-/* Copyright (c) 2008, 2009, 2010 Nicira Networks
+/* Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "bridge.h"
+#include "byte-order.h"
#include <assert.h>
#include <errno.h>
#include <arpa/inet.h>
#include <sys/types.h>
#include <unistd.h>
#include "bitmap.h"
+#include "cfm.h"
+#include "classifier.h"
#include "coverage.h"
#include "dirs.h"
#include "dpif.h"
#include "dynamic-string.h"
#include "flow.h"
#include "hash.h"
+#include "hmap.h"
#include "jsonrpc.h"
#include "list.h"
#include "mac-learning.h"
#include "netdev.h"
+#include "netlink.h"
#include "odp-util.h"
#include "ofp-print.h"
#include "ofpbuf.h"
#include "ofproto/netflow.h"
#include "ofproto/ofproto.h"
+#include "ovsdb-data.h"
#include "packets.h"
#include "poll-loop.h"
-#include "port-array.h"
#include "proc-net-compat.h"
#include "process.h"
#include "sha1.h"
#include "socket-util.h"
#include "stream-ssl.h"
#include "svec.h"
+#include "system-stats.h"
#include "timeval.h"
#include "util.h"
#include "unixctl.h"
#include "vconn.h"
#include "vswitchd/vswitch-idl.h"
#include "xenserver.h"
-#include "xtoxll.h"
+#include "vlog.h"
#include "sflow_api.h"
-#define THIS_MODULE VLM_bridge
-#include "vlog.h"
+VLOG_DEFINE_THIS_MODULE(bridge);
+
+COVERAGE_DEFINE(bridge_flush);
+COVERAGE_DEFINE(bridge_process_flow);
+COVERAGE_DEFINE(bridge_reconfigure);
+
+enum lacp_status {
+ LACP_STATUS_CURRENT, /* Partner is up to date. */
+ LACP_STATUS_EXPIRED, /* Partner is out of date. Attempt to re-sync. */
+ LACP_STATUS_DEFAULTED /* Partner information is unknown. */
+};
struct dst {
uint16_t vlan;
uint16_t dp_ifidx;
};
+struct dst_set {
+ struct dst builtin[32];
+ struct dst *dsts;
+ size_t n, allocated;
+};
+
+static void dst_set_init(struct dst_set *);
+static void dst_set_add(struct dst_set *, const struct dst *);
+static void dst_set_free(struct dst_set *);
+
struct iface {
/* These members are always valid. */
struct port *port; /* Containing port. */
/* These members are valid only after bridge_reconfigure() causes them to
* be initialized. */
+ struct hmap_node dp_ifidx_node; /* In struct bridge's "ifaces" hmap. */
int dp_ifidx; /* Index within kernel datapath. */
struct netdev *netdev; /* Network device. */
bool enabled; /* May be chosen for flows? */
+ bool up; /* Is the interface up? */
+ const char *type; /* Usually same as cfg->type. */
+ struct cfm *cfm; /* Connectivity Fault Management */
const struct ovsrec_interface *cfg;
+
+ /* LACP information. */
+ enum lacp_status lacp_status; /* LACP state machine status. */
+ uint16_t lacp_priority; /* LACP port priority. */
+ struct lacp_info lacp_actor; /* LACP actor information. */
+ struct lacp_info lacp_partner; /* LACP partner information. */
+ long long int lacp_tx; /* Next LACP message transmission time. */
+ long long int lacp_rx; /* Next LACP message receive time. */
+ bool lacp_attached; /* Attached to its aggregator? LACP allows
+ this link to be chosen for flows. */
};
#define BOND_MASK 0xff
tag_type iface_tag; /* Tag associated with iface_idx. */
};
+enum bond_mode {
+ BM_TCP, /* Transport Layer Load Balance. */
+ BM_SLB, /* Source Load Balance. */
+ BM_AB /* Active Backup. */
+};
+
#define MAX_MIRRORS 32
typedef uint32_t mirror_mask_t;
#define MIRROR_MASK_C(X) UINT32_C(X)
struct bridge *bridge;
size_t idx;
char *name;
+ struct uuid uuid; /* UUID of this "mirror" record in database. */
/* Selection criteria. */
struct shash src_ports; /* Name is port name; data is always NULL. */
int out_vlan;
};
+/* Flags for a port's lacp member. */
+#define LACP_ACTIVE 0x01 /* LACP is in active mode. */
+#define LACP_PASSIVE 0x02 /* LACP is in passive mode. */
+#define LACP_NEGOTIATED 0x04 /* LACP has successfully negotiated. */
+
#define FLOOD_PORT ((struct port *) 1) /* The 'flood' output port. */
struct port {
struct bridge *bridge;
size_t n_ifaces, allocated_ifaces;
/* Bonding info. */
- struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */
+ enum bond_mode bond_mode; /* Type of the bond. BM_SLB is the default. */
int active_iface; /* Ifidx on which bcasts accepted, or -1. */
tag_type active_iface_tag; /* Tag for bcast flows. */
tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */
int updelay, downdelay; /* Delay before iface goes up/down, in ms. */
bool bond_compat_is_stale; /* Need to call port_update_bond_compat()? */
bool bond_fake_iface; /* Fake a bond interface for legacy compat? */
- long bond_next_fake_iface_update; /* Next update to fake bond stats. */
+ bool miimon; /* Use miimon instead of carrier? */
+ long long int bond_miimon_interval; /* Miimon status refresh interval. */
+ long long int bond_miimon_next_update; /* Time of next miimon update. */
+ long long int bond_next_fake_iface_update; /* Time of next update. */
+ struct netdev_monitor *monitor; /* Tracks carrier up/down status. */
+
+ /* LACP information. */
+ int lacp; /* LACP status flags. 0 if LACP is off. */
+ uint16_t lacp_key; /* LACP aggregation key. */
+ uint16_t lacp_priority; /* LACP system priority. */
+ bool lacp_need_update; /* Need to update attached interfaces? */
+
+ /* SLB specific bonding info. */
+ struct bond_entry *bond_hash; /* An array of (BOND_MASK + 1) elements. */
int bond_rebalance_interval; /* Interval between rebalances, in ms. */
long long int bond_next_rebalance; /* Next rebalancing time. */
bool is_mirror_output_port; /* Does port mirroring send frames here? */
};
-#define DP_MAX_PORTS 255
struct bridge {
struct list node; /* Node in global list of bridges. */
char *name; /* User-specified arbitrary name. */
struct mac_learning *ml; /* MAC learning table. */
+ uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
const struct ovsrec_bridge *cfg;
/* Kernel datapath information. */
struct dpif *dpif; /* Datapath. */
- struct port_array ifaces; /* Indexed by kernel datapath port number. */
+ struct hmap ifaces; /* Contains "struct iface"s. */
/* Bridge ports. */
struct port **ports;
/* OVSDB IDL used to obtain configuration. */
static struct ovsdb_idl *idl;
+/* Each time this timer expires, the bridge fetches systems and interface
+ * statistics and pushes them into the database. */
+#define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
+static long long int stats_timer = LLONG_MIN;
+
static struct bridge *bridge_create(const struct ovsrec_bridge *br_cfg);
static void bridge_destroy(struct bridge *);
static struct bridge *bridge_lookup(const char *name);
static unixctl_cb_func bridge_unixctl_dump_flows;
+static unixctl_cb_func bridge_unixctl_reconnect;
static int bridge_run_one(struct bridge *);
-static size_t bridge_get_controllers(const struct ovsrec_open_vswitch *ovs_cfg,
- const struct bridge *br,
+static size_t bridge_get_controllers(const struct bridge *br,
struct ovsrec_controller ***controllersp);
-static void bridge_reconfigure_one(const struct ovsrec_open_vswitch *,
- struct bridge *);
-static void bridge_reconfigure_remotes(const struct ovsrec_open_vswitch *,
- struct bridge *,
+static void bridge_reconfigure_one(struct bridge *);
+static void bridge_reconfigure_remotes(struct bridge *,
const struct sockaddr_in *managers,
size_t n_managers);
static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
static unixctl_cb_func bridge_unixctl_fdb_show;
+static void lacp_run(struct bridge *);
+static void lacp_wait(struct bridge *);
+static void lacp_process_packet(const struct ofpbuf *, struct iface *);
+
static void bond_init(void);
static void bond_run(struct bridge *);
static void bond_wait(struct bridge *);
static void port_update_bond_compat(struct port *);
static void port_update_vlan_compat(struct port *);
static void port_update_bonding(struct port *);
+static void port_update_lacp(struct port *);
-static struct mirror *mirror_create(struct bridge *, const char *name);
+static void mirror_create(struct bridge *, struct ovsrec_mirror *);
static void mirror_destroy(struct mirror *);
static void mirror_reconfigure(struct bridge *);
static void mirror_reconfigure_one(struct mirror *, struct ovsrec_mirror *);
static bool vlan_is_mirrored(const struct mirror *, int vlan);
-static struct iface *iface_create(struct port *port,
+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);
-static bool iface_is_internal(const struct bridge *, const char *name);
static void iface_set_mac(struct iface *);
+static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
static void iface_update_qos(struct iface *, const struct ovsrec_qos *);
+static void iface_update_cfm(struct iface *);
+static void iface_refresh_cfm_stats(struct iface *iface);
+static void iface_send_packet(struct iface *, struct ofpbuf *packet);
+static uint8_t iface_get_lacp_state(const struct iface *);
+static void iface_get_lacp_priority(struct iface *, struct lacp_info *);
+static void iface_set_lacp_defaulted(struct iface *);
+static void iface_set_lacp_expired(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);
+
/* Hooks into ofproto processing. */
static struct ofhooks bridge_ofhooks;
bridge_init(const char *remote)
{
/* Create connection to database. */
- idl = ovsdb_idl_create(remote, &ovsrec_idl_class);
+ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
+
+ 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_bridge_col_external_ids);
+
+ ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
+ ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
+
+ ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
+ ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
+ ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
/* Register unixctl commands. */
unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
NULL);
+ unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
+ NULL);
bond_init();
}
+void
+bridge_exit(void)
+{
+ struct bridge *br, *next_br;
+
+ LIST_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
+ bridge_destroy(br);
+ }
+ ovsdb_idl_destroy(idl);
+}
+
/* Performs configuration that is only necessary once at ovs-vswitchd startup,
* but for which the ovs-vswitchd configuration 'cfg' is required. */
static void
}
already_configured_once = true;
+ stats_timer = time_msec() + STATS_INTERVAL;
+
/* Get all the configured bridges' names from 'cfg' into 'bridge_names'. */
svec_init(&bridge_names);
for (i = 0; i < cfg->n_bridges; i++) {
svec_init(&dpif_types);
dp_enumerate_types(&dpif_types);
for (i = 0; i < dpif_types.n; i++) {
- struct dpif *dpif;
- int retval;
size_t j;
dp_enumerate_names(dpif_types.names[i], &dpif_names);
- /* For each dpif... */
+ /* Delete each dpif whose name is not in 'bridge_names'. */
for (j = 0; j < dpif_names.n; j++) {
- retval = dpif_open(dpif_names.names[j], dpif_types.names[i], &dpif);
- if (!retval) {
- struct svec all_names;
- size_t k;
-
- /* ...check whether any of its names is in 'bridge_names'. */
- svec_init(&all_names);
- dpif_get_all_names(dpif, &all_names);
- for (k = 0; k < all_names.n; k++) {
- if (svec_contains(&bridge_names, all_names.names[k])) {
- goto found;
- }
+ if (!svec_contains(&bridge_names, dpif_names.names[j])) {
+ struct dpif *dpif;
+ int retval;
+
+ retval = dpif_open(dpif_names.names[j], dpif_types.names[i],
+ &dpif);
+ if (!retval) {
+ dpif_delete(dpif);
+ dpif_close(dpif);
}
-
- /* No. Delete the dpif. */
- dpif_delete(dpif);
-
- found:
- svec_destroy(&all_names);
- dpif_close(dpif);
}
}
}
svec_destroy(&dpif_types);
}
-#ifdef HAVE_OPENSSL
-static void
-bridge_configure_ssl(const struct ovsrec_ssl *ssl)
-{
- /* XXX SSL should be configurable on a per-bridge basis. */
- if (ssl) {
- stream_ssl_set_private_key_file(ssl->private_key);
- stream_ssl_set_certificate_file(ssl->certificate);
- stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
- }
-}
-#endif
-
-/* Attempt to create the network device 'iface_name' through the netdev
- * library. */
-static int
-set_up_iface(const struct ovsrec_interface *iface_cfg, struct iface *iface,
- bool create)
-{
- struct shash options;
- int error = 0;
- size_t i;
-
- shash_init(&options);
- for (i = 0; i < iface_cfg->n_options; i++) {
- shash_add(&options, iface_cfg->key_options[i],
- xstrdup(iface_cfg->value_options[i]));
- }
-
- if (create) {
- struct netdev_options netdev_options;
-
- memset(&netdev_options, 0, sizeof netdev_options);
- netdev_options.name = iface_cfg->name;
- if (!strcmp(iface_cfg->type, "internal")) {
- /* An "internal" config type maps to a netdev "system" type. */
- netdev_options.type = "system";
- } else {
- netdev_options.type = iface_cfg->type;
- }
- netdev_options.args = &options;
- netdev_options.ethertype = NETDEV_ETH_TYPE_NONE;
-
