X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=vswitchd%2Fbridge.c;h=205577738fe58e0b187b7a8539c6d50bd24d275d;hb=5d9cb63c9160df34d8f282db0e933e3f021f7fb6;hp=10c6fee25e85ae9c28985791d7029b4b8a9b3a63;hpb=c93b1d6a4c7f96c5f75f7ec0972fe62e94d369dc;p=openvswitch
diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c
index 10c6fee2..20557773 100644
--- a/vswitchd/bridge.c
+++ b/vswitchd/bridge.c
@@ -1,28 +1,16 @@
/* Copyright (c) 2008, 2009 Nicira Networks
*
- * This program is free software: you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation, either version 3 of the License, or
- * (at your option) any later version.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
*
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program. If not, see .
- *
- * In addition, as a special exception, Nicira Networks gives permission
- * to link the code of its release of vswitchd with the OpenSSL project's
- * "OpenSSL" library (or with modified versions of it that use the same
- * license as the "OpenSSL" library), and distribute the linked
- * executables. You must obey the GNU General Public License in all
- * respects for all of the code used other than "OpenSSL". If you modify
- * this file, you may extend this exception to your version of the file,
- * but you are not obligated to do so. If you do not wish to do so,
- * delete this exception statement from your version.
+ * http://www.apache.org/licenses/LICENSE-2.0
*
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
*/
#include
@@ -39,9 +27,9 @@
#include
#include
#include
+#include
#include
#include "bitmap.h"
-#include "cfg.h"
#include "coverage.h"
#include "dirs.h"
#include "dpif.h"
@@ -54,20 +42,23 @@
#include "odp-util.h"
#include "ofp-print.h"
#include "ofpbuf.h"
+#include "ofproto/netflow.h"
+#include "ofproto/ofproto.h"
#include "packets.h"
#include "poll-loop.h"
#include "port-array.h"
#include "proc-net-compat.h"
#include "process.h"
-#include "secchan/ofproto.h"
+#include "sha1.h"
+#include "shash.h"
#include "socket-util.h"
-#include "stp.h"
#include "svec.h"
#include "timeval.h"
#include "util.h"
#include "unixctl.h"
#include "vconn.h"
#include "vconn-ssl.h"
+#include "vswitchd/vswitch-idl.h"
#include "xenserver.h"
#include "xtoxll.h"
@@ -79,20 +70,22 @@ struct dst {
uint16_t dp_ifidx;
};
-extern uint64_t mgmt_id;
-
struct iface {
+ /* These members are always valid. */
struct port *port; /* Containing port. */
size_t port_ifidx; /* Index within containing port. */
-
char *name; /* Host network device name. */
- int dp_ifidx; /* Index within kernel datapath. */
-
- uint8_t mac[ETH_ADDR_LEN]; /* Ethernet address (all zeros if unknowns). */
-
tag_type tag; /* Tag associated with this interface. */
- bool enabled; /* May be chosen for flows? */
long long delay_expires; /* Time after which 'enabled' may change. */
+
+ /* These members are valid only after bridge_reconfigure() causes them to
+ * be initialized.*/
+ int dp_ifidx; /* Index within kernel datapath. */
+ struct netdev *netdev; /* Network device. */
+ bool enabled; /* May be chosen for flows? */
+
+ /* This member is only valid *during* bridge_reconfigure(). */
+ const struct ovsrec_interface *cfg;
};
#define BOND_MASK 0xff
@@ -141,22 +134,23 @@ struct port {
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? */
/* Port mirroring info. */
mirror_mask_t src_mirrors; /* Mirrors triggered when packet received. */
mirror_mask_t dst_mirrors; /* Mirrors triggered when packet sent. */
bool is_mirror_output_port; /* Does port mirroring send frames here? */
- /* Spanning tree info. */
- enum stp_state stp_state; /* Always STP_FORWARDING if STP not in use. */
- tag_type stp_state_tag; /* Tag for STP state change. */
+ /* This member is only valid *during* bridge_reconfigure(). */
+ const struct ovsrec_port *cfg;
};
#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, or null not to learn. */
+ struct mac_learning *ml; /* MAC learning table. */
bool sent_config_request; /* Successfully sent config request? */
uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
@@ -169,7 +163,7 @@ struct bridge {
struct ofproto *ofproto; /* OpenFlow switch. */
/* Kernel datapath information. */
- struct dpif dpif; /* Kernel datapath. */
+ struct dpif *dpif; /* Datapath. */
struct port_array ifaces; /* Indexed by kernel datapath port number. */
/* Bridge ports. */
@@ -189,9 +183,8 @@ struct bridge {
/* Port mirroring. */
struct mirror *mirrors[MAX_MIRRORS];
- /* Spanning tree. */
- struct stp *stp;
- long long int stp_last_tick;
+ /* This member is only valid *during* bridge_reconfigure(). */
+ const struct ovsrec_bridge *cfg;
};
/* List of all bridges. */
@@ -203,28 +196,35 @@ enum { DP_MAX = 256 };
static struct bridge *bridge_create(const char *name);
static void bridge_destroy(struct bridge *);
static struct bridge *bridge_lookup(const char *name);
+static unixctl_cb_func bridge_unixctl_dump_flows;
static int bridge_run_one(struct bridge *);
-static void bridge_reconfigure_one(struct bridge *);
-static void bridge_reconfigure_controller(struct bridge *);
-static void bridge_get_all_ifaces(const struct bridge *, struct svec *ifaces);
+static void bridge_reconfigure_one(const struct ovsrec_open_vswitch *,
+ struct bridge *);
+static void bridge_reconfigure_controller(const struct ovsrec_open_vswitch *,
+ struct bridge *);
+static void bridge_get_all_ifaces(const struct bridge *, struct shash *ifaces);
static void bridge_fetch_dp_ifaces(struct bridge *);
static void bridge_flush(struct bridge *);
static void bridge_pick_local_hw_addr(struct bridge *,
uint8_t ea[ETH_ADDR_LEN],
- const char **devname);
+ struct iface **hw_addr_iface);
static uint64_t bridge_pick_datapath_id(struct bridge *,
const uint8_t bridge_ea[ETH_ADDR_LEN],
- const char *devname);
+ struct iface *hw_addr_iface);
+static 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 bond_init(void);
static void bond_run(struct bridge *);
static void bond_wait(struct bridge *);
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_create(struct bridge *, const char *name);
-static void port_reconfigure(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_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);
@@ -232,23 +232,29 @@ static struct port *port_from_dp_ifidx(const struct bridge *,
uint16_t dp_ifidx);
static void port_update_bond_compat(struct port *);
static void port_update_vlan_compat(struct port *);
+static void port_update_bonding(struct port *);
+#if 0
static void mirror_create(struct bridge *, const char *name);
static void mirror_destroy(struct mirror *);
static void mirror_reconfigure(struct bridge *);
static void mirror_reconfigure_one(struct mirror *);
static bool vlan_is_mirrored(const struct mirror *, int vlan);
+#else
+static bool vlan_is_mirrored(const struct mirror *m UNUSED, int vlan UNUSED)
+{
+ return false;
+}
+#endif
-static void brstp_reconfigure(struct bridge *);
-static void brstp_adjust_timers(struct bridge *);
-static void brstp_run(struct bridge *);
-static void brstp_wait(struct bridge *);
-
-static void iface_create(struct port *, const char *name);
+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 *);
/* Hooks into ofproto processing. */
static struct ofhooks bridge_ofhooks;
@@ -270,8 +276,8 @@ bridge_get_ifaces(struct svec *svec)
for (j = 0; j < port->n_ifaces; j++) {
struct iface *iface = port->ifaces[j];
if (iface->dp_ifidx < 0) {
- VLOG_ERR("%s interface not in dp%u, ignoring",
- iface->name, dpif_id(&br->dpif));
+ VLOG_ERR("%s interface not in datapath %s, ignoring",
+ iface->name, dpif_name(br->dpif));
} else {
if (iface->dp_ifidx != ODPP_LOCAL) {
svec_add(svec, iface->name);
@@ -282,42 +288,59 @@ bridge_get_ifaces(struct svec *svec)
}
}
-/* The caller must already have called cfg_read(). */
void
-bridge_init(void)
+bridge_init(const struct ovsrec_open_vswitch *cfg)
{
- int retval;
- int i;
+ struct svec bridge_names;
+ struct svec dpif_names;
+ size_t i;
- bond_init();
+ unixctl_command_register("fdb/show", bridge_unixctl_fdb_show, NULL);
+
+ svec_init(&bridge_names);
+ for (i = 0; i < cfg->n_bridges; i++) {
+ svec_add(&bridge_names, cfg->bridges[i]->name);
+ }
+ svec_sort(&bridge_names);
- for (i = 0; i < DP_MAX; i++) {
- struct dpif dpif;
- char devname[16];
+ svec_init(&dpif_names);
+ dp_enumerate(&dpif_names);
+ for (i = 0; i < dpif_names.n; i++) {
+ const char *dpif_name = dpif_names.names[i];
+ struct dpif *dpif;
+ int retval;
- sprintf(devname, "dp%d", i);
- retval = dpif_open(devname, &dpif);
+ retval = dpif_open(dpif_name, &dpif);
if (!retval) {
- char dpif_name[IF_NAMESIZE];
- if (dpif_get_name(&dpif, dpif_name, sizeof dpif_name)
- || !cfg_has("bridge.%s.port", dpif_name)) {
- dpif_delete(&dpif);
+ struct svec all_names;
+ size_t j;
+
+ svec_init(&all_names);
+ dpif_get_all_names(dpif, &all_names);
+ for (j = 0; j < all_names.n; j++) {
+ if (svec_contains(&bridge_names, all_names.names[j])) {
+ goto found;
+ }
}
- dpif_close(&dpif);
- } else if (retval != ENODEV) {
- VLOG_ERR("failed to delete datapath dp%d: %s",
- i, strerror(retval));
+ dpif_delete(dpif);
+ found:
+ svec_destroy(&all_names);
+ dpif_close(dpif);
}
}
+ svec_destroy(&dpif_names);
- bridge_reconfigure();
+ unixctl_command_register("bridge/dump-flows", bridge_unixctl_dump_flows,
+ NULL);
+
+ bond_init();
+ bridge_reconfigure(cfg);
}
#ifdef HAVE_OPENSSL
static bool
-config_string_change(const char *key, char **valuep)
+config_string_change(const char *value, char **valuep)
{
- const char *value = cfg_get_string(0, "%s", key);
if (value && (!*valuep || strcmp(value, *valuep))) {
free(*valuep);
*valuep = xstrdup(value);
@@ -328,87 +351,222 @@ config_string_change(const char *key, char **valuep)
}
static void
-bridge_configure_ssl(void)
+bridge_configure_ssl(const struct ovsrec_ssl *ssl)
{
/* XXX SSL should be configurable on a per-bridge basis.
