X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=9868ef5b627e92cfd0f2b786f8eaa0395e37cb88;hb=8628b0b71d1cf5b3bd7425e50f0aa1435aa38804;hp=cfd4dcf75b37d530f9354ad224855d0acd2c6a72;hpb=064af42167bf4fc9aaea2702d80ce08074b889c0;p=openvswitch
diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c
index cfd4dcf7..9868ef5b 100644
--- a/vswitchd/bridge.c
+++ b/vswitchd/bridge.c
@@ -1,2319 +1,2045 @@
-/* Copyright (c) 2008, 2009 Nicira Networks
- *
- * This program is free software: you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation, either version 3 of the License, or
- * (at your option) any later version.
+/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
*
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
*
- * You should have received a copy of the GNU General Public License
- * along with this program. If not, see .
- *
- * In addition, as a special exception, Nicira Networks gives permission
- * to link the code of its release of vswitchd with the OpenSSL project's
- * "OpenSSL" library (or with modified versions of it that use the same
- * license as the "OpenSSL" library), and distribute the linked
- * executables. You must obey the GNU General Public License in all
- * respects for all of the code used other than "OpenSSL". If you modify
- * this file, you may extend this exception to your version of the file,
- * but you are not obligated to do so. If you do not wish to do so,
- * delete this exception statement from your version.
+ * http://www.apache.org/licenses/LICENSE-2.0
*
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
*/
#include
#include "bridge.h"
#include
#include
-#include
-#include
#include
-#include
-#include
-#include
#include
-#include
-#include
-#include
-#include
#include "bitmap.h"
-#include "cfg.h"
+#include "bond.h"
+#include "cfm.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 "jsonrpc.h"
+#include "lacp.h"
#include "list.h"
-#include "mac-learning.h"
#include "netdev.h"
-#include "odp-util.h"
#include "ofp-print.h"
#include "ofpbuf.h"
+#include "ofproto/ofproto.h"
#include "poll-loop.h"
-#include "port-array.h"
-#include "proc-net-compat.h"
-#include "process.h"
-#include "secchan/ofproto.h"
+#include "sha1.h"
+#include "shash.h"
#include "socket-util.h"
-#include "stp.h"
-#include "svec.h"
+#include "stream-ssl.h"
+#include "sset.h"
+#include "system-stats.h"
#include "timeval.h"
#include "util.h"
-#include "vconn.h"
-#include "vconn-ssl.h"
+#include "unixctl.h"
+#include "vswitchd/vswitch-idl.h"
#include "xenserver.h"
-#include "xtoxll.h"
-
-#define THIS_MODULE VLM_bridge
#include "vlog.h"
+#include "sflow_api.h"
+#include "vlan-bitmap.h"
-struct dst {
- uint16_t vlan;
- uint16_t dp_ifidx;
-};
+VLOG_DEFINE_THIS_MODULE(bridge);
-extern uint64_t mgmt_id;
+COVERAGE_DEFINE(bridge_reconfigure);
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. */
- int dp_ifidx; /* Index within kernel datapath. */
-
- uint8_t mac[ETH_ADDR_LEN]; /* Ethernet address (all zeros if unknowns). */
-
tag_type tag; /* Tag associated with this interface. */
- bool enabled; /* May be chosen for flows? */
- long long delay_expires; /* Time after which 'enabled' may change. */
-};
-#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. */
+ /* These members are valid only after bridge_reconfigure() causes them to
+ * be initialized. */
+ 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. */
+ const char *type; /* Usually same as cfg->type. */
+ const struct ovsrec_interface *cfg;
};
-#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;
-
- /* Selection criteria. */
- struct svec src_ports;
- struct svec dst_ports;
- int *vlans;
- size_t n_vlans;
-
- /* Output. */
- struct port *out_port;
- int out_vlan;
};
-#define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
struct port {
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. */
+ struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
char *name;
+ 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. */
- struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */
- 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. */
-
- /* 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? */
-
- /* Spanning tree info. */
- enum stp_state stp_state; /* Always STP_FORWARDING if STP not in use. */
- tag_type stp_state_tag; /* Tag for STP state change. */
+ struct list ifaces; /* List of "struct iface"s. */
};
-#define DP_MAX_PORTS 255
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, or null not to learn. */
- bool sent_config_request; /* Successfully sent config request? */
+ char *type; /* Datapath type. */
+ uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
-
- /* Support for remote controllers. */
- char *controller; /* NULL if there is no remote controller;
- * "discover" to do controller discovery;
- * otherwise a vconn name. */
+ const struct ovsrec_bridge *cfg;
/* OpenFlow switch processing. */
struct ofproto *ofproto; /* OpenFlow switch. */
- /* Kernel datapath information. */
- struct dpif dpif; /* Kernel datapath. */
- struct port_array ifaces; /* Indexed by kernel datapath port number. */
-
/* Bridge ports. */
- struct port **ports;
- size_t n_ports, allocated_ports;
-
- /* Bonding. */
- bool has_bonded_ports;
- long long int bond_next_rebalance;
-
- /* Flow tracking. */
- bool flush;
-
- /* Flow statistics gathering. */
- time_t next_stats_request;
+ 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. */
/* Port mirroring. */
- struct mirror *mirrors[MAX_MIRRORS];
+ struct hmap mirrors; /* "struct mirror" indexed by UUID. */
- /* Spanning tree. */
- struct stp *stp;
- long long int stp_last_tick;
+ /* 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;
-/* Maximum number of datapaths. */
-enum { DP_MAX = 256 };
+/* 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;
-static struct bridge *bridge_create(const char *name);
+/* Stores the time after which rate limited statistics may be written to the
+ * database. Only updated when changes to the database require rate limiting.
+ */
+#define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
+static long long int db_limiter = LLONG_MIN;
+
+static void add_del_bridges(const struct ovsrec_open_vswitch *);
+static void bridge_del_ofprotos(void);
+static bool bridge_add_ofprotos(struct bridge *);
+static void bridge_create(const struct ovsrec_bridge *);
static void bridge_destroy(struct bridge *);
static struct bridge *bridge_lookup(const char *name);
-static int bridge_run_one(struct bridge *);
-static void bridge_reconfigure_one(struct bridge *);
-static void bridge_reconfigure_controller(struct bridge *);
-static void bridge_get_all_ifaces(const struct bridge *, struct svec *ifaces);
-static void bridge_fetch_dp_ifaces(struct bridge *);
-static void bridge_flush(struct bridge *);
+static unixctl_cb_func bridge_unixctl_dump_flows;
+static unixctl_cb_func bridge_unixctl_reconnect;
+static size_t bridge_get_controllers(const struct bridge *br,
+ struct ovsrec_controller ***controllersp);
+static void bridge_add_del_ports(struct bridge *);
+static void bridge_add_ofproto_ports(struct bridge *);
+static void bridge_del_ofproto_ports(struct bridge *);
+static void bridge_refresh_ofp_port(struct bridge *);
+static void bridge_configure_datapath_id(struct bridge *);
+static void bridge_configure_netflow(struct bridge *);
+static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
+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],
- const char **devname);
+ struct iface **hw_addr_iface);
static uint64_t bridge_pick_datapath_id(struct bridge *,
const uint8_t bridge_ea[ETH_ADDR_LEN],
- const char *devname);
+ struct iface *hw_addr_iface);
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 void bond_run(struct bridge *);
-static void bond_wait(struct bridge *);
-static void bond_rebalance_port(struct port *);
+static unixctl_cb_func qos_unixctl_show;
-static void port_create(struct bridge *, const char *name);
-static void port_reconfigure(struct port *);
+static struct port *port_create(struct bridge *, const struct ovsrec_port *);
+static void port_add_ifaces(struct port *);
+static void port_del_ifaces(struct port *);
static void port_destroy(struct port *);
static struct port *port_lookup(const struct bridge *, const char *name);
-static struct port *port_from_dp_ifidx(const struct bridge *,
- uint16_t dp_ifidx);
-static void port_update_bond_compat(struct port *);
-static void port_update_vlan_compat(struct port *);
-
-static void mirror_create(struct bridge *, const char *name);
+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 *,
+ uint32_t *bond_stable_ids);
+
+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 *);
-static bool vlan_is_mirrored(const struct mirror *, int vlan);
-
-static void brstp_reconfigure(struct bridge *);
-static void brstp_adjust_timers(struct bridge *);
-static void brstp_run(struct bridge *);
-static void brstp_wait(struct bridge *);
+static bool mirror_configure(struct mirror *, const struct ovsrec_mirror *);
-static void iface_create(struct port *, const char *name);
+static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
+static struct iface *iface_create(struct port *port,
+ const struct ovsrec_interface *if_cfg);
static void iface_destroy(struct iface *);
static struct iface *iface_lookup(const struct bridge *, const char *name);
-static struct iface *iface_from_dp_ifidx(const struct bridge *,
- uint16_t dp_ifidx);
-
-/* Hooks into ofproto processing. */
-static struct ofhooks bridge_ofhooks;
+static struct iface *iface_find(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_configure_qos(struct iface *, const struct ovsrec_qos *);
+static void iface_configure_cfm(struct iface *);
+static bool iface_refresh_cfm_stats(struct iface *iface);
+static bool iface_get_carrier(const struct iface *);
+static bool iface_is_synthetic(const struct iface *);
+
+static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
+ struct shash *);
+static void shash_to_ovs_idl_map(struct shash *,
+ char ***keys, char ***values, size_t *n);
/* Public functions. */
-/* Adds the name of each interface used by a bridge, including local and
- * internal ports, to 'svec'. */
+/* Initializes the bridge module, configuring it to obtain its configuration
+ * from an OVSDB server accessed over 'remote', which should be a string in a
+ * form acceptable to ovsdb_idl_create(). */
void
-bridge_get_ifaces(struct svec *svec)
+bridge_init(const char *remote)
{
- struct bridge *br, *next;
- size_t i, j;
+ /* Create connection to database. */
+ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
- LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ 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);
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (iface->dp_ifidx < 0) {
- VLOG_ERR("%s interface not in dp%u, ignoring",
- iface->name, dpif_id(&br->dpif));
- } else {
- if (iface->dp_ifidx != ODPP_LOCAL) {
- svec_add(svec, iface->name);
- }
- }
- }
- }
- }
-}
+ ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
+ ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
-/* The caller must already have called cfg_read(). */
-void
-bridge_init(void)
-{
- int retval;
- int i;
+ ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
+ ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
- for (i = 0; i < DP_MAX; i++) {
- struct dpif dpif;
- char devname[16];
-
- sprintf(devname, "dp%d", i);
- retval = dpif_open(devname, &dpif);
- if (!retval) {
- char dpif_name[IF_NAMESIZE];
- if (dpif_get_name(&dpif, dpif_name, sizeof dpif_name)
- || !cfg_has("bridge.%s.port", dpif_name)) {
- dpif_delete(&dpif);
- }
- dpif_close(&dpif);
- } else if (retval != ENODEV) {
- VLOG_ERR("failed to delete datapath dp%d: %s",
- i, strerror(retval));
- }
- }
+ 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_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(idl, &ovsrec_interface_col_external_ids);
+
+ 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);
+
+ ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
+
+ ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
+
+ ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
+
+ ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
- bridge_reconfigure();
+ ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
+
+ 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);
+
+ ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
+
+ /* Register unixctl commands. */
+ 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();
+ cfm_init();
}
-#ifdef HAVE_OPENSSL
-static bool
-config_string_change(const char *key, char **valuep)
+void
+bridge_exit(void)
{
- const char *value = cfg_get_string(0, "%s", key);
- if (value && (!*valuep || strcmp(value, *valuep))) {
- free(*valuep);
- *valuep = xstrdup(value);
- return true;
- } else {
- return false;
+ struct bridge *br, *next_br;
+
+ 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
+ * addresses and ports into '*managersp' and '*n_managersp'. The caller is
+ * responsible for freeing '*managersp' (with free()).
+ *
+ * You may be asking yourself "why does ovs-vswitchd care?", because
+ * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
+ * should not be and in fact is not directly involved in that. But
+ * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
+ * it has to tell in-band control where the managers are to enable that.
+ * (Thus, only managers connected in-band are collected.)
+ */
static void
-bridge_configure_ssl(void)
+collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
+ struct sockaddr_in **managersp, size_t *n_managersp)
{
- /* XXX SSL should be configurable on a per-bridge basis.
- * XXX should be possible to de-configure SSL. */
- static char *private_key_file;
- static char *certificate_file;
- static char *cacert_file;
+ struct sockaddr_in *managers = NULL;
+ size_t n_managers = 0;
+ struct sset targets;
+ size_t i;
- if (config_string_change("ssl.private-key", &private_key_file)) {
- vconn_ssl_set_private_key_file(private_key_file);
- }
+ /* 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. */
+ sset_init(&targets);
+ for (i = 0; i < ovs_cfg->n_manager_options; i++) {
+ struct ovsrec_manager *m = ovs_cfg->manager_options[i];
- if (config_string_change("ssl.certificate", &certificate_file)) {
- vconn_ssl_set_certificate_file(certificate_file);
+ if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
+ sset_find_and_delete(&targets, m->target);
+ } else {
+ sset_add(&targets, m->target);
+ }
}
- if (config_string_change("ssl.ca-cert", &cacert_file)) {
- vconn_ssl_set_ca_cert_file(cacert_file,
- cfg_get_bool(0, "ssl.bootstrap-ca-cert"));
+ /* Now extract the targets' IP addresses. */
+ if (!sset_is_empty(&targets)) {
+ const char *target;
+
+ 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))) {
+ n_managers++;
+ }
+ }
}
+ sset_destroy(&targets);
+
+ *managersp = managers;
+ *n_managersp = n_managers;
}
-#endif
-void
-bridge_reconfigure(void)
+static void
+bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
{
- struct svec old_br, new_br, raw_new_br;
+ struct sockaddr_in *managers;
struct bridge *br, *next;
- size_t i, j;
+ int sflow_bridge_number;
+ size_t n_managers;
COVERAGE_INC(bridge_reconfigure);
- /* Collect old bridges. */
- svec_init(&old_br);
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- svec_add(&old_br, br->name);
- }
-
- /* Collect new bridges. */
- svec_init(&raw_new_br);
- cfg_get_subsections(&raw_new_br, "bridge");
- svec_init(&new_br);
- for (i = 0; i < raw_new_br.n; i++) {
- const char *name = raw_new_br.names[i];
- if ((!strncmp(name, "dp", 2) && isdigit(name[2])) ||
- (!strncmp(name, "nl:", 3) && isdigit(name[3]))) {
- VLOG_ERR("%s is not a valid bridge name (bridges may not be "
- "named \"dp\" or \"nl:\" followed by a digit)", name);
- } else {
- svec_add(&new_br, name);
+ /* Create and destroy "struct bridge"s, "struct port"s, and "struct
+ * iface"s according to 'ovs_cfg', with only very minimal configuration
+ * otherwise.
