X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=13eb80bdd144ad575e3511850be48ddc4f64067e;hb=e0edde6fee279cdbbf3c179f5f50adaf0c7c7f1e;hp=1adf06f24f733759d0ecbf335940a3efafad9df2;hpb=15236eb49da2abff8b3d2482e45e0666732f1e18;p=sliver-openvswitch.git diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c index 1adf06f24..13eb80bdd 100644 --- a/vswitchd/bridge.c +++ b/vswitchd/bridge.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks +/* Copyright (c) 2008, 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. @@ -28,9 +28,12 @@ #include "dynamic-string.h" #include "hash.h" #include "hmap.h" +#include "hmapx.h" #include "jsonrpc.h" #include "lacp.h" #include "list.h" +#include "mac-learning.h" +#include "meta-flow.h" #include "netdev.h" #include "ofp-print.h" #include "ofpbuf.h" @@ -39,13 +42,15 @@ #include "sha1.h" #include "shash.h" #include "socket-util.h" +#include "stream.h" #include "stream-ssl.h" #include "sset.h" #include "system-stats.h" #include "timeval.h" #include "util.h" #include "unixctl.h" -#include "vswitchd/vswitch-idl.h" +#include "vlandev.h" +#include "lib/vswitch-idl.h" #include "xenserver.h" #include "vlog.h" #include "sflow_api.h" @@ -55,13 +60,25 @@ VLOG_DEFINE_THIS_MODULE(bridge); COVERAGE_DEFINE(bridge_reconfigure); +/* Configuration of an uninstantiated iface. */ +struct if_cfg { + struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */ + const struct ovsrec_interface *cfg; /* Interface record. */ + const struct ovsrec_port *parent; /* Parent port record. */ +}; + +/* OpenFlow port slated for removal from ofproto. */ +struct ofpp_garbage { + struct list list_node; /* Node in bridge's ofpp_garbage. */ + uint16_t ofp_port; /* Port to be deleted. */ +}; + struct iface { /* These members are always valid. */ struct list port_elem; /* Element in struct port's "ifaces" list. */ struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */ struct port *port; /* Containing port. */ char *name; /* Host network device name. */ - tag_type tag; /* Tag associated with this interface. */ /* These members are valid only after bridge_reconfigure() causes them to * be initialized. */ @@ -77,11 +94,12 @@ struct mirror { struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */ struct bridge *bridge; char *name; + const struct ovsrec_mirror *cfg; }; struct port { - struct bridge *bridge; struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */ + struct bridge *bridge; char *name; const struct ovsrec_port *cfg; @@ -107,6 +125,10 @@ struct bridge { struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */ struct hmap iface_by_name; /* "struct iface"s indexed by name. */ + struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */ + struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation. + Indexed on 'cfg->name'. */ + /* Port mirroring. */ struct hmap mirrors; /* "struct mirror" indexed by UUID. */ @@ -122,6 +144,9 @@ static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges); /* OVSDB IDL used to obtain configuration. */ static struct ovsdb_idl *idl; +/* Most recently processed IDL sequence number. */ +static unsigned int idl_seqno; + /* Each time this timer expires, the bridge fetches systems and interface * statistics and pushes them into the database. */ #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */ @@ -133,9 +158,19 @@ static long long int stats_timer = LLONG_MIN; #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */ static long long int db_limiter = LLONG_MIN; +/* In some datapaths, creating and destroying OpenFlow ports can be extremely + * expensive. This can cause bridge_reconfigure() to take a long time during + * which no other work can be done. To deal with this problem, we limit port + * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per + * call to bridge_reconfigure(). If there is more work to do after the limit + * is reached, 'need_reconfigure', is flagged and it's done on the next loop. + * This allows the rest of the code to catch up on important things like + * forwarding packets. */ +#define OFP_PORT_ACTION_WINDOW 10 +static bool reconfiguring = false; + static void add_del_bridges(const struct ovsrec_open_vswitch *); -static void bridge_del_ofprotos(void); -static bool bridge_add_ofprotos(struct bridge *); +static void bridge_update_ofprotos(void); static void bridge_create(const struct ovsrec_bridge *); static void bridge_destroy(struct bridge *); static struct bridge *bridge_lookup(const char *name); @@ -143,15 +178,17 @@ static unixctl_cb_func bridge_unixctl_dump_flows; static unixctl_cb_func bridge_unixctl_reconnect; static size_t bridge_get_controllers(const struct bridge *br, struct ovsrec_controller ***controllersp); -static void bridge_add_del_ports(struct bridge *); -static void bridge_add_ofproto_ports(struct bridge *); -static void bridge_del_ofproto_ports(struct bridge *); +static void bridge_add_del_ports(struct bridge *, + const unsigned long int *splinter_vlans); static void bridge_refresh_ofp_port(struct bridge *); static void bridge_configure_datapath_id(struct bridge *); static void bridge_configure_flow_eviction_threshold(struct bridge *); static void bridge_configure_netflow(struct bridge *); static void bridge_configure_forward_bpdu(struct bridge *); +static void bridge_configure_mac_idle_time(struct bridge *); static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number); +static void bridge_configure_stp(struct bridge *); +static void bridge_configure_tables(struct bridge *); static void bridge_configure_remotes(struct bridge *, const struct sockaddr_in *managers, size_t n_managers); @@ -161,6 +198,9 @@ static void bridge_pick_local_hw_addr(struct bridge *, static uint64_t bridge_pick_datapath_id(struct bridge *, const uint8_t bridge_ea[ETH_ADDR_LEN], struct iface *hw_addr_iface); +static void bridge_queue_if_cfg(struct bridge *, + const struct ovsrec_interface *, + const struct ovsrec_port *); static uint64_t dpid_from_hash(const void *, size_t nbytes); static bool bridge_has_bond_fake_iface(const struct bridge *, const char *name); @@ -169,7 +209,6 @@ static bool port_is_bond_fake_iface(const struct port *); static unixctl_cb_func qos_unixctl_show; static struct port *port_create(struct bridge *, const struct ovsrec_port *); -static void port_add_ifaces(struct port *); static void port_del_ifaces(struct port *); static void port_destroy(struct port *); static struct port *port_lookup(const struct bridge *, const char *name); @@ -178,19 +217,23 @@ static struct lacp_settings *port_configure_lacp(struct port *, struct lacp_settings *); static void port_configure_bond(struct port *, struct bond_settings *, uint32_t *bond_stable_ids); +static bool port_is_synthetic(const struct port *); static void bridge_configure_mirrors(struct bridge *); static struct mirror *mirror_create(struct bridge *, const struct ovsrec_mirror *); static void mirror_destroy(struct mirror *); -static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *); +static bool mirror_configure(struct mirror *); +static void mirror_refresh_stats(struct mirror *); static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *); -static struct iface *iface_create(struct port *port, - const struct ovsrec_interface *if_cfg); +static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port); +static const char *iface_get_type(const struct ovsrec_interface *, + const struct ovsrec_bridge *); static void iface_destroy(struct iface *); static struct iface *iface_lookup(const struct bridge *, const char *name); static struct iface *iface_find(const char *name); +static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name); static struct iface *iface_from_ofp_port(const struct bridge *, uint16_t ofp_port); static void iface_set_mac(struct iface *); @@ -202,11 +245,28 @@ static void iface_refresh_cfm_stats(struct iface *); static void iface_refresh_stats(struct iface *); static void iface_refresh_status(struct iface *); static bool iface_is_synthetic(const struct iface *); - static void shash_from_ovs_idl_map(char **keys, char **values, size_t n, struct shash *); static void shash_to_ovs_idl_map(struct shash *, char ***keys, char ***values, size_t *n); + +/* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) + * + * This is deprecated. It is only for compatibility with broken device drivers + * in old versions of Linux that do not properly support VLANs when VLAN + * devices are not used. When broken device drivers are no longer in + * widespread use, we will delete these interfaces. */ + +/* True if VLAN splinters are enabled on any interface, false otherwise.*/ +static bool vlan_splinters_enabled_anywhere; + +static bool vlan_splinters_is_enabled(const struct ovsrec_interface *); +static unsigned long int *collect_splinter_vlans( + const struct ovsrec_open_vswitch *); +static void configure_splinter_port(struct port *); +static void add_vlan_splinter_ports(struct bridge *, + const unsigned long int *splinter_vlans, + struct shash *ports); /* Public functions. */ @@ -218,6 +278,7 @@ bridge_init(const char *remote) { /* Create connection to database. */ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true); + idl_seqno = ovsdb_idl_get_seqno(idl); ovsdb_idl_set_lock(idl, "ovs_vswitchd"); ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg); @@ -229,8 +290,11 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version); ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id); + ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status); ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids); + ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status); + ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics); ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge); @@ -238,12 +302,15 @@ bridge_init(const char *remote) ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids); + ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health); ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current); ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids); @@ -257,6 +324,7 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids); ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids); + ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics); ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids); @@ -271,14 +339,16 @@ bridge_init(const char *remote) ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids); /* Register unixctl commands. */ - unixctl_command_register("qos/show", "interface", qos_unixctl_show, NULL); - unixctl_command_register("bridge/dump-flows", "bridge", + unixctl_command_register("qos/show", "interface", 1, 1, + qos_unixctl_show, NULL); + unixctl_command_register("bridge/dump-flows", "bridge", 1, 1, bridge_unixctl_dump_flows, NULL); - unixctl_command_register("bridge/reconnect", "[bridge]", + unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1, bridge_unixctl_reconnect, NULL); lacp_init(); bond_init(); cfm_init(); + stp_init(); } void @@ -334,10 +404,10 @@ collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, SSET_FOR_EACH (target, &targets) { struct sockaddr_in *sin = &managers[n_managers]; - if ((!strncmp(target, "tcp:", 4) - && inet_parse_active(target + 4, JSONRPC_TCP_PORT, sin)) || - (!strncmp(target, "ssl:", 