-/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
+/* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "bridge.h"
-#include "byte-order.h"
-#include <assert.h>
#include <errno.h>
-#include <arpa/inet.h>
-#include <ctype.h>
#include <inttypes.h>
-#include <sys/socket.h>
-#include <net/if.h>
-#include <openflow/openflow.h>
-#include <signal.h>
#include <stdlib.h>
-#include <strings.h>
-#include <sys/stat.h>
-#include <sys/socket.h>
-#include <sys/types.h>
-#include <unistd.h>
#include "bitmap.h"
+#include "bond.h"
#include "cfm.h"
-#include "classifier.h"
#include "coverage.h"
#include "daemon.h"
#include "dirs.h"
-#include "dpif.h"
#include "dynamic-string.h"
-#include "flow.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 "netlink.h"
-#include "odp-util.h"
#include "ofp-print.h"
+#include "ofp-util.h"
#include "ofpbuf.h"
-#include "ofproto/netflow.h"
#include "ofproto/ofproto.h"
-#include "ovsdb-data.h"
-#include "packets.h"
#include "poll-loop.h"
-#include "process.h"
#include "sha1.h"
#include "shash.h"
+#include "smap.h"
#include "socket-util.h"
+#include "stream.h"
#include "stream-ssl.h"
-#include "svec.h"
+#include "sset.h"
#include "system-stats.h"
#include "timeval.h"
#include "util.h"
#include "unixctl.h"
-#include "vconn.h"
-#include "vswitchd/vswitch-idl.h"
+#include "vlandev.h"
+#include "lib/vswitch-idl.h"
#include "xenserver.h"
#include "vlog.h"
#include "sflow_api.h"
+#include "vlan-bitmap.h"
VLOG_DEFINE_THIS_MODULE(bridge);
-COVERAGE_DEFINE(bridge_flush);
-COVERAGE_DEFINE(bridge_process_flow);
-COVERAGE_DEFINE(bridge_process_cfm);
-COVERAGE_DEFINE(bridge_process_lacp);
COVERAGE_DEFINE(bridge_reconfigure);
-COVERAGE_DEFINE(bridge_lacp_update);
-struct dst {
- uint16_t vlan;
- uint16_t dp_ifidx;
+/* 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. */
+ int64_t ofport; /* Requested OpenFlow port number. */
};
-struct dst_set {
- struct dst builtin[32];
- struct dst *dsts;
- size_t n, allocated;
+/* 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. */
};
-static void dst_set_init(struct dst_set *);
-static void dst_set_add(struct dst_set *, const struct dst *);
-static void dst_set_free(struct dst_set *);
-
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. */
- size_t port_ifidx; /* Index within containing port. */
char *name; /* Host network device name. */
- tag_type tag; /* Tag associated with this interface. */
- long long delay_expires; /* Time after which 'enabled' may change. */
/* These members are valid only after bridge_reconfigure() causes them to
* be initialized. */
- struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
- int dp_ifidx; /* Index within kernel datapath. */
+ struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
+ int ofp_port; /* OpenFlow port number, -1 if unknown. */
struct netdev *netdev; /* Network device. */
- bool enabled; /* May be chosen for flows? */
- bool up; /* Is the interface up? */
const char *type; /* Usually same as cfg->type. */
- struct cfm *cfm; /* Connectivity Fault Management */
const struct ovsrec_interface *cfg;
-
- /* LACP information. */
- uint16_t lacp_priority; /* LACP port priority. */
-};
-
-#define BOND_MASK 0xff
-struct bond_entry {
- int iface_idx; /* Index of assigned iface, or -1 if none. */
- uint64_t tx_bytes; /* Count of bytes recently transmitted. */
- tag_type iface_tag; /* Tag associated with iface_idx. */
};
-enum bond_mode {
- BM_TCP, /* Transport Layer Load Balance. */
- BM_SLB, /* Source Load Balance. */
- BM_AB /* Active Backup. */
-};
-
-#define MAX_MIRRORS 32
-typedef uint32_t mirror_mask_t;
-#define MIRROR_MASK_C(X) UINT32_C(X)
-BUILD_ASSERT_DECL(sizeof(mirror_mask_t) * CHAR_BIT >= MAX_MIRRORS);
struct mirror {
+ struct uuid uuid; /* UUID of this "mirror" record in database. */
+ struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
struct bridge *bridge;
- size_t idx;
char *name;
- struct uuid uuid; /* UUID of this "mirror" record in database. */
-
- /* Selection criteria. */
- struct shash src_ports; /* Name is port name; data is always NULL. */
- struct shash dst_ports; /* Name is port name; data is always NULL. */
- int *vlans;
- size_t n_vlans;
-
- /* Output. */
- struct port *out_port;
- int out_vlan;
+ const struct ovsrec_mirror *cfg;
};
-#define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
struct port {
+ struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
struct bridge *bridge;
- size_t port_idx;
- int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
- unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
- * NULL if all VLANs are trunked. */
- const struct ovsrec_port *cfg;
char *name;
- /* Monitoring. */
- struct netdev_monitor *monitor; /* Tracks carrier. NULL if miimon. */
- long long int miimon_interval; /* Miimon status refresh interval. */
- long long int miimon_next_update; /* Time of next miimon update. */
+ const struct ovsrec_port *cfg;
/* An ordinary bridge port has 1 interface.
* A bridge port for bonding has at least 2 interfaces. */
- struct iface **ifaces;
- size_t n_ifaces, allocated_ifaces;
-
- /* Bonding info. */
- enum bond_mode bond_mode; /* Type of the bond. BM_SLB is the default. */
- int active_iface; /* Ifidx on which bcasts accepted, or -1. */
- tag_type active_iface_tag; /* Tag for bcast flows. */
- tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */
- int updelay, downdelay; /* Delay before iface goes up/down, in ms. */
- bool bond_fake_iface; /* Fake a bond interface for legacy compat? */
- long long int bond_next_fake_iface_update; /* Time of next update. */
-
- /* LACP information. */
- struct lacp *lacp; /* LACP object. NULL if LACP is disabled. */
- bool lacp_active; /* True if LACP is active */
- bool lacp_fast; /* True if LACP is in fast mode. */
- uint16_t lacp_priority; /* LACP system priority. */
-
- /* SLB specific bonding info. */
- struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */
- int bond_rebalance_interval; /* Interval between rebalances, in ms. */
- long long int bond_next_rebalance; /* Next rebalancing time. */
-
- /* Port mirroring info. */
- mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
- mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
- bool is_mirror_output_port; /* Does port mirroring send frames here? */
+ struct list ifaces; /* List of "struct iface"s. */
};
struct bridge {
- struct list node; /* Node in global list of bridges. */
+ struct hmap_node node; /* In 'all_bridges'. */
char *name; /* User-specified arbitrary name. */
- struct mac_learning *ml; /* MAC learning table. */
+ char *type; /* Datapath type. */
uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
const struct ovsrec_bridge *cfg;
/* OpenFlow switch processing. */
struct ofproto *ofproto; /* OpenFlow switch. */
- /* Kernel datapath information. */
- struct dpif *dpif; /* Datapath. */
- struct hmap ifaces; /* Contains "struct iface"s. */
-
/* Bridge ports. */
- struct port **ports;
- size_t n_ports, allocated_ports;
- struct shash iface_by_name; /* "struct iface"s indexed by name. */
- struct shash port_by_name; /* "struct port"s indexed by name. */
-
- /* Bonding. */
- bool has_bonded_ports;
+ struct hmap ports; /* "struct port"s indexed by name. */
+ struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
+ struct hmap iface_by_name; /* "struct iface"s indexed by name. */
- /* Flow tracking. */
- bool flush;
+ 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 mirror *mirrors[MAX_MIRRORS];
+ struct hmap mirrors; /* "struct mirror" indexed by UUID. */
+
+ /* Synthetic local port if necessary. */
+ struct ovsrec_port synth_local_port;
+ struct ovsrec_interface synth_local_iface;
+ struct ovsrec_interface *synth_local_ifacep;
};
-/* List of all bridges. */
-static struct list all_bridges = LIST_INITIALIZER(&all_bridges);
+/* All bridges, indexed by name. */
+static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
/* OVSDB IDL used to obtain configuration. */
static struct ovsdb_idl *idl;
-/* 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. */
-static long long int stats_timer = LLONG_MIN;
+/* Most recently processed IDL sequence number. */
+static unsigned int idl_seqno;
-static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
+/* Each time this timer expires, the bridge fetches interface and mirror
+ * statistics and pushes them into the database. */
+#define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
+static long long int iface_stats_timer = 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_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);
static unixctl_cb_func bridge_unixctl_dump_flows;
static unixctl_cb_func bridge_unixctl_reconnect;
-static int bridge_run_one(struct bridge *);
static size_t bridge_get_controllers(const struct bridge *br,
struct ovsrec_controller ***controllersp);
-static void bridge_reconfigure_one(struct bridge *);
-static void bridge_reconfigure_remotes(struct bridge *,
- const struct sockaddr_in *managers,
- size_t n_managers);
-static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
-static void bridge_fetch_dp_ifaces(struct bridge *);
-static void bridge_flush(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_table(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_dp_desc(struct bridge *);
+static void bridge_configure_remotes(struct bridge *,
+ const struct sockaddr_in *managers,
+ size_t n_managers);
static void bridge_pick_local_hw_addr(struct bridge *,
uint8_t ea[ETH_ADDR_LEN],
struct iface **hw_addr_iface);
static uint64_t bridge_pick_datapath_id(struct bridge *,
const uint8_t bridge_ea[ETH_ADDR_LEN],
struct iface *hw_addr_iface);
-static struct iface *bridge_get_local_iface(struct bridge *);
+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);
+static bool port_is_bond_fake_iface(const struct port *);
-static unixctl_cb_func bridge_unixctl_fdb_show;
static unixctl_cb_func qos_unixctl_show;
-static void bond_init(void);
-static void bond_run(struct port *);
-static void bond_wait(struct port *);
-static void bond_rebalance_port(struct port *);
-static void bond_send_learning_packets(struct port *);
-static void bond_enable_slave(struct iface *iface, bool enable);
-
-static void port_run(struct port *);
-static void port_wait(struct port *);
-static struct port *port_create(struct bridge *, const char *name);
-static void port_reconfigure(struct port *, const struct ovsrec_port *);
-static void port_del_ifaces(struct port *, const struct ovsrec_port *);
+static struct port *port_create(struct bridge *, const struct ovsrec_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);
-static struct iface *port_lookup_iface(const struct port *, const char *name);
-static struct port *port_from_dp_ifidx(const struct bridge *,
- uint16_t dp_ifidx);
-static void port_update_bonding(struct port *);
-static void port_update_lacp(struct port *);
-
-static void mirror_create(struct bridge *, struct ovsrec_mirror *);
+static void port_configure(struct port *);
+static struct lacp_settings *port_configure_lacp(struct port *,
+ struct lacp_settings *);
+static void port_configure_bond(struct port *, struct bond_settings *);
+static bool port_is_synthetic(const struct port *);
+
+static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
+static void run_system_stats(void);
+
+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 void mirror_reconfigure(struct bridge *);
-static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
-static bool vlan_is_mirrored(const struct mirror *, int vlan);
-
-static struct iface *iface_create(struct port *port,
- const struct ovsrec_interface *if_cfg);
+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 bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
+static bool iface_is_internal(const struct ovsrec_interface *iface,
+ const struct ovsrec_bridge *br);
+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 iface *iface_from_dp_ifidx(const struct bridge *,
- uint16_t dp_ifidx);
+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 *);
static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
-static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
-static void iface_update_cfm(struct iface *);
-static void iface_refresh_cfm_stats(struct iface *iface);
-static void iface_send_packet(struct iface *, struct ofpbuf *packet);
-static void iface_update_carrier(struct iface *, bool carrier);
+static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
+static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
+static void iface_configure_cfm(struct iface *);
+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 int64_t iface_pick_ofport(const struct ovsrec_interface *);
+
+/* 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);
+
+static void
+bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
+{
+ struct shash iface_hints;
+ static bool initialized = false;
+ int i;
+
+ if (initialized) {
+ return;
+ }
+
+ shash_init(&iface_hints);
+
+ if (cfg) {
+ for (i = 0; i < cfg->n_bridges; i++) {
+ const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
+ int j;
-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);
+ 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 *if_cfg = port_cfg->interfaces[k];
+ struct iface_hint *iface_hint;
-/* Hooks into ofproto processing. */
-static struct ofhooks bridge_ofhooks;
+ iface_hint = xmalloc(sizeof *iface_hint);
+ iface_hint->br_name = br_cfg->name;
+ iface_hint->br_type = br_cfg->datapath_type;
+ iface_hint->ofp_port = iface_pick_ofport(if_cfg);
+
+ shash_add(&iface_hints, if_cfg->name, iface_hint);
+ }
+ }
+ }
+ }
+
+ ofproto_init(&iface_hints);
+
+ shash_destroy_free_data(&iface_hints);
+ initialized = true;
+}
\f
/* Public functions. */
bridge_init(const char *remote)
{
/* Create connection to database. */
- idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
+ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
+ idl_seqno = ovsdb_idl_get_seqno(idl);
+ ovsdb_idl_set_lock(idl, "ovs_vswitchd");
+ ovsdb_idl_verify_write_only(idl);
ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
+ ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
+ ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
+ ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
+ 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);
+ ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
+ 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_mac_in_use);
+ 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_cfm_remote_opstate);
+ ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
- /* Register unixctl commands. */
- unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
- unixctl_command_register("qos/show", qos_unixctl_show, NULL);
- unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
- NULL);
- unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
- NULL);
- lacp_init();
- bond_init();
-}
-
-void
-bridge_exit(void)
-{
- struct bridge *br, *next_br;
-
- LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
- bridge_destroy(br);
- }
- ovsdb_idl_destroy(idl);
-}
-
-/* Performs configuration that is only necessary once at ovs-vswitchd startup,
- * but for which the ovs-vswitchd configuration 'cfg' is required. */
-static void
-bridge_configure_once(const struct ovsrec_open_vswitch *cfg)
-{
- static bool already_configured_once;
- struct svec bridge_names;
- struct svec dpif_names, dpif_types;
- size_t i;
-
- /* Only do this once per ovs-vswitchd run. */
- if (already_configured_once) {
- return;
- }
- already_configured_once = true;
-
- stats_timer = time_msec() + STATS_INTERVAL;
-
- /* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
- svec_init(&bridge_names);
- for (i = 0; i < cfg->n_bridges; i++) {
- svec_add(&bridge_names, cfg->bridges[i]->name);
- }
- svec_sort(&bridge_names);
-
- /* Iterate over all system dpifs and delete any of them that do not appear
- * in 'cfg'. */
- svec_init(&dpif_names);
- svec_init(&dpif_types);
- dp_enumerate_types(&dpif_types);
- for (i = 0; i < dpif_types.n; i++) {
- size_t j;
-
- dp_enumerate_names(dpif_types.names[i], &dpif_names);
+ ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
+ ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
+ ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
+ ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
- /* Delete each dpif whose name is not in 'bridge_names'. */
- for (j = 0; j < dpif_names.n; j++) {
- if (!svec_contains(&bridge_names, dpif_names.names[j])) {
- struct dpif *dpif;
- int retval;
+ ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
- retval = dpif_open(dpif_names.names[j], dpif_types.names[i],
- &dpif);
- if (!retval) {
- dpif_delete(dpif);
- dpif_close(dpif);
- }
- }
- }
- }
- svec_destroy(&bridge_names);
- svec_destroy(&dpif_names);
- svec_destroy(&dpif_types);
-}
+ ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
-/* Callback for iterate_and_prune_ifaces(). */
-static bool
-check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
-{
- if (!iface->netdev) {
- /* We already reported a related error, don't bother duplicating it. */
- return false;
- }
+ ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
+ ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
- if (iface->dp_ifidx < 0) {
- VLOG_ERR("%s interface not in %s, dropping",
- iface->name, dpif_name(br->dpif));
- return false;
- }
+ ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
- VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
- iface->name, iface->dp_ifidx);
- return true;
-}
+ ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
-/* Callback for iterate_and_prune_ifaces(). */
-static bool
-set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
- void *aux OVS_UNUSED)
-{
- /* Set policing attributes. */
- netdev_set_policing(iface->netdev,
- iface->cfg->ingress_policing_rate,
- iface->cfg->ingress_policing_burst);
+ ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
+ ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
+ ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
+ ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
+ ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
- /* Set MAC address of internal interfaces other than the local
- * interface. */
- if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) {
- iface_set_mac(iface);
- }
+ ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
- return true;
+ /* Register unixctl commands. */
+ 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]", 0, 1,
+ bridge_unixctl_reconnect, NULL);
+ lacp_init();
+ bond_init();
+ cfm_init();
+ stp_init();
}
-/* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
- * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
- * deletes from 'br' any ports that no longer have any interfaces. */
-static void
-iterate_and_prune_ifaces(struct bridge *br,
- bool (*cb)(struct bridge *, struct iface *,
- void *aux),
- void *aux)
+void
+bridge_exit(void)
{
- size_t i, j;
-
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; ) {
- struct iface *iface = port->ifaces[j];
- if (cb(br, iface, aux)) {
- j++;
- } else {
- iface_set_ofport(iface->cfg, -1);
- iface_destroy(iface);
- }
- }
+ struct bridge *br, *next_br;
- if (port->n_ifaces) {
- i++;
- } else {
- VLOG_WARN("%s port has no interfaces, dropping", port->name);
- port_destroy(port);
- }
+ HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
+ bridge_destroy(br);
}
+ ovsdb_idl_destroy(idl);
}
/* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
{
struct sockaddr_in *managers = NULL;
size_t n_managers = 0;
- struct shash targets;
+ struct sset targets;
size_t i;
/* Collect all of the potential targets from the "targets" columns of the
* rows pointed to by "manager_options", excluding any that are
* out-of-band. */
- shash_init(&targets);
+ sset_init(&targets);
for (i = 0; i < ovs_cfg->n_manager_options; i++) {
struct ovsrec_manager *m = ovs_cfg->manager_options[i];
if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
- shash_find_and_delete(&targets, m->target);
+ sset_find_and_delete(&targets, m->target);
} else {
- shash_add_once(&targets, m->target, NULL);
+ sset_add(&targets, m->target);
}
}
/* Now extract the targets' IP addresses. */
- if (!shash_is_empty(&targets)) {
- struct shash_node *node;
+ if (!sset_is_empty(&targets)) {
+ const char *target;
- managers = xmalloc(shash_count(&targets) * sizeof *managers);
- SHASH_FOR_EACH (node, &targets) {
- const char *target = node->name;
+ managers = xmalloc(sset_count(&targets) * sizeof *managers);
+ 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++;
}
}
}
- shash_destroy(&targets);
+ sset_destroy(&targets);
*managersp = managers;
*n_managersp = n_managers;
static void
bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
{
- struct shash old_br, new_br;
- struct shash_node *node;
- struct bridge *br, *next;
- struct sockaddr_in *managers;
- size_t n_managers;
- size_t i;
- int sflow_bridge_number;
+ unsigned long int *splinter_vlans;
+ struct bridge *br;
COVERAGE_INC(bridge_reconfigure);
- collect_in_band_managers(ovs_cfg, &managers, &n_managers);
+ ovs_assert(!reconfiguring);
+ reconfiguring = true;
- /* Collect old and new bridges. */
- shash_init(&old_br);
- shash_init(&new_br);
- LIST_FOR_EACH (br, node, &all_bridges) {
- shash_add(&old_br, br->name, br);
- }
- for (i = 0; i < ovs_cfg->n_bridges; i++) {
- const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
- if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
- VLOG_WARN("more than one bridge named %s", br_cfg->name);
- }
+ /* 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 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, splinter_vlans);
}
+ free(splinter_vlans);
- /* Get rid of deleted bridges and add new bridges. */
- LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
- struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
- if (br_cfg) {
- br->cfg = br_cfg;
- } else {
- bridge_destroy(br);
- }
+ /* 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) {
+ bridge_refresh_ofp_port(br);
}
- SHASH_FOR_EACH (node, &new_br) {
- const char *br_name = node->name;
- const struct ovsrec_bridge *br_cfg = node->data;
- br = shash_find_data(&old_br, br_name);
- if (br) {
- /* If the bridge datapath type has changed, we need to tear it
- * down and recreate. */
- if (strcmp(br->cfg->datapath_type, br_cfg->datapath_type)) {
- bridge_destroy(br);
- bridge_create(br_cfg);
- }
- } else {
- bridge_create(br_cfg);
+
+ /* 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);
}
}
- shash_destroy(&old_br);
- shash_destroy(&new_br);
- /* Reconfigure all bridges. */
- LIST_FOR_EACH (br, node, &all_bridges) {
- bridge_reconfigure_one(br);
- }
+ reconfigure_system_stats(ovs_cfg);
+}
- /* Add and delete ports on all datapaths.
