data. */ + ofpbuf_init(©, DP_NETDEV_HEADROOM + execute->packet->size); + ofpbuf_reserve(©, DP_NETDEV_HEADROOM); + ofpbuf_put(©, execute->packet->data, execute->packet->size); - if (mutates) { - /* We need a deep copy of 'packet' since we're going to modify its - * data. */ - ofpbuf_init(©, DP_NETDEV_HEADROOM + packet->size); - ofpbuf_reserve(©, DP_NETDEV_HEADROOM); - ofpbuf_put(©, packet->data, packet->size); - } else { - /* We still need a shallow copy of 'packet', even though we won't - * modify its data, because flow_extract() modifies packet->l2, etc. - * We could probably get away with modifying those but it's more polite - * if we don't. */ - copy = *packet; - } - flow_extract(©, 0, -1, &key); - error = dp_netdev_execute_actions(dp, ©, &key, actions, actions_len); - if (mutates) { - ofpbuf_uninit(©); + flow_extract(©, 0, NULL, -1, &key); + error = dpif_netdev_flow_from_nlattrs(execute->key, execute->key_len, + &key); + if (!error) { + dp_netdev_execute_actions(dp, ©, &key, + execute->actions, execute->actions_len); } + + ofpbuf_uninit(©); return error; } static int -dpif_netdev_recv_get_mask(const struct dpif *dpif, int *listen_mask) +dpif_netdev_recv_set(struct dpif *dpif OVS_UNUSED, bool enable OVS_UNUSED) { - struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); - *listen_mask = dpif_netdev->listen_mask; return 0; } static int -dpif_netdev_recv_set_mask(struct dpif *dpif, int listen_mask) +dpif_netdev_queue_to_priority(const struct dpif *dpif OVS_UNUSED, + uint32_t queue_id, uint32_t *priority) { - struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); - if (!(listen_mask & ~ODPL_ALL)) { - dpif_netdev->listen_mask = listen_mask; - return 0; - } else { - return EINVAL; - } + *priority = queue_id; + return 0; } -static int +static struct dp_netdev_queue * find_nonempty_queue(struct dpif *dpif) { - struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); struct dp_netdev *dp = get_dp_netdev(dpif); - int mask = dpif_netdev->listen_mask; int i; for (i = 0; i < N_QUEUES; i++) { - struct list *queue = &dp->queues[i]; - if (!list_is_empty(queue) && mask & (1u << i)) { - return i; + struct dp_netdev_queue *q = &dp->queues[i]; + if (q->head != q->tail) { + return q; } } - return -1; + return NULL; } static int -dpif_netdev_recv(struct dpif *dpif, struct ofpbuf **bufp) +dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall, + struct ofpbuf *buf) { - int queue_idx = find_nonempty_queue(dpif); - if (queue_idx >= 0) { - struct dp_netdev *dp = get_dp_netdev(dpif); + struct dp_netdev_queue *q = find_nonempty_queue(dpif); + if (q) { + struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK]; + + *upcall = u->upcall; + upcall->packet = buf; - *bufp = ofpbuf_from_list(list_pop_front(&dp->queues[queue_idx])); - dp->queue_len[queue_idx]--; + ofpbuf_uninit(buf); + *buf = u->buf; return 0; } else { @@ -896,25 +989,28 @@ dpif_netdev_recv(struct dpif *dpif, struct ofpbuf **bufp) static void dpif_netdev_recv_wait(struct dpif *dpif) { - if (find_nonempty_queue(dpif) >= 0) { + if (find_nonempty_queue(dpif)) { poll_immediate_wake(); } else { /* No messages ready to be received, and dp_wait() will ensure that we * wake up to queue new messages, so there is nothing to do. */ } } + +static void +dpif_netdev_recv_purge(struct dpif *dpif) +{ + struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); + dp_netdev_purge_queues(dpif_netdev->dp); +} 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) { - time_timespec(&flow->used); + flow->used = time_msec(); flow->packet_count++; flow->byte_count += packet->size; - if (key->dl_type == htons(ETH_TYPE_IP) && key->nw_proto == IPPROTO_TCP) { - struct tcp_header *th = packet->l4; - flow->tcp_ctl |= th->tcp_ctl; - } + flow->tcp_flags |= packet_get_tcp_flags(packet, &flow->key); } static void @@ -927,206 +1023,65 @@ dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, if (packet->size < ETH_HEADER_LEN) { return; } - if (flow_extract(packet, 0, port->port_no, &key) && dp->drop_frags) { - dp->n_frags++; - return; - } - + flow_extract(packet, 0, NULL, odp_port_to_ofp_port(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++; } else { dp->n_missed++; - dp_netdev_output_control(dp, packet, _ODPL_MISS_NR, port->port_no, 0); + dp_netdev_output_userspace(dp, packet, DPIF_UC_MISS, &key, 0); } } static void -dp_netdev_run(void) +dpif_netdev_run(struct dpif *dpif) { - struct shash_node *node; + struct dp_netdev *dp = get_dp_netdev(dpif); + struct dp_netdev_port *port; struct ofpbuf packet; - ofpbuf_init(&packet, DP_NETDEV_HEADROOM + max_mtu); - SHASH_FOR_EACH (node, &dp_netdevs) { - struct dp_netdev *dp = node->data; - struct dp_netdev_port *port; - - LIST_FOR_EACH (port, node, &dp->port_list) { - int error; + ofpbuf_init(&packet, DP_NETDEV_HEADROOM + VLAN_ETH_HEADER_LEN + max_mtu); - /* Reset packet contents. */ - ofpbuf_clear(&packet); - ofpbuf_reserve(&packet, DP_NETDEV_HEADROOM); - - error = netdev_recv(port->netdev, &packet); - if (!error) { - dp_netdev_port_input(dp, port, &packet); - } 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)); - } + LIST_FOR_EACH (port, node, &dp->port_list) { + int error; + + /* Reset packet contents. */ + ofpbuf_clear(&packet); + ofpbuf_reserve(&packet, DP_NETDEV_HEADROOM); + + error = netdev_recv(port->netdev, &packet); + if (!error) { + dp_netdev_port_input(dp, port, &packet); + } 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)); } } ofpbuf_uninit(&packet); } static void -dp_netdev_wait(void) -{ - struct shash_node *node; - - SHASH_FOR_EACH (node, &dp_netdevs) { - struct dp_netdev *dp = node->data; - struct dp_netdev_port *port; - - LIST_FOR_EACH (port, node, &dp->port_list) { - netdev_recv_wait(port->netdev); - } - } -} - - -/* Modify the TCI field of 'packet'. If a VLAN tag is present, its TCI field - * is replaced by 'tci'. If a VLAN tag is not present, one is added with the - * TCI field set to 'tci'. - */ -static void -dp_netdev_set_dl_tci(struct ofpbuf *packet, uint16_t tci) -{ - struct vlan_eth_header *veh; - struct eth_header *eh; - - eh = packet->l2; - if (packet->size >= sizeof(struct vlan_eth_header) - && eh->eth_type == htons(ETH_TYPE_VLAN)) { - veh = packet->l2; - veh->veth_tci = tci; - } else { - /* Insert new 802.1Q header. */ - struct vlan_eth_header tmp; - memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN); - memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN); - tmp.veth_type = htons(ETH_TYPE_VLAN); - tmp.veth_tci = tci; - tmp.veth_next_type = eh->eth_type; - - veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN); - memcpy(veh, &tmp, sizeof tmp); - packet->l2 = (char*)packet->l2 - VLAN_HEADER_LEN; - } -} - -static void -dp_netdev_strip_vlan(struct ofpbuf *packet) +dpif_netdev_wait(struct dpif *dpif) { - struct vlan_eth_header *veh = packet->l2; - if (packet->size >= sizeof *veh - && veh->veth_type == htons(ETH_TYPE_VLAN)) { - struct eth_header tmp; - - memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN); - memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN); - tmp.eth_type = veh->veth_next_type; + struct dp_netdev *dp = get_dp_netdev(dpif); + struct dp_netdev_port *port; - ofpbuf_pull(packet, VLAN_HEADER_LEN); - packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN; - memcpy(packet->data, &tmp, sizeof tmp); + LIST_FOR_EACH (port, node, &dp->port_list) { + netdev_recv_wait(port->netdev); } } static void -dp_netdev_set_dl_src(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN]) +dp_netdev_set_dl(struct ofpbuf *packet, const struct ovs_key_ethernet *eth_key) { struct eth_header *eh = packet->l2; - memcpy(eh->eth_src, dl_addr, sizeof eh->eth_src); -} -static void -dp_netdev_set_dl_dst(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN]) -{ - struct eth_header *eh = packet->l2; - memcpy(eh->eth_dst, dl_addr, sizeof eh->eth_dst); -} - -static bool -is_ip(const struct ofpbuf *packet, const struct flow *key) -{ - return key->dl_type == htons(ETH_TYPE_IP) && packet->l4; -} - -static void -dp_netdev_set_nw_addr(struct ofpbuf *packet, const struct flow *key, - const struct nlattr *a) -{ - if (is_ip(packet, key)) { - struct ip_header *nh = packet->l3; - ovs_be32 ip = nl_attr_get_be32(a); - uint16_t type = nl_attr_type(a); - uint32_t *field; - - field = type == ODPAT_SET_NW_SRC ? &nh->ip_src : &nh->ip_dst; - if (key->nw_proto == IP_TYPE_TCP && packet->l7) { - struct tcp_header *th = packet->l4; - th->tcp_csum = recalc_csum32(th->tcp_csum, *field, ip); - } else if (key->nw_proto == IP_TYPE_UDP && packet->l7) { - struct udp_header *uh = packet->l4; - if (uh->udp_csum) { - uh->udp_csum = recalc_csum32(uh->udp_csum, *field, ip); - if (!uh->udp_csum) { - uh->udp_csum = 0xffff; - } - } - } - nh->ip_csum = recalc_csum32(nh->ip_csum, *field, ip); - *field = ip; - } -} - -static void -dp_netdev_set_nw_tos(struct ofpbuf *packet, const struct flow *key, - uint8_t nw_tos) -{ - if (is_ip(packet, key)) { - struct ip_header *nh = packet->l3; - uint8_t *field = &nh->ip_tos; - - /* Set the DSCP bits and preserve the ECN bits. */ - uint8_t new = nw_tos | (nh->ip_tos & IP_ECN_MASK); - - nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field), - htons((uint16_t) new)); - *field = new; - } -} - -static void -dp_netdev_set_tp_port(struct ofpbuf *packet, const struct flow *key, - const struct nlattr *a) -{ - if (is_ip(packet, key)) { - uint16_t type = nl_attr_type(a); - ovs_be16 port = nl_attr_get_be16(a); - uint16_t *field; - - if (key->nw_proto == IPPROTO_TCP && packet->l7) { - struct tcp_header *th = packet->l4; - field = type == ODPAT_SET_TP_SRC ? &th->tcp_src : &th->tcp_dst; - th->tcp_csum = recalc_csum16(th->tcp_csum, *field, port); - *field = port; - } else if (key->nw_proto == IPPROTO_UDP && packet->l7) { - struct udp_header *uh = packet->l4; - field = type == ODPAT_SET_TP_SRC ? &uh->udp_src : &uh->udp_dst; - uh->udp_csum = recalc_csum16(uh->udp_csum, *field, port); - *field = port; - } else { - return; - } - } + memcpy(eh->eth_src, eth_key->eth_src, sizeof eh->eth_src); + memcpy(eh->eth_dst, eth_key->eth_dst, sizeof eh->eth_dst); } static void @@ -1140,61 +1095,140 @@ dp_netdev_output_port(struct dp_netdev *dp, struct ofpbuf *packet, } static int -dp_netdev_output_control(struct dp_netdev *dp, const struct ofpbuf *packet, - int queue_no, int port_no, uint64_t arg) +dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, + int queue_no, const struct flow *flow, uint64_t arg) { - struct odp_msg *header; - struct ofpbuf *msg; - size_t msg_size; + 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 (dp->queue_len[queue_no] >= MAX_QUEUE_LEN) { + if (q->head - q->tail >= MAX_QUEUE_LEN) { dp->n_lost++; return ENOBUFS; } - msg_size = sizeof *header + packet->size; - msg = ofpbuf_new_with_headroom(msg_size, DPIF_RECV_MSG_PADDING); - header = ofpbuf_put_uninit(msg, sizeof *header); - header->type = queue_no; - header->length = msg_size; - header->port = port_no; - header->arg = arg; - ofpbuf_put(msg, packet->data, packet->size); - list_push_back(&dp->queues[queue_no], &msg->list_node); - dp->queue_len[queue_no]++; + 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); + key_len = buf->size; + ofpbuf_pull(buf, key_len); + ofpbuf_reserve(buf, 2); + ofpbuf_put(buf, packet->data, packet->size); + + upcall = &u->upcall; + upcall->type = queue_no; + upcall->packet = buf; + upcall->key = buf->base; + upcall->key_len = key_len; + upcall->userdata = arg; return 0; } -/* Returns true if 'packet' is an invalid Ethernet+IPv4 ARP packet: one with - * screwy or truncated header fields or one whose inner and outer Ethernet - * address differ. */ -static bool -dp_netdev_is_spoofed_arp(struct ofpbuf *packet, const struct flow *key) +static void +dp_netdev_sample(struct dp_netdev *dp, + struct ofpbuf *packet, struct flow *key, + const struct nlattr *action) { - struct arp_eth_header *arp; - struct eth_header *eth; - ptrdiff_t l3_size; + const struct nlattr *subactions = NULL; + const struct nlattr *a; + size_t left; - if (key->dl_type != htons(ETH_TYPE_ARP)) { - return false; - } + NL_NESTED_FOR_EACH_UNSAFE (a, left, action) { + int type = nl_attr_type(a); - l3_size = (char *) ofpbuf_end(packet) - (char *) packet->l3; - if (l3_size < sizeof(struct arp_eth_header)) { - return true; + switch ((enum ovs_sample_attr) type) { + case OVS_SAMPLE_ATTR_PROBABILITY: + if (random_uint32() >= nl_attr_get_u32(a)) { + return; + } + break; + + case OVS_SAMPLE_ATTR_ACTIONS: + subactions = a; + break; + + case OVS_SAMPLE_ATTR_UNSPEC: + case __OVS_SAMPLE_ATTR_MAX: + default: + NOT_REACHED(); + } } - eth = packet->l2; - arp = packet->l3; - return (arp->ar_hrd != htons(ARP_HRD_ETHERNET) - || arp->ar_pro != htons(ARP_PRO_IP) - || arp->ar_hln != ETH_HEADER_LEN - || arp->ar_pln != 4 - || !eth_addr_equals(arp->ar_sha, eth->eth_src)); + dp_netdev_execute_actions(dp, packet, key, nl_attr_get(subactions), + nl_attr_get_size(subactions)); } -static int +static void +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; + + userdata_attr = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA); + userdata = userdata_attr ? nl_attr_get_u64(userdata_attr) : 0; + dp_netdev_output_userspace(dp, packet, DPIF_UC_ACTION, key, userdata); +} + +static void +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_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: + /* not implemented */ + break; + + case OVS_KEY_ATTR_ETHERNET: + dp_netdev_set_dl(packet, + nl_attr_get_unspec(a, sizeof(struct ovs_key_ethernet))); + break; + + case OVS_KEY_ATTR_IPV4: + ipv4_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv4)); + packet_set_ipv4(packet, ipv4_key->ipv4_src, ipv4_key->ipv4_dst, + ipv4_key->ipv4_tos, ipv4_key->ipv4_ttl); + 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); + break; + + case OVS_KEY_ATTR_UDP: + udp_key = nl_attr_get_unspec(a, sizeof(struct ovs_key_udp)); + packet_set_udp_port(packet, udp_key->udp_src, udp_key->udp_dst); + break; + + case OVS_KEY_ATTR_UNSPEC: + case OVS_KEY_ATTR_ENCAP: + case OVS_KEY_ATTR_ETHERTYPE: + case OVS_KEY_ATTR_IN_PORT: + case OVS_KEY_ATTR_VLAN: + case OVS_KEY_ATTR_ICMP: + case OVS_KEY_ATTR_ICMPV6: + case OVS_KEY_ATTR_ARP: + case OVS_KEY_ATTR_ND: + case __OVS_KEY_ATTR_MAX: + default: + NOT_REACHED(); + } +} + +static void dp_netdev_execute_actions(struct dp_netdev *dp, struct ofpbuf *packet, struct flow *key, const struct nlattr *actions, @@ -1204,95 +1238,114 @@ dp_netdev_execute_actions(struct dp_netdev *dp, unsigned int left; NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) { - switch (nl_attr_type(a)) { - case ODPAT_OUTPUT: - dp_netdev_output_port(dp, packet, nl_attr_get_u32(a)); - break; - - case ODPAT_CONTROLLER: - dp_netdev_output_control(dp, packet, _ODPL_ACTION_NR, - key->in_port, nl_attr_get_u64(a)); - break; - - case ODPAT_SET_DL_TCI: - dp_netdev_set_dl_tci(packet, nl_attr_get_be16(a)); - break; + const struct ovs_action_push_vlan *vlan; + int type = nl_attr_type(a); - case ODPAT_STRIP_VLAN: - dp_netdev_strip_vlan(packet); + switch ((enum ovs_action_attr) type) { + case OVS_ACTION_ATTR_OUTPUT: + dp_netdev_output_port(dp, packet, nl_attr_get_u32(a)); break; - case ODPAT_SET_DL_SRC: - dp_netdev_set_dl_src(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN)); + case OVS_ACTION_ATTR_USERSPACE: + dp_netdev_action_userspace(dp, packet, key, a); break; - case ODPAT_SET_DL_DST: - dp_netdev_set_dl_dst(packet, nl_attr_get_unspec(a, ETH_ADDR_LEN)); + case OVS_ACTION_ATTR_PUSH_VLAN: + vlan = nl_attr_get(a); + eth_push_vlan(packet, vlan->vlan_tci); break; - case ODPAT_SET_NW_SRC: - case ODPAT_SET_NW_DST: - dp_netdev_set_nw_addr(packet, key, a); + case OVS_ACTION_ATTR_POP_VLAN: + eth_pop_vlan(packet); break; - case ODPAT_SET_NW_TOS: - dp_netdev_set_nw_tos(packet, key, nl_attr_get_u8(a)); + case OVS_ACTION_ATTR_SET: + execute_set_action(packet, nl_attr_get(a)); break; - case ODPAT_SET_TP_SRC: - case ODPAT_SET_TP_DST: - dp_netdev_set_tp_port(packet, key, a); + case OVS_ACTION_ATTR_SAMPLE: + dp_netdev_sample(dp, packet, key, a); break; - case ODPAT_DROP_SPOOFED_ARP: - if (dp_netdev_is_spoofed_arp(packet, key)) { - return 0; - } + case OVS_ACTION_ATTR_UNSPEC: + case __OVS_ACTION_ATTR_MAX: + NOT_REACHED(); } } - return 0; } +#define DPIF_NETDEV_CLASS_FUNCTIONS \ + 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, \ + const struct dpif_class dpif_netdev_class = { "netdev", - dp_netdev_run, - dp_netdev_wait, - NULL, /* enumerate */ - dpif_netdev_open, - dpif_netdev_close, - NULL, /* get_all_names */ - dpif_netdev_destroy, - dpif_netdev_get_stats, - dpif_netdev_get_drop_frags, - dpif_netdev_set_drop_frags, - dpif_netdev_port_add, - dpif_netdev_port_del, - dpif_netdev_port_query_by_number, - dpif_netdev_port_query_by_name, - dpif_netdev_port_list, - 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_list, - dpif_netdev_execute, - dpif_netdev_recv_get_mask, - dpif_netdev_recv_set_mask, - NULL, /* get_sflow_probability */ - NULL, /* set_sflow_probability */ - NULL, /* queue_to_priority */ - dpif_netdev_recv, - dpif_netdev_recv_wait, + DPIF_NETDEV_CLASS_FUNCTIONS +}; + +const struct dpif_class dpif_planetlab_class = { + "planetlab", + DPIF_NETDEV_CLASS_FUNCTIONS }; +static void +dpif_dummy_register__(const char *type) +{ + struct dpif_class *class; + + class = xmalloc(sizeof *class); + *class = dpif_netdev_class; + class->type = xstrdup(type); + dp_register_provider(class); +} + void -dpif_dummy_register(void) +dpif_dummy_register(bool override) { - if (!dpif_dummy_class.type) { - dpif_dummy_class = dpif_netdev_class; - dpif_dummy_class.type = "dummy"; - dp_register_provider(&dpif_dummy_class); + if (override) { + struct sset types; + const char *type; + + sset_init(&types); + dp_enumerate_types(&types); + SSET_FOR_EACH (type, &types) { + if (!dp_unregister_provider(type)) { + dpif_dummy_register__(type); + } + } + sset_destroy(&types); } + + dpif_dummy_register__("dummy"); } +