4) - && inet_parse_active(target + 4, JSONRPC_SSL_PORT, sin))) { + if (stream_parse_target_with_default_ports(target, + JSONRPC_TCP_PORT, + JSONRPC_SSL_PORT, + sin)) { n_managers++; } } @@ -351,59 +421,110 @@ collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg, static void bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) { - struct sockaddr_in *managers; - struct bridge *br, *next; - int sflow_bridge_number; - size_t n_managers; + unsigned long int *splinter_vlans; + struct bridge *br; COVERAGE_INC(bridge_reconfigure); - /* Create and destroy "struct bridge"s, "struct port"s, and "struct - * iface"s according to 'ovs_cfg', with only very minimal configuration - * otherwise. + assert(!reconfiguring); + reconfiguring = true; + + /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according + * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal + * configuration otherwise. * - * This is purely an update to bridge data structures. Nothing is pushed + * This is mostly an update to bridge data structures. Nothing is pushed * down to ofproto or lower layers. */ add_del_bridges(ovs_cfg); + splinter_vlans = collect_splinter_vlans(ovs_cfg); HMAP_FOR_EACH (br, node, &all_bridges) { - bridge_add_del_ports(br); + bridge_add_del_ports(br, splinter_vlans); } + free(splinter_vlans); - /* Delete all datapaths and datapath ports that are no longer configured. - * - * The kernel will reject any attempt to add a given port to a datapath if - * that port already belongs to a different datapath, so we must do all - * port deletions before any port additions. A datapath always has a - * "local port" so we must delete not-configured datapaths too. */ - bridge_del_ofprotos(); + /* Delete datapaths that are no longer configured, and create ones which + * don't exist but should. */ + bridge_update_ofprotos(); + + /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */ HMAP_FOR_EACH (br, node, &all_bridges) { - if (br->ofproto) { - bridge_del_ofproto_ports(br); + bridge_refresh_ofp_port(br); + } + + /* Clear database records for "if_cfg"s which haven't been instantiated. */ + HMAP_FOR_EACH (br, node, &all_bridges) { + struct if_cfg *if_cfg; + + HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) { + iface_clear_db_record(if_cfg->cfg); } } +} - /* Create datapaths and datapath ports that are missing. - * - * After this is done, we have our final set of bridges, ports, and - * interfaces. Every "struct bridge" has an ofproto, every "struct port" - * has at least one iface, every "struct iface" has a valid ofp_port and - * netdev. */ - HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) { - if (!br->ofproto && !bridge_add_ofprotos(br)) { - bridge_destroy(br); +static bool +bridge_reconfigure_ofp(void) +{ + long long int deadline; + struct bridge *br; + + time_refresh(); + deadline = time_msec() + OFP_PORT_ACTION_WINDOW; + + /* The kernel will reject any attempt to add a given port to a datapath if + * that port already belongs to a different datapath, so we must do all + * port deletions before any port additions. */ + HMAP_FOR_EACH (br, node, &all_bridges) { + struct ofpp_garbage *garbage, *next; + + LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) { + ofproto_port_del(br->ofproto, garbage->ofp_port); + list_remove(&garbage->list_node); + free(garbage); + + time_refresh(); + if (time_msec() >= deadline) { + return false; + } } } + HMAP_FOR_EACH (br, node, &all_bridges) { - bridge_refresh_ofp_port(br); - bridge_add_ofproto_ports(br); + struct if_cfg *if_cfg, *next; + + HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) { + iface_create(br, if_cfg, -1); + time_refresh(); + if (time_msec() >= deadline) { + return false; + } + } } + return true; +} + +static bool +bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg) +{ + struct sockaddr_in *managers; + int sflow_bridge_number; + size_t n_managers; + struct bridge *br; + bool done; + + assert(reconfiguring); + done = bridge_reconfigure_ofp(); + /* Complete the configuration. */ sflow_bridge_number = 0; collect_in_band_managers(ovs_cfg, &managers, &n_managers); HMAP_FOR_EACH (br, node, &all_bridges) { struct port *port; + /* We need the datapath ID early to allow LACP ports to use it as the + * default system ID. */ + bridge_configure_datapath_id(br); + HMAP_FOR_EACH (port, hmap_node, &br->ports) { struct iface *iface; @@ -416,29 +537,38 @@ bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg) } } bridge_configure_mirrors(br); - bridge_configure_datapath_id(br); bridge_configure_flow_eviction_threshold(br); bridge_configure_forward_bpdu(br); + bridge_configure_mac_idle_time(br); bridge_configure_remotes(br, managers, n_managers); bridge_configure_netflow(br); bridge_configure_sflow(br, &sflow_bridge_number); + bridge_configure_stp(br); + bridge_configure_tables(br); } free(managers); - /* ovs-vswitchd has completed initialization, so allow the process that - * forked us to exit successfully. */ - daemonize_complete(); + if (done) { + /* ovs-vswitchd has completed initialization, so allow the process that + * forked us to exit successfully. */ + daemonize_complete(); + reconfiguring = false; + } + + return done; } -/* Iterate over all ofprotos and delete any of them that do not have a - * configured bridge or that are the wrong type. */ +/* Delete ofprotos which aren't configured or have the wrong type. Create + * ofprotos which don't exist but need to. */ static void -bridge_del_ofprotos(void) +bridge_update_ofprotos(void) { + struct bridge *br, *next; struct sset names; struct sset types; const char *type; + /* Delete ofprotos with no bridge or with the wrong type. */ sset_init(&names); sset_init(&types); ofproto_enumerate_types(&types); @@ -447,7 +577,7 @@ bridge_del_ofprotos(void) ofproto_enumerate_names(type, &names); SSET_FOR_EACH (name, &names) { - struct bridge *br = bridge_lookup(name); + br = bridge_lookup(name); if (!br || strcmp(type, br->type)) { ofproto_delete(name, type); } @@ -455,17 +585,44 @@ bridge_del_ofprotos(void) } sset_destroy(&names); sset_destroy(&types); -} -static bool -bridge_add_ofprotos(struct bridge *br) -{ - int error = ofproto_create(br->name, br->type, &br->ofproto); - if (error) { - VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error)); - return false; + /* Add ofprotos for bridges which don't have one yet. */ + HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) { + struct bridge *br2; + int error; + + if (br->ofproto) { + continue; + } + + /* Remove ports from any datapath with the same name as 'br'. If we + * don't do this, creating 'br''s ofproto will fail because a port with + * the same name as its local port already exists. */ + HMAP_FOR_EACH (br2, node, &all_bridges) { + struct ofproto_port ofproto_port; + + if (!br2->ofproto) { + continue; + } + + if (!ofproto_port_query_by_name(br2->ofproto, br->name, + &ofproto_port)) { + error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port); + if (error) { + VLOG_ERR("failed to delete port %s: %s", ofproto_port.name, + strerror(error)); + } + ofproto_port_destroy(&ofproto_port); + } + } + + error = ofproto_create(br->name, br->type, &br->ofproto); + if (error) { + VLOG_ERR("failed to create bridge %s: %s", br->name, + strerror(error)); + bridge_destroy(br); + } } - return true; } static void @@ -477,6 +634,11 @@ port_configure(struct port *port) struct ofproto_bundle_settings s; struct iface *iface; + if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) { + configure_splinter_port(port); + return; + } + /* Get name. */ s.name = port->name; @@ -489,17 +651,8 @@ port_configure(struct port *port) /* Get VLAN tag. */ s.vlan = -1; - if (cfg->tag) { - if (list_is_short(&port->ifaces)) { - if (*cfg->tag >= 0 && *cfg->tag <= 4095) { - s.vlan = *cfg->tag; - } - } else { - /* It's possible that bonded, VLAN-tagged ports make sense. Maybe - * they even work as-is. But they have not been tested. */ - VLOG_WARN("port %s: VLAN tags not supported on bonded ports", - port->name); - } + if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) { + s.vlan = *cfg->tag; } /* Get VLAN trunks. */ @@ -534,6 +687,8 @@ port_configure(struct port *port) s.vlan_mode = PORT_VLAN_TRUNK; } } + s.use_priority_tags = !strcmp("true", ovsrec_port_get_other_config_value( + cfg, "priority-tags", "")); /* Get LACP settings. */ s.lacp = port_configure_lacp(port, &lacp_settings); @@ -715,6 +870,211 @@ bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number) sset_destroy(&oso.targets); } +static void +port_configure_stp(const struct ofproto *ofproto, struct port *port, + struct ofproto_port_stp_settings *port_s, + int *port_num_counter, unsigned long *port_num_bitmap) +{ + const char *config_str; + struct iface *iface; + + config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-enable", + NULL); + if (config_str && !strcmp(config_str, "false")) { + port_s->enable = false; + return; + } else { + port_s->enable = true; + } + + /* STP over bonds is not supported. */ + if (!list_is_singleton(&port->ifaces)) { + VLOG_ERR("port %s: cannot enable STP on bonds, disabling", + port->name); + port_s->enable = false; + return; + } + + iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); + + /* Internal ports shouldn't participate in spanning tree, so + * skip them. */ + if (!strcmp(iface->type, "internal")) { + VLOG_DBG("port %s: disable STP on internal ports", port->name); + port_s->enable = false; + return; + } + + /* STP on mirror output ports is not supported. */ + if (ofproto_is_mirror_output_bundle(ofproto, port)) { + VLOG_DBG("port %s: disable STP on mirror ports", port->name); + port_s->enable = false; + return; + } + + config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-port-num", + NULL); + if (config_str) { + unsigned long int port_num = strtoul(config_str, NULL, 0); + int port_idx = port_num - 1; + + if (port_num < 1 || port_num > STP_MAX_PORTS) { + VLOG_ERR("port %s: invalid stp-port-num", port->name); + port_s->enable = false; + return; + } + + if (bitmap_is_set(port_num_bitmap, port_idx)) { + VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling", + port->name, port_num); + port_s->enable = false; + return; + } + bitmap_set1(port_num_bitmap, port_idx); + port_s->port_num = port_idx; + } else { + if (*port_num_counter > STP_MAX_PORTS) { + VLOG_ERR("port %s: too many STP ports, disabling", port->name); + port_s->enable = false; + return; + } + + port_s->port_num = (*port_num_counter)++; + } + + config_str = ovsrec_port_get_other_config_value(port->cfg, "stp-path-cost", + NULL); + if (config_str) { + port_s->path_cost = strtoul(config_str, NULL, 10); + } else { + enum netdev_features current; + + if (netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL)) { + /* Couldn't get speed, so assume 100Mb/s. */ + port_s->path_cost = 19; + } else { + unsigned int mbps; + + mbps = netdev_features_to_bps(current) / 1000000; + port_s->path_cost = stp_convert_speed_to_cost(mbps); + } + } + + config_str = ovsrec_port_get_other_config_value(port->cfg, + "stp-port-priority", + NULL); + if (config_str) { + port_s->priority = strtoul(config_str, NULL, 0); + } else { + port_s->priority = STP_DEFAULT_PORT_PRIORITY; + } +} + +/* Set spanning