- error = netdev_open(&netdev_options, &iface->netdev);
-
- if (iface->netdev) {
- netdev_get_carrier(iface->netdev, &iface->enabled);
- }
- } else if (iface->netdev) {
- const char *netdev_type = netdev_get_type(iface->netdev);
- const char *iface_type = iface_cfg->type && strlen(iface_cfg->type)
- ? iface_cfg->type : NULL;
-
- /* An "internal" config type maps to a netdev "system" type. */
- if (iface_type && !strcmp(iface_type, "internal")) {
- iface_type = "system";
- }
-
- if (!iface_type || !strcmp(netdev_type, iface_type)) {
- error = netdev_reconfigure(iface->netdev, &options);
- } else {
- VLOG_WARN("%s: attempting change device type from %s to %s",
- iface_cfg->name, netdev_type, iface_type);
- error = EINVAL;
- }
- }
- shash_destroy_free_data(&options);
-
- return error;
-}
-
-static int
-reconfigure_iface(const struct ovsrec_interface *iface_cfg, struct iface *iface)
-{
- return set_up_iface(iface_cfg, iface, false);
-}
-
+/* Callback for iterate_and_prune_ifaces(). */
static bool
-check_iface_netdev(struct bridge *br OVS_UNUSED, struct iface *iface,
- void *aux OVS_UNUSED)
+check_iface(struct bridge *br, struct iface *iface, void *aux OVS_UNUSED)
{
if (!iface->netdev) {
- int error = set_up_iface(iface->cfg, iface, true);
- if (error) {
- VLOG_WARN("could not open netdev on %s, dropping: %s", iface->name,
- strerror(error));
- return false;
- }
+ /* We already reported a related error, don't bother duplicating it. */
+ return false;
}
- return true;
-}
-
-static bool
-check_iface_dp_ifidx(struct bridge *br, struct iface *iface,
- void *aux OVS_UNUSED)
-{
- if (iface->dp_ifidx >= 0) {
- VLOG_DBG("%s has interface %s on port %d",
- dpif_name(br->dpif),
- iface->name, iface->dp_ifidx);
- return true;
- } else {
+ if (iface->dp_ifidx < 0) {
VLOG_ERR("%s interface not in %s, dropping",
iface->name, dpif_name(br->dpif));
return false;
}
+
+ VLOG_DBG("%s has interface %s on port %d", dpif_name(br->dpif),
+ iface->name, iface->dp_ifidx);
+ return true;
}
+/* Callback for iterate_and_prune_ifaces(). */
static bool
set_iface_properties(struct bridge *br OVS_UNUSED, struct iface *iface,
void *aux OVS_UNUSED)
/* Set MAC address of internal interfaces other than the local
* interface. */
- if (iface->dp_ifidx != ODPP_LOCAL
- && iface_is_internal(br, iface->name)) {
+ if (iface->dp_ifidx != ODPP_LOCAL && !strcmp(iface->type, "internal")) {
iface_set_mac(iface);
}
if (cb(br, iface, aux)) {
j++;
} else {
+ iface_set_ofport(iface->cfg, -1);
iface_destroy(iface);
}
}
* 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
-collect_managers(const struct ovsrec_open_vswitch *ovs_cfg,
- struct sockaddr_in **managersp, size_t *n_managersp)
+collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
+ struct sockaddr_in **managersp, size_t *n_managersp)
{
struct sockaddr_in *managers = NULL;
size_t n_managers = 0;
+ struct shash targets;
+ size_t i;
- if (ovs_cfg->n_managers > 0) {
- size_t i;
+ /* Collect all of the potential targets, as the union of the "managers"
+ * column and the "targets" columns of the rows pointed to by
+ * "manager_options", excluding any that are out-of-band. */
+ shash_init(&targets);
+ for (i = 0; i < ovs_cfg->n_managers; i++) {
+ shash_add_once(&targets, ovs_cfg->managers[i], NULL);
+ }
+ for (i = 0; i < ovs_cfg->n_manager_options; i++) {
+ struct ovsrec_manager *m = ovs_cfg->manager_options[i];
+
+ if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
+ shash_find_and_delete(&targets, m->target);
+ } else {
+ shash_add_once(&targets, m->target, NULL);
+ }
+ }
- managers = xmalloc(ovs_cfg->n_managers * sizeof *managers);
- for (i = 0; i < ovs_cfg->n_managers; i++) {
- const char *name = ovs_cfg->managers[i];
- struct sockaddr_in *sin = &managers[i];
+ /* Now extract the targets' IP addresses. */
+ if (!shash_is_empty(&targets)) {
+ struct shash_node *node;
+
+ managers = xmalloc(shash_count(&targets) * sizeof *managers);
+ SHASH_FOR_EACH (node, &targets) {
+ const char *target = node->name;
+ struct sockaddr_in *sin = &managers[n_managers];
- if ((!strncmp(name, "tcp:", 4)
- && inet_parse_active(name + 4, JSONRPC_TCP_PORT, sin)) ||
- (!strncmp(name, "ssl:", 4)
- && inet_parse_active(name + 4, JSONRPC_SSL_PORT, sin))) {
+ 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++;
}
}
}
+ shash_destroy(&targets);
*managersp = managers;
*n_managersp = n_managers;
COVERAGE_INC(bridge_reconfigure);
- collect_managers(ovs_cfg, &managers, &n_managers);
+ collect_in_band_managers(ovs_cfg, &managers, &n_managers);
/* Collect old and new bridges. */
shash_init(&old_br);
shash_init(&new_br);
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
shash_add(&old_br, br->name, br);
}
for (i = 0; i < ovs_cfg->n_bridges; i++) {
}
/* Get rid of deleted bridges and add new bridges. */
- LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH_SAFE (br, next, node, &all_bridges) {
struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
if (br_cfg) {
br->cfg = br_cfg;
shash_destroy(&old_br);
shash_destroy(&new_br);
-#ifdef HAVE_OPENSSL
- /* Configure SSL. */
- bridge_configure_ssl(ovs_cfg->ssl);
-#endif
-
/* Reconfigure all bridges. */
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- bridge_reconfigure_one(ovs_cfg, br);
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ bridge_reconfigure_one(br);
}
/* 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;
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ struct dpif_port_dump dump;
struct shash want_ifaces;
+ struct dpif_port dpif_port;
- 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 (!shash_find(&want_ifaces, p->devname)
- && strcmp(p->devname, br->name)) {
- int retval = dpif_port_del(br->dpif, p->port);
+ DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
+ if (!shash_find(&want_ifaces, dpif_port.name)
+ && strcmp(dpif_port.name, br->name)) {
+ int retval = dpif_port_del(br->dpif, dpif_port.port_no);
if (retval) {
VLOG_ERR("failed to remove %s interface from %s: %s",
- p->devname, dpif_name(br->dpif),
+ dpif_port.name, dpif_name(br->dpif),
strerror(retval));
}
}
}
shash_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;
+ LIST_FOR_EACH (br, node, &all_bridges) {
struct shash cur_ifaces, want_ifaces;
- struct shash_node *node;
+ struct dpif_port_dump dump;
+ struct dpif_port dpif_port;
/* Get the set of interfaces currently in this datapath. */
- dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports);
shash_init(&cur_ifaces);
- for (i = 0; i < n_dpif_ports; i++) {
- const char *name = dpif_ports[i].devname;
- if (!shash_find(&cur_ifaces, name)) {
- shash_add(&cur_ifaces, name, NULL);
- }
+ DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
+ struct dpif_port *port_info = xmalloc(sizeof *port_info);
+ dpif_port_clone(port_info, &dpif_port);
+ shash_add(&cur_ifaces, dpif_port.name, port_info);
}
- free(dpif_ports);
/* Get the set of interfaces we want on this datapath. */
bridge_get_all_ifaces(br, &want_ifaces);
+ hmap_clear(&br->ifaces);
SHASH_FOR_EACH (node, &want_ifaces) {
const char *if_name = node->name;
struct iface *iface = node->data;
+ struct dpif_port *dpif_port;
+ const char *type;
+ int error;
+
+ type = iface ? iface->type : "internal";
+ dpif_port = shash_find_data(&cur_ifaces, if_name);
+
+ /* If we have a port or a netdev already, and it's not the type we
+ * want, then delete the port (if any) and close the netdev (if
+ * any). */
+ if ((dpif_port && strcmp(dpif_port->type, type))
+ || (iface && iface->netdev
+ && strcmp(type, netdev_get_type(iface->netdev)))) {
+ if (dpif_port) {
+ error = ofproto_port_del(br->ofproto, dpif_port->port_no);
+ if (error) {
+ continue;
+ }
+ dpif_port = NULL;
+ }
+ if (iface) {
+ netdev_close(iface->netdev);
+ iface->netdev = NULL;
+ }
+ }
- if (shash_find(&cur_ifaces, if_name)) {
- /* Already exists, just reconfigure it. */
+ /* If the port doesn't exist or we don't have the netdev open,
+ * we need to do more work. */
+ if (!dpif_port || (iface && !iface->netdev)) {
+ struct netdev_options options;
+ struct netdev *netdev;
+ struct shash args;
+
+ /* First open the network device. */
+ options.name = if_name;
+ options.type = type;
+ options.args = &args;
+ options.ethertype = NETDEV_ETH_TYPE_NONE;
+
+ shash_init(&args);
if (iface) {
- reconfigure_iface(iface->cfg, iface);
+ shash_from_ovs_idl_map(iface->cfg->key_options,
+ iface->cfg->value_options,
+ iface->cfg->n_options, &args);
}
- } else {
- /* Need to add to datapath. */
- bool internal;
- int error;
-
- /* Add to datapath. */
- internal = iface_is_internal(br, if_name);
- error = dpif_port_add(br->dpif, if_name,
- internal ? ODP_PORT_INTERNAL : 0, NULL);
- if (error == EFBIG) {
- VLOG_ERR("ran out of valid port numbers on %s",
- dpif_name(br->dpif));
- break;
- } else if (error) {
- VLOG_ERR("failed to add %s interface to %s: %s",
- if_name, dpif_name(br->dpif), strerror(error));
+ error = netdev_open(&options, &netdev);
+ shash_destroy(&args);
+
+ if (error) {
+ VLOG_WARN("could not open network device %s (%s)",
+ if_name, strerror(error));
+ continue;
+ }
+
+ /* Then add the port if we haven't already. */
+ if (!dpif_port) {
+ error = dpif_port_add(br->dpif, netdev, NULL);
+ if (error) {
+ netdev_close(netdev);
+ if (error == EFBIG) {
+ VLOG_ERR("ran out of valid port numbers on %s",
+ dpif_name(br->dpif));
+ break;
+ } else {
+ VLOG_ERR("failed to add %s interface to %s: %s",
+ if_name, dpif_name(br->dpif),
+ strerror(error));
+ continue;
+ }
+ }
+ }
+
+ /* Update 'iface'. */
+ if (iface) {
+ iface->netdev = netdev;
+ iface->enabled = netdev_get_carrier(iface->netdev);
+ iface->up = iface->enabled;
}
+ } else if (iface && iface->netdev) {
+ struct shash args;
+
+ shash_init(&args);
+ shash_from_ovs_idl_map(iface->cfg->key_options,
+ iface->cfg->value_options,
+ iface->cfg->n_options, &args);
+ netdev_set_config(iface->netdev, &args);
+ shash_destroy(&args);
}
}
- shash_destroy(&cur_ifaces);
shash_destroy(&want_ifaces);
+
+ SHASH_FOR_EACH (node, &cur_ifaces) {
+ struct dpif_port *port_info = node->data;
+ dpif_port_destroy(port_info);
+ free(port_info);
+ }
+ shash_destroy(&cur_ifaces);
}
sflow_bridge_number = 0;
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
uint8_t ea[8];
uint64_t dpid;
struct iface *local_iface;
bridge_fetch_dp_ifaces(br);
- iterate_and_prune_ifaces(br, check_iface_netdev, NULL);
- iterate_and_prune_ifaces(br, check_iface_dp_ifidx, NULL);
+ iterate_and_prune_ifaces(br, check_iface, NULL);
/* Pick local port hardware address, datapath ID. */
bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
br->name, strerror(error));
}
}
+ memcpy(br->ea, ea, ETH_ADDR_LEN);
dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
ofproto_set_datapath_id(br->ofproto, dpid);
opts.collectors.n = nf_cfg->n_targets;
opts.collectors.names = nf_cfg->targets;
if (ofproto_set_netflow(br->ofproto, &opts)) {
- VLOG_ERR("bridge %s: problem setting netflow collectors",
+ VLOG_ERR("bridge %s: problem setting netflow collectors",
br->name);
}
} else {
struct ovsrec_controller **controllers;
struct ofproto_sflow_options oso;
size_t n_controllers;
- size_t i;
memset(&oso, 0, sizeof oso);
oso.agent_device = sflow_cfg->agent;
oso.control_ip = NULL;
- n_controllers = bridge_get_controllers(ovs_cfg, br, &controllers);
+ 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;
* yet; when a controller is configured, resetting the datapath ID will
* immediately disconnect from the controller, so it's better to set
* the datapath ID before the controller. */
- bridge_reconfigure_remotes(ovs_cfg, br, managers, n_managers);
+ bridge_reconfigure_remotes(br, managers, n_managers);
}
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
int j;
port_update_vlan_compat(port);
port_update_bonding(port);
+ port_update_lacp(port);
for (j = 0; j < port->n_ifaces; j++) {
iface_update_qos(port->ifaces[j], port->cfg->qos);
}
}
}
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
iterate_and_prune_ifaces(br, set_iface_properties, NULL);
}
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ struct iface *iface;
+ HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
+ iface_update_cfm(iface);
+ }
+ }
+
free(managers);
}
static const char *
-get_ovsrec_key_value(const char *key, char **keys, char **values, size_t n)
+get_ovsrec_key_value(const struct ovsdb_idl_row *row,
+ const struct ovsdb_idl_column *column,
+ const char *key)
{
- size_t i;
-
- for (i = 0; i < n; i++) {
- if (!strcmp(keys[i], key)) {
- return values[i];
- }
- }
- return NULL;
+ const struct ovsdb_datum *datum;
+ union ovsdb_atom atom;
