#include "hash.h"
#include "hmap.h"
#include "jsonrpc.h"
+#include "lacp.h"
#include "list.h"
#include "mac-learning.h"
#include "netdev.h"
static void dst_set_add(struct dst_set *, const struct dst *);
static void dst_set_free(struct dst_set *);
-enum lacp_status {
- LACP_CURRENT = 0x01, /* Current State. */
- LACP_EXPIRED = 0x02, /* Expired State. */
- LACP_DEFAULTED = 0x04, /* Partner is defaulted. */
- LACP_ATTACHED = 0x08, /* Attached. Interface may be choosen for flows. */
-};
-
struct iface {
/* These members are always valid. */
+ struct list port_elem; /* Element in struct port's "ifaces" list. */
struct port *port; /* Containing port. */
- size_t port_ifidx; /* Index within containing port. */
char *name; /* Host network device name. */
tag_type tag; /* Tag associated with this interface. */
long long delay_expires; /* Time after which 'enabled' may change. */
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 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. */
+ uint16_t lacp_priority; /* LACP port priority. */
};
#define BOND_MASK 0xff
struct bond_entry {
- int iface_idx; /* Index of assigned iface, or -1 if none. */
+ struct iface *iface; /* Assigned iface, or NULL if none. */
uint64_t tx_bytes; /* Count of bytes recently transmitted. */
- tag_type iface_tag; /* Tag associated with iface_idx. */
+ tag_type tag; /* Tag for bond_entry<->iface association. */
};
enum bond_mode {
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 port_idx;
+ struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
+ char *name;
+
int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
* NULL if all VLANs are trunked. */
const struct ovsrec_port *cfg;
- char *name;
+
+ /* Monitoring. */
+ struct netdev_monitor *monitor; /* Tracks carrier. NULL if miimon. */
+ long long int miimon_interval; /* Miimon status refresh interval. */
+ long long int miimon_next_update; /* Time of next miimon update. */
/* An ordinary bridge port has 1 interface.
* A bridge port for bonding has at least 2 interfaces. */
- struct iface **ifaces;
- size_t n_ifaces, allocated_ifaces;
+ struct list ifaces; /* List of "struct iface"s. */
+ size_t n_ifaces; /* list_size(ifaces). */
/* Bonding info. */
enum bond_mode bond_mode; /* Type of the bond. BM_SLB is the default. */
- int active_iface; /* Ifidx on which bcasts accepted, or -1. */
- tag_type active_iface_tag; /* Tag for bcast flows. */
+ struct iface *active_iface; /* iface on which bcasts accepted, or NULL. */
tag_type no_ifaces_tag; /* Tag for flows when all ifaces disabled. */
int updelay, downdelay; /* Delay before iface goes up/down, in ms. */
bool bond_fake_iface; /* Fake a bond interface for legacy compat? */
- 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. */
+ struct lacp *lacp; /* LACP object. NULL if LACP is disabled. */
+ bool lacp_active; /* True if LACP is active */
+ bool lacp_fast; /* True if LACP is in fast mode. */
uint16_t lacp_priority; /* LACP system priority. */
- 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. */
struct hmap ifaces; /* Contains "struct iface"s. */
/* Bridge ports. */
- struct port **ports;
- size_t n_ports, allocated_ports;
+ struct hmap ports; /* "struct port"s indexed by name. */
struct shash iface_by_name; /* "struct iface"s indexed by name. */
- struct shash port_by_name; /* "struct port"s indexed by name. */
/* Bonding. */
bool has_bonded_ports;
static uint64_t bridge_pick_datapath_id(struct bridge *,
const uint8_t bridge_ea[ETH_ADDR_LEN],
struct iface *hw_addr_iface);
-static struct iface *bridge_get_local_iface(struct bridge *);
static uint64_t dpid_from_hash(const void *, size_t nbytes);
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 unixctl_cb_func qos_unixctl_show;
static void bond_init(void);
-static void bond_run(struct bridge *);
-static void bond_wait(struct bridge *);
+static void bond_run(struct port *);
+static void bond_wait(struct port *);
static void bond_rebalance_port(struct port *);
static void bond_send_learning_packets(struct port *);
static void bond_enable_slave(struct iface *iface, bool enable);
+static void port_run(struct port *);
+static void port_wait(struct port *);
static struct port *port_create(struct bridge *, const char *name);
static void port_reconfigure(struct port *, const struct ovsrec_port *);
static void port_del_ifaces(struct port *, const struct ovsrec_port *);
static void port_destroy(struct port *);
static struct port *port_lookup(const struct bridge *, const char *name);
static struct iface *port_lookup_iface(const struct port *, const char *name);
+static struct iface *port_get_an_iface(const struct port *);
static struct port *port_from_dp_ifidx(const struct bridge *,
uint16_t dp_ifidx);
static void port_update_bonding(struct 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_find(const char *name);
static struct iface *iface_from_dp_ifidx(const struct bridge *,
uint16_t dp_ifidx);
static void iface_set_mac(struct iface *);
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 iface_update_carrier(struct iface *);
+static bool iface_get_carrier(const struct iface *);
static void shash_from_ovs_idl_map(char **keys, char **values, size_t n,
struct shash *);
static void shash_to_ovs_idl_map(struct shash *,
char ***keys, char ***values, size_t *n);
-
/* Hooks into ofproto processing. */
static struct ofhooks bridge_ofhooks;
\f
/* Register unixctl commands. */
unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
+ unixctl_command_register("qos/show", qos_unixctl_show, NULL);
unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
NULL);
unixctl_command_register("bridge/reconnect", bridge_unixctl_reconnect,
NULL);
+ lacp_init();
bond_init();
}
void *aux),
void *aux)
{
- size_t i, j;
+ struct port *port, *next_port;
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; ) {
- struct iface *iface = port->ifaces[j];
- if (cb(br, iface, aux)) {
- j++;
- } else {
+ HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
+ struct iface *iface, *next_iface;
+
+ LIST_FOR_EACH_SAFE (iface, next_iface, port_elem, &port->ifaces) {
+ if (!cb(br, iface, aux)) {
iface_set_ofport(iface->cfg, -1);
iface_destroy(iface);
}
}
- if (port->n_ifaces) {
- i++;
- } else {
- VLOG_ERR("%s port has no interfaces, dropping", port->name);
+ if (!port->n_ifaces) {
+ VLOG_WARN("%s port has no interfaces, dropping", port->name);
port_destroy(port);
}
}
struct shash targets;
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. */
+ /* Collect all of the potential targets from the "targets" columns of the
+ * rows pointed to by "manager_options", excluding any that are
+ * out-of-band. */
shash_init(&targets);
- 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];
&& 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",
- dpif_port.name, dpif_name(br->dpif),
- strerror(retval));
+ VLOG_WARN("failed to remove %s interface from %s: %s",
+ dpif_port.name, dpif_name(br->dpif),
+ strerror(retval));
}
}
}
dpif_name(br->dpif));
break;
} else {
- VLOG_ERR("failed to add %s interface to %s: %s",
- if_name, dpif_name(br->dpif),
- strerror(error));
+ VLOG_WARN("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->enabled = iface_get_carrier(iface);
iface->up = iface->enabled;
}
} else if (iface && iface->netdev) {
/* Pick local port hardware address, datapath ID. */
bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
- local_iface = bridge_get_local_iface(br);
+ local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
if (local_iface) {
int error = netdev_set_etheraddr(local_iface->netdev, ea);
if (error) {
"with another vswitch, choose an engine id less "
"than 128", br->name);
}
- if (br->n_ports > 508) {
+ if (hmap_count(&br->ports) > 508) {
VLOG_WARN("bridge %s: netflow port mangling will conflict "
"with another port when more than 508 ports are "
"used", br->name);
bridge_reconfigure_remotes(br, managers, n_managers);
}
LIST_FOR_EACH (br, node, &all_bridges) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- int j;
+ struct port *port;
+
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
+
+ if (port->monitor) {
