X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-netdev.c;h=11f010c84f5877b14caa4e9977493352dd966e00;hb=99ab7312c17cd4ab603eb6194c50b9f1ef3ca1c9;hp=40f59c379338ac519798fc275d53f020806dcad2;hpb=4b60911067a82fbdfa87b7c2824412da20287ed8;p=sliver-openvswitch.git diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c index 40f59c379..11f010c84 100644 --- a/lib/dpif-netdev.c +++ b/lib/dpif-netdev.c @@ -31,6 +31,7 @@ #include #include +#include "classifier.h" #include "csum.h" #include "dpif.h" #include "dpif-provider.h" @@ -42,20 +43,26 @@ #include "netdev.h" #include "netdev-vport.h" #include "netlink.h" +#include "odp-execute.h" #include "odp-util.h" #include "ofp-print.h" #include "ofpbuf.h" #include "packets.h" #include "poll-loop.h" #include "random.h" +#include "seq.h" #include "shash.h" #include "sset.h" #include "timeval.h" +#include "unixctl.h" #include "util.h" #include "vlog.h" VLOG_DEFINE_THIS_MODULE(dpif_netdev); +/* By default, choose a priority in the middle. */ +#define NETDEV_RULE_PRIORITY 0x8000 + /* Configuration parameters. */ enum { MAX_PORTS = 256 }; /* Maximum number of ports. */ enum { MAX_FLOWS = 65536 }; /* Maximum number of flows in flow table. */ @@ -86,9 +93,12 @@ struct dp_netdev { char *name; int open_cnt; bool destroyed; + int max_mtu; /* Maximum MTU of any port added so far. */ struct dp_netdev_queue queues[N_QUEUES]; + struct classifier cls; /* Classifier. */ struct hmap flow_table; /* Flow table. */ + struct seq *queue_seq; /* Incremented whenever a packet is queued. */ /* Statistics. */ long long int n_hit; /* Number of flow table matches. */ @@ -98,28 +108,33 @@ struct dp_netdev { /* Ports. */ struct dp_netdev_port *ports[MAX_PORTS]; struct list port_list; - unsigned int serial; + struct seq *port_seq; /* Incremented whenever a port changes. */ }; /* A port in a netdev-based datapath. */ struct dp_netdev_port { - int port_no; /* Index into dp_netdev's 'ports'. */ + odp_port_t port_no; /* Index into dp_netdev's 'ports'. */ struct list node; /* Element in dp_netdev's 'port_list'. */ struct netdev *netdev; struct netdev_saved_flags *sf; + struct netdev_rx *rx; char *type; /* Port type as requested by user. */ }; /* A flow in dp_netdev's 'flow_table'. */ struct dp_netdev_flow { - struct hmap_node node; /* Element in dp_netdev's 'flow_table'. */ - struct flow key; + /* Packet classification. */ + struct cls_rule cr; /* In owning dp_netdev's 'cls'. */ + + /* Hash table index by unmasked flow.*/ + struct hmap_node node; /* In owning dp_netdev's 'flow_table'. */ + struct flow flow; /* The flow that created this entry. */ /* Statistics. */ 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. */ - uint8_t tcp_flags; /* Bitwise-OR of seen tcp_flags values. */ + uint16_t tcp_flags; /* Bitwise-OR of seen tcp_flags values. */ /* Actions. */ struct nlattr *actions; @@ -130,33 +145,37 @@ struct dp_netdev_flow { struct dpif_netdev { struct dpif dpif; struct dp_netdev *dp; - unsigned int dp_serial; + uint64_t last_port_seq; }; /* All netdev-based datapaths. */ static struct shash dp_netdevs = SHASH_INITIALIZER(&dp_netdevs); -/* Maximum port MTU seen so far. */ -static int max_mtu = ETH_PAYLOAD_MAX; +/* Global lock for all data. */ +static struct ovs_mutex dp_netdev_mutex = OVS_MUTEX_INITIALIZER; -static int get_port_by_number(struct dp_netdev *, uint32_t port_no, +static int get_port_by_number(struct dp_netdev *, odp_port_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, uint32_t port_no); -static int do_del_port(struct dp_netdev *, uint32_t port_no); + const char *type, odp_port_t port_no); +static int do_del_port(struct dp_netdev *, odp_port_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 *, int queue_no, const struct flow *, const struct nlattr *userdata); -static void dp_netdev_execute_actions(struct dp_netdev *, - struct ofpbuf *, struct flow *, +static void dp_netdev_execute_actions(struct dp_netdev *, const struct flow *, + struct ofpbuf *, const struct nlattr *actions, size_t actions_len); +static void dp_netdev_port_input(struct dp_netdev *dp, + struct dp_netdev_port *port, + struct ofpbuf *packet, uint32_t skb_priority, + uint32_t pkt_mark, const struct flow_tnl *tnl); static struct dpif_netdev * dpif_netdev_cast(const struct dpif *dpif) @@ -176,9 +195,12 @@ dpif_netdev_enumerate(struct sset *all_dps) { struct shash_node *node; + ovs_mutex_lock(&dp_netdev_mutex); SHASH_FOR_EACH(node, &dp_netdevs) { sset_add(all_dps, node->name); } + ovs_mutex_unlock(&dp_netdev_mutex); + return 0; } @@ -188,10 +210,17 @@ dpif_netdev_class_is_dummy(const struct dpif_class *class) return class != &dpif_netdev_class; } +static bool +dpif_netdev_class_is_planetlab(const struct dpif_class *class) +{ + return class == &dpif_planetlab_class; +} + static const char * dpif_netdev_port_open_type(const struct dpif_class *class, const char *type) { return strcmp(type, "internal") ? type + : dpif_netdev_class_is_planetlab(class) ? "pltap" : dpif_netdev_class_is_dummy(class) ? "dummy" : "tap"; } @@ -207,17 +236,20 @@ 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->dp_serial = dp->serial; + dpif->last_port_seq = seq_read(dp->port_seq); return &dpif->dpif; } -static int +/* Choose an unused, non-zero port number and return it on success. + * Return ODPP_NONE on failure. */ +static odp_port_t choose_port(struct dp_netdev *dp, const char *name) { - int port_no; + uint32_t port_no; - if (dp->class != &dpif_netdev_class) { + if (dp->class != &dpif_netdev_class && + dp->class != &dpif_planetlab_class) { const char *p; int start_no = 0; @@ -235,7 +267,7 @@ choose_port(struct dp_netdev *dp, const char *name) port_no = start_no + strtol(p, NULL, 10); if (port_no > 0 && port_no < MAX_PORTS && !dp->ports[port_no]) { - return port_no; + return u32_to_odp(port_no); } break; } @@ -244,11 +276,11 @@ choose_port(struct dp_netdev *dp, const char *name) for (port_no = 1; port_no < MAX_PORTS; port_no++) { if (!dp->ports[port_no]) { - return port_no; + return u32_to_odp(port_no); } } - return -1; + return ODPP_NONE; } static int @@ -263,13 +295,17 @@ create_dp_netdev(const char *name, const struct dpif_class *class, dp->class = class; dp->name = xstrdup(name); dp->open_cnt = 0; + dp->max_mtu = ETH_PAYLOAD_MAX; for (i = 0; i < N_QUEUES; i++) { dp->queues[i].head = dp->queues[i].tail = 0; } + dp->queue_seq = seq_create(); + classifier_init(&dp->cls, NULL); hmap_init(&dp->flow_table); list_init(&dp->port_list); + dp->port_seq = seq_create(); - error = do_add_port(dp, name, "internal", OVSP_LOCAL); + error = do_add_port(dp, name, "internal", ODPP_LOCAL); if (error) { dp_netdev_free(dp); return error; @@ -286,28 +322,23 @@ dpif_netdev_open(const struct dpif_class *class, const char *name, bool create, struct dpif **dpifp) { struct dp_netdev *dp; + int error; + ovs_mutex_lock(&dp_netdev_mutex); dp = shash_find_data(&dp_netdevs, name); if (!dp) { - if (!create) { - return ENODEV; - } else { - int error = create_dp_netdev(name, class, &dp); - if (error) { - return error; - } - ovs_assert(dp != NULL); - } + error = create ? create_dp_netdev(name, class, &dp) : ENODEV; } else { - if (dp->class != class) { - return EINVAL; - } else if (create) { - return EEXIST; - } + error = (dp->class != class ? EINVAL + : create ? EEXIST + : 0); } + if (!error) { + *dpifp = create_dpif_netdev(dp); + } + ovs_mutex_unlock(&dp_netdev_mutex); - *dpifp = create_dpif_netdev(dp); - return 0; + return error; } static void @@ -335,7 +366,10 @@ dp_netdev_free(struct dp_netdev *dp) do_del_port(dp, port->port_no); } dp_netdev_purge_queues(dp); + seq_destroy(dp->queue_seq); + classifier_destroy(&dp->cls); hmap_destroy(&dp->flow_table); + seq_destroy(dp->port_seq); free(dp->name); free(dp); } @@ -344,19 +378,28 @@ static void dpif_netdev_close(struct dpif *dpif) { struct dp_netdev *dp = get_dp_netdev(dpif); + + ovs_mutex_lock(&dp_netdev_mutex); + ovs_assert(dp->open_cnt > 0); if (--dp->open_cnt == 0 && dp->destroyed) { shash_find_and_delete(&dp_netdevs, dp->name); dp_netdev_free(dp); } free(dpif); + + ovs_mutex_unlock(&dp_netdev_mutex); } static int dpif_netdev_destroy(struct dpif *dpif) { struct dp_netdev *dp = get_dp_netdev(dpif); + + ovs_mutex_lock(&dp_netdev_mutex); dp->destroyed = true; + ovs_mutex_unlock(&dp_netdev_mutex); + return 0; } @@ -364,20 +407,28 @@ static int dpif_netdev_get_stats(const struct dpif *dpif, struct dpif_dp_stats *stats) { struct dp_netdev *dp = get_dp_netdev(dpif); + + ovs_mutex_lock(&dp_netdev_mutex); stats->n_flows = hmap_count(&dp->flow_table); stats->n_hit = dp->n_hit; stats->n_missed = dp->n_missed; stats->n_lost = dp->n_lost; + stats->n_masks = UINT64_MAX; + stats->n_mask_hit = UINT64_MAX; + ovs_mutex_unlock(&dp_netdev_mutex); + return 0; } static int do_add_port(struct dp_netdev *dp, const char *devname, const char *type, - uint32_t port_no) + odp_port_t port_no) { struct netdev_saved_flags *sf; struct dp_netdev_port *port; struct netdev *netdev; + struct netdev_rx *rx; + enum netdev_flags flags; const char *open_type; int mtu; int error; @@ -390,20 +441,27 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, if (error) { return error; } - /* XXX reject loopback devices */ /* XXX reject non-Ethernet devices */ - error = netdev_listen(netdev); + netdev_get_flags(netdev, &flags); + if (flags & NETDEV_LOOPBACK) { + VLOG_ERR("%s: cannot add a loopback device", devname); + netdev_close(netdev); + return EINVAL; + } + + error = netdev_rx_open(netdev, &rx); if (error && !