* XXX should be possible to de-configure SSL. */
static char *private_key_file;
static char *certificate_file;
static char *cacert_file;
+ struct stat s;
+
+ if (!ssl) {
+ /* XXX We can't un-set SSL settings. */
+ return;
+ }
- if (config_string_change("ssl.private-key", &private_key_file)) {
+ if (config_string_change(ssl->private_key, &private_key_file)) {
vconn_ssl_set_private_key_file(private_key_file);
}
- if (config_string_change("ssl.certificate", &certificate_file)) {
+ if (config_string_change(ssl->certificate, &certificate_file)) {
vconn_ssl_set_certificate_file(certificate_file);
}
- if (config_string_change("ssl.ca-cert", &cacert_file)) {
- vconn_ssl_set_ca_cert_file(cacert_file,
- cfg_get_bool(0, "ssl.bootstrap-ca-cert"));
+ /* We assume that even if the filename hasn't changed, if the CA cert
+ * file has been removed, that we want to move back into
+ * boot-strapping mode. This opens a small security hole, because
+ * the old certificate will still be trusted until vSwitch is
+ * restarted. We may want to address this in vconn's SSL library. */
+ if (config_string_change(ssl->ca_cert, &cacert_file)
+ || (cacert_file && stat(cacert_file, &s) && errno == ENOENT)) {
+ vconn_ssl_set_ca_cert_file(cacert_file, 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, bool create)
+{
+ struct shash_node *node;
+ struct shash options;
+ int error;
+ size_t i;
+
+ /* If a type is not explicitly declared, then assume it's an existing
+ * "system" device. */
+ if (iface_cfg->type[0] == '\0' || !strcmp(iface_cfg->type, "system")) {
+ return 0;
+ }
+
+ 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) {
+ error = netdev_create(iface_cfg->name, iface_cfg->type, &options);
+ } else {
+ /* xxx Check to make sure that the type hasn't changed. */
+ error = netdev_reconfigure(iface_cfg->name, &options);
+ }
+
+ SHASH_FOR_EACH (node, &options) {
+ free(node->data);
+ }
+ shash_destroy(&options);
+
+ return error;
+}
+
+static int
+reconfigure_iface(const struct ovsrec_interface *iface_cfg)
+{
+ return set_up_iface(iface_cfg, false);
+}
+
+
+/* iterate_and_prune_ifaces() callback function that opens the network device
+ * for 'iface', if it is not already open, and retrieves the interface's MAC
+ * address and carrier status. */
+static bool
+init_iface_netdev(struct bridge *br UNUSED, struct iface *iface,
+ void *aux UNUSED)
+{
+ if (iface->netdev) {
+ return true;
+ } else if (!netdev_open(iface->name, NETDEV_ETH_TYPE_NONE,
+ &iface->netdev)) {
+ netdev_get_carrier(iface->netdev, &iface->enabled);
+ return true;
+ } else {
+ /* If the network device can't be opened, then we're not going to try
+ * to do anything with this interface. */
+ return false;
+ }
+}
+
+static bool
+check_iface_dp_ifidx(struct bridge *br, struct iface *iface, void *aux 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 {
+ VLOG_ERR("%s interface not in %s, dropping",
+ iface->name, dpif_name(br->dpif));
+ return false;
+ }
+}
+
+static bool
+set_iface_properties(struct bridge *br UNUSED, struct iface *iface,
+ void *aux UNUSED)
+{
+ /* Set policing attributes. */
+ netdev_set_policing(iface->netdev,
+ iface->cfg->ingress_policing_rate,
+ iface->cfg->ingress_policing_burst);
+
+ /* Set MAC address of internal interfaces other than the local
+ * interface. */
+ if (iface->dp_ifidx != ODPP_LOCAL
+ && iface_is_internal(br, iface->name)) {
+ iface_set_mac(iface);
+ }
+
+ return true;
+}
+
+/* Calls 'cb' for each interfaces in 'br', passing along the 'aux' argument.
+ * Deletes from 'br' all the interfaces for which 'cb' returns false, and then
+ * deletes from 'br' any ports that no longer have any interfaces. */
+static void
+iterate_and_prune_ifaces(struct bridge *br,
+ bool (*cb)(struct bridge *, struct iface *,
+ void *aux),
+ void *aux)
+{
+ size_t i, j;
+
+ for (i = 0; i < br->n_ports; ) {
+ struct port *port = br->ports[i];
+ for (j = 0; j < port->n_ifaces; ) {
+ struct iface *iface = port->ifaces[j];
+ if (cb(br, iface, aux)) {
+ j++;
+ } else {
+ iface_destroy(iface);
+ }
+ }
+
+ if (port->n_ifaces) {
+ i++;
+ } else {
+ VLOG_ERR("%s port has no interfaces, dropping", port->name);
+ port_destroy(port);
+ }
+ }
+}
+
void
-bridge_reconfigure(void)
+bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
{
- struct svec old_br, new_br, raw_new_br;
+ struct shash old_br, new_br;
+ struct shash_node *node;
struct bridge *br, *next;
- size_t i, j;
+ size_t i;
COVERAGE_INC(bridge_reconfigure);
- /* Collect old bridges. */
- svec_init(&old_br);
+ /* Collect old and new bridges. */
+ shash_init(&old_br);
+ shash_init(&new_br);
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- svec_add(&old_br, br->name);
- }
-
- /* Collect new bridges. */
- svec_init(&raw_new_br);
- cfg_get_subsections(&raw_new_br, "bridge");
- svec_init(&new_br);
- for (i = 0; i < raw_new_br.n; i++) {
- const char *name = raw_new_br.names[i];
- if ((!strncmp(name, "dp", 2) && isdigit(name[2])) ||
- (!strncmp(name, "nl:", 3) && isdigit(name[3]))) {
- VLOG_ERR("%s is not a valid bridge name (bridges may not be "
- "named \"dp\" or \"nl:\" followed by a digit)", name);
- } else {
- svec_add(&new_br, name);
+ shash_add(&old_br, br->name, br);
+ }
+ for (i = 0; i < ovs_cfg->n_bridges; i++) {
+ const struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
+ if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
+ VLOG_WARN("more than one bridge named %s", br_cfg->name);
}
}
- svec_destroy(&raw_new_br);
/* Get rid of deleted bridges and add new bridges. */
- svec_sort(&old_br);
- svec_sort(&new_br);
- assert(svec_is_unique(&old_br));
- assert(svec_is_unique(&new_br));
LIST_FOR_EACH_SAFE (br, next, struct bridge, node, &all_bridges) {
- if (!svec_contains(&new_br, br->name)) {
+ struct ovsrec_bridge *br_cfg = shash_find_data(&new_br, br->name);
+ if (br_cfg) {
+ br->cfg = br_cfg;
+ } else {
bridge_destroy(br);
}
}
- for (i = 0; i < new_br.n; i++) {
- const char *name = new_br.names[i];
- if (!svec_contains(&old_br, name)) {
- bridge_create(name);
+ SHASH_FOR_EACH (node, &new_br) {
+ const char *br_name = node->name;
+ const struct ovsrec_bridge *br_cfg = node->data;
+ if (!shash_find_data(&old_br, br_name)) {
+ br = bridge_create(br_name);
+ if (br) {
+ br->cfg = br_cfg;
+ }
}
}
- svec_destroy(&old_br);
- svec_destroy(&new_br);
+ shash_destroy(&old_br);
+ shash_destroy(&new_br);
#ifdef HAVE_OPENSSL
/* Configure SSL. */
- bridge_configure_ssl();
+ bridge_configure_ssl(ovs_cfg->ssl);
#endif
/* Reconfigure all bridges. */
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- bridge_reconfigure_one(br);
+ bridge_reconfigure_one(ovs_cfg, br);
}
/* Add and delete ports on all datapaths.