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ netdev_monitor_add(port->monitor, iface->netdev);
+ }
+ } else {
+ port->miimon_next_update = 0;
+ }
- port_update_bonding(port);
port_update_lacp(port);
+ port_update_bonding(port);
- for (j = 0; j < port->n_ifaces; j++) {
- iface_update_qos(port->ifaces[j], port->cfg->qos);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_update_qos(iface, port->cfg->qos);
}
}
}
struct iface **hw_addr_iface)
{
const char *hwaddr;
- size_t i, j;
+ struct port *port;
int error;
*hw_addr_iface = NULL;
/* Otherwise choose the minimum non-local MAC address among all of the
* interfaces. */
- memset(ea, 0xff, sizeof ea);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ memset(ea, 0xff, ETH_ADDR_LEN);
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
uint8_t iface_ea[ETH_ADDR_LEN];
+ struct iface *candidate;
struct iface *iface;
/* Mirror output ports don't participate. */
}
/* Choose the MAC address to represent the port. */
+ iface = NULL;
if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
/* Find the interface with this Ethernet address (if any) so that
* we can provide the correct devname to the caller. */
- iface = NULL;
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *candidate = port->ifaces[j];
+ LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
uint8_t candidate_ea[ETH_ADDR_LEN];
if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
&& eth_addr_equals(iface_ea, candidate_ea)) {
* scripts always add slaves to a bond in alphabetical order, so
* for compatibility we choose the interface with the name that is
* first in alphabetical order. */
- iface = port->ifaces[0];
- for (j = 1; j < port->n_ifaces; j++) {
- struct iface *candidate = port->ifaces[j];
- if (strcmp(candidate->name, iface->name) < 0) {
+ LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
+ if (!iface || strcmp(candidate->name, iface->name) < 0) {
iface = candidate;
}
}
ovsrec_interface_set_link_state(iface->cfg,
- netdev_get_carrier(iface->netdev)
- ? "up" : "down");
+ iface_get_carrier(iface) ? "up" : "down");
error = netdev_get_mtu(iface->netdev, &mtu);
if (!error && mtu != INT_MAX) {
static void
iface_refresh_cfm_stats(struct iface *iface)
{
- size_t i;
- struct cfm *cfm;
const struct ovsrec_monitor *mon;
+ const struct cfm *cfm;
+ size_t i;
mon = iface->cfg->monitor;
- cfm = iface->cfm;
+ cfm = ofproto_iface_get_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
if (!cfm || !mon) {
return;
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);
}
txn = ovsdb_idl_txn_create(idl);
LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ struct port *port;
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- size_t j;
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
iface_refresh_stats(iface);
iface_refresh_cfm_stats(iface);
iface_refresh_status(iface);
bridge_wait(void)
{
struct bridge *br;
- struct iface *iface;
LIST_FOR_EACH (br, node, &all_bridges) {
+ struct port *port;
+
ofproto_wait(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);
- }
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ port_wait(port);
}
}
ovsdb_idl_wait(idl);
{
COVERAGE_INC(bridge_flush);
br->flush = true;
- mac_learning_flush(br->ml);
-}
-
-/* Returns the 'br' interface for the ODPP_LOCAL port, or null if 'br' has no
- * such interface. */
-static struct iface *
-bridge_get_local_iface(struct bridge *br)
-{
- size_t i, j;
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (iface->dp_ifidx == ODPP_LOCAL) {
- return iface;
- }
- }
- }
-
- return NULL;
}
\f
/* Bridge unixctl user interface functions. */
ds_put_cstr(&ds, " port VLAN MAC Age\n");
LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
- if (e->port < 0 || e->port >= br->n_ports) {
- continue;
- }
+ struct port *port = e->port.p;
ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
- br->ports[e->port]->ifaces[0]->dp_ifidx,
+ port_get_an_iface(port)->dp_ifidx,
e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
}
unixctl_command_reply(conn, 200, ds_cstr(&ds));
ds_destroy(&ds);
}
\f
+/* QoS unixctl user interface functions. */
+
+struct qos_unixctl_show_cbdata {
+ struct ds *ds;
+ struct iface *iface;
+};
+
+static void
+qos_unixctl_show_cb(unsigned int queue_id,
+ const struct shash *details,
+ void *aux)
+{
+ struct qos_unixctl_show_cbdata *data = aux;
+ struct ds *ds = data->ds;
+ struct iface *iface = data->iface;
+ struct netdev_queue_stats stats;
+ struct shash_node *node;
+ int error;
+
+ ds_put_cstr(ds, "\n");
+ if (queue_id) {
+ ds_put_format(ds, "Queue %u:\n", queue_id);
+ } else {
+ ds_put_cstr(ds, "Default:\n");
+ }
+
+ SHASH_FOR_EACH (node, details) {
+ ds_put_format(ds, "\t%s: %s\n", node->name, (char *)node->data);
+ }
+
+ error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
+ if (!error) {
+ if (stats.tx_packets != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
+ }
+
+ if (stats.tx_bytes != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
+ }
+
+ if (stats.tx_errors != UINT64_MAX) {
+ ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
+ }
+ } else {
+ ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
+ queue_id, strerror(error));
+ }
+}
+
+static void
+qos_unixctl_show(struct unixctl_conn *conn,
+ const char *args, void *aux OVS_UNUSED)
+{
+ struct ds ds = DS_EMPTY_INITIALIZER;
+ struct shash sh = SHASH_INITIALIZER(&sh);
+ struct iface *iface;
+ const char *type;
+ struct shash_node *node;
+ struct qos_unixctl_show_cbdata data;
+ int error;
+
+ iface = iface_find(args);
+ if (!iface) {
+ unixctl_command_reply(conn, 501, "no such interface");
+ return;
+ }
+
+ netdev_get_qos(iface->netdev, &type, &sh);
+
+ if (*type != '\0') {
+ ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
+
+ SHASH_FOR_EACH (node, &sh) {
+ ds_put_format(&ds, "%s: %s\n", node->name, (char *)node->data);
+ }
+
+ data.ds = &ds;
+ data.iface = iface;
+ error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
+
+ if (error) {
+ ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
+ }
+ unixctl_command_reply(conn, 200, ds_cstr(&ds));
+ } else {
+ ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
+ unixctl_command_reply(conn, 501, ds_cstr(&ds));
+ }
+
+ shash_destroy_free_data(&sh);
+ ds_destroy(&ds);
+}
+\f
/* Bridge reconfiguration functions. */
static struct bridge *
bridge_create(const struct ovsrec_bridge *br_cfg)
br->ml = mac_learning_create();
eth_addr_nicira_random(br->default_ea);
+ hmap_init(&br->ports);
hmap_init(&br->ifaces);
-
- shash_init(&br->port_by_name);
shash_init(&br->iface_by_name);
br->flush = false;
bridge_destroy(struct bridge *br)
{
if (br) {
+ struct port *port, *next;
int error;
- while (br->n_ports > 0) {
- port_destroy(br->ports[br->n_ports - 1]);
+ HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
+ port_destroy(port);
}
list_remove(&br->node);
error = dpif_delete(br->dpif);
ofproto_destroy(br->ofproto);
mac_learning_destroy(br->ml);
hmap_destroy(&br->ifaces);
- shash_destroy(&br->port_by_name);
+ hmap_destroy(&br->ports);
shash_destroy(&br->iface_by_name);
- free(br->ports);
free(br->name);
free(br);
}
static int
bridge_run_one(struct bridge *br)
{
+ struct port *port;
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);
+
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ port_run(port);
+ }
error = ofproto_run2(br->ofproto, br->flush);
br->flush = false;
- HMAP_FOR_EACH (iface, dp_ifidx_node, &br->ifaces) {
- struct ofpbuf *packet;
-
- if (!iface->cfm) {
- continue;
- }
-
- packet = cfm_run(iface->cfm);
- if (packet) {
- iface_send_packet(iface, packet);
- ofpbuf_uninit(packet);
- free(packet);
- }
- }
-
return error;
}
static void
bridge_reconfigure_one(struct bridge *br)
{
- struct shash old_ports, new_ports;
+ enum ofproto_fail_mode fail_mode;
struct svec snoops, old_snoops;
+ struct port *port, *next;
struct shash_node *node;
- enum ofproto_fail_mode fail_mode;
+ struct shash new_ports;
size_t i;
- /* Collect old ports. */
- shash_init(&old_ports);
- for (i = 0; i < br->n_ports; i++) {
- shash_add(&old_ports, br->ports[i]->name, br->ports[i]);
- }
-
/* Collect new ports. */
shash_init(&new_ports);
for (i = 0; i < br->cfg->n_ports; i++) {
/* Get rid of deleted ports.