+ *
+ * This is purely an update to bridge data structures. Nothing is pushed
+ * down to ofproto or lower layers. */
+ add_del_bridges(ovs_cfg);
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ bridge_add_del_ports(br);
+ }
+
+ /* Delete all datapaths and datapath ports that are no longer configured.
+ *
+ * The kernel will reject any attempt to add a given port to a datapath if
+ * that port already belongs to a different datapath, so we must do all
+ * port deletions before any port additions. A datapath always has a
+ * "local port" so we must delete not-configured datapaths too. */
+ bridge_del_ofprotos();
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ if (br->ofproto) {
+ bridge_del_ofproto_ports(br);
}
}
- svec_destroy(&raw_new_br);
- /* Get rid of deleted bridges and add new bridges. */
- svec_sort(&old_br);
- svec_sort(&new_br);
- assert(svec_is_unique(&old_br));
- assert(svec_is_unique(&new_br));
- LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) {
- if (!svec_contains(&new_br, br->name)) {
+ /* Create datapaths and datapath ports that are missing.
+ *
+ * After this is done, we have our final set of bridges, ports, and
+ * interfaces. Every "struct bridge" has an ofproto, every "struct port"
+ * has at least one iface, every "struct iface" has a valid ofp_port and
+ * netdev. */
+ HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
+ if (!br->ofproto && !bridge_add_ofprotos(br)) {
bridge_destroy(br);
}
}
- for (i = 0; i < new_br.n; i++) {
- const char *name = new_br.names[i];
- if (!svec_contains(&old_br, name)) {
- bridge_create(name);
- }
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ bridge_refresh_ofp_port(br);
+ bridge_add_ofproto_ports(br);
}
- svec_destroy(&old_br);
- svec_destroy(&new_br);
-#ifdef HAVE_OPENSSL
- /* Configure SSL. */
- bridge_configure_ssl();
-#endif
+ /* 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;
- /* Reconfigure all bridges. */
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- bridge_reconfigure_one(br);
- }
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
- /* Add and delete ports on all datapaths.
- *
- * 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, struct bridge, node, &all_bridges) {
- struct odp_port *dpif_ports;
- size_t n_dpif_ports;
- struct svec want_ifaces;
-
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
- bridge_get_all_ifaces(br, &want_ifaces);
- for (i = 0; i < n_dpif_ports; i++) {
- const struct odp_port *p = &dpif_ports[i];
- if (!svec_contains(&want_ifaces, p->devname)
- && strcmp(p->devname, br->name)) {
- int retval = dpif_port_del(&br->dpif, p->port);
- if (retval) {
- VLOG_ERR("failed to remove %s interface from dp%u: %s",
- p->devname, dpif_id(&br->dpif), strerror(retval));
- }
- }
- }
- svec_destroy(&want_ifaces);
- free(dpif_ports);
- }
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- struct odp_port *dpif_ports;
- size_t n_dpif_ports;
- struct svec cur_ifaces, want_ifaces, add_ifaces;
- int next_port_no;
-
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
- svec_init(&cur_ifaces);
- for (i = 0; i < n_dpif_ports; i++) {
- svec_add(&cur_ifaces, dpif_ports[i].devname);
- }
- free(dpif_ports);
- svec_sort_unique(&cur_ifaces);
- bridge_get_all_ifaces(br, &want_ifaces);
- svec_diff(&want_ifaces, &cur_ifaces, &add_ifaces, NULL, NULL);
-
- next_port_no = 1;
- for (i = 0; i < add_ifaces.n; i++) {
- const char *if_name = add_ifaces.names[i];
- for (;;) {
- int internal = cfg_get_bool(0, "iface.%s.internal", if_name);
- int error = dpif_port_add(&br->dpif, if_name, next_port_no++,
- internal ? ODP_PORT_INTERNAL : 0);
- if (error != EEXIST) {
- if (next_port_no >= 256) {
- VLOG_ERR("ran out of valid port numbers on dp%u",
- dpif_id(&br->dpif));
- goto out;
- }
- if (error) {
- VLOG_ERR("failed to add %s interface to dp%u: %s",
- if_name, dpif_id(&br->dpif), strerror(error));
- }
- break;
- }
- }
- }
- out:
- svec_destroy(&cur_ifaces);
- svec_destroy(&want_ifaces);
- svec_destroy(&add_ifaces);
- }
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- uint8_t ea[8];
- uint64_t dpid;
- struct iface *local_iface = NULL;
- const char *devname;
- uint8_t engine_type = br->dpif.minor;
- uint8_t engine_id = br->dpif.minor;
- bool add_id_to_iface = false;
- struct svec nf_hosts;
-
-
- bridge_fetch_dp_ifaces(br);
- 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 (iface->dp_ifidx < 0) {
- VLOG_ERR("%s interface not in dp%u, dropping",
- iface->name, dpif_id(&br->dpif));
- iface_destroy(iface);
- } else {
- if (iface->dp_ifidx == ODPP_LOCAL) {
- local_iface = iface;
- }
- VLOG_DBG("dp%u has interface %s on port %d",
- dpif_id(&br->dpif), iface->name, iface->dp_ifidx);
- j++;
- }
- }
- if (!port->n_ifaces) {
- VLOG_ERR("%s port has no interfaces, dropping", port->name);
- port_destroy(port);
- continue;
- }
- i++;
- }
+ port_configure(port);
- /* Pick local port hardware address, datapath ID. */
- bridge_pick_local_hw_addr(br, ea, &devname);
- if (local_iface) {
- int error = netdev_nodev_set_etheraddr(local_iface->name, 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));
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_configure_cfm(iface);
+ iface_configure_qos(iface, port->cfg->qos);
+ iface_set_mac(iface);
}
}
+ bridge_configure_mirrors(br);
+ bridge_configure_datapath_id(br);
+ bridge_configure_remotes(br, managers, n_managers);
+ bridge_configure_netflow(br);
+ bridge_configure_sflow(br, &sflow_bridge_number);
+ }
+ free(managers);
- dpid = bridge_pick_datapath_id(br, ea, devname);
- ofproto_set_datapath_id(br->ofproto, dpid);
-
- /* Set NetFlow configuration on this bridge. */
- if (cfg_has("netflow.%s.engine-type", br->name)) {
- engine_type = cfg_get_int(0, "netflow.%s.engine-type",
- br->name);
- }
- if (cfg_has("netflow.%s.engine-id", br->name)) {
- engine_id = cfg_get_int(0, "netflow.%s.engine-id", br->name);
- }
- if (cfg_has("netflow.%s.add-id-to-iface", br->name)) {
- add_id_to_iface = cfg_get_bool(0, "netflow.%s.add-id-to-iface",
- br->name);
- }
- if (add_id_to_iface && 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 (add_id_to_iface && br->n_ports > 0x1ff) {
- VLOG_WARN("bridge %s: netflow port mangling will conflict with "
- "another port when 512 or more ports are used",
- br->name);
- }
- svec_init(&nf_hosts);
- cfg_get_all_keys(&nf_hosts, "netflow.%s.host", br->name);
- if (ofproto_set_netflow(br->ofproto, &nf_hosts, engine_type,
- engine_id, add_id_to_iface)) {
- VLOG_ERR("bridge %s: problem setting netflow collectors",
- br->name);
- }
+ /* ovs-vswitchd has completed initialization, so allow the process that
+ * forked us to exit successfully. */
+ daemonize_complete();
+}
- /* 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_controller(br);
- }
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- port_update_vlan_compat(port);
+/* Iterate over all ofprotos and delete any of them that do not have a
+ * configured bridge or that are the wrong type. */
+static void
+bridge_del_ofprotos(void)
+{
+ struct sset names;
+ struct sset types;
+ const char *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) {
+ struct bridge *br = bridge_lookup(name);
+ if (!br || strcmp(type, br->type)) {
+ ofproto_delete(name, type);
+ }
}
}
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- brstp_reconfigure(br);
+ sset_destroy(&names);
+ sset_destroy(&types);
+}
+
+static bool
+bridge_add_ofprotos(struct bridge *br)
+{
+ int error = ofproto_create(br->name, br->type, &br->ofproto);
+ if (error) {
+ VLOG_ERR("failed to create bridge %s: %s", br->name, strerror(error));
+ return false;
}
+ return true;
}
static void
-bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
- const char **devname)
+port_configure(struct port *port)
{
- uint64_t requested_ea;
- size_t i, j;
- int error;
+ 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;
- *devname = NULL;
+ /* Get name. */
+ s.name = port->name;
- /* Did the user request a particular MAC? */
- requested_ea = cfg_get_mac(0, "bridge.%s.mac", br->name);
- if (requested_ea) {
- eth_addr_from_uint64(requested_ea, 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;
- }
+ /* 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;
}
- /* Otherwise choose the minimum MAC address among all of the interfaces.
- * (Xen uses FE:FF:FF:FF:FF:FF for virtual interfaces so this will get the
- * MAC of the physical interface in such an environment.) */
- memset(ea, 0xff, sizeof ea);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->is_mirror_output_port) {
- continue;
- }
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- uint8_t iface_ea[ETH_ADDR_LEN];
- if (iface->dp_ifidx == ODPP_LOCAL
- || cfg_get_bool(0, "iface.%s.internal", iface->name)) {
- continue;
- }
- error = netdev_nodev_get_etheraddr(iface->name, iface_ea);
- if (!error) {
- if (!eth_addr_is_multicast(iface_ea) &&
- !eth_addr_is_reserved(iface_ea) &&
- !eth_addr_is_zero(iface_ea) &&
- memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0) {
- memcpy(ea, iface_ea, ETH_ADDR_LEN);
- *devname = iface->name;
- }
- } else {
- 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));
+ /* Get VLAN tag. */
+ s.vlan = -1;
+ if (cfg->tag) {
+ if (list_is_short(&port->ifaces)) {
+ if (*cfg->tag >= 0 && *cfg->tag <= 4095) {
+ s.vlan = *cfg->tag;
+ VLOG_DBG("port %s: assigning VLAN tag %d", port->name, s.vlan);
}
+ } 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 (eth_addr_is_multicast(ea) || eth_addr_is_vif(ea)) {
- memcpy(ea, br->default_ea, ETH_ADDR_LEN);
- *devname = 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));
+
+ /* Get VLAN trunks. */
+ s.trunks = NULL;
+ if (s.vlan < 0 && cfg->n_trunks) {
+ s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
+ } else if (s.vlan >= 0 && cfg->n_trunks) {
+ VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
+ port->name);
}
-}
-/* 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
- * a network device, then that network device's name must be passed in as
- * 'devname'; if 'bridge_ea' was derived some other way, then 'devname' 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],
- const char *devname)
-{
- /*
- * 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".
- */
- uint64_t dpid;
+ /* Get LACP settings. */
+ s.lacp = port_configure_lacp(port, &lacp_settings);
+ if (s.lacp) {
+ size_t i = 0;
- dpid = cfg_get_dpid(0, "bridge.%s.datapath-id", br->name);
- if (dpid) {
- return dpid;
+ 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;
}
- if (devname) {
- int vlan;
- if (!netdev_get_vlan_vid(devname, &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.
- */
- }
+ /* Get bond settings. */
+ if (s.n_slaves > 1) {
+ s.bond = &bond_settings;
+ s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
+ port_configure_bond(port, &bond_settings, s.bond_stable_ids);
} 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;
+ s.bond = NULL;
+ s.bond_stable_ids = NULL;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ netdev_set_miimon_interval(iface->netdev, 0);
}
}
- return eth_addr_to_uint64(bridge_ea);
-}
-
-static uint64_t
-dpid_from_hash(const void *data, size_t n)
-{
- uint8_t hash[SHA1HashSize];
+ /* Register. */
+ ofproto_bundle_register(port->bridge->ofproto, port, &s);
- BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
- SHA1Bytes(data, n, hash);
- eth_addr_mark_random(hash);
- return eth_addr_to_uint64(hash);
+ /* Clean up. */
+ free(s.slaves);
+ free(s.trunks);
+ free(s.lacp_slaves);
+ free(s.bond_stable_ids);
}
-int
-bridge_run(void)
+/* Pick local port hardware address and datapath ID for 'br'. */
+static void
+bridge_configure_datapath_id(struct bridge *br)
{
- struct bridge *br, *next;
- int retval;
-
- retval = 0;
- LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) {
- int error = bridge_run_one(br);
+ uint8_t ea[ETH_ADDR_LEN];
+ uint64_t dpid;
+ struct iface *local_iface;
+ struct iface *hw_addr_iface;
+ char *dpid_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: datapath was destroyed externally, "
- "forcing reconfiguration", br->name);
- if (!retval) {
- retval = error;
- }
+ VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
+ "Ethernet address: %s",
+ br->name, strerror(error));
}
}
- return retval;
-}
-
-void
-bridge_wait(void)
-{
- struct bridge *br;
+ memcpy(br->ea, ea, ETH_ADDR_LEN);
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- ofproto_wait(br->ofproto);
- if (br->controller) {
- continue;
- }
+ dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
+ ofproto_set_datapath_id(br->ofproto, dpid);
- if (br->ml) {
- mac_learning_wait(br->ml);
- }
- bond_wait(br);
- brstp_wait(br);
- }
+ dpid_string = xasprintf("%016"PRIx64, dpid);
+ ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
+ free(dpid_string);
}
-/* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
- * configuration changes. */
+/* Set NetFlow configuration on 'br'. */
static void
-bridge_flush(struct bridge *br)
+bridge_configure_netflow(struct bridge *br)
{
- COVERAGE_INC(bridge_flush);
- br->flush = true;
- if (br->ml) {
- mac_learning_flush(br->ml);
- }
-}
-
-/* Bridge reconfiguration functions. */
+ struct ovsrec_netflow *cfg = br->cfg->netflow;
+ struct netflow_options opts;
-static struct bridge *
-bridge_create(const char *name)
-{
- struct bridge *br;
- int error;
+ if (!cfg) {
+ ofproto_set_netflow(br->ofproto, NULL);
+ return;
+ }
- assert(!bridge_lookup(name));
- br = xcalloc(1, sizeof *br);
+ memset(&opts, 0, sizeof opts);
- error = dpif_create(name, &br->dpif);
- if (error == EEXIST) {
- error = dpif_open(name, &br->dpif);
- if (error) {
- VLOG_ERR("datapath %s already exists but cannot be opened: %s",
- name, strerror(error));
- free(br);
- return NULL;
- }
- dpif_flow_flush(&br->dpif);
- } else if (error) {
- VLOG_ERR("failed to create datapath %s: %s", name, strerror(error));
- free(br);
- return NULL;
+ /* 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 (cfg->engine_id) {
+ opts.engine_id = *cfg->engine_id;
}
- error = ofproto_create(name, &bridge_ofhooks, br, &br->ofproto);
- if (error) {
- VLOG_ERR("failed to create switch %s: %s", name, strerror(error));
- dpif_delete(&br->dpif);
- dpif_close(&br->dpif);
- free(br);
- return NULL;
+ /* 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;
}
- br->name = xstrdup(name);
- br->ml = mac_learning_create();
- br->sent_config_request = false;
- eth_addr_random(br->default_ea);
-
- port_array_init(&br->ifaces);
-
- br->flush = false;
- br->bond_next_rebalance = time_msec() + 10000;
-
- list_push_back(&all_bridges, &br->node);
+ /* 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);
+ }
+ 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);
+ }
+ }
- VLOG_INFO("created bridge %s on dp%u", br->name, dpif_id(&br->dpif));
+ /* Collectors. */
+ sset_init(&opts.collectors);
+ sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
- return br;
+ /* 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
-bridge_destroy(struct bridge *br)
+bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
{
- if (br) {
- 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;
- 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 dp%u: %s",
- dpif_id(&br->dpif), strerror(error));
- }
- dpif_close(&br->dpif);
- ofproto_destroy(br->ofproto);
- free(br->controller);
- mac_learning_destroy(br->ml);
- port_array_destroy(&br->ifaces);
- free(br->ports);
- free(br->name);
- free(br);
+ if (!cfg) {
+ ofproto_set_sflow(br->ofproto, NULL);
+ return;
}
-}
-static struct bridge *
-bridge_lookup(const char *name)
-{
- struct bridge *br;
+ memset(&oso, 0, sizeof oso);
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- if (!strcmp(br->name, name)) {
- return br;
+ sset_init(&oso.targets);
+ sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
+
+ oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
+ if (cfg->sampling) {
+ oso.sampling_rate = *cfg->sampling;
+ }
+
+ oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
+ if (cfg->polling) {
+ oso.polling_interval = *cfg->polling;
+ }
+
+ 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;
+
+ 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;
}
}
- return NULL;
+ ofproto_set_sflow(br->ofproto, &oso);
+
+ sset_destroy(&oso.targets);
}
-bool
-bridge_exists(const char *name)
+static bool
+bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
{
- return bridge_lookup(name) ? true : false;
+ const struct port *port = port_lookup(br, name);
+ return port && port_is_bond_fake_iface(port);
}
-uint64_t
-bridge_get_datapathid(const char *name)
+static bool
+port_is_bond_fake_iface(const struct port *port)
{
- struct bridge *br = bridge_lookup(name);
- return br ? ofproto_get_datapath_id(br->ofproto) : 0;
+ return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
}
-static int
-bridge_run_one(struct bridge *br)
+static void
+add_del_bridges(const struct ovsrec_open_vswitch *cfg)
{
- int error;
+ struct bridge *br, *next;
+ struct shash new_br;
+ size_t i;
- error = ofproto_run1(br->ofproto);
- if (error) {
- return error;
+ /* Collect new bridges' names and types. */
+ shash_init(&new_br);
+ for (i = 0; i < cfg->n_bridges; i++) {
+ const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
+ if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
+ VLOG_WARN("bridge %s specified twice", br_cfg->name);
+ }
}
- if (br->ml) {
- mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
+ /* 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);
+ }
}
- bond_run(br);
- brstp_run(br);
- error = ofproto_run2(br->ofproto, br->flush);
- br->flush = false;
+ /* 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);
+ }
+ }
- return error;
+ shash_destroy(&new_br);
}
-static const char *
-bridge_get_controller(const struct bridge *br)
+/* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
+ * iface".