@@ -419,107 +577,92 @@ bridge_reconfigure(void)
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
struct odp_port *dpif_ports;
size_t n_dpif_ports;
- struct svec want_ifaces;
+ struct shash want_ifaces;
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
+ dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports);
bridge_get_all_ifaces(br, &want_ifaces);
for (i = 0; i < n_dpif_ports; i++) {
const struct odp_port *p = &dpif_ports[i];
- if (!svec_contains(&want_ifaces, p->devname)
+ if (!shash_find(&want_ifaces, p->devname)
&& strcmp(p->devname, br->name)) {
- int retval = dpif_port_del(&br->dpif, p->port);
+ int retval = dpif_port_del(br->dpif, p->port);
if (retval) {
- VLOG_ERR("failed to remove %s interface from dp%u: %s",
- p->devname, dpif_id(&br->dpif), strerror(retval));
+ VLOG_ERR("failed to remove %s interface from %s: %s",
+ p->devname, dpif_name(br->dpif),
+ strerror(retval));
}
}
}
- svec_destroy(&want_ifaces);
+ 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;
- struct svec cur_ifaces, want_ifaces, add_ifaces;
- int next_port_no;
+ struct shash cur_ifaces, want_ifaces;
+ struct shash_node *node;
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
- svec_init(&cur_ifaces);
+ /* 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++) {
- svec_add(&cur_ifaces, dpif_ports[i].devname);
+ const char *name = dpif_ports[i].devname;
+ if (!shash_find(&cur_ifaces, name)) {
+ shash_add(&cur_ifaces, name, NULL);
+ }
}
free(dpif_ports);
- svec_sort_unique(&cur_ifaces);
+
+ /* Get the set of interfaces we want on this datapath. */
bridge_get_all_ifaces(br, &want_ifaces);
- svec_diff(&want_ifaces, &cur_ifaces, &add_ifaces, NULL, NULL);
-
- next_port_no = 1;
- for (i = 0; i < add_ifaces.n; i++) {
- const char *if_name = add_ifaces.names[i];
- for (;;) {
- int internal = cfg_get_bool(0, "iface.%s.internal", if_name);
- int error = dpif_port_add(&br->dpif, if_name, next_port_no++,
- internal ? ODP_PORT_INTERNAL : 0);
- if (error != EEXIST) {
- if (next_port_no >= 256) {
- VLOG_ERR("ran out of valid port numbers on dp%u",
- dpif_id(&br->dpif));
- goto out;
- }
- if (error) {
- VLOG_ERR("failed to add %s interface to dp%u: %s",
- if_name, dpif_id(&br->dpif), strerror(error));
- }
+
+ SHASH_FOR_EACH (node, &want_ifaces) {
+ const char *if_name = node->name;
+ struct iface *iface = node->data;
+
+ if (shash_find(&cur_ifaces, if_name)) {
+ /* Already exists, just reconfigure it. */
+ if (iface) {
+ reconfigure_iface(iface->cfg);
+ }
+ } 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));
}
}
}
- out:
- svec_destroy(&cur_ifaces);
- svec_destroy(&want_ifaces);
- svec_destroy(&add_ifaces);
+ shash_destroy(&cur_ifaces);
+ shash_destroy(&want_ifaces);
}
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
uint8_t ea[8];
uint64_t dpid;
- struct iface *local_iface = NULL;
- const char *devname;
- uint8_t engine_type = br->dpif.minor;
- uint8_t engine_id = br->dpif.minor;
- bool add_id_to_iface = false;
- struct svec nf_hosts;
+ struct iface *local_iface;
+ struct iface *hw_addr_iface;
bridge_fetch_dp_ifaces(br);
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
+ iterate_and_prune_ifaces(br, init_iface_netdev, NULL);
- for (j = 0; j < port->n_ifaces; ) {
- struct iface *iface = port->ifaces[j];
- if (iface->dp_ifidx < 0) {
- VLOG_ERR("%s interface not in dp%u, dropping",
- iface->name, dpif_id(&br->dpif));
- iface_destroy(iface);
- } else {
- if (iface->dp_ifidx == ODPP_LOCAL) {
- local_iface = iface;
- }
- VLOG_DBG("dp%u has interface %s on port %d",
- dpif_id(&br->dpif), iface->name, iface->dp_ifidx);
- j++;
- }
- }
- if (!port->n_ifaces) {
- VLOG_ERR("%s port has no interfaces, dropping", port->name);
- port_destroy(port);
- continue;
- }
- i++;
- }
+ iterate_and_prune_ifaces(br, check_iface_dp_ifidx, NULL);
/* Pick local port hardware address, datapath ID. */
- bridge_pick_local_hw_addr(br, ea, &devname);
+ bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
+ local_iface = bridge_get_local_iface(br);
if (local_iface) {
- int error = netdev_nodev_set_etheraddr(local_iface->name, ea);
+ int error = netdev_set_etheraddr(local_iface->netdev, ea);
if (error) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
@@ -528,37 +671,56 @@ bridge_reconfigure(void)
}
}
- dpid = bridge_pick_datapath_id(br, ea, devname);
+ dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
ofproto_set_datapath_id(br->ofproto, dpid);
/* Set NetFlow configuration on this bridge. */
- if (cfg_has("netflow.%s.engine-type", br->name)) {
- engine_type = cfg_get_int(0, "netflow.%s.engine-type",
- br->name);
- }
- if (cfg_has("netflow.%s.engine-id", br->name)) {
- engine_id = cfg_get_int(0, "netflow.%s.engine-id", br->name);
- }
- if (cfg_has("netflow.%s.add-id-to-iface", br->name)) {
- add_id_to_iface = cfg_get_bool(0, "netflow.%s.add-id-to-iface",
- br->name);
- }
- if (add_id_to_iface && engine_id > 0x7f) {
- VLOG_WARN("bridge %s: netflow port mangling may conflict with "
- "another vswitch, choose an engine id less than 128",
- br->name);
- }
- if (add_id_to_iface && br->n_ports > 0x1ff) {
- VLOG_WARN("bridge %s: netflow port mangling will conflict with "
- "another port when 512 or more ports are used",
- br->name);
- }
- svec_init(&nf_hosts);
- cfg_get_all_keys(&nf_hosts, "netflow.%s.host", br->name);
- if (ofproto_set_netflow(br->ofproto, &nf_hosts, engine_type,
- engine_id, add_id_to_iface)) {
- VLOG_ERR("bridge %s: problem setting netflow collectors",
- br->name);
+ if (br->cfg->netflow) {
+ struct ovsrec_netflow *nf_cfg = br->cfg->netflow;
+ struct netflow_options opts;
+
+ memset(&opts, 0, sizeof opts);
+
+ dpif_get_netflow_ids(br->dpif, &opts.engine_type, &opts.engine_id);
+ if (nf_cfg->engine_type) {
+ opts.engine_type = nf_cfg->engine_type;
+ }
+ if (nf_cfg->engine_id) {
+ opts.engine_id = nf_cfg->engine_id;
+ }
+
+ opts.active_timeout = nf_cfg->active_timeout;
+ if (!opts.active_timeout) {
+ opts.active_timeout = -1;
+ } else if (opts.active_timeout < 0) {
+ VLOG_WARN("bridge %s: active timeout interval set to negative "
+ "value, using default instead (%d seconds)", br->name,
+ NF_ACTIVE_TIMEOUT_DEFAULT);
+ opts.active_timeout = -1;
+ }
+
+ opts.add_id_to_iface = nf_cfg->add_id_to_interface;
+ if (opts.add_id_to_iface) {
+ if (opts.engine_id > 0x7f) {
+ VLOG_WARN("bridge %s: netflow port mangling may conflict "
+ "with another vswitch, choose an engine id less "
+ "than 128", br->name);
+ }
+ if (br->n_ports > 508) {
+ VLOG_WARN("bridge %s: netflow port mangling will conflict "
+ "with another port when more than 508 ports are "
+ "used", br->name);
+ }
+ }
+
+ opts.collectors.n = nf_cfg->n_targets;
+ opts.collectors.names = nf_cfg->targets;
+ if (ofproto_set_netflow(br->ofproto, &opts)) {
+ VLOG_ERR("bridge %s: problem setting netflow collectors",
+ br->name);
+ }
+ } else {
+ ofproto_set_netflow(br->ofproto, NULL);
}
/* Update the controller and related settings. It would be more
@@ -570,33 +732,32 @@ bridge_reconfigure(void)
* yet; when a controller is configured, resetting the datapath ID will
* immediately disconnect from the controller, so it's better to set
* the datapath ID before the controller. */
- bridge_reconfigure_controller(br);
+ bridge_reconfigure_controller(ovs_cfg, br);
}
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
+
port_update_vlan_compat(port);
+ port_update_bonding(port);
}
}
LIST_FOR_EACH (br, struct bridge, node, &all_bridges) {
- brstp_reconfigure(br);
+ iterate_and_prune_ifaces(br, set_iface_properties, NULL);
}
}
static void
bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
- const char **devname)
+ struct iface **hw_addr_iface)
{
- uint64_t requested_ea;
size_t i, j;
int error;
- *devname = NULL;
+ *hw_addr_iface = NULL;
/* Did the user request a particular MAC? */
- requested_ea = cfg_get_mac(0, "bridge.%s.mac", br->name);
- if (requested_ea) {
- eth_addr_from_uint64(requested_ea, ea);
+ if (br->cfg->hwaddr && eth_addr_from_string(br->cfg->hwaddr, ea)) {
if (eth_addr_is_multicast(ea)) {
VLOG_ERR("bridge %s: cannot set MAC address to multicast "
"address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
@@ -613,35 +774,74 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
memset(ea, 0xff, sizeof ea);
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
+ uint8_t iface_ea[ETH_ADDR_LEN];
+ struct iface *iface;
+
+ /* Mirror output ports don't participate. */
if (port->is_mirror_output_port) {
continue;
}
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- uint8_t iface_ea[ETH_ADDR_LEN];
+
+ /* Choose the MAC address to represent the port. */
+ if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
+ /* Find the interface with this Ethernet address (if any) so that
+ * we can provide the correct devname to the caller. */
+ iface = NULL;
+ for (j = 0; j < port->n_ifaces; j++) {
+ struct iface *candidate = port->ifaces[j];
+ uint8_t candidate_ea[ETH_ADDR_LEN];
+ if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
+ && eth_addr_equals(iface_ea, candidate_ea)) {
+ iface = candidate;
+ }
+ }
+ } else {
+ /* Choose the interface whose MAC address will represent the port.
