X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-netdev.c;h=eb8d66d9cae4e10e5ec2f47e5b2091c4beb5c11a;hb=fea393b1d6b2729a784b898dbdd48d30d42e3ff7;hp=435f90900190fdaed73ef73f8187d5f8741dd3f9;hpb=b3907fbc6c74ddad7507d0f7abb1f5a2528cd2be;p=sliver-openvswitch.git diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index 435f90900..eb8d66d9c 100644 --- a/lib/dpif-netdev.c +++ b/lib/dpif-netdev.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010 Nicira Networks. + * Copyright (c) 2009, 2010, 2011 Nicira Networks. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -35,15 +36,18 @@ #include "dpif.h" #include "dpif-provider.h" #include "dummy.h" +#include "dynamic-string.h" #include "flow.h" #include "hmap.h" #include "list.h" #include "netdev.h" +#include "netlink.h" #include "odp-util.h" #include "ofp-print.h" #include "ofpbuf.h" #include "packets.h" #include "poll-loop.h" +#include "random.h" #include "shash.h" #include "timeval.h" #include "util.h" @@ -52,8 +56,6 @@ 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. */ @@ -61,6 +63,17 @@ 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_queue { + struct dpif_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; @@ -68,19 +81,15 @@ 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; @@ -100,14 +109,14 @@ 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. */ + ovs_be16 tcp_ctl; /* Bitwise-OR of seen tcp_ctl values. */ /* Actions. */ - union odp_action *actions; - unsigned int n_actions; + struct nlattr *actions; + size_t actions_len; }; /* Interface to netdev-based datapath. */ @@ -135,11 +144,13 @@ 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, uint32_t arg); -static int dp_netdev_execute_actions(struct dp_netdev *, - struct ofpbuf *, struct flow *, - const union odp_action *, int n); +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_class dpif_dummy_class; @@ -185,13 +196,12 @@ 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; @@ -233,19 +243,32 @@ 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 dpif_upcall *upcall = q->upcalls[q->tail++ & QUEUE_MASK]; + + ofpbuf_delete(upcall->packet); + free(upcall); + } + } +} + +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); @@ -272,37 +295,13 @@ 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) -{ - struct dp_netdev *dp = get_dp_netdev(dpif); - dp->drop_frags = drop_frags; return 0; } @@ -311,14 +310,13 @@ 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; int mtu; int error; /* XXX reject devices already in some dp_netdev. */ - if (type[0] == '\0' || !strcmp(type, "system")) { + if (type[0] == '\0' || !strcmp(type, "system") || !strcmp(type, "dummy")) { internal = false; } else if (!strcmp(type, "internal")) { internal = true; @@ -328,22 +326,27 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, } /* 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"; + type = "dummy"; } else if (internal) { - netdev_options.type = "tap"; + type = "tap"; } - error = netdev_open(&netdev_options, &netdev); + error = netdev_open(devname, 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); @@ -355,14 +358,13 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, port->netdev = netdev; port->internal = internal; - 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; @@ -389,7 +391,7 @@ 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 @@ -440,7 +442,6 @@ 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)); @@ -453,18 +454,17 @@ 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->internal ? "internal" : "system"); + 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; @@ -472,14 +472,14 @@ 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; @@ -487,11 +487,17 @@ 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) { @@ -518,23 +524,48 @@ 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->internal ? "internal" : "system"; + 