- *
- * The kernel will reject any attempt to add a given port to a datapath if
+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. */
- LIST_FOR_EACH (br, node, &all_bridges) {
- struct dpif_port_dump dump;
- struct shash want_ifaces;
- struct dpif_port dpif_port;
-
- bridge_get_all_ifaces(br, &want_ifaces);
- DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
- if (!shash_find(&want_ifaces, dpif_port.name)
- && strcmp(dpif_port.name, br->name)) {
- int retval = dpif_port_del(br->dpif, dpif_port.port_no);
- if (retval) {
- VLOG_WARN("failed to remove %s interface from %s: %s",
- dpif_port.name, dpif_name(br->dpif),
- strerror(retval));
- }
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ struct ofpp_garbage *garbage, *next;
+
+ LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
+ /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
+ * internal datastructures may not be consistent. Eventually, when
+ * port additions and deletions are cheaper, these calls should be
+ * removed. */
+ bridge_run_fast();
+ ofproto_port_del(br->ofproto, garbage->ofp_port);
+ list_remove(&garbage->list_node);
+ free(garbage);
+
+ time_refresh();
+ if (time_msec() >= deadline) {
+ return false;
}
+ bridge_run_fast();
}
- shash_destroy(&want_ifaces);
- }
- LIST_FOR_EACH (br, node, &all_bridges) {
- struct shash cur_ifaces, want_ifaces;
- struct dpif_port_dump dump;
- struct dpif_port dpif_port;
-
- /* Get the set of interfaces currently in this datapath. */
- shash_init(&cur_ifaces);
- DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
- struct dpif_port *port_info = xmalloc(sizeof *port_info);
- dpif_port_clone(port_info, &dpif_port);
- shash_add(&cur_ifaces, dpif_port.name, port_info);
- }
-
- /* Get the set of interfaces we want on this datapath. */
- bridge_get_all_ifaces(br, &want_ifaces);
-
- hmap_clear(&br->ifaces);
- SHASH_FOR_EACH (node, &want_ifaces) {
- const char *if_name = node->name;
- struct iface *iface = node->data;
- struct dpif_port *dpif_port;
- const char *type;
- int error;
-
- type = iface ? iface->type : "internal";
- dpif_port = shash_find_data(&cur_ifaces, if_name);
-
- /* If we have a port or a netdev already, and it's not the type we
- * want, then delete the port (if any) and close the netdev (if
- * any). */
- if ((dpif_port && strcmp(dpif_port->type, type))
- || (iface && iface->netdev
- && strcmp(type, netdev_get_type(iface->netdev)))) {
- if (dpif_port) {
- error = ofproto_port_del(br->ofproto, dpif_port->port_no);
- if (error) {
- continue;
- }
- dpif_port = NULL;
- }
- if (iface) {
- netdev_close(iface->netdev);
- iface->netdev = NULL;
- }
- }
-
- /* If the port doesn't exist or we don't have the netdev open,
- * we need to do more work. */
- if (!dpif_port || (iface && !iface->netdev)) {
- struct netdev_options options;
- struct netdev *netdev;
- struct shash args;
-
- /* First open the network device. */
- options.name = if_name;
- options.type = type;
- options.args = &args;
- options.ethertype = NETDEV_ETH_TYPE_NONE;
-
- shash_init(&args);
- if (iface) {
- shash_from_ovs_idl_map(iface->cfg->key_options,
- iface->cfg->value_options,
- iface->cfg->n_options, &args);
- }
- error = netdev_open(&options, &netdev);
- shash_destroy(&args);
-
- if (error) {
- VLOG_WARN("could not open network device %s (%s)",
- if_name, strerror(error));
- continue;
- }
+ }
- /* Then add the port if we haven't already. */
- if (!dpif_port) {
- error = dpif_port_add(br->dpif, netdev, NULL);
- if (error) {
- netdev_close(netdev);
- if (error == EFBIG) {
- VLOG_ERR("ran out of valid port numbers on %s",
- dpif_name(br->dpif));
- break;
- } else {
- VLOG_WARN("failed to add %s interface to %s: %s",
- if_name, dpif_name(br->dpif),
- strerror(error));
- continue;
- }
- }
- }
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ struct if_cfg *if_cfg, *next;
- /* Update 'iface'. */
- if (iface) {
- iface->netdev = netdev;
- iface->enabled = netdev_get_carrier(iface->netdev);
- iface->up = iface->enabled;
- }
- } else if (iface && iface->netdev) {
- struct shash args;
-
- shash_init(&args);
- shash_from_ovs_idl_map(iface->cfg->key_options,
- iface->cfg->value_options,
- iface->cfg->n_options, &args);
- netdev_set_config(iface->netdev, &args);
- shash_destroy(&args);
+ 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;
}
}
- shash_destroy(&want_ifaces);
-
- SHASH_FOR_EACH (node, &cur_ifaces) {
- struct dpif_port *port_info = node->data;
- dpif_port_destroy(port_info);
- free(port_info);
- }
- shash_destroy(&cur_ifaces);
}
- sflow_bridge_number = 0;
- LIST_FOR_EACH (br, node, &all_bridges) {
- uint8_t ea[8];
- uint64_t dpid;
- struct iface *local_iface;
- struct iface *hw_addr_iface;
- char *dpid_string;
-
- bridge_fetch_dp_ifaces(br);
-
- iterate_and_prune_ifaces(br, check_iface, NULL);
-
- /* Pick local port hardware address, datapath ID. */
- bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
- local_iface = bridge_get_local_iface(br);
- if (local_iface) {
- int error = netdev_set_etheraddr(local_iface->netdev, ea);
- if (error) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
- "Ethernet address: %s",
- br->name, strerror(error));
- }
- }
- memcpy(br->ea, ea, ETH_ADDR_LEN);
- dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
- ofproto_set_datapath_id(br->ofproto, dpid);
+ return true;
+}
- dpid_string = xasprintf("%016"PRIx64, dpid);
- ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
- free(dpid_string);
+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;
- /* Set NetFlow configuration on this bridge. */
- if (br->cfg->netflow) {
- struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
- struct netflow_options opts;
+ ovs_assert(reconfiguring);
+ done = bridge_reconfigure_ofp();
- memset(&opts, 0, sizeof opts);
+ /* 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;
- dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
- if (nf_cfg->engine_type) {
- opts.engine_type = *nf_cfg->engine_type;
- }
- if (nf_cfg->engine_id) {
- opts.engine_id = *nf_cfg->engine_id;
- }
+ /* We need the datapath ID early to allow LACP ports to use it as the
+ * default system ID. */
+ bridge_configure_datapath_id(br);
- opts.active_timeout = nf_cfg->active_timeout;
- if (!opts.active_timeout) {
- opts.active_timeout = -1;
- } else if (opts.active_timeout < 0) {
- VLOG_WARN("bridge %s: active timeout interval set to negative "
- "value, using default instead (%d seconds)", br->name,
- NF_ACTIVE_TIMEOUT_DEFAULT);
- opts.active_timeout = -1;
- }
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
- opts.add_id_to_iface = nf_cfg->add_id_to_interface;
- if (opts.add_id_to_iface) {
- if (opts.engine_id > 0x7f) {
- VLOG_WARN("bridge %s: netflow port mangling may conflict "
- "with another vswitch, choose an engine id less "
- "than 128", br->name);
- }
- if (br->n_ports > 508) {
- VLOG_WARN("bridge %s: netflow port mangling will conflict "
- "with another port when more than 508 ports are "
- "used", br->name);
- }
- }
+ port_configure(port);
- opts.collectors.n = nf_cfg->n_targets;
- opts.collectors.names = nf_cfg->targets;
- if (ofproto_set_netflow(br->ofproto, &opts)) {
- VLOG_ERR("bridge %s: problem setting netflow collectors",
- br->name);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_configure_cfm(iface);
+ iface_configure_qos(iface, port->cfg->qos);
+ iface_set_mac(iface);
}
- } else {
- ofproto_set_netflow(br->ofproto, NULL);
}
+ bridge_configure_mirrors(br);
+ bridge_configure_flow_eviction_threshold(br);
+ bridge_configure_forward_bpdu(br);
+ bridge_configure_mac_table(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);
+ bridge_configure_dp_desc(br);
+ }
+ free(managers);
- /* Set sFlow configuration on this bridge. */
- if (br->cfg->sflow) {
- const struct ovsrec_sflow *sflow_cfg = br->cfg->sflow;
- struct ovsrec_controller **controllers;
- struct ofproto_sflow_options oso;
- size_t n_controllers;
+ if (done) {
+ /* ovs-vswitchd has completed initialization, so allow the process that
+ * forked us to exit successfully. */
+ daemonize_complete();
+ reconfiguring = false;
- memset(&oso, 0, sizeof oso);
+ VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
+ }
- oso.targets.n = sflow_cfg->n_targets;
- oso.targets.names = sflow_cfg->targets;
+ return done;
+}
- oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
- if (sflow_cfg->sampling) {
- oso.sampling_rate = *sflow_cfg->sampling;
- }
+/* Delete ofprotos which aren't configured or have the wrong type. Create
+ * ofprotos which don't exist but need to. */
+static void
+bridge_update_ofprotos(void)
+{
+ struct bridge *br, *next;
+ struct sset names;
+ struct sset types;
+ const char *type;
- oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
- if (sflow_cfg->polling) {
- oso.polling_interval = *sflow_cfg->polling;
+ /* Delete ofprotos with no bridge or with the wrong type. */
+ sset_init(&names);
+ sset_init(&types);
+ ofproto_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ const char *name;
+
+ ofproto_enumerate_names(type, &names);
+ SSET_FOR_EACH (name, &names) {
+ br = bridge_lookup(name);
+ if (!br || strcmp(type, br->type)) {
+ ofproto_delete(name, type);
}
+ }
+ }
+ sset_destroy(&names);
+ sset_destroy(&types);
- oso.header_len = SFL_DEFAULT_HEADER_SIZE;
- if (sflow_cfg->header) {
- oso.header_len = *sflow_cfg->header;
- }
+ /* Add ofprotos for bridges which don't have one yet. */
+ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
+ struct bridge *br2;
+ int error;
- oso.sub_id = sflow_bridge_number++;
- oso.agent_device = sflow_cfg->agent;
+ if (br->ofproto) {
+ continue;
+ }
- oso.control_ip = NULL;
- n_controllers = bridge_get_controllers(br, &controllers);
- for (i = 0; i < n_controllers; i++) {
- if (controllers[i]->local_ip) {
- oso.control_ip = controllers[i]->local_ip;
- break;
+ /* 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);
}
- ofproto_set_sflow(br->ofproto, &oso);
+ }
- /* Do not destroy oso.targets because it is owned by sflow_cfg. */
- } else {
- ofproto_set_sflow(br->ofproto, NULL);
+ 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);
}
+ }
+}
- /* Update the controller and related settings. It would be more
- * straightforward to call this from bridge_reconfigure_one(), but we
- * can't do it there for two reasons. First, and most importantly, at
- * that point we don't know the dp_ifidx of any interfaces that have
- * been added to the bridge (because we haven't actually added them to
- * the datapath). Second, at that point we haven't set the datapath ID
- * yet; when a controller is configured, resetting the datapath ID will
- * immediately disconnect from the controller, so it's better to set
- * the datapath ID before the controller. */
- bridge_reconfigure_remotes(br, managers, n_managers);
- }
- LIST_FOR_EACH (br, node, &all_bridges) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- int j;
+static void
+port_configure(struct port *port)
+{
+ const struct ovsrec_port *cfg = port->cfg;
+ struct bond_settings bond_settings;
+ struct lacp_settings lacp_settings;
+ struct ofproto_bundle_settings s;
+ struct iface *iface;
- if (port->monitor) {
- for (j = 0; j < port->n_ifaces; j++) {
- netdev_monitor_add(port->monitor, port->ifaces[j]->netdev);
- }
- } else {
- port->miimon_next_update = 0;
- }
+ if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
+ configure_splinter_port(port);
+ return;
+ }
+
+ /* Get name. */
+ s.name = port->name;
- port_update_lacp(port);
- port_update_bonding(port);
+ /* Get slaves. */
+ s.n_slaves = 0;
+ s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ s.slaves[s.n_slaves++] = iface->ofp_port;
+ }
- for (j = 0; j < port->n_ifaces; j++) {
- iface_update_qos(port->ifaces[j], port->cfg->qos);
+ /* Get VLAN tag. */
+ s.vlan = -1;
+ if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
+ s.vlan = *cfg->tag;
+ }
+
+ /* Get VLAN trunks. */
+ s.trunks = NULL;
+ if (cfg->n_trunks) {
+ s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
+ }
+
+ /* Get VLAN mode. */
+ if (cfg->vlan_mode) {
+ if (!strcmp(cfg->vlan_mode, "access")) {
+ s.vlan_mode = PORT_VLAN_ACCESS;
+ } else if (!strcmp(cfg->vlan_mode, "trunk")) {
+ s.vlan_mode = PORT_VLAN_TRUNK;
+ } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
+ s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
+ } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
+ s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
+ } else {
+ /* This "can't happen" because ovsdb-server should prevent it. */
+ VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
+ s.vlan_mode = PORT_VLAN_TRUNK;
+ }
+ } else {
+ if (s.vlan >= 0) {
+ s.vlan_mode = PORT_VLAN_ACCESS;
+ if (cfg->n_trunks) {
+ VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
+ port->name);
}
+ } else {
+ s.vlan_mode = PORT_VLAN_TRUNK;
}
}
- LIST_FOR_EACH (br, node, &all_bridges) {
- iterate_and_prune_ifaces(br, set_iface_properties, NULL);
+ s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
+ false);
+
+ /* Get LACP settings. */
+ s.lacp = port_configure_lacp(port, &lacp_settings);
+ if (s.lacp) {
+ size_t i = 0;
+
+ s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_configure_lacp(iface, &s.lacp_slaves[i++]);
+ }
+ } else {
+ s.lacp_slaves = NULL;
}
- LIST_FOR_EACH (br, node, &all_bridges) {
- struct iface *iface;
- HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
- iface_update_cfm(iface);
+ /* Get bond settings. */
+ if (s.n_slaves > 1) {
+ s.bond = &bond_settings;
+ port_configure_bond(port, &bond_settings);
+ } else {
+ s.bond = NULL;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ netdev_set_miimon_interval(iface->netdev, 0);
}
}
- free(managers);
+ /* Register. */
+ ofproto_bundle_register(port->bridge->ofproto, port, &s);
- /* ovs-vswitchd has completed initialization, so allow the process that
- * forked us to exit successfully. */
- daemonize_complete();
+ /* Clean up. */
+ free(s.slaves);
+ free(s.trunks);
+ free(s.lacp_slaves);
}
-static const char *
-get_ovsrec_key_value(const struct ovsdb_idl_row *row,
- const struct ovsdb_idl_column *column,
- const char *key)
+/* Pick local port hardware address and datapath ID for 'br'. */
+static void
+bridge_configure_datapath_id(struct bridge *br)
{
- const struct ovsdb_datum *datum;
- union ovsdb_atom atom;
- unsigned int idx;
+ uint8_t ea[ETH_ADDR_LEN];
+ uint64_t dpid;
+ struct iface *local_iface;
+ struct iface *hw_addr_iface;
+ char *dpid_string;
- 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;
+ bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
+ local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
+ if (local_iface) {
+ int error = netdev_set_etheraddr(local_iface->netdev, ea);
+ if (error) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
+ "Ethernet address: %s",
+ br->name, strerror(error));
+ }
+ }
+ memcpy(br->ea, ea, ETH_ADDR_LEN);
+
+ dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
+ if (dpid != ofproto_get_datapath_id(br->ofproto)) {
+ VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
+ ofproto_set_datapath_id(br->ofproto, dpid);
+ }
+
+ dpid_string = xasprintf("%016"PRIx64, dpid);
+ ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
+ free(dpid_string);
}
-static const char *
-bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
+/* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
+ * 'br'. */
+static uint32_t
+bridge_get_allowed_versions(struct bridge *br)
{
- return get_ovsrec_key_value(&br_cfg->header_,
- &ovsrec_bridge_col_other_config, key);
+ if (!br->cfg->n_protocols)
+ return 0;
+
+ return ofputil_versions_from_strings(br->cfg->protocols,
+ br->cfg->n_protocols);
}
+/* Set NetFlow configuration on 'br'. */
static void
-bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
- struct iface **hw_addr_iface)
+bridge_configure_netflow(struct bridge *br)
{
- const char *hwaddr;
- size_t i, j;
- int error;
+ struct ovsrec_netflow *cfg = br->cfg->netflow;
+ struct netflow_options opts;
- *hw_addr_iface = NULL;
-
- /* Did the user request a particular MAC? */
- hwaddr = bridge_get_other_config(br->cfg, "hwaddr");
- if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
- if (eth_addr_is_multicast(ea)) {
- VLOG_ERR("bridge %s: cannot set MAC address to multicast "
- "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
- } else if (eth_addr_is_zero(ea)) {
- VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
- } else {
- return;
- }
+ if (!cfg) {
+ ofproto_set_netflow(br->ofproto, NULL);
+ return;
}
- /* Otherwise choose the minimum non-local MAC address among all of the
- * interfaces. */
- memset(ea, 0xff, ETH_ADDR_LEN);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- uint8_t iface_ea[ETH_ADDR_LEN];
- struct iface *iface;
-
- /* Mirror output ports don't participate. */
- if (port->is_mirror_output_port) {
- continue;
- }
-
- /* Choose the MAC address to represent the port. */
- if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
- /* Find the interface with this Ethernet address (if any) so that
- * we can provide the correct devname to the caller. */
- iface = NULL;
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *candidate = port->ifaces[j];
- uint8_t candidate_ea[ETH_ADDR_LEN];
- if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
- && eth_addr_equals(iface_ea, candidate_ea)) {
- iface = candidate;
- }
- }
- } else {
- /* Choose the interface whose MAC address will represent the port.
- * The Linux kernel bonding code always chooses the MAC address of
- * the first slave added to a bond, and the Fedora networking
- * scripts always add slaves to a bond in alphabetical order, so
- * for compatibility we choose the interface with the name that is
- * first in alphabetical order. */
- iface = port->ifaces[0];
- for (j = 1; j < port->n_ifaces; j++) {
- struct iface *candidate = port->ifaces[j];
- if (strcmp(candidate->name, iface->name) < 0) {
- iface = candidate;
- }
- }
-
- /* The local port doesn't count (since we're trying to choose its
- * MAC address anyway). */
- if (iface->dp_ifidx == ODPP_LOCAL) {
- continue;
- }
-
- /* Grab MAC. */
- error = netdev_get_etheraddr(iface->netdev, iface_ea);
- if (error) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
- iface->name, strerror(error));
- continue;
- }
- }
+ memset(&opts, 0, sizeof opts);
- /* Compare against our current choice. */
- if (!eth_addr_is_multicast(iface_ea) &&
- !eth_addr_is_local(iface_ea) &&
- !eth_addr_is_reserved(iface_ea) &&
- !eth_addr_is_zero(iface_ea) &&
- eth_addr_compare_3way(iface_ea, ea) < 0)
- {
- memcpy(ea, iface_ea, ETH_ADDR_LEN);
- *hw_addr_iface = iface;
- }
+ /* Get default NetFlow configuration from datapath.
+ * Apply overrides from 'cfg'. */
+ ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
+ if (cfg->engine_type) {
+ opts.engine_type = *cfg->engine_type;
}
- if (eth_addr_is_multicast(ea)) {
- 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));
- } else {
- VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
- br->name, ETH_ADDR_ARGS(ea));
+ if (cfg->engine_id) {
+ opts.engine_id = *cfg->engine_id;
}
-}
-
-/* Choose and returns the datapath ID for bridge 'br' given that the bridge
- * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
- * an interface on 'br', then that interface must be passed in as
- * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
- * 'hw_addr_iface' must be passed in as a null pointer. */
-static uint64_t
-bridge_pick_datapath_id(struct bridge *br,
- const uint8_t bridge_ea[ETH_ADDR_LEN],
- struct iface *hw_addr_iface)
-{
- /*
- * The procedure for choosing a bridge MAC address will, in the most
- * ordinary case, also choose a unique MAC that we can use as a datapath
- * ID. In some special cases, though, multiple bridges will end up with
- * the same MAC address. This is OK for the bridges, but it will confuse
- * the OpenFlow controller, because each datapath needs a unique datapath
- * ID.
- *
- * Datapath IDs must be unique. It is also very desirable that they be
- * stable from one run to the next, so that policy set on a datapath
- * "sticks".
- */
- const char *datapath_id;
- uint64_t dpid;
- datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
- if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
- return dpid;
+ /* Configure active timeout interval. */
+ opts.active_timeout = cfg->active_timeout;
+ if (!opts.active_timeout) {
+ opts.active_timeout = -1;
+ } else if (opts.active_timeout < 0) {
+ VLOG_WARN("bridge %s: active timeout interval set to negative "
+ "value, using default instead (%d seconds)", br->name,
+ NF_ACTIVE_TIMEOUT_DEFAULT);
+ opts.active_timeout = -1;
}
- 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.
- */
+ /* Add engine ID to interface number to disambiguate bridgs? */
+ opts.add_id_to_iface = cfg->add_id_to_interface;
+ if (opts.add_id_to_iface) {
+ if (opts.engine_id > 0x7f) {
+ VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
+ "another vswitch, choose an engine id less than 128",
+ br->name);
}
- } else {
- /*
- * 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
- * natural unique identifier at all.
- *
- * When the host is a XenServer, we handle this case by hashing the
- * host's UUID with the name of the bridge. Names of bridges are
- * persistent across XenServer reboots, although they can be reused if
- * an internal network is destroyed and then a new one is later
- * created, so this is fairly effective.
- *
- * When the host is not a XenServer, we punt by using a random MAC
- * address on each run.
- */
- const char *host_uuid = xenserver_get_host_uuid();
- if (host_uuid) {
- char *combined = xasprintf("%s,%s", host_uuid, br->name);
- dpid = dpid_from_hash(combined, strlen(combined));
- free(combined);
- return dpid;
+ if (hmap_count(&br->ports) > 508) {
+ VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
+ "another port when more than 508 ports are used",
+ br->name);
}
}
- return eth_addr_to_uint64(bridge_ea);
-}
-
-static uint64_t
-dpid_from_hash(const void *data, size_t n)
-{
- uint8_t hash[SHA1_DIGEST_SIZE];
+ /* Collectors. */
+ sset_init(&opts.collectors);
+ sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
- BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
- sha1_bytes(data, n, hash);
- eth_addr_mark_random(hash);
- return eth_addr_to_uint64(hash);
+ /* Configure. */
+ if (ofproto_set_netflow(br->ofproto, &opts)) {
+ VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
+ }
+ sset_destroy(&opts.collectors);
}
+/* Set sFlow configuration on 'br'. */
static void
-iface_refresh_status(struct iface *iface)
+bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
{
- struct shash sh;
-
- enum netdev_flags flags;
- uint32_t current;
- int64_t bps;
- int mtu;
- int64_t mtu_64;
- int error;
+ const struct ovsrec_sflow *cfg = br->cfg->sflow;
+ struct ovsrec_controller **controllers;
+ struct ofproto_sflow_options oso;
+ size_t n_controllers;
+ size_t i;
- shash_init(&sh);
+ if (!cfg) {
+ ofproto_set_sflow(br->ofproto, NULL);
+ return;
+ }
- if (!netdev_get_status(iface->netdev, &sh)) {
- size_t n;
- char **keys, **values;
+ memset(&oso, 0, sizeof oso);
- shash_to_ovs_idl_map(&sh, &keys, &values, &n);
- ovsrec_interface_set_status(iface->cfg, keys, values, n);
+ sset_init(&oso.targets);
+ sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
- free(keys);
- free(values);
- } else {
- ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
+ oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
+ if (cfg->sampling) {
+ oso.sampling_rate = *cfg->sampling;
}
- shash_destroy_free_data(&sh);
-
- error = netdev_get_flags(iface->netdev, &flags);
- if (!error) {
- ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
- }
- else {
- ovsrec_interface_set_admin_state(iface->cfg, NULL);
+ oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
+ if (cfg->polling) {
+ oso.polling_interval = *cfg->polling;
}
- error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
- if (!error) {
- ovsrec_interface_set_duplex(iface->cfg,
- netdev_features_is_full_duplex(current)
- ? "full" : "half");
- /* warning: uint64_t -> int64_t conversion */
- bps = netdev_features_to_bps(current);
- ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
- }
- else {
- ovsrec_interface_set_duplex(iface->cfg, NULL);
- ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
+ oso.header_len = SFL_DEFAULT_HEADER_SIZE;
+ if (cfg->header) {
+ oso.header_len = *cfg->header;
}
+ oso.sub_id = (*sflow_bridge_number)++;
+ oso.agent_device = cfg->agent;
- ovsrec_interface_set_link_state(iface->cfg,
- netdev_get_carrier(iface->netdev)
- ? "up" : "down");
-
- error = netdev_get_mtu(iface->netdev, &mtu);
- if (!error && mtu != INT_MAX) {
- mtu_64 = mtu;
- ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
- }
- else {
- ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
+ oso.control_ip = NULL;
+ n_controllers = bridge_get_controllers(br, &controllers);
+ for (i = 0; i < n_controllers; i++) {
+ if (controllers[i]->local_ip) {
+ oso.control_ip = controllers[i]->local_ip;
+ break;
+ }
}
+ ofproto_set_sflow(br->ofproto, &oso);
+
+ sset_destroy(&oso.targets);
}
static void
-iface_refresh_cfm_stats(struct iface *iface)
+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)
{
- size_t i;
- struct cfm *cfm;
- const struct ovsrec_monitor *mon;
+ const char *config_str;
+ struct iface *iface;
- mon = iface->cfg->monitor;
- cfm = iface->cfm;
+ if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
+ port_s->enable = false;
+ return;
+ } else {
+ port_s->enable = true;
+ }
- if (!cfm || !mon) {
+ /* 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;
}
- for (i = 0; i < mon->n_remote_mps; i++) {
- const struct ovsrec_maintenance_point *mp;
- const struct remote_mp *rmp;
+ iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
- mp = mon->remote_mps[i];
- rmp = cfm_get_remote_mp(cfm, mp->mpid);
+ /* 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;
+ }
- ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
+ /* 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;
}
- if (hmap_is_empty(&cfm->x_remote_mps)) {
- ovsrec_monitor_set_unexpected_remote_mpids(mon, NULL, 0);
- } else {
- size_t length;
- struct remote_mp *rmp;
- int64_t *x_remote_mps;
+ config_str = smap_get(&port->cfg->other_config, "stp-port-num");
+ if (config_str) {
+ unsigned long int port_num = strtoul(config_str, NULL, 0);
+ int port_idx = port_num - 1;
- length = hmap_count(&cfm->x_remote_mps);
- x_remote_mps = xzalloc(length * sizeof *x_remote_mps);
+ if (port_num < 1 || port_num > STP_MAX_PORTS) {
+ VLOG_ERR("port %s: invalid stp-port-num", port->name);
+ port_s->enable = false;
+ return;
+ }
- i = 0;
- HMAP_FOR_EACH (rmp, node, &cfm->x_remote_mps) {
- x_remote_mps[i++] = rmp->mpid;
+ 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;
}
- ovsrec_monitor_set_unexpected_remote_mpids(mon, x_remote_mps, length);
- free(x_remote_mps);
+ port_s->port_num = (*port_num_counter)++;
}
- if (hmap_is_empty(&cfm->x_remote_maids)) {
- ovsrec_monitor_set_unexpected_remote_maids(mon, NULL, 0);
+ config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
+ if (config_str) {
+ port_s->path_cost = strtoul(config_str, NULL, 10);
} else {
- size_t length;
- char **x_remote_maids;
- struct remote_maid *rmaid;
-
- length = hmap_count(&cfm->x_remote_maids);
- x_remote_maids = xzalloc(length * sizeof *x_remote_maids);
+ enum netdev_features current;
+ unsigned int mbps;
- i = 0;
- HMAP_FOR_EACH (rmaid, node, &cfm->x_remote_maids) {
- size_t j;
+ netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
+ mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
+ port_s->path_cost = stp_convert_speed_to_cost(mbps);
+ }
- x_remote_maids[i] = xzalloc(CCM_MAID_LEN * 2 + 1);
+ config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
+ if (config_str) {
+ port_s->priority = strtoul(config_str, NULL, 0);
+ } else {
+ port_s->priority = STP_DEFAULT_PORT_PRIORITY;
+ }
+}
- for (j = 0; j < CCM_MAID_LEN; j++) {
- snprintf(&x_remote_maids[i][j * 2], 3, "%02hhx",
- rmaid->maid[j]);
+/* 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 = smap_get(&br->cfg->other_config, "stp-system-id");
+ 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));
}
- i++;
+ } else {
+ br_s.system_id = eth_addr_to_uint64(br->ea);
}
- ovsrec_monitor_set_unexpected_remote_maids(mon, x_remote_maids, length);
- for (i = 0; i < length; i++) {
- free(x_remote_maids[i]);
+ config_str = smap_get(&br->cfg->other_config, "stp-priority");
+ if (config_str) {
+ br_s.priority = strtoul(config_str, NULL, 0);
+ } else {
+ br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
}
- free(x_remote_maids);
- }
- ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
-}
+ config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
+ if (config_str) {
+ br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
+ } else {
+ br_s.hello_time = STP_DEFAULT_HELLO_TIME;
+ }
-static void
-iface_refresh_stats(struct iface *iface)
-{
- struct iface_stat {
- char *name;
- int offset;
- };
- static const struct iface_stat iface_stats[] = {
- { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
- { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
- { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
- { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
- { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
- { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
- { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
- { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
- { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
- { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
- { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
- { "collisions", offsetof(struct netdev_stats, collisions) },
- };
- enum { N_STATS = ARRAY_SIZE(iface_stats) };
- const struct iface_stat *s;
-
- char *keys[N_STATS];
- int64_t values[N_STATS];
- int n;
+ config_str = smap_get(&br->cfg->other_config, "stp-max-age");
+ if (config_str) {
+ br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
+ } else {
+ br_s.max_age = STP_DEFAULT_MAX_AGE;
+ }
- struct netdev_stats stats;
+ config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
+ if (config_str) {
+ br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
+ } else {
+ br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
+ }
- /* Intentionally ignore return value, since errors will set 'stats' to
- * all-1s, and we will deal with that correctly below. */
- netdev_get_stats(iface->netdev, &stats);
+ /* Configure STP on the bridge. */
+ if (ofproto_set_stp(br->ofproto, &br_s)) {
+ VLOG_ERR("bridge %s: could not enable STP", br->name);
+ return;
+ }
- n = 0;
- for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
- uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
- if (value != UINT64_MAX) {
- keys[n] = s->name;
- values[n] = value;
- n++;
+ /* 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;
+ }
+ }
}
- }
- ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
+ 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 void
-refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
+static bool
+bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
{
- struct ovsdb_datum datum;
- struct shash stats;
-
- shash_init(&stats);
- get_system_stats(&stats);
-
- ovsdb_datum_from_shash(&datum, &stats);
- ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
- &datum);
+ const struct port *port = port_lookup(br, name);
+ return port && port_is_bond_fake_iface(port);
}
-static inline const char *
-nx_role_to_str(enum nx_role role)
+static bool
+port_is_bond_fake_iface(const struct port *port)
{
- switch (role) {
- case NX_ROLE_OTHER:
- return "other";
- case NX_ROLE_MASTER:
- return "master";
- case NX_ROLE_SLAVE:
- return "slave";
- default:
- return "*** INVALID ROLE ***";
- }
+ return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
}
static void
-bridge_refresh_controller_status(const struct bridge *br)
+add_del_bridges(const struct ovsrec_open_vswitch *cfg)
{
- struct shash info;
- const struct ovsrec_controller *cfg;
+ struct bridge *br, *next;
+ struct shash new_br;
+ size_t i;
- ofproto_get_ofproto_controller_info(br->ofproto, &info);
+ /* Collect new bridges' names and types. */
+ shash_init(&new_br);
+ for (i = 0; i < cfg->n_bridges; i++) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
- OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
- struct ofproto_controller_info *cinfo =
- shash_find_data(&info, cfg->target);
+ if (strchr(br_cfg->name, '/')) {
+ /* Prevent remote ovsdb-server users from accessing arbitrary
+ * directories, e.g. consider a bridge named "../../../etc/". */
+ VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
+ br_cfg->name);
+ } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
+ VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
+ }
+ }
- if (cinfo) {
- ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
- ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
- ovsrec_controller_set_status(cfg, (char **) cinfo->pairs.keys,
- (char **) cinfo->pairs.values,
- cinfo->pairs.n);
- } else {
- ovsrec_controller_set_is_connected(cfg, false);
- ovsrec_controller_set_role(cfg, NULL);
- ovsrec_controller_set_status(cfg, NULL, NULL, 0);
+ /* Get rid of deleted bridges or those whose types have changed.