+ * The Linux kernel bonding code always chooses the MAC address of
+ * the first slave added to a bond, and the Fedora networking
+ * scripts always add slaves to a bond in alphabetical order, so
+ * for compatibility we choose the interface with the name that is
+ * first in alphabetical order. */
+ iface = port->ifaces[0];
+ for (j = 1; j < port->n_ifaces; j++) {
+ struct iface *candidate = port->ifaces[j];
+ if (strcmp(candidate->name, iface->name) < 0) {
+ iface = candidate;
+ }
+ }
+
+ /* The local port doesn't count (since we're trying to choose its
+ * MAC address anyway). Other internal ports don't count because
+ * we really want a physical MAC if we can get it, and internal
+ * ports typically have randomly generated MACs. */
if (iface->dp_ifidx == ODPP_LOCAL
- || cfg_get_bool(0, "iface.%s.internal", iface->name)) {
+ || !strcmp(iface->cfg->type, "internal")) {
continue;
}
- error = netdev_nodev_get_etheraddr(iface->name, iface_ea);
- if (!error) {
- if (!eth_addr_is_multicast(iface_ea) &&
- !eth_addr_is_reserved(iface_ea) &&
- !eth_addr_is_zero(iface_ea) &&
- memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0) {
- memcpy(ea, iface_ea, ETH_ADDR_LEN);
- *devname = iface->name;
- }
- } else {
+
+ /* Grab MAC. */
+ error = netdev_get_etheraddr(iface->netdev, iface_ea);
+ if (error) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_ERR_RL(&rl, "failed to obtain Ethernet address of %s: %s",
iface->name, strerror(error));
+ continue;
}
}
+
+ /* Compare against our current choice. */
+ if (!eth_addr_is_multicast(iface_ea) &&
+ !eth_addr_is_reserved(iface_ea) &&
+ !eth_addr_is_zero(iface_ea) &&
+ memcmp(iface_ea, ea, ETH_ADDR_LEN) < 0)
+ {
+ memcpy(ea, iface_ea, ETH_ADDR_LEN);
+ *hw_addr_iface = iface;
+ }
}
if (eth_addr_is_multicast(ea) || eth_addr_is_vif(ea)) {
memcpy(ea, br->default_ea, ETH_ADDR_LEN);
- *devname = NULL;
+ *hw_addr_iface = NULL;
VLOG_WARN("bridge %s: using default bridge Ethernet "
"address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
} else {
@@ -652,13 +852,13 @@ bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
/* Choose and returns the datapath ID for bridge 'br' given that the bridge
* Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
- * a network device, then that network device's name must be passed in as
- * 'devname'; if 'bridge_ea' was derived some other way, then 'devname' must be
- * passed in as a null pointer. */
+ * an interface on 'br', then that interface must be passed in as
+ * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
+ * 'hw_addr_iface' must be passed in as a null pointer. */
static uint64_t
bridge_pick_datapath_id(struct bridge *br,
const uint8_t bridge_ea[ETH_ADDR_LEN],
- const char *devname)
+ struct iface *hw_addr_iface)
{
/*
* The procedure for choosing a bridge MAC address will, in the most
@@ -674,14 +874,14 @@ bridge_pick_datapath_id(struct bridge *br,
*/
uint64_t dpid;
- dpid = cfg_get_dpid(0, "bridge.%s.datapath-id", br->name);
- if (dpid) {
+ if (br->cfg->datapath_id
+ && dpid_from_string(br->cfg->datapath_id, &dpid)) {
return dpid;
}
- if (devname) {
+ if (hw_addr_iface) {
int vlan;
- if (!netdev_get_vlan_vid(devname, &vlan)) {
+ if (!netdev_get_vlan_vid(hw_addr_iface->netdev, &vlan)) {
/*
* A bridge whose MAC address is taken from a VLAN network device
* (that is, a network device created with vconfig(8) or similar
@@ -732,10 +932,10 @@ bridge_pick_datapath_id(struct bridge *br,
static uint64_t
dpid_from_hash(const void *data, size_t n)
{
- uint8_t hash[SHA1HashSize];
+ uint8_t hash[SHA1_DIGEST_SIZE];
BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
- SHA1Bytes(data, n, hash);
+ sha1_bytes(data, n, hash);
eth_addr_mark_random(hash);
return eth_addr_to_uint64(hash);
}
@@ -772,11 +972,8 @@ bridge_wait(void)
continue;
}
- if (br->ml) {
- mac_learning_wait(br->ml);
- }
+ mac_learning_wait(br->ml);
bond_wait(br);
- brstp_wait(br);
}
}
@@ -787,9 +984,55 @@ bridge_flush(struct bridge *br)
{
COVERAGE_INC(bridge_flush);
br->flush = true;
- if (br->ml) {
- mac_learning_flush(br->ml);
+ 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;
+}
+
+/* Bridge unixctl user interface functions. */
+static void
+bridge_unixctl_fdb_show(struct unixctl_conn *conn,
+ const char *args, void *aux UNUSED)
+{
+ struct ds ds = DS_EMPTY_INITIALIZER;
+ const struct bridge *br;
+ const struct mac_entry *e;
+
+ br = bridge_lookup(args);
+ if (!br) {
+ unixctl_command_reply(conn, 501, "no such bridge");
+ return;
+ }
+
+ ds_put_cstr(&ds, " port VLAN MAC Age\n");
+ LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) {
+ if (e->port < 0 || e->port >= br->n_ports) {
+ continue;
+ }
+ ds_put_format(&ds, "%5d %4d "ETH_ADDR_FMT" %3d\n",
+ br->ports[e->port]->ifaces[0]->dp_ifidx,
+ e->vlan, ETH_ADDR_ARGS(e->mac), mac_entry_age(e));
}
+ unixctl_command_reply(conn, 200, ds_cstr(&ds));
+ ds_destroy(&ds);
}
/* Bridge reconfiguration functions. */
@@ -801,29 +1044,20 @@ bridge_create(const char *name)
int error;
assert(!bridge_lookup(name));
- br = xcalloc(1, sizeof *br);
+ br = xzalloc(sizeof *br);
- error = dpif_create(name, &br->dpif);
- if (error == EEXIST) {
- error = dpif_open(name, &br->dpif);
- if (error) {
- VLOG_ERR("datapath %s already exists but cannot be opened: %s",
- name, strerror(error));
- free(br);
- return NULL;
- }
- dpif_flow_flush(&br->dpif);
- } else if (error) {
- VLOG_ERR("failed to create datapath %s: %s", name, strerror(error));
+ error = dpif_create_and_open(name, &br->dpif);
+ if (error) {
free(br);
return NULL;
}
+ dpif_flow_flush(br->dpif);
error = ofproto_create(name, &bridge_ofhooks, br, &br->ofproto);
if (error) {
VLOG_ERR("failed to create switch %s: %s", name, strerror(error));
- dpif_delete(&br->dpif);
- dpif_close(&br->dpif);
+ dpif_delete(br->dpif);
+ dpif_close(br->dpif);
free(br);
return NULL;
}
@@ -840,7 +1074,7 @@ bridge_create(const char *name)
list_push_back(&all_bridges, &br->node);
- VLOG_INFO("created bridge %s on dp%u", br->name, dpif_id(&br->dpif));
+ VLOG_INFO("created bridge %s on %s", br->name, dpif_name(br->dpif));
return br;
}
@@ -855,12 +1089,12 @@ bridge_destroy(struct bridge *br)
port_destroy(br->ports[br->n_ports - 1]);
}
list_remove(&br->node);
- error = dpif_delete(&br->dpif);
+ error = dpif_delete(br->dpif);
if (error && error != ENOENT) {
- VLOG_ERR("failed to delete dp%u: %s",
- dpif_id(&br->dpif), strerror(error));
+ VLOG_ERR("failed to delete %s: %s",
+ dpif_name(br->dpif), strerror(error));
}
- dpif_close(&br->dpif);
+ dpif_close(br->dpif);
ofproto_destroy(br->ofproto);
free(br->controller);
mac_learning_destroy(br->ml);
@@ -897,6 +1131,28 @@ bridge_get_datapathid(const char *name)
return br ? ofproto_get_datapath_id(br->ofproto) : 0;
}
+/* Handle requests for a listing of all flows known by the OpenFlow
+ * stack, including those normally hidden. */
+static void
+bridge_unixctl_dump_flows(struct unixctl_conn *conn,
+ const char *args, void *aux UNUSED)
+{
+ struct bridge *br;
+ struct ds results;
+
+ br = bridge_lookup(args);
+ if (!br) {
+ unixctl_command_reply(conn, 501, "Unknown bridge");
+ return;
+ }
+
+ ds_init(&results);
+ ofproto_get_all_flows(br->ofproto, &results);
+
+ unixctl_command_reply(conn, 200, ds_cstr(&results));
+ ds_destroy(&results);
+}
+
static int
bridge_run_one(struct bridge *br)
{
@@ -907,11 +1163,8 @@ bridge_run_one(struct bridge *br)
return error;
}
- if (br->ml) {
- mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
- }
+ mac_learning_run(br->ml, ofproto_get_revalidate_set(br->ofproto));
bond_run(br);
- brstp_run(br);
error = ofproto_run2(br->ofproto, br->flush);
br->flush = false;
@@ -919,103 +1172,113 @@ bridge_run_one(struct bridge *br)
return error;
}
-static const char *
-bridge_get_controller(const struct bridge *br)
+static const struct ovsrec_controller *
+bridge_get_controller(const struct ovsrec_open_vswitch *ovs_cfg,
+ const struct bridge *br)
{
- const char *controller;
+ const struct ovsrec_controller *controller;
+
+ controller = (br->cfg->controller ? br->cfg->controller
+ : ovs_cfg->controller ? ovs_cfg->controller
+ : NULL);
+
+ if (controller && !strcmp(controller->target, "none")) {
+ return NULL;
+ }
+
+ return controller;
+}
- controller = cfg_get_string(0, "bridge.%s.controller", br->name);
- if (!controller) {
- controller = cfg_get_string(0, "mgmt.controller");
+static bool
+check_duplicate_ifaces(struct bridge *br, struct iface *iface, void *ifaces_)
+{
+ struct svec *ifaces = ifaces_;
+ if (!svec_contains(ifaces, iface->name)) {
+ svec_add(ifaces, iface->name);
+ svec_sort(ifaces);
+ return true;
+ } else {
+ VLOG_ERR("bridge %s: %s interface is on multiple ports, "
+ "removing from %s",
+ br->name, iface->name, iface->port->name);
+ return false;
}
- return controller && controller[0] ? controller : NULL;
}
static void
-bridge_reconfigure_one(struct bridge *br)
+bridge_reconfigure_one(const struct ovsrec_open_vswitch *ovs_cfg,
+ struct bridge *br)
{
- struct svec old_ports, new_ports, ifaces;
+ struct shash old_ports, new_ports;
+ struct svec ifaces;
struct svec listeners, old_listeners;
struct svec snoops, old_snoops;
- size_t i, j;
+ struct shash_node *node;
+ uint64_t mgmt_id;
+ size_t i;
/* Collect old ports. */
- svec_init(&old_ports);
+ shash_init(&old_ports);
for (i = 0; i < br->n_ports; i++) {
- svec_add(&old_ports, br->ports[i]->name);
+ shash_add(&old_ports, br->ports[i]->name, br->ports[i]);
}
- svec_sort(&old_ports);
- assert(svec_is_unique(&old_ports));
/* Collect new ports. */
- svec_init(&new_ports);
- cfg_get_all_keys(&new_ports, "bridge.%s.port", br->name);
- svec_sort(&new_ports);
- if (bridge_get_controller(br) && !svec_contains(&new_ports, br->name)) {
- svec_add(&new_ports, br->name);
- svec_sort(&new_ports);
+ shash_init(&new_ports);
+ for (i = 0; i < br->cfg->n_ports; i++) {
+ const char *name = br->cfg->ports[i]->name;
+ if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
+ VLOG_WARN("bridge %s: %s specified twice as bridge port",
+ br->name, name);
+ }
}
- if (!svec_is_unique(&new_ports)) {
- VLOG_WARN("bridge %s: %s specified twice as bridge port",
- br->name, svec_get_duplicate(&new_ports));
- svec_unique(&new_ports);
+
+ /* If we have a controller, then we need a local port. Complain if the
+ * user didn't specify one.