+ unsigned int idx;
+
+ datum = ovsdb_idl_get(row, column, OVSDB_TYPE_STRING, OVSDB_TYPE_STRING);
+ atom.string = (char *) key;
+ idx = ovsdb_datum_find_key(datum, &atom, OVSDB_TYPE_STRING);
+ return idx == UINT_MAX ? NULL : datum->values[idx].string;
}
static const char *
bridge_get_other_config(const struct ovsrec_bridge *br_cfg, const char *key)
{
- return get_ovsrec_key_value(key,
- br_cfg->key_other_config,
- br_cfg->value_other_config,
- br_cfg->n_other_config);
+ return get_ovsrec_key_value(&br_cfg->header_,
+ &ovsrec_bridge_col_other_config, key);
}
static void
!eth_addr_is_local(iface_ea) &&
!eth_addr_is_reserved(iface_ea) &&
!eth_addr_is_zero(iface_ea) &&
- memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0)
+ eth_addr_compare_3way(iface_ea, ea) < 0)
{
memcpy(ea, iface_ea, ETH_ADDR_LEN);
*hw_addr_iface = iface;
return eth_addr_to_uint64(hash);
}
+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;
+
+ shash_init(&sh);
+
+ 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 {
+ ovsrec_interface_set_status(iface->cfg, NULL, NULL, 0);
+ }
+
+ shash_destroy_free_data(&sh);
+
+ error = netdev_get_flags(iface->netdev, &flags);
+ if (!error) {
+ ovsrec_interface_set_admin_state(iface->cfg, flags & NETDEV_UP ? "up" : "down");
+ }
+ else {
+ ovsrec_interface_set_admin_state(iface->cfg, NULL);
+ }
+
+ 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);
+ }
+
+
+ ovsrec_interface_set_link_state(iface->cfg,
+ netdev_get_carrier(iface->netdev)
+ ? "up" : "down");
+
+ error = netdev_get_mtu(iface->netdev, &mtu);
+ if (!error && mtu != INT_MAX) {
+ mtu_64 = mtu;
+ ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
+ }
+ else {
+ ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
+ }
+}
+
+static void
+iface_refresh_cfm_stats(struct iface *iface)
+{
+ size_t i;
+ struct cfm *cfm;
+ const struct ovsrec_monitor *mon;
+
+ mon = iface->cfg->monitor;
+ cfm = iface->cfm;
+
+ if (!cfm || !mon) {
+ return;
+ }
+
+ for (i = 0; i < mon->n_remote_mps; i++) {
+ const struct ovsrec_maintenance_point *mp;
+ const struct remote_mp *rmp;
+
+ mp = mon->remote_mps[i];
+ rmp = cfm_get_remote_mp(cfm, mp->mpid);
+
+ ovsrec_maintenance_point_set_fault(mp, &rmp->fault, 1);
+ }
+
+ if (hmap_is_empty(&cfm->x_remote_mps)) {
+ ovsrec_monitor_set_unexpected_remote_mpids(mon, NULL, 0);
+ } else {
+ size_t length;
+ struct remote_mp *rmp;
+ int64_t *x_remote_mps;
+
+ length = hmap_count(&cfm->x_remote_mps);
+ x_remote_mps = xzalloc(length * sizeof *x_remote_mps);
+
+ i = 0;
+ HMAP_FOR_EACH (rmp, node, &cfm->x_remote_mps) {
+ x_remote_mps[i++] = rmp->mpid;
+ }
+
+ ovsrec_monitor_set_unexpected_remote_mpids(mon, x_remote_mps, length);
+ free(x_remote_mps);
+ }
+
+ if (hmap_is_empty(&cfm->x_remote_maids)) {
+ ovsrec_monitor_set_unexpected_remote_maids(mon, NULL, 0);
+ } else {
+ size_t length;
+ char **x_remote_maids;
+ struct remote_maid *rmaid;
+
+ length = hmap_count(&cfm->x_remote_maids);
+ x_remote_maids = xzalloc(length * sizeof *x_remote_maids);
+
+ i = 0;
+ HMAP_FOR_EACH (rmaid, node, &cfm->x_remote_maids) {
+ size_t j;
+
+ x_remote_maids[i] = xzalloc(CCM_MAID_LEN * 2 + 1);
+
+ for (j = 0; j < CCM_MAID_LEN; j++) {
+ snprintf(&x_remote_maids[i][j * 2], 3, "%02hhx",
+ rmaid->maid[j]);
+ }
+ i++;
+ }
+ ovsrec_monitor_set_unexpected_remote_maids(mon, x_remote_maids, length);
+
+ for (i = 0; i < length; i++) {
+ free(x_remote_maids[i]);
+ }
+ free(x_remote_maids);
+ }
+
+ ovsrec_monitor_set_fault(mon, &cfm->fault, 1);
+}
+
+static void
+iface_refresh_stats(struct iface *iface)
+{
+ struct iface_stat {
+ char *name;
+ int offset;
+ };
+ static const struct iface_stat iface_stats[] = {
+ { "rx_packets", offsetof(struct netdev_stats, rx_packets) },
+ { "tx_packets", offsetof(struct netdev_stats, tx_packets) },
+ { "rx_bytes", offsetof(struct netdev_stats, rx_bytes) },
+ { "tx_bytes", offsetof(struct netdev_stats, tx_bytes) },
+ { "rx_dropped", offsetof(struct netdev_stats, rx_dropped) },
+ { "tx_dropped", offsetof(struct netdev_stats, tx_dropped) },
+ { "rx_errors", offsetof(struct netdev_stats, rx_errors) },
+ { "tx_errors", offsetof(struct netdev_stats, tx_errors) },
+ { "rx_frame_err", offsetof(struct netdev_stats, rx_frame_errors) },
+ { "rx_over_err", offsetof(struct netdev_stats, rx_over_errors) },
+ { "rx_crc_err", offsetof(struct netdev_stats, rx_crc_errors) },
+ { "collisions", offsetof(struct netdev_stats, collisions) },
+ };
+ enum { N_STATS = ARRAY_SIZE(iface_stats) };
+ const struct iface_stat *s;
+
+ char *keys[N_STATS];
+ int64_t values[N_STATS];
+ int n;
+
+ struct netdev_stats stats;
+
+ /* 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);
+
+ n = 0;
+ for (s = iface_stats; s < &iface_stats[N_STATS]; s++) {
+ uint64_t value = *(uint64_t *) (((char *) &stats) + s->offset);
+ if (value != UINT64_MAX) {
+ keys[n] = s->name;
+ values[n] = value;
+ n++;
+ }
+ }
+
+ ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
+}
+
+static void
+refresh_system_stats(const struct ovsrec_open_vswitch *cfg)
+{
+ 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 inline const char *
+nx_role_to_str(enum nx_role role)
+{
+ switch (role) {
+ case NX_ROLE_OTHER:
+ return "other";
+ case NX_ROLE_MASTER:
+ return "master";
+ case NX_ROLE_SLAVE:
+ return "slave";
+ default:
+ return "*** INVALID ROLE ***";
+ }
+}
+
+static void
+bridge_refresh_controller_status(const struct bridge *br)
+{
+ struct shash info;
+ const struct ovsrec_controller *cfg;
+
+ 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);
+ }
+ }
+
+ ofproto_free_ofproto_controller_info(&info);
+}
+
void
bridge_run(void)
{
+ const struct ovsrec_open_vswitch *cfg;
+
bool datapath_destroyed;
+ bool database_changed;
struct bridge *br;
/* Let each bridge do the work that it needs to do. */
datapath_destroyed = false;
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
int error = bridge_run_one(br);
if (error) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
}
/* (Re)configure if necessary. */
- if (ovsdb_idl_run(idl) || datapath_destroyed) {
- const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
+ database_changed = ovsdb_idl_run(idl);
+ cfg = ovsrec_open_vswitch_first(idl);
+#ifdef HAVE_OPENSSL
+ /* Re-configure SSL. We do this on every trip through the main loop,
+ * instead of just when the database changes, because the contents of the
+ * key and certificate files can change without the database changing.
+ *
+ * We do this before bridge_reconfigure() because that function might
+ * initiate SSL connections and thus requires SSL to be configured. */
+ if (cfg && cfg->ssl) {
+ const struct ovsrec_ssl *ssl = cfg->ssl;
+
+ stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
+ stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
+ }
+#endif
+ if (database_changed || datapath_destroyed) {
if (cfg) {
struct ovsdb_idl_txn *txn = ovsdb_idl_txn_create(idl);
bridge_reconfigure(&null_cfg);
}
}
+
+ /* Refresh system and interface stats if necessary. */
+ if (time_msec() >= stats_timer) {
+ if (cfg) {
+ struct ovsdb_idl_txn *txn;
+
+ txn = ovsdb_idl_txn_create(idl);
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ size_t i;
+
+ for (i = 0; i < br->n_ports; i++) {
+ struct port *port = br->ports[i];
+ size_t j;
+
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *iface = port->ifaces[j];
+ iface_refresh_stats(iface);
+ iface_refresh_cfm_stats(iface);
+ iface_refresh_status(iface);
+ }
+ }
+ bridge_refresh_controller_status(br);
+ }
+ refresh_system_stats(cfg);
+ ovsdb_idl_txn_commit(txn);
+ ovsdb_idl_txn_destroy(txn); /* XXX */
+ }
+
+ stats_timer = time_msec() + STATS_INTERVAL;
+ }
}
void
bridge_wait(void)
{
struct bridge *br;
+ struct iface *iface;
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
ofproto_wait(br->ofproto);
- if (ofproto_has_controller(br->ofproto)) {
+ if (ofproto_has_primary_controller(br->ofproto)) {
continue;
}
mac_learning_wait(br->ml);
+ lacp_wait(br);
bond_wait(br);
+
+ HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
+ if (iface->cfm) {
+ cfm_wait(iface->cfm);
+ }
+ }
}
ovsdb_idl_wait(idl);
+ poll_timer_wait_until(stats_timer);
}
/* Forces 'br' to revalidate all of its flows. This is appropriate when 'br''s
}
ds_put_cstr(&ds, " port VLAN MAC Age\n");
- LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) {
+ LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
if (e->port < 0 || e->port >= br->n_ports) {
continue;
}
br->ml = mac_learning_create();
eth_addr_nicira_random(br->default_ea);
- port_array_init(&br->ifaces);
+ hmap_init(&br->ifaces);
shash_init(&br->port_by_name);
shash_init(&br->iface_by_name);
dpif_close(br->dpif);
ofproto_destroy(br->ofproto);
mac_learning_destroy(br->ml);
- port_array_destroy(&br->ifaces);
+ hmap_destroy(&br->ifaces);
shash_destroy(&br->port_by_name);
shash_destroy(&br->iface_by_name);
free(br->ports);
{
struct bridge *br;
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
if (!strcmp(br->name, name)) {
return br;
}
{
struct bridge *br;
struct ds results;
-
+
br = bridge_lookup(args);
if (!br) {
unixctl_command_reply(conn, 501, "Unknown bridge");
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 {
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ ofproto_reconnect_controllers(br->ofproto);
+ }
+ }
+ unixctl_command_reply(conn, 200, NULL);
+}
+
static int
bridge_run_one(struct bridge *br)
{
int error;
+ struct iface *iface;
error = ofproto_run1(br->ofproto);
if (error) {
}
mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
+ lacp_run(br);
bond_run(br);
error = ofproto_run2(br->ofproto, br->flush);
br->flush = false;
- return error;
-}
+ HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
+ struct ofpbuf *packet;
-static size_t
-bridge_get_controllers(const struct ovsrec_open_vswitch *ovs_cfg,
- const struct bridge *br,
- struct ovsrec_controller ***controllersp)
+ if (!iface->cfm) {
+ continue;
+ }
+
+ packet = cfm_run(iface->cfm);
+ if (packet) {
+ iface_send_packet(iface, packet);
+ ofpbuf_uninit(packet);
+ free(packet);
+ }
+ }
+
+ return error;
+}
+
+static size_t
+bridge_get_controllers(const struct bridge *br,
+ struct ovsrec_controller ***controllersp)
{
struct ovsrec_controller **controllers;
size_t n_controllers;
- if (br->cfg->n_controller) {
- controllers = br->cfg->controller;
- n_controllers = br->cfg->n_controller;
- } else {
- controllers = ovs_cfg->controller;
- n_controllers = ovs_cfg->n_controller;
- }
+ controllers = br->cfg->controller;
+ n_controllers = br->cfg->n_controller;
if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
controllers = NULL;
}
static void
-bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg,
- struct bridge *br)
+bridge_reconfigure_one(struct bridge *br)
{
struct shash old_ports, new_ports;
- struct svec listeners, old_listeners;
struct svec snoops, old_snoops;
struct shash_node *node;
+ enum ofproto_fail_mode fail_mode;
size_t i;
/* Collect old ports. */
* user didn't specify one.
*
* XXX perhaps we should synthesize a port ourselves in this case. */
- if (bridge_get_controllers(ovs_cfg, br, NULL)) {
+ if (bridge_get_controllers(br, NULL)) {
char local_name[IF_NAMESIZE];
int error;
shash_destroy(&old_ports);
shash_destroy(&new_ports);
+ /* Set the fail-mode */
+ fail_mode = !br->cfg->fail_mode
+ || !strcmp(br->cfg->fail_mode, "standalone")
+ ? OFPROTO_FAIL_STANDALONE
+ : OFPROTO_FAIL_SECURE;
+ if (ofproto_get_fail_mode(br->ofproto) != fail_mode
+ && !ofproto_has_primary_controller(br->ofproto)) {
+ ofproto_flush_flows(br->ofproto);
+ }
+ ofproto_set_fail_mode(br->ofproto, fail_mode);
+
/* Delete all flows if we're switching from connected to standalone or vice
* versa. (XXX Should we delete all flows if we are switching from one
* controller to another?) */
- /* Configure OpenFlow management listener. */
- svec_init(&listeners);
- svec_add_nocopy(&listeners, xasprintf("punix:%s/%s.mgmt",
- ovs_rundir, br->name));
- svec_init(&old_listeners);
- ofproto_get_listeners(br->ofproto, &old_listeners);
- if (!svec_equal(&listeners, &old_listeners)) {
- ofproto_set_listeners(br->ofproto, &listeners);
- }
- svec_destroy(&listeners);
- svec_destroy(&old_listeners);
-
/* Configure OpenFlow controller connection snooping. */
svec_init(&snoops);
svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop",
- ovs_rundir, br->name));
+ ovs_rundir(), br->name));
svec_init(&old_snoops);
ofproto_get_snoops(br->ofproto, &old_snoops);
if (!svec_equal(&snoops, &old_snoops)) {
mirror_reconfigure(br);
}
+/* Initializes 'oc' appropriately as a management service controller for
+ * 'br'.