+ * Update 'cfg' of bridges that still exist. */
+ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
+ br->cfg = shash_find_data(&new_br, br->name);
+ if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
+ br->cfg->datapath_type))) {
+ bridge_destroy(br);
}
}
- ofproto_free_ofproto_controller_info(&info);
+ /* Add new bridges. */
+ for (i = 0; i < cfg->n_bridges; i++) {
+ const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
+ struct bridge *br = bridge_lookup(br_cfg->name);
+ if (!br) {
+ bridge_create(br_cfg);
+ }
+ }
+
+ shash_destroy(&new_br);
}
-void
-bridge_run(void)
+static void
+iface_set_ofp_port(struct iface *iface, int ofp_port)
{
- const struct ovsrec_open_vswitch *cfg;
+ struct bridge *br = iface->port->bridge;
- bool datapath_destroyed;
- bool database_changed;
- struct bridge *br;
-
- /* Let each bridge do the work that it needs to do. */
- datapath_destroyed = false;
- LIST_FOR_EACH (br, node, &all_bridges) {
- int error = bridge_run_one(br);
- 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;
- }
- }
+ ovs_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);
+}
- /* (Re)configure if necessary. */
- database_changed = ovsdb_idl_run(idl);
- cfg = ovsrec_open_vswitch_first(idl);
-#ifdef HAVE_OPENSSL
- /* Re-configure SSL. We do this on every trip through the main loop,
- * instead of just when the database changes, because the contents of the
- * key and certificate files can change without the database changing.
- *
- * We do this before bridge_reconfigure() because that function might
- * initiate SSL connections and thus requires SSL to be configured. */
- if (cfg && cfg->ssl) {
- const struct ovsrec_ssl *ssl = cfg->ssl;
+/* 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)
+{
+ int error;
- stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
- stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
+ error = netdev_set_config(netdev, &iface_cfg->options);
+ if (error) {
+ VLOG_WARN("could not configure network device %s (%s)",
+ iface_cfg->name, strerror(error));
}
-#endif
- if (database_changed || datapath_destroyed) {
- if (cfg) {
- struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
-
- bridge_configure_once(cfg);
- bridge_reconfigure(cfg);
+ return error;
+}
- ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
- ovsdb_idl_txn_commit(txn);
- ovsdb_idl_txn_destroy(txn); /* XXX */
- } 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;
+/* 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)
+{
+ const char *name = ofproto_port->name;
+ const char *type = ofproto_port->type;
+ uint16_t ofp_port = ofproto_port->ofp_port;
- bridge_reconfigure(&null_cfg);
+ 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;
}
- }
-
- /* Refresh system and interface stats if necessary. */
- if (time_msec() >= stats_timer) {
- if (cfg) {
- struct ovsdb_idl_txn *txn;
-
- txn = ovsdb_idl_txn_create(idl);
- LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- size_t j;
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- iface_refresh_stats(iface);
- iface_refresh_cfm_stats(iface);
- iface_refresh_status(iface);
- }
- }
- bridge_refresh_controller_status(br);
- }
- refresh_system_stats(cfg);
- ovsdb_idl_txn_commit(txn);
- ovsdb_idl_txn_destroy(txn); /* XXX */
+ /* 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;
}
-
- stats_timer = time_msec() + STATS_INTERVAL;
+ } 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));
}
}
-void
-bridge_wait(void)
+/* 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 bridge *br;
+ struct ofproto_port_dump dump;
+ struct ofproto_port ofproto_port;
+ struct port *port, *port_next;
- LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ /* 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;
- ofproto_wait(br->ofproto);
- if (ofproto_has_primary_controller(br->ofproto)) {
- continue;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface->ofp_port = -1;
}
+ }
- mac_learning_wait(br->ml);
-
- for (i = 0; i < br->n_ports; i++) {
- port_wait(br->ports[i]);
+ /* 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) {
+ 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);
}
}
- ovsdb_idl_wait(idl);
- poll_timer_wait_until(stats_timer);
-}
-/* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
- * configuration changes. */
-static void
-bridge_flush(struct bridge *br)
-{
- COVERAGE_INC(bridge_flush);
- br->flush = true;
- mac_learning_flush(br->ml);
-}
+ /* 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;
-/* Returns the 'br' interface for the ODPP_LOCAL port, or null if 'br' has no
- * such interface. */
-static struct iface *
-bridge_get_local_iface(struct bridge *br)
-{
- size_t i, j;
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (iface->dp_ifidx == ODPP_LOCAL) {
- return iface;
+ 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);
}
}
- }
-
- return NULL;
-}
-\f
-/* Bridge unixctl user interface functions. */
-static void
-bridge_unixctl_fdb_show(struct unixctl_conn *conn,
- const char *args, void *aux OVS_UNUSED)
-{
- struct ds ds = DS_EMPTY_INITIALIZER;
- const struct bridge *br;
- const struct mac_entry *e;
-
- br = bridge_lookup(args);
- if (!br) {
- unixctl_command_reply(conn, 501, "no such bridge");
- return;
- }
- ds_put_cstr(&ds, " port VLAN MAC Age\n");
- LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
- if (e->port < 0 || e->port >= br->n_ports) {
- continue;
+ if (list_is_empty(&port->ifaces)) {
+ port_destroy(port);
}
- ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
- br->ports[e->port]->ifaces[0]->dp_ifidx,
- e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
}
- unixctl_command_reply(conn, 200, ds_cstr(&ds));
- ds_destroy(&ds);
}
-\f
-/* QoS unixctl user interface functions. */
-
-struct qos_unixctl_show_cbdata {
- struct ds *ds;
- struct iface *iface;
-};
-static void
-qos_unixctl_show_cb(unsigned int queue_id,
- const struct shash *details,
- void *aux)
+/* Opens a network device for 'if_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 if_cfg *if_cfg,
+ int *ofp_portp, struct netdev **netdevp)
{
- struct qos_unixctl_show_cbdata *data = aux;
- struct ds *ds = data->ds;
- struct iface *iface = data->iface;
- struct netdev_queue_stats stats;
- struct shash_node *node;
+ const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
+ const struct ovsrec_port *port_cfg = if_cfg->parent;
+ struct netdev *netdev;
int error;
- ds_put_cstr(ds, "\n");
- if (queue_id) {
- ds_put_format(ds, "Queue %u:\n", queue_id);
- } else {
- ds_put_cstr(ds, "Default:\n");
+ 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;
}
- SHASH_FOR_EACH (node, details) {
- ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
+ error = iface_set_netdev_config(iface_cfg, netdev);
+ if (error) {
+ goto error;
}
- error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
- if (!error) {
- if (stats.tx_packets != UINT64_MAX) {
- ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
- }
+ if (*ofp_portp < 0) {
+ uint16_t ofp_port = if_cfg->ofport;
- if (stats.tx_bytes != UINT64_MAX) {
- ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
+ error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
+ if (error) {
+ goto error;
}
+ *ofp_portp = ofp_port;
- if (stats.tx_errors != UINT64_MAX) {
- ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
- }
+ VLOG_INFO("bridge %s: added interface %s on port %d",
+ br->name, iface_cfg->name, *ofp_portp);
} else {
- ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
- queue_id, strerror(error));
+ VLOG_DBG("bridge %s: interface %s is on port %d",
+ br->name, iface_cfg->name, *ofp_portp);
}
-}
-
-static void
-qos_unixctl_show(struct unixctl_conn *conn,
- const char *args, void *aux OVS_UNUSED)
-{
- struct ds ds = DS_EMPTY_INITIALIZER;
- struct shash sh = SHASH_INITIALIZER(&sh);
- struct iface *iface;
- const char *type;
- struct shash_node *node;
- struct qos_unixctl_show_cbdata data;
- int error;
- iface = iface_find(args);
- if (!iface) {
- unixctl_command_reply(conn, 501, "no such interface");
- return;
+ if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
+ || iface_is_internal(iface_cfg, br->cfg)) {
+ netdev_turn_flags_on(netdev, NETDEV_UP, true);
}
- netdev_get_qos(iface->netdev, &type, &sh);
-
- if (*type != '\0') {
- ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
-
- SHASH_FOR_EACH (node, &sh) {
- ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
- }
-
- data.ds = &ds;
- data.iface = iface;
- error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
-
- if (error) {
- ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
- }
- unixctl_command_reply(conn, 200, ds_cstr(&ds));
- } else {
- ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
- unixctl_command_reply(conn, 501, ds_cstr(&ds));
- }
+ *netdevp = netdev;
+ return 0;
- shash_destroy_free_data(&sh);
- ds_destroy(&ds);
+error:
+ *netdevp = NULL;
+ netdev_close(netdev);
+ return error;
}
-\f
-/* Bridge reconfiguration functions. */
-static struct bridge *
-bridge_create(const struct ovsrec_bridge *br_cfg)
+
+/* 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)
{
- struct bridge *br;
- int error;
+ const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
+ const struct ovsrec_port *port_cfg = if_cfg->parent;
- assert(!bridge_lookup(br_cfg->name));
- br = xzalloc(sizeof *br);
+ struct netdev *netdev;
+ struct iface *iface;
+ struct port *port;
+ int error;
+ bool ok = true;
- error = dpif_create_and_open(br_cfg->name, br_cfg->datapath_type,
- &br->dpif);
+ /* Do the bits that can fail up front.
+ *
+ * It's a bit dangerous to call bridge_run_fast() here as ofproto's
+ * internal datastructures may not be consistent. Eventually, when port
+ * additions and deletions are cheaper, these calls should be removed. */
+ bridge_run_fast();
+ ovs_assert(!iface_lookup(br, iface_cfg->name));
+ error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
+ bridge_run_fast();
if (error) {
- free(br);
- return NULL;
+ iface_set_ofport(iface_cfg, -1);
+ iface_clear_db_record(iface_cfg);
+ ok = false;
+ goto done;
}
- dpif_flow_flush(br->dpif);
- error = ofproto_create(br_cfg->name, br_cfg->datapath_type, &bridge_ofhooks,
- br, &br->ofproto);
- if (error) {
- VLOG_ERR("failed to create switch %s: %s", br_cfg->name,
- strerror(error));
- dpif_delete(br->dpif);
- dpif_close(br->dpif);
- free(br);
- return NULL;
+ /* Get or create the port structure. */
+ port = port_lookup(br, port_cfg->name);
+ if (!port) {
+ port = port_create(br, port_cfg);
}
- br->name = xstrdup(br_cfg->name);
- br->cfg = br_cfg;
- br->ml = mac_learning_create();
- eth_addr_nicira_random(br->default_ea);
-
- hmap_init(&br->ifaces);
-
- shash_init(&br->port_by_name);
- shash_init(&br->iface_by_name);
+ /* 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;
- br->flush = false;
+ iface_set_ofp_port(iface, ofp_port);
- list_push_back(&all_bridges, &br->node);
+ /* Populate initial status in database. */
+ iface_refresh_stats(iface);
+ iface_refresh_status(iface);
- VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
+ /* Add bond fake iface if necessary. */
+ if (port_is_bond_fake_iface(port)) {
+ struct ofproto_port ofproto_port;
- return br;
-}
+ if (ofproto_port_query_by_name(br->ofproto, port->name,
+ &ofproto_port)) {
+ struct netdev *netdev;
+ int error;
-static void
-bridge_destroy(struct bridge *br)
-{
- if (br) {
- int error;
+ error = netdev_open(port->name, "internal", &netdev);
+ if (!error) {
+ uint16_t fake_ofp_port = if_cfg->ofport;
- while (br->n_ports > 0) {
- port_destroy(br->ports[br->n_ports - 1]);
- }
- list_remove(&br->node);
- error = dpif_delete(br->dpif);
- if (error && error != ENOENT) {
- VLOG_ERR("failed to delete %s: %s",
- dpif_name(br->dpif), strerror(error));
+ ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
+ netdev_close(netdev);
+ } else {
+ VLOG_WARN("could not open network device %s (%s)",
+ port->name, strerror(error));
+ }
+ } else {
+ /* Already exists, nothing to do. */
+ ofproto_port_destroy(&ofproto_port);
}
- dpif_close(br->dpif);
- ofproto_destroy(br->ofproto);
- mac_learning_destroy(br->ml);
- hmap_destroy(&br->ifaces);
- shash_destroy(&br->port_by_name);
- shash_destroy(&br->iface_by_name);
- free(br->ports);
- free(br->name);
- free(br);
}
-}
-static struct bridge *
-bridge_lookup(const char *name)
-{
- struct bridge *br;
+done:
+ hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
+ free(if_cfg);
- LIST_FOR_EACH (br, node, &all_bridges) {
- if (!strcmp(br->name, name)) {
- return br;
- }
- }
- return NULL;
+ return ok;
}
-/* Handle requests for a listing of all flows known by the OpenFlow
- * stack, including those normally hidden. */
+/* Set Flow eviction threshold */
static void
-bridge_unixctl_dump_flows(struct unixctl_conn *conn,
- const char *args, void *aux OVS_UNUSED)
+bridge_configure_flow_eviction_threshold(struct bridge *br)
{
- struct bridge *br;
- struct ds results;
+ const char *threshold_str;
+ unsigned threshold;
- br = bridge_lookup(args);
- if (!br) {
- unixctl_command_reply(conn, 501, "Unknown bridge");
- return;
+ threshold_str = smap_get(&br->cfg->other_config,
+ "flow-eviction-threshold");
+ if (threshold_str) {
+ threshold = strtoul(threshold_str, NULL, 10);
+ } else {
+ threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
}
-
- ds_init(&results);
- ofproto_get_all_flows(br->ofproto, &results);
-
- unixctl_command_reply(conn, 200, ds_cstr(&results));
- ds_destroy(&results);
+ ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
}
-/* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
- * connections and reconnect. If BRIDGE is not specified, then all bridges
- * drop their controller connections and reconnect. */
+/* Set forward BPDU option. */
static void
-bridge_unixctl_reconnect(struct unixctl_conn *conn,
- const char *args, void *aux OVS_UNUSED)
+bridge_configure_forward_bpdu(struct bridge *br)
{
- struct bridge *br;
- if (args[0] != '\0') {
- br = bridge_lookup(args);
- if (!br) {
- unixctl_command_reply(conn, 501, "Unknown bridge");
- return;
- }
- ofproto_reconnect_controllers(br->ofproto);
- } else {
- LIST_FOR_EACH (br, node, &all_bridges) {
- ofproto_reconnect_controllers(br->ofproto);
- }
- }
- unixctl_command_reply(conn, 200, NULL);
+ ofproto_set_forward_bpdu(br->ofproto,
+ smap_get_bool(&br->cfg->other_config,
+ "forward-bpdu",
+ false));
}
-static int
-bridge_run_one(struct bridge *br)
+/* Set MAC learning table configuration for 'br'. */
+static void
+bridge_configure_mac_table(struct bridge *br)
{
- size_t i;
- int error;
+ const char *idle_time_str;
+ int idle_time;
- error = ofproto_run1(br->ofproto);
- if (error) {
- return error;
- }
+ const char *mac_table_size_str;
+ int mac_table_size;
- mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
-
- for (i = 0; i < br->n_ports; i++) {
- port_run(br->ports[i]);
- }
+ idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
+ idle_time = (idle_time_str && atoi(idle_time_str)
+ ? atoi(idle_time_str)
+ : MAC_ENTRY_DEFAULT_IDLE_TIME);
- error = ofproto_run2(br->ofproto, br->flush);
- br->flush = false;
+ mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
+ mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
+ ? atoi(mac_table_size_str)
+ : MAC_DEFAULT_MAX);
- return error;
-}
-
-static size_t
-bridge_get_controllers(const struct bridge *br,
- struct ovsrec_controller ***controllersp)
-{
- struct ovsrec_controller **controllers;
- size_t n_controllers;
-
- controllers = br->cfg->controller;
- n_controllers = br->cfg->n_controller;
-
- if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
- controllers = NULL;
- n_controllers = 0;
- }
-
- if (controllersp) {
- *controllersp = controllers;
- }
- return n_controllers;
-}
-
-static void
-bridge_reconfigure_one(struct bridge *br)
-{
- struct shash old_ports, new_ports;
- struct svec snoops, old_snoops;
- struct shash_node *node;
- enum ofproto_fail_mode fail_mode;
- size_t i;
-
- /* Collect old ports. */
- shash_init(&old_ports);
- for (i = 0; i < br->n_ports; i++) {
- shash_add(&old_ports, br->ports[i]->name, br->ports[i]);
- }
-
- /* Collect new ports. */
- shash_init(&new_ports);
- for (i = 0; i < br->cfg->n_ports; i++) {
- const char *name = br->cfg->ports[i]->name;
- if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
- VLOG_WARN("bridge %s: %s specified twice as bridge port",
- br->name, name);
- }
- }
-
- /* If we have a controller, then we need a local port. Complain if the
- * user didn't specify one.
- *
- * XXX perhaps we should synthesize a port ourselves in this case. */
- if (bridge_get_controllers(br, NULL)) {
- char local_name[IF_NAMESIZE];
- int error;
-
- error = dpif_port_get_name(br->dpif, ODPP_LOCAL,
- local_name, sizeof local_name);
- if (!error && !shash_find(&new_ports, local_name)) {
- VLOG_WARN("bridge %s: controller specified but no local port "
- "(port named %s) defined",
- br->name, local_name);
- }
- }
-
- /* Get rid of deleted ports.
- * Get rid of deleted interfaces on ports that still exist. */
- SHASH_FOR_EACH (node, &old_ports) {
- struct port *port = node->data;
- const struct ovsrec_port *port_cfg;
-
- port_cfg = shash_find_data(&new_ports, node->name);
- if (!port_cfg) {
- port_destroy(port);
- } else {
- port_del_ifaces(port, port_cfg);
- }
- }
-
- /* Create new ports.
- * Add new interfaces to existing ports.
- * Reconfigure existing ports. */
- SHASH_FOR_EACH (node, &new_ports) {
- struct port *port = shash_find_data(&old_ports, node->name);
- if (!port) {
- port = port_create(br, node->name);
- }
-
- port_reconfigure(port, node->data);
- if (!port->n_ifaces) {
- VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
- br->name, port->name);
- port_destroy(port);
- }
- }
- shash_destroy(&old_ports);
- shash_destroy(&new_ports);
-
- /* Set the fail-mode */
- fail_mode = !br->cfg->fail_mode
- || !strcmp(br->cfg->fail_mode, "standalone")
- ? OFPROTO_FAIL_STANDALONE
- : OFPROTO_FAIL_SECURE;
- if (ofproto_get_fail_mode(br->ofproto) != fail_mode
- && !ofproto_has_primary_controller(br->ofproto)) {
- ofproto_flush_flows(br->ofproto);
- }
- ofproto_set_fail_mode(br->ofproto, fail_mode);
-
- /* Delete all flows if we're switching from connected to standalone or vice
- * versa. (XXX Should we delete all flows if we are switching from one
- * controller to another?) */
-
- /* Configure OpenFlow controller connection snooping. */
- svec_init(&snoops);
- svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop",
- ovs_rundir(), br->name));
- svec_init(&old_snoops);
- ofproto_get_snoops(br->ofproto, &old_snoops);
- if (!svec_equal(&snoops, &old_snoops)) {
- ofproto_set_snoops(br->ofproto, &snoops);
- }
- svec_destroy(&snoops);
- svec_destroy(&old_snoops);
-
- mirror_reconfigure(br);
-}
-
-/* Initializes 'oc' appropriately as a management service controller for
- * 'br'.