+ *
+ * XXX perhaps we should synthesize a port ourselves in this case. */
+ if (bridge_get_controller(ovs_cfg, br)) {
+ char local_name[IF_NAMESIZE];
+ int error;
+
+ error = dpif_port_get_name(br->dpif, ODPP_LOCAL,
+ local_name, sizeof local_name);
+ if (!error && !shash_find(&new_ports, local_name)) {
+ VLOG_WARN("bridge %s: controller specified but no local port "
+ "(port named %s) defined",
+ br->name, local_name);
+ }
}
+ dpid_from_string(ovs_cfg->management_id, &mgmt_id);
ofproto_set_mgmt_id(br->ofproto, mgmt_id);
/* Get rid of deleted ports and add new ports. */
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
- if (!svec_contains(&new_ports, port->name)) {
- port_destroy(port);
- } else {
- i++;
+ SHASH_FOR_EACH (node, &old_ports) {
+ if (!shash_find(&new_ports, node->name)) {
+ port_destroy(node->data);
}
}
- for (i = 0; i < new_ports.n; i++) {
- const char *name = new_ports.names[i];
- if (!svec_contains(&old_ports, name)) {
- port_create(br, name);
+ SHASH_FOR_EACH (node, &new_ports) {
+ struct port *port = shash_find_data(&old_ports, node->name);
+ if (!port) {
+ port = port_create(br, node->name);
}
+ port_reconfigure(port, node->data);
}
- svec_destroy(&old_ports);
- svec_destroy(&new_ports);
-
- /* Reconfigure all ports. */
- for (i = 0; i < br->n_ports; i++) {
- port_reconfigure(br->ports[i]);
- }
+ shash_destroy(&old_ports);
+ shash_destroy(&new_ports);
/* Check and delete duplicate interfaces. */
svec_init(&ifaces);
- for (i = 0; i < br->n_ports; ) {
- struct port *port = br->ports[i];
- for (j = 0; j < port->n_ifaces; ) {
- struct iface *iface = port->ifaces[j];
- if (svec_contains(&ifaces, iface->name)) {
- VLOG_ERR("bridge %s: %s interface is on multiple ports, "
- "removing from %s",
- br->name, iface->name, port->name);
- iface_destroy(iface);
- } else {
- svec_add(&ifaces, iface->name);
- svec_sort(&ifaces);
- j++;
- }
- }
- if (!port->n_ifaces) {
- VLOG_ERR("%s port has no interfaces, dropping", port->name);
- port_destroy(port);
- } else {
- i++;
- }
- }
+ iterate_and_prune_ifaces(br, check_duplicate_ifaces, &ifaces);
svec_destroy(&ifaces);
/* Delete all flows if we're switching from connected to standalone or vice
* versa. (XXX Should we delete all flows if we are switching from one
* controller to another?) */
+#if 0
/* Configure OpenFlow management listeners. */
svec_init(&listeners);
cfg_get_all_strings(&listeners, "bridge.%s.openflow.listeners", br->name);
@@ -1057,125 +1320,118 @@ bridge_reconfigure_one(struct bridge *br)
}
svec_destroy(&snoops);
svec_destroy(&old_snoops);
+#else
+ /* Default 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);
+
+ /* Default snoop. */
+ svec_init(&snoops);
+ svec_add_nocopy(&snoops, xasprintf("punix:%s/%s.snoop",
+ ovs_rundir, br->name));
+ svec_init(&old_snoops);
+ ofproto_get_snoops(br->ofproto, &old_snoops);
+ if (!svec_equal(&snoops, &old_snoops)) {
+ ofproto_set_snoops(br->ofproto, &snoops);
+ }
+ svec_destroy(&snoops);
+ svec_destroy(&old_snoops);
+#endif
+#if 0
mirror_reconfigure(br);
+#endif
}
static void
-bridge_reconfigure_controller(struct bridge *br)
+bridge_reconfigure_controller(const struct ovsrec_open_vswitch *ovs_cfg,
+ struct bridge *br)
{
char *pfx = xasprintf("bridge.%s.controller", br->name);
- const char *controller;
+ const struct ovsrec_controller *c;
- controller = bridge_get_controller(br);
- if ((br->controller != NULL) != (controller != NULL)) {
+ c = bridge_get_controller(ovs_cfg, br);
+ if ((br->controller != NULL) != (c != NULL)) {
ofproto_flush_flows(br->ofproto);
}
free(br->controller);
- br->controller = controller ? xstrdup(controller) : NULL;
+ br->controller = c ? xstrdup(c->target) : NULL;
- if (controller) {
- const char *fail_mode;
+ if (c) {
int max_backoff, probe;
int rate_limit, burst_limit;
- if (!strcmp(controller, "discover")) {
+ if (!strcmp(c->target, "discover")) {
ofproto_set_discovery(br->ofproto, true,
- cfg_get_string(0, "%s.accept-regex", pfx),
- cfg_get_bool(0, "%s.update-resolv.conf",
- pfx));
+ c->discover_accept_regex,
+ c->discover_update_resolv_conf);
} else {
- struct netdev *netdev;
+ struct iface *local_iface;
+ struct in_addr ip;
bool in_band;
- int error;
- in_band = (!cfg_is_valid(CFG_BOOL | CFG_REQUIRED,
- "%s.in-band", pfx)
- || cfg_get_bool(0, "%s.in-band", pfx));
+ in_band = (!c->connection_mode
+ || !strcmp(c->connection_mode, "out-of-band"));
ofproto_set_discovery(br->ofproto, false, NULL, NULL);
ofproto_set_in_band(br->ofproto, in_band);
- error = netdev_open(br->name, NETDEV_ETH_TYPE_NONE, &netdev);
- if (!error) {
- if (cfg_is_valid(CFG_IP | CFG_REQUIRED, "%s.ip", pfx)) {
- struct in_addr ip, mask, gateway;
- ip.s_addr = cfg_get_ip(0, "%s.ip", pfx);
- mask.s_addr = cfg_get_ip(0, "%s.netmask", pfx);
- gateway.s_addr = cfg_get_ip(0, "%s.gateway", pfx);
-
- netdev_turn_flags_on(netdev, NETDEV_UP, true);
- if (!mask.s_addr) {
- mask.s_addr = guess_netmask(ip.s_addr);
- }
- if (!netdev_set_in4(netdev, ip, mask)) {
- VLOG_INFO("bridge %s: configured IP address "IP_FMT", "
- "netmask "IP_FMT,
- br->name, IP_ARGS(&ip.s_addr),
- IP_ARGS(&mask.s_addr));
- }
+ 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 ip, mask, gateway;
- if (gateway.s_addr) {
- if (!netdev_add_router(gateway)) {
- VLOG_INFO("bridge %s: configured gateway "IP_FMT,
- br->name, IP_ARGS(&gateway.s_addr));
- }
+ 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;
+ }
+
+ 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 (gateway.s_addr) {
+ if (!netdev_add_router(netdev, gateway)) {
+ VLOG_INFO("bridge %s: configured gateway "IP_FMT,
+ br->name, IP_ARGS(&gateway.s_addr));
}
}
- netdev_close(netdev);
}
}
- fail_mode = cfg_get_string(0, "%s.fail-mode", pfx);
- if (!fail_mode) {
- fail_mode = cfg_get_string(0, "mgmt.fail-mode");
- }
ofproto_set_failure(br->ofproto,
- (!fail_mode
- || !strcmp(fail_mode, "standalone")
- || !strcmp(fail_mode, "open")));
-
- probe = cfg_get_int(0, "%s.inactivity-probe", pfx);
- ofproto_set_probe_interval(br->ofproto,
- probe ? probe : cfg_get_int(0, "mgmt.inactivity-probe"));
-
- max_backoff = cfg_get_int(0, "%s.max-backoff", pfx);
- if (!max_backoff) {
- max_backoff = cfg_get_int(0, "mgmt.max-backoff");
- if (!max_backoff) {
- max_backoff = 15;
- }
- }
- ofproto_set_max_backoff(br->ofproto, max_backoff);
+ (!c->fail_mode
+ || !strcmp(c->fail_mode, "standalone")
+ || !strcmp(c->fail_mode, "open")));
- rate_limit = cfg_get_int(0, "%s.rate-limit", pfx);
- if (!rate_limit) {
- rate_limit = cfg_get_int(0, "mgmt.rate-limit");
- }
- burst_limit = cfg_get_int(0, "%s.burst-limit", pfx);
- if (!burst_limit) {
- burst_limit = cfg_get_int(0, "mgmt.burst-limit");
- }
- ofproto_set_rate_limit(br->ofproto, rate_limit, burst_limit);
-
- ofproto_set_stp(br->ofproto, cfg_get_bool(0, "%s.stp", pfx));
-
- if (cfg_has("%s.commands.acl", pfx)) {
- struct svec command_acls;
- char *command_acl;