+ *
+ * The kernel will reject any attempt to add a given port to a datapath if that
+ * port already belongs to a different datapath, so we must do all port
+ * deletions before any port additions. */
+static void
+bridge_del_ofproto_ports(struct bridge *br)
{
- const char *controller;
+ struct ofproto_port_dump dump;
+ struct ofproto_port ofproto_port;
- controller = cfg_get_string(0, "bridge.%s.controller", br->name);
- if (!controller) {
- controller = cfg_get_string(0, "mgmt.controller");
+ OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
+ const char *name = ofproto_port.name;
+ struct iface *iface;
+ const char *type;
+ int error;
+
+ /* Ignore the local port. We can't change it anyhow. */
+ if (!strcmp(name, br->name)) {
+ continue;
+ }
+
+ /* Get the type that 'ofproto_port' should have (ordinarily the
+ * type of its corresponding iface) or NULL if it should be
+ * deleted. */
+ iface = iface_lookup(br, name);
+ type = (iface ? iface->type
+ : bridge_has_bond_fake_iface(br, name) ? "internal"
+ : NULL);
+
+ /* If it's the wrong type then delete the ofproto port. */
+ if (type
+ && !strcmp(ofproto_port.type, type)
+ && (!iface || !iface->netdev
+ || !strcmp(netdev_get_type(iface->netdev), type))) {
+ continue;
+ }
+ error = ofproto_port_del(br->ofproto, ofproto_port.ofp_port);
+ if (error) {
+ VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
+ br->name, name, strerror(error));
+ }
+ if (iface) {
+ netdev_close(iface->netdev);
+ iface->netdev = NULL;
+ }
}
- return controller && controller[0] ? controller : NULL;
}
static void
-bridge_reconfigure_one(struct bridge *br)
+iface_set_ofp_port(struct iface *iface, int ofp_port)
{
- struct svec old_ports, new_ports, ifaces;
- struct svec listeners, old_listeners;
- struct svec snoops, old_snoops;
- size_t i, j;
+ struct bridge *br = iface->port->bridge;
- /* Collect old ports. */
- svec_init(&old_ports);
- for (i = 0; i < br->n_ports; i++) {
- svec_add(&old_ports, br->ports[i]->name);
- }
- svec_sort(&old_ports);
- assert(svec_is_unique(&old_ports));
+ 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);
+}
- /* Collect new ports. */
- svec_init(&new_ports);
- cfg_get_all_keys(&new_ports, "bridge.%s.port", br->name);
- svec_sort(&new_ports);
- if (bridge_get_controller(br) && !svec_contains(&new_ports, br->name)) {
- svec_add(&new_ports, br->name);
- svec_sort(&new_ports);
- }
- if (!svec_is_unique(&new_ports)) {
- VLOG_WARN("bridge %s: %s specified twice as bridge port",
- br->name, svec_get_duplicate(&new_ports));
- svec_unique(&new_ports);
- }
+static void
+bridge_refresh_ofp_port(struct bridge *br)
+{
+ struct ofproto_port_dump dump;
+ struct ofproto_port ofproto_port;
+ struct port *port;
- ofproto_set_mgmt_id(br->ofproto, mgmt_id);
+ /* Clear all the "ofp_port"es. */
+ hmap_clear(&br->ifaces);
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
- /* Get rid of deleted ports and add new ports. */
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
- if (!svec_contains(&new_ports, port->name)) {
- port_destroy(port);
- } else {
- i++;
- }
- }
- for (i = 0; i < new_ports.n; i++) {
- const char *name = new_ports.names[i];
- if (!svec_contains(&old_ports, name)) {
- port_create(br, name);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface->ofp_port = -1;
}
}
- svec_destroy(&old_ports);
- svec_destroy(&new_ports);
-
- /* Reconfigure all ports. */
- for (i = 0; i < br->n_ports; i++) {
- port_reconfigure(br->ports[i]);
- }
- /* Check and delete duplicate interfaces. */
- svec_init(&ifaces);
- 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 (svec_contains(&ifaces, iface->name)) {
- VLOG_ERR("bridge %s: %s interface is on multiple ports, "
- "removing from %s",
- br->name, iface->name, port->name);
- iface_destroy(iface);
+ /* Obtain the correct "ofp_port"s from ofproto. */
+ OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
+ struct iface *iface = iface_lookup(br, ofproto_port.name);
+ if (iface) {
+ if (iface->ofp_port >= 0) {
+ VLOG_WARN("bridge %s: interface %s reported twice",
+ br->name, ofproto_port.name);
+ } else if (iface_from_ofp_port(br, ofproto_port.ofp_port)) {
+ VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
+ br->name, ofproto_port.ofp_port);
} else {
- svec_add(&ifaces, iface->name);
- svec_sort(&ifaces);
- j++;
+ iface_set_ofp_port(iface, ofproto_port.ofp_port);
}
}
- if (!port->n_ifaces) {
- VLOG_ERR("%s port has no interfaces, dropping", port->name);
- port_destroy(port);
- } else {
- i++;
- }
}
- svec_destroy(&ifaces);
+}
- /* 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?) */
+/* Add an ofproto port for any "struct iface" that doesn't have one.
+ * Delete any "struct iface" for which this fails.
+ * Delete any "struct port" that thereby ends up with no ifaces. */
+static void
+bridge_add_ofproto_ports(struct bridge *br)
+{
+ struct port *port, *next_port;
- /* Configure OpenFlow management listeners. */
- svec_init(&listeners);
- cfg_get_all_strings(&listeners, "bridge.%s.openflow.listeners", br->name);
- if (!listeners.n) {
- svec_add_nocopy(&listeners, xasprintf("punix:%s/%s.mgmt",
- ovs_rundir, br->name));
- } else if (listeners.n == 1 && !strcmp(listeners.names[0], "none")) {
- svec_clear(&listeners);
- }
- svec_sort_unique(&listeners);
+ HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
+ struct iface *iface, *next_iface;
+ struct ofproto_port ofproto_port;
- svec_init(&old_listeners);
- ofproto_get_listeners(br->ofproto, &old_listeners);
- svec_sort_unique(&old_listeners);
+ LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
+ struct shash args;
+ int error;
- if (!svec_equal(&listeners, &old_listeners)) {
- ofproto_set_listeners(br->ofproto, &listeners);
- }
- svec_destroy(&listeners);
- svec_destroy(&old_listeners);
+ /* Open the netdev or reconfigure it. */
+ shash_init(&args);
+ shash_from_ovs_idl_map(iface->cfg->key_options,
+ iface->cfg->value_options,
+ iface->cfg->n_options, &args);
+ if (!iface->netdev) {
+ struct netdev_options options;
+ options.name = iface->name;
+ options.type = iface->type;
+ options.args = &args;
+ options.ethertype = NETDEV_ETH_TYPE_NONE;
+ error = netdev_open(&options, &iface->netdev);
+ } else {
+ error = netdev_set_config(iface->netdev, &args);
+ }
+ shash_destroy(&args);
+ if (error) {
+ VLOG_WARN("could not %s network device %s (%s)",
+ iface->netdev ? "reconfigure" : "open",
+ iface->name, strerror(error));
+ }
- /* Configure OpenFlow controller connection snooping. */
- svec_init(&snoops);
- cfg_get_all_strings(&snoops, "bridge.%s.openflow.snoops", br->name);
- if (!snoops.n) {
- svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop",
- ovs_rundir, br->name));
- } else if (snoops.n == 1 && !strcmp(snoops.names[0], "none")) {
- svec_clear(&snoops);
- }
- svec_sort_unique(&snoops);
+ /* Add the port, if necessary. */
+ if (iface->netdev && iface->ofp_port < 0) {
+ uint16_t ofp_port;
+ int error;
- svec_init(&old_snoops);
- ofproto_get_snoops(br->ofproto, &old_snoops);
- svec_sort_unique(&old_snoops);
+ error = ofproto_port_add(br->ofproto, iface->netdev,
+ &ofp_port);
+ if (!error) {
+ iface_set_ofp_port(iface, ofp_port);
+ } else {
+ netdev_close(iface->netdev);
+ iface->netdev = NULL;
+ }
+ }
- if (!svec_equal(&snoops, &old_snoops)) {
- ofproto_set_snoops(br->ofproto, &snoops);
+ /* Delete the iface if */
+ if (iface->netdev && iface->ofp_port >= 0) {
+ VLOG_DBG("bridge %s: interface %s is on port %d",
+ br->name, iface->name, iface->ofp_port);
+ } else {
+ if (iface->netdev) {
+ VLOG_ERR("bridge %s: missing %s interface, dropping",
+ br->name, iface->name);
+ } else {
+ /* We already reported a related error, don't bother
+ * duplicating it. */
+ }
+ iface_set_ofport(iface->cfg, -1);
+ iface_destroy(iface);
+ }
+ }
+ if (list_is_empty(&port->ifaces)) {
+ VLOG_WARN("%s port has no interfaces, dropping", port->name);
+ port_destroy(port);
+ continue;
+ }
+
+ /* Add bond fake iface if necessary. */
+ if (port_is_bond_fake_iface(port)) {
+ if (ofproto_port_query_by_name(br->ofproto, port->name,
+ &ofproto_port)) {
+ struct netdev_options options;
+ struct netdev *netdev;
+ int error;
+
+ options.name = port->name;
+ options.type = "internal";
+ options.args = NULL;
+ options.ethertype = NETDEV_ETH_TYPE_NONE;
+ error = netdev_open(&options, &netdev);
+ if (!error) {
+ ofproto_port_add(br->ofproto, netdev, NULL);
+ netdev_close(netdev);
+ } else {
+ VLOG_WARN("could not open network device %s (%s)",
+ port->name, strerror(error));
+ }
+ } else {
+ /* Already exists, nothing to do. */
+ ofproto_port_destroy(&ofproto_port);
+ }
+ }
}
- svec_destroy(&snoops);
- svec_destroy(&old_snoops);
+}
+
+static const char *
+get_ovsrec_key_value(const struct ovsdb_idl_row *row,
+ const struct ovsdb_idl_column *column,
+ const char *key)
+{
+ const struct ovsdb_datum *datum;
+ union ovsdb_atom atom;
+ unsigned int idx;
- mirror_reconfigure(br);
+ datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
+ atom.string = (char *) key;
+ idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
+ return idx == UINT_MAX ? NULL : datum->values[idx].string;
+}
+
+static const char *
+bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
+{
+ return get_ovsrec_key_value(&br_cfg->header_,
+ &ovsrec_bridge_col_other_config, key);
}
static void
-bridge_reconfigure_controller(struct bridge *br)
+bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
+ struct iface **hw_addr_iface)
{
- char *pfx = xasprintf("bridge.%s.controller", br->name);
- const char *controller;
+ const char *hwaddr;
+ struct port *port;
+ int error;
+
+ *hw_addr_iface = NULL;
- controller = bridge_get_controller(br);
- if ((br->controller != NULL) != (controller != NULL)) {
- ofproto_flush_flows(br->ofproto);
+ /* 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;
+ }
}
- free(br->controller);
- br->controller = controller ? xstrdup(controller) : NULL;
- if (controller) {
- const char *fail_mode;
- int max_backoff, probe;
- int rate_limit, burst_limit;
+ /* Otherwise choose the minimum non-local MAC address among all of the
+ * interfaces. */
+ memset(ea, 0xff, ETH_ADDR_LEN);
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ uint8_t iface_ea[ETH_ADDR_LEN];
+ struct iface *candidate;
+ struct iface *iface;
- if (!strcmp(controller, "discover")) {
- ofproto_set_discovery(br->ofproto, true,
- cfg_get_string(0, "%s.accept-regex", pfx),
- cfg_get_bool(0, "%s.update-resolv.conf",
- pfx));
+ /* Mirror output ports don't participate. */
+ if (ofproto_is_mirror_output_bundle(br->ofproto, port)) {
+ continue;
+ }
+
+ /* 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 {
- struct netdev *netdev;
- bool in_band;
- int error;
+ /* 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;
+ }
+ }
- in_band = (!cfg_is_valid(CFG_BOOL | CFG_REQUIRED,
- "%s.in-band", pfx)
- || cfg_get_bool(0, "%s.in-band", pfx));
- ofproto_set_discovery(br->ofproto, false, NULL, NULL);
- ofproto_set_in_band(br->ofproto, in_band);
-
- error = netdev_open(br->name, NETDEV_ETH_TYPE_NONE, &netdev);
- if (!error) {
- if (cfg_is_valid(CFG_IP | CFG_REQUIRED, "%s.ip", pfx)) {
- struct in_addr ip, mask, gateway;
- ip.s_addr = cfg_get_ip(0, "%s.ip", pfx);
- mask.s_addr = cfg_get_ip(0, "%s.netmask", pfx);
- gateway.s_addr = cfg_get_ip(0, "%s.gateway", pfx);
-
- netdev_turn_flags_on(netdev, NETDEV_UP, true);
- if (!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));
- }
+ /* The local port doesn't count (since we're trying to choose its
+ * MAC address anyway). */
+ if (iface->ofp_port == OFPP_LOCAL) {
+ continue;
+ }
- if (gateway.s_addr) {
- if (!netdev_add_router(gateway)) {
- VLOG_INFO("bridge %s: configured gateway "IP_FMT,
- br->name, IP_ARGS(&gateway.s_addr));
- }
- }
- }
- netdev_close(netdev);
+ /* 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;
}
}
- fail_mode = cfg_get_string(0, "%s.fail-mode", pfx);
- if (!fail_mode) {
- fail_mode = cfg_get_string(0, "mgmt.fail-mode");
+ /* 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;
}
- ofproto_set_failure(br->ofproto,
- (!fail_mode
- || !strcmp(fail_mode, "standalone")
- || !strcmp(fail_mode, "open")));
+ }
+ 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));
+ }
+}
- probe = cfg_get_int(0, "%s.inactivity-probe", pfx);
- ofproto_set_probe_interval(br->ofproto,
- probe ? probe : cfg_get_int(0, "mgmt.inactivity-probe"));
+/* 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;
- max_backoff = cfg_get_int(0, "%s.max-backoff", pfx);
- if (!max_backoff) {
- max_backoff = cfg_get_int(0, "mgmt.max-backoff");
- if (!max_backoff) {
- max_backoff = 15;
- }
- }
- ofproto_set_max_backoff(br->ofproto, max_backoff);
+ datapath_id = bridge_get_other_config(br->cfg, "datapath-id");
+ if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
+ return dpid;
+ }
- rate_limit = cfg_get_int(0, "%s.rate-limit", pfx);
- if (!rate_limit) {
- rate_limit = cfg_get_int(0, "mgmt.rate-limit");
+ 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.