(error == EOPNOTSUPP && dpif_netdev_class_is_dummy(dp->class))) { VLOG_ERR("%s: cannot receive packets on this network device (%s)", - devname, strerror(errno)); + devname, ovs_strerror(errno)); netdev_close(netdev); return error; } error = netdev_turn_flags_on(netdev, NETDEV_PROMISC, &sf); if (error) { + netdev_rx_close(rx); netdev_close(netdev); return error; } @@ -412,67 +470,84 @@ do_add_port(struct dp_netdev *dp, const char *devname, const char *type, port->port_no = port_no; port->netdev = netdev; port->sf = sf; + port->rx = rx; port->type = xstrdup(type); error = netdev_get_mtu(netdev, &mtu); - if (!error && mtu > max_mtu) { - max_mtu = mtu; + if (!error && mtu > dp->max_mtu) { + dp->max_mtu = mtu; } list_push_back(&dp->port_list, &port->node); - dp->ports[port_no] = port; - dp->serial++; + dp->ports[odp_to_u32(port_no)] = port; + seq_change(dp->port_seq); return 0; } static int dpif_netdev_port_add(struct dpif *dpif, struct netdev *netdev, - uint32_t *port_nop) + odp_port_t *port_nop) { struct dp_netdev *dp = get_dp_netdev(dpif); - int port_no; + char namebuf[NETDEV_VPORT_NAME_BUFSIZE]; + const char *dpif_port; + odp_port_t port_no; + int error; - if (*port_nop != UINT32_MAX) { - if (*port_nop >= MAX_PORTS) { - return EFBIG; - } else if (dp->ports[*port_nop]) { - return EBUSY; + ovs_mutex_lock(&dp_netdev_mutex); + dpif_port = netdev_vport_get_dpif_port(netdev, namebuf, sizeof namebuf); + if (*port_nop != ODPP_NONE) { + uint32_t port_idx = odp_to_u32(*port_nop); + if (port_idx >= MAX_PORTS) { + error = EFBIG; + } else if (dp->ports[port_idx]) { + error = EBUSY; + } else { + error = 0; + port_no = *port_nop; } - port_no = *port_nop; } else { - port_no = choose_port(dp, netdev_vport_get_dpif_port(netdev)); + port_no = choose_port(dp, dpif_port); + error = port_no == ODPP_NONE ? EFBIG : 0; } - if (port_no >= 0) { + if (!error) { *port_nop = port_no; - return do_add_port(dp, netdev_vport_get_dpif_port(netdev), - netdev_get_type(netdev), port_no); + error = do_add_port(dp, dpif_port, netdev_get_type(netdev), port_no); } - return EFBIG; + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } static int -dpif_netdev_port_del(struct dpif *dpif, uint32_t port_no) +dpif_netdev_port_del(struct dpif *dpif, odp_port_t port_no) { struct dp_netdev *dp = get_dp_netdev(dpif); - return port_no == OVSP_LOCAL ? EINVAL : do_del_port(dp, port_no); + int error; + + ovs_mutex_lock(&dp_netdev_mutex); + error = port_no == ODPP_LOCAL ? EINVAL : do_del_port(dp, port_no); + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } static bool -is_valid_port_number(uint32_t port_no) +is_valid_port_number(odp_port_t port_no) { - return port_no < MAX_PORTS; + return odp_to_u32(port_no) < MAX_PORTS; } static int get_port_by_number(struct dp_netdev *dp, - uint32_t port_no, struct dp_netdev_port **portp) + odp_port_t port_no, struct dp_netdev_port **portp) { if (!is_valid_port_number(port_no)) { *portp = NULL; return EINVAL; } else { - *portp = dp->ports[port_no]; + *portp = dp->ports[odp_to_u32(port_no)]; return *portp ? 0 : ENOENT; } } @@ -484,7 +559,7 @@ get_port_by_name(struct dp_netdev *dp, struct dp_netdev_port *port; LIST_FOR_EACH (port, node, &dp->port_list) { - if (!strcmp(netdev_vport_get_dpif_port(port->netdev), devname)) { + if (!strcmp(netdev_get_name(port->netdev), devname)) { *portp = port; return 0; } @@ -493,7 +568,7 @@ get_port_by_name(struct dp_netdev *dp, } static int -do_del_port(struct dp_netdev *dp, uint32_t port_no) +do_del_port(struct dp_netdev *dp, odp_port_t port_no) { struct dp_netdev_port *port; int error; @@ -504,11 +579,12 @@ do_del_port(struct dp_netdev *dp, uint32_t port_no) } list_remove(&port->node); - dp->ports[port->port_no] = NULL; - dp->serial++; + dp->ports[odp_to_u32(port_no)] = NULL; + seq_change(dp->port_seq); netdev_close(port->netdev); netdev_restore_flags(port->sf); + netdev_rx_close(port->rx); free(port->type); free(port); @@ -519,23 +595,26 @@ static void answer_port_query(const struct dp_netdev_port *port, struct dpif_port *dpif_port) { - dpif_port->name = xstrdup(netdev_vport_get_dpif_port(port->netdev)); + 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, uint32_t port_no, +dpif_netdev_port_query_by_number(const struct dpif *dpif, odp_port_t port_no, struct dpif_port *dpif_port) { struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_port *port; int error; + ovs_mutex_lock(&dp_netdev_mutex); error = get_port_by_number(dp, port_no, &port); if (!error && dpif_port) { answer_port_query(port, dpif_port); } + ovs_mutex_unlock(&dp_netdev_mutex); + return error; } @@ -547,34 +626,42 @@ dpif_netdev_port_query_by_name(const struct dpif *dpif, const char *devname, struct dp_netdev_port *port; int error; + ovs_mutex_lock(&dp_netdev_mutex); error = get_port_by_name(dp, devname, &port); if (!error && dpif_port) { answer_port_query(port, dpif_port); } + ovs_mutex_unlock(&dp_netdev_mutex); + return error; } -static int +static uint32_t 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) +dp_netdev_free_flow(struct dp_netdev *dp, struct dp_netdev_flow *netdev_flow) { - hmap_remove(&dp->flow_table, &flow->node); - free(flow->actions); - free(flow); + ovs_rwlock_wrlock(&dp->cls.rwlock); + classifier_remove(&dp->cls, &netdev_flow->cr); + ovs_rwlock_unlock(&dp->cls.rwlock); + cls_rule_destroy(&netdev_flow->cr); + + hmap_remove(&dp->flow_table, &netdev_flow->node); + free(netdev_flow->actions); + free(netdev_flow); } static void dp_netdev_flow_flush(struct dp_netdev *dp) { - struct dp_netdev_flow *flow, *next; + struct dp_netdev_flow *netdev_flow, *next; - HMAP_FOR_EACH_SAFE (flow, next, node, &dp->flow_table) { - dp_netdev_free_flow(dp, flow); + HMAP_FOR_EACH_SAFE (netdev_flow, next, node, &dp->flow_table) { + dp_netdev_free_flow(dp, netdev_flow); } } @@ -582,12 +669,16 @@ static int dpif_netdev_flow_flush(struct