* Get rid of deleted interfaces on ports that still exist. */
- SHASH_FOR_EACH (node, &old_ports) {
- struct port *port = node->data;
+ HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
const struct ovsrec_port *port_cfg;
- port_cfg = shash_find_data(&new_ports, node->name);
+ port_cfg = shash_find_data(&new_ports, port->name);
if (!port_cfg) {
port_destroy(port);
} else {
* Add new interfaces to existing ports.
* Reconfigure existing ports. */
SHASH_FOR_EACH (node, &new_ports) {
- struct port *port = shash_find_data(&old_ports, node->name);
+ struct port *port = port_lookup(br, node->name);
if (!port) {
port = port_create(br, node->name);
}
port_destroy(port);
}
}
- shash_destroy(&old_ports);
shash_destroy(&new_ports);
/* Set the fail-mode */
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;
}
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);
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);
+ local_iface = iface_from_dp_ifidx(br, ODPP_LOCAL);
if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
return;
}
static void
bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
{
- size_t i, j;
+ struct port *port;
shash_init(ifaces);
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
shash_add_once(ifaces, iface->name, iface);
}
if (port->n_ifaces > 1 && port->cfg->bond_fake_iface) {
{
struct dpif_port_dump dump;
struct dpif_port dpif_port;
- size_t i, j;
+ struct port *port;
/* Reset all interface numbers. */
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ struct iface *iface;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
iface->dp_ifidx = -1;
}
}
static bool
bond_is_tcp_hash(const struct port *port)
{
- return port->bond_mode == BM_TCP && port->lacp & LACP_NEGOTIATED;
+ return port->bond_mode == BM_TCP && lacp_negotiated(port->lacp);
}
static int
}
}
-static int
+static struct iface *
bond_choose_iface(const struct port *port)
{
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
- size_t i, best_down_slave = -1;
- long long next_delay_expiration = LLONG_MAX;
-
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
+ struct iface *best_down_slave;
+ struct iface *iface;
+ best_down_slave = NULL;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
if (iface->enabled) {
- return i;
- } else if (iface->delay_expires < next_delay_expiration
- && (iface->lacp_status & LACP_ATTACHED
- || !(port->lacp & LACP_NEGOTIATED))) {
- best_down_slave = i;
- next_delay_expiration = iface->delay_expires;
+ return iface;
+ } else if ((!best_down_slave
+ || iface->delay_expires < best_down_slave->delay_expires)
+ && lacp_slave_may_enable(port->lacp, iface)) {
+ best_down_slave = iface;
}
}
- if (best_down_slave != -1) {
- struct iface *iface = port->ifaces[best_down_slave];
-
+ if (best_down_slave) {
VLOG_INFO_RL(&rl, "interface %s: skipping remaining %lli ms updelay "
- "since no other interface is up", iface->name,
- iface->delay_expires - time_msec());
- bond_enable_slave(iface, true);
+ "since no other interface is up",
+ best_down_slave->name,
+ best_down_slave->delay_expires - time_msec());
+ bond_enable_slave(best_down_slave, true);
}
return best_down_slave;
assert(port->n_ifaces);
if (port->n_ifaces == 1) {
- iface = port->ifaces[0];
+ iface = port_get_an_iface(port);
} else if (port->bond_mode == BM_AB) {
- if (port->active_iface < 0) {
+ iface = port->active_iface;
+ if (!iface) {
*tags |= port->no_ifaces_tag;
return false;
}
- iface = port->ifaces[port->active_iface];
} else {
struct bond_entry *e = lookup_bond_entry(port, flow, vlan);
- if (e->iface_idx < 0 || e->iface_idx >= port->n_ifaces
- || !port->ifaces[e->iface_idx]->enabled) {
+ if (!e->iface || !e->iface->enabled) {
/* XXX select interface properly. The current interface selection
* is only good for testing the rebalancing code. */
- e->iface_idx = bond_choose_iface(port);
- if (e->iface_idx < 0) {
+ e->iface = bond_choose_iface(port);
+ if (!e->iface) {
*tags |= port->no_ifaces_tag;
return false;
}
- e->iface_tag = tag_create_random();
+ e->tag = tag_create_random();
}
- *tags |= e->iface_tag;
- iface = port->ifaces[e->iface_idx];
+ *tags |= e->tag;
+ iface = e->iface;
}
*dp_ifidx = iface->dp_ifidx;
*tags |= iface->tag; /* Currently only used for bonding. */
{
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
struct port *port = iface->port;
- bool up = iface->up;
+ bool up = iface->up && lacp_slave_may_enable(port->lacp, iface);
int updelay, downdelay;
updelay = port->updelay;
downdelay = port->downdelay;
- if (iface->port->lacp & LACP_NEGOTIATED) {
+ if (lacp_negotiated(port->lacp)) {
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_status & LACP_ATTACHED
- && (iface->lacp_partner.state & LACP_STATE_SYNC
- || iface->lacp_status & LACP_DEFAULTED);
- }
-
-
if ((up == iface->enabled) == (iface->delay_expires == LLONG_MAX)) {
/* Nothing to do. */
return;
iface->delay_expires = LLONG_MAX;
VLOG_INFO_RL(&rl, "interface %s: will not be %s",
iface->name, up ? "disabled" : "enabled");
- } else if (up && port->active_iface < 0) {
+ } else if (up && !port->active_iface) {
bond_enable_slave(iface, true);
if (updelay) {
VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no "
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
port->active_iface = bond_choose_iface(port);
- port->active_iface_tag = tag_create_random();
- if (port->active_iface >= 0) {
+ if (port->active_iface) {
VLOG_INFO_RL(&rl, "port %s: active interface is now %s",
- port->name, port->ifaces[port->active_iface]->name);
+ port->name, port->active_iface->name);
} else {
VLOG_WARN_RL(&rl, "port %s: all ports disabled, no active interface",
port->name);
if (!iface->enabled) {
VLOG_WARN("interface %s: disabled", iface->name);
ofproto_revalidate(br->ofproto, iface->tag);
- if (iface->port_ifidx == port->active_iface) {
- ofproto_revalidate(br->ofproto,
- port->active_iface_tag);
-
+ if (iface == port->active_iface) {
/* Disabling a slave can lead to another slave being immediately
* enabled if there will be no active slaves but one is waiting
* on an updelay. In this case we do not need to run most of the
bond_send_learning_packets(port);
} else {
VLOG_WARN("interface %s: enabled", iface->name);
- if (port->active_iface < 0 && !moving_active_iface) {
+ if (!port->active_iface && !moving_active_iface) {
ofproto_revalidate(br->ofproto, port->no_ifaces_tag);
bond_choose_active_iface(port);
bond_send_learning_packets(port);
{
struct netdev_stats bond_stats;
struct netdev *bond_dev;
- size_t i;
+ struct iface *iface;
memset(&bond_stats, 0, sizeof bond_stats);
- for (i = 0; i < port->n_ifaces; i++) {
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
struct netdev_stats slave_stats;
- if (!netdev_get_stats(port->ifaces[i]->netdev, &slave_stats)) {
+ if (!netdev_get_stats(iface->netdev, &slave_stats)) {
/* XXX: We swap the stats here because they are swapped back when