+ *
+ * The caller must free oc->target when it is no longer needed. */
+static void
+bridge_ofproto_controller_for_mgmt(const struct bridge *br,
+ struct ofproto_controller *oc)
+{
+ oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
+ oc->max_backoff = 0;
+ oc->probe_interval = 60;
+ oc->band = OFPROTO_OUT_OF_BAND;
+ oc->accept_re = NULL;
+ oc->update_resolv_conf = false;
+ oc->rate_limit = 0;
+ oc->burst_limit = 0;
+}
+
+/* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
+static void
+bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
+ struct ofproto_controller *oc)
+{
+ oc->target = c->target;
+ oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
+ oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
+ oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
+ ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
+ oc->accept_re = c->discover_accept_regex;
+ oc->update_resolv_conf = c->discover_update_resolv_conf;
+ oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
+ oc->burst_limit = (c->controller_burst_limit
+ ? *c->controller_burst_limit : 0);
+}
+
+/* Configures the IP stack for 'br''s local interface properly according to the
+ * configuration in 'c'. */
+static void
+bridge_configure_local_iface_netdev(struct bridge *br,
+ struct ovsrec_controller *c)
+{
+ struct netdev *netdev;
+ struct in_addr mask, gateway;
+
+ struct iface *local_iface;
+ struct in_addr ip;
+
+ /* Controller discovery does its own TCP/IP configuration later. */
+ if (strcmp(c->target, "discover")) {
+ return;
+ }
+
+ /* If there's no local interface or no IP address, give up. */
+ local_iface = bridge_get_local_iface(br);
+ if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
+ return;
+ }
+
+ /* Bring up the local interface. */
+ netdev = local_iface->netdev;
+ netdev_turn_flags_on(netdev, NETDEV_UP, true);
+
+ /* Configure the IP address and netmask. */
+ if (!c->local_netmask
+ || !inet_aton(c->local_netmask, &mask)
+ || !mask.s_addr) {
+ mask.s_addr = guess_netmask(ip.s_addr);
+ }
+ if (!netdev_set_in4(netdev, ip, mask)) {
+ VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
+ br->name, IP_ARGS(&ip.s_addr), IP_ARGS(&mask.s_addr));
+ }
+
+ /* Configure the default gateway. */
+ if (c->local_gateway
+ && inet_aton(c->local_gateway, &gateway)
+ && gateway.s_addr) {
+ if (!netdev_add_router(netdev, gateway)) {
+ VLOG_INFO("bridge %s: configured gateway "IP_FMT,
+ br->name, IP_ARGS(&gateway.s_addr));
+ }
+ }
+}
+
static void
-bridge_reconfigure_remotes(const struct ovsrec_open_vswitch *ovs_cfg,
- struct bridge *br,
+bridge_reconfigure_remotes(struct bridge *br,
const struct sockaddr_in *managers,
size_t n_managers)
{
+ const char *disable_ib_str, *queue_id_str;
+ bool disable_in_band = false;
+ int queue_id;
+
struct ovsrec_controller **controllers;
size_t n_controllers;
+ bool had_primary;
- ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
+ struct ofproto_controller *ocs;
+ size_t n_ocs;
+ size_t i;
- n_controllers = bridge_get_controllers(ovs_cfg, br, &controllers);
- if (ofproto_has_controller(br->ofproto) != (n_controllers != 0)) {
- ofproto_flush_flows(br->ofproto);
+ /* 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;
}
- if (!n_controllers) {
- union ofp_action action;
- flow_t flow;
-
- /* Clear out controllers. */
- ofproto_set_controllers(br->ofproto, NULL, 0);
+ /* 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);
- /* 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, OVSFW_ALL, 0, &action, 1, 0);
+ if (disable_in_band) {
+ ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
} else {
- struct ofproto_controller *ocs;
- size_t i;
-
- ocs = xmalloc(n_controllers * sizeof *ocs);
- for (i = 0; i < n_controllers; i++) {
- struct ovsrec_controller *c = controllers[i];
- struct ofproto_controller *oc = &ocs[i];
+ ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
+ }
+ had_primary = ofproto_has_primary_controller(br->ofproto);
- if (strcmp(c->target, "discover")) {
- struct iface *local_iface;
- struct in_addr ip;
+ n_controllers = bridge_get_controllers(br, &controllers);
- local_iface = bridge_get_local_iface(br);
- if (local_iface && c->local_ip
- && inet_aton(c->local_ip, &ip)) {
- struct netdev *netdev = local_iface->netdev;
- struct in_addr mask, gateway;
+ ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
+ n_ocs = 0;
- if (!c->local_netmask
- || !inet_aton(c->local_netmask, &mask)) {
- mask.s_addr = 0;
- }
- if (!c->local_gateway
- || !inet_aton(c->local_gateway, &gateway)) {
- gateway.s_addr = 0;
- }
+ bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
+ for (i = 0; i < n_controllers; i++) {
+ struct ovsrec_controller *c = controllers[i];
- 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));
- }
+ if (!strncmp(c->target, "punix:", 6)
+ || !strncmp(c->target, "unix:", 5)) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- if (gateway.s_addr) {
- if (!netdev_add_router(netdev, gateway)) {
- VLOG_INFO("bridge %s: configured gateway "IP_FMT,
- br->name, IP_ARGS(&gateway.s_addr));
- }
- }
- }
- }
+ /* Prevent remote ovsdb-server users from accessing arbitrary Unix
+ * domain sockets and overwriting arbitrary local files. */
+ VLOG_ERR_RL(&rl, "%s: not adding Unix domain socket controller "
+ "\"%s\" due to possibility for remote exploit",
+ dpif_name(br->dpif), c->target);
+ continue;
+ }
- 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->fail = (!c->fail_mode
- || !strcmp(c->fail_mode, "standalone")
- || !strcmp(c->fail_mode, "open")
- ? OFPROTO_FAIL_STANDALONE
- : OFPROTO_FAIL_SECURE);
- oc->band = (!c->connection_mode
- || !strcmp(c->connection_mode, "in-band")
- ? OFPROTO_IN_BAND
- : OFPROTO_OUT_OF_BAND);
- oc->accept_re = c->discover_accept_regex;
- oc->update_resolv_conf = c->discover_update_resolv_conf;
- oc->rate_limit = (c->controller_rate_limit
- ? *c->controller_rate_limit : 0);
- oc->burst_limit = (c->controller_burst_limit
- ? *c->controller_burst_limit : 0);
+ 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;
}
- ofproto_set_controllers(br->ofproto, ocs, n_controllers);
- free(ocs);
+ n_ocs++;
+ }
+
+ ofproto_set_controllers(br->ofproto, ocs, n_ocs);
+ free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
+ free(ocs);
+
+ if (had_primary != ofproto_has_primary_controller(br->ofproto)) {
+ ofproto_flush_flows(br->ofproto);
+ }
+
+ /* If there are no controllers and the bridge is in standalone
+ * mode, set up a flow that matches every packet and directs
+ * them to OFPP_NORMAL (which goes to us). Otherwise, the
+ * switch is in secure mode and we won't pass any traffic until
+ * a controller has been defined and it tells us to do so. */
+ if (!n_controllers
+ && ofproto_get_fail_mode(br->ofproto) == OFPROTO_FAIL_STANDALONE) {
+ union ofp_action action;
+ struct cls_rule rule;
+
+ memset(&action, 0, sizeof action);
+ action.type = htons(OFPAT_OUTPUT);
+ action.output.len = htons(sizeof action);
+ action.output.port = htons(OFPP_NORMAL);
+ cls_rule_init_catchall(&rule, 0);
+ ofproto_add_flow(br->ofproto, &rule, &action, 1);
}
}
static void
bridge_fetch_dp_ifaces(struct bridge *br)
{
- struct odp_port *dpif_ports;
- size_t n_dpif_ports;
+ struct dpif_port_dump dump;
+ struct dpif_port dpif_port;
size_t i, j;
/* Reset all interface numbers. */
iface->dp_ifidx = -1;
}
}
- port_array_clear(&br->ifaces);
+ hmap_clear(&br->ifaces);
- 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);
+ DPIF_PORT_FOR_EACH (&dpif_port, &dump, br->dpif) {
+ struct iface *iface = iface_lookup(br, dpif_port.name);
if (iface) {
if (iface->dp_ifidx >= 0) {
VLOG_WARN("%s reported interface %s twice",
- dpif_name(br->dpif), p->devname);
- } else if (iface_from_dp_ifidx(br, p->port)) {
+ dpif_name(br->dpif), dpif_port.name);
+ } else if (iface_from_dp_ifidx(br, dpif_port.port_no)) {
VLOG_WARN("%s reported interface %"PRIu16" twice",
- dpif_name(br->dpif), p->port);
+ dpif_name(br->dpif), dpif_port.port_no);
} else {
- port_array_set(&br->ifaces, p->port, iface);
- iface->dp_ifidx = p->port;
+ iface->dp_ifidx = dpif_port.port_no;
+ hmap_insert(&br->ifaces, &iface->dp_ifidx_node,
+ hash_int(iface->dp_ifidx, 0));
}
- if (iface->cfg) {
- int64_t ofport = (iface->dp_ifidx >= 0
- ? odp_port_to_ofp_port(iface->dp_ifidx)
- : -1);
- ovsrec_interface_set_ofport(iface->cfg, &ofport, 1);
- }
+ iface_set_ofport(iface->cfg,
+ (iface->dp_ifidx >= 0
+ ? odp_port_to_ofp_port(iface->dp_ifidx)
+ : -1));
}
}
- free(dpif_ports);
}
\f
/* Bridge packet processing functions. */
+static bool
+bond_is_tcp_hash(const struct port *port)
+{
+ return port->bond_mode == BM_TCP && port->lacp & LACP_NEGOTIATED;
+}
+
static int
-bond_hash(const uint8_t mac[ETH_ADDR_LEN])
+bond_hash_src(const uint8_t mac[ETH_ADDR_LEN], uint16_t vlan)
{
- return hash_bytes(mac, ETH_ADDR_LEN, 0) & BOND_MASK;
+ return hash_bytes(mac, ETH_ADDR_LEN, vlan) & BOND_MASK;
+}
+
+static int bond_hash_tcp(const struct flow *flow, uint16_t vlan)
+{
+ struct flow hash_flow;
+
+ memcpy(&hash_flow, flow, sizeof hash_flow);
+ hash_flow.vlan_tci = 0;
+
+ /* The symmetric quality of this hash function is not required, but
+ * flow_hash_symmetric_l4 already exists, and is sufficient for our
+ * purposes, so we use it out of convenience. */
+ return flow_hash_symmetric_l4(&hash_flow, vlan) & BOND_MASK;
}
static struct bond_entry *
-lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN])
+lookup_bond_entry(const struct port *port, const struct flow *flow,
+ uint16_t vlan)
{
- return &port->bond_hash[bond_hash(mac)];
+ assert(port->bond_mode != BM_AB);
+
+ if (bond_is_tcp_hash(port)) {
+ return &port->bond_hash[bond_hash_tcp(flow, vlan)];
+ } else {
+ return &port->bond_hash[bond_hash_src(flow->dl_src, vlan)];
+ }
}
static int
if (iface->enabled) {
return i;
- } else if (iface->delay_expires < next_delay_expiration) {
+ } else if (iface->delay_expires < next_delay_expiration
+ && (iface->lacp_attached
+ || !(port->lacp & LACP_NEGOTIATED))) {
best_down_slave = i;
next_delay_expiration = iface->delay_expires;
}
}
static bool
-choose_output_iface(const struct port *port, const uint8_t *dl_src,
- uint16_t *dp_ifidx, tag_type *tags)
+choose_output_iface(const struct port *port, const struct flow *flow,
+ uint16_t vlan, uint16_t *dp_ifidx, tag_type *tags)
{
struct iface *iface;
assert(port->n_ifaces);
if (port->n_ifaces == 1) {
iface = port->ifaces[0];
+ } else if (port->bond_mode == BM_AB) {
+ if (port->active_iface < 0) {
+ *tags |= port->no_ifaces_tag;
+ return false;
+ }
+ iface = port->ifaces[port->active_iface];
} else {
- struct bond_entry *e = lookup_bond_entry(port, dl_src);
+ struct bond_entry *e = lookup_bond_entry(port, flow, vlan);
if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces
|| !port->ifaces[e->iface_idx]->enabled) {
/* XXX select interface properly. The current interface selection
}
static void
-bond_link_status_update(struct iface *iface, bool carrier)
+bond_link_status_update(struct iface *iface)
{
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
struct port *port = iface->port;
+ bool up = iface->up;
+ int updelay, downdelay;
+
+ updelay = port->updelay;
+ downdelay = port->downdelay;
- if ((carrier == iface->enabled) == (iface->delay_expires == LLONG_MAX)) {
+ if (iface->port->lacp & LACP_NEGOTIATED) {
+ downdelay = 0;
+ updelay = 0;
+ }
+
+ if (iface->port->lacp && up) {
+ /* The interface is up if it's attached to an aggregator and its
+ * partner is synchronized. The only exception is defaulted links.