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 @@ -571,136 +602,98 @@ 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->n_actions > 0) { - unsigned int n = MIN(odp_flow->n_actions, flow->n_actions); - memcpy(odp_flow->actions, flow->actions, - n * sizeof *odp_flow->actions); - odp_flow->n_actions = flow->n_actions; - } - - 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 = TCP_FLAGS(flow->tcp_ctl); } 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 union odp_action *actions, int n_actions, - bool *mutates) -{ - unsigned int i; - - *mutates = false; - for (i = 0; i < n_actions; i++) { - const union odp_action *a = &actions[i]; - switch (a->type) { - case ODPAT_OUTPUT: - if (a->output.port >= MAX_PORTS) { - return EINVAL; - } - break; - - case ODPAT_CONTROLLER: - break; - - case ODPAT_SET_DL_TCI: - *mutates = true; - if (a->dl_tci.tci & htons(VLAN_CFI)) { - return EINVAL; - } - break; +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) +{ + 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 (a->nw_tos.nw_tos & 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; + } - 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) { - size_t n_bytes; - bool mutates; - int error; - - if (odp_flow->n_actions >= 4096 / sizeof *odp_flow->actions) { - return EINVAL; - } - error = dpif_netdev_validate_actions(odp_flow->actions, - odp_flow->n_actions, &mutates); - if (error) { - return error; - } - - n_bytes = odp_flow->n_actions * sizeof *flow->actions; - flow->actions = xrealloc(flow->actions, n_bytes); - flow->n_actions = odp_flow->n_actions; - memcpy(flow->actions, odp_flow->actions, n_bytes); + 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) +add_flow(struct dpif *dpif, 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; @@ -713,26 +706,36 @@ 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; } static int -dpif_netdev_flow_put(struct dpif *dpif, struct odp_flow_put *put) +dpif_netdev_flow_put(struct dpif *dpif, enum dpif_flow_put_flags flags, + const struct nlattr *nl_key, size_t nl_key_len, + const struct nlattr *actions, size_t actions_len, + struct dpif_flow_stats *stats) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; struct flow key; + int error; + + error = dpif_netdev_flow_from_nlattrs(nl_key, nl_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 (flags & DPIF_FP_CREATE) { if (hmap_count(&dp->flow_table) < MAX_FLOWS) { - return add_flow(dpif, &put->flow); + if (stats) { + memset(stats, 0, sizeof *stats); + } + return add_flow(dpif, &key, actions, actions_len); } else { return EFBIG; } @@ -740,10 +743,15 @@ 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 (flags & DPIF_FP_MODIFY) { + int error = set_flow_actions(flow, actions, actions_len); + if (!error) { + if (stats) { + get_dpif_flow_stats(flow, stats); + } + if (flags & DPIF_FP_ZERO_STATS) { + clear_stats(flow); + } } return error; } else { @@ -752,18 +760,26 @@ 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 nlattr *nl_key, size_t nl_key_len, + struct dpif_flow_stats *stats) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_flow *flow; struct flow key; + int error; + + error = dpif_netdev_flow_from_nlattrs(nl_key, nl_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 (stats) { + get_dpif_flow_stats(flow, stats); + } dp_netdev_free_flow(dp, flow); return 0; } else { @@ -771,34 +787,88 @@ 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; + } - odp_flow_key_from_flow(&flows[i].key, &flow->key); - answer_flow_query(flow, 0, &flows[i]); - i++; + 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; } - return hmap_count(&dp->flow_table); + + if (actions) { + free(state->actions); + state->actions = xmemdup(flow->actions, flow->actions_len); + + *actions = state->actions; + *actions_len = flow->actions_len; + } + + if (stats) { + get_dpif_flow_stats(flow, &state->stats); + *stats = &state->stats; + } + + 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 union odp_action actions[], int n_actions, + const struct nlattr *key_attrs, size_t key_len, + const struct nlattr *actions, size_t actions_len, const struct ofpbuf *packet) { struct dp_netdev *dp = get_dp_netdev(dpif); struct ofpbuf copy; - bool mutates; struct flow key; int error; @@ -806,29 +876,19 @@ dpif_netdev_execute(struct dpif *dpif, return EINVAL; } - error = dpif_netdev_validate_actions(actions, n_actions, &mutates); - if (error) { - return error; - } + /* Make a deep copy of 'packet', because we might modify its data. */ + ofpbuf_init(©, DP_NETDEV_HEADROOM + packet->size); + ofpbuf_reserve(©, DP_NETDEV_HEADROOM); + ofpbuf_put(©, packet->data, 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); - copy.data = (char*)copy.base + 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, n_actions); - if (mutates) { - ofpbuf_uninit(©); + flow_extract(©, 0, 0, -1, &key); + error = dpif_netdev_flow_from_nlattrs(key_attrs, key_len, &key); + if (!error) { + dp_netdev_execute_actions(dp, ©, &key, + actions, actions_len); } + + ofpbuf_uninit(©); return error; } @@ -844,15 +904,11 @@ static int dpif_netdev_recv_set_mask(struct dpif *dpif, int listen_mask) { 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; - } + dpif_netdev->listen_mask = listen_mask; + return 0; } -static int +static struct dp_netdev_queue * find_nonempty_queue(struct dpif *dpif) { struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); @@ -861,23 +917,22 @@ find_nonempty_queue(struct dpif *dpif) 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 && mask & (1u << i)) { + 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) { - int queue_idx = find_nonempty_queue(dpif); - if (queue_idx >= 0) { - struct dp_netdev *dp = get_dp_netdev(dpif); - - *bufp = ofpbuf_from_list(list_pop_front(&dp->queues[queue_idx])); - dp->queue_len[queue_idx]--; + 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); return 0; } else { @@ -888,19 +943,26 @@ 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) { - 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) { @@ -919,102 +981,60 @@ 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, 0, port->port_no, &key); flow = dp_netdev_lookup_flow(dp, &key); if (flow) { dp_netdev_flow_used(flow, &key, packet); dp_netdev_execute_actions(dp, packet, &key, - flow->actions, flow->n_actions); + 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. */ - packet.data = (char*)packet.base + DP_NETDEV_HEADROOM; - packet.size = 0; - - error = netdev_recv(port->netdev, &packet); - if (!error) { - dp_netdev_port_input(dp, port, &packet); - } else if (error != EAGAIN && error != EOPNOTSUPP) { - 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) +dpif_netdev_wait(struct dpif *dpif) { - 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; + struct dp_netdev *dp = get_dp_netdev(dpif); + struct dp_netdev_port *port; - veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN); - memcpy(veh, &tmp, sizeof tmp); - packet->l2 = (char*)packet->l2 - VLAN_HEADER_LEN; + LIST_FOR_EACH (port, node, &dp->port_list) { + netdev_recv_wait(port->netdev); } } static void -dp_netdev_strip_vlan(struct ofpbuf *packet) +dp_netdev_pop_vlan(struct ofpbuf *packet) { struct vlan_eth_header *veh = packet->l2; if (packet->size >= sizeof *veh @@ -1025,95 +1045,111 @@ dp_netdev_strip_vlan(struct ofpbuf *packet) memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN); tmp.eth_type = veh->veth_next_type; - packet->size -= VLAN_HEADER_LEN; - packet->data = (char*)packet->data + VLAN_HEADER_LEN; + ofpbuf_pull(packet, VLAN_HEADER_LEN); packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN; memcpy(packet->data, &tmp, sizeof tmp); } } 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); + + 