dpif *dpif) { struct dp_netdev *dp = get_dp_netdev(dpif); + + ovs_mutex_lock(&dp_netdev_mutex); dp_netdev_flow_flush(dp); + ovs_mutex_unlock(&dp_netdev_mutex); + return 0; } struct dp_netdev_port_state { - uint32_t port_no; + odp_port_t port_no; char *name; }; @@ -604,20 +695,26 @@ dpif_netdev_port_dump_next(const struct dpif *dpif, void *state_, { struct dp_netdev_port_state *state = state_; struct dp_netdev *dp = get_dp_netdev(dpif); - uint32_t port_no; + uint32_t port_idx; - for (port_no = state->port_no; port_no < MAX_PORTS; port_no++) { - struct dp_netdev_port *port = dp->ports[port_no]; + ovs_mutex_lock(&dp_netdev_mutex); + for (port_idx = odp_to_u32(state->port_no); + port_idx < MAX_PORTS; port_idx++) { + struct dp_netdev_port *port = dp->ports[port_idx]; if (port) { free(state->name); - state->name = xstrdup(netdev_vport_get_dpif_port(port->netdev)); + 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; + state->port_no = u32_to_odp(port_idx + 1); + ovs_mutex_unlock(&dp_netdev_mutex); + return 0; } } + ovs_mutex_unlock(&dp_netdev_mutex); + return EOF; } @@ -634,61 +731,94 @@ static int dpif_netdev_port_poll(const struct dpif *dpif_, char **devnamep OVS_UNUSED) { struct dpif_netdev *dpif = dpif_netdev_cast(dpif_); - if (dpif->dp_serial != dpif->dp->serial) { - dpif->dp_serial = dpif->dp->serial; - return ENOBUFS; + uint64_t new_port_seq; + int error; + + ovs_mutex_lock(&dp_netdev_mutex); + new_port_seq = seq_read(dpif->dp->port_seq); + if (dpif->last_port_seq != new_port_seq) { + dpif->last_port_seq = new_port_seq; + error = ENOBUFS; } else { - return EAGAIN; + error = EAGAIN; } + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } static void dpif_netdev_port_poll_wait(const struct dpif *dpif_) { struct dpif_netdev *dpif = dpif_netdev_cast(dpif_); - if (dpif->dp_serial != dpif->dp->serial) { - poll_immediate_wake(); - } + + ovs_mutex_lock(&dp_netdev_mutex); + seq_wait(dpif->dp->port_seq, dpif->last_port_seq); + ovs_mutex_unlock(&dp_netdev_mutex); +} + +static struct dp_netdev_flow * +dp_netdev_lookup_flow(const struct dp_netdev *dp, const struct flow *flow) +{ + struct cls_rule *cr; + + ovs_rwlock_wrlock(&dp->cls.rwlock); + cr = classifier_lookup(&dp->cls, flow, NULL); + ovs_rwlock_unlock(&dp->cls.rwlock); + + return (cr + ? CONTAINER_OF(cr, struct dp_netdev_flow, cr) + : NULL); } static struct dp_netdev_flow * -dp_netdev_lookup_flow(const struct dp_netdev *dp, const struct flow *key) +dp_netdev_find_flow(const struct dp_netdev *dp, const struct flow *flow) { - struct dp_netdev_flow *flow; + struct dp_netdev_flow *netdev_flow; - HMAP_FOR_EACH_WITH_HASH (flow, node, flow_hash(key, 0), &dp->flow_table) { - if (flow_equal(&flow->key, key)) { - return flow; + HMAP_FOR_EACH_WITH_HASH (netdev_flow, node, flow_hash(flow, 0), + &dp->flow_table) { + if (flow_equal(&netdev_flow->flow, flow)) { + return netdev_flow; } } return NULL; } static void -get_dpif_flow_stats(struct dp_netdev_flow *flow, struct dpif_flow_stats *stats) +get_dpif_flow_stats(struct dp_netdev_flow *netdev_flow, + struct dpif_flow_stats *stats) { - stats->n_packets = flow->packet_count; - stats->n_bytes = flow->byte_count; - stats->used = flow->used; - stats->tcp_flags = flow->tcp_flags; + stats->n_packets = netdev_flow->packet_count; + stats->n_bytes = netdev_flow->byte_count; + stats->used = netdev_flow->used; + stats->tcp_flags = netdev_flow->tcp_flags; } static int -dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len, - struct flow *flow) +dpif_netdev_flow_mask_from_nlattrs(const struct nlattr *key, uint32_t key_len, + const struct nlattr *mask_key, + uint32_t mask_key_len, struct flow *flow, + struct flow *mask) { - if (odp_flow_key_to_flow(key, key_len, flow) != ODP_FIT_PERFECT) { + odp_port_t in_port; + + if (odp_flow_key_to_flow(key, key_len, flow) + || (mask_key + && odp_flow_key_to_mask(mask_key, mask_key_len, mask, 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. */ + * and odp_flow_key_to_flow() disagree on the acceptable form of a flow + * or odp_flow_key_from_mask() and odp_flow_key_to_mask() disagree on + * the acceptable form of a mask. 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); + odp_flow_format(key, key_len, mask_key, mask_key_len, NULL, &s, + true); VLOG_ERR("internal error parsing flow key %s", ds_cstr(&s)); ds_destroy(&s); } @@ -696,22 +826,34 @@ dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len, return EINVAL; } - if (flow->in_port < OFPP_MAX - ? flow->in_port >= MAX_PORTS - : flow->in_port != OFPP_LOCAL && flow->in_port != OFPP_NONE) { + if (mask_key) { + /* Force unwildcard the in_port. */ + mask->in_port.odp_port = u32_to_odp(UINT32_MAX); + } + + in_port = flow->in_port.odp_port; + if (!is_valid_port_number(in_port) && in_port != ODPP_NONE) { return EINVAL; } return 0; } +static int +dpif_netdev_flow_from_nlattrs(const struct nlattr *key, uint32_t key_len, + struct flow *flow) +{ + return dpif_netdev_flow_mask_from_nlattrs(key, key_len, NULL, 0, flow, + NULL); +} + static int 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 dp_netdev_flow *netdev_flow; struct