* reported by the internal device. The reason for this is
* internal devices normally represent packets going into the system
}
static void
-bond_link_carrier_update(struct iface *iface, bool carrier)
+bond_run(struct port *port)
{
- if (carrier == iface->up) {
+ struct iface *iface;
+
+ if (port->n_ifaces < 2) {
return;
}
- if (iface->lacp_status & LACP_CURRENT) {
- iface_set_lacp_expired(iface);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ bond_link_status_update(iface);
}
- iface->up = carrier;
- iface->lacp_tx = 0;
-}
-
-static void
-bond_run(struct bridge *br)
-{
- size_t i, j;
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
-
- if (port->n_ifaces >= 2) {
- 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) {
- bond_enable_slave(iface, !iface->enabled);
- }
- }
-
- if (port->bond_fake_iface
- && time_msec() >= port->bond_next_fake_iface_update) {
- bond_update_fake_iface_stats(port);
- port->bond_next_fake_iface_update = time_msec() + 1000;
- }
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ if (time_msec() >= iface->delay_expires) {
+ bond_enable_slave(iface, !iface->enabled);
}
}
+
+ if (port->bond_fake_iface
+ && time_msec() >= port->bond_next_fake_iface_update) {
+ bond_update_fake_iface_stats(port);
+ port->bond_next_fake_iface_update = time_msec() + 1000;
+ }
}
static void
-bond_wait(struct bridge *br)
+bond_wait(struct port *port)
{
- size_t i, j;
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
- if (port->n_ifaces < 2) {
- continue;
- }
+ struct iface *iface;
- if (port->monitor) {
- netdev_monitor_poll_wait(port->monitor);
- }
+ if (port->n_ifaces < 2) {
+ return;
+ }
- if (port->miimon) {
- poll_timer_wait_until(port->bond_miimon_next_update);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ if (iface->delay_expires != LLONG_MAX) {
+ poll_timer_wait_until(iface->delay_expires);
}
+ }
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (iface->delay_expires != LLONG_MAX) {
- poll_timer_wait_until(iface->delay_expires);
- }
- }
- if (port->bond_fake_iface) {
- poll_timer_wait_until(port->bond_next_fake_iface_update);
- }
+ if (port->bond_fake_iface) {
+ poll_timer_wait_until(port->bond_next_fake_iface_update);
}
}
static bool
port_is_floodable(const struct port *port)
{
- int i;
+ struct iface *iface;
- for (i = 0; i < port->n_ifaces; i++) {
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
if (!ofproto_port_is_floodable(port->bridge->ofproto,
- port->ifaces[i]->dp_ifidx)) {
+ iface->dp_ifidx)) {
return false;
}
}
return true;
}
+/* Returns the tag for 'port''s active iface, or 'port''s no_ifaces_tag if
+ * there is no active iface. */
+static tag_type
+port_get_active_iface_tag(const struct port *port)
+{
+ return (port->active_iface
+ ? port->active_iface->tag
+ : port->no_ifaces_tag);
+}
+
+/* Returns an arbitrary interface within 'port'.
+ *
+ * 'port' must have at least one interface. */
+static struct iface *
+port_get_an_iface(const struct port *port)
+{
+ return CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
+}
+
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,
mirror_mask_t mirrors = in_port->src_mirrors;
struct dst dst;
int flow_vlan;
- size_t i;
flow_vlan = vlan_tci_to_vid(flow->vlan_tci);
if (flow_vlan == 0) {
}
if (out_port == FLOOD_PORT) {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ struct port *port;
+
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
if (port != in_port
&& port_is_floodable(port)
&& port_includes_vlan(port, vlan)
dst_set_add(set, &dst);
}
} else {
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ struct port *port;
+
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
if (port_includes_vlan(port, m->out_vlan)
&& set_dst(&dst, flow, in_port, port, tags))
{
update_learning_table(struct bridge *br, const struct flow *flow, int vlan,
struct port *in_port)
{
- enum grat_arp_lock_type lock_type;
- tag_type rev_tag;
+ struct mac_entry *mac;
- /* We don't want to learn from gratuitous ARP packets that are reflected
- * back over bond slaves so we lock the learning table. */
- lock_type = !is_gratuitous_arp(flow) ? GRAT_ARP_LOCK_NONE :
- (in_port->n_ifaces == 1) ? GRAT_ARP_LOCK_SET :
- GRAT_ARP_LOCK_CHECK;
+ if (!mac_learning_may_learn(br->ml, flow->dl_src, vlan)) {
+ return;
+ }
+
+ mac = mac_learning_insert(br->ml, flow->dl_src, vlan);
+ if (is_gratuitous_arp(flow)) {
+ /* We don't want to learn from gratuitous ARP packets that are
+ * reflected back over bond slaves so we lock the learning table. */
+ if (in_port->n_ifaces == 1) {
+ mac_entry_set_grat_arp_lock(mac);
+ } else if (mac_entry_is_grat_arp_locked(mac)) {
+ return;
+ }
+ }
- rev_tag = mac_learning_learn(br->ml, flow->dl_src, vlan, in_port->port_idx,
- lock_type);
- if (rev_tag) {
+ if (mac_entry_is_new(mac) || mac->port.p != in_port) {
/* The log messages here could actually be useful in debugging,
* so keep the rate limit relatively high. */
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30,
- 300);
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
"on port %s in VLAN %d",
br->name, ETH_ADDR_ARGS(flow->dl_src),
in_port->name, vlan);
- ofproto_revalidate(br->ofproto, rev_tag);
+
+ mac->port.p = in_port;
+ ofproto_revalidate(br->ofproto, mac_learning_changed(br->ml, mac));
}
}
}
/* When using LACP, do not accept packets from disabled interfaces. */
- if (in_port->lacp & LACP_NEGOTIATED && !in_iface->enabled) {
+ if (lacp_negotiated(in_port->lacp) && !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;
+ if (in_port->n_ifaces > 1 && !lacp_negotiated(in_port->lacp)) {
+ struct mac_entry *mac;
if (eth_addr_is_multicast(flow->dl_dst)) {
- *tags |= in_port->active_iface_tag;
- if (in_port->active_iface != in_iface->port_ifidx) {
+ *tags |= port_get_active_iface_tag(in_port);
+ if (in_port->active_iface != in_iface) {
/* Drop all multicast packets on inactive slaves. */
return false;
}
* to the exception is if we locked the learning table to avoid
* reflections on bond slaves. If this is the case, just drop the
* packet now. */
- src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan,
- &is_grat_arp_locked);
- if (src_idx != -1 && src_idx != in_port->port_idx &&
- (!is_gratuitous_arp(flow) || is_grat_arp_locked)) {
+ mac = mac_learning_lookup(br->ml, flow->dl_src, vlan, NULL);
+ if (mac && mac->port.p != in_port &&
+ (!is_gratuitous_arp(flow) || mac_entry_is_grat_arp_locked(mac))) {
return false;
}
}
{
struct port *in_port;
struct port *out_port;
+ struct mac_entry *mac;
int vlan;
- int out_port_idx;
/* Check whether we should drop packets in this flow. */
if (!is_admissible(br, flow, packet != NULL, tags, &vlan, &in_port)) {
}
/* Determine output port. */
- out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan, tags,
- NULL);
- if (out_port_idx >= 0 && out_port_idx < br->n_ports) {