+ probe = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
+ ofproto_set_probe_interval(br->ofproto, probe);
- svec_init(&command_acls);
- cfg_get_all_strings(&command_acls, "%s.commands.acl", pfx);
- command_acl = svec_join(&command_acls, ",", "");
+ max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
+ ofproto_set_max_backoff(br->ofproto, max_backoff);
- ofproto_set_remote_execution(br->ofproto, command_acl,
- cfg_get_string(0, "%s.commands.dir",
- pfx));
+ rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
+ burst_limit = c->controller_burst_limit ? *c->controller_burst_limit : 0;
+ ofproto_set_rate_limit(br->ofproto, rate_limit, burst_limit);
- svec_destroy(&command_acls);
- free(command_acl);
- } else {
- ofproto_set_remote_execution(br->ofproto, NULL, NULL);
- }
+ ofproto_set_remote_execution(br->ofproto, NULL, NULL); /* XXX */
} else {
union ofp_action action;
flow_t flow;
@@ -1194,7 +1450,6 @@ bridge_reconfigure_controller(struct bridge *br)
ofproto_set_max_backoff(br->ofproto, 1);
ofproto_set_probe_interval(br->ofproto, 5);
ofproto_set_failure(br->ofproto, false);
- ofproto_set_stp(br->ofproto, false);
}
free(pfx);
@@ -1202,20 +1457,21 @@ bridge_reconfigure_controller(struct bridge *br)
}
static void
-bridge_get_all_ifaces(const struct bridge *br, struct svec *ifaces)
+bridge_get_all_ifaces(const struct bridge *br, struct shash *ifaces)
{
size_t i, j;
- svec_init(ifaces);
+ 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];
- svec_add(ifaces, iface->name);
+ shash_add_once(ifaces, iface->name, iface);
+ }
+ if (port->n_ifaces > 1 && port->cfg->bond_fake_iface) {
+ shash_add_once(ifaces, port->name, NULL);
}
}
- svec_sort(ifaces);
- assert(svec_is_unique(ifaces));
}
/* For robustness, in case the administrator moves around datapath ports behind
@@ -1243,17 +1499,17 @@ bridge_fetch_dp_ifaces(struct bridge *br)
}
port_array_clear(&br->ifaces);
- dpif_port_list(&br->dpif, &dpif_ports, &n_dpif_ports);
+ dpif_port_list(br->dpif, &dpif_ports, &n_dpif_ports);
for (i = 0; i < n_dpif_ports; i++) {
struct odp_port *p = &dpif_ports[i];
struct iface *iface = iface_lookup(br, p->devname);
if (iface) {
if (iface->dp_ifidx >= 0) {
- VLOG_WARN("dp%u reported interface %s twice",
- dpif_id(&br->dpif), p->devname);
+ VLOG_WARN("%s reported interface %s twice",
+ dpif_name(br->dpif), p->devname);
} else if (iface_from_dp_ifidx(br, p->port)) {
- VLOG_WARN("dp%u reported interface %"PRIu16" twice",
- dpif_id(&br->dpif), p->port);
+ VLOG_WARN("%s reported interface %"PRIu16" twice",
+ dpif_name(br->dpif), p->port);
} else {
port_array_set(&br->ifaces, p->port, iface);
iface->dp_ifidx = p->port;
@@ -1280,13 +1536,31 @@ lookup_bond_entry(const struct port *port, const uint8_t mac[ETH_ADDR_LEN])
static int
bond_choose_iface(const struct port *port)
{
- size_t i;
+ 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++) {
- if (port->ifaces[i]->enabled) {
+ struct iface *iface = port->ifaces[i];
+
+ if (iface->enabled) {
return i;
+ } else if (iface->delay_expires < next_delay_expiration) {
+ best_down_slave = i;
+ next_delay_expiration = iface->delay_expires;
}
}
- return -1;
+
+ if (best_down_slave != -1) {
+ struct iface *iface = port->ifaces[best_down_slave];
+
+ VLOG_INFO_RL(&rl, "interface %s: skipping remaining %lli ms updelay "
+ "since no other interface is up", iface->name,
+ iface->delay_expires - time_msec());
+ bond_enable_slave(iface, true);
+ }
+
+ return best_down_slave;
}
static bool
@@ -1310,6 +1584,7 @@ choose_output_iface(const struct port *port, const uint8_t *dl_src,
return false;
}
e->iface_tag = tag_create_random();
+ ((struct port *) port)->bond_compat_is_stale = true;
}
*tags |= e->iface_tag;
iface = port->ifaces[e->iface_idx];
@@ -1335,6 +1610,12 @@ bond_link_status_update(struct iface *iface, bool carrier)
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) {
+ bond_enable_slave(iface, true);
+ if (port->updelay) {
+ VLOG_INFO_RL(&rl, "interface %s: skipping %d ms updelay since no "
+ "other interface is up", iface->name, port->updelay);
+ }
} else {
int delay = carrier ? port->updelay : port->downdelay;
iface->delay_expires = time_msec() + delay;
@@ -1371,6 +1652,12 @@ bond_enable_slave(struct iface *iface, bool enable)
struct port *port = iface->port;
struct bridge *br = port->bridge;
+ /* This acts as a recursion check. If the act of disabling a slave
+ * causes a different slave to be enabled, the flag will allow us to
+ * skip redundant work when we reenter this function. It must be
+ * cleared on exit to keep things safe with multiple bonds. */
+ static bool moving_active_iface = false;
+
iface->delay_expires = LLONG_MAX;
if (enable == iface->enabled) {
return;
@@ -1378,23 +1665,34 @@ bond_enable_slave(struct iface *iface, bool enable)
iface->enabled = enable;
if (!iface->enabled) {
- VLOG_WARN("interface %s: enabled", iface->name);
+ VLOG_WARN("interface %s: disabled", iface->name);
ofproto_revalidate(br->ofproto, iface->tag);
if (iface->port_ifidx == port->active_iface) {
ofproto_revalidate(br->ofproto,
port->active_iface_tag);
+
+ /* Disabling a slave can lead to another slave being immediately
+ * enabled if there will be no active slaves but one is waiting
+ * on an updelay. In this case we do not need to run most of the
+ * code for the newly enabled slave since there was no period
+ * without an active slave and it is redundant with the disabling
+ * path. */
+ moving_active_iface = true;
bond_choose_active_iface(port);
}
bond_send_learning_packets(port);
} else {
- VLOG_WARN("interface %s: disabled", iface->name);
- if (port->active_iface < 0) {
+ VLOG_WARN("interface %s: enabled", iface->name);
+ if (port->active_iface < 0 && !moving_active_iface) {
ofproto_revalidate(br->ofproto, port->no_ifaces_tag);
bond_choose_active_iface(port);
bond_send_learning_packets(port);
}
iface->tag = tag_create_random();
}
+
+ moving_active_iface = false;
+ port->bond_compat_is_stale = true;
}
static void
@@ -1404,15 +1702,20 @@ bond_run(struct bridge *br)
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
- if (port->n_ifaces < 2) {
- continue;
- }
- for (j = 0; j < port->n_ifaces; j++) {
- struct iface *iface = port->ifaces[j];
- if (time_msec() >= iface->delay_expires) {
- bond_enable_slave(iface, !iface->enabled);
+
+ if (port->n_ifaces >= 2) {
+ 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_compat_is_stale) {
+ port->bond_compat_is_stale = false;
+ port_update_bond_compat(port);
+ }
}
}
@@ -1440,18 +1743,6 @@ set_dst(struct dst *p, const flow_t *flow,
const struct port *in_port, const struct port *out_port,
tag_type *tags)
{
- /* STP handling.
- *
- * XXX This uses too many tags: any broadcast flow will get one tag per
- * destination port, and thus a broadcast on a switch of any size is likely
- * to have all tag bits set. We should figure out a way to be smarter.