+ */
}
- burst_limit = cfg_get_int(0, "%s.burst-limit", pfx);
- if (!burst_limit) {
- burst_limit = cfg_get_int(0, "mgmt.burst-limit");
+ } 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;
}
- ofproto_set_rate_limit(br->ofproto, rate_limit, burst_limit);
+ }
- ofproto_set_stp(br->ofproto, cfg_get_bool(0, "%s.stp", pfx));
+ 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];
+
+ BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
+ sha1_bytes(data, n, hash);
+ eth_addr_mark_random(hash);
+ return eth_addr_to_uint64(hash);
+}
- if (cfg_has("%s.commands.acl", pfx)) {
- struct svec command_acls;
- char *command_acl;
+static void
+iface_refresh_status(struct iface *iface)
+{
+ struct shash sh;
+
+ enum netdev_flags flags;
+ uint32_t current;
+ int64_t bps;
+ int mtu;
+ int64_t mtu_64;
+ int error;
- svec_init(&command_acls);
- cfg_get_all_strings(&command_acls, "%s.commands.acl", pfx);
- command_acl = svec_join(&command_acls, ",", "");
+ if (iface_is_synthetic(iface)) {
+ return;
+ }
- ofproto_set_remote_execution(br->ofproto, command_acl,
- cfg_get_string(0, "%s.commands.dir",
- pfx));
+ shash_init(&sh);
- svec_destroy(&command_acls);
- free(command_acl);
- } else {
- ofproto_set_remote_execution(br->ofproto, NULL, NULL);
- }
+ if (!netdev_get_status(iface->netdev, &sh)) {
+ size_t n;
+ char **keys, **values;
+
+ shash_to_ovs_idl_map(&sh, &keys, &values, &n);
+ ovsrec_interface_set_status(iface->cfg, keys, values, n);
+
+ free(keys);
+ free(values);
} else {
- union ofp_action action;
- flow_t flow;
+ ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
+ }
+
+ shash_destroy_free_data(&sh);
- /* Set up a flow that matches every packet and directs them to
- * OFPP_NORMAL (which goes to us). */
- memset(&action, 0, sizeof action);
- action.type = htons(OFPAT_OUTPUT);
- action.output.len = htons(sizeof action);
- action.output.port = htons(OFPP_NORMAL);
- memset(&flow, 0, sizeof flow);
- ofproto_add_flow(br->ofproto, &flow, OFPFW_ALL, 0,
- &action, 1, 0);
+ 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);
+ }
- ofproto_set_in_band(br->ofproto, false);
- ofproto_set_max_backoff(br->ofproto, 1);
- ofproto_set_probe_interval(br->ofproto, 5);
- ofproto_set_failure(br->ofproto, false);
- ofproto_set_stp(br->ofproto, false);
+ 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);
}
- free(pfx);
- ofproto_set_controller(br->ofproto, br->controller);
+ ovsrec_interface_set_link_state(iface->cfg,
+ iface_get_carrier(iface) ? "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);
+ }
}
-static void
-bridge_get_all_ifaces(const struct bridge *br, struct svec *ifaces)
+/* Writes 'iface''s CFM statistics to the database. Returns true if anything
+ * changed, false otherwise. */
+static bool
+iface_refresh_cfm_stats(struct iface *iface)
{
- size_t i, j;
+ const struct ovsrec_interface *cfg = iface->cfg;
+ bool changed = false;
+ int fault;
- svec_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];
- svec_add(ifaces, iface->name);
- }
+ fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
+ iface->ofp_port);
+
+ if (fault < 0) {
+ return false;
+ }
+
+ if (cfg->n_cfm_fault != 1 || cfg->cfm_fault[0] != fault) {
+ bool fault_bool = fault;
+ ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
+ changed = true;
}
- svec_sort(ifaces);
- assert(svec_is_unique(ifaces));
+
+ return changed;
}
-/* 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)
+static bool
+iface_refresh_lacp_stats(struct iface *iface)
{
- struct odp_port *dpif_ports;
- size_t n_dpif_ports;
- size_t i, j;
+ struct ofproto *ofproto = iface->port->bridge->ofproto;
+ int old = iface->cfg->lacp_current ? *iface->cfg->lacp_current : -1;
+ int new = ofproto_port_is_lacp_current(ofproto, iface->ofp_port);
- /* 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;
- }
+ if (old != new) {
+ bool current = new;
+ ovsrec_interface_set_lacp_current(iface->cfg, ¤t, new >= 0);
}
- port_array_clear(&br->ifaces);
+ return old != new;
+}
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
- for (i = 0; i < n_dpif_ports; i++) {
- struct odp_port *p = &dpif_ports[i];
- struct iface *iface = iface_lookup(br, p->devname);
- if (iface) {
- if (iface->dp_ifidx >= 0) {
- VLOG_WARN("dp%u reported interface %s twice",
- dpif_id(&br->dpif), p->devname);
- } else if (iface_from_dp_ifidx(br, p->port)) {
- VLOG_WARN("dp%u reported interface %"PRIu16" twice",
- dpif_id(&br->dpif), p->port);
- } else {
- port_array_set(&br->ifaces, p->port, iface);
- iface->dp_ifidx = p->port;
- }
- }
+static void
+iface_refresh_stats(struct iface *iface)
+{
+#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;
+
+ struct netdev_stats stats;
+
+ if (iface_is_synthetic(iface)) {
+ return;
}
- free(dpif_ports);
+
+ /* 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);
+
+ /* Copy statistics into values[] array. */
+ i = 0;
+#define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
+ IFACE_STATS;
+#undef IFACE_STAT
+ assert(i == ARRAY_SIZE(keys));
+
+ ovsrec_interface_set_statistics(iface->cfg, keys, values, ARRAY_SIZE(keys));
+#undef IFACE_STATS
}
-
-/* Bridge packet processing functions. */
-static struct bond_entry *
-lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN])
+static void
+refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
{
- size_t h = hash_bytes(mac, ETH_ADDR_LEN, 0);
- return &port->bond_hash[h & BOND_MASK];
+ 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);
}
-static int
-bond_choose_iface(const struct port *port)
+static inline const char *
+nx_role_to_str(enum nx_role role)
{
- size_t i;
- for (i = 0; i < port->n_ifaces; i++) {
- if (port->ifaces[i]->enabled) {
- return i;
- }
+ 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 -1;
}
-static bool
-choose_output_iface(const struct port *port, const flow_t *flow,
- uint16_t *dp_ifidx, tag_type *tags)
+static void
+bridge_refresh_controller_status(const struct bridge *br)
{
- struct iface *iface;
+ struct shash info;
+ const struct ovsrec_controller *cfg;
- assert(port->n_ifaces);
- if (port->n_ifaces == 1) {
- iface = port->ifaces[0];
- } else {
- struct bond_entry *e = lookup_bond_entry(port, flow->dl_src);
- 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();
+ ofproto_get_ofproto_controller_info(br->ofproto, &info);
+
+ OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
+ struct ofproto_controller_info *cinfo =
+ shash_find_data(&info, cfg->target);
+
+ 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);
}
- *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;
+
+ ofproto_free_ofproto_controller_info(&info);
}
-static void
-bond_link_status_update(struct iface *iface, bool carrier)
+void
+bridge_run(void)
{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
- struct port *port = iface->port;
+ const struct ovsrec_open_vswitch *cfg;
- if ((carrier == iface->enabled) == (iface->delay_expires == LLONG_MAX)) {
- /* Nothing to do. */
- return;
- }
- VLOG_INFO_RL(&rl, "interface %s: carrier %s",
- iface->name, carrier ? "detected" : "dropped");
- if (carrier == iface->enabled) {
- iface->delay_expires = LLONG_MAX;
- VLOG_INFO_RL(&rl, "interface %s: will not be %s",
- iface->name, carrier ? "disabled" : "enabled");
- } else {
- int delay = carrier ? port->updelay : port->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,
- carrier ? "enabled" : "disabled",
- carrier ? "up" : "down",
- delay);
+ bool datapath_destroyed;
+ bool database_changed;
+ struct bridge *br;
+
+ /* Let each bridge do the work that it needs to do. */
+ datapath_destroyed = false;
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ int error = ofproto_run(br->ofproto);
+ if (error) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "bridge %s: datapath was destroyed externally, "
+ "forcing reconfiguration", br->name);
+ datapath_destroyed = true;
}
}
-}
-static void
-bond_choose_active_iface(struct port *port)
-{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
+ /* (Re)configure if necessary. */
+ database_changed = ovsdb_idl_run(idl);
+ cfg = ovsrec_open_vswitch_first(idl);
- 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);
+ /* 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;
+
+ stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
+ stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
}
-}
-static void
-bond_run(struct bridge *br)
-{
- size_t i, j;
+ if (database_changed || datapath_destroyed) {
+ if (cfg) {
+ struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->n_ifaces < 2) {
- continue;
+ bridge_reconfigure(cfg);
+
+ 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;
+
+ bridge_reconfigure(&null_cfg);
}
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (time_msec() >= iface->delay_expires) {
- iface->delay_expires = LLONG_MAX;
- iface->enabled = !iface->enabled;
- VLOG_WARN("interface %s: %s",
- iface->name,
- iface->enabled ? "enabled" : "disabled");
- if (!iface->enabled) {
- ofproto_revalidate(br->ofproto, iface->tag);
- if (iface->port_ifidx == port->active_iface) {
- ofproto_revalidate(br->ofproto,
- port->active_iface_tag);
- bond_choose_active_iface(port);
- }
- } else {
- if (port->active_iface < 0) {
- ofproto_revalidate(br->ofproto, port->no_ifaces_tag);
- bond_choose_active_iface(port);
+ }
+
+ /* 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);
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ struct port *port;
+
+ 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);
}
- iface->tag = tag_create_random();
}
+ bridge_refresh_controller_status(br);
}
+ refresh_system_stats(cfg);
+ ovsdb_idl_txn_commit(txn);
+ ovsdb_idl_txn_destroy(txn); /* XXX */
}
- }
-}
-
-static void
-bond_wait(struct bridge *br)
-{
- size_t i, j;
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->n_ifaces < 2) {
- continue;
- }
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (iface->delay_expires != LLONG_MAX) {
- poll_timer_wait(iface->delay_expires - time_msec());
- }
- }
+ stats_timer = time_msec() + STATS_INTERVAL;
}
-}
-static bool
-set_dst(struct dst *p, const flow_t *flow,
- const struct port *in_port, const struct port *out_port,
- tag_type *tags)
-{
- /* STP handling.
- *
- * XXX This uses too many tags: any broadcast flow will get one tag per
- * destination port, and thus a broadcast on a switch of any size is likely
- * to have all tag bits set. We should figure out a way to be smarter.
- *
- * This is OK when STP is disabled, because stp_state_tag is 0 then. */
- *tags |= out_port->stp_state_tag;
- if (!(out_port->stp_state & (STP_DISABLED | STP_FORWARDING))) {
- return false;
- }
+ if (time_msec() >= db_limiter) {
+ struct ovsdb_idl_txn *txn;
+ bool changed = false;
- p->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
- : in_port->vlan >= 0 ? in_port->vlan
- : ntohs(flow->dl_vlan));
- return choose_output_iface(out_port, flow, &p->dp_ifidx, tags);
-}
+ txn = ovsdb_idl_txn_create(idl);
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ struct port *port;
-static void
-swap_dst(struct dst *p, struct dst *q)
-{
- struct dst tmp = *p;
- *p = *q;
- *q = tmp;
-}
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
-/* 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 *dsts, size_t n_dsts, int vlan)
-{
- struct dst *first = dsts;
- struct dst *last = dsts + n_dsts;
-
- 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;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ changed = iface_refresh_cfm_stats(iface) || changed;
+ changed = iface_refresh_lacp_stats(iface) || changed;
+ }
}
}
- /* Same invariants, plus one additional:
- * - first->vlan != vlan.