- *
- * The caller must free oc->target when it is no longer needed. */
-static void
-bridge_ofproto_controller_for_mgmt(const struct bridge *br,
- struct ofproto_controller *oc)
-{
- oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
- oc->max_backoff = 0;
- oc->probe_interval = 60;
- oc->band = OFPROTO_OUT_OF_BAND;
- oc->rate_limit = 0;
- oc->burst_limit = 0;
-}
-
-/* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
-static void
-bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
- struct ofproto_controller *oc)
-{
- 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;
- oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
- ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
- oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
- oc->burst_limit = (c->controller_burst_limit
- ? *c->controller_burst_limit : 0);
-}
-
-/* Configures the IP stack for 'br''s local interface properly according to the
- * configuration in 'c'. */
-static void
-bridge_configure_local_iface_netdev(struct bridge *br,
- struct ovsrec_controller *c)
-{
- struct netdev *netdev;
- struct in_addr mask, gateway;
-
- struct iface *local_iface;
- struct in_addr ip;
-
- /* If there's no local interface or no IP address, give up. */
- local_iface = bridge_get_local_iface(br);
- if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
- return;
- }
-
- /* Bring up the local interface. */
- netdev = local_iface->netdev;
- netdev_turn_flags_on(netdev, NETDEV_UP, true);
-
- /* Configure the IP address and netmask. */
- if (!c->local_netmask
- || !inet_aton(c->local_netmask, &mask)
- || !mask.s_addr) {
- mask.s_addr = guess_netmask(ip.s_addr);
- }
- if (!netdev_set_in4(netdev, ip, mask)) {
- VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
- br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
- }
-
- /* Configure the default gateway. */
- if (c->local_gateway
- && inet_aton(c->local_gateway, &gateway)
- && gateway.s_addr) {
- if (!netdev_add_router(netdev, gateway)) {
- VLOG_INFO("bridge %s: configured gateway "IP_FMT,
- br->name, IP_ARGS(&gateway.s_addr));
- }
- }
-}
-
-static void
-bridge_reconfigure_remotes(struct bridge *br,
- const struct sockaddr_in *managers,
- size_t n_managers)
-{
- const char *disable_ib_str, *queue_id_str;
- bool disable_in_band = false;
- int queue_id;
-
- struct ovsrec_controller **controllers;
- size_t n_controllers;
- bool had_primary;
-
- struct ofproto_controller *ocs;
- size_t n_ocs;
- 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");
- 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 = queue_id_str ? strtol(queue_id_str, NULL, 10) : -1;
- ofproto_set_in_band_queue(br->ofproto, queue_id);
-
- if (disable_in_band) {
- ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
- } else {
- ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
- }
- had_primary = ofproto_has_primary_controller(br->ofproto);
-
- n_controllers = bridge_get_controllers(br, &controllers);
-
- ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
- n_ocs = 0;
-
- bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
- for (i = 0; i < n_controllers; i++) {
- struct ovsrec_controller *c = controllers[i];
-
- if (!strncmp(c->target, "punix:", 6)
- || !strncmp(c->target, "unix:", 5)) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
-
- /* Prevent remote ovsdb-server users from accessing arbitrary Unix
- * domain sockets and overwriting arbitrary local files. */
- VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
- "\"%s\" due to possibility for remote exploit",
- dpif_name(br->dpif), c->target);
- continue;
- }
-
- bridge_configure_local_iface_netdev(br, c);
- bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
- if (disable_in_band) {
- ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
- }
- n_ocs++;
- }
-
- ofproto_set_controllers(br->ofproto, ocs, n_ocs);
- free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
- free(ocs);
-
- if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
- ofproto_flush_flows(br->ofproto);
- }
-
- /* If there are no controllers and the bridge is in standalone
- * mode, set up a flow that matches every packet and directs
- * them to OFPP_NORMAL (which goes to us). Otherwise, the
- * switch is in secure mode and we won't pass any traffic until
- * a controller has been defined and it tells us to do so. */
- if (!n_controllers
- && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
- union ofp_action action;
- struct cls_rule rule;
-
- memset(&action, 0, sizeof action);
- action.type = htons(OFPAT_OUTPUT);
- action.output.len = htons(sizeof action);
- action.output.port = htons(OFPP_NORMAL);
- cls_rule_init_catchall(&rule, 0);
- ofproto_add_flow(br->ofproto, &rule, &action, 1);
- }
-}
-
-static void
-bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
-{
- size_t i, j;
-
- shash_init(ifaces);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- shash_add_once(ifaces, iface->name, iface);
- }
- if (port->n_ifaces > 1 && port->cfg->bond_fake_iface) {
- shash_add_once(ifaces, port->name, NULL);
- }
- }
-}
-
-/* For robustness, in case the administrator moves around datapath ports behind
- * our back, we re-check all the datapath port numbers here.
- *
- * This function will set the 'dp_ifidx' members of interfaces that have
- * disappeared to -1, so only call this function from a context where those
- * 'struct iface's will be removed from the bridge. Otherwise, the -1
- * 'dp_ifidx'es will cause trouble later when we try to send them to the
- * datapath, which doesn't support UINT16_MAX+1 ports. */
-static void
-bridge_fetch_dp_ifaces(struct bridge *br)
-{
- struct dpif_port_dump dump;
- struct dpif_port dpif_port;
- size_t i, j;
-
- /* Reset all interface numbers. */
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- iface->dp_ifidx = -1;
- }
- }
- hmap_clear(&br->ifaces);
-
- DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
- struct iface *iface = iface_lookup(br, dpif_port.name);
- if (iface) {
- if (iface->dp_ifidx >= 0) {
- VLOG_WARN("%s reported interface %s twice",
- dpif_name(br->dpif), dpif_port.name);
- } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
- VLOG_WARN("%s reported interface %"PRIu16" twice",
- dpif_name(br->dpif), dpif_port.port_no);
- } else {
- iface->dp_ifidx = dpif_port.port_no;
- hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
- hash_int(iface->dp_ifidx, 0));
- }
-
- iface_set_ofport(iface->cfg,
- (iface->dp_ifidx >= 0
- ? odp_port_to_ofp_port(iface->dp_ifidx)
- : -1));
- }
- }
-}
-\f
-/* Bridge packet processing functions. */
-
-static bool
-bond_is_tcp_hash(const struct port *port)
-{
- return port->bond_mode == BM_TCP && lacp_negotiated(port->lacp);
-}
-
-static int
-bond_hash_src(const uint8_t mac[ETH_ADDR_LEN], uint16_t vlan)
-{
- return hash_bytes(mac, ETH_ADDR_LEN, vlan) & BOND_MASK;
-}
-
-static int bond_hash_tcp(const struct flow *flow, uint16_t vlan)
-{
- struct flow hash_flow;
-
- memcpy(&hash_flow, flow, sizeof hash_flow);
- hash_flow.vlan_tci = 0;
-
- /* The symmetric quality of this hash function is not required, but
- * flow_hash_symmetric_l4 already exists, and is sufficient for our
- * purposes, so we use it out of convenience. */
- return flow_hash_symmetric_l4(&hash_flow, vlan) & BOND_MASK;
-}
-
-static struct bond_entry *
-lookup_bond_entry(const struct port *port, const struct flow *flow,
- uint16_t vlan)
-{
- assert(port->bond_mode != BM_AB);
-
- if (bond_is_tcp_hash(port)) {
- return &port->bond_hash[bond_hash_tcp(flow, vlan)];
- } else {
- return &port->bond_hash[bond_hash_src(flow->dl_src, vlan)];
- }
-}
-
-static int
-bond_choose_iface(const struct port *port)
-{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
- size_t i, best_down_slave = -1;
- long long next_delay_expiration = LLONG_MAX;
-
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
-
- if (iface->enabled) {
- return i;
- } else if (iface->delay_expires < next_delay_expiration
- && lacp_slave_may_enable(port->lacp, iface)) {
- best_down_slave = i;
- next_delay_expiration = iface->delay_expires;
- }
- }
-
- if (best_down_slave != -1) {
- struct iface *iface = port->ifaces[best_down_slave];
-
- VLOG_INFO_RL(&rl, "interface %s: skipping remaining %lli ms updelay "
- "since no other interface is up", iface->name,
- iface->delay_expires - time_msec());
- bond_enable_slave(iface, true);
- }
-
- return best_down_slave;
-}
-
-static bool
-choose_output_iface(const struct port *port, const struct flow *flow,
- uint16_t vlan, uint16_t *dp_ifidx, tag_type *tags)
-{
- struct iface *iface;
-
- assert(port->n_ifaces);
- if (port->n_ifaces == 1) {
- iface = port->ifaces[0];
- } else if (port->bond_mode == BM_AB) {
- if (port->active_iface < 0) {
- *tags |= port->no_ifaces_tag;
- return false;
- }
- iface = port->ifaces[port->active_iface];
- } else {
- struct bond_entry *e = lookup_bond_entry(port, flow, vlan);
- if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces
- || !port->ifaces[e->iface_idx]->enabled) {
- /* XXX select interface properly. The current interface selection
- * is only good for testing the rebalancing code. */
- e->iface_idx = bond_choose_iface(port);
- if (e->iface_idx < 0) {
- *tags |= port->no_ifaces_tag;
- return false;
- }
- e->iface_tag = tag_create_random();
- }
- *tags |= e->iface_tag;
- iface = port->ifaces[e->iface_idx];
- }
- *dp_ifidx = iface->dp_ifidx;
- *tags |= iface->tag; /* Currently only used for bonding. */
- return true;
-}
-
-static void
-bond_link_status_update(struct iface *iface)
-{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
- struct port *port = iface->port;
- bool up = iface->up && lacp_slave_may_enable(port->lacp, iface);
- int updelay, downdelay;
-
- updelay = port->updelay;
- downdelay = port->downdelay;
-
- if (lacp_negotiated(port->lacp)) {
- downdelay = 0;
- updelay = 0;
- }
-
- if ((up == iface->enabled) == (iface->delay_expires == LLONG_MAX)) {
- /* Nothing to do. */
- return;
- }
- VLOG_INFO_RL(&rl, "interface %s: link state %s",
- iface->name, up ? "up" : "down");
- if (up == iface->enabled) {
- iface->delay_expires = LLONG_MAX;
- VLOG_INFO_RL(&rl, "interface %s: will not be %s",
- iface->name, up ? "disabled" : "enabled");
- } else if (up && port->active_iface < 0) {
- bond_enable_slave(iface, true);
- if (updelay) {
- VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no "
- "other interface is up", iface->name, updelay);
- }
- } else {
- int delay = up ? updelay : downdelay;
- iface->delay_expires = time_msec() + delay;
- if (delay) {
- VLOG_INFO_RL(&rl,
- "interface %s: will be %s if it stays %s for %d ms",
- iface->name,
- up ? "enabled" : "disabled",
- up ? "up" : "down",
- delay);
- }
- }
-}
-
-static void
-bond_choose_active_iface(struct port *port)
-{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
-
- port->active_iface = bond_choose_iface(port);
- port->active_iface_tag = tag_create_random();
- if (port->active_iface >= 0) {
- VLOG_INFO_RL(&rl, "port %s: active interface is now %s",
- port->name, port->ifaces[port->active_iface]->name);
- } else {
- VLOG_WARN_RL(&rl, "port %s: all ports disabled, no active interface",
- port->name);
- }
-}
-
-static void
-bond_enable_slave(struct iface *iface, bool enable)
-{
- struct port *port = iface->port;
- struct bridge *br = port->bridge;
-
- /* This acts as a recursion check. If the act of disabling a slave
- * causes a different slave to be enabled, the flag will allow us to
- * skip redundant work when we reenter this function. It must be
- * cleared on exit to keep things safe with multiple bonds. */
- static bool moving_active_iface = false;
-
- iface->delay_expires = LLONG_MAX;
- if (enable == iface->enabled) {
- return;
- }
-
- iface->enabled = enable;
- if (!iface->enabled) {
- VLOG_WARN("interface %s: disabled", iface->name);
- ofproto_revalidate(br->ofproto, iface->tag);
- if (iface->port_ifidx == port->active_iface) {
- ofproto_revalidate(br->ofproto,
- port->active_iface_tag);
-
- /* Disabling a slave can lead to another slave being immediately
- * enabled if there will be no active slaves but one is waiting
- * on an updelay. In this case we do not need to run most of the
- * code for the newly enabled slave since there was no period
- * without an active slave and it is redundant with the disabling
- * path. */
- moving_active_iface = true;
- bond_choose_active_iface(port);
- }
- bond_send_learning_packets(port);
- } else {
- VLOG_WARN("interface %s: enabled", iface->name);
- if (port->active_iface < 0 && !moving_active_iface) {
- ofproto_revalidate(br->ofproto, port->no_ifaces_tag);
- bond_choose_active_iface(port);
- bond_send_learning_packets(port);
- }
- iface->tag = tag_create_random();
- }
-
- moving_active_iface = false;
-}
-
-/* Attempts to make the sum of the bond slaves' statistics appear on the fake
- * bond interface. */
-static void
-bond_update_fake_iface_stats(struct port *port)
-{
- struct netdev_stats bond_stats;
- struct netdev *bond_dev;
- size_t i;
-
- memset(&bond_stats, 0, sizeof bond_stats);
-
- for (i = 0; i < port->n_ifaces; i++) {
- struct netdev_stats slave_stats;
-
- if (!netdev_get_stats(port->ifaces[i]->netdev, &slave_stats)) {
- /* XXX: We swap the stats here because they are swapped back when
- * reported by the internal device. The reason for this is
- * internal devices normally represent packets going into the system
- * but when used as fake bond device they represent packets leaving
- * the system. We really should do this in the internal device
- * itself because changing it here reverses the counts from the
- * perspective of the switch. However, the internal device doesn't
- * know what type of device it represents so we have to do it here
- * for now. */
- bond_stats.tx_packets += slave_stats.rx_packets;
- bond_stats.tx_bytes += slave_stats.rx_bytes;
- bond_stats.rx_packets += slave_stats.tx_packets;
- bond_stats.rx_bytes += slave_stats.tx_bytes;
- }
- }
-
- if (!netdev_open_default(port->name, &bond_dev)) {
- netdev_set_stats(bond_dev, &bond_stats);
- netdev_close(bond_dev);
- }
-}
-
-static void
-bond_run(struct port *port)
-{
- size_t i;
-
- if (port->n_ifaces < 2) {
- return;
- }
-
- for (i = 0; i < port->n_ifaces; i++) {
- bond_link_status_update(port->ifaces[i]);
- }
-
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- if (time_msec() >= iface->delay_expires) {
- bond_enable_slave(iface, !iface->enabled);
- }
- }
-
- if (port->bond_fake_iface
- && time_msec() >= port->bond_next_fake_iface_update) {
- bond_update_fake_iface_stats(port);
- port->bond_next_fake_iface_update = time_msec() + 1000;
- }
-}
-
-static void
-bond_wait(struct port *port)
-{
- size_t i;
-
- if (port->n_ifaces < 2) {
- return;
- }
-
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- if (iface->delay_expires != LLONG_MAX) {
- poll_timer_wait_until(iface->delay_expires);
- }
- }
-
- if (port->bond_fake_iface) {
- poll_timer_wait_until(port->bond_next_fake_iface_update);
- }
-}
-
-static bool
-set_dst(struct dst *dst, const struct flow *flow,
- const struct port *in_port, const struct port *out_port,
- tag_type *tags)
-{
- dst->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
- : in_port->vlan >= 0 ? in_port->vlan
- : flow->vlan_tci == 0 ? OFP_VLAN_NONE
- : vlan_tci_to_vid(flow->vlan_tci));
- return choose_output_iface(out_port, flow, dst->vlan,
- &dst->dp_ifidx, tags);
+ ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
}
static void
-swap_dst(struct dst *p, struct dst *q)
+bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
+ struct iface **hw_addr_iface)
{
- struct dst tmp = *p;
- *p = *q;
- *q = tmp;
-}
+ struct hmapx mirror_output_ports;
+ const char *hwaddr;
+ struct port *port;
+ bool found_addr = false;
+ int error;
+ int i;
-/* Moves all the dsts with vlan == 'vlan' to the front of the 'n_dsts' in
- * 'dsts'. (This may help performance by reducing the number of VLAN changes
- * that we push to the datapath. We could in fact fully sort the array by
- * vlan, but in most cases there are at most two different vlan tags so that's
- * possibly overkill.) */
-static void
-partition_dsts(struct dst_set *set, int vlan)
-{
- struct dst *first = set->dsts;
- struct dst *last = set->dsts + set->n;
-
- while (first != last) {
- /* Invariants:
- * - All dsts < first have vlan == 'vlan'.
- * - All dsts >= last have vlan != 'vlan'.
- * - first < last. */
- while (first->vlan == vlan) {
- if (++first == last) {
- return;
- }
- }
+ *hw_addr_iface = NULL;
- /* Same invariants, plus one additional:
- * - first->vlan != vlan.
- */
- while (last[-1].vlan != vlan) {
- if (--last == first) {
- return;
- }
+ /* Did the user request a particular MAC? */
+ hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
+ if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
+ if (eth_addr_is_multicast(ea)) {
+ VLOG_ERR("bridge %s: cannot set MAC address to multicast "
+ "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
+ } else if (eth_addr_is_zero(ea)) {
+ VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
+ } else {
+ return;
}
-
- /* Same invariants, plus one additional:
- * - last[-1].vlan == vlan.*/
- swap_dst(first++, --last);
- }
-}
-
-static int
-mirror_mask_ffs(mirror_mask_t mask)
-{
- BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
- return ffs(mask);
-}
-
-static void
-dst_set_init(struct dst_set *set)
-{
- set->dsts = set->builtin;
- set->n = 0;
- set->allocated = ARRAY_SIZE(set->builtin);
-}
-
-static void
-dst_set_add(struct dst_set *set, const struct dst *dst)
-{
- if (set->n >= set->allocated) {
- size_t new_allocated;
- struct dst *new_dsts;
-
- new_allocated = set->allocated * 2;
- new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
- memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
-
- dst_set_free(set);
-
- set->dsts = new_dsts;
- set->allocated = new_allocated;
- }
- set->dsts[set->n++] = *dst;
-}
-
-static void
-dst_set_free(struct dst_set *set)
-{
- if (set->dsts != set->builtin) {
- free(set->dsts);
}
-}
-static bool
-dst_is_duplicate(const struct dst_set *set, const struct dst *test)
-{
- size_t i;
- for (i = 0; i < set->n; i++) {
- if (set->dsts[i].vlan == test->vlan
- && set->dsts[i].dp_ifidx == test->dp_ifidx) {
- return true;
+ /* 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);
}
}
- return false;
-}
-
-static bool
-port_trunks_vlan(const struct port *port, uint16_t vlan)
-{
- return (port->vlan < 0
- && (!port->trunks || bitmap_is_set(port->trunks, vlan)));
-}
-
-static bool
-port_includes_vlan(const struct port *port, uint16_t vlan)
-{
- return vlan == port->vlan || port_trunks_vlan(port, vlan);
-}
-static bool
-port_is_floodable(const struct port *port)
-{
- int i;
+ /* Otherwise choose the minimum non-local MAC address among all of the
+ * interfaces. */
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ uint8_t iface_ea[ETH_ADDR_LEN];
+ struct iface *candidate;
+ struct iface *iface;
- for (i = 0; i < port->n_ifaces; i++) {
- if (!ofproto_port_is_floodable(port->bridge->ofproto,
- port->ifaces[i]->dp_ifidx)) {
- return false;
+ /* Mirror output ports don't participate. */
+ if (hmapx_contains(&mirror_output_ports, port->cfg)) {
+ continue;
}
- }
- return true;
-}
-
-static void
-compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
- const struct port *in_port, const struct port *out_port,
- struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
-{
- mirror_mask_t mirrors = in_port->src_mirrors;
- struct dst dst;
- int flow_vlan;
- size_t i;
- flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
- if (flow_vlan == 0) {
- flow_vlan = OFP_VLAN_NONE;
- }
-
- if (out_port == FLOOD_PORT) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port != in_port
- && port_is_floodable(port)
- && port_includes_vlan(port, vlan)
- && !port->is_mirror_output_port
- && set_dst(&dst, flow, in_port, port, tags)) {
- mirrors |= port->dst_mirrors;
- dst_set_add(set, &dst);
- }
- }
- *nf_output_iface = NF_OUT_FLOOD;
- } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
- dst_set_add(set, &dst);
- *nf_output_iface = dst.dp_ifidx;
- mirrors |= out_port->dst_mirrors;
- }
-
- while (mirrors) {
- struct mirror *m = br->mirrors[mirror_mask_ffs(mirrors) - 1];
- if (!m->n_vlans || vlan_is_mirrored(m, vlan)) {
- if (m->out_port) {
- if (set_dst(&dst, flow, in_port, m->out_port, tags)
- && !dst_is_duplicate(set, &dst)) {
- dst_set_add(set, &dst);
+ /* Choose the MAC address to represent the port. */
+ iface = NULL;
+ if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
+ /* Find the interface with this Ethernet address (if any) so that
+ * we can provide the correct devname to the caller. */
+ LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
+ uint8_t candidate_ea[ETH_ADDR_LEN];
+ if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
+ && eth_addr_equals(iface_ea, candidate_ea)) {
+ iface = candidate;
}
- } else {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port_includes_vlan(port, m->out_vlan)
- && set_dst(&dst, flow, in_port, port, tags))
- {
- if (port->vlan < 0) {
- dst.vlan = m->out_vlan;
- }
- if (dst_is_duplicate(set, &dst)) {
- continue;
- }
-
- /* Use the vlan tag on the original flow instead of
- * the one passed in the vlan parameter. This ensures
- * that we compare the vlan from before any implicit
- * tagging tags place. This is necessary because
- * dst->vlan is the final vlan, after removing implicit
- * tags. */
- if (port == in_port && dst.vlan == flow_vlan) {
- /* Don't send out input port on same VLAN. */
- continue;
- }
- dst_set_add(set, &dst);
- }
+ }
+ } else {
+ /* Choose the interface whose MAC address will represent the port.
+ * The Linux kernel bonding code always chooses the MAC address of
+ * the first slave added to a bond, and the Fedora networking
+ * scripts always add slaves to a bond in alphabetical order, so
+ * for compatibility we choose the interface with the name that is
+ * first in alphabetical order. */
+ LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
+ if (!iface || strcmp(candidate->name, iface->name) < 0) {
+ iface = candidate;
}
}
- }
- mirrors &= mirrors - 1;
- }
-
- partition_dsts(set, flow_vlan);
-}
-static void OVS_UNUSED
-print_dsts(const struct dst_set *set)
-{
- size_t i;
+ /* The local port doesn't count (since we're trying to choose its
+ * MAC address anyway). */
+ if (iface->ofp_port == OFPP_LOCAL) {
+ continue;
+ }
- for (i = 0; i < set->n; i++) {
- const struct dst *dst = &set->dsts[i];
+ /* Grab MAC. */
+ error = netdev_get_etheraddr(iface->netdev, iface_ea);
+ if (error) {
+ continue;
+ }
+ }
- printf(">p%"PRIu16, dst->dp_ifidx);
- if (dst->vlan != OFP_VLAN_NONE) {
- printf("v%"PRIu16, dst->vlan);
+ /* Compare against our current choice. */
+ if (!eth_addr_is_multicast(iface_ea) &&
+ !eth_addr_is_local(iface_ea) &&
+ !eth_addr_is_reserved(iface_ea) &&
+ !eth_addr_is_zero(iface_ea) &&
+ (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
+ {
+ memcpy(ea, iface_ea, ETH_ADDR_LEN);
+ *hw_addr_iface = iface;
+ found_addr = true;
}
}
+
+ if (!found_addr) {
+ memcpy(ea, br->default_ea, ETH_ADDR_LEN);
+ *hw_addr_iface = NULL;
+ }
+
+ hmapx_destroy(&mirror_output_ports);
}
-static void
-compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
- const struct port *in_port, const struct port *out_port,
- tag_type *tags, struct ofpbuf *actions,
- uint16_t *nf_output_iface)
+/* Choose and returns the datapath ID for bridge 'br' given that the bridge
+ * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
+ * an interface on 'br', then that interface must be passed in as
+ * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
+ * 'hw_addr_iface' must be passed in as a null pointer. */
+static uint64_t
+bridge_pick_datapath_id(struct bridge *br,
+ const uint8_t bridge_ea[ETH_ADDR_LEN],
+ struct iface *hw_addr_iface)
{
- struct dst_set set;
- uint16_t cur_vlan;
- size_t i;
-
- dst_set_init(&set);
- compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
- nf_output_iface);
+ /*
+ * The procedure for choosing a bridge MAC address will, in the most
+ * ordinary case, also choose a unique MAC that we can use as a datapath
+ * ID. In some special cases, though, multiple bridges will end up with
+ * the same MAC address. This is OK for the bridges, but it will confuse
+ * the OpenFlow controller, because each datapath needs a unique datapath
+ * ID.
+ *
+ * Datapath IDs must be unique. It is also very desirable that they be
+ * stable from one run to the next, so that policy set on a datapath
+ * "sticks".