- out_port = br->ports[out_port_idx];
+ mac = mac_learning_lookup(br->ml, flow->dl_dst, vlan, tags);
+ if (mac) {
+ out_port = mac->port.p;
} else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
/* If we are revalidating but don't have a learning entry then
* eject the flow. Installing a flow that floods packets opens
iface = iface_from_dp_ifidx(br, flow->in_port);
- if (cfm_should_process_flow(flow)) {
+ if (flow->dl_type == htons(ETH_TYPE_LACP)) {
- if (iface && packet && iface->cfm) {
- COVERAGE_INC(bridge_process_cfm);
- cfm_process_heartbeat(iface->cfm, packet);
- }
- return false;
- } else if (flow->dl_type == htons(ETH_TYPE_LACP)) {
+ if (iface && iface->port->lacp && packet) {
+ const struct lacp_pdu *pdu = parse_lacp_packet(packet);
- if (iface && packet) {
- COVERAGE_INC(bridge_process_lacp);
- lacp_process_packet(packet, iface);
+ if (pdu) {
+ COVERAGE_INC(bridge_process_lacp);
+ lacp_process_pdu(iface->port->lacp, iface, pdu);
+ }
}
return false;
}
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_)
+ uint64_t n_bytes, void *br_)
{
struct bridge *br = br_;
const struct nlattr *a;
bridge_account_checkpoint_ofhook_cb(void *br_)
{
struct bridge *br = br_;
+ struct port *port;
long long int now;
- size_t i;
if (!br->has_bonded_ports) {
return;
}
now = time_msec();
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
if (port->n_ifaces > 1 && port->bond_mode != BM_AB
&& now >= port->bond_next_rebalance) {
port->bond_next_rebalance = now + port->bond_rebalance_interval;
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;
+/* Bonding functions. */
- if (!iface->port->lacp) {
- return;
- }
+/* Statistics for a single interface on a bonded port, used for load-based
+ * bond rebalancing. */
+struct slave_balance {
+ struct iface *iface; /* The interface. */
+ uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */
- pdu = parse_lacp_packet(packet);
- if (!pdu) {
- return;
- }
+ /* All the "bond_entry"s that are assigned to this interface, in order of
+ * increasing tx_bytes. */
+ struct bond_entry **hashes;
+ size_t n_hashes;
+};
- iface->lacp_status |= LACP_CURRENT;
- iface->lacp_status &= ~(LACP_EXPIRED | LACP_DEFAULTED);
- iface->lacp_rx = time_msec() + LACP_SLOW_TIME_RX;
+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";
- iface->lacp_actor.state = iface_get_lacp_state(iface);
- if (memcmp(&iface->lacp_actor, &pdu->partner, sizeof pdu->partner)) {
- iface->lacp_tx = 0;
+ switch (bm) {
+ case BM_SLB: return bm_slb;
+ case BM_AB: return bm_ab;
+ case BM_TCP: return bm_tcp;
}
- if (memcmp(&iface->lacp_partner, &pdu->actor, sizeof pdu->actor)) {
- iface->port->lacp_need_update = true;
- iface->lacp_partner = pdu->actor;
- }
+ NOT_REACHED();
+ return NULL;
}
-static void
-lacp_update_ifaces(struct port *port)
-{
- size_t i;
- struct iface *lead;
- struct lacp_info lead_pri;
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
-
- port->lacp_need_update = false;
- COVERAGE_INC(bridge_lacp_update);
-
- if (!port->lacp) {
- return;
- }
-
- VLOG_DBG_RL(&rl, "port %s: re-evaluating LACP link status", port->name);
-
- lead = NULL;
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- struct lacp_info pri;
-
- iface->lacp_status |= LACP_ATTACHED;
- 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)) {
- VLOG_WARN_RL(&rl, "iface %s: Loopback detected. Interface is "
- "connected to its own bridge", iface->name);
- iface->lacp_status &= ~LACP_ATTACHED;
- continue;
- }
-
- if (iface->lacp_status & LACP_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_DEFAULTED
- || lead->lacp_partner.key != iface->lacp_partner.key
- || !eth_addr_equals(lead->lacp_partner.sysid,
- iface->lacp_partner.sysid)) {
- iface->lacp_status &= ~LACP_ATTACHED;
- }
- }
-}
-
-static bool
-lacp_iface_may_tx(const struct iface *iface)
-{
- return iface->port->lacp & LACP_ACTIVE
- || iface->lacp_status & (LACP_CURRENT | LACP_EXPIRED);
-}
-
-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_CURRENT) {
- iface_set_lacp_expired(iface);
- } else if (iface->lacp_status & LACP_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_CURRENT | LACP_EXPIRED)) {
- poll_timer_wait_until(iface->lacp_rx);
- }
- }
- }
-}
-\f
-/* Bonding functions. */
-
-/* Statistics for a single interface on a bonded port, used for load-based
- * bond rebalancing. */
-struct slave_balance {
- struct iface *iface; /* The interface. */
- uint64_t tx_bytes; /* Sum of hashes[*]->tx_bytes. */
-
- /* All the "bond_entry"s that are assigned to this interface, in order of
- * increasing tx_bytes. */
- struct bond_entry **hashes;
- size_t n_hashes;
-};
-
-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. */
-static int
-compare_bond_entries(const void *a_, const void *b_)
+/* Sorts pointers to pointers to bond_entries in ascending order by the
+ * interface to which they are assigned, and within a single interface in
+ * ascending order of bytes transmitted. */
+static int
+compare_bond_entries(const void *a_, const void *b_)
{
const struct bond_entry *const *ap = a_;
const struct bond_entry *const *bp = b_;
const struct bond_entry *a = *ap;
const struct bond_entry *b = *bp;
- if (a->iface_idx != b->iface_idx) {
- return a->iface_idx > b->iface_idx ? 1 : -1;
+ if (a->iface != b->iface) {
+ return a->iface > b->iface ? 1 : -1;
} else if (a->tx_bytes != b->tx_bytes) {
return a->tx_bytes > b->tx_bytes ? 1 : -1;
} else {
to->tx_bytes += delta;
/* Arrange for flows to be revalidated. */
- ofproto_revalidate(port->bridge->ofproto, hash->iface_tag);
- hash->iface_idx = to->iface->port_ifidx;
- hash->iface_tag = tag_create_random();
+ ofproto_revalidate(port->bridge->ofproto, hash->tag);
+ hash->iface = to->iface;
+ hash->tag = tag_create_random();
}
static void
struct bond_entry *hashes[BOND_MASK + 1];
struct slave_balance *b, *from, *to;
struct bond_entry *e;
+ struct iface *iface;
size_t i;
assert(port->bond_mode != BM_AB);
* 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 = bals = xmalloc(n_bals * sizeof *bals);
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ b->iface = iface;
b->tx_bytes = 0;
b->hashes = NULL;
b->n_hashes = 0;
+ b++;
}
+ assert(b == &bals[n_bals]);
for (i = 0; i <= BOND_MASK; i++) {
hashes[i] = &port->bond_hash[i];
}
qsort(hashes, BOND_MASK + 1, sizeof *hashes, compare_bond_entries);
for (i = 0; i <= BOND_MASK; i++) {
e = hashes[i];
- if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) {
- b = &bals[e->iface_idx];
- b->tx_bytes += e->tx_bytes;
- if (!b->hashes) {
- b->hashes = &hashes[i];
+ if (!e->iface) {
+ continue;
+ }
+
+ for (b = bals; b < &bals[n_bals]; b++) {
+ if (b->iface == e->iface) {
+ b->tx_bytes += e->tx_bytes;
+ if (!b->hashes) {
+ b->hashes = &hashes[i];
+ }
+ b->n_hashes++;
+ break;
}
- b->n_hashes++;
}
}
qsort(bals, n_bals, sizeof *bals, compare_slave_balance);