+ * They are not required to have synchronized partners because they
+ * have no partners at all. However, they will only be attached if
+ * negotiations failed on all interfaces in the bond. */
+ up = iface->lacp_attached
+ && (iface->lacp_partner.state & LACP_STATE_SYNC
+ || iface->lacp_status == LACP_STATUS_DEFAULTED);
+ }
+
+
+ if ((up == 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) {
+ VLOG_INFO_RL(&rl, "interface %s: link state %s",
+ iface->name, up ? "up" : "down");
+ if (up == iface->enabled) {
iface->delay_expires = LLONG_MAX;
VLOG_INFO_RL(&rl, "interface %s: will not be %s",
- iface->name, carrier ? "disabled" : "enabled");
- } else if (carrier && port->active_iface < 0) {
+ iface->name, up ? "disabled" : "enabled");
+ } else if (up && port->active_iface < 0) {
bond_enable_slave(iface, true);
- if (port->updelay) {
+ if (updelay) {
VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no "
- "other interface is up", iface->name, port->updelay);
+ "other interface is up", iface->name, updelay);
}
} else {
- int delay = carrier ? port->updelay : port->downdelay;
+ int delay = up ? updelay : downdelay;
iface->delay_expires = time_msec() + delay;
if (delay) {
VLOG_INFO_RL(&rl,
"interface %s: will be %s if it stays %s for %d ms",
iface->name,
- carrier ? "enabled" : "disabled",
- carrier ? "up" : "down",
+ up ? "enabled" : "disabled",
+ up ? "up" : "down",
delay);
}
}
}
}
+static void
+bond_link_carrier_update(struct iface *iface, bool carrier)
+{
+ if (carrier == iface->up) {
+ return;
+ }
+
+ if (iface->lacp_status == LACP_STATUS_CURRENT) {
+ iface_set_lacp_expired(iface);
+ }
+
+ iface->up = carrier;
+ iface->lacp_tx = 0;
+ iface->port->bond_compat_is_stale = true;
+}
+
static void
bond_run(struct bridge *br)
{
struct port *port = br->ports[i];
if (port->n_ifaces >= 2) {
+ char *devname;
+
+ if (port->monitor) {
+ assert(!port->miimon);
+
+ /* Track carrier going up and down on interfaces. */
+ while (!netdev_monitor_poll(port->monitor, &devname)) {
+ struct iface *iface;
+
+ iface = port_lookup_iface(port, devname);
+ if (iface) {
+ bool up = netdev_get_carrier(iface->netdev);
+ bond_link_carrier_update(iface, up);
+ }
+ free(devname);
+ }
+ } else {
+ assert(port->miimon);
+
+ if (time_msec() >= port->bond_miimon_next_update) {
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *iface = port->ifaces[j];
+ bool up = netdev_get_miimon(iface->netdev);
+ bond_link_carrier_update(iface, up);
+ }
+ port->bond_miimon_next_update = time_msec() +
+ port->bond_miimon_interval;
+ }
+ }
+
+ for (j = 0; j < port->n_ifaces; j++) {
+ bond_link_status_update(port->ifaces[j]);
+ }
+
for (j = 0; j < port->n_ifaces; j++) {
struct iface *iface = port->ifaces[j];
if (time_msec() >= iface->delay_expires) {
if (port->n_ifaces < 2) {
continue;
}
+
+ if (port->monitor) {
+ netdev_monitor_poll_wait(port->monitor);
+ }
+
+ if (port->miimon) {
+ poll_timer_wait_until(port->bond_miimon_next_update);
+ }
+
for (j = 0; j < port->n_ifaces; j++) {
struct iface *iface = port->ifaces[j];
if (iface->delay_expires != LLONG_MAX) {
}
static bool
-set_dst(struct dst *p, const flow_t *flow,
+set_dst(struct dst *dst, const struct flow *flow,
const struct port *in_port, const struct port *out_port,
tag_type *tags)
{
- p->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
+ dst->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->dl_src, &p->dp_ifidx, tags);
+ : flow->vlan_tci == 0 ? OFP_VLAN_NONE
+ : vlan_tci_to_vid(flow->vlan_tci));
+ return choose_output_iface(out_port, flow, dst->vlan,
+ &dst->dp_ifidx, tags);
}
static void
* 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)
+partition_dsts(struct dst_set *set, int vlan)
{
- struct dst *first = dsts;
- struct dst *last = dsts + n_dsts;
+ struct dst *first = set->dsts;
+ struct dst *last = set->dsts + set->n;
while (first != last) {
/* Invariants:
return ffs(mask);
}
+static void
+dst_set_init(struct dst_set *set)
+{
+ set->dsts = set->builtin;
+ set->n = 0;
+ set->allocated = ARRAY_SIZE(set->builtin);
+}
+
+static void
+dst_set_add(struct dst_set *set, const struct dst *dst)
+{
+ if (set->n >= set->allocated) {
+ size_t new_allocated;
+ struct dst *new_dsts;
+
+ new_allocated = set->allocated * 2;
+ new_dsts = xmalloc(new_allocated * sizeof *new_dsts);
+ memcpy(new_dsts, set->dsts, set->n * sizeof *new_dsts);
+
+ dst_set_free(set);
+
+ set->dsts = new_dsts;
+ set->allocated = new_allocated;
+ }
+ set->dsts[set->n++] = *dst;
+}
+
+static void
+dst_set_free(struct dst_set *set)
+{
+ if (set->dsts != set->builtin) {
+ free(set->dsts);
+ }
+}
+
static bool
-dst_is_duplicate(const struct dst *dsts, size_t n_dsts,
- const struct dst *test)
+dst_is_duplicate(const struct dst_set *set, 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) {
+ for (i = 0; i < set->n; i++) {
+ if (set->dsts[i].vlan == test->vlan
+ && set->dsts[i].dp_ifidx == test->dp_ifidx) {
return true;
}
}
return vlan == port->vlan || port_trunks_vlan(port, vlan);
}
-static size_t
-compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan,
+static bool
+port_is_floodable(const struct port *port)
+{
+ int i;
+
+ for (i = 0; i < port->n_ifaces; i++) {
+ if (!ofproto_port_is_floodable(port->bridge->ofproto,
+ port->ifaces[i]->dp_ifidx)) {
+ return false;
+ }
+ }
+ return true;
+}
+
+static void
+compose_dsts(const struct bridge *br, const struct flow *flow, uint16_t vlan,
const struct port *in_port, const struct port *out_port,
- struct dst dsts[], tag_type *tags, uint16_t *nf_output_iface)
+ struct dst_set *set, tag_type *tags, uint16_t *nf_output_iface)
{
mirror_mask_t mirrors = in_port->src_mirrors;
- struct dst *dst = dsts;
+ struct dst dst;
+ int flow_vlan;
size_t i;
+ flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
+ if (flow_vlan == 0) {
+ flow_vlan = OFP_VLAN_NONE;
+ }
+
if (out_port == FLOOD_PORT) {
- /* 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)
+ if (port != in_port
+ && port_is_floodable(port)
+ && port_includes_vlan(port, vlan)
&& !port->is_mirror_output_port
- && set_dst(dst, flow, in_port, port, tags)) {
+ && set_dst(&dst, flow, in_port, port, tags)) {
mirrors |= port->dst_mirrors;
- dst++;
+ dst_set_add(set, &dst);
}
}
*nf_output_iface = NF_OUT_FLOOD;
- } else if (out_port && set_dst(dst, flow, in_port, out_port, tags)) {
- *nf_output_iface = dst->dp_ifidx;
+ } else if (out_port && set_dst(&dst, flow, in_port, out_port, tags)) {
+ dst_set_add(set, &dst);
+ *nf_output_iface = dst.dp_ifidx;
mirrors |= out_port->dst_mirrors;
- dst++;
}
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++;
+ if (set_dst(&dst, flow, in_port, m->out_port, tags)
+ && !dst_is_duplicate(set, &dst)) {
+ dst_set_add(set, &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))
+ && set_dst(&dst, flow, in_port, port, tags))
{
- int flow_vlan;
-
if (port->vlan < 0) {
- dst->vlan = m->out_vlan;
+ dst.vlan = m->out_vlan;
}
- if (dst_is_duplicate(dsts, dst - dsts, dst)) {
+ if (dst_is_duplicate(set, &dst)) {
continue;
}
* tagging tags place. This is necessary because
* dst->vlan is the final vlan, after removing implicit
* tags. */
- flow_vlan = ntohs(flow->dl_vlan);
- if (flow_vlan == 0) {
- flow_vlan = OFP_VLAN_NONE;
- }
- if (port == in_port && dst->vlan == flow_vlan) {
+ if (port == in_port && dst.vlan == flow_vlan) {
/* Don't send out input port on same VLAN. */
continue;
}
- dst++;
+ dst_set_add(set, &dst);
}
}
}
mirrors &= mirrors - 1;
}
- partition_dsts(dsts, dst - dsts, ntohs(flow->dl_vlan));
- return dst - dsts;
+ partition_dsts(set, flow_vlan);
}
static void OVS_UNUSED
-print_dsts(const struct dst *dsts, size_t n)
+print_dsts(const struct dst_set *set)
{
- for (; n--; dsts++) {
- printf(">p%"PRIu16, dsts->dp_ifidx);
- if (dsts->vlan != OFP_VLAN_NONE) {
- printf("v%"PRIu16, dsts->vlan);
+ size_t i;
+
+ for (i = 0; i < set->n; i++) {
+ const struct dst *dst = &set->dsts[i];
+
+ printf(">p%"PRIu16, dst->dp_ifidx);
+ if (dst->vlan != OFP_VLAN_NONE) {
+ printf("v%"PRIu16, dst->vlan);
}
}
}
static void
-compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan,
+compose_actions(struct bridge *br, const struct flow *flow, uint16_t vlan,
const struct port *in_port, const struct port *out_port,
- tag_type *tags, struct odp_actions *actions,
+ tag_type *tags, struct ofpbuf *actions,
uint16_t *nf_output_iface)
{
- struct dst dsts[DP_MAX_PORTS * (MAX_MIRRORS + 1)];
- size_t n_dsts;
- const struct dst *p;
+ struct dst_set set;
uint16_t cur_vlan;
+ size_t i;
- n_dsts = compose_dsts(br, flow, vlan, in_port, out_port, dsts, tags,
- nf_output_iface);
+ dst_set_init(&set);
+ compose_dsts(br, flow, vlan, in_port, out_port, &set, tags,
+ nf_output_iface);
- 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);
+ cur_vlan = vlan_tci_to_vid(flow->vlan_tci);
+ if (cur_vlan == 0) {
+ cur_vlan = OFP_VLAN_NONE;
+ }
+ for (i = 0; i < set.n; i++) {
+ const struct dst *dst = &set.dsts[i];
+ if (dst->vlan != cur_vlan) {
+ if (dst->vlan == OFP_VLAN_NONE) {
+ nl_msg_put_flag(actions, ODP_ACTION_ATTR_STRIP_VLAN);
} else {
- a = odp_actions_add(actions, ODPAT_SET_VLAN_VID);
- a->vlan_vid.vlan_vid = htons(p->vlan);
+ ovs_be16 tci;
+ tci = htons(dst->vlan & VLAN_VID_MASK);
+ tci |= flow->vlan_tci & htons(VLAN_PCP_MASK);
+ nl_msg_put_be16(actions, ODP_ACTION_ATTR_SET_DL_TCI, tci);
}
- cur_vlan = p->vlan;
+ cur_vlan = dst->vlan;
}
- a = odp_actions_add(actions, ODPAT_OUTPUT);
- a->output.port = p->dp_ifidx;
+ nl_msg_put_u32(actions, ODP_ACTION_ATTR_OUTPUT, dst->dp_ifidx);
}
+ dst_set_free(&set);
}
/* Returns the effective vlan of a packet, taking into account both the
* 802.1Q header and implicitly tagged ports. A value of 0 indicates that
* the packet is untagged and -1 indicates it has an invalid header and
* should be dropped. */
-static int flow_get_vlan(struct bridge *br, const flow_t *flow,
+static int flow_get_vlan(struct bridge *br, const struct flow *flow,
struct port *in_port, bool have_packet)
{
- /* 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.) */
- int vlan = ntohs(flow->dl_vlan);
- if (vlan == OFP_VLAN_NONE) {
- vlan = 0;
- }
+ int vlan = vlan_tci_to_vid(flow->vlan_tci);
if (in_port->vlan >= 0) {
if (vlan) {
/* XXX support double tagging? */
if (have_packet) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %"PRIu16" tagged "
+ VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
"packet received on port %s configured with "
"implicit VLAN %"PRIu16,
- br->name, ntohs(flow->dl_vlan),
- in_port->name, in_port->vlan);
+ br->name, vlan, in_port->name, in_port->vlan);
}
return -1;
}
* migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
* indicate this; newer upstream kernels use gratuitous ARP requests. */
static bool
-is_gratuitous_arp(const flow_t *flow)
+is_gratuitous_arp(const struct flow *flow)
{
return (flow->dl_type == htons(ETH_TYPE_ARP)
&& eth_addr_is_broadcast(flow->dl_dst)
}
static void
-update_learning_table(struct bridge *br, const flow_t *flow, int vlan,
+update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
struct port *in_port)
{
enum grat_arp_lock_type lock_type;
* so in one special case.
*/
static bool
-is_admissible(struct bridge *br, const flow_t *flow, bool have_packet,
+is_admissible(struct bridge *br, const struct flow *flow, bool have_packet,
tag_type *tags, int *vlanp, struct port **in_portp)
{
struct iface *in_iface;
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
- "interface %"PRIu16, br->name, flow->in_port);
+ "interface %"PRIu16, br->name, flow->in_port);
}
*in_portp = NULL;
return false;
}
- /* Packets received on bonds need special attention to avoid duplicates. */
- if (in_port->n_ifaces > 1) {
+ /* When using LACP, do not accept packets from disabled interfaces. */
+ if (in_port->lacp & LACP_NEGOTIATED && !in_iface->enabled) {
+ return false;
+ }
+
+ /* Packets received on non-LACP bonds need special attention to avoid
+ * duplicates. */
+ if (in_port->n_ifaces > 1 && !(in_port->lacp & LACP_NEGOTIATED)) {
int src_idx;
bool is_grat_arp_locked;
* 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,
+process_flow(struct bridge *br, const struct flow *flow,
+ const struct ofpbuf *packet, struct ofpbuf *actions,
tag_type *tags, uint16_t *nf_output_iface)
{
struct port *in_port;
return true;
}
-/* 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_)
+static bool
+bridge_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet,
+ struct ofpbuf *actions, tag_type *tags,
+ uint16_t *nf_output_iface, void *br_)
{
- struct bridge *br = br_;
struct iface *iface;
- struct port *port;
+ struct bridge *br = br_;
- iface = iface_from_dp_ifidx(br, ofp_port_to_odp_port(opp->port_no));
- if (!iface) {
- return;
- }
- port = iface->port;
+ COVERAGE_INC(bridge_process_flow);
- 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);
- port_destroy(port);
- }
+ iface = iface_from_dp_ifidx(br, flow->in_port);
- bridge_flush(br);
- } else {
- if (port->n_ifaces > 1) {
- bool up = !(opp->state & OFPPS_LINK_DOWN);
- bond_link_status_update(iface, up);
- port_update_bond_compat(port);
+ if (cfm_should_process_flow(flow)) {
+ if (packet && iface->cfm) {
+ cfm_process_heartbeat(iface->cfm, packet);
+ }
+ return false;
+ } else if (flow->dl_type == htons(ETH_TYPE_LACP)) {
+ if (packet) {
+ lacp_process_packet(packet, iface);
}
+ return false;
}
-}
-static bool
-bridge_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet,
- struct odp_actions *actions, tag_type *tags,
- uint16_t *nf_output_iface, void *br_)
-{
- struct bridge *br = br_;
-
- COVERAGE_INC(bridge_process_flow);
return process_flow(br, flow, packet, actions, tags, nf_output_iface);
}
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_account_flow_ofhook_cb(const struct flow *flow, tag_type tags,
+ const struct nlattr *actions,
+ size_t actions_len,
+ unsigned long long int n_bytes, void *br_)
{
struct bridge *br = br_;
- const union odp_action *a;
+ const struct nlattr *a;
struct port *in_port;
- tag_type tags = 0;
+ tag_type dummy = 0;
+ unsigned int left;
int vlan;
- /* Feed information from the active flows back into the learning table
- * to ensure that table is always in sync with what is actually flowing
- * through the datapath. */
- if (is_admissible(br, flow, false, &tags, &vlan, &in_port)) {
+ /* Feed information from the active flows back into the learning table to
+ * ensure that table is always in sync with what is actually flowing
+ * through the datapath.