- */
- while (last[-1].vlan != vlan) {
- if (--last == first) {
- return;
- }
+ if (changed) {
+ db_limiter = time_msec() + DB_LIMIT_INTERVAL;
}
- /* Same invariants, plus one additional:
- * - last[-1].vlan == vlan.*/
- swap_dst(first++, --last);
+ ovsdb_idl_txn_commit(txn);
+ ovsdb_idl_txn_destroy(txn);
}
}
-static int
-mirror_mask_ffs(mirror_mask_t mask)
+void
+bridge_wait(void)
{
- BUILD_ASSERT_DECL(sizeof(unsigned int) >= sizeof(mask));
- return ffs(mask);
-}
+ struct bridge *br;
-static bool
-dst_is_duplicate(const struct dst *dsts, size_t n_dsts,
- const struct dst *test)
-{
- size_t i;
- for (i = 0; i < n_dsts; i++) {
- if (dsts[i].vlan == test->vlan && dsts[i].dp_ifidx == test->dp_ifidx) {
- return true;
- }
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_wait(br->ofproto);
}
- return false;
-}
+ ovsdb_idl_wait(idl);
+ poll_timer_wait_until(stats_timer);
-static bool
-port_trunks_vlan(const struct port *port, uint16_t vlan)
-{
- return port->vlan < 0 && bitmap_is_set(port->trunks, vlan);
+ if (db_limiter > time_msec()) {
+ poll_timer_wait_until(db_limiter);
+ }
}
+
+/* QoS unixctl user interface functions. */
-static bool
-port_includes_vlan(const struct port *port, uint16_t vlan)
-{
- return vlan == port->vlan || port_trunks_vlan(port, vlan);
-}
+struct qos_unixctl_show_cbdata {
+ struct ds *ds;
+ struct iface *iface;
+};
-static size_t
-compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan,
- const struct port *in_port, const struct port *out_port,
- struct dst dsts[], tag_type *tags)
-{
- mirror_mask_t mirrors = in_port->src_mirrors;
- struct dst *dst = dsts;
- size_t i;
+static void
+qos_unixctl_show_cb(unsigned int queue_id,
+ const struct shash *details,
+ void *aux)
+{
+ 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;
+ int error;
- *tags |= in_port->stp_state_tag;
- if (out_port == FLOOD_PORT) {
- /* XXX use ODP_FLOOD if no vlans or bonding. */
- /* XXX even better, define each VLAN as a datapath port group */
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port != in_port && port_includes_vlan(port, vlan)
- && !port->is_mirror_output_port
- && set_dst(dst, flow, in_port, port, tags)) {
- mirrors |= port->dst_mirrors;
- dst++;
- }
- }
- } else if (out_port && set_dst(dst, flow, in_port, out_port, tags)) {
- mirrors |= out_port->dst_mirrors;
- dst++;
+ ds_put_cstr(ds, "\n");
+ if (queue_id) {
+ ds_put_format(ds, "Queue %u:\n", queue_id);
+ } else {
+ ds_put_cstr(ds, "Default:\n");
}
- 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(dsts, dst - dsts, dst)) {
- dst++;
- }
- } 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)
- && !dst_is_duplicate(dsts, dst - dsts, dst))
- {
- if (port->vlan < 0) {
- dst->vlan = m->out_vlan;
- }
- if (dst->dp_ifidx == flow->in_port
- && dst->vlan == vlan) {
- /* Don't send out input port on same VLAN. */
- continue;
- }
- dst++;
- }
- }
- }
- }
- mirrors &= mirrors - 1;
+ SHASH_FOR_EACH (node, details) {
+ ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
}
- partition_dsts(dsts, dst - dsts, ntohs(flow->dl_vlan));
- return dst - dsts;
-}
+ 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);
+ }
-static void UNUSED
-print_dsts(const struct dst *dsts, size_t n)
-{
- for (; n--; dsts++) {
- printf(">p%"PRIu16, dsts->dp_ifidx);
- if (dsts->vlan != OFP_VLAN_NONE) {
- printf("v%"PRIu16, dsts->vlan);
+ if (stats.tx_bytes != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
}
- }
-}
-static void
-compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan,
- const struct port *in_port, const struct port *out_port,
- tag_type *tags, struct odp_actions *actions)
-{
- struct dst dsts[DP_MAX_PORTS * (MAX_MIRRORS + 1)];
- size_t n_dsts;
- const struct dst *p;
- uint16_t cur_vlan;
-
- n_dsts = compose_dsts(br, flow, vlan, in_port, out_port, dsts, tags);
-
- cur_vlan = ntohs(flow->dl_vlan);
- for (p = dsts; p < &dsts[n_dsts]; p++) {
- union odp_action *a;
- if (p->vlan != cur_vlan) {
- if (p->vlan == OFP_VLAN_NONE) {
- odp_actions_add(actions, ODPAT_STRIP_VLAN);
- } else {
- a = odp_actions_add(actions, ODPAT_SET_VLAN_VID);
- a->vlan_vid.vlan_vid = htons(p->vlan);
- }
- cur_vlan = p->vlan;
+ if (stats.tx_errors != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
}
- a = odp_actions_add(actions, ODPAT_OUTPUT);
- a->output.port = p->dp_ifidx;
+ } else {
+ ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
+ queue_id, strerror(error));
}
}
-static bool
-is_bcast_arp_reply(const flow_t *flow, const struct ofpbuf *packet)
+static void
+qos_unixctl_show(struct unixctl_conn *conn,
+ const char *args, void *aux OVS_UNUSED)
{
- struct arp_eth_header *arp = (struct arp_eth_header *) packet->data;
- return (flow->dl_type == htons(ETH_TYPE_ARP)
- && eth_addr_is_broadcast(flow->dl_dst)
- && packet->size >= sizeof(struct arp_eth_header)
- && arp->ar_op == ARP_OP_REQUEST);
-}
+ 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;
-/* 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 flow_t *flow,
- const struct ofpbuf *packet, struct odp_actions *actions,
- tag_type *tags)
-{
- struct iface *in_iface;
- struct port *in_port;
- struct port *out_port = NULL; /* By default, drop the packet/flow. */
- 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 (packet != NULL) {
- /* 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);
+ iface = iface_find(args);
+ if (!iface) {
+ unixctl_command_reply(conn, 501, "no such interface");
+ return;
+ }
- VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
- "interface %"PRIu16, br->name, flow->in_port);
- }
+ netdev_get_qos(iface->netdev, &type, &sh);
- /* Return without adding any actions, to drop packets on this flow. */
- return true;
- }
- in_port = in_iface->port;
+ if (*type != '\0') {
+ ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
- /* Figure out what VLAN this packet belongs to.
- *
- * Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet
- * belongs to VLAN 0, so we should treat both cases identically. (In the
- * former case, the packet has an 802.1Q header that specifies VLAN 0,
- * presumably to allow a priority to be specified. In the latter case, the
- * packet does not have any 802.1Q header.) */
- vlan = ntohs(flow->dl_vlan);
- if (vlan == OFP_VLAN_NONE) {
- vlan = 0;
- }
- if (in_port->vlan >= 0) {
- if (vlan) {
- /* XXX support double tagging? */
- if (packet != NULL) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %"PRIu16" tagged "
- "packet received on port %s configured with "
- "implicit VLAN %"PRIu16,
- br->name, ntohs(flow->dl_vlan),
- in_port->name, in_port->vlan);
- }
- goto done;
- }
- vlan = in_port->vlan;
- } else {
- if (!port_includes_vlan(in_port, vlan)) {
- 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);
- goto done;
- }
- }
-
- /* Drop frames for ports that STP wants entirely killed (both for
- * forwarding and for learning). Later, after we do learning, we'll drop
- * the frames that STP wants to do learning but not forwarding on. */
- if (in_port->stp_state & (STP_LISTENING | STP_BLOCKING)) {
- goto done;
- }
-
- /* Drop frames for reserved multicast addresses. */
- if (eth_addr_is_reserved(flow->dl_dst)) {
- goto done;
- }
-
- /* Drop frames on ports reserved for mirroring. */
- if (in_port->is_mirror_output_port) {
- 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);
- goto done;
- }
-
- /* Drop multicast and broadcast packets on inactive bonded interfaces, to
- * avoid receiving duplicates. */
- if (in_port->n_ifaces > 1 && eth_addr_is_multicast(flow->dl_dst)) {
- *tags |= in_port->active_iface_tag;
- if (in_port->active_iface != in_iface->port_ifidx) {
- goto done;
- }
- }
-
- /* MAC learning. */
- out_port = FLOOD_PORT;
- if (br->ml) {
- int out_port_idx;
- bool may_learn;
-
- if (!packet) {
- /* Don't try to learn from revalidation. */
- may_learn = false;
- } else if (in_port->n_ifaces > 1) {
- /* If the packet arrived on a bonded port, don't learn from it
- * unless we haven't learned any port at all for that address
- * (because we probably sent the packet on one bonded interface and
- * got it back on the other). Broadcast ARP replies are an
- * exception to this rule: the host has moved to another switch. */
- int src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan);
- may_learn = (src_idx < 0
- || src_idx == in_port->port_idx
- || is_bcast_arp_reply(flow, packet));
- } else {
- may_learn = true;
- }
-
- /* Learn source MAC. */
- if (may_learn) {
- tag_type rev_tag = mac_learning_learn(br->ml, flow->dl_src,
- vlan, in_port->port_idx);
- 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);
- }
+ SHASH_FOR_EACH (node, &sh) {
+ ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
}
- /* Determine output port. */
- out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan,
- tags);
- if (out_port_idx >= 0 && out_port_idx < br->n_ports) {
- out_port = br->ports[out_port_idx];
+ 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));
}
- /* Don't send packets out their input ports. Don't forward frames that STP
- * wants us to discard. */
- if (in_port == out_port || in_port->stp_state == STP_LEARNING) {
- out_port = NULL;
- }
+ shash_destroy_free_data(&sh);
+ ds_destroy(&ds);
+}
+
+/* Bridge reconfiguration functions. */
+static void
+bridge_create(const struct ovsrec_bridge *br_cfg)
+{
+ struct bridge *br;
-done:
- compose_actions(br, flow, vlan, in_port, out_port, tags, actions);
+ assert(!bridge_lookup(br_cfg->name));
+ br = xzalloc(sizeof *br);
- /*
- * We send out only a single packet, instead of setting up a flow, if the
- * packet is an ARP directed to broadcast that arrived on a bonded
- * interface. In such a situation ARP requests and replies must be handled
- * differently, but OpenFlow unfortunately can't distinguish them.
- */
- return (in_port->n_ifaces < 2
- || flow->dl_type != htons(ETH_TYPE_ARP)
- || !eth_addr_is_broadcast(flow->dl_dst));
+ br->name = xstrdup(br_cfg->name);
+ br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
+ br->cfg = br_cfg;
+
+ /* 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);
+
+ hmap_init(&br->ports);
+ hmap_init(&br->ifaces);
+ hmap_init(&br->iface_by_name);
+ hmap_init(&br->mirrors);
+
+ hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
}
-/* Careful: 'opp' is in host byte order and opp->port_no is an OFP port
- * number. */
static void
-bridge_port_changed_ofhook_cb(enum ofp_port_reason reason,
- const struct ofp_phy_port *opp,
- void *br_)
+bridge_destroy(struct bridge *br)
{
- struct bridge *br = br_;
- struct iface *iface;
- struct port *port;
-
- iface = iface_from_dp_ifidx(br, ofp_port_to_odp_port(opp->port_no));
- if (!iface) {
- return;
- }
- port = iface->port;
+ if (br) {
+ struct mirror *mirror, *next_mirror;
+ struct port *port, *next_port;
- if (reason == OFPPR_DELETE) {
- VLOG_WARN("bridge %s: interface %s deleted unexpectedly",
- br->name, iface->name);
- iface_destroy(iface);
- if (!port->n_ifaces) {
- VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
- br->name, port->name);
+ HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
port_destroy(port);
}
-
- bridge_flush(br);
- } else {
- memcpy(iface->mac, opp->hw_addr, ETH_ADDR_LEN);
- if (port->n_ifaces > 1) {
- bool up = !(opp->state & OFPPS_LINK_DOWN);
- bond_link_status_update(iface, up);
- port_update_bond_compat(port);
+ HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
+ mirror_destroy(mirror);
}
+ 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);
+ free(br->name);
+ free(br->type);
+ free(br);
}
}
-static bool
-bridge_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet,
- struct odp_actions *actions, tag_type *tags, void *br_)
+static struct bridge *
+bridge_lookup(const char *name)
{
- struct bridge *br = br_;
+ struct bridge *br;
-#if 0
- if (flow->dl_type == htons(OFP_DL_TYPE_NOT_ETH_TYPE)
- && eth_addr_equals(flow->dl_dst, stp_eth_addr)) {
- brstp_receive(br, flow, payload);
- return true;
+ HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
+ if (!strcmp(br->name, name)) {
+ return br;
+ }
}
-#endif
-
- COVERAGE_INC(bridge_process_flow);
- return process_flow(br, flow, packet, actions, tags);
+ return NULL;
}
+/* Handle requests for a listing of all flows known by the OpenFlow
+ * stack, including those normally hidden. */
static void
-bridge_account_flow_ofhook_cb(const flow_t *flow,
- const union odp_action *actions,
- size_t n_actions, unsigned long long int n_bytes,
- void *br_)
+bridge_unixctl_dump_flows(struct unixctl_conn *conn,
+ const char *args, void *aux OVS_UNUSED)
{
- struct bridge *br = br_;
- const union odp_action *a;
+ struct bridge *br;
+ struct ds results;
- if (!br->has_bonded_ports) {
+ br = bridge_lookup(args);
+ if (!br) {
+ unixctl_command_reply(conn, 501, "Unknown bridge");
return;
}
- for (a = actions; a < &actions[n_actions]; a++) {
- if (a->type == ODPAT_OUTPUT) {
- struct port *port = port_from_dp_ifidx(br, a->output.port);
- if (port && port->n_ifaces >= 2) {
- struct bond_entry *e = lookup_bond_entry(port, flow->dl_src);
- e->tx_bytes += n_bytes;
- }
+ ds_init(&results);
+ ofproto_get_all_flows(br->ofproto, &results);
+
+ unixctl_command_reply(conn, 200, ds_cstr(&results));
+ ds_destroy(&results);
+}
+
+/* "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,
+ const char *args, void *aux OVS_UNUSED)
+{
+ 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 {
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ ofproto_reconnect_controllers(br->ofproto);
}
}
+ unixctl_command_reply(conn, 200, NULL);
}
-static void
-bridge_account_checkpoint_ofhook_cb(void *br_)
+static size_t
+bridge_get_controllers(const struct bridge *br,
+ struct ovsrec_controller ***controllersp)
{
- struct bridge *br = br_;
- size_t i;
+ struct ovsrec_controller **controllers;
+ size_t n_controllers;
- if (!br->has_bonded_ports) {
- return;
+ controllers = br->cfg->controller;
+ n_controllers = br->cfg->n_controller;
+
+ if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
+ controllers = NULL;
+ n_controllers = 0;
}
- /* The current ofproto implementation calls this callback at least once a
- * second, so this timer implementation is sufficient. */
- if (time_msec() < br->bond_next_rebalance) {
- return;
+ if (controllersp) {
+ *controllersp = controllers;
}
- br->bond_next_rebalance = time_msec() + 10000;
+ return n_controllers;
+}
+
+/* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
+ * those configured in 'br->cfg'. */
+static void
+bridge_add_del_ports(struct bridge *br)
+{
+ struct port *port, *next;
+ struct shash_node *node;
+ struct shash new_ports;
+ size_t i;
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->n_ifaces > 1) {
- bond_rebalance_port(port);
+ /* 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);
-static struct ofhooks bridge_ofhooks = {
- bridge_port_changed_ofhook_cb,
- bridge_normal_ofhook_cb,
- bridge_account_flow_ofhook_cb,
- bridge_account_checkpoint_ofhook_cb,
-};
-
-/* 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;
-};
+ br->synth_local_port.interfaces = &br->synth_local_ifacep;
+ br->synth_local_port.n_interfaces = 1;
+ br->synth_local_port.name = br->name;
-/* 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_)
-{
- 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;
+ br->synth_local_iface.name = br->name;
+ br->synth_local_iface.type = "internal";
+
+ br->synth_local_ifacep = &br->synth_local_iface;
+
+ shash_add(&new_ports, br->name, &br->synth_local_port);
}
-}
-/* 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_)
-{
- 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;
+ /* Get rid of deleted ports.