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 -dp_netdev_set_dl_dst(struct ofpbuf *packet, const uint8_t dl_addr[ETH_ADDR_LEN]) +dp_netdev_set_ip_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr) { - struct eth_header *eh = packet->l2; - memcpy(eh->eth_dst, dl_addr, sizeof eh->eth_dst); + struct ip_header *nh = packet->l3; + + if (nh->ip_proto == IPPROTO_TCP && packet->l7) { + struct tcp_header *th = packet->l4; + th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr); + } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) { + struct udp_header *uh = packet->l4; + if (uh->udp_csum) { + uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr); + if (!uh->udp_csum) { + uh->udp_csum = htons(0xffff); + } + } + } + nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr); + *addr = new_addr; } -static bool -is_ip(const struct ofpbuf *packet, const struct flow *key) +static void +dp_netdev_set_ip_tos(struct ip_header *nh, uint8_t new_tos) { - return key->dl_type == htons(ETH_TYPE_IP) && packet->l4; + uint8_t *field = &nh->ip_tos; + + nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field), + htons((uint16_t) new_tos)); + *field = new_tos; } static void -dp_netdev_set_nw_addr(struct ofpbuf *packet, struct flow *key, - const struct odp_action_nw_addr *a) -{ - if (is_ip(packet, key)) { - struct ip_header *nh = packet->l3; - uint32_t *field; - - field = a->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, a->nw_addr); - } 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, a->nw_addr); - if (!uh->udp_csum) { - uh->udp_csum = 0xffff; - } - } - } - nh->ip_csum = recalc_csum32(nh->ip_csum, *field, a->nw_addr); - *field = a->nw_addr; +dp_netdev_set_ip_ttl(struct ip_header *nh, uint8_t new_ttl) +{ + uint8_t *field = &nh->ip_ttl; + + nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8), + htons(new_ttl << 8)); + *field = new_ttl; +} + +static void +dp_netdev_set_ipv4(struct ofpbuf *packet, const struct ovs_key_ipv4 *ipv4_key) +{ + struct ip_header *nh = packet->l3; + + if (nh->ip_src != ipv4_key->ipv4_src) { + dp_netdev_set_ip_addr(packet, &nh->ip_src, ipv4_key->ipv4_src); + } + if (nh->ip_dst != ipv4_key->ipv4_dst) { + dp_netdev_set_ip_addr(packet, &nh->ip_dst, ipv4_key->ipv4_dst); + } + if (nh->ip_tos != ipv4_key->ipv4_tos) { + dp_netdev_set_ip_tos(nh, ipv4_key->ipv4_tos); + } + if (nh->ip_ttl != ipv4_key->ipv4_ttl) { + dp_netdev_set_ip_ttl(nh, ipv4_key->ipv4_ttl); } } static void -dp_netdev_set_nw_tos(struct ofpbuf *packet, struct flow *key, - const struct odp_action_nw_tos *a) +dp_netdev_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum) { - if (is_ip(packet, key)) { - struct ip_header *nh = packet->l3; - uint8_t *field = &nh->ip_tos; + *csum = recalc_csum16(*csum, *port, new_port); + *port = new_port; +} - /* Set the DSCP bits and preserve the ECN bits. */ - uint8_t new = a->nw_tos | (nh->ip_tos & IP_ECN_MASK); +static void +dp_netdev_set_tcp_port(struct ofpbuf *packet, const struct ovs_key_tcp *tcp_key) +{ + struct tcp_header *th = packet->l4; - nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t)*field), - htons((uint16_t)a->nw_tos)); - *field = new; + if (th->tcp_src != tcp_key->tcp_src) { + dp_netdev_set_port(&th->tcp_src, tcp_key->tcp_src, &th->tcp_csum); + } + if (th->tcp_dst != tcp_key->tcp_dst) { + dp_netdev_set_port(&th->tcp_dst, tcp_key->tcp_dst, &th->tcp_csum); } } static void -dp_netdev_set_tp_port(struct ofpbuf *packet, struct flow *key, - const struct odp_action_tp_port *a) -{ - if (is_ip(packet, key)) { - uint16_t *field; - if (key->nw_proto == IPPROTO_TCP && packet->l7) { - struct tcp_header *th = packet->l4; - field = a->type == ODPAT_SET_TP_SRC ? &th->tcp_src : &th->tcp_dst; - th->tcp_csum = recalc_csum16(th->tcp_csum, *field, a->tp_port); - *field = a->tp_port; - } else if (key->nw_proto == IPPROTO_UDP && packet->l7) { - struct udp_header *uh = packet->l4; - field = a->type == ODPAT_SET_TP_SRC ? &uh->udp_src : &uh->udp_dst; - uh->udp_csum = recalc_csum16(uh->udp_csum, *field, a->tp_port); - *field = a->tp_port; - } else { - return; - } +dp_netdev_set_udp_port(struct ofpbuf *packet, const struct ovs_key_udp *udp_key) +{ + struct udp_header *uh = packet->l4; + + if (uh->udp_src != udp_key->udp_src) { + dp_netdev_set_port(&uh->udp_src, udp_key->udp_src, &uh->udp_csum); + } + if (uh->udp_dst != udp_key->udp_dst) { + dp_netdev_set_port(&uh->udp_dst, udp_key->udp_dst, &uh->udp_csum); } } @@ -1128,150 +1164,212 @@ 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, uint32_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 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]++; + buf = ofpbuf_new(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 = xzalloc(sizeof *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; } -/* 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); + + 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; - l3_size = (char *) ofpbuf_end(packet) - (char *) packet->l3; - if (l3_size < sizeof(struct arp_eth_header)) { - return true; + 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 -dp_netdev_execute_actions(struct dp_netdev *dp, +static void +dp_netdev_action_userspace(struct dp_netdev *dp, struct ofpbuf *packet, struct flow *key, - const union odp_action *actions, int n_actions) + const struct nlattr *a) { - int i; - for (i = 0; i < n_actions; i++) { - const union odp_action *a = &actions[i]; + const struct nlattr *userdata_attr; + uint64_t userdata; - switch (a->type) { - case ODPAT_OUTPUT: - dp_netdev_output_port(dp, packet, a->output.port); - break; + 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); +} - case ODPAT_CONTROLLER: - dp_netdev_output_control(dp, packet, _ODPL_ACTION_NR, - key->in_port, a->controller.arg); - break; +static void +execute_set_action(struct ofpbuf *packet, const struct nlattr *a) +{ + enum ovs_key_attr type = nl_attr_type(a); + switch (type) { + case OVS_KEY_ATTR_TUN_ID: + case OVS_KEY_ATTR_PRIORITY: + /* 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: + dp_netdev_set_ipv4(packet, + nl_attr_get_unspec(a, sizeof(struct ovs_key_ipv4))); + break; + + case OVS_KEY_ATTR_TCP: + dp_netdev_set_tcp_port(packet, + nl_attr_get_unspec(a, sizeof(struct ovs_key_tcp))); + break; + + case OVS_KEY_ATTR_UDP: + dp_netdev_set_udp_port(packet, + nl_attr_get_unspec(a, sizeof(struct ovs_key_udp))); + break; + + case OVS_KEY_ATTR_UNSPEC: + case OVS_KEY_ATTR_ENCAP: + case OVS_KEY_ATTR_ETHERTYPE: + case OVS_KEY_ATTR_IPV6: + 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(); + } +} - case ODPAT_SET_DL_TCI: - dp_netdev_set_dl_tci(packet, a->dl_tci.tci); - break; +static void +dp_netdev_execute_actions(struct dp_netdev *dp, + struct ofpbuf *packet, struct flow *key, + const struct nlattr *actions, + size_t actions_len) +{ + const struct nlattr *a; + 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); - 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, a->dl_addr.dl_addr); + 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, a->dl_addr.dl_addr); + 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->nw_addr); + case OVS_ACTION_ATTR_POP_VLAN: + dp_netdev_pop_vlan(packet); break; - case ODPAT_SET_NW_TOS: - dp_netdev_set_nw_tos(packet, key, &a->nw_tos); + 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->tp_port); + 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; } 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_run, + dpif_netdev_wait, 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_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_list, + dpif_netdev_flow_dump_start, + dpif_netdev_flow_dump_next, + dpif_netdev_flow_dump_done, dpif_netdev_execute, + NULL, /* operate */ 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_recv_purge, }; void