* historical data to decay to <1% in 7 rebalancing runs. */
for (e = &port->bond_hash[0]; e <= &port->bond_hash[BOND_MASK]; e++) {
e->tx_bytes /= 2;
+ if (!e->tx_bytes) {
+ e->iface = NULL;
+ }
}
exit:
struct ofpbuf packet;
int error, n_packets, n_errors;
- if (!port->n_ifaces || port->active_iface < 0 || bond_is_tcp_hash(port)) {
+ if (!port->n_ifaces || !port->active_iface || bond_is_tcp_hash(port)) {
return;
}
ofpbuf_init(&packet, 128);
error = n_packets = n_errors = 0;
LIST_FOR_EACH (e, lru_node, &br->ml->lrus) {
- union ofp_action actions[2], *a;
- uint16_t dp_ifidx;
tag_type tags = 0;
+ uint16_t dp_ifidx;
struct flow flow;
int retval;
- if (e->port == port->port_idx) {
+ if (e->port.p == port) {
continue;
}
continue;
}
- /* Compose actions. */
- memset(actions, 0, sizeof actions);
- a = actions;
- if (e->vlan) {
- a->vlan_vid.type = htons(OFPAT_SET_VLAN_VID);
- a->vlan_vid.len = htons(sizeof *a);
- a->vlan_vid.vlan_vid = htons(e->vlan);
- a++;
- }
- a->output.type = htons(OFPAT_OUTPUT);
- a->output.len = htons(sizeof *a);
- a->output.port = htons(odp_port_to_ofp_port(dp_ifidx));
- a++;
-
/* Send packet. */
n_packets++;
- retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions,
- &packet);
+ retval = ofproto_send_packet(br->ofproto, dp_ifidx, e->vlan, &packet);
if (retval) {
error = retval;
n_errors++;
ds_put_cstr(&ds, "bridge\tbond\ttype\tslaves\n");
LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ struct port *port;
- for (i = 0; i < br->n_ports; i++) {
- const struct port *port = br->ports[i];
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
if (port->n_ifaces > 1) {
- size_t j;
+ struct iface *iface;
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) {
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ if (&iface->port_elem != list_front(&port->ifaces)) {
ds_put_cstr(&ds, ", ");
}
ds_put_cstr(&ds, iface->name);
const struct bridge *br;
LIST_FOR_EACH (br, node, &all_bridges) {
- size_t i;
+ struct port *port;
- for (i = 0; i < br->n_ports; i++) {
- struct port *port = br->ports[i];
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
if (!strcmp(port->name, name) && port->n_ifaces > 1) {
return port;
}
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)
{
struct ds ds = DS_EMPTY_INITIALIZER;
const struct port *port;
- size_t j;
+ struct iface *iface;
port = bond_find(args);
if (!port) {
if (port->lacp) {
ds_put_format(&ds, "lacp: %s\n",
- port->lacp & LACP_ACTIVE ? "active" : "passive");
+ port->lacp_active ? "active" : "passive");
} else {
ds_put_cstr(&ds, "lacp: off\n");
}
ds_put_format(&ds, "bond-detect-mode: %s\n",
- port->miimon ? "miimon" : "carrier");
+ port->monitor ? "carrier" : "miimon");
- if (port->miimon) {
+ if (!port->monitor) {
ds_put_format(&ds, "bond-miimon-interval: %lld\n",
- port->bond_miimon_interval);
+ port->miimon_interval);
}
ds_put_format(&ds, "updelay: %d ms\n", port->updelay);
port->bond_next_rebalance - time_msec());
}
- for (j = 0; j < port->n_ifaces; j++) {
- const struct iface *iface = port->ifaces[j];
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
struct bond_entry *be;
struct flow flow;
/* Basic info. */
ds_put_format(&ds, "\nslave %s: %s\n",
iface->name, iface->enabled ? "enabled" : "disabled");
- if (j == port->active_iface) {
+ if (iface == port->active_iface) {
ds_put_cstr(&ds, "\tactive slave\n");
}
if (iface->delay_expires != LLONG_MAX) {
iface->delay_expires - time_msec());
}
- if (port->lacp) {
- ds_put_cstr(&ds, "\tstatus: ");
-
- if (iface->lacp_status & LACP_CURRENT) {
- ds_put_cstr(&ds, "current ");
- }
-
- if (iface->lacp_status & LACP_EXPIRED) {
- ds_put_cstr(&ds, "expired ");
- }
-
- if (iface->lacp_status & LACP_DEFAULTED) {
- ds_put_cstr(&ds, "defaulted ");
- }
-
- if (iface->lacp_status & 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");
- }
-
if (port->bond_mode == BM_AB) {
continue;
}
int hash = be - port->bond_hash;
struct mac_entry *me;
- if (be->iface_idx != j) {
+ if (be->iface != iface) {
continue;
}
memcpy(flow.dl_src, me->mac, ETH_ADDR_LEN);
if (bond_hash_src(me->mac, me->vlan) == hash
- && me->port != port->port_idx
+ && me->port.p != port
&& choose_output_iface(port, &flow, me->vlan,
&dp_ifidx, &tags)
&& dp_ifidx == iface->dp_ifidx)
}
entry = &port->bond_hash[hash];
- ofproto_revalidate(port->bridge->ofproto, entry->iface_tag);
- entry->iface_idx = iface->port_ifidx;
- entry->iface_tag = tag_create_random();
+ ofproto_revalidate(port->bridge->ofproto, entry->tag);
+ entry->iface = iface;
+ entry->tag = tag_create_random();
unixctl_command_reply(conn, 200, "migrated");
}
return;
}
- if (port->active_iface != iface->port_ifidx) {
- ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
- port->active_iface = iface->port_ifidx;
- port->active_iface_tag = tag_create_random();
+ if (port->active_iface != iface) {
+ ofproto_revalidate(port->bridge->ofproto,
+ port_get_active_iface_tag(port));
+ port->active_iface = iface;
VLOG_INFO("port %s: active interface is now %s",
port->name, iface->name);
bond_send_learning_packets(port);
\f
/* Port functions. */
+static void
+lacp_send_pdu_cb(void *aux, const struct lacp_pdu *pdu)
+{
+ struct iface *iface = aux;
+ uint8_t ea[ETH_ADDR_LEN];
+ int error;
+
+ error = netdev_get_etheraddr(iface->netdev, ea);
+ if (!error) {
+ struct ofpbuf packet;
+ struct lacp_pdu *packet_pdu;
+
+ ofpbuf_init(&packet, 0);
+ packet_pdu = compose_packet(&packet, eth_addr_lacp, ea, ETH_TYPE_LACP,
+ sizeof *packet_pdu);
+ memcpy(packet_pdu, pdu, sizeof *packet_pdu);
+ ofproto_send_packet(iface->port->bridge->ofproto,
+ iface->dp_ifidx, 0, &packet);
+ ofpbuf_uninit(&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));
+ }
+}
+
+static void
+port_run(struct port *port)
+{
+ if (port->monitor) {
+ char *devname;
+
+ /* 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) {
+ iface_update_carrier(iface);
+ }
+ free(devname);
+ }
+ } else if (time_msec() >= port->miimon_next_update) {
+ struct iface *iface;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ iface_update_carrier(iface);
+ }
+ port->miimon_next_update = time_msec() + port->miimon_interval;
+ }
+
+ if (port->lacp) {
+ struct iface *iface;
+
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ lacp_slave_enable(port->lacp, iface, iface->enabled);
+ }
+
+ lacp_run(port->lacp, lacp_send_pdu_cb);
+ }
+
+ bond_run(port);
+}
+
+static void
+port_wait(struct port *port)
+{
+ if (port->monitor) {
+ netdev_monitor_poll_wait(port->monitor);
+ } else {
+ poll_timer_wait_until(port->miimon_next_update);
+ }
+
+ if (port->lacp) {
+ lacp_wait(port->lacp);
+ }
+
+ bond_wait(port);
+}
+
static struct port *
port_create(struct bridge *br, const char *name)
{
port = xzalloc(sizeof *port);
port->bridge = br;
- port->port_idx = br->n_ports;
port->vlan = -1;
port->trunks = NULL;
port->name = xstrdup(name);
- port->active_iface = -1;