+ * Get rid of deleted interfaces on ports that still exist.
+ * Update 'cfg' of 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);
+ }
+ }
+
+ /* Create new ports.
+ * Add new interfaces to existing ports. */
+ SHASH_FOR_EACH (node, &new_ports) {
+ struct port *port = port_lookup(br, node->name);
+ if (!port) {
+ struct ovsrec_port *cfg = node->data;
+ port = port_create(br, cfg);
+ }
+ port_add_ifaces(port);
+ if (list_is_empty(&port->ifaces)) {
+ VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
+ br->name, port->name);
+ port_destroy(port);
+ }
}
+ 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
-swap_bals(struct slave_balance *a, struct slave_balance *b)
+bridge_ofproto_controller_for_mgmt(const struct bridge *br,
+ struct ofproto_controller *oc)
{
- struct slave_balance tmp = *a;
- *a = *b;
- *b = tmp;
+ 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;
}
-/* 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. */
+/* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
static void
-resort_bals(struct slave_balance *p,
- struct slave_balance bals[], size_t n_bals)
+bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
+ struct ofproto_controller *oc)
{
- 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);
- }
- }
+ 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
-log_bals(const struct slave_balance *bals, size_t n_bals, struct port *port)
+bridge_configure_local_iface_netdev(struct bridge *br,
+ struct ovsrec_controller *c)
{
- if (VLOG_IS_DBG_ENABLED()) {
- struct ds ds = DS_EMPTY_INITIALIZER;
- const struct slave_balance *b;
+ struct netdev *netdev;
+ struct in_addr mask, gateway;
- for (b = bals; b < bals + n_bals; b++) {
- size_t i;
+ struct iface *local_iface;
+ struct in_addr ip;
- if (b > bals) {
- ds_put_char(&ds, ',');
- }
- ds_put_format(&ds, " %s %"PRIu64"kB",
- b->iface->name, b->tx_bytes / 1024);
+ /* 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 (!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, ")");
- }
+ /* 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));
}
- VLOG_DBG("bond %s:%s", port->name, ds_cstr(&ds));
- ds_destroy(&ds);
}
}
-/* Shifts 'hash' from 'from' to 'to' within 'port'. */
static void
-bond_shift_load(struct slave_balance *from, struct slave_balance *to,
- struct bond_entry *hash)
+bridge_configure_remotes(struct bridge *br,
+ const struct sockaddr_in *managers, size_t n_managers)
{
- struct port *port = from->iface->port;
- uint64_t delta = hash->tx_bytes;
+ const char *disable_ib_str, *queue_id_str;
+ bool disable_in_band = false;
+ int queue_id;
- 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);
+ struct ovsrec_controller **controllers;
+ size_t n_controllers;
- /* 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 (from->hashes[0] == hash) {
- from->hashes++;
+ 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_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 {
- int i = hash - from->hashes[0];
- memmove(from->hashes + i, from->hashes + i + 1,
- (from->n_hashes - (i + 1)) * sizeof *from->hashes);
+ ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
}
- from->n_hashes--;
- /* Shift load away from 'from' to 'to'. */
- from->tx_bytes -= delta;
- to->tx_bytes += delta;
+ n_controllers = bridge_get_controllers(br, &controllers);
- /* 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();
+ 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];
-static void
-bond_rebalance_port(struct port *port)
-{
- struct slave_balance bals[DP_MAX_PORTS];
- size_t n_bals;
- struct bond_entry *hashes[BOND_MASK + 1];
- struct slave_balance *b, *from, *to;
- struct bond_entry *e;
- size_t i;
+ if (!strncmp(c->target, "punix:", 6)
+ || !strncmp(c->target, "unix:", 5)) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- /* 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;
- 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];
- }
- b->n_hashes++;
+ /* Prevent remote ovsdb-server users from accessing arbitrary Unix
+ * domain sockets and overwriting arbitrary local files. */
+ VLOG_ERR_RL(&rl, "bridge %s: not adding Unix domain socket "
+ "controller \"%s\" due to possibility for remote "
+ "exploit", br->name, c->target);
+ continue;
}
- }
- qsort(bals, n_bals, sizeof *bals, compare_slave_balance);
- log_bals(bals, n_bals, port);
- /* 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) {
- return;
+ 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++;
}
- /* 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 not cause 'to''s load to exceed 'from''s load.
- *
- * 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. */
- size_t i;
-
- for (i = 0; i < from->n_hashes; i++) {
- uint64_t delta = from->hashes[i]->tx_bytes;
- if (to->tx_bytes + delta < from->tx_bytes - delta) {
- break;
- }
- }
- if (i < from->n_hashes) {
- bond_shift_load(from, to, from->hashes[i]);
-
- /* 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++;
- }
- }
- }
+ ofproto_set_controllers(br->ofproto, ocs, n_ocs);
+ free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
+ free(ocs);
- /* 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;
+ /* 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);
+
+ /* Configure OpenFlow controller connection snooping. */
+ if (!ofproto_has_snoops(br->ofproto)) {
+ struct sset snoops;
+
+ 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);
}
}
/* Port functions. */
-static void
-port_create(struct bridge *br, const char *name)
+static struct port *
+port_create(struct bridge *br, const struct ovsrec_port *cfg)
{
struct port *port;
- port = xcalloc(1, sizeof *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;
- port->stp_state = STP_DISABLED;
- port->stp_state_tag = 0;
+ port->name = xstrdup(cfg->name);
+ port->cfg = cfg;
+ list_init(&port->ifaces);
- if (br->n_ports >= br->allocated_ports) {
- br->ports = x2nrealloc(br->ports, &br->allocated_ports,
- sizeof *br->ports);
- }
- br->ports[br->n_ports++] = port;
+ hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
VLOG_INFO("created port %s on bridge %s", port->name, br->name);
- bridge_flush(br);
+
+ 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_reconfigure(struct port *port)
+port_del_ifaces(struct port *port)
{
- bool bonded = cfg_has_section("bonding.%s", port->name);
- struct svec old_ifaces, new_ifaces;
- unsigned long *trunks;
- int vlan;
+ struct iface *iface, *next;
+ struct sset new_ifaces;
size_t i;
- /* Collect old and new interfaces. */
- svec_init(&old_ifaces);
- svec_init(&new_ifaces);
- for (i = 0; i < port->n_ifaces; i++) {
- svec_add(&old_ifaces, port->ifaces[i]->name);
- }
- svec_sort(&old_ifaces);
- if (bonded) {
- cfg_get_all_keys(&new_ifaces, "bonding.%s.slave", port->name);
- if (!new_ifaces.n) {
- VLOG_ERR("port %s: no interfaces specified for bonded port",
- port->name);
- } else if (new_ifaces.n == 1) {
- VLOG_WARN("port %s: only 1 interface specified for bonded port",
- port->name);
- }
-
- port->updelay = cfg_get_int(0, "bonding.%s.updelay", port->name);
- if (port->updelay < 0) {
- port->updelay = 0;
- }
- port->downdelay = cfg_get_int(0, "bonding.%s.downdelay", port->name);
- if (port->downdelay < 0) {
- port->downdelay = 0;
+ /* Collect list of new interfaces. */
+ sset_init(&new_ifaces);
+ for (i = 0; i < port->cfg->n_interfaces; i++) {
+ const char *name = port->cfg->interfaces[i]->name;
+ const char *type = port->cfg->interfaces[i]->name;
+ if (strcmp(type, "null")) {
+ sset_add(&new_ifaces, name);
}
- } else {
- svec_init(&new_ifaces);
- svec_add(&new_ifaces, port->name);
}
- /* Get rid of deleted interfaces and add new interfaces. */
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- if (!svec_contains(&new_ifaces, iface->name)) {
+ /* Get rid of deleted interfaces. */
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ if (!sset_contains(&new_ifaces, iface->name)) {
iface_destroy(iface);
- } else {
- i++;
- }
- }
- for (i = 0; i < new_ifaces.n; i++) {
- const char *name = new_ifaces.names[i];
- if (!svec_contains(&old_ifaces, name)) {
- iface_create(port, name);
}
}
- /* Get VLAN tag. */
- vlan = -1;
- if (cfg_has("vlan.%s.tag", port->name)) {
- if (!bonded) {
- vlan = cfg_get_vlan(0, "vlan.%s.tag", port->name);
- if (vlan >= 0 && vlan <= 4095) {
- VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
- }
- } 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);
+ sset_destroy(&new_ifaces);
+}
+
+/* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
+ * existing ones. */
+static void
+port_add_ifaces(struct port *port)
+{
+ struct shash new_ifaces;
+ struct shash_node *node;
+ size_t i;
+
+ /* Collect new ifaces. */
+ shash_init(&new_ifaces);
+ for (i = 0; i < port->cfg->n_interfaces; i++) {
+ const struct ovsrec_interface *cfg = port->cfg->interfaces[i];
+ if (strcmp(cfg->type, "null")
+ && !shash_add_once(&new_ifaces, cfg->name, cfg)) {
+ VLOG_WARN("port %s: %s specified twice as port interface",
+ port->name, cfg->name);
+ iface_set_ofport(cfg, -1);
}
}
- if (port->vlan != vlan) {
- port->vlan = vlan;
- bridge_flush(port->bridge);
- }
- /* Get trunked VLANs. */
- trunks = NULL;
- if (vlan < 0) {
- size_t n_trunks, n_errors;
- size_t i;
+ /* Create new interfaces.
+ * Update interface types and 'cfg' members. */
+ SHASH_FOR_EACH (node, &new_ifaces) {
+ const struct ovsrec_interface *cfg = node->data;
+ const char *iface_name = node->name;
+ struct iface *iface;
- trunks = bitmap_allocate(4096);
- n_trunks = cfg_count("vlan.%s.trunks", port->name);
- n_errors = 0;
- for (i = 0; i < n_trunks; i++) {
- int trunk = cfg_get_vlan(i, "vlan.%s.trunks", port->name);
- if (trunk >= 0) {
- bitmap_set1(trunks, trunk);
- } else {
- n_errors++;
- }
- }
- if (n_errors) {
- VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
- port->name, n_trunks);
- }
- if (n_errors == n_trunks) {
- if (n_errors) {
- VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
- port->name);
- }
- bitmap_set_multiple(trunks, 0, 4096, 1);
- }
- } else {
- if (cfg_has("vlan.%s.trunks", port->name)) {
- VLOG_ERR("ignoring vlan.%s.trunks in favor of vlan.%s.vlan",
- port->name, port->name);
+ iface = iface_lookup(port->bridge, iface_name);
+ if (!iface) {
+ iface = iface_create(port, cfg);
+ } else {
+ iface->cfg = cfg;
}
- }
- if (trunks == NULL
- ? port->trunks != NULL
- : port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
- bridge_flush(port->bridge);
- }
- bitmap_free(port->trunks);
- port->trunks = trunks;
- svec_destroy(&old_ifaces);
- svec_destroy(&new_ifaces);
+ /* Determine interface type. The local port always has type
+ * "internal". Other ports take their type from the database and
+ * default to "system" if none is specified. */
+ iface->type = (!strcmp(iface_name, port->bridge->name) ? "internal"
+ : cfg->type[0] ? cfg->type
+ : "system");
+ }
+ shash_destroy(&new_ifaces);
}
static void
@@ -2321,47 +2047,32 @@ port_destroy(struct port *port)
{
if (port) {
struct bridge *br = port->bridge;
- struct port *del;
- size_t i;
-
- proc_net_compat_update_vlan(port->name, NULL, 0);
+ struct iface *iface, *next;
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m && m->out_port == port) {
- mirror_destroy(m);
- }
+ if (br->ofproto) {
+ ofproto_bundle_unregister(br->ofproto, port);
}
- while (port->n_ifaces > 0) {
- iface_destroy(port->ifaces[port->n_ifaces - 1]);
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ iface_destroy(iface);
}
- del = br->ports[port->port_idx] = br->ports[--br->n_ports];
- del->port_idx = port->port_idx;
+ hmap_remove(&br->ports, &port->hmap_node);
+
+ VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
- free(port->ifaces);
- bitmap_free(port->trunks);
free(port->name);
free(port);
- bridge_flush(br);
}
}
-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;
-}
-
static struct port *
port_lookup(const struct bridge *br, const char *name)
{
- size_t i;
+ struct port *port;
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
+ &br->ports) {
if (!strcmp(port->name, name)) {
return port;
}
@@ -2369,135 +2080,182 @@ port_lookup(const struct bridge *br, const char *name)
return NULL;
}
-static void
-port_update_bonding(struct port *port)
+static bool
+enable_lacp(struct port *port, bool *activep)
{
- if (port->n_ifaces < 2) {
- /* Not a bonded port. */
- if (port->bond_hash) {
- free(port->bond_hash);
- port->bond_hash = NULL;
- proc_net_compat_update_bond(port->name, NULL);
- }
+ if (!port->cfg->lacp) {
+ /* XXX when LACP implementation has been sufficiently tested, enable by
+ * default and make active on bonded ports. */
+ return false;
+ } else if (!strcmp(port->cfg->lacp, "off")) {
+ return false;
+ } else if (!strcmp(port->cfg->lacp, "active")) {
+ *activep = true;
+ return true;
+ } else if (!strcmp(port->cfg->lacp, "passive")) {
+ *activep = false;
+ return true;
} else {
- if (!port->bond_hash) {
- size_t i;
-
- 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->no_ifaces_tag = tag_create_random();
- bond_choose_active_iface(port);
- }
- port_update_bond_compat(port);
+ VLOG_WARN("port %s: unknown LACP mode %s",
+ port->name, port->cfg->lacp);
+ return false;
+ }
+}
+
+static struct lacp_settings *
+port_configure_lacp(struct port *port, struct lacp_settings *s)
+{
+ const char *lacp_time;
+ long long int custom_time;
+ int priority;
+
+ if (!enable_lacp(port, &s->active)) {
+ return NULL;
+ }
+
+ s->name = port->name;
+ memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
+
+ /* Prefer bondable links if unspecified. */
+ priority = atoi(get_port_other_config(port->cfg, "lacp-system-priority",
+ "0"));
+ s->priority = (priority > 0 && priority <= UINT16_MAX
+ ? priority
+ : UINT16_MAX - !list_is_short(&port->ifaces));
+
+ s->heartbeat = !strcmp(get_port_other_config(port->cfg,
+ "lacp-heartbeat",
+ "false"), "true");
+
+
+ lacp_time = get_port_other_config(port->cfg, "lacp-time", "slow");
+ custom_time = atoi(lacp_time);
+ if (!strcmp(lacp_time, "fast")) {
+ s->lacp_time = LACP_TIME_FAST;
+ } else if (!strcmp(lacp_time, "slow")) {
+ s->lacp_time = LACP_TIME_SLOW;
+ } else if (custom_time > 0) {
+ s->lacp_time = LACP_TIME_CUSTOM;
+ s->custom_time = custom_time;
+ } else {
+ s->lacp_time = LACP_TIME_SLOW;
}
+
+ return s;
}
static void
-port_update_bond_compat(struct port *port)
+iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
{
- struct compat_bond bond;
- size_t i;
+ int priority, portid, key;
- if (port->n_ifaces < 2) {
- return;
+ portid = atoi(get_interface_other_config(iface->cfg, "lacp-port-id", "0"));
+ priority = atoi(get_interface_other_config(iface->cfg,
+ "lacp-port-priority", "0"));
+ key = atoi(get_interface_other_config(iface->cfg, "lacp-aggregation-key",
+ "0"));
+
+ if (portid <= 0 || portid > UINT16_MAX) {
+ portid = iface->ofp_port;
}
- bond.up = false;
- bond.updelay = port->updelay;
- bond.downdelay = port->downdelay;
- bond.n_slaves = port->n_ifaces;
- bond.slaves = xmalloc(port->n_ifaces * sizeof *bond.slaves);
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- struct compat_bond_slave *slave = &bond.slaves[i];
- slave->name = iface->name;
- slave->up = ((iface->enabled && iface->delay_expires == LLONG_MAX) ||
- (!iface->enabled && iface->delay_expires != LLONG_MAX));
- if (slave->up) {
- bond.up = true;
- }
- memcpy(slave->mac, iface->mac, ETH_ADDR_LEN);
+ if (priority <= 0 || priority > UINT16_MAX) {
+ priority = UINT16_MAX;
+ }
+
+ if (key < 0 || key > UINT16_MAX) {
+ key = 0;
}
- proc_net_compat_update_bond(port->name, &bond);
- free(bond.slaves);
+
+ s->name = iface->name;
+ s->id = portid;
+ s->priority = priority;
+ s->key = key;
}
static void
-port_update_vlan_compat(struct port *port)
+port_configure_bond(struct port *port, struct bond_settings *s,
+ uint32_t *bond_stable_ids)
{
- struct bridge *br = port->bridge;
- char *vlandev_name = NULL;
-
- if (port->vlan > 0) {
- /* Figure out the name that the VLAN device should actually have, if it
- * existed. This takes some work because the VLAN device would not
- * have port->name in its name; rather, it would have the trunk port's
- * name, and 'port' would be attached to a bridge that also had the
- * VLAN device one of its ports. So we need to find a trunk port that
- * includes port->vlan.