+ *
+ * We test that 'tags' is nonzero to ensure that only flows that include an
+ * OFPP_NORMAL action are used for learning. This works because
+ * bridge_normal_ofhook_cb() always sets a nonzero tag value. */
+ if (tags && is_admissible(br, flow, false, &dummy, &vlan, &in_port)) {
update_learning_table(br, flow, vlan, in_port);
}
+ /* Account for bond slave utilization. */
if (!br->has_bonded_ports) {
return;
}
-
- for (a = actions; a < &actions[n_actions]; a++) {
- if (a->type == ODPAT_OUTPUT) {
- struct port *out_port = port_from_dp_ifidx(br, a->output.port);
- if (out_port && out_port->n_ifaces >= 2) {
- struct bond_entry *e = lookup_bond_entry(out_port,
- flow->dl_src);
+ NL_ATTR_FOR_EACH_UNSAFE (a, left, actions, actions_len) {
+ if (nl_attr_type(a) == ODP_ACTION_ATTR_OUTPUT) {
+ struct port *out_port = port_from_dp_ifidx(br, nl_attr_get_u32(a));
+ if (out_port && out_port->n_ifaces >= 2 &&
+ out_port->bond_mode == BM_SLB) {
+ uint16_t vlan = (flow->vlan_tci
+ ? vlan_tci_to_vid(flow->vlan_tci)
+ : OFP_VLAN_NONE);
+ struct bond_entry *e = lookup_bond_entry(out_port, flow, vlan);
e->tx_bytes += n_bytes;
}
}
now = time_msec();
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
- if (port->n_ifaces > 1 && now >= port->bond_next_rebalance) {
+ if (port->n_ifaces > 1 && port->bond_mode == BM_SLB
+ && now >= port->bond_next_rebalance) {
port->bond_next_rebalance = now + port->bond_rebalance_interval;
bond_rebalance_port(port);
}
}
static struct ofhooks bridge_ofhooks = {
- bridge_port_changed_ofhook_cb,
bridge_normal_ofhook_cb,
bridge_account_flow_ofhook_cb,
bridge_account_checkpoint_ofhook_cb,
};
\f
+/* LACP functions. */
+
+static void
+lacp_process_packet(const struct ofpbuf *packet, struct iface *iface)
+{
+ const struct lacp_pdu *pdu;
+
+ if (!iface->port->lacp) {
+ return;
+ }
+
+ pdu = parse_lacp_packet(packet);
+ if (!pdu) {
+ return;
+ }
+
+ iface->lacp_status = LACP_STATUS_CURRENT;
+ iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX;
+
+ iface->lacp_actor.state = iface_get_lacp_state(iface);
+ if (memcmp(&iface->lacp_actor, &pdu->partner, sizeof pdu->partner)) {
+ iface->lacp_tx = 0;
+ }
+
+ if (memcmp(&iface->lacp_partner, &pdu->actor, sizeof pdu->actor)) {
+ iface->port->lacp_need_update = true;
+ iface->lacp_partner = pdu->actor;
+ }
+}
+
+static void
+lacp_update_ifaces(struct port *port)
+{
+ size_t i;
+ struct iface *lead;
+ struct lacp_info lead_pri;
+
+ port->lacp_need_update = false;
+
+ if (!port->lacp) {
+ return;
+ }
+
+ lead = NULL;
+ for (i = 0; i < port->n_ifaces; i++) {
+ struct iface *iface = port->ifaces[i];
+ struct lacp_info pri;
+
+ iface->lacp_attached = true;
+ ofproto_revalidate(port->bridge->ofproto, iface->tag);
+
+ /* Don't allow loopback interfaces to send traffic or lead. */
+ if (eth_addr_equals(iface->lacp_partner.sysid,
+ iface->lacp_actor.sysid)) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
+ VLOG_WARN_RL(&rl, "iface %s: Loopback detected. Interface is "
+ "connected to its own bridge", iface->name);
+ iface->lacp_attached = false;
+ continue;
+ }
+
+ if (iface->lacp_status == LACP_STATUS_DEFAULTED) {
+ continue;
+ }
+
+ iface_get_lacp_priority(iface, &pri);
+
+ if (!lead || memcmp(&pri, &lead_pri, sizeof pri) < 0) {
+ lead = iface;
+ lead_pri = pri;
+ }
+ }
+
+ if (!lead) {
+ port->lacp &= ~LACP_NEGOTIATED;
+ return;
+ }
+
+ port->lacp |= LACP_NEGOTIATED;
+
+ for (i = 0; i < port->n_ifaces; i++) {
+ struct iface *iface = port->ifaces[i];
+
+ if (iface->lacp_status == LACP_STATUS_DEFAULTED
+ || lead->lacp_partner.key != iface->lacp_partner.key
+ || !eth_addr_equals(lead->lacp_partner.sysid,
+ iface->lacp_partner.sysid)) {
+ iface->lacp_attached = false;
+ }
+ }
+}
+
+static bool
+lacp_iface_may_tx(const struct iface *iface)
+{
+ return iface->port->lacp & LACP_ACTIVE
+ || iface->lacp_status != LACP_STATUS_DEFAULTED;
+}
+
+static void
+lacp_run(struct bridge *br)
+{
+ size_t i, j;
+ struct ofpbuf packet;
+
+ ofpbuf_init(&packet, ETH_HEADER_LEN + LACP_PDU_LEN);
+
+ for (i = 0; i < br->n_ports; i++) {
+ struct port *port = br->ports[i];
+
+ if (!port->lacp) {
+ continue;
+ }
+
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *iface = port->ifaces[j];
+
+ if (time_msec() > iface->lacp_rx) {
+ if (iface->lacp_status == LACP_STATUS_CURRENT) {
+ iface_set_lacp_expired(iface);
+ } else if (iface->lacp_status == LACP_STATUS_EXPIRED) {
+ iface_set_lacp_defaulted(iface);
+ }
+ }
+ }
+
+ if (port->lacp_need_update) {
+ lacp_update_ifaces(port);
+ }
+
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *iface = port->ifaces[j];
+ uint8_t ea[ETH_ADDR_LEN];
+ int error;
+
+ if (time_msec() < iface->lacp_tx || !lacp_iface_may_tx(iface)) {
+ continue;
+ }
+
+ error = netdev_get_etheraddr(iface->netdev, ea);
+ if (!error) {
+ iface->lacp_actor.state = iface_get_lacp_state(iface);
+ compose_lacp_packet(&packet, &iface->lacp_actor,
+ &iface->lacp_partner, ea);
+ iface_send_packet(iface, &packet);
+ } else {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
+ VLOG_ERR_RL(&rl, "iface %s: failed to obtain Ethernet address "
+ "(%s)", iface->name, strerror(error));
+ }
+
+ iface->lacp_tx = time_msec() +
+ (iface->lacp_partner.state & LACP_STATE_TIME
+ ? LACP_FAST_TIME_TX
+ : LACP_SLOW_TIME_TX);
+ }
+ }
+ ofpbuf_uninit(&packet);
+}
+
+static void
+lacp_wait(struct bridge *br)
+{
+ size_t i, j;
+
+ for (i = 0; i < br->n_ports; i++) {
+ struct port *port = br->ports[i];
+
+ if (!port->lacp) {
+ continue;
+ }
+
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *iface = port->ifaces[j];
+
+ if (lacp_iface_may_tx(iface)) {
+ poll_timer_wait_until(iface->lacp_tx);
+ }
+
+ if (iface->lacp_status != LACP_STATUS_DEFAULTED) {
+ poll_timer_wait_until(iface->lacp_rx);
+ }
+ }
+ }
+}
+\f
/* Bonding functions. */
/* Statistics for a single interface on a bonded port, used for load-based
size_t n_hashes;
};
+static const char *
+bond_mode_to_string(enum bond_mode bm) {
+ static char *bm_slb = "balance-slb";
+ static char *bm_ab = "active-backup";
+ static char *bm_tcp = "balance-tcp";
+
+ switch (bm) {
+ case BM_SLB: return bm_slb;
+ case BM_AB: return bm_ab;
+ case BM_TCP: return bm_tcp;
+ }
+
+ NOT_REACHED();
+ return NULL;
+}
+
/* 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. */
struct port *port = from->iface->port;
uint64_t delta = hash->tx_bytes;
+ assert(port->bond_mode == BM_SLB);
+
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)",
static void
bond_rebalance_port(struct port *port)
{
- struct slave_balance bals[DP_MAX_PORTS];
+ struct slave_balance *bals;
size_t n_bals;
struct bond_entry *hashes[BOND_MASK + 1];
struct slave_balance *b, *from, *to;
struct bond_entry *e;
size_t i;
+ assert(port->bond_mode != BM_AB);
+
/* 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
* become contiguous in memory, and then we point each 'hashes' members of
* a slave_balance structure to the start of a contiguous group. */
n_bals = port->n_ifaces;
+ bals = xmalloc(n_bals * sizeof *bals);
for (b = bals; b < &bals[n_bals]; b++) {
b->iface = port->ifaces[b - bals];
b->tx_bytes = 0;
while (!bals[n_bals - 1].iface->enabled) {
n_bals--;
if (!n_bals) {
- return;
+ goto exit;
}
}
* 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;
bool order_swapped;
for (i = 0; i < from->n_hashes; i++) {
for (e = &port->bond_hash[0]; e <= &port->bond_hash[BOND_MASK]; e++) {
e->tx_bytes /= 2;
}
+
+exit:
+ free(bals);
}
static void
struct ofpbuf packet;
int error, n_packets, n_errors;
- if (!port->n_ifaces || port->active_iface < 0) {
+ if (!port->n_ifaces || port->active_iface < 0 || bond_is_tcp_hash(port)) {
return;
}
ofpbuf_init(&packet, 128);
error = n_packets = n_errors = 0;
- LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) {
+ LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
union ofp_action actions[2], *a;
uint16_t dp_ifidx;
tag_type tags = 0;
- flow_t flow;
+ struct flow flow;
int retval;
- if (e->port == port->port_idx
- || !choose_output_iface(port, e->mac, &dp_ifidx, &tags)) {
+ if (e->port == port->port_idx) {
+ continue;
+ }
+
+ compose_benign_packet(&packet, "Open vSwitch Bond Failover", 0xf177,
+ e->mac);
+ flow_extract(&packet, 0, ODPP_NONE, &flow);
+
+ if (!choose_output_iface(port, &flow, e->vlan, &dp_ifidx, &tags)) {
continue;
}
/* Send packet. */
n_packets++;
- compose_benign_packet(&packet, "Open vSwitch Bond Failover", 0xf177,
- e->mac);
- flow_extract(&packet, 0, ODPP_NONE, &flow);
retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions,
&packet);
if (retval) {
struct ds ds = DS_EMPTY_INITIALIZER;
const struct bridge *br;
- ds_put_cstr(&ds, "bridge\tbond\tslaves\n");
+ ds_put_cstr(&ds, "bridge\tbond\ttype\tslaves\n");
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
size_t i;
for (i = 0; i < br->n_ports; i++) {
if (port->n_ifaces > 1) {
size_t j;
- ds_put_format(&ds, "%s\t%s\t", br->name, port->name);
+ ds_put_format(&ds, "%s\t%s\t%s\t", br->name, port->name,
+ bond_mode_to_string(port->bond_mode));
for (j = 0; j < port->n_ifaces; j++) {
const struct iface *iface = port->ifaces[j];
if (j) {
{
const struct bridge *br;
- LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
+ LIST_FOR_EACH (br, node, &all_bridges) {
size_t i;
for (i = 0; i < br->n_ports; i++) {
return NULL;
}
+static void
+ds_put_lacp_state(struct ds *ds, uint8_t state)
+{
+ if (state & LACP_STATE_ACT) {
+ ds_put_cstr(ds, "activity ");
+ }
+
+ if (state & LACP_STATE_TIME) {
+ ds_put_cstr(ds, "timeout ");
+ }
+
+ if (state & LACP_STATE_AGG) {
+ ds_put_cstr(ds, "aggregation ");
+ }
+
+ if (state & LACP_STATE_SYNC) {
+ ds_put_cstr(ds, "synchronized ");
+ }
+
+ if (state & LACP_STATE_COL) {
+ ds_put_cstr(ds, "collecting ");
+ }
+
+ if (state & LACP_STATE_DIST) {
+ ds_put_cstr(ds, "distributing ");
+ }
+
+ if (state & LACP_STATE_DEF) {
+ ds_put_cstr(ds, "defaulted ");
+ }
+
+ if (state & LACP_STATE_EXP) {
+ ds_put_cstr(ds, "expired ");
+ }
+}
+
static void
bond_unixctl_show(struct unixctl_conn *conn,
const char *args, void *aux OVS_UNUSED)
const struct port *port;
size_t j;
- port = bond_find(args);
- if (!port) {
- unixctl_command_reply(conn, 501, "no such bond");
- return;
+ port = bond_find(args);
+ if (!port) {
+ unixctl_command_reply(conn, 501, "no such bond");
+ return;
+ }
+
+ ds_put_format(&ds, "bond_mode: %s\n",
+ bond_mode_to_string(port->bond_mode));
+
+ if (port->lacp) {
+ ds_put_format(&ds, "\tlacp: %s\n",
+ port->lacp & LACP_ACTIVE ? "active" : "passive");
+ } else {
+ ds_put_cstr(&ds, "\tlacp: off\n");
+ }
+
+ if (port->bond_mode != BM_AB) {
+ ds_put_format(&ds, "bond-hash-algorithm: %s\n",
+ bond_is_tcp_hash(port) ? "balance-tcp" : "balance-slb");
+ }
+
+
+ ds_put_format(&ds, "bond-detect-mode: %s\n",
+ port->miimon ? "miimon" : "carrier");
+
+ if (port->miimon) {
+ ds_put_format(&ds, "bond-miimon-interval: %lld\n",
+ port->bond_miimon_interval);
}
ds_put_format(&ds, "updelay: %d ms\n", port->updelay);
ds_put_format(&ds, "downdelay: %d ms\n", port->downdelay);
- ds_put_format(&ds, "next rebalance: %lld ms\n",
- port->bond_next_rebalance - time_msec());
+
+ if (port->bond_mode != BM_AB) {
+ ds_put_format(&ds, "next rebalance: %lld ms\n",
+ port->bond_next_rebalance - time_msec());
+ }
+
for (j = 0; j < port->n_ifaces; j++) {
const struct iface *iface = port->ifaces[j];
struct bond_entry *be;
+ struct flow flow;
/* Basic info. */
ds_put_format(&ds, "slave %s: %s\n",
iface->delay_expires - time_msec());
}
+ if (port->lacp) {
+ ds_put_cstr(&ds, "\tstatus: ");
+
+ if (iface->lacp_status == LACP_STATUS_CURRENT) {
+ ds_put_cstr(&ds, "current ");