+ port->active_iface = NULL;
+ list_init(&port->ifaces);
- if (br->n_ports >= br->allocated_ports) {
- br->ports = x2nrealloc(br->ports, &br->allocated_ports,
- sizeof *br->ports);
- }
- br->ports[br->n_ports++] = port;
- shash_add_assert(&br->port_by_name, port->name, port);
+ hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
VLOG_INFO("created port %s on bridge %s", port->name, br->name);
bridge_flush(br);
static void
port_del_ifaces(struct port *port, const struct ovsrec_port *cfg)
{
+ struct iface *iface, *next;
struct shash new_ifaces;
size_t i;
}
/* Get rid of deleted interfaces. */
- for (i = 0; i < port->n_ifaces; ) {
- if (!shash_find(&new_ifaces, cfg->interfaces[i]->name)) {
- iface_destroy(port->ifaces[i]);
- } else {
- i++;
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ if (!shash_find(&new_ifaces, iface->name)) {
+ iface_destroy(iface);
}
}
shash_destroy(&new_ifaces);
}
+/* Expires all MAC learning entries associated with 'port' and forces ofproto
+ * to revalidate every flow. */
+static void
+port_flush_macs(struct port *port)
+{
+ struct bridge *br = port->bridge;
+ struct mac_learning *ml = br->ml;
+ struct mac_entry *mac, *next_mac;
+
+ bridge_flush(br);
+ LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
+ if (mac->port.p == port) {
+ mac_learning_expire(ml, mac);
+ }
+ }
+}
+
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, miimon_next_update, lacp_priority;
+ bool need_flush = false;
unsigned long *trunks;
int vlan;
size_t i;
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);
+ netdev_monitor_destroy(port->monitor);
+ port->monitor = NULL;
+
+ if (strcmp(detect_mode, "miimon")) {
+ port->monitor = netdev_monitor_create();
+
+ if (strcmp(detect_mode, "carrier")) {
+ VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
+ "defaulting to carrier", port->name, detect_mode);
+ }
}
- port->bond_miimon_interval = atoi(
+ port->miimon_interval = atoi(
get_port_other_config(cfg, "bond-miimon-interval", "200"));
- if (port->bond_miimon_interval < 100) {
- port->bond_miimon_interval = 100;
+ if (port->miimon_interval < 100) {
+ port->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;
+ miimon_next_update = time_msec() + port->miimon_interval;
+ if (port->miimon_next_update > miimon_next_update) {
+ port->miimon_next_update = miimon_next_update;
}
if (!port->cfg->bond_mode ||
}
shash_destroy(&new_ifaces);
+ port->lacp_fast = !strcmp(get_port_other_config(cfg, "lacp-time", "slow"),
+ "fast");
+
lacp_priority =
atoi(get_port_other_config(cfg, "lacp-system-priority", "0"));
if (!port->cfg->lacp) {
/* XXX when LACP implementation has been sufficiently tested, enable by
* default and make active on bonded ports. */
- port->lacp = 0;
+ lacp_destroy(port->lacp);
+ port->lacp = NULL;
} else if (!strcmp(port->cfg->lacp, "off")) {
- port->lacp = 0;
+ lacp_destroy(port->lacp);
+ port->lacp = NULL;
} else if (!strcmp(port->cfg->lacp, "active")) {
- port->lacp = LACP_ACTIVE;
+ if (!port->lacp) {
+ port->lacp = lacp_create();
+ }
+ port->lacp_active = true;
} else if (!strcmp(port->cfg->lacp, "passive")) {
- port->lacp = LACP_PASSIVE;
+ if (!port->lacp) {
+ port->lacp = lacp_create();
+ }
+ port->lacp_active = false;
} else {
VLOG_WARN("port %s: unknown LACP mode %s",
port->name, port->cfg->lacp);
- port->lacp = 0;
+ lacp_destroy(port->lacp);
+ port->lacp = NULL;
}
/* Get VLAN tag. */
}
if (port->vlan != vlan) {
port->vlan = vlan;
- bridge_flush(port->bridge);
+ need_flush = true;
}
/* Get trunked VLANs. */
if (trunks == NULL
? port->trunks != NULL
: port->trunks == NULL || !bitmap_equal(trunks, port->trunks, 4096)) {
- bridge_flush(port->bridge);
+ need_flush = true;
}
bitmap_free(port->trunks);
port->trunks = trunks;
+
+ if (need_flush) {
+ port_flush_macs(port);
+ }
}
static void
{
if (port) {
struct bridge *br = port->bridge;
- struct port *del;
+ struct iface *iface, *next;
int i;
for (i = 0; i < MAX_MIRRORS; i++) {
}
}
- while (port->n_ifaces > 0) {
- iface_destroy(port->ifaces[port->n_ifaces - 1]);
+ LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
+ iface_destroy(iface);
}
- shash_find_and_delete_assert(&br->port_by_name, port->name);
-
- del = br->ports[port->port_idx] = br->ports[--br->n_ports];
- del->port_idx = port->port_idx;
+ hmap_remove(&br->ports, &port->hmap_node);
VLOG_INFO("destroyed port %s on bridge %s", port->name, br->name);
+ port_flush_macs(port);
+
+ lacp_destroy(port->lacp);
netdev_monitor_destroy(port->monitor);
- free(port->ifaces);
bitmap_free(port->trunks);
+ free(port->bond_hash);
free(port->name);
free(port);
- bridge_flush(br);
}
}
static struct port *
port_lookup(const struct bridge *br, const char *name)
{
- return shash_find_data(&br->port_by_name, name);
+ struct port *port;
+
+ HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
+ &br->ports) {
+ if (!strcmp(port->name, name)) {
+ return port;
+ }
+ }
+ return NULL;
}
static struct iface *
static void
port_update_lacp(struct port *port)
{
- size_t i;
- bool key_changed;
-
- if (!port->lacp || port->n_ifaces < 1) {
- for (i = 0; i < port->n_ifaces; i++) {
- iface_set_lacp_defaulted(port->ifaces[i]);
- }
- return;
- }
-
- key_changed = true;
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
+ if (port->lacp) {
+ struct iface *iface;
- if (iface->dp_ifidx <= 0 || iface->dp_ifidx > UINT16_MAX) {
- port->lacp = 0;
- return;
- }
+ lacp_configure(port->lacp, port->name,
+ port->bridge->ea, port->lacp_priority,
+ port->lacp_active, port->lacp_fast);
- if (iface->dp_ifidx == port->lacp_key) {
- key_changed = false;
+ LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
+ lacp_slave_register(port->lacp, iface, iface->name,
+ iface->dp_ifidx, iface->lacp_priority);
}
}
-
- 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. */
free(port->bond_hash);
port->bond_hash = NULL;
port->bond_fake_iface = false;
+ port->active_iface = NULL;
+ port->no_ifaces_tag = 0;
} else {
size_t i;
port->bond_hash = xcalloc(BOND_MASK + 1, sizeof *port->bond_hash);
for (i = 0; i <= BOND_MASK; i++) {
struct bond_entry *e = &port->bond_hash[i];
- e->iface_idx = -1;
+ e->iface = NULL;
e->tx_bytes = 0;
}
- port->no_ifaces_tag = tag_create_random();
- bond_choose_active_iface(port);
port->bond_next_rebalance
= time_msec() + port->bond_rebalance_interval;
-
- 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_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);
- }
+ if (!port->no_ifaces_tag) {
+ port->no_ifaces_tag = tag_create_random();
}
- }
-}
-\f
-/* Interface functions. */
-
-static void
-iface_set_lacp_defaulted(struct iface *iface)
-{
- memset(&iface->lacp_partner, 0, sizeof iface->lacp_partner);
-
- iface->lacp_status |= LACP_DEFAULTED;
- iface->lacp_status &= ~(LACP_CURRENT | LACP_EXPIRED);
- iface->lacp_tx = 0;
- iface->port->lacp_need_update = true;
-}
-
-static void
-iface_set_lacp_expired(struct iface *iface)