+ */
+ const char *datapath_id;
+ uint64_t dpid;
- cur_vlan = vlan_tci_to_vid(flow->vlan_tci);
- if (cur_vlan == 0) {
- cur_vlan = OFP_VLAN_NONE;
- }
- for (i = 0; i < set.n; i++) {
- const struct dst *dst = &set.dsts[i];
- if (dst->vlan != cur_vlan) {
- if (dst->vlan == OFP_VLAN_NONE) {
- nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
- } else {
- ovs_be16 tci;
- tci = htons(dst->vlan & VLAN_VID_MASK);
- tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
- nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
- }
- cur_vlan = dst->vlan;
- }
- nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->dp_ifidx);
+ datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
+ if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
+ return dpid;
}
- dst_set_free(&set);
-}
-/* Returns the effective vlan of a packet, taking into account both the
- * 802.1Q header and implicitly tagged ports. A value of 0 indicates that
- * the packet is untagged and -1 indicates it has an invalid header and
- * should be dropped. */
-static int flow_get_vlan(struct bridge *br, const struct flow *flow,
- struct port *in_port, bool have_packet)
-{
- int vlan = vlan_tci_to_vid(flow->vlan_tci);
- if (in_port->vlan >= 0) {
- if (vlan) {
- /* XXX support double tagging? */
- if (have_packet) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
- "packet received on port %s configured with "
- "implicit VLAN %"PRIu16,
- br->name, vlan, in_port->name, in_port->vlan);
- }
- return -1;
- }
- vlan = in_port->vlan;
- } else {
- if (!port_includes_vlan(in_port, vlan)) {
- if (have_packet) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
- "packet received on port %s not configured for "
- "trunking VLAN %d",
- br->name, vlan, in_port->name, vlan);
- }
- return -1;
+ if (!hw_addr_iface) {
+ /*
+ * A purely internal bridge, that is, one that has no non-virtual
+ * 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
+ * host's UUID with the name of the bridge. Names of bridges are
+ * persistent across XenServer reboots, although they can be reused if
+ * an internal network is destroyed and then a new one is later
+ * created, so this is fairly effective.
+ *
+ * When the host is not a XenServer, we punt by using a random MAC
+ * address on each run.
+ */
+ const char *host_uuid = xenserver_get_host_uuid();
+ if (host_uuid) {
+ char *combined = xasprintf("%s,%s", host_uuid, br->name);
+ dpid = dpid_from_hash(combined, strlen(combined));
+ free(combined);
+ return dpid;
}
}
- return vlan;
+ return eth_addr_to_uint64(bridge_ea);
}
-/* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
- * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
- * indicate this; newer upstream kernels use gratuitous ARP requests. */
-static bool
-is_gratuitous_arp(const struct flow *flow)
+static uint64_t
+dpid_from_hash(const void *data, size_t n)
{
- return (flow->dl_type == htons(ETH_TYPE_ARP)
- && eth_addr_is_broadcast(flow->dl_dst)
- && (flow->nw_proto == ARP_OP_REPLY
- || (flow->nw_proto == ARP_OP_REQUEST
- && flow->nw_src == flow->nw_dst)));
-}
+ uint8_t hash[SHA1_DIGEST_SIZE];
-static void
-update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
- struct port *in_port)
-{
- enum grat_arp_lock_type lock_type;
- tag_type rev_tag;
-
- /* We don't want to learn from gratuitous ARP packets that are reflected
- * back over bond slaves so we lock the learning table. */
- lock_type = !is_gratuitous_arp(flow) ? GRAT_ARP_LOCK_NONE :
- (in_port->n_ifaces == 1) ? GRAT_ARP_LOCK_SET :
- GRAT_ARP_LOCK_CHECK;
-
- rev_tag = mac_learning_learn(br->ml, flow->dl_src, vlan, in_port->port_idx,
- lock_type);
- if (rev_tag) {
- /* The log messages here could actually be useful in debugging,
- * so keep the rate limit relatively high. */
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30,
- 300);
- VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
- "on port %s in VLAN %d",
- br->name, ETH_ADDR_ARGS(flow->dl_src),
- in_port->name, vlan);
- ofproto_revalidate(br->ofproto, rev_tag);
- }
+ BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
+ sha1_bytes(data, n, hash);
+ eth_addr_mark_random(hash);
+ return eth_addr_to_uint64(hash);
}
-/* Determines whether packets in 'flow' within 'br' should be forwarded or
- * dropped. Returns true if they may be forwarded, false if they should be
- * dropped.
- *
- * If 'have_packet' is true, it indicates that the caller is processing a
- * received packet. If 'have_packet' is false, then the caller is just
- * revalidating an existing flow because configuration has changed. Either
- * way, 'have_packet' only affects logging (there is no point in logging errors
- * during revalidation).
- *
- * Sets '*in_portp' to the input port. This will be a null pointer if
- * flow->in_port does not designate a known input port (in which case
- * is_admissible() returns false).
- *
- * When returning true, sets '*vlanp' to the effective VLAN of the input
- * packet, as returned by flow_get_vlan().
- *
- * May also add tags to '*tags', although the current implementation only does
- * so in one special case.
- */
-static bool
-is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
- tag_type *tags, int *vlanp, struct port **in_portp)
+static void
+iface_refresh_status(struct iface *iface)
{
- struct iface *in_iface;
- struct port *in_port;
- int vlan;
-
- /* Find the interface and port structure for the received packet. */
- in_iface = iface_from_dp_ifidx(br, flow->in_port);
- if (!in_iface) {
- /* No interface? Something fishy... */
- if (have_packet) {
- /* Odd. A few possible reasons here:
- *
- * - We deleted an interface but there are still a few packets
- * queued up from it.
- *
- * - Someone externally added an interface (e.g. with "ovs-dpctl
- * add-if") that we don't know about.
- *
- * - Packet arrived on the local port but the local port is not
- * one of our bridge ports.
- */
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
-
- VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
- "interface %"PRIu16, br->name, flow->in_port);
- }
-
- *in_portp = NULL;
- return false;
- }
- *in_portp = in_port = in_iface->port;
- *vlanp = vlan = flow_get_vlan(br, flow, in_port, have_packet);
- if (vlan < 0) {
- return false;
- }
-
- /* Drop frames for reserved multicast addresses. */
- if (eth_addr_is_reserved(flow->dl_dst)) {
- return false;
- }
+ struct smap smap;
- /* Drop frames on ports reserved for mirroring. */
- if (in_port->is_mirror_output_port) {
- if (have_packet) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
- "%s, which is reserved exclusively for mirroring",
- br->name, in_port->name);
- }
- return false;
- }
+ enum netdev_features current;
+ int64_t bps;
+ int mtu;
+ int64_t mtu_64;
+ uint8_t mac[ETH_ADDR_LEN];
+ int error;
- /* When using LACP, do not accept packets from disabled interfaces. */
- if (lacp_negotiated(in_port->lacp) && !in_iface->enabled) {
- return false;
+ if (iface_is_synthetic(iface)) {
+ return;
}
- /* Packets received on non-LACP bonds need special attention to avoid
- * duplicates. */
- if (in_port->n_ifaces > 1 && !lacp_negotiated(in_port->lacp)) {
- int src_idx;
- bool is_grat_arp_locked;
-
- if (eth_addr_is_multicast(flow->dl_dst)) {
- *tags |= in_port->active_iface_tag;
- if (in_port->active_iface != in_iface->port_ifidx) {
- /* Drop all multicast packets on inactive slaves. */
- return false;
- }
- }
+ smap_init(&smap);
- /* Drop all packets for which we have learned a different input
- * port, because we probably sent the packet on one slave and got
- * it back on the other. Gratuitous ARP packets are an exception
- * to this rule: the host has moved to another switch. The exception
- * to the exception is if we locked the learning table to avoid
- * reflections on bond slaves. If this is the case, just drop the
- * packet now. */
- src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan,
- &is_grat_arp_locked);
- if (src_idx != -1 && src_idx != in_port->port_idx &&
- (!is_gratuitous_arp(flow) || is_grat_arp_locked)) {
- return false;
- }
+ if (!netdev_get_status(iface->netdev, &smap)) {
+ ovsrec_interface_set_status(iface->cfg, &smap);
+ } else {
+ ovsrec_interface_set_status(iface->cfg, NULL);
}
- return true;
-}
-
-/* If the composed actions may be applied to any packet in the given 'flow',
- * returns true. Otherwise, the actions should only be applied to 'packet', or
- * not at all, if 'packet' was NULL. */
-static bool
-process_flow(struct bridge *br, const struct flow *flow,
- const struct ofpbuf *packet, struct ofpbuf *actions,
- tag_type *tags, uint16_t *nf_output_iface)
-{
- struct port *in_port;
- struct port *out_port;
- int vlan;
- int out_port_idx;
-
- /* Check whether we should drop packets in this flow. */
- if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
- out_port = NULL;
- goto done;
- }
+ smap_destroy(&smap);
- /* Learn source MAC (but don't try to learn from revalidation). */
- if (packet) {
- update_learning_table(br, flow, vlan, in_port);
+ error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
+ bps = !error ? netdev_features_to_bps(current, 0) : 0;
+ if (bps) {
+ ovsrec_interface_set_duplex(iface->cfg,
+ netdev_features_is_full_duplex(current)
+ ? "full" : "half");
+ ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
+ } else {
+ ovsrec_interface_set_duplex(iface->cfg, NULL);
+ ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
}
- /* Determine output port. */
- out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan, tags,
- NULL);
- if (out_port_idx >= 0 && out_port_idx < br->n_ports) {
- out_port = br->ports[out_port_idx];
- } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
- /* If we are revalidating but don't have a learning entry then
- * eject the flow. Installing a flow that floods packets opens
- * up a window of time where we could learn from a packet reflected
- * on a bond and blackhole packets before the learning table is
- * updated to reflect the correct port. */
- return false;
+ error = netdev_get_mtu(iface->netdev, &mtu);
+ if (!error) {
+ mtu_64 = mtu;
+ ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
} else {
- out_port = FLOOD_PORT;
+ ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
}
- /* Don't send packets out their input ports. */
- if (in_port == out_port) {
- out_port = NULL;
- }
+ error = netdev_get_etheraddr(iface->netdev, mac);
+ if (!error) {
+ char mac_string[32];
-done:
- if (in_port) {
- compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
- nf_output_iface);
+ sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
+ ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
+ } else {
+ ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
}
-
- return true;
}
-static bool
-bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
- struct ofpbuf *actions, tag_type *tags,
- uint16_t *nf_output_iface, void *br_)
+/* Writes 'iface''s CFM statistics to the database. 'iface' must not be
+ * synthetic. */
+static void
+iface_refresh_cfm_stats(struct iface *iface)
{
- struct bridge *br = br_;
+ const struct ovsrec_interface *cfg = iface->cfg;
+ struct ofproto_cfm_status status;
+
+ if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
+ iface->ofp_port, &status)) {
+ ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
+ ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
+ ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
+ ovsrec_interface_set_cfm_health(cfg, NULL, 0);
+ ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
+ } else {
+ const char *reasons[CFM_FAULT_N_REASONS];
+ int64_t cfm_health = status.health;
+ bool faulted = status.faults != 0;
+ size_t i, j;
+
+ ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
+
+ j = 0;
+ for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
+ int reason = 1 << i;
+ if (status.faults & reason) {
+ reasons[j++] = cfm_fault_reason_to_str(reason);
+ }
+ }
+ ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
+
+ if (status.remote_opstate >= 0) {
+ const char *remote_opstate = status.remote_opstate ? "up" : "down";
+ ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
+ } else {
+ ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
+ }
- COVERAGE_INC(bridge_process_flow);
- return process_flow(br, flow, packet, actions, tags, nf_output_iface);
+ ovsrec_interface_set_cfm_remote_mpids(cfg,
+ (const int64_t *)status.rmps,
+ status.n_rmps);
+ if (cfm_health >= 0) {
+ ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
+ } else {
+ ovsrec_interface_set_cfm_health(cfg, NULL, 0);
+ }
+ }
}
-static bool
-bridge_special_ofhook_cb(const struct flow *flow,
- const struct ofpbuf *packet, void *br_)
+static void
+iface_refresh_stats(struct iface *iface)
{
- struct iface *iface;
- struct bridge *br = br_;
-
- iface = iface_from_dp_ifidx(br, flow->in_port);
+#define IFACE_STATS \
+ IFACE_STAT(rx_packets, "rx_packets") \
+ IFACE_STAT(tx_packets, "tx_packets") \
+ IFACE_STAT(rx_bytes, "rx_bytes") \
+ IFACE_STAT(tx_bytes, "tx_bytes") \
+ IFACE_STAT(rx_dropped, "rx_dropped") \
+ IFACE_STAT(tx_dropped, "tx_dropped") \
+ IFACE_STAT(rx_errors, "rx_errors") \
+ IFACE_STAT(tx_errors, "tx_errors") \
+ IFACE_STAT(rx_frame_errors, "rx_frame_err") \
+ IFACE_STAT(rx_over_errors, "rx_over_err") \
+ IFACE_STAT(rx_crc_errors, "rx_crc_err") \
+ IFACE_STAT(collisions, "collisions")
+
+#define IFACE_STAT(MEMBER, NAME) NAME,
+ static char *keys[] = { IFACE_STATS };
+#undef IFACE_STAT
+ int64_t values[ARRAY_SIZE(keys)];
+ int i;
- if (cfm_should_process_flow(flow)) {
+ struct netdev_stats stats;
- if (iface && packet && iface->cfm) {
- COVERAGE_INC(bridge_process_cfm);
- cfm_process_heartbeat(iface->cfm, packet);
- }
- return false;
- } else if (flow->dl_type == htons(ETH_TYPE_LACP)) {
+ if (iface_is_synthetic(iface)) {
+ return;
+ }
- if (iface && iface->port->lacp && packet) {
- const struct lacp_pdu *pdu = parse_lacp_packet(packet);
+ /* Intentionally ignore return value, since errors will set 'stats' to
+ * all-1s, and we will deal with that correctly below. */
+ netdev_get_stats(iface->netdev, &stats);
- if (pdu) {
- COVERAGE_INC(bridge_process_lacp);
- lacp_process_pdu(iface->port->lacp, iface, pdu);
- }
- }
- return false;
- }
+ /* Copy statistics into values[] array. */
+ i = 0;
+#define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
+ IFACE_STATS;
+#undef IFACE_STAT
+ ovs_assert(i == ARRAY_SIZE(keys));
- return true;
+ ovsrec_interface_set_statistics(iface->cfg, keys, values,
+ ARRAY_SIZE(keys));
+#undef IFACE_STATS
}
static void
-bridge_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
- const struct nlattr *actions,
- size_t actions_len,
- uint64_t n_bytes, void *br_)
+br_refresh_stp_status(struct bridge *br)
{
- struct bridge *br = br_;
- const struct nlattr *a;
- struct port *in_port;
- tag_type dummy = 0;
- unsigned int left;
- int vlan;
-
- /* Feed information from the active flows back into the learning table to
- * ensure that table is always in sync with what is actually flowing
- * through the datapath.
- *
- * We test that 'tags' is nonzero to ensure that only flows that include an
- * OFPP_NORMAL action are used for learning. This works because
- * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
- if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
- update_learning_table(br, flow, vlan, in_port);
- }
+ struct smap smap = SMAP_INITIALIZER(&smap);
+ struct ofproto *ofproto = br->ofproto;
+ struct ofproto_stp_status status;
- /* Account for bond slave utilization. */
- if (!br->has_bonded_ports) {
+ if (ofproto_get_stp_status(ofproto, &status)) {
return;
}
- NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
- if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
- struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
- if (out_port && out_port->n_ifaces >= 2 &&
- out_port->bond_mode != BM_AB) {
- uint16_t vlan = (flow->vlan_tci
- ? vlan_tci_to_vid(flow->vlan_tci)
- : OFP_VLAN_NONE);
- struct bond_entry *e = lookup_bond_entry(out_port, flow, vlan);
- e->tx_bytes += n_bytes;
- }
- }
+
+ if (!status.enabled) {
+ ovsrec_bridge_set_status(br->cfg, NULL);
+ return;
}
+
+ smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
+ STP_ID_ARGS(status.bridge_id));
+ smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
+ STP_ID_ARGS(status.designated_root));
+ smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
+
+ ovsrec_bridge_set_status(br->cfg, &smap);
+ smap_destroy(&smap);
}
static void
-bridge_account_checkpoint_ofhook_cb(void *br_)
+port_refresh_stp_status(struct port *port)
{
- struct bridge *br = br_;
- long long int now;
- size_t i;
+ struct ofproto *ofproto = port->bridge->ofproto;
+ struct iface *iface;
+ struct ofproto_port_stp_status status;
+ char *keys[3];
+ int64_t int_values[3];
+ struct smap smap;
- if (!br->has_bonded_ports) {
+ if (port_is_synthetic(port)) {
return;
}
- now = time_msec();
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->n_ifaces > 1 && port->bond_mode != BM_AB
- && now >= port->bond_next_rebalance) {
- port->bond_next_rebalance = now + port->bond_rebalance_interval;
- bond_rebalance_port(port);
- }
+ /* STP doesn't currently support bonds. */
+ if (!list_is_singleton(&port->ifaces)) {
+ ovsrec_port_set_status(port->cfg, NULL);
+ return;
}
-}
-static struct ofhooks bridge_ofhooks = {
- bridge_normal_ofhook_cb,
- bridge_special_ofhook_cb,
- bridge_account_flow_ofhook_cb,
- bridge_account_checkpoint_ofhook_cb,
-};
-\f
-/* Bonding functions. */
-
-/* Statistics for a single interface on a bonded port, used for load-based
- * bond rebalancing. */
-struct slave_balance {
- struct iface *iface; /* The interface. */
- uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */
-
- /* All the "bond_entry"s that are assigned to this interface, in order of
- * increasing tx_bytes. */
- struct bond_entry **hashes;
- size_t n_hashes;
-};
+ iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
-static const char *
-bond_mode_to_string(enum bond_mode bm) {
- static char *bm_slb = "balance-slb";
- static char *bm_ab = "active-backup";
- static char *bm_tcp = "balance-tcp";
+ if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
+ return;
+ }
- switch (bm) {
- case BM_SLB: return bm_slb;
- case BM_AB: return bm_ab;
- case BM_TCP: return bm_tcp;
+ if (!status.enabled) {
+ ovsrec_port_set_status(port->cfg, NULL);
+ ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
+ return;
}
- NOT_REACHED();
- return NULL;
+ /* Set Status column. */
+ smap_init(&smap);
+ smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
+ smap_add(&smap, "stp_state", stp_state_name(status.state));
+ smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
+ smap_add(&smap, "stp_role", stp_role_name(status.role));
+ ovsrec_port_set_status(port->cfg, &smap);
+ smap_destroy(&smap);
+
+ /* 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));
}
-/* Sorts pointers to pointers to bond_entries in ascending order by the
- * interface to which they are assigned, and within a single interface in
- * ascending order of bytes transmitted. */
-static int
-compare_bond_entries(const void *a_, const void *b_)
+static bool
+enable_system_stats(const struct ovsrec_open_vswitch *cfg)
{
- const struct bond_entry *const *ap = a_;
- const struct bond_entry *const *bp = b_;
- const struct bond_entry *a = *ap;
- const struct bond_entry *b = *bp;
- if (a->iface_idx != b->iface_idx) {
- return a->iface_idx > b->iface_idx ? 1 : -1;
- } else if (a->tx_bytes != b->tx_bytes) {
- return a->tx_bytes > b->tx_bytes ? 1 : -1;
- } else {
- return 0;
- }
+ return smap_get_bool(&cfg->other_config, "enable-statistics", false);
}
-/* Sorts slave_balances so that enabled ports come first, and otherwise in
- * *descending* order by number of bytes transmitted. */
-static int
-compare_slave_balance(const void *a_, const void *b_)
+static void
+reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
{
- const struct slave_balance *a = a_;
- const struct slave_balance *b = b_;
- if (a->iface->enabled != b->iface->enabled) {
- return a->iface->enabled ? -1 : 1;
- } else if (a->tx_bytes != b->tx_bytes) {
- return a->tx_bytes > b->tx_bytes ? -1 : 1;
- } else {
- return 0;
+ bool enable = enable_system_stats(cfg);
+
+ system_stats_enable(enable);
+ if (!enable) {
+ ovsrec_open_vswitch_set_statistics(cfg, NULL);
}
}
static void
-swap_bals(struct slave_balance *a, struct slave_balance *b)
+run_system_stats(void)
{
- struct slave_balance tmp = *a;
- *a = *b;
- *b = tmp;
+ const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
+ struct smap *stats;
+
+ stats = system_stats_run();
+ if (stats && cfg) {
+ struct ovsdb_idl_txn *txn;
+ struct ovsdb_datum datum;
+
+ txn = ovsdb_idl_txn_create(idl);
+ ovsdb_datum_from_smap(&datum, stats);
+ ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
+ &datum);
+ ovsdb_idl_txn_commit(txn);
+ ovsdb_idl_txn_destroy(txn);
+
+ free(stats);
+ }
}
-/* Restores the 'n_bals' slave_balance structures in 'bals' to sorted order
- * given that 'p' (and only 'p') might be in the wrong location.
- *
- * This function invalidates 'p', since it might now be in a different memory
- * location. */
-static void
-resort_bals(struct slave_balance *p,
- struct slave_balance bals[], size_t n_bals)
+static inline const char *
+ofp12_controller_role_to_str(enum ofp12_controller_role role)
{
- if (n_bals > 1) {
- for (; p > bals && p->tx_bytes > p[-1].tx_bytes; p--) {
- swap_bals(p, p - 1);
- }
- for (; p < &bals[n_bals - 1] && p->tx_bytes < p[1].tx_bytes; p++) {
- swap_bals(p, p + 1);
- }
+ switch (role) {
+ case OFPCR12_ROLE_EQUAL:
+ return "other";
+ case OFPCR12_ROLE_MASTER:
+ return "master";
+ case OFPCR12_ROLE_SLAVE:
+ return "slave";
+ case OFPCR12_ROLE_NOCHANGE:
+ default:
+ return "*** INVALID ROLE ***";
}
}
static void
-log_bals(const struct slave_balance *bals, size_t n_bals, struct port *port)
+refresh_controller_status(void)
{
- if (VLOG_IS_DBG_ENABLED()) {
- struct ds ds = DS_EMPTY_INITIALIZER;
- const struct slave_balance *b;
+ struct bridge *br;
+ struct shash info;
+ const struct ovsrec_controller *cfg;
+
+ shash_init(&info);
- for (b = bals; b < bals + n_bals; b++) {
+ /* Accumulate status for controllers on all bridges. */
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_get_ofproto_controller_info(br->ofproto, &info);
+ }
+
+ /* Update each controller in the database with current status. */
+ OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
+ struct ofproto_controller_info *cinfo =
+ shash_find_data(&info, cfg->target);
+
+ if (cinfo) {
+ struct smap smap = SMAP_INITIALIZER(&smap);
+ const char **values = cinfo->pairs.values;
+ const char **keys = cinfo->pairs.keys;
size_t i;
- if (b > bals) {
- ds_put_char(&ds, ',');
+ for (i = 0; i < cinfo->pairs.n; i++) {
+ smap_add(&smap, keys[i], values[i]);
}
- ds_put_format(&ds, " %s %"PRIu64"kB",
- b->iface->name, b->tx_bytes / 1024);
- if (!b->iface->enabled) {
- ds_put_cstr(&ds, " (disabled)");
- }
- if (b->n_hashes > 0) {
- ds_put_cstr(&ds, " (");
- for (i = 0; i < b->n_hashes; i++) {
- const struct bond_entry *e = b->hashes[i];
- if (i > 0) {
- ds_put_cstr(&ds, " + ");
- }
- ds_put_format(&ds, "h%td: %"PRIu64"kB",
- e - port->bond_hash, e->tx_bytes / 1024);
- }
- ds_put_cstr(&ds, ")");
- }
+ ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
+ ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
+ cinfo->role));
+ ovsrec_controller_set_status(cfg, &smap);
+ smap_destroy(&smap);
+ } else {
+ ovsrec_controller_set_is_connected(cfg, false);
+ ovsrec_controller_set_role(cfg, NULL);
+ ovsrec_controller_set_status(cfg, NULL);
}
- VLOG_DBG("bond %s:%s", port->name, ds_cstr(&ds));
- ds_destroy(&ds);
}
+
+ ofproto_free_ofproto_controller_info(&info);
}
-/* Shifts 'hash' from 'from' to 'to' within 'port'. */
static void
-bond_shift_load(struct slave_balance *from, struct slave_balance *to,
- int hash_idx)
+refresh_instant_stats(void)
{
- struct bond_entry *hash = from->hashes[hash_idx];
- struct port *port = from->iface->port;
- uint64_t delta = hash->tx_bytes;
+ static struct ovsdb_idl_txn *txn = NULL;
- assert(port->bond_mode != BM_AB);
+ if (!txn) {
+ struct bridge *br;
- VLOG_INFO("bond %s: shift %"PRIu64"kB of load (with hash %td) "
- "from %s to %s (now carrying %"PRIu64"kB and "
- "%"PRIu64"kB load, respectively)",
- port->name, delta / 1024, hash - port->bond_hash,
- from->iface->name, to->iface->name,
- (from->tx_bytes - delta) / 1024,
- (to->tx_bytes + delta) / 1024);
+ txn = ovsdb_idl_txn_create(idl);
- /* Delete element from from->hashes.