flow key; int error; @@ -720,110 +862,138 @@ dpif_netdev_flow_get(const struct dpif *dpif, return error; } - flow = dp_netdev_lookup_flow(dp, &key); - if (!flow) { - return ENOENT; + ovs_mutex_lock(&dp_netdev_mutex); + netdev_flow = dp_netdev_find_flow(dp, &key); + if (netdev_flow) { + if (stats) { + get_dpif_flow_stats(netdev_flow, stats); + } + if (actionsp) { + *actionsp = ofpbuf_clone_data(netdev_flow->actions, + netdev_flow->actions_len); + } + } else { + error = ENOENT; } + ovs_mutex_unlock(&dp_netdev_mutex); - if (stats) { - get_dpif_flow_stats(flow, stats); - } - if (actionsp) { - *actionsp = ofpbuf_clone_data(flow->actions, flow->actions_len); - } - return 0; + return error; } static int -set_flow_actions(struct dp_netdev_flow *flow, +set_flow_actions(struct dp_netdev_flow *netdev_flow, const struct nlattr *actions, size_t actions_len) { - flow->actions = xrealloc(flow->actions, actions_len); - flow->actions_len = actions_len; - memcpy(flow->actions, actions, actions_len); + netdev_flow->actions = xrealloc(netdev_flow->actions, actions_len); + netdev_flow->actions_len = actions_len; + memcpy(netdev_flow->actions, actions, actions_len); return 0; } static int -dp_netdev_flow_add(struct dp_netdev *dp, const struct flow *key, - const struct nlattr *actions, size_t actions_len) +dp_netdev_flow_add(struct dp_netdev *dp, const struct flow *flow, + const struct flow_wildcards *wc, + const struct nlattr *actions, + size_t actions_len) { - struct dp_netdev_flow *flow; + struct dp_netdev_flow *netdev_flow; + struct match match; int error; - flow = xzalloc(sizeof *flow); - flow->key = *key; + netdev_flow = xzalloc(sizeof *netdev_flow); + netdev_flow->flow = *flow; - error = set_flow_actions(flow, actions, actions_len); + match_init(&match, flow, wc); + cls_rule_init(&netdev_flow->cr, &match, NETDEV_RULE_PRIORITY); + ovs_rwlock_wrlock(&dp->cls.rwlock); + classifier_insert(&dp->cls, &netdev_flow->cr); + ovs_rwlock_unlock(&dp->cls.rwlock); + + error = set_flow_actions(netdev_flow, actions, actions_len); if (error) { - free(flow); + ovs_rwlock_wrlock(&dp->cls.rwlock); + classifier_remove(&dp->cls, &netdev_flow->cr); + ovs_rwlock_unlock(&dp->cls.rwlock); + cls_rule_destroy(&netdev_flow->cr); + + free(netdev_flow); return error; } - hmap_insert(&dp->flow_table, &flow->node, flow_hash(&flow->key, 0)); + hmap_insert(&dp->flow_table, &netdev_flow->node, flow_hash(flow, 0)); return 0; } static void -clear_stats(struct dp_netdev_flow *flow) +clear_stats(struct dp_netdev_flow *netdev_flow) { - flow->used = 0; - flow->packet_count = 0; - flow->byte_count = 0; - flow->tcp_flags = 0; + netdev_flow->used = 0; + netdev_flow->packet_count = 0; + netdev_flow->byte_count = 0; + netdev_flow->tcp_flags = 0; } static int 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; + struct dp_netdev_flow *netdev_flow; + struct flow flow; + struct flow_wildcards wc; int error; - error = dpif_netdev_flow_from_nlattrs(put->key, put->key_len, &key); + error = dpif_netdev_flow_mask_from_nlattrs(put->key, put->key_len, + put->mask, put->mask_len, &flow, &wc.masks); if (error) { return error; } - flow = dp_netdev_lookup_flow(dp, &key); - if (!flow) { + ovs_mutex_lock(&dp_netdev_mutex); + netdev_flow = dp_netdev_lookup_flow(dp, &flow); + if (!netdev_flow) { if (put->flags & DPIF_FP_CREATE) { if (hmap_count(&dp->flow_table) < MAX_FLOWS) { if (put->stats) { memset(put->stats, 0, sizeof *put->stats); } - return dp_netdev_flow_add(dp, &key, put->actions, - put->actions_len); + error = dp_netdev_flow_add(dp, &flow, &wc, put->actions, + put->actions_len); } else { - return EFBIG; + error = EFBIG; } } else { - return ENOENT; + error = ENOENT; } } else { - if (put->flags & DPIF_FP_MODIFY) { - int error = set_flow_actions(flow, put->actions, put->actions_len); + if (put->flags & DPIF_FP_MODIFY + && flow_equal(&flow, &netdev_flow->flow)) { + error = set_flow_actions(netdev_flow, put->actions, + put->actions_len); if (!error) { if (put->stats) { - get_dpif_flow_stats(flow, put->stats); + get_dpif_flow_stats(netdev_flow, put->stats); } if (put->flags & DPIF_FP_ZERO_STATS) { - clear_stats(flow); + clear_stats(netdev_flow); } } - return error; + } else if (put->flags & DPIF_FP_CREATE) { + error = EEXIST; } else { - return EEXIST; + /* Overlapping flow. */ + error = EINVAL; } } + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } static int 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 dp_netdev_flow *netdev_flow; struct flow key; int error; @@ -832,16 +1002,19 @@ dpif_netdev_flow_del(struct dpif *dpif, const struct dpif_flow_del *del) return error; } - flow = dp_netdev_lookup_flow(dp, &key); - if (flow) { + ovs_mutex_lock(&dp_netdev_mutex); + netdev_flow = dp_netdev_find_flow(dp, &key); + if (netdev_flow) { if (del->stats) { - get_dpif_flow_stats(flow, del->stats); + get_dpif_flow_stats(netdev_flow, del->stats); } - dp_netdev_free_flow(dp, flow); - return 0; + dp_netdev_free_flow(dp, netdev_flow); } else { - return ENOENT; + error = ENOENT; } + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } struct dp_netdev_flow_state { @@ -849,6 +1022,7 @@ struct dp_netdev_flow_state { uint32_t offset; struct nlattr *actions; struct odputil_keybuf keybuf; + struct odputil_keybuf maskbuf; struct dpif_flow_stats stats; }; @@ -867,44 +1041,63 @@ dpif_netdev_flow_dump_start(const struct dpif *dpif OVS_UNUSED, void **statep) static int dpif_netdev_flow_dump_next(const struct dpif *dpif, void *state_, const struct nlattr **key, size_t *key_len, + const struct nlattr **mask, size_t *mask_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; + struct dp_netdev_flow *netdev_flow; struct hmap_node *node; + ovs_mutex_lock(&dp_netdev_mutex); node = hmap_at_position(&dp->flow_table, &state->bucket, &state->offset); if (!node) { + ovs_mutex_unlock(&dp_netdev_mutex); return EOF; } - flow = CONTAINER_OF(node, struct dp_netdev_flow, node); + netdev_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, flow->key.in_port); + odp_flow_key_from_flow(&buf, &netdev_flow->flow, + netdev_flow->flow.in_port.odp_port); *key = buf.data; *key_len = buf.size; } + if (key && mask) { + struct ofpbuf buf; + struct flow_wildcards wc; + + ofpbuf_use_stack(&buf, &state->maskbuf, sizeof state->maskbuf); + minimask_expand(&netdev_flow->cr.match.mask, &wc); + odp_flow_key_from_mask(&buf, &wc.masks, &netdev_flow->flow, + odp_to_u32(wc.masks.in_port.odp_port)); + + *mask = buf.data; + *mask_len = buf.size; + } + if (actions) { free(state->actions); - state->actions = xmemdup(flow->actions, flow->actions_len); + state->actions = xmemdup(netdev_flow->actions, + netdev_flow->actions_len); *actions = state->actions; - *actions_len = flow->actions_len; + *actions_len = netdev_flow->actions_len; } if (stats) { - get_dpif_flow_stats(flow, &state->stats); + get_dpif_flow_stats(netdev_flow, &state->stats); *stats = &state->stats; } + ovs_mutex_unlock(&dp_netdev_mutex); return 0; } @@ -922,8 +1115,7 @@ static int dpif_netdev_execute(struct dpif *dpif, const struct dpif_execute *execute) { struct dp_netdev *dp = get_dp_netdev(dpif); - struct ofpbuf copy; - struct flow key; + struct flow md; int error; if (execute->packet->size < ETH_HEADER_LEN || @@ -931,20 +1123,24 @@ dpif_netdev_execute(struct dpif *dpif, const struct dpif_execute *execute) return EINVAL; } - /* 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); - - flow_extract(©, 0, 0, NULL, -1, &key); - error = dpif_netdev_flow_from_nlattrs(execute->key, execute->key_len, - &key); + /* Get packet metadata. */ + error = dpif_netdev_flow_from_nlattrs(execute->key, execute->key_len, &md); if (!error) { - dp_netdev_execute_actions(dp, ©, &key, + struct ofpbuf *copy; + struct flow key; + + /* Make a deep copy of 'packet', because we might modify its data. */ + copy = ofpbuf_clone_with_headroom(execute->packet, DP_NETDEV_HEADROOM); + + /* Extract flow key. */ + flow_extract(copy, md.skb_priority, md.pkt_mark, &md.tunnel, + &md.in_port, &key); + ovs_mutex_lock(&dp_netdev_mutex); + dp_netdev_execute_actions(dp, &key, copy, execute->actions, execute->actions_len); + ovs_mutex_unlock(&dp_netdev_mutex); + ofpbuf_delete(copy); } - - ofpbuf_uninit(©); return error; } @@ -981,7 +1177,11 @@ static int dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall, struct ofpbuf *buf) { - struct dp_netdev_queue *q = find_nonempty_queue(dpif); + struct dp_netdev_queue *q; + int error; + + ovs_mutex_lock(&dp_netdev_mutex); + q = find_nonempty_queue(dpif); if (q) { struct dp_netdev_upcall *u = &q->upcalls[q->tail++ & QUEUE_MASK]; @@ -991,55 +1191,70 @@ dpif_netdev_recv(struct dpif *dpif, struct dpif_upcall *upcall, ofpbuf_uninit(buf); *buf = u->buf; - return 0; + error = 0; } else { - return EAGAIN; + error = EAGAIN; } + ovs_mutex_unlock(&dp_netdev_mutex); + + return error; } static void dpif_netdev_recv_wait(struct dpif *dpif) { + struct dp_netdev *dp = get_dp_netdev(dpif); + uint64_t seq; + + ovs_mutex_lock(&dp_netdev_mutex); + seq = seq_read(dp->queue_seq); 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. */ + seq_wait(dp->queue_seq, seq); } + ovs_mutex_unlock(&dp_netdev_mutex); } static void dpif_netdev_recv_purge(struct dpif *dpif) { struct dpif_netdev *dpif_netdev = dpif_netdev_cast(dpif); + ovs_mutex_lock(&dp_netdev_mutex); dp_netdev_purge_queues(dpif_netdev->dp); + ovs_mutex_unlock(&dp_netdev_mutex); } static void -dp_netdev_flow_used(struct dp_netdev_flow *flow, const struct ofpbuf *packet) +dp_netdev_flow_used(struct dp_netdev_flow *netdev_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, &flow->key); + netdev_flow->used = time_msec(); + netdev_flow->packet_count++; + netdev_flow->byte_count += packet->size; + netdev_flow->tcp_flags |= packet_get_tcp_flags(packet, &netdev_flow->flow); } static void dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, - struct ofpbuf *packet) + struct ofpbuf *packet, uint32_t skb_priority, + uint32_t pkt_mark, const struct flow_tnl *tnl) { - struct dp_netdev_flow *flow; + struct dp_netdev_flow *netdev_flow; struct flow key; + union flow_in_port in_port_; if (packet->size < ETH_HEADER_LEN) { return; } - flow_extract(packet, 0, 0, NULL, port->port_no, &key); - flow = dp_netdev_lookup_flow(dp, &key); - if (flow) { - dp_netdev_flow_used(flow, packet); - dp_netdev_execute_actions(dp, packet, &key, - flow->actions, flow->actions_len); + in_port_.odp_port = port->port_no; + flow_extract(packet, skb_priority, pkt_mark, tnl, &in_port_, &key); + netdev_flow = dp_netdev_lookup_flow(dp, &key); + if (netdev_flow) { + dp_netdev_flow_used(netdev_flow, packet); + dp_netdev_execute_actions(dp, &key, packet, + netdev_flow->actions, + netdev_flow->actions_len); dp->n_hit++; } else { dp->n_missed++; @@ -1050,11 +1265,14 @@ dp_netdev_port_input(struct dp_netdev *dp, struct dp_netdev_port *port, static void dpif_netdev_run(struct dpif *dpif) { - struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_port *port; + struct dp_netdev *dp; struct ofpbuf packet; - ofpbuf_init(&packet, DP_NETDEV_HEADROOM + VLAN_ETH_HEADER_LEN + max_mtu); + ovs_mutex_lock(&dp_netdev_mutex); + dp = get_dp_netdev(dpif); + ofpbuf_init(&packet, + DP_NETDEV_HEADROOM + VLAN_ETH_HEADER_LEN + dp->max_mtu); LIST_FOR_EACH (port, node, &dp->port_list) { int error; @@ -1063,47 +1281,45 @@ dpif_netdev_run(struct dpif *dpif) ofpbuf_clear(&packet); ofpbuf_reserve(&packet, DP_NETDEV_HEADROOM); - error = netdev_recv(port->netdev, &packet); + error = port->rx ? netdev_rx_recv(port->rx, &packet) : EOPNOTSUPP; if (!error) { - dp_netdev_port_input(dp, port, &packet); + dp_netdev_port_input(dp, port, &packet, 0, 0, NULL); } 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_vport_get_dpif_port(port->netdev), - strerror(error)); + netdev_get_name(port->netdev), ovs_strerror(error)); } } ofpbuf_uninit(&packet); + ovs_mutex_unlock(&dp_netdev_mutex); } static void dpif_netdev_wait(struct dpif *dpif) { - struct dp_netdev *dp = get_dp_netdev(dpif); struct dp_netdev_port *port; - LIST_FOR_EACH (port, node, &dp->port_list) { - netdev_recv_wait(port->netdev); - } -} - -static void -dp_netdev_set_dl(struct ofpbuf *packet, const struct ovs_key_ethernet *eth_key) -{ - struct eth_header *eh = packet->l2; - - 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_output_port(struct dp_netdev *dp, struct ofpbuf *packet, - uint32_t out_port) -{ - struct dp_netdev_port *p = dp->ports[out_port]; - if (p) { - netdev_send(p->netdev, packet); + /* There is a race here, if thread A calls dpif_netdev_wait(dpif) and + * thread B calls dpif_port_add(dpif) or dpif_port_remove(dpif) before + * A makes it to poll_block(). + * + * But I think it doesn't matter: + * + * - In the dpif_port_add() case, A will not wake up when a packet + * arrives on the new port, but this would also happen if the + * ordering were reversed. + * + * - In the dpif_port_remove() case, A might wake up spuriously, but + * that is harmless. */ + + ovs_mutex_lock(&dp_netdev_mutex); + LIST_FOR_EACH (port, node, &get_dp_netdev(dpif)->port_list) { + if (port->rx) { + netdev_rx_wait(port->rx); + } } + ovs_mutex_unlock(&dp_netdev_mutex); } static int @@ -1128,7 +1344,7 @@ dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, ofpbuf_init(buf, buf_size); /* Put ODP flow. */ - odp_flow_key_from_flow(buf, flow, flow->in_port); + odp_flow_key_from_flow(buf, flow, flow->in_port.odp_port); upcall->key = buf->data; upcall->key_len = buf->size; @@ -1148,6 +1364,8 @@ dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, buf->size = packet->size; upcall->packet = buf; + seq_change(dp->queue_seq); + return 0; } else { dp->n_lost++; @@ -1155,207 +1373,128 @@ dp_netdev_output_userspace(struct dp_netdev *dp, const struct ofpbuf *packet, } } +struct dp_netdev_execute_aux { + struct dp_netdev *dp; + const struct flow *key; +}; + static void -dp_netdev_sample(struct dp_netdev *dp, - struct ofpbuf *packet, struct flow *key, - const struct nlattr *action) +dp_netdev_action_output(void *aux_, struct ofpbuf *packet, + const struct flow *flow OVS_UNUSED, + odp_port_t out_port) { - const struct nlattr *subactions = NULL; - const struct nlattr *a; - size_t left; - - 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; - - case OVS_SAMPLE_ATTR_UNSPEC: - case __OVS_SAMPLE_ATTR_MAX: - default: - NOT_REACHED(); - } + struct dp_netdev_execute_aux *aux = aux_; + struct dp_netdev_port *p = aux->dp->ports[odp_to_u32(out_port)]; + if (p) { + netdev_send(p->netdev, packet); } - - dp_netdev_execute_actions(dp, packet, key, nl_attr_get(subactions), - nl_attr_get_size(subactions)); } static void -dp_netdev_action_userspace(struct dp_netdev *dp, - struct ofpbuf *packet, struct flow *key, - const struct nlattr *a) +dp_netdev_action_userspace(void *aux_, struct ofpbuf *packet, + const struct flow *flow OVS_UNUSED, + const struct nlattr *a) { + struct dp_netdev_execute_aux *aux = aux_; const struct nlattr *userdata; userdata = nl_attr_find_nested(a, OVS_USERSPACE_ATTR_USERDATA); - 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_ipv6 *ipv6_key; - const struct ovs_key_tcp *tcp_key; - const struct ovs_key_udp *udp_key; - - switch (type) { - case OVS_KEY_ATTR_PRIORITY: - case OVS_KEY_ATTR_SKB_MARK: - case OVS_KEY_ATTR_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_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); - 