+ } else if (iface->lacp_status == LACP_STATUS_EXPIRED) {
+ ds_put_cstr(&ds, "expired ");
+ } else {
+ ds_put_cstr(&ds, "defaulted ");
+ }
+
+ if (iface->lacp_attached) {
+ ds_put_cstr(&ds, "attached ");
+ }
+
+ ds_put_cstr(&ds, "\n");
+
+ ds_put_cstr(&ds, "\n\tactor sysid: ");
+ ds_put_format(&ds, ETH_ADDR_FMT,
+ ETH_ADDR_ARGS(iface->lacp_actor.sysid));
+ ds_put_cstr(&ds, "\n");
+
+ ds_put_format(&ds, "\tactor sys_priority: %u\n",
+ ntohs(iface->lacp_actor.sys_priority));
+
+ ds_put_format(&ds, "\tactor portid: %u\n",
+ ntohs(iface->lacp_actor.portid));
+
+ ds_put_format(&ds, "\tactor port_priority: %u\n",
+ ntohs(iface->lacp_actor.port_priority));
+
+ ds_put_format(&ds, "\tactor key: %u\n",
+ ntohs(iface->lacp_actor.key));
+
+ ds_put_cstr(&ds, "\tactor state: ");
+ ds_put_lacp_state(&ds, iface_get_lacp_state(iface));
+ ds_put_cstr(&ds, "\n\n");
+
+ ds_put_cstr(&ds, "\tpartner sysid: ");
+ ds_put_format(&ds, ETH_ADDR_FMT,
+ ETH_ADDR_ARGS(iface->lacp_partner.sysid));
+ ds_put_cstr(&ds, "\n");
+
+ ds_put_format(&ds, "\tpartner sys_priority: %u\n",
+ ntohs(iface->lacp_partner.sys_priority));
+
+ ds_put_format(&ds, "\tpartner portid: %u\n",
+ ntohs(iface->lacp_partner.portid));
+
+ ds_put_format(&ds, "\tpartner port_priority: %u\n",
+ ntohs(iface->lacp_partner.port_priority));
+
+ ds_put_format(&ds, "\tpartner key: %u\n",
+ ntohs(iface->lacp_partner.key));
+
+ ds_put_cstr(&ds, "\tpartner state: ");
+ ds_put_lacp_state(&ds, iface->lacp_partner.state);
+ ds_put_cstr(&ds, "\n\n");
+ }
+
+ if (port->bond_mode == BM_AB) {
+ continue;
+ }
+
/* Hashes. */
+ memset(&flow, 0, sizeof flow);
for (be = port->bond_hash; be <= &port->bond_hash[BOND_MASK]; be++) {
int hash = be - port->bond_hash;
struct mac_entry *me;
ds_put_format(&ds, "\thash %d: %"PRIu64" kB load\n",
hash, be->tx_bytes / 1024);
+ if (port->bond_mode != BM_SLB) {
+ continue;
+ }
+
/* MACs. */
- LIST_FOR_EACH (me, struct mac_entry, lru_node,
- &port->bridge->ml->lrus) {
+ LIST_FOR_EACH (me, lru_node, &port->bridge->ml->lrus) {
uint16_t dp_ifidx;
tag_type tags = 0;
- if (bond_hash(me->mac) == hash
+
+ memcpy(flow.dl_src, me->mac, ETH_ADDR_LEN);
+ if (bond_hash_src(me->mac, me->vlan) == hash
&& me->port != port->port_idx
- && choose_output_iface(port, me->mac, &dp_ifidx, &tags)
+ && choose_output_iface(port, &flow, me->vlan,
+ &dp_ifidx, &tags)
&& dp_ifidx == iface->dp_ifidx)
{
ds_put_format(&ds, "\t\t"ETH_ADDR_FMT"\n",
char *args = (char *) args_;
char *save_ptr = NULL;
char *bond_s, *hash_s, *slave_s;
- uint8_t mac[ETH_ADDR_LEN];
struct port *port;
struct iface *iface;
struct bond_entry *entry;
return;
}
- if (sscanf(hash_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac))
- == ETH_ADDR_SCAN_COUNT) {
- hash = bond_hash(mac);
- } else if (strspn(hash_s, "0123456789") == strlen(hash_s)) {
+ if (port->bond_mode != BM_SLB) {
+ unixctl_command_reply(conn, 501, "not an SLB bond");
+ return;
+ }
+
+ if (strspn(hash_s, "0123456789") == strlen(hash_s)) {
hash = atoi(hash_s) & BOND_MASK;
} else {
unixctl_command_reply(conn, 501, "bad hash");
}
static void
-bond_unixctl_hash(struct unixctl_conn *conn, const char *args,
+bond_unixctl_hash(struct unixctl_conn *conn, const char *args_,
void *aux OVS_UNUSED)
{
- uint8_t mac[ETH_ADDR_LEN];
- uint8_t hash;
- char *hash_cstr;
+ char *args = (char *) args_;
+ uint8_t mac[ETH_ADDR_LEN];
+ uint8_t hash;
+ char *hash_cstr;
+ unsigned int vlan;
+ char *mac_s, *vlan_s;
+ char *save_ptr = NULL;
+
+ mac_s = strtok_r(args, " ", &save_ptr);
+ vlan_s = strtok_r(NULL, " ", &save_ptr);
+
+ if (vlan_s) {
+ if (sscanf(vlan_s, "%u", &vlan) != 1) {
+ unixctl_command_reply(conn, 501, "invalid vlan");
+ return;
+ }
+ } else {
+ vlan = OFP_VLAN_NONE;
+ }
- if (sscanf(args, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac))
- == ETH_ADDR_SCAN_COUNT) {
- hash = bond_hash(mac);
+ if (sscanf(mac_s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac))
+ == ETH_ADDR_SCAN_COUNT) {
+ hash = bond_hash_src(mac, vlan);
- hash_cstr = xasprintf("%u", hash);
- unixctl_command_reply(conn, 200, hash_cstr);
- free(hash_cstr);
- } else {
- unixctl_command_reply(conn, 501, "invalid mac");
- }
+ hash_cstr = xasprintf("%u", hash);
+ unixctl_command_reply(conn, 200, hash_cstr);
+ free(hash_cstr);
+ } else {
+ unixctl_command_reply(conn, 501, "invalid mac");
+ }
}
static void
get_port_other_config(const struct ovsrec_port *port, const char *key,
const char *default_value)
{
- const char *value = get_ovsrec_key_value(key,
- port->key_other_config,
- port->value_other_config,
- port->n_other_config);
+ 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;
}
static void
port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
{
+ const char *detect_mode;
struct shash new_ifaces;
- long long int next_rebalance;
+ long long int next_rebalance, miimon_next_update, lacp_priority;
unsigned long *trunks;
int vlan;
size_t i;
if (port->updelay < 0) {
port->updelay = 0;
}
- port->updelay = cfg->bond_downdelay;
+ port->downdelay = cfg->bond_downdelay;
if (port->downdelay < 0) {
port->downdelay = 0;
}
port->bond_next_rebalance = next_rebalance;
}
+ detect_mode = get_port_other_config(cfg, "bond-detect-mode",
+ "carrier");
+
+ if (!strcmp(detect_mode, "carrier")) {
+ port->miimon = false;
+ } else if (!strcmp(detect_mode, "miimon")) {
+ port->miimon = true;
+ } else {
+ port->miimon = false;
+ VLOG_WARN("port %s: unsupported bond-detect-mode %s, defaulting to "
+ "carrier", port->name, detect_mode);
+ }
+
+ port->bond_miimon_interval = atoi(
+ get_port_other_config(cfg, "bond-miimon-interval", "200"));
+ if (port->bond_miimon_interval < 100) {
+ port->bond_miimon_interval = 100;
+ }
+ miimon_next_update = time_msec() + port->bond_miimon_interval;
+ if (port->bond_miimon_next_update > miimon_next_update) {
+ port->bond_miimon_next_update = miimon_next_update;
+ }
+
+ if (!port->cfg->bond_mode ||
+ !strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_SLB))) {
+ port->bond_mode = BM_SLB;
+ } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_AB))) {
+ port->bond_mode = BM_AB;
+ } else if (!strcmp(port->cfg->bond_mode, bond_mode_to_string(BM_TCP))) {
+ port->bond_mode = BM_TCP;
+ } else {
+ port->bond_mode = BM_SLB;
+ VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
+ port->name, port->cfg->bond_mode,
+ bond_mode_to_string(port->bond_mode));
+ }
+
/* Add new interfaces and update 'cfg' member of existing ones. */
shash_init(&new_ifaces);
for (i = 0; i < cfg->n_interfaces; i++) {
if (!shash_add_once(&new_ifaces, if_cfg->name, NULL)) {
VLOG_WARN("port %s: %s specified twice as port interface",
port->name, if_cfg->name);
+ iface_set_ofport(if_cfg, -1);
continue;
}
}
iface->cfg = if_cfg;
} else {
- iface_create(port, if_cfg);
+ iface = iface_create(port, if_cfg);
+ }
+
+ /* Determine interface type. The local port always has type
+ * "internal". Other ports take their type from the database and
+ * default to "system" if none is specified. */
+ iface->type = (!strcmp(if_cfg->name, port->bridge->name) ? "internal"
+ : if_cfg->type[0] ? if_cfg->type
+ : "system");
+
+ lacp_priority =
+ atoi(get_interface_other_config(if_cfg, "lacp-port-priority",
+ "0"));
+
+ if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) {
+ iface->lacp_priority = UINT16_MAX;
+ } else {
+ iface->lacp_priority = lacp_priority;
}
}
shash_destroy(&new_ifaces);
+ lacp_priority =
+ atoi(get_port_other_config(cfg, "lacp-system-priority", "0"));
+
+ if (lacp_priority <= 0 || lacp_priority > UINT16_MAX) {
+ /* Prefer bondable links if unspecified. */
+ port->lacp_priority = port->n_ifaces > 1 ? UINT16_MAX - 1 : UINT16_MAX;
+ } else {
+ port->lacp_priority = lacp_priority;
+ }
+
+ if (!port->cfg->lacp) {
+ /* XXX when LACP implementation has been sufficiently tested, enable by
+ * default and make active on bonded ports. */
+ port->lacp = 0;
+ } else if (!strcmp(port->cfg->lacp, "off")) {
+ port->lacp = 0;
+ } else if (!strcmp(port->cfg->lacp, "active")) {
+ port->lacp = LACP_ACTIVE;
+ } else if (!strcmp(port->cfg->lacp, "passive")) {
+ port->lacp = LACP_PASSIVE;
+ } else {
+ VLOG_WARN("port %s: unknown LACP mode %s",
+ port->name, port->cfg->lacp);
+ port->lacp = 0;
+ }
+
/* Get VLAN tag. */
vlan = -1;
if (cfg->tag) {
trunks = NULL;
if (vlan < 0 && cfg->n_trunks) {
size_t n_errors;
- size_t i;
trunks = bitmap_allocate(4096);
n_errors = 0;
del = br->ports[port->port_idx] = br->ports[--br->n_ports];
del->port_idx = port->port_idx;
+ VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
+
+ netdev_monitor_destroy(port->monitor);
free(port->ifaces);
bitmap_free(port->trunks);
free(port->name);
return iface && iface->port == port ? iface : NULL;
}
+static void
+port_update_lacp(struct port *port)
+{
+ size_t i;
+ bool key_changed;
+
+ if (!port->lacp || port->n_ifaces < 1) {
+ return;
+ }
+
+ key_changed = true;
+ for (i = 0; i < port->n_ifaces; i++) {
+ struct iface *iface = port->ifaces[i];
+
+ if (iface->dp_ifidx <= 0 || iface->dp_ifidx > UINT16_MAX) {
+ port->lacp = 0;
+ return;
+ }
+
+ if (iface->dp_ifidx == port->lacp_key) {
+ key_changed = false;
+ }
+ }
+
+ if (key_changed) {
+ port->lacp_key = port->ifaces[0]->dp_ifidx;
+ }
+
+ for (i = 0; i < port->n_ifaces; i++) {
+ struct iface *iface = port->ifaces[i];
+
+ iface->lacp_actor.sys_priority = htons(port->lacp_priority);
+ memcpy(&iface->lacp_actor.sysid, port->bridge->ea, ETH_ADDR_LEN);
+
+ iface->lacp_actor.port_priority = htons(iface->lacp_priority);
+ iface->lacp_actor.portid = htons(iface->dp_ifidx);
+ iface->lacp_actor.key = htons(port->lacp_key);
+
+ iface->lacp_tx = 0;
+ }
+ port->lacp_need_update = true;
+}
+
static void
port_update_bonding(struct port *port)
{
+ if (port->monitor) {
+ netdev_monitor_destroy(port->monitor);
+ port->monitor = NULL;
+ }
if (port->n_ifaces < 2) {
/* Not a bonded port. */
if (port->bond_hash) {
free(port->bond_hash);
port->bond_hash = NULL;
port->bond_compat_is_stale = true;
- port->bond_fake_iface = false;
}
+
+ port->bond_fake_iface = false;
} else {
- if (!port->bond_hash) {
- size_t i;
+ size_t i;
+ if (port->bond_mode != BM_AB && !port->bond_hash) {
port->bond_hash = xcalloc(BOND_MASK + 1, sizeof *port->bond_hash);
for (i = 0; i <= BOND_MASK; i++) {
struct bond_entry *e = &port->bond_hash[i];
if (port->cfg->bond_fake_iface) {
port->bond_next_fake_iface_update = time_msec();
}
+ } else if (port->bond_mode == BM_AB) {
+ free(port->bond_hash);
+ port->bond_hash = NULL;
}
port->bond_compat_is_stale = true;
port->bond_fake_iface = port->cfg->bond_fake_iface;
+
+ if (!port->miimon) {
+ port->monitor = netdev_monitor_create();
+ for (i = 0; i < port->n_ifaces; i++) {
+ netdev_monitor_add(port->monitor, port->ifaces[i]->netdev);
+ }
+ }
}
}
struct compat_bond bond;
size_t i;
- if (port->n_ifaces < 2) {
+ if (port->n_ifaces < 2 || port->bond_mode != BM_SLB) {
proc_net_compat_update_bond(port->name, NULL);
return;
}
/* We need to make the same determination as the Linux bonding
* code to determine whether a slave should be consider "up".
- * The Linux function bond_miimon_inspect() supports four
+ * The Linux function bond_miimon_inspect() supports four
* BOND_LINK_* states:
- *
+ *
* - BOND_LINK_UP: carrier detected, updelay has passed.
* - BOND_LINK_FAIL: carrier lost, downdelay in progress.
* - BOND_LINK_DOWN: carrier lost, downdelay has passed.
* - BOND_LINK_BACK: carrier detected, updelay in progress.
*
- * The function bond_info_show_slave() only considers BOND_LINK_UP
+ * The function bond_info_show_slave() only considers BOND_LINK_UP
* to be "up" and anything else to be "down".