- *
- * There might be more than one candidate. This doesn't happen on
- * XenServer, so if it happens we just pick the first choice in
- * alphabetical order instead of creating multiple VLAN devices. */
- size_t i;
- for (i = 0; i < br->n_ports; i++) {
- struct port *p = br->ports[i];
- if (port_trunks_vlan(p, port->vlan)
- && p->n_ifaces
- && (!vlandev_name || strcmp(p->name, vlandev_name) <= 0))
- {
- const uint8_t *ea = p->ifaces[0]->mac;
- if (!eth_addr_is_multicast(ea) &&
- !eth_addr_is_reserved(ea) &&
- !eth_addr_is_zero(ea)) {
- vlandev_name = p->name;
- }
- }
+ const char *detect_s;
+ struct iface *iface;
+ int miimon_interval;
+ size_t i;
+
+ s->name = port->name;
+ s->balance = BM_SLB;
+ if (port->cfg->bond_mode
+ && !bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
+ VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
+ port->name, port->cfg->bond_mode,
+ bond_mode_to_string(s->balance));
+ }
+
+ miimon_interval = atoi(get_port_other_config(port->cfg,
+ "bond-miimon-interval", "0"));
+ if (miimon_interval <= 0) {
+ miimon_interval = 200;
+ }
+
+ detect_s = get_port_other_config(port->cfg, "bond-detect-mode", "carrier");
+ if (!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;
+ }
+
+ s->up_delay = MAX(0, port->cfg->bond_updelay);
+ s->down_delay = MAX(0, port->cfg->bond_downdelay);
+ s->basis = atoi(get_port_other_config(port->cfg, "bond-hash-basis", "0"));
+ s->rebalance_interval = atoi(
+ get_port_other_config(port->cfg, "bond-rebalance-interval", "10000"));
+ if (s->rebalance_interval < 1000) {
+ s->rebalance_interval = 1000;
+ }
+
+ s->fake_iface = port->cfg->bond_fake_iface;
+
+ i = 0;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ long long stable_id;
+
+ stable_id = atoll(get_interface_other_config(iface->cfg,
+ "bond-stable-id", "0"));
+ if (stable_id <= 0 || stable_id >= UINT32_MAX) {
+ stable_id = iface->ofp_port;
}
+ bond_stable_ids[i++] = stable_id;
+
+ netdev_set_miimon_interval(iface->netdev, miimon_interval);
}
- proc_net_compat_update_vlan(port->name, vlandev_name, port->vlan);
}
/* Interface functions. */
-static void
-iface_create(struct port *port, const char *name)
+static struct iface *
+iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
{
- enum netdev_flags flags;
+ struct bridge *br = port->bridge;
struct iface *iface;
+ char *name = if_cfg->name;
- iface = xcalloc(1, sizeof *iface);
+ iface = xzalloc(sizeof *iface);
iface->port = port;
- iface->port_ifidx = port->n_ifaces;
iface->name = xstrdup(name);
- iface->dp_ifidx = -1;
+ iface->ofp_port = -1;
iface->tag = tag_create_random();
- iface->enabled = true;
- iface->delay_expires = LLONG_MAX;
-
- netdev_nodev_get_etheraddr(name, iface->mac);
+ iface->netdev = NULL;
+ iface->cfg = if_cfg;
- if (!netdev_nodev_get_flags(name, &flags)) {
- iface->enabled = (flags & NETDEV_UP) != 0;
- }
+ hmap_insert(&br->iface_by_name, &iface->name_node, hash_string(name, 0));
- 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) {
- port->bridge->has_bonded_ports = true;
- }
+ list_push_back(&port->ifaces, &iface->port_elem);
VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
- port_update_bonding(port);
- bridge_flush(port->bridge);
+ return iface;
}
static void
@@ -2506,553 +2264,444 @@ iface_destroy(struct iface *iface)
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->dp_ifidx >= 0) {
- port_array_set(&br->ifaces, iface->dp_ifidx, NULL);
+ if (br->ofproto && iface->ofp_port >= 0) {
+ ofproto_port_unregister(br->ofproto, iface->ofp_port);
}
- del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces];
- del->port_ifidx = iface->port_ifidx;
+ if (iface->ofp_port >= 0) {
+ hmap_remove(&br->ifaces, &iface->ofp_port_node);
+ }
- free(iface->name);
- free(iface);
+ list_remove(&iface->port_elem);
+ hmap_remove(&br->iface_by_name, &iface->name_node);
- if (del_active) {
- ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
- bond_choose_active_iface(port);
- }
+ netdev_close(iface->netdev);
- port_update_bonding(port);
- bridge_flush(port->bridge);
+ free(iface->name);
+ free(iface);
}
}
static struct iface *
iface_lookup(const struct bridge *br, const char *name)
{
- size_t i, j;
+ struct iface *iface;
- 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 (!strcmp(iface->name, name)) {
- return 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 *
-iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
+iface_find(const char *name)
{
- return port_array_get(&br->ifaces, dp_ifidx);
-}
-
-/* Port mirroring. */
+ const struct bridge *br;
-static void
-mirror_reconfigure(struct bridge *br)
-{
- struct svec old_mirrors, new_mirrors;
- size_t i;
+ HMAP_FOR_EACH (br, node, &all_bridges) {
+ struct iface *iface = iface_lookup(br, name);
- /* Collect old and new mirrors. */
- svec_init(&old_mirrors);
- svec_init(&new_mirrors);
- cfg_get_subsections(&new_mirrors, "mirror.%s", br->name);
- for (i = 0; i < MAX_MIRRORS; i++) {
- if (br->mirrors[i]) {
- svec_add(&old_mirrors, br->mirrors[i]->name);
+ if (iface) {
+ return iface;
}
}
+ return NULL;
+}
- /* Get rid of deleted mirrors and add new mirrors. */
- svec_sort(&old_mirrors);
- assert(svec_is_unique(&old_mirrors));
- svec_sort(&new_mirrors);
- assert(svec_is_unique(&new_mirrors));
- for (i = 0; i < MAX_MIRRORS; i++) {
- struct mirror *m = br->mirrors[i];
- if (m && !svec_contains(&new_mirrors, m->name)) {
- mirror_destroy(m);
- }
- }
- for (i = 0; i < new_mirrors.n; i++) {
- const char *name = new_mirrors.names[i];
- if (!svec_contains(&old_mirrors, name)) {
- mirror_create(br, name);
- }
- }
- svec_destroy(&old_mirrors);
- svec_destroy(&new_mirrors);
+static struct iface *
+iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
+{
+ struct iface *iface;
- /* Reconfigure all mirrors. */
- for (i = 0; i < MAX_MIRRORS; i++) {
- if (br->mirrors[i]) {
- mirror_reconfigure_one(br->mirrors[i]);
+ HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
+ hash_int(ofp_port, 0), &br->ifaces) {
+ if (iface->ofp_port == ofp_port) {
+ return iface;
}
}
+ return NULL;
+}
- /* 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;
+/* Set Ethernet address of 'iface', if one is specified in the configuration
+ * file. */
+static void
+iface_set_mac(struct iface *iface)
+{
+ uint8_t ea[ETH_ADDR_LEN];
+
+ 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 {
+ int error = netdev_set_etheraddr(iface->netdev, ea);
+ if (error) {
+ VLOG_ERR("interface %s: setting MAC failed (%s)",
+ iface->name, strerror(error));
+ }
}
}
}
+/* Sets the ofport column of 'if_cfg' to 'ofport'. */
static void
-mirror_create(struct bridge *br, const char *name)
+iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
{
- 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, name);
- return;
- }
- if (!br->mirrors[i]) {
- break;
- }
+ if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
+ ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
}
-
- VLOG_INFO("created port mirror %s on bridge %s", name, br->name);
- bridge_flush(br);
-
- br->mirrors[i] = m = xcalloc(1, sizeof *m);
- m->bridge = br;
- m->idx = i;
- m->name = xstrdup(name);
- svec_init(&m->src_ports);
- svec_init(&m->dst_ports);
- m->vlans = NULL;
- m->n_vlans = 0;
- m->out_vlan = -1;
- m->out_port = NULL;
}
+/* 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
-mirror_destroy(struct mirror *m)
+shash_from_ovs_idl_map(char **keys, char **values, size_t n,
+ struct shash *shash)
{
- 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);
- }
-
- svec_destroy(&m->src_ports);
- svec_destroy(&m->dst_ports);
- free(m->vlans);
-
- m->bridge->mirrors[m->idx] = NULL;
- free(m);
+ size_t i;
- bridge_flush(br);
+ shash_init(shash);
+ for (i = 0; i < n; i++) {
+ shash_add(shash, keys[i], values[i]);
}
}
+/* Creates 'keys' and 'values' arrays from 'shash'.
+ *
+ * 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.