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_MPLS: - set_mpls_lse(packet, nl_attr_get_be32(a)); - break; - - case OVS_KEY_ATTR_UNSPEC: - case OVS_KEY_ATTR_ENCAP: - case OVS_KEY_ATTR_ETHERTYPE: - 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(); - } + dp_netdev_output_userspace(aux->dp, packet, DPIF_UC_ACTION, aux->key, + userdata); } 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) { - int type = nl_attr_type(a); - - switch ((enum ovs_action_attr) type) { - case OVS_ACTION_ATTR_OUTPUT: - dp_netdev_output_port(dp, packet, nl_attr_get_u32(a)); - break; - - case OVS_ACTION_ATTR_USERSPACE: - dp_netdev_action_userspace(dp, packet, key, a); - break; - - case OVS_ACTION_ATTR_PUSH_VLAN: { - const struct ovs_action_push_vlan *vlan = nl_attr_get(a); - eth_push_vlan(packet, vlan->vlan_tci); - break; - } +dp_netdev_execute_actions(struct dp_netdev *dp, const struct flow *key, + struct ofpbuf *packet, + const struct nlattr *actions, size_t actions_len) +{ + struct dp_netdev_execute_aux aux = {dp, key}; + struct flow md = *key; /* Packet metadata, may be modified by actions. */ + + odp_execute_actions(&aux, packet, &md, actions, actions_len, + dp_netdev_action_output, dp_netdev_action_userspace); +} + +#define DPIF_NETDEV_CLASS_FUNCTIONS \ + dpif_netdev_enumerate, \ + dpif_netdev_port_open_type, \ + dpif_netdev_open, \ + dpif_netdev_close, \ + dpif_netdev_destroy, \ + dpif_netdev_run, \ + dpif_netdev_wait, \ + dpif_netdev_get_stats, \ + dpif_netdev_port_add, \ + dpif_netdev_port_del, \ + dpif_netdev_port_query_by_number, \ + dpif_netdev_port_query_by_name, \ + dpif_netdev_get_max_ports, \ + NULL, /* port_get_pid */ \ + dpif_netdev_port_dump_start, \ + dpif_netdev_port_dump_next, \ + dpif_netdev_port_dump_done, \ + dpif_netdev_port_poll, \ + dpif_netdev_port_poll_wait, \ + dpif_netdev_flow_get, \ + dpif_netdev_flow_put, \ + dpif_netdev_flow_del, \ + dpif_netdev_flow_flush, \ + dpif_netdev_flow_dump_start, \ + dpif_netdev_flow_dump_next, \ + dpif_netdev_flow_dump_done, \ + dpif_netdev_execute, \ + NULL, /* operate */ \ + dpif_netdev_recv_set, \ + dpif_netdev_queue_to_priority, \ + dpif_netdev_recv, \ + dpif_netdev_recv_wait, \ + dpif_netdev_recv_purge, \ - case OVS_ACTION_ATTR_POP_VLAN: - eth_pop_vlan(packet); - break; +const struct dpif_class dpif_netdev_class = { + "netdev", + DPIF_NETDEV_CLASS_FUNCTIONS +}; - case OVS_ACTION_ATTR_PUSH_MPLS: { - const struct ovs_action_push_mpls *mpls = nl_attr_get(a); - push_mpls(packet, mpls->mpls_ethertype, mpls->mpls_lse); - break; - } +const struct dpif_class dpif_planetlab_class = { + "planetlab", + DPIF_NETDEV_CLASS_FUNCTIONS +}; - case OVS_ACTION_ATTR_POP_MPLS: - pop_mpls(packet, nl_attr_get_be16(a)); - break; +static void +dpif_dummy_change_port_number(struct unixctl_conn *conn, int argc OVS_UNUSED, + const char *argv[], void *aux OVS_UNUSED) +{ + struct dp_netdev_port *port; + struct dp_netdev *dp; + int port_no; - case OVS_ACTION_ATTR_SET: - execute_set_action(packet, nl_attr_get(a)); - break; + dp = shash_find_data(&dp_netdevs, argv[1]); + if (!dp || !dpif_netdev_class_is_dummy(dp->class)) { + unixctl_command_reply_error(conn, "unknown datapath or not a dummy"); + return; + } - case OVS_ACTION_ATTR_SAMPLE: - dp_netdev_sample(dp, packet, key, a); - break; + if (get_port_by_name(dp, argv[2], &port)) { + unixctl_command_reply_error(conn, "unknown port"); + return; + } - case OVS_ACTION_ATTR_UNSPEC: - case __OVS_ACTION_ATTR_MAX: - NOT_REACHED(); - } + port_no = atoi(argv[3]); + if (port_no <= 0 || port_no >= MAX_PORTS) { + unixctl_command_reply_error(conn, "bad port number"); + return; + } + if (dp->ports[port_no]) { + unixctl_command_reply_error(conn, "port number already in use"); + return; } + dp->ports[odp_to_u32(port->port_no)] = NULL; + dp->ports[port_no] = port; + port->port_no = u32_to_odp(port_no); + seq_change(dp->port_seq); + unixctl_command_reply(conn, NULL); } -const struct dpif_class dpif_netdev_class = { - "netdev", - dpif_netdev_enumerate, - dpif_netdev_port_open_type, - dpif_netdev_open, - dpif_netdev_close, - dpif_netdev_destroy, - dpif_netdev_run, - dpif_netdev_wait, - dpif_netdev_get_stats, - dpif_netdev_port_add, - dpif_netdev_port_del, - dpif_netdev_port_query_by_number, - dpif_netdev_port_query_by_name, - dpif_netdev_get_max_ports, - NULL, /* port_get_pid */ - dpif_netdev_port_dump_start, - dpif_netdev_port_dump_next, - dpif_netdev_port_dump_done, - dpif_netdev_port_poll, - dpif_netdev_port_poll_wait, - dpif_netdev_flow_get, - dpif_netdev_flow_put, - dpif_netdev_flow_del, - dpif_netdev_flow_flush, - dpif_netdev_flow_dump_start, - dpif_netdev_flow_dump_next, - dpif_netdev_flow_dump_done, - dpif_netdev_execute, - NULL, /* operate */ - dpif_netdev_recv_set, - dpif_netdev_queue_to_priority, - dpif_netdev_recv, - dpif_netdev_recv_wait, - dpif_netdev_recv_purge, -}; - static void dpif_dummy_register__(const char *type) { @@ -1385,4 +1524,9 @@ dpif_dummy_register(bool override) } dpif_dummy_register__("dummy"); + + unixctl_command_register("dpif-dummy/change-port-number", + "DP PORT NEW-NUMBER", + 3, 3, dpif_dummy_change_port_number, NULL); } +