- *
- * This is OK when STP is disabled, because stp_state_tag is 0 then. */
- *tags |= out_port->stp_state_tag;
- if (!(out_port->stp_state & (STP_DISABLED | STP_FORWARDING))) {
- return false;
- }
-
p->vlan = (out_port->vlan >= 0 ? OFP_VLAN_NONE
: in_port->vlan >= 0 ? in_port->vlan
: ntohs(flow->dl_vlan));
@@ -1538,13 +1829,12 @@ port_includes_vlan(const struct port *port, uint16_t vlan)
static size_t
compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan,
const struct port *in_port, const struct port *out_port,
- struct dst dsts[], tag_type *tags)
+ struct dst dsts[], tag_type *tags, uint16_t *nf_output_iface)
{
mirror_mask_t mirrors = in_port->src_mirrors;
struct dst *dst = dsts;
size_t i;
- *tags |= in_port->stp_state_tag;
if (out_port == FLOOD_PORT) {
/* XXX use ODP_FLOOD if no vlans or bonding. */
/* XXX even better, define each VLAN as a datapath port group */
@@ -1557,7 +1847,9 @@ compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan,
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;
mirrors |= out_port->dst_mirrors;
dst++;
}
@@ -1574,14 +1866,28 @@ compose_dsts(const struct bridge *br, const flow_t *flow, uint16_t vlan,
for (i = 0; i < br->n_ports; i++) {
struct port *port = br->ports[i];
if (port_includes_vlan(port, m->out_vlan)
- && set_dst(dst, flow, in_port, port, tags)
- && !dst_is_duplicate(dsts, dst - dsts, dst))
+ && set_dst(dst, flow, in_port, port, tags))
{
+ int flow_vlan;
+
if (port->vlan < 0) {
dst->vlan = m->out_vlan;
}
- if (dst->dp_ifidx == flow->in_port
- && dst->vlan == vlan) {
+ if (dst_is_duplicate(dsts, dst - dsts, dst)) {
+ continue;
+ }
+
+ /* Use the vlan tag on the original flow instead of
+ * the one passed in the vlan parameter. This ensures
+ * that we compare the vlan from before any implicit
+ * tagging tags place. This is necessary because
+ * dst->vlan is the final vlan, after removing implicit
+ * tags. */
+ flow_vlan = ntohs(flow->dl_vlan);
+ if (flow_vlan == 0) {
+ flow_vlan = OFP_VLAN_NONE;
+ }
+ if (port == in_port && dst->vlan == flow_vlan) {
/* Don't send out input port on same VLAN. */
continue;
}
@@ -1611,14 +1917,16 @@ print_dsts(const struct dst *dsts, size_t n)
static void
compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan,
const struct port *in_port, const struct port *out_port,
- tag_type *tags, struct odp_actions *actions)
+ tag_type *tags, struct odp_actions *actions,
+ uint16_t *nf_output_iface)
{
struct dst dsts[DP_MAX_PORTS * (MAX_MIRRORS + 1)];
size_t n_dsts;
const struct dst *p;
uint16_t cur_vlan;
- n_dsts = compose_dsts(br, flow, vlan, in_port, out_port, dsts, tags);
+ n_dsts = compose_dsts(br, flow, vlan, in_port, out_port, dsts, tags,
+ nf_output_iface);
cur_vlan = ntohs(flow->dl_vlan);
for (p = dsts; p < &dsts[n_dsts]; p++) {
@@ -1637,14 +1945,77 @@ compose_actions(struct bridge *br, const flow_t *flow, uint16_t vlan,
}
}
+/* 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,
+ 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;
+ }
+ 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 "
+ "packet received on port %s configured with "
+ "implicit VLAN %"PRIu16,
+ br->name, ntohs(flow->dl_vlan),
+ in_port->name, in_port->vlan);
+ }
+ return -1;
+ }
+ vlan = in_port->vlan;
+ } else {
+ if (!port_includes_vlan(in_port, vlan)) {
+ if (have_packet) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
+ "packet received on port %s not configured for "
+ "trunking VLAN %d",
+ br->name, vlan, in_port->name, vlan);
+ }
+ return -1;
+ }
+ }
+
+ return vlan;
+}
+
+static void
+update_learning_table(struct bridge *br, const flow_t *flow, int vlan,
+ struct port *in_port)
+{
+ tag_type rev_tag = mac_learning_learn(br->ml, flow->dl_src,
+ vlan, in_port->port_idx);
+ if (rev_tag) {
+ /* The log messages here could actually be useful in debugging,
+ * so keep the rate limit relatively high. */
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30,
+ 300);
+ VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
+ "on port %s in VLAN %d",
+ br->name, ETH_ADDR_ARGS(flow->dl_src),
+ in_port->name, vlan);
+ ofproto_revalidate(br->ofproto, rev_tag);
+ }
+}
+
static bool
-is_bcast_arp_reply(const flow_t *flow, const struct ofpbuf *packet)
+is_bcast_arp_reply(const flow_t *flow)
{
- struct arp_eth_header *arp = (struct arp_eth_header *) packet->data;
return (flow->dl_type == htons(ETH_TYPE_ARP)
- && eth_addr_is_broadcast(flow->dl_dst)
- && packet->size >= sizeof(struct arp_eth_header)
- && arp->ar_op == ARP_OP_REQUEST);
+ && flow->nw_proto == ARP_OP_REPLY
+ && eth_addr_is_broadcast(flow->dl_dst));
}
/* If the composed actions may be applied to any packet in the given 'flow',
@@ -1653,12 +2024,13 @@ is_bcast_arp_reply(const flow_t *flow, const struct ofpbuf *packet)
static bool
process_flow(struct bridge *br, const flow_t *flow,
const struct ofpbuf *packet, struct odp_actions *actions,
- tag_type *tags)
+ tag_type *tags, uint16_t *nf_output_iface)
{
struct iface *in_iface;
struct port *in_port;
struct port *out_port = NULL; /* By default, drop the packet/flow. */
int vlan;
+ int out_port_idx;
/* Find the interface and port structure for the received packet. */
in_iface = iface_from_dp_ifidx(br, flow->in_port);
@@ -1686,47 +2058,8 @@ process_flow(struct bridge *br, const flow_t *flow,
return true;
}
in_port = in_iface->port;
-
- /* Figure out what VLAN this packet belongs to.
- *
- * Note that dl_vlan of 0 and of OFP_VLAN_NONE both mean that the packet
- * belongs to VLAN 0, so we should treat both cases identically. (In the
- * former case, the packet has an 802.1Q header that specifies VLAN 0,
- * presumably to allow a priority to be specified. In the latter case, the
- * packet does not have any 802.1Q header.) */
- vlan = ntohs(flow->dl_vlan);
- if (vlan == OFP_VLAN_NONE) {
- vlan = 0;
- }
- if (in_port->vlan >= 0) {
- if (vlan) {
- /* XXX support double tagging? */
- if (packet != NULL) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %"PRIu16" tagged "
- "packet received on port %s configured with "
- "implicit VLAN %"PRIu16,
- br->name, ntohs(flow->dl_vlan),
- in_port->name, in_port->vlan);
- }
- goto done;
- }
- vlan = in_port->vlan;
- } else {
- if (!port_includes_vlan(in_port, vlan)) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- VLOG_WARN_RL(&rl, "bridge %s: dropping VLAN %d tagged "
- "packet received on port %s not configured for "
- "trunking VLAN %d",
- br->name, vlan, in_port->name, vlan);
- goto done;
- }
- }
-
- /* Drop frames for ports that STP wants entirely killed (both for
- * forwarding and for learning). Later, after we do learning, we'll drop
- * the frames that STP wants to do learning but not forwarding on. */
- if (in_port->stp_state & (STP_LISTENING | STP_BLOCKING)) {
+ vlan = flow_get_vlan(br, flow, in_port, !!packet);
+ if (vlan < 0) {
goto done;
}
@@ -1744,76 +2077,60 @@ process_flow(struct bridge *br, const flow_t *flow,
goto done;
}
- /* Multicast (and broadcast) packets on bonds need special attention, to
- * avoid receiving duplicates. */
- if (in_port->n_ifaces > 1 && eth_addr_is_multicast(flow->dl_dst)) {
- *tags |= in_port->active_iface_tag;
- if (in_port->active_iface != in_iface->port_ifidx) {
- /* Drop all multicast packets on inactive slaves. */
- goto done;
- } else {
- /* Drop all multicast packets for which we have learned a different
- * input port, because we probably sent the packet on one slaves
- * and got it back on the active slave. Broadcast ARP replies are
- * an exception to this rule: the host has moved to another
- * switch. */
- int src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan);
- if (src_idx != -1
- && src_idx != in_port->port_idx
- && !is_bcast_arp_reply(flow, packet)) {
+ /* Packets received on bonds need special attention to avoid duplicates. */
+ if (in_port->n_ifaces > 1) {
+ int src_idx;
+
+ if (eth_addr_is_multicast(flow->dl_dst)) {
+ *tags |= in_port->active_iface_tag;
+ if (in_port->active_iface != in_iface->port_ifidx) {
+ /* Drop all multicast packets on inactive slaves. */
goto done;
}
}
+
+ /* Drop all packets for which we have learned a different input
+ * port, because we probably sent the packet on one slave and got
+ * it back on the other. Broadcast ARP replies are an exception
+ * to this rule: the host has moved to another switch. */
+ src_idx = mac_learning_lookup(br->ml, flow->dl_src, vlan);
+ if (src_idx != -1 && src_idx != in_port->port_idx &&
+ !is_bcast_arp_reply(flow)) {
+ goto done;
+ }
}
/* MAC learning. */
out_port = FLOOD_PORT;
- if (br->ml) {
- int out_port_idx;
-
- /* Learn source MAC (but don't try to learn from revalidation). */
- if (packet) {
- tag_type rev_tag = mac_learning_learn(br->ml, flow->dl_src,
- vlan, in_port->port_idx);
- if (rev_tag) {
- /* The log messages here could actually be useful in debugging,
- * so keep the rate limit relatively high. */
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30,
- 300);
- VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
- "on port %s in VLAN %d",
- br->name, ETH_ADDR_ARGS(flow->dl_src),
- in_port->name, vlan);
- ofproto_revalidate(br->ofproto, rev_tag);
- }
- }
-
- /* Determine output port. */
- out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan,
- tags);
- if (out_port_idx >= 0 && out_port_idx < br->n_ports) {
- out_port = br->ports[out_port_idx];
- }
+ /* Learn source MAC (but don't try to learn from revalidation). */
+ if (packet) {
+ update_learning_table(br, flow, vlan, in_port);
+ }
+
+ /* Determine output port. */
+ out_port_idx = mac_learning_lookup_tag(br->ml, flow->dl_dst, vlan,
+ tags);
+ if (out_port_idx >= 0 && out_port_idx < br->n_ports) {
+ out_port = br->ports[out_port_idx];
+ } else if (!packet && !eth_addr_is_multicast(flow->dl_dst)) {
+ /* If we are revalidating but don't have a learning entry then
+ * eject the flow. Installing a flow that floods packets opens
+ * up a window of time where we could learn from a packet reflected
+ * on a bond and blackhole packets before the learning table is
+ * updated to reflect the correct port. */
+ return false;
}
- /* Don't send packets out their input ports. Don't forward frames that STP
- * wants us to discard. */
- if (in_port == out_port || in_port->stp_state == STP_LEARNING) {
+ /* Don't send packets out their input ports. */
+ if (in_port == out_port) {
out_port = NULL;
}
done:
- compose_actions(br, flow, vlan, in_port, out_port, tags, actions);
+ compose_actions(br, flow, vlan, in_port, out_port, tags, actions,
+ nf_output_iface);
- /*
- * We send out only a single packet, instead of setting up a flow, if the
- * packet is an ARP directed to broadcast that arrived on a bonded
- * interface. In such a situation ARP requests and replies must be handled
- * differently, but OpenFlow unfortunately can't distinguish them.