-{
- iface->lacp_status &= ~LACP_CURRENT;
- iface->lacp_status |= LACP_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_ATTACHED) {
- state |= LACP_STATE_SYNC;
- }
-
- if (iface->lacp_status & LACP_DEFAULTED) {
- state |= LACP_STATE_DEF;
- }
-
- if (iface->lacp_status & LACP_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));
+ if (!port->active_iface) {
+ bond_choose_active_iface(port);
+ }
- flow_extract(packet, 0, ODPP_NONE, &flow);
+ port->bond_fake_iface = port->cfg->bond_fake_iface;
+ if (port->bond_fake_iface) {
+ port->bond_next_fake_iface_update = time_msec();
+ }
- 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);
}
}
+\f
+/* Interface functions. */
static struct iface *
iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
iface = xzalloc(sizeof *iface);
iface->port = port;
- iface->port_ifidx = port->n_ifaces;
iface->name = xstrdup(name);
iface->dp_ifidx = -1;
iface->tag = tag_create_random();
iface->delay_expires = LLONG_MAX;
iface->netdev = NULL;
iface->cfg = if_cfg;
- iface_set_lacp_defaulted(iface);
-
- 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);
- }
- port->ifaces[port->n_ifaces++] = iface;
+ list_push_back(&port->ifaces, &iface->port_elem);
+ port->n_ifaces++;
+
if (port->n_ifaces > 1) {
br->has_bonded_ports = true;
}
if (iface) {
struct port *port = iface->port;
struct bridge *br = port->bridge;
- bool del_active = port->active_iface == iface->port_ifidx;
- struct iface *del;
+ bool del_active = port->active_iface == iface;
+
+ if (port->bond_hash) {
+ struct bond_entry *e;
+ for (e = port->bond_hash; e <= &port->bond_hash[BOND_MASK]; e++) {
+ if (e->iface == iface) {
+ e->iface = NULL;
+ }
+ }
+ }
- if (port->monitor) {
+ if (iface->port->lacp) {
+ lacp_slave_unregister(iface->port->lacp, iface);
+ }
+
+ if (port->monitor && iface->netdev) {
netdev_monitor_remove(port->monitor, iface->netdev);
}
hmap_remove(&br->ifaces, &iface->dp_ifidx_node);
}
- del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces];
- del->port_ifidx = iface->port_ifidx;
+ list_remove(&iface->port_elem);
+ port->n_ifaces--;
netdev_close(iface->netdev);
if (del_active) {
- ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
bond_choose_active_iface(port);
bond_send_learning_packets(port);
}
- cfm_destroy(iface->cfm);
-
free(iface->name);
free(iface);
return shash_find_data(&br->iface_by_name, name);
}
+static struct iface *
+iface_find(const char *name)
+{
+ const struct bridge *br;
+
+ LIST_FOR_EACH (br, node, &all_bridges) {
+ struct iface *iface = iface_lookup(br, name);
+
+ if (iface) {
+ return iface;
+ }
+ }
+ return NULL;
+}
+
static struct iface *
iface_from_dp_ifidx(const struct bridge *br, uint16_t dp_ifidx)
{
}
}
+static void
+iface_update_carrier(struct iface *iface)
+{
+ bool carrier = iface_get_carrier(iface);
+ if (carrier == iface->up) {
+ return;
+ }
+
+ iface->up = carrier;
+ if (iface->port->lacp) {
+ lacp_slave_carrier_changed(iface->port->lacp, iface);
+ }
+}
+
static void
iface_update_qos(struct iface *iface, const struct ovsrec_qos *qos)
{
iface_update_cfm(struct iface *iface)
{
size_t i;
- struct cfm *cfm;
+ struct cfm cfm;
uint16_t *remote_mps;
struct ovsrec_monitor *mon;
- uint8_t ea[ETH_ADDR_LEN], maid[CCM_MAID_LEN];
+ uint8_t maid[CCM_MAID_LEN];
mon = iface->cfg->monitor;
if (!mon) {
- cfm_destroy(iface->cfm);
- iface->cfm = NULL;
- return;
- }
-
- if (netdev_get_etheraddr(iface->netdev, ea)) {
- VLOG_WARN("interface %s: Failed to get ethernet address. "
- "Skipping Monitor.", iface->name);
+ ofproto_iface_clear_cfm(iface->port->bridge->ofproto, iface->dp_ifidx);
return;
}
return;
}
- if (!iface->cfm) {
- iface->cfm = cfm_create();
- }
-
- cfm = iface->cfm;
- cfm->mpid = mon->mpid;
- cfm->interval = mon->interval ? *mon->interval : 1000;
+ 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);
+ 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);
+
+ ofproto_iface_set_cfm(iface->port->bridge->ofproto, iface->dp_ifidx,
+ &cfm, remote_mps, mon->n_remote_mps);
free(remote_mps);
+}
- if (!cfm_configure(iface->cfm)) {
- cfm_destroy(iface->cfm);
- iface->cfm = NULL;
- }
+/* Read carrier or miimon status directly from 'iface''s netdev, according to
+ * how 'iface''s port is configured.
+ *
+ * Returns true if 'iface' is up, false otherwise. */
+static bool
+iface_get_carrier(const struct iface *iface)
+{
+ return (iface->port->monitor
+ ? netdev_get_carrier(iface->netdev)
+ : netdev_get_miimon(iface->netdev));
}
\f
/* Port mirroring. */
mirror_reconfigure(struct bridge *br)
{
unsigned long *rspan_vlans;
+ struct port *port;
int i;
/* Get rid of deleted mirrors. */
}
/* Update port reserved status. */
- for (i = 0; i < br->n_ports; i++) {
- br->ports[i]->is_mirror_output_port = false;
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ port->is_mirror_output_port = false;
}
for (i = 0; i < MAX_MIRRORS; i++) {
struct mirror *m = br->mirrors[i];
}
if (mac_learning_set_flood_vlans(br->ml, rspan_vlans)) {
bridge_flush(br);
+ mac_learning_flush(br->ml);
}
}
VLOG_INFO("created port mirror %s on bridge %s", cfg->name, br->name);
bridge_flush(br);
+ mac_learning_flush(br->ml);
br->mirrors[i] = m = xzalloc(sizeof *m);
m->bridge = br;
{
if (m) {
struct bridge *br = m->bridge;
- size_t i;
+ struct port *port;
- for (i = 0; i < br->n_ports; i++) {
- br->ports[i]->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
- br->ports[i]->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
+ HMAP_FOR_EACH (port, hmap_node, &br->ports) {
+ port->src_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
+ port->dst_mirrors &= ~(MIRROR_MASK_C(1) << m->idx);
}
shash_destroy(&m->src_ports);
free(m);
bridge_flush(br);
+ mac_learning_flush(br->ml);
}
}
struct shash src_ports, dst_ports;
mirror_mask_t mirror_bit;
struct port *out_port;
+ struct port *port;
int out_vlan;
size_t n_vlans;
int *vlans;
- size_t i;
/* Set name. */
if (strcmp(cfg->name, m->name)) {
shash_init(&src_ports);
shash_init(&dst_ports);
if (cfg->select_all) {
- for (i = 0; i < m->bridge->n_ports; i++) {
- const char *name = m->bridge->ports[i]->name;
- shash_add_once(&src_ports, name, NULL);
- shash_add_once(&dst_ports, name, NULL);
+ HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
+ shash_add_once(&src_ports, port->name, NULL);
+ shash_add_once(&dst_ports, port->name, NULL);
}
vlans = NULL;
n_vlans = 0;
|| m->out_port != out_port
|| m->out_vlan != out_vlan) {
bridge_flush(m->bridge);
+ mac_learning_flush(m->bridge->ml);
}
shash_swap(&m->src_ports, &src_ports);
shash_swap(&m->dst_ports, &dst_ports);
/* Update ports. */
mirror_bit = MIRROR_MASK_C(1) << m->idx;
- for (i = 0; i < m->bridge->n_ports; i++) {
- struct port *port = m->bridge->ports[i];
-
+ HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
if (shash_find(&m->src_ports, port->name)
|| (m->n_vlans
&& (!port->vlan