- *
- * We don't bother to add the element to to->hashes because not only would
- * it require more work, the only purpose it would be to allow that hash to
- * be migrated to another slave in this rebalancing run, and there is no
- * point in doing that. */
- if (hash_idx == 0) {
- from->hashes++;
- } else {
- memmove(from->hashes + hash_idx, from->hashes + hash_idx + 1,
- (from->n_hashes - (hash_idx + 1)) * sizeof *from->hashes);
+ 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) {
+ enum netdev_flags flags;
+ const char *link_state;
+ int64_t link_resets;
+ int current, error;
+
+ if (iface_is_synthetic(iface)) {
+ continue;
+ }
+
+ current = ofproto_port_is_lacp_current(br->ofproto,
+ iface->ofp_port);
+ if (current >= 0) {
+ bool bl = current;
+ ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
+ } else {
+ ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
+ }
+
+ error = netdev_get_flags(iface->netdev, &flags);
+ if (!error) {
+ const char *state = flags & NETDEV_UP ? "up" : "down";
+ ovsrec_interface_set_admin_state(iface->cfg, state);
+ } else {
+ ovsrec_interface_set_admin_state(iface->cfg, NULL);
+ }
+
+ 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);
+
+ iface_refresh_cfm_stats(iface);
+ }
+ }
}
- from->n_hashes--;
- /* Shift load away from 'from' to 'to'. */
- from->tx_bytes -= delta;
- to->tx_bytes += delta;
+ if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
+ ovsdb_idl_txn_destroy(txn);
+ txn = NULL;
+ }
+}
+
+/* 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 sset types;
+ const char *type;
+ struct bridge *br;
+
+ sset_init(&types);
+ ofproto_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ ofproto_type_run_fast(type);
+ }
+ sset_destroy(&types);
- /* Arrange for flows to be revalidated. */
- ofproto_revalidate(port->bridge->ofproto, hash->iface_tag);
- hash->iface_idx = to->iface->port_ifidx;
- hash->iface_tag = tag_create_random();
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_run_fast(br->ofproto);
+ }
}
-static void
-bond_rebalance_port(struct port *port)
+void
+bridge_run(void)
{
- struct slave_balance *bals;
- size_t n_bals;
- struct bond_entry *hashes[BOND_MASK + 1];
- struct slave_balance *b, *from, *to;
- struct bond_entry *e;
- size_t i;
+ static struct ovsrec_open_vswitch null_cfg;
+ const struct ovsrec_open_vswitch *cfg;
+ struct ovsdb_idl_txn *reconf_txn = NULL;
+ struct sset types;
+ const char *type;
+
+ bool vlan_splinters_changed;
+ struct bridge *br;
- assert(port->bond_mode != BM_AB);
+ ovsrec_open_vswitch_init(&null_cfg);
- /* Sets up 'bals' to describe each of the port's interfaces, sorted in
- * descending order of tx_bytes, so that bals[0] represents the most
- * heavily loaded slave and bals[n_bals - 1] represents the least heavily
- * loaded slave.
- *
- * The code is a bit tricky: to avoid dynamically allocating a 'hashes'
- * array for each slave_balance structure, we sort our local array of
- * hashes in order by slave, so that all of the hashes for a given slave
- * become contiguous in memory, and then we point each 'hashes' members of
- * a slave_balance structure to the start of a contiguous group. */
- n_bals = port->n_ifaces;
- bals = xmalloc(n_bals * sizeof *bals);
- for (b = bals; b < &bals[n_bals]; b++) {
- b->iface = port->ifaces[b - bals];
- b->tx_bytes = 0;
- b->hashes = NULL;
- b->n_hashes = 0;
- }
- for (i = 0; i <= BOND_MASK; i++) {
- hashes[i] = &port->bond_hash[i];
- }
- qsort(hashes, BOND_MASK + 1, sizeof *hashes, compare_bond_entries);
- for (i = 0; i <= BOND_MASK; i++) {
- e = hashes[i];
- if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) {
- b = &bals[e->iface_idx];
- b->tx_bytes += e->tx_bytes;
- if (!b->hashes) {
- b->hashes = &hashes[i];
+ /* (Re)configure if necessary. */
+ if (!reconfiguring) {
+ 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;
+
+ 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);
}
- b->n_hashes++;
+ return;
+ } else if (!ovsdb_idl_has_lock(idl)) {
+ return;
}
}
- qsort(bals, n_bals, sizeof *bals, compare_slave_balance);
- log_bals(bals, n_bals, port);
+ cfg = ovsrec_open_vswitch_first(idl);
+
+ /* Initialize the ofproto library. This only needs to run once, but
+ * it must be done after the configuration is set. If the
+ * initialization has already occurred, bridge_init_ofproto()
+ * returns immediately. */
+ bridge_init_ofproto(cfg);
+
+ /* Let each datapath type do the work that it needs to do. */
+ sset_init(&types);
+ ofproto_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ ofproto_type_run(type);
+ }
+ sset_destroy(&types);
- /* Discard slaves that aren't enabled (which were sorted to the back of the
- * array earlier). */
- while (!bals[n_bals - 1].iface->enabled) {
- n_bals--;
- if (!n_bals) {
- goto exit;
- }
+ /* Let each bridge do the work that it needs to do. */
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_run(br->ofproto);
}
- /* Shift load from the most-loaded slaves to the least-loaded slaves. */
- to = &bals[n_bals - 1];
- for (from = bals; from < to; ) {
- uint64_t overload = from->tx_bytes - to->tx_bytes;
- if (overload < to->tx_bytes >> 5 || overload < 100000) {
- /* The extra load on 'from' (and all less-loaded slaves), compared
- * to that of 'to' (the least-loaded slave), is less than ~3%, or
- * it is less than ~1Mbps. No point in rebalancing. */
- break;
- } else if (from->n_hashes == 1) {
- /* 'from' only carries a single MAC hash, so we can't shift any
- * load away from it, even though we want to. */
- from++;
- } else {
- /* 'from' is carrying significantly more load than 'to', and that
- * load is split across at least two different hashes. Pick a hash
- * to migrate to 'to' (the least-loaded slave), given that doing so
- * must decrease the ratio of the load on the two slaves by at
- * least 0.1.
- *
- * The sort order we use means that we prefer to shift away the
- * smallest hashes instead of the biggest ones. There is little
- * reason behind this decision; we could use the opposite sort
- * order to shift away big hashes ahead of small ones. */
- bool order_swapped;
-
- for (i = 0; i < from->n_hashes; i++) {
- double old_ratio, new_ratio;
- uint64_t delta = from->hashes[i]->tx_bytes;
-
- if (delta == 0 || from->tx_bytes - delta == 0) {
- /* Pointless move. */
- continue;
- }
+ /* Re-configure SSL. We do this on every trip through the main loop,
+ * instead of just when the database changes, because the contents of the
+ * key and certificate files can change without the database changing.
+ *
+ * We do this before bridge_reconfigure() because that function might
+ * initiate SSL connections and thus requires SSL to be configured. */
+ if (cfg && cfg->ssl) {
+ const struct ovsrec_ssl *ssl = cfg->ssl;
- order_swapped = from->tx_bytes - delta < to->tx_bytes + delta;
+ stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
+ stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
+ }
- if (to->tx_bytes == 0) {
- /* Nothing on the new slave, move it. */
+ 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;
}
+ }
+ }
- old_ratio = (double)from->tx_bytes / to->tx_bytes;
- new_ratio = (double)(from->tx_bytes - delta) /
- (to->tx_bytes + delta);
+ 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);
+ }
+ }
+ }
- if (new_ratio == 0) {
- /* Should already be covered but check to prevent division
- * by zero. */
- continue;
- }
+ 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 {
+ bridge_reconfigure_continue(&null_cfg);
+ }
+ }
- if (new_ratio < 1) {
- new_ratio = 1 / new_ratio;
- }
+ if (reconf_txn) {
+ ovsdb_idl_txn_commit(reconf_txn);
+ ovsdb_idl_txn_destroy(reconf_txn);
+ reconf_txn = NULL;
+ }
- if (old_ratio - new_ratio > 0.1) {
- /* Would decrease the ratio, move it. */
- break;
+ /* Refresh interface and mirror stats if necessary. */
+ if (time_msec() >= iface_stats_timer) {
+ if (cfg) {
+ struct ovsdb_idl_txn *txn;
+
+ 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;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_refresh_stats(iface);
+ iface_refresh_status(iface);
+ }
}
- }
- if (i < from->n_hashes) {
- bond_shift_load(from, to, i);
- /* If the result of the migration changed the relative order of
- * 'from' and 'to' swap them back to maintain invariants. */
- if (order_swapped) {
- swap_bals(from, to);
+ HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
+ mirror_refresh_stats(m);
}
- /* Re-sort 'bals'. Note that this may make 'from' and 'to'
- * point to different slave_balance structures. It is only
- * valid to do these two operations in a row at all because we
- * know that 'from' will not move past 'to' and vice versa. */
- resort_bals(from, bals, n_bals);
- resort_bals(to, bals, n_bals);
- } else {
- from++;
}
+ refresh_controller_status();
+ ovsdb_idl_txn_commit(txn);
+ ovsdb_idl_txn_destroy(txn); /* XXX */
}
- }
- /* Implement exponentially weighted moving average. A weight of 1/2 causes
- * historical data to decay to <1% in 7 rebalancing runs. */
- for (e = &port->bond_hash[0]; e <= &port->bond_hash[BOND_MASK]; e++) {
- e->tx_bytes /= 2;
+ iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
}
-exit:
- free(bals);
+ run_system_stats();
+ refresh_instant_stats();
}
-static void
-bond_send_learning_packets(struct port *port)
+void
+bridge_wait(void)
{
- struct bridge *br = port->bridge;
- struct mac_entry *e;
- struct ofpbuf packet;
- int error, n_packets, n_errors;
-
- if (!port->n_ifaces || port->active_iface < 0 || bond_is_tcp_hash(port)) {
- return;
- }
+ struct sset types;
+ const char *type;
- ofpbuf_init(&packet, 128);
- error = n_packets = n_errors = 0;
- LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
- union ofp_action actions[2], *a;
- uint16_t dp_ifidx;
- tag_type tags = 0;
- struct flow flow;
- int retval;
+ ovsdb_idl_wait(idl);
- if (e->port == port->port_idx) {
- continue;
- }
+ if (reconfiguring) {
+ poll_immediate_wake();
+ }
- compose_benign_packet(&packet, "Open vSwitch Bond Failover", 0xf177,
- e->mac);
- flow_extract(&packet, 0, ODPP_NONE, &flow);
+ sset_init(&types);
+ ofproto_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ ofproto_type_wait(type);
+ }
+ sset_destroy(&types);
- if (!choose_output_iface(port, &flow, e->vlan, &dp_ifidx, &tags)) {
- continue;
- }
+ if (!hmap_is_empty(&all_bridges)) {
+ struct bridge *br;
- /* Compose actions. */
- memset(actions, 0, sizeof actions);
- a = actions;
- if (e->vlan) {
- a->vlan_vid.type = htons(OFPAT_SET_VLAN_VID);
- a->vlan_vid.len = htons(sizeof *a);
- a->vlan_vid.vlan_vid = htons(e->vlan);
- a++;
- }
- a->output.type = htons(OFPAT_OUTPUT);
- a->output.len = htons(sizeof *a);
- a->output.port = htons(odp_port_to_ofp_port(dp_ifidx));
- a++;
-
- /* Send packet. */
- n_packets++;
- retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions,
- &packet);
- if (retval) {
- error = retval;
- n_errors++;
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_wait(br->ofproto);
}
+ poll_timer_wait_until(iface_stats_timer);
}
- ofpbuf_uninit(&packet);
- if (n_errors) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bond %s: %d errors sending %d gratuitous learning "
- "packets, last error was: %s",
- port->name, n_errors, n_packets, strerror(error));
- } else {
- VLOG_DBG("bond %s: sent %d gratuitous learning packets",
- port->name, n_packets);
+ system_stats_wait();
+}
+
+/* Adds some memory usage statistics for bridges into 'usage', for use with
+ * memory_report(). */
+void
+bridge_get_memory_usage(struct simap *usage)
+{
+ struct bridge *br;
+
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_get_memory_usage(br->ofproto, usage);
}
}
\f
-/* Bonding unixctl user interface functions. */
+/* QoS unixctl user interface functions. */
+
+struct qos_unixctl_show_cbdata {
+ struct ds *ds;
+ struct iface *iface;
+};
static void
-bond_unixctl_list(struct unixctl_conn *conn,
- const char *args OVS_UNUSED, void *aux OVS_UNUSED)
+qos_unixctl_show_cb(unsigned int queue_id,
+ const struct smap *details,
+ void *aux)
{
- struct ds ds = DS_EMPTY_INITIALIZER;
- const struct bridge *br;
-
- ds_put_cstr(&ds, "bridge\tbond\ttype\tslaves\n");
+ struct qos_unixctl_show_cbdata *data = aux;
+ struct ds *ds = data->ds;
+ struct iface *iface = data->iface;
+ struct netdev_queue_stats stats;
+ struct smap_node *node;
+ int error;
- LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ ds_put_cstr(ds, "\n");
+ if (queue_id) {
+ ds_put_format(ds, "Queue %u:\n", queue_id);
+ } else {
+ ds_put_cstr(ds, "Default:\n");
+ }
- for (i = 0; i < br->n_ports; i++) {
- const struct port *port = br->ports[i];
- if (port->n_ifaces > 1) {
- size_t j;
-
- ds_put_format(&ds, "%s\t%s\t%s\t", br->name, port->name,
- bond_mode_to_string(port->bond_mode));
- for (j = 0; j < port->n_ifaces; j++) {
- const struct iface *iface = port->ifaces[j];
- if (j) {
- ds_put_cstr(&ds, ", ");
- }
- ds_put_cstr(&ds, iface->name);
- }
- ds_put_char(&ds, '\n');
- }
- }
+ SMAP_FOR_EACH (node, details) {
+ ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
}
- unixctl_command_reply(conn, 200, ds_cstr(&ds));
- ds_destroy(&ds);
-}
-static struct port *
-bond_find(const char *name)
-{
- const struct bridge *br;
+ error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
+ if (!error) {
+ if (stats.tx_packets != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
+ }
- LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ if (stats.tx_bytes != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
+ }
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (!strcmp(port->name, name) && port->n_ifaces > 1) {
- return port;
- }
+ if (stats.tx_errors != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
}
+ } else {
+ ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
+ queue_id, strerror(error));
}
- return NULL;
}
static void
-bond_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;
- const struct port *port;
- size_t j;
+ struct smap smap = SMAP_INITIALIZER(&smap);
+ struct iface *iface;
+ const char *type;
+ struct smap_node *node;
+ struct qos_unixctl_show_cbdata data;
+ int error;
- port = bond_find(args);
- if (!port) {
- unixctl_command_reply(conn, 501, "no such bond");
+ iface = iface_find(argv[1]);
+ if (!iface) {
+ unixctl_command_reply_error(conn, "no such interface");
return;
}
- ds_put_format(&ds, "bond_mode: %s\n",
- bond_mode_to_string(port->bond_mode));
+ netdev_get_qos(iface->netdev, &type, &smap);
+
+ if (*type != '\0') {
+ ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
+
+ SMAP_FOR_EACH (node, &smap) {
+ ds_put_format(&ds, "%s: %s\n", node->key, node->value);
+ }
+
+ data.ds = &ds;
+ data.iface = iface;
+ error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
- if (port->lacp) {
- ds_put_format(&ds, "lacp: %s\n",
- port->lacp_active ? "active" : "passive");
+ if (error) {
+ ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
+ }
+ unixctl_command_reply(conn, ds_cstr(&ds));
} else {
- ds_put_cstr(&ds, "lacp: off\n");
+ ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
+ unixctl_command_reply_error(conn, ds_cstr(&ds));
}
- if (port->bond_mode != BM_AB) {
- ds_put_format(&ds, "bond-hash-algorithm: %s\n",
- bond_is_tcp_hash(port) ? "balance-tcp" : "balance-slb");
- }
+ smap_destroy(&smap);
+ ds_destroy(&ds);
+}
+\f
+/* Bridge reconfiguration functions. */
+static void
+bridge_create(const struct ovsrec_bridge *br_cfg)
+{
+ struct bridge *br;
+
+ ovs_assert(!bridge_lookup(br_cfg->name));
+ br = xzalloc(sizeof *br);
+ br->name = xstrdup(br_cfg->name);
+ br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
+ br->cfg = br_cfg;
- ds_put_format(&ds, "bond-detect-mode: %s\n",
- port->monitor ? "carrier" : "miimon");
+ /* Derive the default Ethernet address from the bridge's UUID. This should
+ * be unique and it will be stable between ovs-vswitchd runs. */
+ memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
+ eth_addr_mark_random(br->default_ea);
- if (!port->monitor) {
- ds_put_format(&ds, "bond-miimon-interval: %lld\n",
- port->miimon_interval);
- }
+ hmap_init(&br->ports);
+ hmap_init(&br->ifaces);
+ hmap_init(&br->iface_by_name);
+ hmap_init(&br->mirrors);
- ds_put_format(&ds, "updelay: %d ms\n", port->updelay);
- ds_put_format(&ds, "downdelay: %d ms\n", port->downdelay);
+ hmap_init(&br->if_cfg_todo);
+ list_init(&br->ofpp_garbage);
- if (port->bond_mode != BM_AB) {
- ds_put_format(&ds, "next rebalance: %lld ms\n",
- port->bond_next_rebalance - time_msec());
- }
+ hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
+}
- for (j = 0; j < port->n_ifaces; j++) {
- const struct iface *iface = port->ifaces[j];
- struct bond_entry *be;
- struct flow flow;
+static void
+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;
- /* Basic info. */
- ds_put_format(&ds, "\nslave %s: %s\n",
- iface->name, iface->enabled ? "enabled" : "disabled");
- if (j == port->active_iface) {
- ds_put_cstr(&ds, "\tactive slave\n");
+ HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
+ port_destroy(port);
}
- if (iface->delay_expires != LLONG_MAX) {
- ds_put_format(&ds, "\t%s expires in %lld ms\n",
- iface->enabled ? "downdelay" : "updelay",
- iface->delay_expires - time_msec());
+ HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
+ mirror_destroy(mirror);
}
-
- if (port->bond_mode == BM_AB) {
- continue;
+ 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);
}
- /* Hashes. */
- memset(&flow, 0, sizeof flow);
- for (be = port->bond_hash; be <= &port->bond_hash[BOND_MASK]; be++) {
- int hash = be - port->bond_hash;
- struct mac_entry *me;
-
- if (be->iface_idx != j) {
- continue;
- }
-
- ds_put_format(&ds, "\thash %d: %"PRIu64" kB load\n",
- hash, be->tx_bytes / 1024);
+ 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);
+ }
+}
- if (port->bond_mode != BM_SLB) {
- continue;
- }
+static struct bridge *
+bridge_lookup(const char *name)
+{
+ struct bridge *br;
- /* MACs. */
- LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) {
- uint16_t dp_ifidx;
- tag_type tags = 0;
-
- memcpy(flow.dl_src, me->mac, ETH_ADDR_LEN);
- if (bond_hash_src(me->mac, me->vlan) == hash
- && me->port != port->port_idx
- && choose_output_iface(port, &flow, me->vlan,
- &dp_ifidx, &tags)
- && dp_ifidx == iface->dp_ifidx)
- {
- ds_put_format(&ds, "\t\t"ETH_ADDR_FMT"\n",
- ETH_ADDR_ARGS(me->mac));
- }
- }
+ HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
+ if (!strcmp(br->name, name)) {
+ return br;
}
}
- unixctl_command_reply(conn, 200, ds_cstr(&ds));
- ds_destroy(&ds);
+ return NULL;
}
+/* Handle requests for a listing of all flows known by the OpenFlow
+ * stack, including those normally hidden. */
static void
-bond_unixctl_migrate(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)
{
- char *args = (char *) args_;
- char *save_ptr = NULL;
- char *bond_s, *hash_s, *slave_s;
- struct port *port;
- struct iface *iface;
- struct bond_entry *entry;
- int hash;
-
- bond_s = strtok_r(args, " ", &save_ptr);
- hash_s = strtok_r(NULL, " ", &save_ptr);
- slave_s = strtok_r(NULL, " ", &save_ptr);
- if (!slave_s) {
- unixctl_command_reply(conn, 501,
- "usage: bond/migrate BOND HASH SLAVE");
- return;
- }
+ struct bridge *br;
+ struct ds results;
- port = bond_find(bond_s);
- if (!port) {
- unixctl_command_reply(conn, 501, "no such bond");
+ br = bridge_lookup(argv[1]);
+ if (!br) {
+ unixctl_command_reply_error(conn, "Unknown bridge");
return;
}
- if (port->bond_mode != BM_SLB) {
- unixctl_command_reply(conn, 501, "not an SLB bond");
- return;
- }
+ ds_init(&results);
+ ofproto_get_all_flows(br->ofproto, &results);
+
+ unixctl_command_reply(conn, ds_cstr(&results));
+ ds_destroy(&results);
+}
- if (strspn(hash_s, "0123456789") == strlen(hash_s)) {
- hash = atoi(hash_s) & BOND_MASK;
+/* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
+ * 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, int argc,
+ const char *argv[], void *aux OVS_UNUSED)
+{
+ struct bridge *br;
+ if (argc > 1) {
+ br = bridge_lookup(argv[1]);
+ if (!br) {
+ unixctl_command_reply_error(conn, "Unknown bridge");
+ return;
+ }
+ ofproto_reconnect_controllers(br->ofproto);
} else {
- unixctl_command_reply(conn, 501, "bad hash");
- return;
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_reconnect_controllers(br->ofproto);
+ }
}
+ unixctl_command_reply(conn, NULL);
+}
- iface = port_lookup_iface(port, slave_s);
- if (!iface) {
- unixctl_command_reply(conn, 501, "no such slave");
- return;
+static size_t
+bridge_get_controllers(const struct bridge *br,
+ struct ovsrec_controller ***controllersp)
+{
+ struct ovsrec_controller **controllers;
+ size_t n_controllers;
+
+ controllers = br->cfg->controller;
+ n_controllers = br->cfg->n_controller;
+
+ if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
+ controllers = NULL;
+ n_controllers = 0;
}
- if (!iface->enabled) {
- unixctl_command_reply(conn, 501, "cannot migrate to disabled slave");
- return;
+ if (controllersp) {
+ *controllersp = controllers;
}
+ return n_controllers;
+}
- entry = &port->bond_hash[hash];
- ofproto_revalidate(port->bridge->ofproto, entry->iface_tag);
- entry->iface_idx = iface->port_ifidx;
- entry->iface_tag = tag_create_random();
- unixctl_command_reply(conn, 200, "migrated");
+static void
+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;
+ if_cfg->ofport = iface_pick_ofport(cfg);
+ 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
-bond_unixctl_set_active_slave(struct unixctl_conn *conn, const char *args_,
- void *aux OVS_UNUSED)
+bridge_add_del_ports(struct bridge *br,
+ const unsigned long int *splinter_vlans)
{
- char *args = (char *) args_;
- char *save_ptr = NULL;
- char *bond_s, *slave_s;
- struct port *port;
- struct iface *iface;
+ struct shash_node *port_node;
+ struct port *port, *next;
+ struct shash new_ports;
+ size_t i;
- bond_s = strtok_r(args, " ", &save_ptr);
- slave_s = strtok_r(NULL, " ", &save_ptr);
- if (!slave_s) {
- unixctl_command_reply(conn, 501,
- "usage: bond/set-active-slave BOND SLAVE");
- return;
- }
+ ovs_assert(hmap_is_empty(&br->if_cfg_todo));
- port = bond_find(bond_s);
- if (!port) {
- unixctl_command_reply(conn, 501, "no such bond");
- return;
+ /* Collect new ports. */
+ shash_init(&new_ports);
+ for (i = 0; i < br->cfg->n_ports; i++) {
+ const char *name = br->cfg->ports[i]->name;
+ if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
+ VLOG_WARN("bridge %s: %s specified twice as bridge port",
+ br->name, name);
+ }
}
+ if (bridge_get_controllers(br, NULL)
+ && !shash_find(&new_ports, br->name)) {
+ VLOG_WARN("bridge %s: no port named %s, synthesizing one",
+ br->name, br->name);
- iface = port_lookup_iface(port, slave_s);
- if (!iface) {
- unixctl_command_reply(conn, 501, "no such slave");
- return;
- }
+ ovsrec_interface_init(&br->synth_local_iface);
+ ovsrec_port_init(&br->synth_local_port);
- if (!iface->enabled) {
- unixctl_command_reply(conn, 501, "cannot make disabled slave active");
- return;
- }
+ br->synth_local_port.interfaces = &br->synth_local_ifacep;
+ br->synth_local_port.n_interfaces = 1;
+ br->synth_local_port.name = br->name;
- if (port->active_iface != iface->port_ifidx) {
- ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
- port->active_iface = iface->port_ifidx;
- port->active_iface_tag = tag_create_random();
- VLOG_INFO("port %s: active interface is now %s",
- port->name, iface->name);
- bond_send_learning_packets(port);
- unixctl_command_reply(conn, 200, "done");
- } else {
- unixctl_command_reply(conn, 200, "no change");
- }
-}
+ br->synth_local_iface.name = br->name;
+ br->synth_local_iface.type = "internal";
-static void
-enable_slave(struct unixctl_conn *conn, const char *args_, bool enable)
-{
- char *args = (char *) args_;
- char *save_ptr = NULL;
- char *bond_s, *slave_s;
- struct port *port;
- struct iface *iface;
+ br->synth_local_ifacep = &br->synth_local_iface;
- bond_s = strtok_r(args, " ", &save_ptr);
- slave_s = strtok_r(NULL, " ", &save_ptr);
- if (!slave_s) {
- unixctl_command_reply(conn, 501,
- "usage: bond/enable/disable-slave BOND SLAVE");
- return;
+ shash_add(&new_ports, br->name, &br->synth_local_port);
}
- port = bond_find(bond_s);
- if (!port) {
- unixctl_command_reply(conn, 501, "no such bond");
- return;
+ if (splinter_vlans) {
+ add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
}
- iface = port_lookup_iface(port, slave_s);
- if (!iface) {
- unixctl_command_reply(conn, 501, "no such slave");
- return;
+ /* Get rid of deleted ports.