tree configuration on 'br'. */ +static void +bridge_configure_stp(struct bridge *br) +{ + if (!br->cfg->stp_enable) { + ofproto_set_stp(br->ofproto, NULL); + } else { + struct ofproto_stp_settings br_s; + const char *config_str; + struct port *port; + int port_num_counter; + unsigned long *port_num_bitmap; + + config_str = ovsrec_bridge_get_other_config_value(br->cfg, + "stp-system-id", + NULL); + if (config_str) { + uint8_t ea[ETH_ADDR_LEN]; + + if (eth_addr_from_string(config_str, ea)) { + br_s.system_id = eth_addr_to_uint64(ea); + } else { + br_s.system_id = eth_addr_to_uint64(br->ea); + VLOG_ERR("bridge %s: invalid stp-system-id, defaulting " + "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea)); + } + } else { + br_s.system_id = eth_addr_to_uint64(br->ea); + } + + config_str = ovsrec_bridge_get_other_config_value(br->cfg, + "stp-priority", + NULL); + if (config_str) { + br_s.priority = strtoul(config_str, NULL, 0); + } else { + br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY; + } + + config_str = ovsrec_bridge_get_other_config_value(br->cfg, + "stp-hello-time", + NULL); + if (config_str) { + br_s.hello_time = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.hello_time = STP_DEFAULT_HELLO_TIME; + } + + config_str = ovsrec_bridge_get_other_config_value(br->cfg, + "stp-max-age", + NULL); + if (config_str) { + br_s.max_age = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.max_age = STP_DEFAULT_MAX_AGE; + } + + config_str = ovsrec_bridge_get_other_config_value(br->cfg, + "stp-forward-delay", + NULL); + if (config_str) { + br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000; + } else { + br_s.fwd_delay = STP_DEFAULT_FWD_DELAY; + } + + /* Configure STP on the bridge. */ + if (ofproto_set_stp(br->ofproto, &br_s)) { + VLOG_ERR("bridge %s: could not enable STP", br->name); + return; + } + + /* Users must either set the port number with the "stp-port-num" + * configuration on all ports or none. If manual configuration + * is not done, then we allocate them sequentially. */ + port_num_counter = 0; + port_num_bitmap = bitmap_allocate(STP_MAX_PORTS); + HMAP_FOR_EACH (port, hmap_node, &br->ports) { + struct ofproto_port_stp_settings port_s; + struct iface *iface; + + port_configure_stp(br->ofproto, port, &port_s, + &port_num_counter, port_num_bitmap); + + /* As bonds are not supported, just apply configuration to + * all interfaces. */ + LIST_FOR_EACH (iface, port_elem, &port->ifaces) { + if (ofproto_port_set_stp(br->ofproto, iface->ofp_port, + &port_s)) { + VLOG_ERR("port %s: could not enable STP", port->name); + continue; + } + } + } + + if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS + && port_num_counter) { + VLOG_ERR("bridge %s: must manually configure all STP port " + "IDs or none, disabling", br->name); + ofproto_set_stp(br->ofproto, NULL); + } + bitmap_free(port_num_bitmap); + } +} + static bool bridge_has_bond_fake_iface(const struct bridge *br, const char *name) { @@ -773,75 +1133,109 @@ add_del_bridges(const struct ovsrec_open_vswitch *cfg) shash_destroy(&new_br); } -/* Delete each ofproto port on 'br' that doesn't have a corresponding "struct - * iface". - * - * The kernel will reject any attempt to add a given port to a datapath if that - * port already belongs to a different datapath, so we must do all port - * deletions before any port additions. */ static void -bridge_del_ofproto_ports(struct bridge *br) +iface_set_ofp_port(struct iface *iface, int ofp_port) { - struct ofproto_port_dump dump; - struct ofproto_port ofproto_port; + struct bridge *br = iface->port->bridge; - OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) { - const char *name = ofproto_port.name; - struct iface *iface; - const char *type; - int error; + assert(iface->ofp_port < 0 && ofp_port >= 0); + iface->ofp_port = ofp_port; + hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0)); + iface_set_ofport(iface->cfg, ofp_port); +} - /* Ignore the local port. We can't change it anyhow. */ - if (!strcmp(name, br->name)) { - continue; - } +/* Configures 'netdev' based on the "options" column in 'iface_cfg'. + * Returns 0 if successful, otherwise a positive errno value. */ +static int +iface_set_netdev_config(const struct ovsrec_interface *iface_cfg, + struct netdev *netdev) +{ + struct shash args; + int error; - /* Get the type that 'ofproto_port' should have (ordinarily the - * type of its corresponding iface) or NULL if it should be - * deleted. */ - iface = iface_lookup(br, name); - type = (iface ? iface->type - : bridge_has_bond_fake_iface(br, name) ? "internal" - : NULL); + shash_init(&args); + shash_from_ovs_idl_map(iface_cfg->key_options, + iface_cfg->value_options, + iface_cfg->n_options, &args); + error = netdev_set_config(netdev, &args); + shash_destroy(&args); - /* If it's the wrong type then delete the ofproto port. */ - if (type - && !strcmp(ofproto_port.type, type) - && (!iface || !iface->netdev - || !strcmp(netdev_get_type(iface->netdev), type))) { - continue; - } - error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port); - if (error) { - VLOG_WARN("bridge %s: failed to remove %s interface (%s)", - br->name, name, strerror(error)); - } - if (iface) { - netdev_close(iface->netdev); - iface->netdev = NULL; - } + if (error) { + VLOG_WARN("could not configure network device %s (%s)", + iface_cfg->name, strerror(error)); } + return error; } -static void -iface_set_ofp_port(struct iface *iface, int ofp_port) +/* This function determines whether 'ofproto_port', which is attached to + * br->ofproto's datapath, is one that we want in 'br'. + * + * If it is, it returns true, after creating an iface (if necessary), + * configuring the iface's netdev according to the iface's options, and setting + * iface's ofp_port member to 'ofproto_port->ofp_port'. + * + * If, on the other hand, 'port' should be removed, it returns false. The + * caller should later detach the port from br->ofproto. */ +static bool +bridge_refresh_one_ofp_port(struct bridge *br, + const struct ofproto_port *ofproto_port) { - struct bridge *br = iface->port->bridge; + const char *name = ofproto_port->name; + const char *type = ofproto_port->type; + uint16_t ofp_port = ofproto_port->ofp_port; - assert(iface->ofp_port < 0 && ofp_port >= 0); - iface->ofp_port = ofp_port; - hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0)); - iface_set_ofport(iface->cfg, ofp_port); + struct iface *iface = iface_lookup(br, name); + if (iface) { + /* Check that the name-to-number mapping is one-to-one. */ + if (iface->ofp_port >= 0) { + VLOG_WARN("bridge %s: interface %s reported twice", + br->name, name); + return false; + } else if (iface_from_ofp_port(br, ofp_port)) { + VLOG_WARN("bridge %s: interface %"PRIu16" reported twice", + br->name, ofp_port); + return false; + } + + /* There's a configured interface named 'name'. */ + if (strcmp(type, iface->type) + || iface_set_netdev_config(iface->cfg, iface->netdev)) { + /* It's the wrong type, or it's the right type but can't be + * configured as the user requested, so we must destroy it. */ + return false; + } else { + /* It's the right type and configured correctly. keep it. */ + iface_set_ofp_port(iface, ofp_port); + return true; + } + } else if (bridge_has_bond_fake_iface(br, name) + && !strcmp(type, "internal")) { + /* It's a bond fake iface. Keep it. */ + return true; + } else { + /* There's no configured interface named 'name', but there might be an + * interface of that name queued to be created. + * + * If there is, and it has the correct type, then try to configure it + * and add it. If that's successful, we'll keep it. Otherwise, we'll + * delete it and later try to re-add it. */ + struct if_cfg *if_cfg = if_cfg_lookup(br, name); + return (if_cfg + && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg)) + && iface_create(br, if_cfg, ofp_port)); + } } +/* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be + * consistent with the ofp_ports in "br->ofproto". */ static void bridge_refresh_ofp_port(struct bridge *br) { struct ofproto_port_dump dump; struct ofproto_port ofproto_port; - struct port *port; + struct port *port, *port_next; - /* Clear all the "ofp_port"es. */ + /* Clear each "struct iface"s ofp_port so we can get its correct value. */ hmap_clear(&br->ifaces); HMAP_FOR_EACH (port, hmap_node, &br->ports) { struct iface *iface; @@ -851,154 +1245,172 @@ bridge_refresh_ofp_port(struct bridge *br) } } - /* Obtain the correct "ofp_port"s from ofproto. */ + /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which + * already exist in the datapath and promote them to full fledged "struct + * iface"s. Mark ports in the datapath which don't belong as garbage. */ OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) { - struct iface *iface = iface_lookup(br, ofproto_port.name); - if (iface) { - if (iface->ofp_port >= 0) { - VLOG_WARN("bridge %s: interface %s reported twice", - br->name, ofproto_port.name); - } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) { - VLOG_WARN("bridge %s: interface %"PRIu16" reported twice", - br->name, ofproto_port.ofp_port); - } else { - iface_set_ofp_port(iface, ofproto_port.ofp_port); + if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) { + struct ofpp_garbage *garbage = xmalloc(sizeof *garbage); + garbage->ofp_port = ofproto_port.ofp_port; + list_push_front(&br->ofpp_garbage, &garbage->list_node); + } + } + + /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't + * deserve to have "struct iface"s. Demote these to "if_cfg"s so that + * later they can be added to ofproto. */ + HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) { + struct iface *iface, *iface_next; + + LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) { + if (iface->ofp_port < 0) { + bridge_queue_if_cfg(br, iface->cfg, port->cfg); + iface_destroy(iface); } } + + if (list_is_empty(&port->ifaces)) { + port_destroy(port); + } } } -/* Add an ofproto port for any "struct iface" that doesn't have one. - * Delete any "struct iface" for which this fails. - * Delete any "struct port" that thereby ends up with no ifaces. */ -static void -bridge_add_ofproto_ports(struct bridge *br) +/* Opens a network device for 'iface_cfg' and configures it. If '*ofp_portp' + * is negative, adds the network device to br->ofproto and stores the OpenFlow + * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp' + * untouched. + * + * If successful, returns 0 and stores the network device in '*netdevp'. On + * failure, returns a positive errno value and stores NULL in '*netdevp'. */ +static int +iface_do_create(const struct bridge *br, + const struct ovsrec_interface *iface_cfg, + const struct ovsrec_port *port_cfg, + int *ofp_portp, struct netdev **netdevp) { - struct port *port, *next_port; + struct netdev *netdev; + int error; - HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) { - struct iface *iface, *next_iface; - struct ofproto_port ofproto_port; + error = netdev_open(iface_cfg->name, + iface_get_type(iface_cfg, br->cfg), &netdev); + if (error) { + VLOG_WARN("could not open network device %s (%s)", + iface_cfg->name, strerror(error)); + goto error; + } - LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) { - int error; + error = iface_set_netdev_config(iface_cfg, netdev); + if (error) { + goto error; + } - /* Open the netdev. */ - if (!iface->netdev) { - error = netdev_open(iface->name, iface->type, &iface->netdev); - if (error) { - VLOG_WARN("could not open network device %s (%s)", - iface->name, strerror(error)); - } - } else { - error = 0; - } + if (*ofp_portp < 0) { + uint16_t ofp_port; - /* Configure the netdev. */ - if (iface->netdev) { - struct shash args; + error = ofproto_port_add(br->ofproto, netdev, &ofp_port); + if (error) { + goto error; + } + *ofp_portp = ofp_port; - shash_init(&args); - shash_from_ovs_idl_map(iface->cfg->key_options, - iface->cfg->value_options, - iface->cfg->n_options, &args); - error = netdev_set_config(iface->netdev, &args); - shash_destroy(&args); + VLOG_INFO("bridge %s: added interface %s on port %d", + br->name, iface_cfg->name, *ofp_portp); + } else { + VLOG_DBG("bridge %s: interface %s is on port %d", + br->name, iface_cfg->name, *ofp_portp); + } - if (error) { - VLOG_WARN("could not configure network device %s (%s)", - iface->name, strerror(error)); - netdev_close(iface->netdev); - iface->netdev = NULL; - } - } + if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) { + netdev_turn_flags_on(netdev, NETDEV_UP, true); + } - /* Add the port, if necessary. */ - if (iface->netdev && iface->ofp_port < 0) { - uint16_t ofp_port; - int error; + *netdevp = netdev; + return 0; - error = ofproto_port_add(br->ofproto, iface->netdev, - &ofp_port); - if (!error) { - iface_set_ofp_port(iface, ofp_port); - } else { - netdev_close(iface->netdev); - iface->netdev = NULL; - } - } +error: + *netdevp = NULL; + netdev_close(netdev); + return error; +} - /* Populate stats columns in new Interface rows. */ - if (iface->netdev && !iface->cfg->mtu) { - iface_refresh_stats(iface); - iface_refresh_status(iface); - } +/* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow + * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is + * automatically allocated for the iface. Takes ownership of and + * deallocates 'if_cfg'. + * + * Return true if an iface is successfully created, false otherwise. */ +static bool +iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port) +{ + const struct ovsrec_interface *iface_cfg = if_cfg->cfg; + const struct ovsrec_port *port_cfg = if_cfg->parent; - /* Delete the iface if we failed. */ - if (iface->netdev && iface->ofp_port >= 0) { - VLOG_DBG("bridge %s: interface %s is on port %d", - br->name, iface->name, iface->ofp_port); - } else { - if (iface->netdev) { - VLOG_ERR("bridge %s: missing %s interface, dropping", - br->name, iface->name); - } else { - /* We already reported a related error, don't bother - * duplicating it. */ - } - iface_clear_db_record(iface->cfg); - iface_destroy(iface); - } - } - if (list_is_empty(&port->ifaces)) { - VLOG_WARN("%s port has no interfaces, dropping", port->name); - port_destroy(port); - continue; - } + struct netdev *netdev; + struct iface *iface; + struct port *port; + int error; - /* Add bond fake iface if necessary. */ - if (port_is_bond_fake_iface(port)) { - if (ofproto_port_query_by_name(br->ofproto, port->name, - &ofproto_port)) { - struct netdev *netdev; - int error; + /* Get rid of 'if_cfg' itself. We already copied out the interesting + * bits. */ + hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); + free(if_cfg); - error = netdev_open(port->name, "internal", &netdev); - if (!error) { - ofproto_port_add(br->ofproto, netdev, NULL); - netdev_close(netdev); - } else { - VLOG_WARN("could not open network device %s (%s)", - port->name, strerror(error)); - } + /* Do the bits that can fail up front. */ + assert(!iface_lookup(br, iface_cfg->name)); + error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev); + if (error) { + iface_clear_db_record(iface_cfg); + return false; + } + + /* Get or create the port structure. */ + port = port_lookup(br, port_cfg->name); + if (!port) { + port = port_create(br, port_cfg); + } + + /* Create the iface structure. */ + iface = xzalloc(sizeof *iface); + list_push_back(&port->ifaces, &iface->port_elem); + hmap_insert(&br->iface_by_name, &iface->name_node, + hash_string(iface_cfg->name, 0)); + iface->port = port; + iface->name = xstrdup(iface_cfg->name); + iface->ofp_port = -1; + iface->netdev = netdev; + iface->type = iface_get_type(iface_cfg, br->cfg); + iface->cfg = iface_cfg; + + iface_set_ofp_port(iface, ofp_port); + + /* Populate initial status in database. */ + iface_refresh_stats(iface); + iface_refresh_status(iface); + + /* Add bond fake iface if necessary. */ + if (port_is_bond_fake_iface(port)) { + struct ofproto_port ofproto_port; + + if (ofproto_port_query_by_name(br->ofproto, port->name, + &ofproto_port)) { + struct netdev *netdev; + int error; + + error = netdev_open(port->name, "internal", &netdev); + if (!error) { + ofproto_port_add(br->ofproto, netdev, NULL); + netdev_close(netdev); } else { - /* Already exists, nothing to do. */ - ofproto_port_destroy(&ofproto_port); + VLOG_WARN("could not open network device %s (%s)", + port->name, strerror(error)); } + } else { + /* Already exists, nothing to do. */ + ofproto_port_destroy(&ofproto_port); } } -} -static const char * -get_ovsrec_key_value(const struct ovsdb_idl_row *row, - const struct ovsdb_idl_column *column, - const char *key) -{ - const struct ovsdb_datum *datum; - union ovsdb_atom atom; - unsigned int idx; - - datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING); - atom.string = (char *) key; - idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING); - return idx == UINT_MAX ? NULL : datum->values[idx].string; -} - -static const char * -bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key) -{ - return get_ovsrec_key_value(&br_cfg->header_, - &ovsrec_bridge_col_other_config, key); + return true; } /* Set Flow eviction threshold */ @@ -1008,7 +1420,10 @@ bridge_configure_flow_eviction_threshold(struct bridge *br) const char *threshold_str; unsigned threshold; - threshold_str = bridge_get_other_config(br->cfg, "flow-eviction-threshold"); + threshold_str = + ovsrec_bridge_get_other_config_value(br->cfg, + "flow-eviction-threshold", + NULL); if (threshold_str) { threshold = strtoul(threshold_str, NULL, 10); } else { @@ -1024,26 +1439,46 @@ bridge_configure_forward_bpdu(struct bridge *br) const char *forward_bpdu_str; bool forward_bpdu = false; - forward_bpdu_str = bridge_get_other_config(br->cfg, "forward-bpdu"); + forward_bpdu_str = ovsrec_bridge_get_other_config_value(br->cfg, + "forward-bpdu", + NULL); if (forward_bpdu_str && !strcmp(forward_bpdu_str, "true")) { forward_bpdu = true; } ofproto_set_forward_bpdu(br->ofproto, forward_bpdu); } +/* Set MAC aging time for 'br'. */ +static void +bridge_configure_mac_idle_time(struct bridge *br) +{ + const char *idle_time_str; + int idle_time; + + idle_time_str = ovsrec_bridge_get_other_config_value(br->cfg, + "mac-aging-time", + NULL); + idle_time = (idle_time_str && atoi(idle_time_str) + ? atoi(idle_time_str) + : MAC_ENTRY_DEFAULT_IDLE_TIME); + ofproto_set_mac_idle_time(br->ofproto, idle_time); +} + static void bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface **hw_addr_iface) { + struct hmapx mirror_output_ports; const char *hwaddr; struct port *port; bool found_addr = false; int error; + int i; *hw_addr_iface = NULL; /* Did the user request a particular MAC? */ - hwaddr = bridge_get_other_config(br->cfg, "hwaddr"); + hwaddr = ovsrec_bridge_get_other_config_value(br->cfg, "hwaddr", NULL); if (hwaddr && eth_addr_from_string(hwaddr, ea)) { if (eth_addr_is_multicast(ea)) { VLOG_ERR("bridge %s: cannot set MAC address to multicast " @@ -1055,6 +1490,18 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], } } + /* Mirror output ports don't participate in picking the local hardware + * address. ofproto can't help us find out whether a given port is a + * mirror output because we haven't configured mirrors yet, so we need to + * accumulate them ourselves. */ + hmapx_init(&mirror_output_ports); + for (i = 0; i < br->cfg->n_mirrors; i++) { + struct ovsrec_mirror *m = br->cfg->mirrors[i]; + if (m->output_port) { + hmapx_add(&mirror_output_ports, m->output_port); + } + } + /* Otherwise choose the minimum non-local MAC address among all of the * interfaces. */ HMAP_FOR_EACH (port, hmap_node, &br->ports) { @@ -1063,7 +1510,7 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], struct iface *iface; /* Mirror output ports don't participate. */ - if (ofproto_is_mirror_output_bundle(br->ofproto, port)) { + if (hmapx_contains(&mirror_output_ports, port->cfg)) { continue; } @@ -1121,11 +1568,14 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN], VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); } else { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10); memcpy(ea, br->default_ea, ETH_ADDR_LEN); *hw_addr_iface = NULL; - VLOG_WARN("bridge %s: using default bridge Ethernet " - "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); + VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet " + "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea)); } + + hmapx_destroy(&mirror_output_ports); } /* Choose and returns the datapath ID for bridge 'br' given that the bridge @@ -1153,38 +1603,16 @@ bridge_pick_datapath_id(struct bridge *br, const char *datapath_id; uint64_t dpid; - datapath_id = bridge_get_other_config(br->cfg, "datapath-id"); + datapath_id = ovsrec_bridge_get_other_config_value(br->cfg, "datapath-id", + NULL); if (datapath_id && dpid_from_string(datapath_id, &dpid)) { return dpid; } - if (hw_addr_iface) { - int vlan; - if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) { - /* - * A bridge whose MAC address is taken from a VLAN network device - * (that is, a network device created with vconfig(8) or similar - * tool) will have the same MAC address as a bridge on the VLAN - * device's physical network device. - * - * Handle this case by hashing the physical network device MAC - * along with the VLAN identifier. - */ - uint8_t buf[ETH_ADDR_LEN + 2]; - memcpy(buf, bridge_ea, ETH_ADDR_LEN); - buf[ETH_ADDR_LEN] = vlan >> 8; - buf[ETH_ADDR_LEN + 1] = vlan; - return dpid_from_hash(buf, sizeof buf); - } else { - /* - * Assume that this bridge's MAC address is unique, since it - * doesn't fit any of the cases we handle specially. - */ - } - } else { + if (!hw_addr_iface) { /* * A purely internal bridge, that is, one that has no non-virtual - * network devices on it at all, is more difficult because it has no + * network devices on it at all, is difficult because it has no * natural unique identifier at all. * * When the host is a XenServer, we handle this case by hashing the @@ -1224,8 +1652,8 @@ iface_refresh_status(struct iface *iface) { struct shash sh; + enum netdev_features current; enum netdev_flags flags; - uint32_t current; int64_t bps; int mtu; int64_t mtu_64; @@ -1237,7 +1665,7 @@ iface_refresh_status(struct iface *iface) shash_init(&sh); - if (!netdev_get_status(iface->netdev, &sh)) { + if (!netdev_get_drv_info(iface->netdev, &sh)) { size_t n; char **keys, **values; @@ -1254,7 +1682,8 @@ iface_refresh_status(struct iface *iface) error = netdev_get_flags(iface->netdev, &flags); if (!error) { - ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down"); + ovsrec_interface_set_admin_state(iface->cfg, + flags & NETDEV_UP ? "up" : "down"); } else { ovsrec_interface_set_admin_state(iface->cfg, NULL); @@ -1274,11 +1703,6 @@ iface_refresh_status(struct iface *iface) ovsrec_interface_set_link_speed(iface->cfg, NULL, 0); } - ovsrec_interface_set_link_state(iface->cfg, - (netdev_get_carrier(iface->netdev) - ? "up" - : "down")); - error = netdev_get_mtu(iface->netdev, &mtu); if (!error) { mtu_64 = mtu; @@ -1297,6 +1721,7 @@ iface_refresh_cfm_stats(struct iface *iface) int fault, error; const uint64_t *rmps; size_t n_rmps; + int health; if (iface_is_synthetic(iface)) { return; @@ -1305,10 +1730,23 @@ iface_refresh_cfm_stats(struct iface *iface) fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto, iface->ofp_port); if (fault >= 0) { + const char *reasons[CFM_FAULT_N_REASONS]; bool fault_bool = fault; + size_t i, j; + + j = 0; + for (i = 0; i < CFM_FAULT_N_REASONS; i++) { + int reason = 1 << i; + if (fault & reason) { + reasons[j++] = cfm_fault_reason_to_str(reason); + } + } + ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1); + ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j); } else { ovsrec_interface_set_cfm_fault(cfg, NULL, 0); + ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0); } error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto, @@ -1319,6 +1757,15 @@ iface_refresh_cfm_stats(struct iface *iface) } else { ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0); } + + health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto, + iface->ofp_port); + if (health >= 0) { + int64_t cfm_health = health; + ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1); + } else { + ovsrec_interface_set_cfm_health(cfg, NULL, 0); + } } static void @@ -1361,19 +1808,113 @@ iface_refresh_stats(struct iface *iface) #undef IFACE_STAT assert(i == ARRAY_SIZE(keys)); - ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys)); + ovsrec_interface_set_statistics(iface->cfg, keys, values, + ARRAY_SIZE(keys)); #undef IFACE_STATS } +static void +br_refresh_stp_status(struct bridge *br) +{ + struct ofproto *ofproto = br->ofproto; + struct ofproto_stp_status status; + char *keys[3], *values[3]; + size_t i; + + if (ofproto_get_stp_status(ofproto, &status)) { + return; + } + + if (!status.enabled) { + ovsrec_bridge_set_status(br->cfg, NULL, NULL, 0); + return; + } + + keys[0] = "stp_bridge_id", + values[0] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.bridge_id)); + keys[1] = "stp_designated_root", + values[1] = xasprintf(STP_ID_FMT, STP_ID_ARGS(status.designated_root)); + keys[2] = "stp_root_path_cost", + values[2] = xasprintf("%d", status.root_path_cost); + + ovsrec_bridge_set_status(br->cfg, keys, values, ARRAY_SIZE(values)); + + for (i = 0; i < ARRAY_SIZE(values); i++) { + free(values[i]); + } +} + +static void +port_refresh_stp_status(struct port *port) +{ + struct ofproto *ofproto = port->bridge->ofproto; + struct iface *iface; + struct ofproto_port_stp_status status; + char *keys[4]; + char *str_values[4]; + int64_t int_values[3]; + size_t i; + + if (port_is_synthetic(port)) { + return; + } + + /* STP doesn't currently support bonds. */ + if (!list_is_singleton(&port->ifaces)) { + ovsrec_port_set_status(port->cfg, NULL, NULL, 0); + return; + } + + iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem); + + if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) { + return; + } + + if (!status.enabled) { + ovsrec_port_set_status(port->cfg, NULL, NULL, 0); + ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0); + return; + } + + /* Set Status column. */ + keys[0] = "stp_port_id"; + str_values[0] = xasprintf(STP_PORT_ID_FMT, status.port_id); + keys[1] = "stp_state"; + str_values[1] = xstrdup(stp_state_name(status.state)); + keys[2] = "stp_sec_in_state"; + str_values[2] = xasprintf("%u", status.sec_in_state); + keys[3] = "stp_role"; + str_values[3] = xstrdup(stp_role_name(status.role)); + + ovsrec_port_set_status(port->cfg, keys, str_values, + ARRAY_SIZE(str_values)); + + for (i = 0; i < ARRAY_SIZE(str_values); i++) { + free(str_values[i]); + } + + /* Set Statistics column. */ + keys[0] = "stp_tx_count"; + int_values[0] = status.tx_count; + keys[1] = "stp_rx_count"; + int_values[1] = status.rx_count; + keys[2] = "stp_error_count"; + int_values[2] = status.error_count; + + ovsrec_port_set_statistics(port->cfg, keys, int_values, + ARRAY_SIZE(int_values)); +} + static bool enable_system_stats(const struct ovsrec_open_vswitch *cfg) { const char *enable; /* Use other-config:enable-system-stats by preference. */ - enable = get_ovsrec_key_value(&cfg->header_, - &ovsrec_open_vswitch_col_other_config, - "enable-statistics"); + enable = ovsrec_open_vswitch_get_other_config_value(cfg, + "enable-statistics", + NULL); if (enable) { return !strcmp(enable, "true"); } @@ -1473,43 +2014,56 @@ refresh_cfm_stats(void) } } +/* Performs periodic activity required by bridges that needs to be done with + * the least possible latency. + * + * It makes sense to call this function a couple of times per poll loop, to + * provide a significant performance boost on some benchmarks with ofprotos + * that use the ofproto-dpif implementation. */ +void +bridge_run_fast(void) +{ + struct bridge *br; + + HMAP_FOR_EACH (br, node, &all_bridges) { + ofproto_run_fast(br->ofproto); + } +} + void bridge_run(void) { + static const struct ovsrec_open_vswitch null_cfg; const struct ovsrec_open_vswitch *cfg; + struct ovsdb_idl_txn *reconf_txn = NULL; - bool datapath_destroyed; - bool database_changed; + bool vlan_splinters_changed; struct bridge *br; /* (Re)configure if necessary. */ - database_changed = ovsdb_idl_run(idl); - if (ovsdb_idl_is_lock_contended(idl)) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); - struct bridge *br, *next_br; + if (!reconfiguring) { + ovsdb_idl_run(idl); - VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, " - "disabling this process until it goes away"); + if (ovsdb_idl_is_lock_contended(idl)) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + struct bridge *br, *next_br; - HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) { - bridge_destroy(br); + VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, " + "disabling this process until it goes away"); + + HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) { + bridge_destroy(br); + } + return; + } else if (!ovsdb_idl_has_lock(idl)) { + return; } - return; - } else if (!ovsdb_idl_has_lock(idl)) { - return; } cfg = ovsrec_open_vswitch_first(idl); /* Let each bridge do the work that it needs to do. */ - datapath_destroyed = false; HMAP_FOR_EACH (br, node, &all_bridges) { - int error = ofproto_run(br->ofproto); - if (error) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, " - "forcing reconfiguration", br->name); - datapath_destroyed = true; - } + ofproto_run(br->ofproto); } /* Re-configure SSL. We do this on every trip through the main loop, @@ -1525,24 +2079,51 @@ bridge_run(void) stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert); } - if (database_changed || datapath_destroyed) { - if (cfg) { - struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl); + if (!reconfiguring) { + /* If VLAN splinters are in use, then we need to reconfigure if VLAN + * usage has changed. */ + vlan_splinters_changed = false; + if (vlan_splinters_enabled_anywhere) { + HMAP_FOR_EACH (br, node, &all_bridges) { + if (ofproto_has_vlan_usage_changed(br->ofproto)) { + vlan_splinters_changed = true; + break; + } + } + } - bridge_reconfigure(cfg); + if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) { + idl_seqno = ovsdb_idl_get_seqno(idl); + if (cfg) { + reconf_txn = ovsdb_idl_txn_create(idl); + bridge_reconfigure(cfg); + } else { + /* We still need to reconfigure to avoid dangling pointers to + * now-destroyed ovsrec structures inside bridge data. */ + bridge_reconfigure(&null_cfg); + } + } + } - ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg); - ovsdb_idl_txn_commit(txn); - ovsdb_idl_txn_destroy(txn); /* XXX */ + if (reconfiguring) { + if (cfg) { + if (!reconf_txn) { + reconf_txn = ovsdb_idl_txn_create(idl); + } + if (bridge_reconfigure_continue(cfg)) { + ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg); + } } else { - /* We still need to reconfigure to avoid dangling pointers to - * now-destroyed ovsrec structures inside bridge data. */ - static const struct ovsrec_open_vswitch null_cfg; - - bridge_reconfigure(&null_cfg); + bridge_reconfigure_continue(&null_cfg); } } + if (reconf_txn) { + ovsdb_idl_txn_commit(reconf_txn); + ovsdb_idl_txn_destroy(reconf_txn); + reconf_txn = NULL; + } + /* Refresh system and interface stats if necessary. */ if (time_msec() >= stats_timer) { if (cfg) { @@ -1551,6 +2132,7 @@ bridge_run(void) txn = ovsdb_idl_txn_create(idl); HMAP_FOR_EACH (br, node, &all_bridges) { struct port *port; + struct mirror *m; HMAP_FOR_EACH (port, hmap_node, &br->ports) { struct iface *iface; @@ -1560,6 +2142,11 @@ bridge_run(void) iface_refresh_status(iface); } } + + HMAP_FOR_EACH (m, hmap_node, &br->mirrors) { + mirror_refresh_stats(m); + } + } refresh_system_stats(cfg); refresh_controller_status(); @@ -1576,8 +2163,17 @@ bridge_run(void) txn = ovsdb_idl_txn_create(idl); HMAP_FOR_EACH (br, node, &all_bridges) { struct iface *iface; + struct port *port; + + br_refresh_stp_status(br); + + HMAP_FOR_EACH (port, hmap_node, &br->ports) { + port_refresh_stp_status(port); + } HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) { + const char *link_state; + int64_t link_resets; int current; if (iface_is_synthetic(iface)) { @@ -1592,6 +2188,12 @@ bridge_run(void) } else { ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0); } + + link_state = netdev_get_carrier(iface->netdev) ? "up" : "down"; + ovsrec_interface_set_link_state(iface->cfg, link_state); + + link_resets = netdev_get_carrier_resets(iface->netdev); + ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1); } } @@ -1608,6 +2210,11 @@ void bridge_wait(void) { ovsdb_idl_wait(idl); + + if (reconfiguring) { + poll_immediate_wake(); + } + if (!hmap_is_empty(&all_bridges)) { struct bridge *br; @@ -1672,8 +2279,8 @@ qos_unixctl_show_cb(unsigned int queue_id, } static void -qos_unixctl_show(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED, + const char *argv[], void *aux OVS_UNUSED) { struct ds ds = DS_EMPTY_INITIALIZER; struct shash sh = SHASH_INITIALIZER(&sh); @@ -1683,9 +2290,9 @@ qos_unixctl_show(struct unixctl_conn *conn, struct qos_unixctl_show_cbdata data; int error; - iface = iface_find(args); + iface = iface_find(argv[1]); if (!iface) { - unixctl_command_reply(conn, 501, "no such interface"); + unixctl_command_reply_error(conn, "no such interface"); return; } @@ -1705,10 +2312,10 @@ qos_unixctl_show(struct unixctl_conn *conn, if (error) { ds_put_format(&ds, "failed to dump queues: %s", strerror(error)); } - unixctl_command_reply(conn, 200, ds_cstr(&ds)); + unixctl_command_reply(conn, ds_cstr(&ds)); } else { ds_put_format(&ds, "QoS not configured on %s\n", iface->name); - unixctl_command_reply(conn, 501, ds_cstr(&ds)); + unixctl_command_reply_error(conn, ds_cstr(&ds)); } shash_destroy_free_data(&sh); @@ -1738,6 +2345,9 @@ bridge_create(const struct ovsrec_bridge *br_cfg) hmap_init(&br->iface_by_name); hmap_init(&br->mirrors); + hmap_init(&br->if_cfg_todo); + list_init(&br->ofpp_garbage); + hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0)); } @@ -1747,6 +2357,8 @@ bridge_destroy(struct bridge *br) if (br) { struct mirror *mirror, *next_mirror; struct port *port, *next_port; + struct if_cfg *if_cfg, *next_if_cfg; + struct ofpp_garbage *garbage, *next_garbage; HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) { port_destroy(port); @@ -1754,12 +2366,23 @@ bridge_destroy(struct bridge *br) HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) { mirror_destroy(mirror); } + HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) { + hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node); + free(if_cfg); + } + LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node, + &br->ofpp_garbage) { + list_remove(&garbage->list_node); + free(garbage); + } + hmap_remove(&all_bridges, &br->node); ofproto_destroy(br->ofproto); hmap_destroy(&br->ifaces); hmap_destroy(&br->ports); hmap_destroy(&br->iface_by_name); hmap_destroy(&br->mirrors); + hmap_destroy(&br->if_cfg_todo); free(br->name); free(br->type); free(br); @@ -1782,22 +2405,22 @@ bridge_lookup(const char *name) /* Handle requests for a listing of all flows known by the OpenFlow * stack, including those normally hidden. */ static void -bridge_unixctl_dump_flows(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED, + const char *argv[], void *aux OVS_UNUSED) { struct bridge *br; struct ds results; - br = bridge_lookup(args); + br = bridge_lookup(argv[1]); if (!br) { - unixctl_command_reply(conn, 501, "Unknown bridge"); + unixctl_command_reply_error(conn, "Unknown bridge"); return; } ds_init(&results); ofproto_get_all_flows(br->ofproto, &results); - unixctl_command_reply(conn, 200, ds_cstr(&results)); + unixctl_command_reply(conn, ds_cstr(&results)); ds_destroy(&results); } @@ -1805,14 +2428,14 @@ bridge_unixctl_dump_flows(struct unixctl_conn *conn, * connections and reconnect. If BRIDGE is not specified, then all bridges * drop their controller connections and reconnect. */ static void -bridge_unixctl_reconnect(struct unixctl_conn *conn, - const char *args, void *aux OVS_UNUSED) +bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc, + const char *argv[], void *aux OVS_UNUSED) { struct bridge *br; - if (args[0] != '\0') { - br = bridge_lookup(args); + if (argc > 1) { + br = bridge_lookup(argv[1]); if (!br) { - unixctl_command_reply(conn, 501, "Unknown bridge"); + unixctl_command_reply_error(conn, "Unknown bridge"); return; } ofproto_reconnect_controllers(br->ofproto); @@ -1821,7 +2444,7 @@ bridge_unixctl_reconnect(struct unixctl_conn *conn, ofproto_reconnect_controllers(br->ofproto); } } - unixctl_command_reply(conn, 200, NULL); + unixctl_command_reply(conn, NULL); } static size_t @@ -1845,16 +2468,33 @@ bridge_get_controllers(const struct bridge *br, return n_controllers; } -/* Adds and deletes "struct port"s and "struct iface"s under 'br' to match - * those configured in 'br->cfg'. */ static void -bridge_add_del_ports(struct bridge *br) +bridge_queue_if_cfg(struct bridge *br, + const struct ovsrec_interface *cfg, + const struct ovsrec_port *parent) +{ + struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg); + + if_cfg->cfg = cfg; + if_cfg->parent = parent; + hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node, + hash_string(if_cfg->cfg->name, 0)); +} + +/* Deletes "struct port"s and "struct iface"s under 'br' which aren't + * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which + * 'br' needs to complete its configuration. */ +static void +bridge_add_del_ports(struct bridge *br, + const unsigned long int *splinter_vlans) { + struct shash_node *port_node; struct port *port, *next; - struct shash_node *node; struct shash new_ports; size_t i; + assert(hmap_is_empty(&br->if_cfg_todo)); + /* Collect new ports. */ shash_init(&new_ports); for (i = 0; i < br->cfg->n_ports; i++) { @@ -1881,9 +2521,12 @@ bridge_add_del_ports(struct bridge *br) shash_add(&new_ports, br->name, &br->synth_local_port); } + if (splinter_vlans) { + add_vlan_splinter_ports(br, splinter_vlans, &new_ports); + } + /* Get rid of deleted ports. - * Get rid of deleted interfaces on ports that still exist. - * Update 'cfg' of ports that still exist. */ + * Get rid of deleted interfaces on ports that still exist. */ HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) { port->cfg = shash_find_data(&new_ports, port->name); if (!port->cfg) { @@ -1893,21 +2536,25 @@ bridge_add_del_ports(struct bridge *br) } } - /* Create new ports. - * Add new interfaces to existing ports. */ - SHASH_FOR_EACH (node, &new_ports) { - struct port *port = port_lookup(br, node->name); - if (!port) { - struct ovsrec_port *cfg = node->data; - port = port_create(br, cfg); - } - port_add_ifaces(port); - if (list_is_empty(&port->ifaces)) { - VLOG_WARN("bridge %s: port %s has no interfaces, dropping", - br->name, port->name); - port_destroy(port); + /* Update iface->cfg and iface->type in interfaces that still exist. + * Add new interfaces to creation queue. */ + SHASH_FOR_EACH (port_node, &new_ports) { + const struct ovsrec_port *port = port_node->data; + size_t i; + + for (i = 0; i < port->n_interfaces; i++) { + const struct ovsrec_interface *cfg = port->interfaces[i]; + struct iface *iface = iface_lookup(br, cfg->name); + + if (iface) { + iface->cfg = cfg; + iface->type = iface_get_type(cfg, br->cfg); + } else { + bridge_queue_if_cfg(br, cfg, port); + } } } + shash_destroy(&new_ports); } @@ -1925,6 +2572,7 @@ bridge_ofproto_controller_for_mgmt(const struct bridge *br, oc->band = OFPROTO_OUT_OF_BAND; oc->rate_limit = 0; oc->burst_limit = 0; + oc->enable_async_msgs = true; } /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */ @@ -1932,6 +2580,8 @@ static void bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, struct ofproto_controller *oc) { + const char *config_str; + oc->target = c->target; oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8; oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5; @@ -1940,6 +2590,18 @@ bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c, oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0; oc->burst_limit = (c->controller_burst_limit ? *c->controller_burst_limit : 0); + oc->enable_async_msgs = (!c->enable_async_messages + || *c->enable_async_messages); + config_str = ovsrec_controller_get_other_config_value(c, "dscp", NULL); + + oc->dscp = DSCP_DEFAULT; + if (config_str) { + int dscp = atoi(config_str); + + if (dscp >= 0 && dscp <= 63) { + oc->dscp = dscp; + } + } } /* Configures the IP stack for 'br''s local interface properly according to the @@ -1986,6 +2648,26 @@ bridge_configure_local_iface_netdev(struct bridge *br, } } +/* Returns true if 'a' and 'b' are the same except that any number of slashes + * in either string are treated as equal to any number of slashes in the other, + * e.g. "x///y" is equal to "x/y". */ +static bool +equal_pathnames(const char *a, const char *b) +{ + while (*a == *b) { + if (*a == '/') { + a += strspn(a, "/"); + b += strspn(b, "/"); + } else if (*a == '\0') { + return true; + } else { + a++; + b++; + } + } + return false; +} + static void bridge_configure_remotes(struct bridge *br, const struct sockaddr_in *managers, size_t n_managers) @@ -2004,13 +2686,17 @@ bridge_configure_remotes(struct bridge *br, size_t i; /* Check if we should disable in-band control on this bridge. */ - disable_ib_str = bridge_get_other_config(br->cfg, "disable-in-band"); + disable_ib_str = ovsrec_bridge_get_other_config_value(br->cfg, + "disable-in-band", + NULL); if (disable_ib_str && !strcmp(disable_ib_str, "true")) { disable_in_band = true; } /* Set OpenFlow queue ID for in-band control. */ - queue_id_str = bridge_get_other_config(br->cfg, "in-band-queue"); + queue_id_str = ovsrec_bridge_get_other_config_value(br->cfg, + "in-band-queue", + NULL); queue_id = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1; ofproto_set_in_band_queue(br->ofproto, queue_id); @@ -2032,13 +2718,26 @@ bridge_configure_remotes(struct bridge *br, if (!strncmp(c->target, "punix:", 6) || !strncmp(c->target, "unix:", 5)) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + char *whitelist; + + whitelist = xasprintf("unix:%s/%s.controller", + ovs_rundir(), br->name); + if (!equal_pathnames(c->target, whitelist)) { + /* Prevent remote ovsdb-server users from accessing arbitrary + * Unix domain sockets and overwriting arbitrary local + * files. */ + VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket " + "controller \"%s\" due to possibility for remote " + "exploit. Instead, specify whitelisted \"%s\" or " + "connect to \"unix:%s/%s.mgmt\" (which is always " + "available without special configuration).", + br->name, c->target, whitelist, + ovs_rundir(), br->name); + free(whitelist); + continue; + } - /* Prevent remote ovsdb-server users from accessing arbitrary Unix - * domain sockets and overwriting arbitrary local files. */ - VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket " - "controller \"%s\" due to possibility for remote " - "exploit", br->name, c->target); - continue; + free(whitelist); } bridge_configure_local_iface_netdev(br, c); @@ -2071,6 +2770,66 @@ bridge_configure_remotes(struct bridge *br, sset_destroy(&snoops); } } + +static void +bridge_configure_tables(struct bridge *br) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + int n_tables; + int i, j; + + n_tables = ofproto_get_n_tables(br->ofproto); + j = 0; + for (i = 0; i < n_tables; i++) { + struct ofproto_table_settings s; + + s.name = NULL; + s.max_flows = UINT_MAX; + s.groups = NULL; + s.n_groups = 0; + + if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) { + struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++]; + + s.name = cfg->name; + if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) { + s.max_flows = *cfg->flow_limit; + } + if (cfg->overflow_policy + && !strcmp(cfg->overflow_policy, "evict")) { + size_t k; + + s.groups = xmalloc(cfg->n_groups * sizeof *s.groups); + for (k = 0; k < cfg->n_groups; k++) { + const char *string = cfg->groups[k]; + char *msg; + + msg = mf_parse_subfield__(&s.groups[k], &string); + if (msg) { + VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing " + "'groups' (%s)", br->name, i, msg); + free(msg); + } else if (*string) { + VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' " + "element '%s' contains trailing garbage", + br->name, i, cfg->groups[k]); + } else { + s.n_groups++; + } + } + } + } + + ofproto_configure_table(br->ofproto, i, &s); + + free(s.groups); + } + for (; j < br->cfg->n_flow_tables; j++) { + VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table " + "%"PRId64" not supported by this datapath", br->name, + br->cfg->key_flow_tables[j]); + } +} /* Port functions. */ @@ -2086,34 +2845,9 @@ port_create(struct bridge *br, const struct ovsrec_port *cfg) list_init(&port->ifaces); hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0)); - - VLOG_INFO("created port %s on bridge %s", port->name, br->name); - return port; } -static const char * -get_port_other_config(const struct ovsrec_port *port, const char *key, - const char *default_value) -{ - const char *value; - - value = get_ovsrec_key_value(&port->header_, &ovsrec_port_col_other_config, - key); - return value ? value : default_value; -} - -static const char * -get_interface_other_config(const struct ovsrec_interface *iface, - const char *key, const char *default_value) -{ - const char *value; - - value = get_ovsrec_key_value(&iface->header_, - &ovsrec_interface_col_other_config, key); - return value ? value : default_value; -} - /* Deletes interfaces from 'port' that are no longer configured for it. */ static void port_del_ifaces(struct port *port) @@ -2126,7 +2860,7 @@ port_del_ifaces(struct port *port) sset_init(&new_ifaces); for (i = 0; i < port->cfg->n_interfaces; i++) { const char *name = port->cfg->interfaces[i]->name; - const char *type = port->cfg->interfaces[i]->name; + const char *type = port->cfg->interfaces[i]->type; if (strcmp(type, "null")) { sset_add(&new_ifaces, name); } @@ -2142,51 +2876,6 @@ port_del_ifaces(struct port *port) sset_destroy(&new_ifaces); } -/* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of - * existing ones. */ -static void -port_add_ifaces(struct port *port) -{ - struct shash new_ifaces; - struct shash_node *node; - size_t i; - - /* Collect new ifaces. */ - shash_init(&new_ifaces); - for (i = 0; i < port->cfg->n_interfaces; i++) { - const struct ovsrec_interface *cfg = port->cfg->interfaces[i]; - if (strcmp(cfg->type, "null") - && !shash_add_once(&new_ifaces, cfg->name, cfg)) { - VLOG_WARN("port %s: %s specified twice as port interface", - port->name, cfg->name); - iface_clear_db_record(cfg); - } - } - - /* Create new interfaces. - * Update interface types and 'cfg' members. */ - SHASH_FOR_EACH (node, &new_ifaces) { - const struct ovsrec_interface *cfg = node->data; - const char *iface_name = node->name; - struct iface *iface; - - iface = iface_lookup(port->bridge, iface_name); - if (!iface) { - iface = iface_create(port, cfg); - } else { - iface->cfg = cfg; - } - - /* Determine interface type. The local port always has type - * "internal". Other ports take their type from the database and - * default to "system" if none is specified. */ - iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal" - : cfg->type[0] ? cfg->type - : "system"); - } - shash_destroy(&new_ifaces); -} - static void port_destroy(struct port *port) { @@ -2203,9 +2892,6 @@ port_destroy(struct port *port) } hmap_remove(&br->ports, &port->hmap_node); - - VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name); - free(port->name); free(port); } @@ -2250,8 +2936,7 @@ enable_lacp(struct port *port, bool *activep) static struct lacp_settings * port_configure_lacp(struct port *port, struct lacp_settings *s) { - const char *lacp_time; - long long int custom_time; + const char *lacp_time, *system_id; int priority; if (!enable_lacp(port, &s->active)) { @@ -2259,33 +2944,36 @@ port_configure_lacp(struct port *port, struct lacp_settings *s) } s->name = port->name; - memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN); + + system_id = ovsrec_port_get_other_config_value(port->cfg, "lacp-system-id", + NULL); + if (system_id) { + if (sscanf(system_id, ETH_ADDR_SCAN_FMT, + ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) { + VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet" + " address.", port->name, system_id); + return NULL; + } + } else { + memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN); + } + + if (eth_addr_is_zero(s->id)) { + VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name); + return NULL; + } /* Prefer bondable links if unspecified. */ - priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority", - "0")); + priority = atoi(ovsrec_port_get_other_config_value(port->cfg, + "lacp-system-priority", + "0")); s->priority = (priority > 0 && priority <= UINT16_MAX ? priority : UINT16_MAX - !list_is_short(&port->ifaces)); - s->heartbeat = !strcmp(get_port_other_config(port->cfg, - "lacp-heartbeat", - "false"), "true"); - - - lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow"); - custom_time = atoi(lacp_time); - if (!strcmp(lacp_time, "fast")) { - s->lacp_time = LACP_TIME_FAST; - } else if (!strcmp(lacp_time, "slow")) { - s->lacp_time = LACP_TIME_SLOW; - } else if (custom_time > 0) { - s->lacp_time = LACP_TIME_CUSTOM; - s->custom_time = custom_time; - } else { - s->lacp_time = LACP_TIME_SLOW; - } - + lacp_time = ovsrec_port_get_other_config_value(port->cfg, "lacp-time", + "slow"); + s->fast = !strcasecmp(lacp_time, "fast"); return s; } @@ -2294,11 +2982,16 @@ iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s) { int priority, portid, key; - portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0")); - priority = atoi(get_interface_other_config(iface->cfg, - "lacp-port-priority", "0")); - key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key", - "0")); + portid = atoi(ovsrec_interface_get_other_config_value(iface->cfg, + "lacp-port-id", + "0")); + priority = + atoi(ovsrec_interface_get_other_config_value(iface->cfg, + "lacp-port-priority", + "0")); + key = atoi(ovsrec_interface_get_other_config_value(iface->cfg, + "lacp-aggregation-key", + "0")); if (portid <= 0 || portid > UINT16_MAX) { portid = iface->ofp_port; @@ -2328,12 +3021,22 @@ port_configure_bond(struct port *port, struct bond_settings *s, size_t i; s->name = port->name; - s->balance = BM_SLB; - if (port->cfg->bond_mode - && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) { - VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", - port->name, port->cfg->bond_mode, - bond_mode_to_string(s->balance)); + s->balance = BM_AB; + if (port->cfg->bond_mode) { + if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) { + VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s", + port->name, port->cfg->bond_mode, + bond_mode_to_string(s->balance)); + } + } else { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + + /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to + * active-backup. At some point we should remove this warning. */ + VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that" + " in previous versions, the default bond_mode was" + " balance-slb", port->name, + bond_mode_to_string(s->balance)); } if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) { VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, " @@ -2341,13 +3044,16 @@ port_configure_bond(struct port *port, struct bond_settings *s, port->name); } - miimon_interval = atoi(get_port_other_config(port->cfg, - "bond-miimon-interval", "0")); + miimon_interval = + atoi(ovsrec_port_get_other_config_value(port->cfg, + "bond-miimon-interval", "0")); if (miimon_interval <= 0) { miimon_interval = 200; } - detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier"); + detect_s = ovsrec_port_get_other_config_value(port->cfg, + "bond-detect-mode", + "carrier"); if (!strcmp(detect_s, "carrier")) { miimon_interval = 0; } else if (strcmp(detect_s, "miimon")) { @@ -2358,10 +3064,14 @@ port_configure_bond(struct port *port, struct bond_settings *s, s->up_delay = MAX(0, port->cfg->bond_updelay); s->down_delay = MAX(0, port->cfg->bond_downdelay); - s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0")); + s->basis = atoi(ovsrec_port_get_other_config_value(port->cfg, + "bond-hash-basis", + "0")); s->rebalance_interval = atoi( - get_port_other_config(port->cfg, "bond-rebalance-interval", "10000")); - if (s->rebalance_interval < 1000) { + ovsrec_port_get_other_config_value(port->cfg, + "bond-rebalance-interval", + "10000")); + if (s->rebalance_interval && s->rebalance_interval < 1000) { s->rebalance_interval = 1000; } @@ -2371,8 +3081,10 @@ port_configure_bond(struct port *port, struct bond_settings *s, LIST_FOR_EACH (iface, port_elem, &port->ifaces) { long long stable_id; - stable_id = atoll(get_interface_other_config(iface->cfg, - "bond-stable-id", "0")); + stable_id = + atoll(ovsrec_interface_get_other_config_value(iface->cfg, + "bond-stable-id", + "0")); if (stable_id <= 0 || stable_id >= UINT32_MAX) { stable_id = iface->ofp_port; } @@ -2381,31 +3093,28 @@ port_configure_bond(struct port *port, struct bond_settings *s, netdev_set_miimon_interval(iface->netdev, miimon_interval); } } + +/* Returns true if 'port' is synthetic, that is, if we constructed it locally + * instead of obtaining it from the database. */ +static bool +port_is_synthetic(const struct port *port) +{ + return ovsdb_idl_row_is_synthetic(&port->cfg->header_); +} /* Interface functions. */ -static struct iface * -iface_create(struct port *port, const struct ovsrec_interface *if_cfg) +/* Returns the correct network device type for interface 'iface' in bridge + * 'br'. */ +static const char * +iface_get_type(const struct ovsrec_interface *iface, + const struct ovsrec_bridge *br) { - struct bridge *br = port->bridge; - struct iface *iface; - char *name = if_cfg->name; - - iface = xzalloc(sizeof *iface); - iface->port = port; - iface->name = xstrdup(name); - iface->ofp_port = -1; - iface->tag = tag_create_random(); - iface->netdev = NULL; - iface->cfg = if_cfg; - - hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0)); - - list_push_back(&port->ifaces, &iface->port_elem); - - VLOG_DBG("attached network device %s to port %s", iface->name, port->name); - - return iface; + /* The local port always has type "internal". Other ports take their type + * from the database and default to "system" if none is specified. */ + return (!strcmp(iface->name, br->name) ? "internal" + : iface->type[0] ? iface->type + : "system"); } static void @@ -2463,6 +3172,21 @@ iface_find(const char *name) return NULL; } +static struct if_cfg * +if_cfg_lookup(const struct bridge *br, const char *name) +{ + struct if_cfg *if_cfg; + + HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0), + &br->if_cfg_todo) { + if (!strcmp(if_cfg->cfg->name, name)) { + return if_cfg; + } + } + + return NULL; +} + static struct iface * iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port) { @@ -2529,6 +3253,7 @@ iface_clear_db_record(const struct ovsrec_interface *if_cfg) ovsrec_interface_set_link_state(if_cfg, NULL); ovsrec_interface_set_mtu(if_cfg, NULL, 0); ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0); + ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0); ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0); ovsrec_interface_set_lacp_current(if_cfg, NULL, 0); ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0); @@ -2611,6 +3336,10 @@ iface_delete_queues(unsigned int queue_id, static void iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) { + struct ofpbuf queues_buf; + + ofpbuf_init(&queues_buf, 0); + if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) { netdev_set_qos(iface->netdev, NULL, NULL); } else { @@ -2641,6 +3370,15 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) queue_zero = true; } + if (queue->n_dscp == 1) { + struct ofproto_port_queue *port_queue; + + port_queue = ofpbuf_put_uninit(&queues_buf, + sizeof *port_queue); + port_queue->queue = queue_id; + port_queue->dscp = queue->dscp[0]; + } + shash_from_ovs_idl_map(queue->key_other_config, queue->value_other_config, queue->n_other_config, &details); @@ -2648,17 +3386,25 @@ iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos) shash_destroy(&details); } if (!queue_zero) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); - VLOG_WARN_RL(&rl, "interface %s: QoS configured without a default " - "queue (queue 0). Packets not directed to a " - "correctly configured queue may be dropped.", - iface->name); + shash_init(&details); + netdev_set_queue(iface->netdev, 0, &details); + shash_destroy(&details); } } + if (iface->ofp_port >= 0) { + const struct ofproto_port_queue *port_queues = queues_buf.data; + size_t n_queues = queues_buf.size / sizeof *port_queues; + + ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port, + port_queues, n_queues); + } + netdev_set_policing(iface->netdev, iface->cfg->ingress_policing_rate, iface->cfg->ingress_policing_burst); + + ofpbuf_uninit(&queues_buf); } static void @@ -2666,6 +3412,7 @@ iface_configure_cfm(struct iface *iface) { const struct ovsrec_interface *cfg = iface->cfg; const char *extended_str, *opstate_str; + const char *cfm_ccm_vlan; struct cfm_settings s; if (!cfg->n_cfm_mpid) { @@ -2674,17 +3421,36 @@ iface_configure_cfm(struct iface *iface) } s.mpid = *cfg->cfm_mpid; - s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval", - "0")); + s.interval = atoi(ovsrec_interface_get_other_config_value(iface->cfg, + "cfm_interval", + "0")); + cfm_ccm_vlan = ovsrec_interface_get_other_config_value(iface->cfg, + "cfm_ccm_vlan", + "0"); + s.ccm_pcp = atoi(ovsrec_interface_get_other_config_value(iface->cfg, + "cfm_ccm_pcp", + "0")); if (s.interval <= 0) { s.interval = 1000; } - extended_str = get_interface_other_config(iface->cfg, "cfm_extended", - "false"); + if (!strcasecmp("random", cfm_ccm_vlan)) { + s.ccm_vlan = CFM_RANDOM_VLAN; + } else { + s.ccm_vlan = atoi(cfm_ccm_vlan); + if (s.ccm_vlan == CFM_RANDOM_VLAN) { + s.ccm_vlan = 0; + } + } + + extended_str = ovsrec_interface_get_other_config_value(iface->cfg, + "cfm_extended", + "false"); s.extended = !strcasecmp("true", extended_str); - opstate_str = get_interface_other_config(iface->cfg, "cfm_opstate", "up"); + opstate_str = ovsrec_interface_get_other_config_value(iface->cfg, + "cfm_opstate", + "up"); s.opup = !strcasecmp("up", opstate_str); ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s); @@ -2697,6 +3463,7 @@ iface_is_synthetic(const struct iface *iface) { return ovsdb_idl_row_is_synthetic(&iface->cfg->header_); } + /* Port mirroring. */ @@ -2739,7 +3506,8 @@ bridge_configure_mirrors(struct bridge *br) if (!m) { m = mirror_create(br, cfg); } - if (!mirror_configure(m, cfg)) { + m->cfg = cfg; + if (!mirror_configure(m)) { mirror_destroy(m); } } @@ -2805,8 +3573,9 @@ mirror_collect_ports(struct mirror *m, } static bool -mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) +mirror_configure(struct mirror *m) { + const struct ovsrec_mirror *cfg = m->cfg; struct ofproto_mirror_settings s; /* Set name. */ @@ -2882,3 +3651,305 @@ mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg) return true; } + +/* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.) + * + * This is deprecated. It is only for compatibility with broken device drivers + * in old versions of Linux that do not properly support VLANs when VLAN + * devices are not used. When broken device drivers are no longer in + * widespread use, we will delete these interfaces. */ + +static void **blocks; +static size_t n_blocks, allocated_blocks; + +/* Adds 'block' to a list of blocks that have to be freed with free() when the + * VLAN splinters are reconfigured. */ +static void +register_block(void *block) +{ + if (n_blocks >= allocated_blocks) { + blocks = x2nrealloc(blocks, &allocated_blocks, sizeof *blocks); + } + blocks[n_blocks++] = block; +} + +/* Frees all of the blocks registered with register_block(). */ +static void +free_registered_blocks(void) +{ + size_t i; + + for (i = 0; i < n_blocks; i++) { + free(blocks[i]); + } + n_blocks = 0; +} + +/* Returns true if VLAN splinters are enabled on 'iface_cfg', false + * otherwise. */ +static bool +vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg) +{ + const char *value; + + value = ovsrec_interface_get_other_config_value(iface_cfg, + "enable-vlan-splinters", + ""); + return !strcmp(value, "true"); +} + +/* Figures out the set of VLANs that are in use for the purpose of VLAN + * splinters. + * + * If VLAN splinters are enabled on at least one interface and any VLANs are in + * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and + * 4095 will not be set). The caller is responsible for freeing the bitmap, + * with free(). + * + * If VLANs splinters are not enabled on any interface or if no VLANs are in + * use, returns NULL. + * + * Updates 'vlan_splinters_enabled_anywhere'. */ +static unsigned long int * +collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg) +{ + unsigned long int *splinter_vlans; + struct sset splinter_ifaces; + const char *real_dev_name; + struct shash *real_devs; + struct shash_node *node; + struct bridge *br; + size_t i; + + /* Free space allocated for synthesized ports and interfaces, since we're + * in the process of reconstructing all of them. */ + free_registered_blocks(); + + splinter_vlans = bitmap_allocate(4096); + sset_init(&splinter_ifaces); + vlan_splinters_enabled_anywhere = false; + for (i = 0; i < ovs_cfg->n_bridges; i++) { + struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i]; + size_t j; + + for (j = 0; j < br_cfg->n_ports; j++) { + struct ovsrec_port *port_cfg = br_cfg->ports[j]; + int k; + + for (k = 0; k < port_cfg->n_interfaces; k++) { + struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k]; + + if (vlan_splinters_is_enabled(iface_cfg)) { + vlan_splinters_enabled_anywhere = true; + sset_add(&splinter_ifaces, iface_cfg->name); + vlan_bitmap_from_array__(port_cfg->trunks, + port_cfg->n_trunks, + splinter_vlans); + } + } + + if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) { + bitmap_set1(splinter_vlans, *port_cfg->tag); + } + } + } + + if (!vlan_splinters_enabled_anywhere) { + free(splinter_vlans); + sset_destroy(&splinter_ifaces); + return NULL; + } + + HMAP_FOR_EACH (br, node, &all_bridges) { + if (br->ofproto) { + ofproto_get_vlan_usage(br->ofproto, splinter_vlans); + } + } + + /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on + * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN + * device to be created for it. */ + bitmap_set0(splinter_vlans, 0); + bitmap_set0(splinter_vlans, 4095); + + /* Delete all VLAN devices that we don't need. */ + vlandev_refresh(); + real_devs = vlandev_get_real_devs(); + SHASH_FOR_EACH (node, real_devs) { + const struct vlan_real_dev *real_dev = node->data; + const struct vlan_dev *vlan_dev; + bool real_dev_has_splinters; + + real_dev_has_splinters = sset_contains(&splinter_ifaces, + real_dev->name); + HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) { + if (!real_dev_has_splinters + || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) { + struct netdev *netdev; + + if (!netdev_open(vlan_dev->name, "system", &netdev)) { + if (!netdev_get_in4(netdev, NULL, NULL) || + !netdev_get_in6(netdev, NULL)) { + vlandev_del(vlan_dev->name); + } else { + /* It has an IP address configured, so we don't own + * it. Don't delete it. */ + } + netdev_close(netdev); + } + } + + } + } + + /* Add all VLAN devices that we need. */ + SSET_FOR_EACH (real_dev_name, &splinter_ifaces) { + int vid; + + BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) { + if (!vlandev_get_name(real_dev_name, vid)) { + vlandev_add(real_dev_name, vid); + } + } + } + + vlandev_refresh(); + + sset_destroy(&splinter_ifaces); + + if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) { + free(splinter_vlans); + return NULL; + } + return splinter_vlans; +} + +/* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to + * ofproto. */ +static void +configure_splinter_port(struct port *port) +{ + struct ofproto *ofproto = port->bridge->ofproto; + uint16_t realdev_ofp_port; + const char *realdev_name; + struct iface *vlandev, *realdev; + + ofproto_bundle_unregister(port->bridge->ofproto, port); + + vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface, + port_elem); + + realdev_name = ovsrec_port_get_other_config_value(port->cfg, + "realdev", NULL); + realdev = iface_lookup(port->bridge, realdev_name); + realdev_ofp_port = realdev ? realdev->ofp_port : 0; + + ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port, + *port->cfg->tag); +} + +static struct ovsrec_port * +synthesize_splinter_port(const char *real_dev_name, + const char *vlan_dev_name, int vid) +{ + struct ovsrec_interface *iface; + struct ovsrec_port *port; + + iface = xzalloc(sizeof *iface); + iface->name = xstrdup(vlan_dev_name); + iface->type = "system"; + + port = xzalloc(sizeof *port); + port->interfaces = xmemdup(&iface, sizeof iface); + port->n_interfaces = 1; + port->name = xstrdup(vlan_dev_name); + port->vlan_mode = "splinter"; + port->tag = xmalloc(sizeof *port->tag); + *port->tag = vid; + port->key_other_config = xmalloc(sizeof *port->key_other_config); + port->key_other_config[0] = "realdev"; + port->value_other_config = xmalloc(sizeof *port->value_other_config); + port->value_other_config[0] = xstrdup(real_dev_name); + port->n_other_config = 1; + + register_block(iface); + register_block(iface->name); + register_block(port); + register_block(port->interfaces); + register_block(port->name); + register_block(port->tag); + register_block(port->key_other_config); + register_block(port->value_other_config); + register_block(port->value_other_config[0]); + + return port; +} + +/* For each interface with 'br' that has VLAN splinters enabled, adds a + * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a + * 1-bit in the 'splinter_vlans' bitmap. */ +static void +add_vlan_splinter_ports(struct bridge *br, + const unsigned long int *splinter_vlans, + struct shash *ports) +{ + size_t i; + + /* We iterate through 'br->cfg->ports' instead of 'ports' here because + * we're modifying 'ports'. */ + for (i = 0; i < br->cfg->n_ports; i++) { + const char *name = br->cfg->ports[i]->name; + struct ovsrec_port *port_cfg = shash_find_data(ports, name); + size_t j; + + for (j = 0; j < port_cfg->n_interfaces; j++) { + struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j]; + + if (vlan_splinters_is_enabled(iface_cfg)) { + const char *real_dev_name; + uint16_t vid; + + real_dev_name = iface_cfg->name; + BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) { + const char *vlan_dev_name; + + vlan_dev_name = vlandev_get_name(real_dev_name, vid); + if (vlan_dev_name + && !shash_find(ports, vlan_dev_name)) { + shash_add(ports, vlan_dev_name, + synthesize_splinter_port( + real_dev_name, vlan_dev_name, vid)); + } + } + } + } + } +} + +static void +mirror_refresh_stats(struct mirror *m) +{ + struct ofproto *ofproto = m->bridge->ofproto; + uint64_t tx_packets, tx_bytes; + char *keys[2]; + int64_t values[2]; + size_t stat_cnt = 0; + + if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) { + ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0); + return; + } + + if (tx_packets != UINT64_MAX) { + keys[stat_cnt] = "tx_packets"; + values[stat_cnt] = tx_packets; + stat_cnt++; + } + if (tx_bytes != UINT64_MAX) { + keys[stat_cnt] = "tx_bytes"; + values[stat_cnt] = tx_bytes; + stat_cnt++; + } + + ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt); +}