+ */
static void
-prune_ports(struct mirror *m, struct svec *ports)
+shash_to_ovs_idl_map(struct shash *shash,
+ char ***keys, char ***values, size_t *n)
{
- struct svec tmp;
- size_t i;
+ size_t i, count;
+ char **k, **v;
+ struct shash_node *sn;
- svec_sort_unique(ports);
+ count = shash_count(shash);
- svec_init(&tmp);
- for (i = 0; i < ports->n; i++) {
- const char *name = ports->names[i];
- if (port_lookup(m->bridge, name)) {
- svec_add(&tmp, name);
- } else {
- VLOG_WARN("mirror.%s.%s: cannot match on nonexistent port %s",
- m->bridge->name, m->name, name);
- }
+ 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++;
}
- svec_swap(ports, &tmp);
- svec_destroy(&tmp);
+
+ *n = count;
+ *keys = k;
+ *values = v;
}
-static size_t
-prune_vlans(struct mirror *m, struct svec *vlan_strings, int **vlans)
-{
- size_t n_vlans, i;
+struct iface_delete_queues_cbdata {
+ struct netdev *netdev;
+ const struct ovsdb_datum *queues;
+};
- /* This isn't perfect: it won't combine "0" and "00", and the textual sort
- * order won't give us numeric sort order. But that's good enough for what
- * we need right now. */
- svec_sort_unique(vlan_strings);
+static bool
+queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
+{
+ union ovsdb_atom atom;
- *vlans = xmalloc(sizeof *vlans * vlan_strings->n);
- n_vlans = 0;
- for (i = 0; i < vlan_strings->n; i++) {
- const char *name = vlan_strings->names[i];
- int vlan;
- if (!str_to_int(name, 10, &vlan) || vlan < 0 || vlan > 4095) {
- VLOG_WARN("mirror.%s.%s.select.vlan: ignoring invalid VLAN %s",
- m->bridge->name, m->name, name);
- } else {
- (*vlans)[n_vlans++] = vlan;
- }
- }
- return n_vlans;
+ atom.integer = target;
+ return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
}
-static bool
-vlan_is_mirrored(const struct mirror *m, int vlan)
+static void
+iface_delete_queues(unsigned int queue_id,
+ const struct shash *details OVS_UNUSED, void *cbdata_)
{
- size_t i;
+ struct iface_delete_queues_cbdata *cbdata = cbdata_;
- for (i = 0; i < m->n_vlans; i++) {
- if (m->vlans[i] == vlan) {
- return true;
- }
+ if (!queue_ids_include(cbdata->queues, queue_id)) {
+ netdev_delete_queue(cbdata->netdev, queue_id);
}
- return false;
}
-static bool
-port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
+static void
+iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
{
- size_t i;
+ 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;
+ size_t i;
- for (i = 0; i < m->n_vlans; i++) {
- if (port_trunks_vlan(p, m->vlans[i])) {
- return true;
+ /* 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);
+
+ /* Deconfigure queues that were deleted. */
+ cbdata.netdev = iface->netdev;
+ cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
+ OVSDB_TYPE_UUID);
+ netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
+
+ /* Configure queues for 'iface'. */
+ 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);
}
}
- return false;
+
+ netdev_set_policing(iface->netdev,
+ iface->cfg->ingress_policing_rate,
+ iface->cfg->ingress_policing_burst);
}
static void
-mirror_reconfigure_one(struct mirror *m)
-{
- char *pfx = xasprintf("mirror.%s.%s", m->bridge->name, m->name);
- struct svec src_ports, dst_ports, ports;
- struct svec vlan_strings;
- mirror_mask_t mirror_bit;
- const char *out_port_name;
- struct port *out_port;
- int out_vlan;
- size_t n_vlans;
- int *vlans;
- size_t i;
- bool mirror_all_ports;
-
- /* Get output port. */
- out_port_name = cfg_get_key(0, "mirror.%s.%s.output.port",
- m->bridge->name, m->name);
- if (out_port_name) {
- out_port = port_lookup(m->bridge, out_port_name);
- if (!out_port) {
- VLOG_ERR("%s.output.port: bridge %s does not have a port "
- "named %s", pfx, m->bridge->name, out_port_name);
- mirror_destroy(m);
- free(pfx);
- return;
- }
- out_vlan = -1;
+iface_configure_cfm(struct iface *iface)
+{
+ const struct ovsrec_interface *cfg = iface->cfg;
+ struct cfm_settings s;
+ uint16_t remote_mpid;
- if (cfg_has("%s.output.vlan", pfx)) {
- VLOG_ERR("%s.output.port and %s.output.vlan both specified; "
- "ignoring %s.output.vlan", pfx, pfx, pfx);
- }
- } else if (cfg_has("%s.output.vlan", pfx)) {
- out_port = NULL;
- out_vlan = cfg_get_vlan(0, "%s.output.vlan", pfx);
- } else {
- VLOG_ERR("%s: neither %s.output.port nor %s.output.vlan specified, "
- "but exactly one is required; disabling port mirror %s",
- pfx, pfx, pfx, pfx);
- mirror_destroy(m);
- free(pfx);
+ if (!cfg->n_cfm_mpid || !cfg->n_cfm_remote_mpid) {
+ ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
return;
}
- /* Get all the ports, and drop duplicates and ports that don't exist. */
- svec_init(&src_ports);
- svec_init(&dst_ports);
- svec_init(&ports);
- cfg_get_all_keys(&src_ports, "%s.select.src-port", pfx);
- cfg_get_all_keys(&dst_ports, "%s.select.dst-port", pfx);
- cfg_get_all_keys(&ports, "%s.select.port", pfx);
- svec_append(&src_ports, &ports);
- svec_append(&dst_ports, &ports);
- svec_destroy(&ports);
- prune_ports(m, &src_ports);
- prune_ports(m, &dst_ports);
-
- /* Get all the vlans, and drop duplicate and invalid vlans. */
- svec_init(&vlan_strings);
- cfg_get_all_keys(&vlan_strings, "%s.select.vlan", pfx);
- n_vlans = prune_vlans(m, &vlan_strings, &vlans);
- svec_destroy(&vlan_strings);
-
- /* Update mirror data. */
- if (!svec_equal(&m->src_ports, &src_ports)
- || !svec_equal(&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);
- }
- svec_swap(&m->src_ports, &src_ports);
- svec_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;
-
- /* If no selection criteria have been given, mirror for all ports. */
- mirror_all_ports = (!m->src_ports.n) && (!m->dst_ports.n) && (!m->n_vlans);
-
- /* 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 (mirror_all_ports
- || svec_contains(&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.mpid = *cfg->cfm_mpid;
+ remote_mpid = *cfg->cfm_remote_mpid;
+ s.remote_mpids = &remote_mpid;
+ s.n_remote_mpids = 1;
- if (mirror_all_ports || svec_contains(&m->dst_ports, port->name)) {
- port->dst_mirrors |= mirror_bit;
- } else {
- port->dst_mirrors &= ~mirror_bit;
- }
+ s.interval = atoi(get_interface_other_config(iface->cfg, "cfm_interval",
+ "0"));
+ if (s.interval <= 0) {
+ s.interval = 1000;
}
- /* Clean up. */
- svec_destroy(&src_ports);
- svec_destroy(&dst_ports);
- free(pfx);
+ ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
}
-
-/* Spanning tree protocol. */
-
-static void brstp_update_port_state(struct port *);
-static void
-brstp_send_bpdu(struct ofpbuf *pkt, int port_no, void *br_)
+/* Read carrier or miimon status directly from 'iface''s netdev, according to
+ * how 'iface''s port is configured.
+ *
+ * Returns true if 'iface' is up, false otherwise. */
+static bool
+iface_get_carrier(const struct iface *iface)
{
- struct bridge *br = br_;
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- struct iface *iface = iface_from_dp_ifidx(br, port_no);
- if (!iface) {
- VLOG_WARN_RL(&rl, "%s: cannot send BPDU on unknown port %d",
- br->name, port_no);
- } else if (eth_addr_is_zero(iface->mac)) {
- VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d with unknown MAC",
- br->name, port_no);
- } else {
- union ofp_action action;
- struct eth_header *eth = pkt->l2;
- flow_t flow;
+ /* XXX */
+ return netdev_get_carrier(iface->netdev);
+}
- memcpy(eth->eth_src, iface->mac, ETH_ADDR_LEN);
+/* 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_);
+}
+
+/* Port mirroring. */
- memset(&action, 0, sizeof action);
- action.type = htons(OFPAT_OUTPUT);
- action.output.len = htons(sizeof action);
- action.output.port = htons(port_no);
+static struct mirror *
+mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
+{
+ struct mirror *m;
- flow_extract(pkt, ODPP_NONE, &flow);
- ofproto_send_packet(br->ofproto, &flow, &action, 1, pkt);
+ HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
+ if (uuid_equals(uuid, &m->uuid)) {
+ return m;
+ }
}
- ofpbuf_delete(pkt);
+ return NULL;
}
static void
-brstp_reconfigure(struct bridge *br)
+bridge_configure_mirrors(struct bridge *br)
{
+ const struct ovsdb_datum *mc;
+ unsigned long *flood_vlans;
+ struct mirror *m, *next;
size_t i;
- if (!cfg_get_bool(0, "stp.%s.enabled", br->name)) {
- if (br->stp) {
- stp_destroy(br->stp);
- br->stp = NULL;
+ /* Get rid of deleted mirrors. */
+ mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
+ HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
+ union ovsdb_atom atom;
- bridge_flush(br);
- }
- } else {
- uint64_t bridge_address, bridge_id;
- int bridge_priority;
-
- bridge_address = cfg_get_mac(0, "stp.%s.address", br->name);
- if (!bridge_address) {
- if (br->stp) {
- bridge_address = (stp_get_bridge_id(br->stp)
- & ((UINT64_C(1) << 48) - 1));
- } else {
- uint8_t mac[ETH_ADDR_LEN];
- eth_addr_random(mac);
- bridge_address = eth_addr_to_uint64(mac);
- }
+ atom.uuid = m->uuid;
+ if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
+ mirror_destroy(m);
}
+ }
- if (cfg_is_valid(CFG_INT | CFG_REQUIRED, "stp.%s.priority",
- br->name)) {
- bridge_priority = cfg_get_int(0, "stp.%s.priority", br->name);
- } else {
- bridge_priority = STP_DEFAULT_BRIDGE_PRIORITY;
+ /* Add new mirrors and reconfigure existing ones. */
+ for (i = 0; i < br->cfg->n_mirrors; i++) {
+ const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
+ struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
+ if (!m) {
+ m = mirror_create(br, cfg);
}
-
- bridge_id = bridge_address | ((uint64_t) bridge_priority << 48);
- if (!br->stp) {
- br->stp = stp_create(br->name, bridge_id, brstp_send_bpdu, br);
- br->stp_last_tick = time_msec();
- bridge_flush(br);
- } else {
- if (bridge_id != stp_get_bridge_id(br->stp)) {
- stp_set_bridge_id(br->stp, bridge_id);
- bridge_flush(br);
- }
+ if (!mirror_configure(m, cfg)) {
+ mirror_destroy(m);
}
+ }
- for (i = 0; i < br->n_ports; i++) {
- struct port *p = br->ports[i];
- int dp_ifidx;
- struct stp_port *sp;
- int path_cost, priority;
- bool enable;
+ /* Update flooded vlans (for RSPAN). */
+ 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);
+}
- if (!p->n_ifaces) {
- continue;
- }
- dp_ifidx = p->ifaces[0]->dp_ifidx;
- if (dp_ifidx < 0 || dp_ifidx >= STP_MAX_PORTS) {
- continue;
- }
+static struct mirror *
+mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
+{
+ struct mirror *m;
- sp = stp_get_port(br->stp, dp_ifidx);
- enable = (!cfg_is_valid(CFG_BOOL | CFG_REQUIRED,
- "stp.%s.port.%s.enabled",
- br->name, p->name)
- || cfg_get_bool(0, "stp.%s.port.%s.enabled",
- br->name, p->name));
- if (p->is_mirror_output_port) {
- enable = false;
- }
- if (enable != (stp_port_get_state(sp) != STP_DISABLED)) {
- bridge_flush(br); /* Might not be necessary. */
- if (enable) {
- stp_port_enable(sp);
- } else {
- stp_port_disable(sp);
- }
- }
+ m = xzalloc(sizeof *m);
+ m->uuid = cfg->header_.uuid;
+ hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
+ m->bridge = br;
+ m->name = xstrdup(cfg->name);
+
+ return m;
+}
- path_cost = cfg_get_int(0, "stp.%s.port.%s.path-cost",
- br->name, p->name);
- stp_port_set_path_cost(sp, path_cost ? path_cost : 19 /* XXX */);
+static void
+mirror_destroy(struct mirror *m)
+{
+ if (m) {
+ struct bridge *br = m->bridge;
- priority = (cfg_is_valid(CFG_INT | CFG_REQUIRED,
- "stp.%s.port.%s.priority",
- br->name, p->name)
- ? cfg_get_int(0, "stp.%s.port.%s.priority",
- br->name, p->name)
- : STP_DEFAULT_PORT_PRIORITY);
- stp_port_set_priority(sp, priority);
+ if (br->ofproto) {
+ ofproto_mirror_unregister(br->ofproto, m);
}
- brstp_adjust_timers(br);
- }
- for (i = 0; i < br->n_ports; i++) {
- brstp_update_port_state(br->ports[i]);
+ hmap_remove(&br->mirrors, &m->hmap_node);
+ free(m->name);
+ free(m);
}
}
static void
-brstp_update_port_state(struct port *p)
+mirror_collect_ports(struct mirror *m,
+ struct ovsrec_port **in_ports, int n_in_ports,
+ void ***out_portsp, size_t *n_out_portsp)
{
- struct bridge *br = p->bridge;
- enum stp_state state;
-
- /* Figure out new state. */
- state = STP_DISABLED;
- if (br->stp && p->n_ifaces > 0) {
- int dp_ifidx = p->ifaces[0]->dp_ifidx;
- if (dp_ifidx >= 0 && dp_ifidx < STP_MAX_PORTS) {
- state = stp_port_get_state(stp_get_port(br->stp, dp_ifidx));
- }
- }
+ void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
+ size_t n_out_ports = 0;
+ size_t i;
- /* Update state. */
- if (p->stp_state != state) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
- VLOG_INFO_RL(&rl, "port %s: STP state changed from %s to %s",
- p->name, stp_state_name(p->stp_state),
- stp_state_name(state));
- if (p->stp_state == STP_DISABLED) {
- bridge_flush(br);
+ 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 {
- ofproto_revalidate(p->bridge->ofproto, p->stp_state_tag);
+ VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
+ "port %s", m->bridge->name, m->name, name);
}
- p->stp_state = state;
- p->stp_state_tag = (p->stp_state == STP_DISABLED ? 0
- : tag_create_random());
}
+ *out_portsp = out_ports;
+ *n_out_portsp = n_out_ports;
}
-static void
-brstp_adjust_timers(struct bridge *br)
+static bool
+mirror_configure(struct mirror *m, const struct ovsrec_mirror *cfg)
{
- int hello_time = cfg_get_int(0, "stp.%s.hello-time", br->name);
- int max_age = cfg_get_int(0, "stp.%s.max-age", br->name);
- int forward_delay = cfg_get_int(0, "stp.%s.forward-delay", br->name);
+ struct ofproto_mirror_settings s;
- stp_set_hello_time(br->stp, hello_time ? hello_time : 2000);
- stp_set_max_age(br->stp, max_age ? max_age : 20000);
- stp_set_forward_delay(br->stp, forward_delay ? forward_delay : 15000);
-}
+ /* Set name. */
+ if (strcmp(cfg->name, m->name)) {
+ free(m->name);
+ m->name = xstrdup(cfg->name);
+ }
+ s.name = m->name;
-static void
-brstp_run(struct bridge *br)
-{
- if (br->stp) {
- long long int now = time_msec();
- long long int elapsed = now - br->stp_last_tick;
- struct stp_port *sp;
+ /* Get output port or VLAN. */
+ if (cfg->output_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);
+ return false;
+ }
+ s.out_vlan = UINT16_MAX;
- if (elapsed > 0) {
- stp_tick(br->stp, MIN(INT_MAX, elapsed));
- br->stp_last_tick = now;
+ if (cfg->output_vlan) {
+ VLOG_ERR("bridge %s: mirror %s specifies both output port and "
+ "output vlan; ignoring output vlan",
+ m->bridge->name, m->name);
}
- while (stp_get_changed_port(br->stp, &sp)) {
- struct port *p = port_from_dp_ifidx(br, stp_port_no(sp));
- if (p) {
- brstp_update_port_state(p);
- }
+ } else if (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);
+ return false;
+ }
+
+ /* Get port selection. */
+ if (cfg->select_all) {
+ 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;
}
+
+ s.srcs = ports;
+ s.n_srcs = n_ports;
+
+ s.dsts = ports;
+ s.n_dsts = n_ports;
+ } else {
+ /* 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,
+ &s.srcs, &s.n_srcs);
+ mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
+ &s.dsts, &s.n_dsts);
+
}
-}
-static void
-brstp_wait(struct bridge *br)
-{
- if (br->stp) {
- poll_timer_wait(1000);
+ /* 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;
}