+ * 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) {
+ port_destroy(port);
+ } else {
+ port_del_ifaces(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);
+ const char *type = iface_get_type(cfg, br->cfg);
+
+ if (iface) {
+ iface->cfg = cfg;
+ iface->type = type;
+ } else if (!strcmp(type, "null")) {
+ VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
+ " may be removed in February 2013. Please email"
+ " dev@openvswitch.org with concerns.",
+ cfg->name);
+ } else {
+ bridge_queue_if_cfg(br, cfg, port);
+ }
+ }
}
- bond_enable_slave(iface, enable);
- unixctl_command_reply(conn, 501, enable ? "enabled" : "disabled");
+ shash_destroy(&new_ports);
}
+/* Initializes 'oc' appropriately as a management service controller for
+ * 'br'.
+ *
+ * The caller must free oc->target when it is no longer needed. */
static void
-bond_unixctl_enable_slave(struct unixctl_conn *conn, const char *args,
- void *aux OVS_UNUSED)
+bridge_ofproto_controller_for_mgmt(const struct bridge *br,
+ struct ofproto_controller *oc)
{
- enable_slave(conn, args, true);
+ oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
+ oc->max_backoff = 0;
+ oc->probe_interval = 60;
+ 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'. */
static void
-bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args,
- void *aux OVS_UNUSED)
+bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
+ struct ofproto_controller *oc)
{
- enable_slave(conn, args, false);
+ int dscp;
+
+ 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;
+ oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
+ ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
+ 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);
+ dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
+ if (dscp < 0 || dscp > 63) {
+ dscp = DSCP_DEFAULT;
+ }
+ oc->dscp = dscp;
}
+/* Configures the IP stack for 'br''s local interface properly according to the
+ * configuration in 'c'. */
static void
-bond_unixctl_hash(struct unixctl_conn *conn, const char *args_,
- void *aux OVS_UNUSED)
+bridge_configure_local_iface_netdev(struct bridge *br,
+ struct ovsrec_controller *c)
{
- char *args = (char *) args_;
- uint8_t mac[ETH_ADDR_LEN];
- uint8_t hash;
- char *hash_cstr;
- unsigned int vlan;
- char *mac_s, *vlan_s;
- char *save_ptr = NULL;
-
- mac_s = strtok_r(args, " ", &save_ptr);
- vlan_s = strtok_r(NULL, " ", &save_ptr);
-
- if (vlan_s) {
- if (sscanf(vlan_s, "%u", &vlan) != 1) {
- unixctl_command_reply(conn, 501, "invalid vlan");
- return;
- }
- } else {
- vlan = OFP_VLAN_NONE;
+ struct netdev *netdev;
+ struct in_addr mask, gateway;
+
+ struct iface *local_iface;
+ struct in_addr ip;
+
+ /* If there's no local interface or no IP address, give up. */
+ local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
+ if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
+ return;
}
- if (sscanf(mac_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac))
- == ETH_ADDR_SCAN_COUNT) {
- hash = bond_hash_src(mac, vlan);
+ /* Bring up the local interface. */
+ netdev = local_iface->netdev;
+ netdev_turn_flags_on(netdev, NETDEV_UP, true);
- hash_cstr = xasprintf("%u", hash);
- unixctl_command_reply(conn, 200, hash_cstr);
- free(hash_cstr);
- } else {
- unixctl_command_reply(conn, 501, "invalid mac");
+ /* Configure the IP address and netmask. */
+ if (!c->local_netmask
+ || !inet_aton(c->local_netmask, &mask)
+ || !mask.s_addr) {
+ mask.s_addr = guess_netmask(ip.s_addr);
+ }
+ if (!netdev_set_in4(netdev, ip, mask)) {
+ VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
+ br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
+ }
+
+ /* Configure the default gateway. */
+ if (c->local_gateway
+ && inet_aton(c->local_gateway, &gateway)
+ && gateway.s_addr) {
+ if (!netdev_add_router(netdev, gateway)) {
+ VLOG_INFO("bridge %s: configured gateway "IP_FMT,
+ br->name, IP_ARGS(gateway.s_addr));
+ }
}
}
-static void
-bond_init(void)
+/* 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".
+ *
+ * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
+ * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
+ * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
+ * 'b' against a prefix of 'a'.
+ */
+static bool
+equal_pathnames(const char *a, const char *b, size_t b_stoplen)
{
- unixctl_command_register("bond/list", bond_unixctl_list, NULL);
- unixctl_command_register("bond/show", bond_unixctl_show, NULL);
- unixctl_command_register("bond/migrate", bond_unixctl_migrate, NULL);
- unixctl_command_register("bond/set-active-slave",
- bond_unixctl_set_active_slave, NULL);
- unixctl_command_register("bond/enable-slave", bond_unixctl_enable_slave,
- NULL);
- unixctl_command_register("bond/disable-slave", bond_unixctl_disable_slave,
- NULL);
- unixctl_command_register("bond/hash", bond_unixctl_hash, NULL);
+ const char *b_start = b;
+ for (;;) {
+ if (b - b_start >= b_stoplen) {
+ return true;
+ } else if (*a != *b) {
+ return false;
+ } else if (*a == '/') {
+ a += strspn(a, "/");
+ b += strspn(b, "/");
+ } else if (*a == '\0') {
+ return true;
+ } else {
+ a++;
+ b++;
+ }
+ }
}
-\f
-/* Port functions. */
static void
-lacp_send_pdu_cb(void *aux, const struct lacp_pdu *pdu)
+bridge_configure_remotes(struct bridge *br,
+ const struct sockaddr_in *managers, size_t n_managers)
{
- struct iface *iface = aux;
- uint8_t ea[ETH_ADDR_LEN];
- int error;
+ bool disable_in_band;
- error = netdev_get_etheraddr(iface->netdev, ea);
- if (!error) {
- struct ofpbuf packet;
+ struct ovsrec_controller **controllers;
+ size_t n_controllers;
+
+ enum ofproto_fail_mode fail_mode;
+
+ struct ofproto_controller *ocs;
+ size_t n_ocs;
+ size_t i;
+
+ /* Check if we should disable in-band control on this bridge. */
+ disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
+ false);
+
+ /* Set OpenFlow queue ID for in-band control. */
+ ofproto_set_in_band_queue(br->ofproto,
+ smap_get_int(&br->cfg->other_config,
+ "in-band-queue", -1));
- ofpbuf_init(&packet, ETH_HEADER_LEN + LACP_PDU_LEN);
- compose_lacp_packet(&packet, ea, pdu);
- iface_send_packet(iface, &packet);
- ofpbuf_uninit(&packet);
+ if (disable_in_band) {
+ ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
} else {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
- VLOG_ERR_RL(&rl, "iface %s: failed to obtain Ethernet address "
- "(%s)", iface->name, strerror(error));
+ ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
}
-}
-static void
-port_run(struct port *port)
-{
- size_t i;
+ n_controllers = bridge_get_controllers(br, &controllers);
- if (port->monitor) {
- char *devname;
+ ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
+ n_ocs = 0;
- /* Track carrier going up and down on interfaces. */
- while (!netdev_monitor_poll(port->monitor, &devname)) {
- struct iface *iface;
+ bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
+ for (i = 0; i < n_controllers; i++) {
+ struct ovsrec_controller *c = controllers[i];
- iface = port_lookup_iface(port, devname);
- if (iface) {
- iface_update_carrier(iface, netdev_get_carrier(iface->netdev));
+ 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;
+
+ if (!strncmp(c->target, "unix:", 5)) {
+ /* Connect to a listening socket */
+ whitelist = xasprintf("unix:%s/", ovs_rundir());
+ if (strchr(c->target, '/') &&
+ !equal_pathnames(c->target, whitelist,
+ strlen(whitelist))) {
+ /* Absolute path specified, but not in ovs_rundir */
+ VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
+ "controller \"%s\" due to possibility for "
+ "remote exploit. Instead, specify socket "
+ "in 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;
+ }
+ } else {
+ whitelist = xasprintf("punix:%s/%s.controller",
+ ovs_rundir(), br->name);
+ if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
+ /* 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 of "
+ "overwriting local files. 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;
+ }
}
- free(devname);
- }
- } else if (time_msec() >= port->miimon_next_update) {
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- iface_update_carrier(iface, netdev_get_miimon(iface->netdev));
+ free(whitelist);
}
- port->miimon_next_update = time_msec() + port->miimon_interval;
- }
- if (port->lacp) {
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- lacp_slave_enable(port->lacp, iface, iface->enabled);
+ bridge_configure_local_iface_netdev(br, c);
+ bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
+ if (disable_in_band) {
+ ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
}
-
- lacp_run(port->lacp, lacp_send_pdu_cb);
+ n_ocs++;
}
- bond_run(port);
+ ofproto_set_controllers(br->ofproto, ocs, n_ocs,
+ bridge_get_allowed_versions(br));
+ free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
+ free(ocs);
+
+ /* Set the fail-mode. */
+ fail_mode = !br->cfg->fail_mode
+ || !strcmp(br->cfg->fail_mode, "standalone")
+ ? OFPROTO_FAIL_STANDALONE
+ : OFPROTO_FAIL_SECURE;
+ ofproto_set_fail_mode(br->ofproto, fail_mode);
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
+ /* Configure OpenFlow controller connection snooping. */
+ if (!ofproto_has_snoops(br->ofproto)) {
+ struct sset snoops;
- if (iface->cfm) {
- struct ofpbuf *packet = cfm_run(iface->cfm);
- if (packet) {
- iface_send_packet(iface, packet);
- ofpbuf_uninit(packet);
- free(packet);
- }
- }
+ sset_init(&snoops);
+ sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
+ ovs_rundir(), br->name));
+ ofproto_set_snoops(br->ofproto, &snoops);
+ sset_destroy(&snoops);
}
}
static void
-port_wait(struct port *port)
+bridge_configure_tables(struct bridge *br)
{
- size_t i;
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ int n_tables;
+ int i, j;
- if (port->monitor) {
- netdev_monitor_poll_wait(port->monitor);
- } else {
- poll_timer_wait_until(port->miimon_next_update);
- }
+ n_tables = ofproto_get_n_tables(br->ofproto);
+ j = 0;
+ for (i = 0; i < n_tables; i++) {
+ struct ofproto_table_settings s;
- if (port->lacp) {
- lacp_wait(port->lacp);
- }
+ s.name = NULL;
+ s.max_flows = UINT_MAX;
+ s.groups = NULL;
+ s.n_groups = 0;
- bond_wait(port);
+ 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++];
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- if (iface->cfm) {
- cfm_wait(iface->cfm);
+ 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]);
+ }
+}
+
+static void
+bridge_configure_dp_desc(struct bridge *br)
+{
+ ofproto_set_dp_desc(br->ofproto,
+ smap_get(&br->cfg->other_config, "dp-desc"));
}
+\f
+/* Port functions. */
static struct port *
-port_create(struct bridge *br, const char *name)
+port_create(struct bridge *br, const struct ovsrec_port *cfg)
{
struct port *port;
port = xzalloc(sizeof *port);
port->bridge = br;
- port->port_idx = br->n_ports;
- port->vlan = -1;
- port->trunks = NULL;
- port->name = xstrdup(name);
- port->active_iface = -1;
-
- if (br->n_ports >= br->allocated_ports) {
- br->ports = x2nrealloc(br->ports, &br->allocated_ports,
- sizeof *br->ports);
- }
- br->ports[br->n_ports++] = port;
- shash_add_assert(&br->port_by_name, port->name, port);
-
- VLOG_INFO("created port %s on bridge %s", port->name, br->name);
- bridge_flush(br);
+ port->name = xstrdup(cfg->name);
+ port->cfg = cfg;
+ list_init(&port->ifaces);
+ hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
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, const struct ovsrec_port *cfg)
+port_del_ifaces(struct port *port)
{
- struct shash new_ifaces;
+ struct iface *iface, *next;
+ struct sset new_ifaces;
size_t i;
/* Collect list of new interfaces. */
- shash_init(&new_ifaces);
- for (i = 0; i < cfg->n_interfaces; i++) {
- const char *name = cfg->interfaces[i]->name;
- shash_add_once(&new_ifaces, name, NULL);
+ 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]->type;
+ if (strcmp(type, "null")) {
+ sset_add(&new_ifaces, name);
+ }
}
/* Get rid of deleted interfaces. */
- for (i = 0; i < port->n_ifaces; ) {
- if (!shash_find(&new_ifaces, cfg->interfaces[i]->name)) {
- iface_destroy(port->ifaces[i]);
- } else {
- i++;
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ if (!sset_contains(&new_ifaces, iface->name)) {
+ iface_destroy(iface);
}
}
- shash_destroy(&new_ifaces);
+ sset_destroy(&new_ifaces);
}
static void
-port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
+port_destroy(struct port *port)
{
- const char *detect_mode;
- struct shash new_ifaces;
- long long int next_rebalance, miimon_next_update, lacp_priority;
- unsigned long *trunks;
- int vlan;
- size_t i;
-
- port->cfg = cfg;
-
- /* Update settings. */
- port->updelay = cfg->bond_updelay;
- if (port->updelay < 0) {
- port->updelay = 0;
- }
- port->downdelay = cfg->bond_downdelay;
- if (port->downdelay < 0) {
- port->downdelay = 0;
- }
- port->bond_rebalance_interval = atoi(
- get_port_other_config(cfg, "bond-rebalance-interval", "10000"));
- if (port->bond_rebalance_interval < 1000) {
- port->bond_rebalance_interval = 1000;
- }
- next_rebalance = time_msec() + port->bond_rebalance_interval;
- if (port->bond_next_rebalance > next_rebalance) {
- port->bond_next_rebalance = next_rebalance;
- }
-
- detect_mode = get_port_other_config(cfg, "bond-detect-mode",
- "carrier");
-
- netdev_monitor_destroy(port->monitor);
- port->monitor = NULL;
-
- if (strcmp(detect_mode, "miimon")) {
- port->monitor = netdev_monitor_create();
-
- if (strcmp(detect_mode, "carrier")) {
- VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
- "defaulting to carrier", port->name, detect_mode);
- }
- }
-
- port->miimon_interval = atoi(
- get_port_other_config(cfg, "bond-miimon-interval", "200"));
- if (port->miimon_interval < 100) {
- port->miimon_interval = 100;
- }
- miimon_next_update = time_msec() + port->miimon_interval;
- if (port->miimon_next_update > miimon_next_update) {
- port->miimon_next_update = miimon_next_update;
- }
-
- if (!port->cfg->bond_mode ||
- !strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_SLB))) {
- port->bond_mode = BM_SLB;
- } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_AB))) {
- port->bond_mode = BM_AB;
- } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_TCP))) {
- port->bond_mode = BM_TCP;
- } else {
- port->bond_mode = BM_SLB;
- VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
- port->name, port->cfg->bond_mode,
- bond_mode_to_string(port->bond_mode));
- }
-
- /* Add new interfaces and update 'cfg' member of existing ones. */
- shash_init(&new_ifaces);
- for (i = 0; i < cfg->n_interfaces; i++) {
- const struct ovsrec_interface *if_cfg = cfg->interfaces[i];
- struct iface *iface;
+ if (port) {
+ struct bridge *br = port->bridge;
+ struct iface *iface, *next;
- if (!shash_add_once(&new_ifaces, if_cfg->name, NULL)) {
- VLOG_WARN("port %s: %s specified twice as port interface",
- port->name, if_cfg->name);
- iface_set_ofport(if_cfg, -1);
- continue;
+ if (br->ofproto) {
+ ofproto_bundle_unregister(br->ofproto, port);
}
- iface = iface_lookup(port->bridge, if_cfg->name);
- if (iface) {
- if (iface->port != port) {
- VLOG_ERR("bridge %s: %s interface is on multiple ports, "
- "removing from %s",
- port->bridge->name, if_cfg->name, iface->port->name);
- continue;
- }
- iface->cfg = if_cfg;
- } else {
- iface = iface_create(port, if_cfg);
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ iface_destroy(iface);
}
- /* 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(if_cfg->name, port->bridge->name) ? "internal"
- : if_cfg->type[0] ? if_cfg->type
- : "system");
-
- lacp_priority =
- atoi(get_interface_other_config(if_cfg, "lacp-port-priority",
- "0"));
-
- if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) {
- iface->lacp_priority = UINT16_MAX;
- } else {
- iface->lacp_priority = lacp_priority;
- }
+ hmap_remove(&br->ports, &port->hmap_node);
+ free(port->name);
+ free(port);
}
- shash_destroy(&new_ifaces);
-
- port->lacp_fast = !strcmp(get_port_other_config(cfg, "lacp-time", "slow"),
- "fast");
+}
- lacp_priority =
- atoi(get_port_other_config(cfg, "lacp-system-priority", "0"));
+static struct port *
+port_lookup(const struct bridge *br, const char *name)
+{
+ struct port *port;
- if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) {
- /* Prefer bondable links if unspecified. */
- port->lacp_priority = port->n_ifaces > 1 ? UINT16_MAX - 1 : UINT16_MAX;
- } else {
- port->lacp_priority = lacp_priority;
+ HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
+ &br->ports) {
+ if (!strcmp(port->name, name)) {
+ return port;
+ }
}
+ return NULL;
+}
+static bool
+enable_lacp(struct port *port, bool *activep)
+{
if (!port->cfg->lacp) {
/* XXX when LACP implementation has been sufficiently tested, enable by
* default and make active on bonded ports. */
- lacp_destroy(port->lacp);
- port->lacp = NULL;
+ return false;
} else if (!strcmp(port->cfg->lacp, "off")) {
- lacp_destroy(port->lacp);
- port->lacp = NULL;
+ return false;
} else if (!strcmp(port->cfg->lacp, "active")) {
- if (!port->lacp) {
- port->lacp = lacp_create();
- }
- port->lacp_active = true;
+ *activep = true;
+ return true;
} else if (!strcmp(port->cfg->lacp, "passive")) {
- if (!port->lacp) {
- port->lacp = lacp_create();
- }
- port->lacp_active = false;
+ *activep = false;
+ return true;
} else {
VLOG_WARN("port %s: unknown LACP mode %s",
port->name, port->cfg->lacp);
- lacp_destroy(port->lacp);
- port->lacp = NULL;
+ return false;
}
+}
- /* Get VLAN tag. */
- vlan = -1;
- if (cfg->tag) {
- if (port->n_ifaces < 2) {
- vlan = *cfg->tag;
- if (vlan >= 0 && vlan <= 4095) {
- VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
- } else {
- vlan = -1;
- }
- } 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 (port->vlan != vlan) {
- port->vlan = vlan;
- bridge_flush(port->bridge);
+static struct lacp_settings *
+port_configure_lacp(struct port *port, struct lacp_settings *s)
+{
+ const char *lacp_time, *system_id;
+ int priority;
+
+ if (!enable_lacp(port, &s->active)) {
+ return NULL;
}
- /* Get trunked VLANs. */
- trunks = NULL;
- if (vlan < 0 && cfg->n_trunks) {
- size_t n_errors;
+ s->name = port->name;
- trunks = bitmap_allocate(4096);
- n_errors = 0;
- for (i = 0; i < cfg->n_trunks; i++) {
- int trunk = cfg->trunks[i];
- if (trunk >= 0) {
- bitmap_set1(trunks, trunk);
- } else {
- n_errors++;
- }
+ system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
+ 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;
}
- if (n_errors) {
- VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
- port->name, cfg->n_trunks);
- }
- if (n_errors == cfg->n_trunks) {
- VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
- port->name);
- bitmap_free(trunks);
- trunks = NULL;
- }
- } else if (vlan >= 0 && cfg->n_trunks) {
- VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
- port->name);
+ } else {
+ memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
}
- if (trunks == NULL
- ? port->trunks != NULL
- : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
- bridge_flush(port->bridge);
+
+ if (eth_addr_is_zero(s->id)) {
+ VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
+ return NULL;
}
- bitmap_free(port->trunks);
- port->trunks = trunks;
+
+ /* Prefer bondable links if unspecified. */
+ priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
+ 0);
+ s->priority = (priority > 0 && priority <= UINT16_MAX
+ ? priority
+ : UINT16_MAX - !list_is_short(&port->ifaces));
+
+ lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
+ s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
+ return s;
}
static void
-port_destroy(struct port *port)
+iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
{
- if (port) {
- struct bridge *br = port->bridge;
- struct port *del;
- int i;
-
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m && m->out_port == port) {
- mirror_destroy(m);
- }
- }
-
- while (port->n_ifaces > 0) {
- iface_destroy(port->ifaces[port->n_ifaces - 1]);
- }
+ int priority, portid, key;
- shash_find_and_delete_assert(&br->port_by_name, port->name);
+ portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
+ priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
+ 0);
+ key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
- del = br->ports[port->port_idx] = br->ports[--br->n_ports];
- del->port_idx = port->port_idx;
-
- VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
-
- netdev_monitor_destroy(port->monitor);
- free(port->ifaces);
- bitmap_free(port->trunks);
- free(port->name);
- free(port);
- bridge_flush(br);
+ if (portid <= 0 || portid > UINT16_MAX) {
+ portid = iface->ofp_port;
}
-}
-static struct port *
-port_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
-{
- struct iface *iface = iface_from_dp_ifidx(br, dp_ifidx);
- return iface ? iface->port : NULL;
-}
+ if (priority <= 0 || priority > UINT16_MAX) {
+ priority = UINT16_MAX;
+ }
-static struct port *
-port_lookup(const struct bridge *br, const char *name)
-{
- return shash_find_data(&br->port_by_name, name);
-}
+ if (key < 0 || key > UINT16_MAX) {
+ key = 0;
+ }
-static struct iface *
-port_lookup_iface(const struct port *port, const char *name)
-{
- struct iface *iface = iface_lookup(port->bridge, name);
- return iface && iface->port == port ? iface : NULL;
+ s->name = iface->name;
+ s->id = portid;
+ s->priority = priority;
+ s->key = key;
}
static void
-port_update_lacp(struct port *port)
+port_configure_bond(struct port *port, struct bond_settings *s)
{
- if (port->lacp) {
- size_t i;
-
- lacp_configure(port->lacp, port->name,
- port->bridge->ea, port->lacp_priority,
- port->lacp_active, port->lacp_fast);
+ const char *detect_s;
+ struct iface *iface;
+ int miimon_interval;
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- lacp_slave_register(port->lacp, iface, iface->name,
- iface->dp_ifidx, iface->lacp_priority);
+ s->name = port->name;
+ 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));
}
- }
-}
-
-static void
-port_update_bonding(struct port *port)
-{
- if (port->n_ifaces < 2) {
- /* Not a bonded port. */
- free(port->bond_hash);
- port->bond_hash = NULL;
- port->bond_fake_iface = false;
- port->active_iface = -1;
- port->no_ifaces_tag = 0;
} else {
- size_t i;
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
- if (port->bond_mode != BM_AB && !port->bond_hash) {
- port->bond_hash = xcalloc(BOND_MASK + 1, sizeof *port->bond_hash);
- for (i = 0; i <= BOND_MASK; i++) {
- struct bond_entry *e = &port->bond_hash[i];
- e->iface_idx = -1;
- e->tx_bytes = 0;
- }
- port->bond_next_rebalance
- = time_msec() + port->bond_rebalance_interval;
- } else if (port->bond_mode == BM_AB) {
- free(port->bond_hash);
- port->bond_hash = NULL;
- }
+ /* 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, "
+ "please use another bond type or disable flood_vlans",
+ port->name);
+ }
- if (!port->no_ifaces_tag) {
- port->no_ifaces_tag = tag_create_random();
- }
+ miimon_interval = smap_get_int(&port->cfg->other_config,
+ "bond-miimon-interval", 0);
+ if (miimon_interval <= 0) {
+ miimon_interval = 200;
+ }
- if (port->active_iface < 0) {
- bond_choose_active_iface(port);
- }
+ detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
+ if (!detect_s || !strcmp(detect_s, "carrier")) {
+ miimon_interval = 0;
+ } else if (strcmp(detect_s, "miimon")) {
+ VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
+ "defaulting to carrier", port->name, detect_s);
+ miimon_interval = 0;
+ }
- port->bond_fake_iface = port->cfg->bond_fake_iface;
- if (port->bond_fake_iface) {
- port->bond_next_fake_iface_update = time_msec();
- }
+ s->up_delay = MAX(0, port->cfg->bond_updelay);
+ s->down_delay = MAX(0, port->cfg->bond_downdelay);
+ s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
+ s->rebalance_interval = smap_get_int(&port->cfg->other_config,
+ "bond-rebalance-interval", 10000);
+ if (s->rebalance_interval && s->rebalance_interval < 1000) {
+ s->rebalance_interval = 1000;
+ }
+
+ s->fake_iface = port->cfg->bond_fake_iface;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ 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_);
+}
\f
/* Interface functions. */
-static void
-iface_send_packet(struct iface *iface, struct ofpbuf *packet)
+static bool
+iface_is_internal(const struct ovsrec_interface *iface,
+ const struct ovsrec_bridge *br)
{
- struct flow flow;
- union ofp_action action;
-
- memset(&action, 0, sizeof action);
- action.output.type = htons(OFPAT_OUTPUT);
- action.output.len = htons(sizeof action);
- action.output.port = htons(odp_port_to_ofp_port(iface->dp_ifidx));
-
- flow_extract(packet, 0, ODPP_NONE, &flow);
-
- if (ofproto_send_packet(iface->port->bridge->ofproto, &flow, &action, 1,
- packet)) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "interface %s: Failed to send packet.", iface->name);
- }
+ /* The local port and "internal" ports are always "internal". */
+ return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
}
-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->port_ifidx = port->n_ifaces;
- iface->name = xstrdup(name);
- iface->dp_ifidx = -1;
- iface->tag = tag_create_random();
- iface->delay_expires = LLONG_MAX;
- iface->netdev = NULL;
- iface->cfg = if_cfg;
-
- shash_add_assert(&br->iface_by_name, iface->name, iface);
+ const char *type;
- if (port->n_ifaces >= port->allocated_ifaces) {
- port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces,
- sizeof *port->ifaces);
- }
- port->ifaces[port->n_ifaces++] = iface;
- if (port->n_ifaces > 1) {
- br->has_bonded_ports = true;
+ /* The local port always has type "internal". Other ports take
+ * their type from the database and default to "system" if none is
+ * specified. */
+ if (iface_is_internal(iface, br)) {
+ type = "internal";
+ } else {
+ type = iface->type[0] ? iface->type : "system";
}
- VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
-
- bridge_flush(br);
-
- return iface;
+ return ofproto_port_open_type(br->datapath_type, type);
}
static void
if (iface) {
struct port *port = iface->port;
struct bridge *br = port->bridge;
- bool del_active = port->active_iface == iface->port_ifidx;
- struct iface *del;
-
- if (iface->port->lacp) {
- lacp_slave_unregister(iface->port->lacp, iface);
- }
- if (port->monitor && iface->netdev) {
- netdev_monitor_remove(port->monitor, iface->netdev);
+ if (br->ofproto && iface->ofp_port >= 0) {
+ ofproto_port_unregister(br->ofproto, iface->ofp_port);
}
- shash_find_and_delete_assert(&br->iface_by_name, iface->name);
-
- if (iface->dp_ifidx >= 0) {
- hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
+ if (iface->ofp_port >= 0) {
+ hmap_remove(&br->ifaces, &iface->ofp_port_node);
}
- del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces];
- del->port_ifidx = iface->port_ifidx;
+ list_remove(&iface->port_elem);
+ hmap_remove(&br->iface_by_name, &iface->name_node);
netdev_close(iface->netdev);
- if (del_active) {
- ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
- bond_choose_active_iface(port);
- bond_send_learning_packets(port);
- }
-
- cfm_destroy(iface->cfm);
-
free(iface->name);
free(iface);
-
- bridge_flush(port->bridge);
}
}
static struct iface *
iface_lookup(const struct bridge *br, const char *name)
{
- return shash_find_data(&br->iface_by_name, name);
+ struct iface *iface;
+
+ HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
+ &br->iface_by_name) {
+ if (!strcmp(iface->name, name)) {
+ return iface;
+ }
+ }
+
+ return NULL;
}
static struct iface *
{
const struct bridge *br;
- LIST_FOR_EACH (br, node, &all_bridges) {
+ HMAP_FOR_EACH (br, node, &all_bridges) {
struct iface *iface = iface_lookup(br, name);
if (iface) {
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_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
+iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
{
struct iface *iface;
- HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
- hash_int(dp_ifidx, 0), &br->ifaces) {
- if (iface->dp_ifidx == dp_ifidx) {
+ HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
+ hash_int(ofp_port, 0), &br->ifaces) {
+ if (iface->ofp_port == ofp_port) {
return iface;
}
}
{
uint8_t ea[ETH_ADDR_LEN];
- if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
- if (eth_addr_is_multicast(ea)) {
+ if (!strcmp(iface->type, "internal")
+ && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
+ if (iface->ofp_port == OFPP_LOCAL) {
+ VLOG_ERR("interface %s: ignoring mac in Interface record "
+ "(use Bridge record to set local port's mac)",
+ iface->name);
+ } else if (eth_addr_is_multicast(ea)) {
VLOG_ERR("interface %s: cannot set MAC to multicast address",
iface->name);
- } else if (iface->dp_ifidx == ODPP_LOCAL) {
- VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead",
- iface->name, iface->name);
} else {
int error = netdev_set_etheraddr(iface->netdev, ea);
if (error) {
static void
iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
{
- if (if_cfg) {
+ if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
}
}
-/* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
- *
- * The value strings in '*shash' are taken directly from values[], not copied,
- * so the caller should not modify or free them. */
-static void
-shash_from_ovs_idl_map(char **keys, char **values, size_t n,
- struct shash *shash)
-{
- size_t i;
-
- shash_init(shash);
- for (i = 0; i < n; i++) {
- shash_add(shash, keys[i], values[i]);
- }
-}
-
-/* Creates 'keys' and 'values' arrays from 'shash'.