*/
slave->up = iface->enabled && iface->delay_expires == LLONG_MAX;
\f
/* Interface functions. */
+static void
+iface_set_lacp_defaulted(struct iface *iface)
+{
+ memset(&iface->lacp_partner, 0xff, sizeof iface->lacp_partner);
+ iface->lacp_partner.state = 0;
+
+ iface->lacp_status = LACP_STATUS_DEFAULTED;
+ iface->lacp_tx = 0;
+ iface->port->lacp_need_update = true;
+}
+
+static void
+iface_set_lacp_expired(struct iface *iface)
+{
+ iface->lacp_status = LACP_STATUS_EXPIRED;
+ iface->lacp_partner.state |= LACP_STATE_TIME;
+ iface->lacp_partner.state &= ~LACP_STATE_SYNC;
+
+ iface->lacp_rx = time_msec() + LACP_FAST_TIME_RX;
+ iface->lacp_tx = 0;
+}
+
+static uint8_t
+iface_get_lacp_state(const struct iface *iface)
+{
+ uint8_t state = 0;
+
+ if (iface->port->lacp & LACP_ACTIVE) {
+ state |= LACP_STATE_ACT;
+ }
+
+ if (iface->lacp_status == LACP_STATUS_DEFAULTED) {
+ state |= LACP_STATE_DEF;
+ } else if (iface->lacp_attached) {
+ state |= LACP_STATE_SYNC;
+ }
+
+ if (iface->lacp_status == LACP_STATUS_EXPIRED) {
+ state |= LACP_STATE_EXP;
+ }
+
+ if (iface->port->n_ifaces > 1) {
+ state |= LACP_STATE_AGG;
+ }
+
+ if (iface->enabled) {
+ state |= LACP_STATE_COL | LACP_STATE_DIST;
+ }
+
+ return state;
+}
+
+/* Given 'iface', populates 'priority' with data representing its LACP link
+ * priority. If two priority objects populated by this function are compared
+ * using memcmp, the higher priority link will be less than the lower priority
+ * link. */
+static void
+iface_get_lacp_priority(struct iface *iface, struct lacp_info *priority)
+{
+ uint16_t partner_priority, actor_priority;
+
+ /* Choose the lacp_info of the higher priority system by comparing their
+ * system priorities and mac addresses. */
+ actor_priority = ntohs(iface->lacp_actor.sys_priority);
+ partner_priority = ntohs(iface->lacp_partner.sys_priority);
+ if (actor_priority < partner_priority) {
+ *priority = iface->lacp_actor;
+ } else if (partner_priority < actor_priority) {
+ *priority = iface->lacp_partner;
+ } else if (eth_addr_compare_3way(iface->lacp_actor.sysid,
+ iface->lacp_partner.sysid) < 0) {
+ *priority = iface->lacp_actor;
+ } else {
+ *priority = iface->lacp_partner;
+ }
+
+ /* Key and state are not used in priority comparisons. */
+ priority->key = 0;
+ priority->state = 0;
+}
+
+static void
+iface_send_packet(struct iface *iface, struct ofpbuf *packet)
+{
+ struct flow flow;
+ union ofp_action action;
+
+ memset(&action, 0, sizeof action);
+ action.output.type = htons(OFPAT_OUTPUT);
+ action.output.len = htons(sizeof action);
+ action.output.port = htons(odp_port_to_ofp_port(iface->dp_ifidx));
+
+ flow_extract(packet, 0, ODPP_NONE, &flow);
+
+ if (ofproto_send_packet(iface->port->bridge->ofproto, &flow, &action, 1,
+ packet)) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_WARN_RL(&rl, "interface %s: Failed to send packet.", iface->name);
+ }
+}
+
static struct iface *
iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
{
struct bridge *br = port->bridge;
struct iface *iface;
char *name = if_cfg->name;
- int error;
iface = xzalloc(sizeof *iface);
iface->port = port;
iface->delay_expires = LLONG_MAX;
iface->netdev = NULL;
iface->cfg = if_cfg;
+ iface_set_lacp_defaulted(iface);
- shash_add_assert(&br->iface_by_name, iface->name, iface);
-
- /* Attempt to create the network interface in case it doesn't exist yet. */
- if (!iface_is_internal(br, iface->name)) {
- error = set_up_iface(if_cfg, iface, true);
- if (error) {
- VLOG_WARN("could not create iface %s: %s", iface->name,
- strerror(error));
-
- shash_find_and_delete_assert(&br->iface_by_name, iface->name);
- free(iface->name);
- free(iface);
- return NULL;
- }
+ if (port->lacp & LACP_ACTIVE) {
+ iface_set_lacp_expired(iface);
}
+ shash_add_assert(&br->iface_by_name, iface->name, iface);
+
if (port->n_ifaces >= port->allocated_ifaces) {
port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces,
sizeof *port->ifaces);
bool del_active = port->active_iface == iface->port_ifidx;
struct iface *del;
+ if (port->monitor) {
+ netdev_monitor_remove(port->monitor, iface->netdev);
+ }
+
shash_find_and_delete_assert(&br->iface_by_name, iface->name);
if (iface->dp_ifidx >= 0) {
- port_array_set(&br->ifaces, iface->dp_ifidx, NULL);
+ hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
}
del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces];
bond_send_learning_packets(port);
}
+ cfm_destroy(iface->cfm);
+
free(iface->name);
free(iface);
static struct iface *
iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
-{
- return port_array_get(&br->ifaces, dp_ifidx);
-}
-
-/* Returns true if 'iface' is the name of an "internal" interface on bridge
- * 'br', that is, an interface that is entirely simulated within the datapath.
- * The local port (ODPP_LOCAL) is always an internal interface. Other local
- * interfaces are created by setting "iface.<iface>.internal = true".
- *
- * In addition, we have a kluge-y feature that creates an internal port with
- * the name of a bonded port if "bonding.<bondname>.fake-iface = true" is set.
- * This feature needs to go away in the long term. Until then, this is one
- * reason why this function takes a name instead of a struct iface: the fake
- * interfaces created this way do not have a struct iface. */
-static bool
-iface_is_internal(const struct bridge *br, const char *if_name)
{
struct iface *iface;
- struct port *port;
- if (!strcmp(if_name, br->name)) {
- return true;
- }
-
- iface = iface_lookup(br, if_name);
- if (iface && !strcmp(iface->cfg->type, "internal")) {
- return true;
- }
-
- port = port_lookup(br, if_name);
- if (port && port->n_ifaces > 1 && port->cfg->bond_fake_iface) {
- return true;
+ HMAP_FOR_EACH_IN_BUCKET (iface, dp_ifidx_node,
+ hash_int(dp_ifidx, 0), &br->ifaces) {
+ if (iface->dp_ifidx == dp_ifidx) {
+ return iface;
+ }
}
- return false;
+ return NULL;
}
/* Set Ethernet address of 'iface', if one is specified in the configuration
}
}
+/* Sets the ofport column of 'if_cfg' to 'ofport'. */
+static void
+iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
+{
+ if (if_cfg) {
+ ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
+ }
+}
+
+/* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
+ *
+ * The value strings in '*shash' are taken directly from values[], not copied,
+ * so the caller should not modify or free them. */
static void
shash_from_ovs_idl_map(char **keys, char **values, size_t n,
struct shash *shash)
}
}
+/* 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
+shash_to_ovs_idl_map(struct shash *shash,
+ char ***keys, char ***values, size_t *n)
+{
+ size_t i, count;
+ char **k, **v;
+ struct shash_node *sn;
+
+ count = shash_count(shash);
+
+ k = xmalloc(count * sizeof *k);
+ v = xmalloc(count * sizeof *v);
+
+ i = 0;
+ SHASH_FOR_EACH(sn, shash) {
+ k[i] = sn->name;
+ v[i] = sn->data;
+ i++;
+ }
+
+ *n = count;
+ *keys = k;
+ *values = v;
+}
+
struct iface_delete_queues_cbdata {
struct netdev *netdev;
- const int64_t *queue_ids;
- size_t n_queue_ids;
+ const struct ovsdb_datum *queues;
};
static bool
-queue_ids_include(const int64_t *ids, size_t n, int64_t target)
+queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
{
- size_t low = 0;
- size_t high = n;
+ union ovsdb_atom atom;
- while (low < high) {
- size_t mid = low + (high - low) / 2;
- if (target > ids[mid]) {
- high = mid;
- } else if (target < ids[mid]) {
- low = mid + 1;
- } else {
- return true;
- }
- }
- return false;
+ atom.integer = target;
+ return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
}
static void
{
struct iface_delete_queues_cbdata *cbdata = cbdata_;
- if (!queue_ids_include(cbdata->queue_ids, cbdata->n_queue_ids, queue_id)) {
+ if (!queue_ids_include(cbdata->queues, queue_id)) {
netdev_delete_queue(cbdata->netdev, queue_id);
}
}
/* Deconfigure queues that were deleted. */
cbdata.netdev = iface->netdev;
- cbdata.queue_ids = qos->key_queues;
- cbdata.n_queue_ids = qos->n_queues;
+ 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'. */
}
}
}
+
+static void
+iface_update_cfm(struct iface *iface)
+{
+ size_t i;
+ struct cfm *cfm;
+ uint16_t *remote_mps;
+ struct ovsrec_monitor *mon;
+ uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN];
+
+ mon = iface->cfg->monitor;
+
+ if (!mon) {
+ return;
+ }
+
+ if (netdev_get_etheraddr(iface->netdev, ea)) {
+ VLOG_WARN("interface %s: Failed to get ethernet address. "
+ "Skipping Monitor.", iface->name);
+ return;
+ }
+
+ if (!cfm_generate_maid(mon->md_name, mon->ma_name, maid)) {
+ VLOG_WARN("interface %s: Failed to generate MAID.", iface->name);
+ return;
+ }
+
+ if (!iface->cfm) {
+ iface->cfm = cfm_create();
+ }
+
+ cfm = iface->cfm;
+ cfm->mpid = mon->mpid;
+ cfm->interval = mon->interval ? *mon->interval : 1000;
+
+ memcpy(cfm->eth_src, ea, sizeof cfm->eth_src);
+ memcpy(cfm->maid, maid, sizeof cfm->maid);
+
+ remote_mps = xzalloc(mon->n_remote_mps * sizeof *remote_mps);
+ for(i = 0; i < mon->n_remote_mps; i++) {
+ remote_mps[i] = mon->remote_mps[i]->mpid;
+ }
+ cfm_update_remote_mps(cfm, remote_mps, mon->n_remote_mps);
+ free(remote_mps);
+
+ if (!cfm_configure(iface->cfm)) {
+ cfm_destroy(iface->cfm);
+ iface->cfm = NULL;
+ }
+}
\f
/* Port mirroring. */
+static struct mirror *
+mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
+{
+ int i;
+
+ for (i = 0; i < MAX_MIRRORS; i++) {
+ struct mirror *m = br->mirrors[i];
+ if (m && uuid_equals(uuid, &m->uuid)) {
+ return m;
+ }
+ }
+ return NULL;
+}
+
static void
mirror_reconfigure(struct bridge *br)
{
- struct shash old_mirrors, new_mirrors;
- struct shash_node *node;
unsigned long *rspan_vlans;
int i;
- /* Collect old mirrors. */
- shash_init(&old_mirrors);
+ /* Get rid of deleted mirrors. */
for (i = 0; i < MAX_MIRRORS; i++) {
- if (br->mirrors[i]) {
- shash_add(&old_mirrors, br->mirrors[i]->name, br->mirrors[i]);
+ struct mirror *m = br->mirrors[i];
+ if (m) {
+ const struct ovsdb_datum *mc;
+ union ovsdb_atom atom;
+
+ mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
+ atom.uuid = br->mirrors[i]->uuid;
+ if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
+ mirror_destroy(m);
+ }
}
}
- /* Collect new mirrors. */
- shash_init(&new_mirrors);
+ /* Add new mirrors and reconfigure existing ones. */
for (i = 0; i < br->cfg->n_mirrors; i++) {
struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
- if (!shash_add_once(&new_mirrors, cfg->name, cfg)) {
- VLOG_WARN("bridge %s: %s specified twice as mirror",
- br->name, cfg->name);
- }
- }
-
- /* Get rid of deleted mirrors and add new mirrors. */
- SHASH_FOR_EACH (node, &old_mirrors) {
- if (!shash_find(&new_mirrors, node->name)) {
- mirror_destroy(node->data);
- }
- }
- SHASH_FOR_EACH (node, &new_mirrors) {
- struct mirror *mirror = shash_find_data(&old_mirrors, node->name);
- if (!mirror) {
- mirror = mirror_create(br, node->name);
- if (!mirror) {
- break;
- }
+ struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
+ if (m) {
+ mirror_reconfigure_one(m, cfg);
+ } else {
+ mirror_create(br, cfg);
}
- mirror_reconfigure_one(mirror, node->data);
}
- shash_destroy(&old_mirrors);
- shash_destroy(&new_mirrors);
/* Update port reserved status. */
for (i = 0; i < br->n_ports; i++) {
}
}
-static struct mirror *
-mirror_create(struct bridge *br, const char *name)
+static void
+mirror_create(struct bridge *br, struct ovsrec_mirror *cfg)
{
struct mirror *m;
size_t i;
for (i = 0; ; i++) {
if (i >= MAX_MIRRORS) {
VLOG_WARN("bridge %s: maximum of %d port mirrors reached, "
- "cannot create %s", br->name, MAX_MIRRORS, name);
- return NULL;
+ "cannot create %s", br->name, MAX_MIRRORS, cfg->name);
+ return;
}
if (!br->mirrors[i]) {
break;
}
}
- VLOG_INFO("created port mirror %s on bridge %s", name, br->name);
+ VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
bridge_flush(br);
br->mirrors[i] = m = xzalloc(sizeof *m);
m->bridge = br;
m->idx = i;
- m->name = xstrdup(name);
+ m->name = xstrdup(cfg->name);
shash_init(&m->src_ports);
shash_init(&m->dst_ports);
m->vlans = NULL;
m->out_vlan = -1;
m->out_port = NULL;
- return m;
+ mirror_reconfigure_one(m, cfg);
}
static void
free(m->vlans);
m->bridge->mirrors[m->idx] = NULL;
+ free(m->name);
free(m);
bridge_flush(br);
int *vlans;
size_t i;
+ /* Set name. */
+ if (strcmp(cfg->name, m->name)) {
+ free(m->name);
+ m->name = xstrdup(cfg->name);
+ }
+
/* Get output port. */
if (cfg->output_port) {
out_port = port_lookup(m->bridge, cfg->output_port->name);