+/* Clears all of the fields in 'if_cfg' that indicate interface status, and
+ * sets the "ofport" field to -1.
*
- * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
- * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
- * are populated with with strings taken directly from 'shash' and thus have
- * the same ownership of the key-value pairs in shash.
- */
+ * This is appropriate when 'if_cfg''s interface cannot be created or is
+ * otherwise invalid. */
static void
-shash_to_ovs_idl_map(struct shash *shash,
- char ***keys, char ***values, size_t *n)
+iface_clear_db_record(const struct ovsrec_interface *if_cfg)
{
- size_t i, count;
- char **k, **v;
- struct shash_node *sn;
-
- count = shash_count(shash);
-
- k = xmalloc(count * sizeof *k);
- v = xmalloc(count * sizeof *v);
-
- i = 0;
- SHASH_FOR_EACH(sn, shash) {
- k[i] = sn->name;
- v[i] = sn->data;
- i++;
+ if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
+ ovsrec_interface_set_status(if_cfg, NULL);
+ ovsrec_interface_set_admin_state(if_cfg, NULL);
+ ovsrec_interface_set_duplex(if_cfg, NULL);
+ ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
+ ovsrec_interface_set_link_state(if_cfg, NULL);
+ ovsrec_interface_set_mac_in_use(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);
}
-
- *n = count;
- *keys = k;
- *values = v;
}
struct iface_delete_queues_cbdata {
static void
iface_delete_queues(unsigned int queue_id,
- const struct shash *details OVS_UNUSED, void *cbdata_)
+ const struct smap *details OVS_UNUSED, void *cbdata_)
{
struct iface_delete_queues_cbdata *cbdata = cbdata_;
}
static void
-iface_update_carrier(struct iface *iface, bool carrier)
+iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
{
- if (carrier == iface->up) {
- return;
- }
+ struct ofpbuf queues_buf;
- iface->up = carrier;
- if (iface->port->lacp) {
- lacp_slave_carrier_changed(iface->port->lacp, iface);
- }
-}
+ ofpbuf_init(&queues_buf, 0);
-static void
-iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
-{
- if (!qos || qos->type[0] == '\0') {
+ if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
netdev_set_qos(iface->netdev, NULL, NULL);
} else {
struct iface_delete_queues_cbdata cbdata;
- struct shash details;
+ bool queue_zero;
size_t i;
/* Configure top-level Qos for 'iface'. */
- shash_from_ovs_idl_map(qos->key_other_config, qos->value_other_config,
- qos->n_other_config, &details);
- netdev_set_qos(iface->netdev, qos->type, &details);
- shash_destroy(&details);
+ netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
/* Deconfigure queues that were deleted. */
cbdata.netdev = iface->netdev;
netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
/* Configure queues for 'iface'. */
+ queue_zero = false;
for (i = 0; i < qos->n_queues; i++) {
const struct ovsrec_queue *queue = qos->value_queues[i];
unsigned int queue_id = qos->key_queues[i];
- shash_from_ovs_idl_map(queue->key_other_config,
- queue->value_other_config,
- queue->n_other_config, &details);
- netdev_set_queue(iface->netdev, queue_id, &details);
- shash_destroy(&details);
+ if (queue_id == 0) {
+ 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];
+ }
+
+ netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
}
+ if (!queue_zero) {
+ struct smap details;
+
+ smap_init(&details);
+ netdev_set_queue(iface->netdev, 0, &details);
+ smap_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
-iface_update_cfm(struct iface *iface)
+iface_configure_cfm(struct iface *iface)
{
- size_t i;
- struct cfm *cfm;
- uint16_t *remote_mps;
- struct ovsrec_monitor *mon;
- uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN];
-
- mon = iface->cfg->monitor;
+ const struct ovsrec_interface *cfg = iface->cfg;
+ const char *opstate_str;
+ const char *cfm_ccm_vlan;
+ struct cfm_settings s;
+ struct smap netdev_args;
- if (!mon) {
- cfm_destroy(iface->cfm);
- iface->cfm = NULL;
+ if (!cfg->n_cfm_mpid) {
+ ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
return;
}
- if (netdev_get_etheraddr(iface->netdev, ea)) {
- VLOG_WARN("interface %s: Failed to get ethernet address. "
- "Skipping Monitor.", iface->name);
- return;
+ s.check_tnl_key = false;
+ smap_init(&netdev_args);
+ if (!netdev_get_config(iface->netdev, &netdev_args)) {
+ const char *key = smap_get(&netdev_args, "key");
+ const char *in_key = smap_get(&netdev_args, "in_key");
+
+ s.check_tnl_key = (key && !strcmp(key, "flow"))
+ || (in_key && !strcmp(in_key, "flow"));
}
+ smap_destroy(&netdev_args);
- if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
- VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
- return;
+ s.mpid = *cfg->cfm_mpid;
+ s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
+ cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
+ s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
+
+ if (s.interval <= 0) {
+ s.interval = 1000;
}
- if (!iface->cfm) {
- iface->cfm = cfm_create();
+ if (!cfm_ccm_vlan) {
+ s.ccm_vlan = 0;
+ } else 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;
+ }
}
- cfm = iface->cfm;
- cfm->mpid = mon->mpid;
- cfm->interval = mon->interval ? *mon->interval : 1000;
+ s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
+ false);
- memcpy(cfm->eth_src, ea, sizeof cfm->eth_src);
- memcpy(cfm->maid, maid, sizeof cfm->maid);
+ opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
+ s.opup = !opstate_str || !strcasecmp("up", opstate_str);
- remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
- for(i = 0; i < mon->n_remote_mps; i++) {
- remote_mps[i] = mon->remote_mps[i]->mpid;
- }
- cfm_update_remote_mps(cfm, remote_mps, mon->n_remote_mps);
- free(remote_mps);
+ ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
+}
- if (!cfm_configure(iface->cfm)) {
- cfm_destroy(iface->cfm);
- iface->cfm = NULL;
- }
+/* Returns true if 'iface' is synthetic, that is, if we constructed it locally
+ * instead of obtaining it from the database. */
+static bool
+iface_is_synthetic(const struct iface *iface)
+{
+ return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
+}
+
+static int64_t
+iface_pick_ofport(const struct ovsrec_interface *cfg)
+{
+ int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
+ return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
}
+
\f
/* Port mirroring. */
static struct mirror *
mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
{
- int i;
+ struct mirror *m;
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m && uuid_equals(uuid, &m->uuid)) {
+ HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
+ if (uuid_equals(uuid, &m->uuid)) {
return m;
}
}
}
static void
-mirror_reconfigure(struct bridge *br)
+bridge_configure_mirrors(struct bridge *br)
{
- unsigned long *rspan_vlans;
- int i;
+ const struct ovsdb_datum *mc;
+ unsigned long *flood_vlans;
+ struct mirror *m, *next;
+ size_t i;
/* Get rid of deleted mirrors. */
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m) {
- const struct ovsdb_datum *mc;
- union ovsdb_atom atom;
-
- mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
- atom.uuid = br->mirrors[i]->uuid;
- if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
- mirror_destroy(m);
- }
+ mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
+ HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
+ union ovsdb_atom atom;
+
+ atom.uuid = m->uuid;
+ if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
+ mirror_destroy(m);
}
}
/* Add new mirrors and reconfigure existing ones. */
for (i = 0; i < br->cfg->n_mirrors; i++) {
- struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
+ const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
- if (m) {
- mirror_reconfigure_one(m, cfg);
- } else {
- mirror_create(br, cfg);
+ if (!m) {
+ m = mirror_create(br, cfg);
}
- }
-
- /* Update port reserved status. */
- for (i = 0; i < br->n_ports; i++) {
- br->ports[i]->is_mirror_output_port = false;
- }
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m && m->out_port) {
- m->out_port->is_mirror_output_port = true;
+ m->cfg = cfg;
+ if (!mirror_configure(m)) {
+ mirror_destroy(m);
}
}
/* Update flooded vlans (for RSPAN). */
- rspan_vlans = NULL;
- if (br->cfg->n_flood_vlans) {
- rspan_vlans = bitmap_allocate(4096);
-
- for (i = 0; i < br->cfg->n_flood_vlans; i++) {
- int64_t vlan = br->cfg->flood_vlans[i];
- if (vlan >= 0 && vlan < 4096) {
- bitmap_set1(rspan_vlans, vlan);
- VLOG_INFO("bridge %s: disabling learning on vlan %"PRId64,
- br->name, vlan);
- } else {
- VLOG_ERR("bridge %s: invalid value %"PRId64 "for flood VLAN",
- br->name, vlan);
- }
- }
- }
- if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
- bridge_flush(br);
- }
+ flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
+ br->cfg->n_flood_vlans);
+ ofproto_set_flood_vlans(br->ofproto, flood_vlans);
+ bitmap_free(flood_vlans);
}
-static void
-mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
+static struct mirror *
+mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
{
struct mirror *m;
- size_t i;
-
- for (i = 0; ; i++) {
- if (i >= MAX_MIRRORS) {
- VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
- "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
- return;
- }
- if (!br->mirrors[i]) {
- break;
- }
- }
-
- VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
- bridge_flush(br);
- br->mirrors[i] = m = xzalloc(sizeof *m);
+ m = xzalloc(sizeof *m);
+ m->uuid = cfg->header_.uuid;
+ hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
m->bridge = br;
- m->idx = i;
m->name = xstrdup(cfg->name);
- shash_init(&m->src_ports);
- shash_init(&m->dst_ports);
- m->vlans = NULL;
- m->n_vlans = 0;
- m->out_vlan = -1;
- m->out_port = NULL;
-
- mirror_reconfigure_one(m, cfg);
+
+ return m;
}
static void
{
if (m) {
struct bridge *br = m->bridge;
- size_t i;
- for (i = 0; i < br->n_ports; i++) {
- br->ports[i]->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
- br->ports[i]->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
+ if (br->ofproto) {
+ ofproto_mirror_unregister(br->ofproto, m);
}
- shash_destroy(&m->src_ports);
- shash_destroy(&m->dst_ports);
- free(m->vlans);
-
- m->bridge->mirrors[m->idx] = NULL;
+ hmap_remove(&br->mirrors, &m->hmap_node);
free(m->name);
free(m);
-
- bridge_flush(br);
}
}
static void
-mirror_collect_ports(struct mirror *m, struct ovsrec_port **ports, int n_ports,
- struct shash *names)
+mirror_collect_ports(struct mirror *m,
+ struct ovsrec_port **in_ports, int n_in_ports,
+ void ***out_portsp, size_t *n_out_portsp)
{
+ void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
+ size_t n_out_ports = 0;
size_t i;
- for (i = 0; i < n_ports; i++) {
- const char *name = ports[i]->name;
- if (port_lookup(m->bridge, name)) {
- shash_add_once(names, name, NULL);
+ for (i = 0; i < n_in_ports; i++) {
+ const char *name = in_ports[i]->name;
+ struct port *port = port_lookup(m->bridge, name);
+ if (port) {
+ out_ports[n_out_ports++] = port;
} else {
VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
"port %s", m->bridge->name, m->name, name);
}
}
-}
-
-static size_t
-mirror_collect_vlans(struct mirror *m, const struct ovsrec_mirror *cfg,
- int **vlans)
-{
- size_t n_vlans;
- size_t i;
-
- *vlans = xmalloc(sizeof **vlans * cfg->n_select_vlan);
- n_vlans = 0;
- for (i = 0; i < cfg->n_select_vlan; i++) {
- int64_t vlan = cfg->select_vlan[i];
- if (vlan < 0 || vlan > 4095) {
- VLOG_WARN("bridge %s: mirror %s selects invalid VLAN %"PRId64,
- m->bridge->name, m->name, vlan);
- } else {
- (*vlans)[n_vlans++] = vlan;
- }
- }
- return n_vlans;
-}
-
-static bool
-vlan_is_mirrored(const struct mirror *m, int vlan)
-{
- size_t i;
-
- for (i = 0; i < m->n_vlans; i++) {
- if (m->vlans[i] == vlan) {
- return true;
- }
- }
- return false;
+ *out_portsp = out_ports;
+ *n_out_portsp = n_out_ports;
}
static bool
-port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
+mirror_configure(struct mirror *m)
{
- size_t i;
-
- for (i = 0; i < m->n_vlans; i++) {
- if (port_trunks_vlan(p, m->vlans[i])) {
- return true;
- }
- }
- return false;
-}
-
-static void
-mirror_reconfigure_one(struct mirror *m, struct ovsrec_mirror *cfg)
-{
- struct shash src_ports, dst_ports;
- mirror_mask_t mirror_bit;
- struct port *out_port;
- int out_vlan;
- size_t n_vlans;
- int *vlans;
- size_t i;
+ const struct ovsrec_mirror *cfg = m->cfg;
+ struct ofproto_mirror_settings s;
/* Set name. */
if (strcmp(cfg->name, m->name)) {
free(m->name);
m->name = xstrdup(cfg->name);
}
+ s.name = m->name;
- /* Get output port. */
+ /* Get output port or VLAN. */
if (cfg->output_port) {
- out_port = port_lookup(m->bridge, cfg->output_port->name);
- if (!out_port) {
+ s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
+ if (!s.out_bundle) {
VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
m->bridge->name, m->name);
- mirror_destroy(m);
- return;
+ return false;
}
- out_vlan = -1;
+ s.out_vlan = UINT16_MAX;
if (cfg->output_vlan) {
VLOG_ERR("bridge %s: mirror %s specifies both output port and "
m->bridge->name, m->name);
}
} else if (cfg->output_vlan) {
- out_port = NULL;
- out_vlan = *cfg->output_vlan;
+ /* The database should prevent invalid VLAN values. */
+ s.out_bundle = NULL;
+ s.out_vlan = *cfg->output_vlan;
} else {
VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
m->bridge->name, m->name);
- mirror_destroy(m);
- return;
+ return false;
}
- shash_init(&src_ports);
- shash_init(&dst_ports);
+ /* Get port selection. */
if (cfg->select_all) {
- for (i = 0; i < m->bridge->n_ports; i++) {
- const char *name = m->bridge->ports[i]->name;
- shash_add_once(&src_ports, name, NULL);
- shash_add_once(&dst_ports, name, NULL);
+ size_t n_ports = hmap_count(&m->bridge->ports);
+ void **ports = xmalloc(n_ports * sizeof *ports);
+ struct port *port;
+ size_t i;
+
+ i = 0;
+ HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
+ ports[i++] = port;
}
- vlans = NULL;
- n_vlans = 0;
+
+ s.srcs = ports;
+ s.n_srcs = n_ports;
+
+ s.dsts = ports;
+ s.n_dsts = n_ports;
} else {
- /* Get ports, and drop duplicates and ports that don't exist. */
+ /* Get ports, dropping ports that don't exist.
+ * The IDL ensures that there are no duplicates. */
mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
- &src_ports);
+ &s.srcs, &s.n_srcs);
mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
- &dst_ports);
-
- /* Get all the vlans, and drop duplicate and invalid vlans. */
- n_vlans = mirror_collect_vlans(m, cfg, &vlans);
- }
-
- /* Update mirror data. */
- if (!shash_equal_keys(&m->src_ports, &src_ports)
- || !shash_equal_keys(&m->dst_ports, &dst_ports)
- || m->n_vlans != n_vlans
- || memcmp(m->vlans, vlans, sizeof *vlans * n_vlans)
- || m->out_port != out_port
- || m->out_vlan != out_vlan) {
- bridge_flush(m->bridge);
- }
- shash_swap(&m->src_ports, &src_ports);
- shash_swap(&m->dst_ports, &dst_ports);
- free(m->vlans);
- m->vlans = vlans;
- m->n_vlans = n_vlans;
- m->out_port = out_port;
- m->out_vlan = out_vlan;
-
- /* Update ports. */
- mirror_bit = MIRROR_MASK_C(1) << m->idx;
- for (i = 0; i < m->bridge->n_ports; i++) {
- struct port *port = m->bridge->ports[i];
-
- if (shash_find(&m->src_ports, port->name)
- || (m->n_vlans
- && (!port->vlan
- ? port_trunks_any_mirrored_vlan(m, port)
- : vlan_is_mirrored(m, port->vlan)))) {
- port->src_mirrors |= mirror_bit;
- } else {
- port->src_mirrors &= ~mirror_bit;
+ &s.dsts, &s.n_dsts);
+ }
+
+ /* Get VLAN selection. */
+ s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
+
+ /* Configure. */
+ ofproto_mirror_register(m->bridge->ofproto, m, &s);
+
+ /* Clean up. */
+ if (s.srcs != s.dsts) {
+ free(s.dsts);
+ }
+ free(s.srcs);
+ free(s.src_vlans);
+
+ return true;
+}
+\f
+/* 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 struct ovsrec_port **recs;
+static size_t n_recs, allocated_recs;
+
+/* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
+ * splinters are reconfigured. */
+static void
+register_rec(struct ovsrec_port *rec)
+{
+ if (n_recs >= allocated_recs) {
+ recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
+ }
+ recs[n_recs++] = rec;
+}
+
+/* Frees all of the ports registered with register_reg(). */
+static void
+free_registered_recs(void)
+{
+ size_t i;
+
+ for (i = 0; i < n_recs; i++) {
+ struct ovsrec_port *port = recs[i];
+ size_t j;
+
+ for (j = 0; j < port->n_interfaces; j++) {
+ struct ovsrec_interface *iface = port->interfaces[j];
+ free(iface->name);
+ free(iface);
}
- if (shash_find(&m->dst_ports, port->name)) {
- port->dst_mirrors |= mirror_bit;
- } else {
- port->dst_mirrors &= ~mirror_bit;
+ smap_destroy(&port->other_config);
+ free(port->interfaces);
+ free(port->name);
+ free(port->tag);
+ free(port);
+ }
+ n_recs = 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)
+{
+ return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
+ false);
+}
+
+/* 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_recs();
+
+ 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);
+ }
}
}
- /* Clean up. */
- shash_destroy(&src_ports);
- shash_destroy(&dst_ports);
+ 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 = smap_get(&port->cfg->other_config, "realdev");
+ 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 = xmalloc(sizeof *iface);
+ ovsrec_interface_init(iface);
+ iface->name = xstrdup(vlan_dev_name);
+ iface->type = "system";
+
+ port = xmalloc(sizeof *port);
+ ovsrec_port_init(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;
+
+ smap_add(&port->other_config, "realdev", real_dev_name);
+
+ register_rec(port);
+ 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);
}