- */
- return (in_port->n_ifaces < 2
- || flow->dl_type != htons(ETH_TYPE_ARP)
- || !eth_addr_is_broadcast(flow->dl_dst));
+ return true;
}
/* Careful: 'opp' is in host byte order and opp->port_no is an OFP port
@@ -1845,7 +2162,6 @@ bridge_port_changed_ofhook_cb(enum ofp_port_reason reason,
bridge_flush(br);
} else {
- memcpy(iface->mac, opp->hw_addr, ETH_ADDR_LEN);
if (port->n_ifaces > 1) {
bool up = !(opp->state & OFPPS_LINK_DOWN);
bond_link_status_update(iface, up);
@@ -1856,20 +2172,13 @@ bridge_port_changed_ofhook_cb(enum ofp_port_reason reason,
static bool
bridge_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet,
- struct odp_actions *actions, tag_type *tags, void *br_)
+ struct odp_actions *actions, tag_type *tags,
+ uint16_t *nf_output_iface, void *br_)
{
struct bridge *br = br_;
-#if 0
- if (flow->dl_type == htons(OFP_DL_TYPE_NOT_ETH_TYPE)
- && eth_addr_equals(flow->dl_dst, stp_eth_addr)) {
- brstp_receive(br, flow, payload);
- return true;
- }
-#endif
-
COVERAGE_INC(bridge_process_flow);
- return process_flow(br, flow, packet, actions, tags);
+ return process_flow(br, flow, packet, actions, tags, nf_output_iface);
}
static void
@@ -1879,17 +2188,30 @@ bridge_account_flow_ofhook_cb(const flow_t *flow,
void *br_)
{
struct bridge *br = br_;
+ struct port *in_port;
const union odp_action *a;
+ /* 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. */
+ in_port = port_from_dp_ifidx(br, flow->in_port);
+ if (in_port) {
+ int vlan = flow_get_vlan(br, flow, in_port, false);
+ if (vlan >= 0) {
+ update_learning_table(br, flow, vlan, in_port);
+ }
+ }
+
if (!br->has_bonded_ports) {
return;
}
for (a = actions; a < &actions[n_actions]; a++) {
if (a->type == ODPAT_OUTPUT) {
- struct port *port = port_from_dp_ifidx(br, a->output.port);
- if (port && port->n_ifaces >= 2) {
- struct bond_entry *e = lookup_bond_entry(port, flow->dl_src);
+ 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);
e->tx_bytes += n_bytes;
}
}
@@ -2044,8 +2366,9 @@ log_bals(const struct slave_balance *bals, size_t n_bals, struct port *port)
/* Shifts 'hash' from 'from' to 'to' within 'port'. */
static void
bond_shift_load(struct slave_balance *from, struct slave_balance *to,
- struct bond_entry *hash)
+ int hash_idx)
{
+ struct bond_entry *hash = from->hashes[hash_idx];
struct port *port = from->iface->port;
uint64_t delta = hash->tx_bytes;
@@ -2063,12 +2386,11 @@ bond_shift_load(struct slave_balance *from, struct slave_balance *to,
* it require more work, the only purpose it would be to allow that hash to
* be migrated to another slave in this rebalancing run, and there is no
* point in doing that. */
- if (from->hashes[0] == hash) {
+ if (hash_idx == 0) {
from->hashes++;
} else {
- int i = hash - from->hashes[0];
- memmove(from->hashes + i, from->hashes + i + 1,
- (from->n_hashes - (i + 1)) * sizeof *from->hashes);
+ memmove(from->hashes + hash_idx, from->hashes + hash_idx + 1,
+ (from->n_hashes - (hash_idx + 1)) * sizeof *from->hashes);
}
from->n_hashes--;
@@ -2153,22 +2475,60 @@ bond_rebalance_port(struct port *port)
/* 'from' is carrying significantly more load than 'to', and that
* load is split across at least two different hashes. Pick a hash
* to migrate to 'to' (the least-loaded slave), given that doing so
- * must not cause 'to''s load to exceed 'from''s load.
+ * must decrease the ratio of the load on the two slaves by at
+ * least 0.1.
*
* The sort order we use means that we prefer to shift away the
* smallest hashes instead of the biggest ones. There is little
* reason behind this decision; we could use the opposite sort
* order to shift away big hashes ahead of small ones. */
size_t i;
+ bool order_swapped;
for (i = 0; i < from->n_hashes; i++) {
+ double old_ratio, new_ratio;
uint64_t delta = from->hashes[i]->tx_bytes;
- if (to->tx_bytes + delta < from->tx_bytes - delta) {
+
+ if (delta == 0 || from->tx_bytes - delta == 0) {
+ /* Pointless move. */
+ continue;
+ }
+
+ order_swapped = from->tx_bytes - delta < to->tx_bytes + delta;
+
+ if (to->tx_bytes == 0) {
+ /* Nothing on the new slave, move it. */
+ break;
+ }
+
+ old_ratio = (double)from->tx_bytes / to->tx_bytes;
+ new_ratio = (double)(from->tx_bytes - delta) /
+ (to->tx_bytes + delta);
+
+ if (new_ratio == 0) {
+ /* Should already be covered but check to prevent division
+ * by zero. */
+ continue;
+ }
+
+ if (new_ratio < 1) {
+ new_ratio = 1 / new_ratio;
+ }
+
+ if (old_ratio - new_ratio > 0.1) {
+ /* Would decrease the ratio, move it. */
break;
}
}
if (i < from->n_hashes) {
- bond_shift_load(from, to, from->hashes[i]);
+ bond_shift_load(from, to, i);
+ port->bond_compat_is_stale = true;
+
+ /* If the result of the migration changed the relative order of
+ * 'from' and 'to' swap them back to maintain invariants. */
+ if (order_swapped) {
+ swap_bals(from, to);
+ }
/* Re-sort 'bals'. Note that this may make 'from' and 'to'
* point to different slave_balance structures. It is only
@@ -2197,17 +2557,14 @@ bond_send_learning_packets(struct port *port)
struct ofpbuf packet;
int error, n_packets, n_errors;
- if (!port->n_ifaces || port->active_iface < 0 || !br->ml) {
+ if (!port->n_ifaces || port->active_iface < 0) {
return;
}
ofpbuf_init(&packet, 128);
error = n_packets = n_errors = 0;
LIST_FOR_EACH (e, struct mac_entry, lru_node, &br->ml->lrus) {
- static const char s[] = "Open vSwitch Bond Failover";
union ofp_action actions[2], *a;
- struct eth_header *eth;
- struct llc_snap_header *llc_snap;
uint16_t dp_ifidx;
tag_type tags = 0;
flow_t flow;
@@ -2218,23 +2575,6 @@ bond_send_learning_packets(struct port *port)
continue;
}
- /* Compose packet to send. */
- ofpbuf_clear(&packet);
- eth = ofpbuf_put_zeros(&packet, ETH_HEADER_LEN);
- llc_snap = ofpbuf_put_zeros(&packet, LLC_SNAP_HEADER_LEN);
- ofpbuf_put(&packet, s, sizeof s); /* Includes null byte. */
- ofpbuf_put(&packet, e->mac, ETH_ADDR_LEN);
-
- memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
- memcpy(eth->eth_src, e->mac, ETH_ADDR_LEN);
- eth->eth_type = htons(packet.size - ETH_HEADER_LEN);
-
- llc_snap->llc.llc_dsap = LLC_DSAP_SNAP;
- llc_snap->llc.llc_ssap = LLC_SSAP_SNAP;
- llc_snap->llc.llc_cntl = LLC_CNTL_SNAP;
- memcpy(llc_snap->snap.snap_org, "\x00\x23\x20", 3);
- llc_snap->snap.snap_type = htons(0xf177); /* Random number. */
-
/* Compose actions. */
memset(actions, 0, sizeof actions);
a = actions;
@@ -2251,6 +2591,8 @@ bond_send_learning_packets(struct port *port)
/* Send packet. */
n_packets++;
+ compose_benign_packet(&packet, "Open vSwitch Bond Failover", 0xf177,
+ e->mac);
flow_extract(&packet, ODPP_NONE, &flow);
retval = ofproto_send_packet(br->ofproto, &flow, actions, a - actions,
&packet);
@@ -2275,7 +2617,8 @@ bond_send_learning_packets(struct port *port)
/* Bonding unixctl user interface functions. */
static void
-bond_unixctl_list(struct unixctl_conn *conn, const char *args UNUSED)
+bond_unixctl_list(struct unixctl_conn *conn,
+ const char *args UNUSED, void *aux UNUSED)
{
struct ds ds = DS_EMPTY_INITIALIZER;
const struct bridge *br;
@@ -2325,7 +2668,8 @@ bond_find(const char *name)
}
static void
-bond_unixctl_show(struct unixctl_conn *conn, const char *args)
+bond_unixctl_show(struct unixctl_conn *conn,
+ const char *args, void *aux UNUSED)
{
struct ds ds = DS_EMPTY_INITIALIZER;
const struct port *port;
@@ -2366,14 +2710,10 @@ bond_unixctl_show(struct unixctl_conn *conn, const char *args)
continue;
}
- ds_put_format(&ds, "\thash %d: %lld kB load\n",
+ ds_put_format(&ds, "\thash %d: %"PRIu64" kB load\n",
hash, be->tx_bytes / 1024);
/* MACs. */
- if (!port->bridge->ml) {
- break;
- }
-
LIST_FOR_EACH (me, struct mac_entry, lru_node,
&port->bridge->ml->lrus) {
uint16_t dp_ifidx;
@@ -2394,7 +2734,8 @@ bond_unixctl_show(struct unixctl_conn *conn, const char *args)
}
static void
-bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_)
+bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_,
+ void *aux UNUSED)
{
char *args = (char *) args_;
char *save_ptr = NULL;
@@ -2420,8 +2761,8 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_)
return;
}
- if (sscanf(hash_s, "%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8":%"SCNx8,
- &mac[0], &mac[1], &mac[2], &mac[3], &mac[4], &mac[5]) == 6) {
+ 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)) {
hash = atoi(hash_s) & BOND_MASK;
@@ -2445,11 +2786,13 @@ bond_unixctl_migrate(struct unixctl_conn *conn, const char *args_)
ofproto_revalidate(port->bridge->ofproto, entry->iface_tag);
entry->iface_idx = iface->port_ifidx;
entry->iface_tag = tag_create_random();
+ port->bond_compat_is_stale = true;
unixctl_command_reply(conn, 200, "migrated");
}
static void
-bond_unixctl_set_active_slave(struct unixctl_conn *conn, const char *args_)
+bond_unixctl_set_active_slave(struct unixctl_conn *conn, const char *args_,
+ void *aux UNUSED)
{
char *args = (char *) args_;
char *save_ptr = NULL;
@@ -2529,45 +2872,68 @@ enable_slave(struct unixctl_conn *conn, const char *args_, bool enable)
}
static void
-bond_unixctl_enable_slave(struct unixctl_conn *conn, const char *args)
+bond_unixctl_enable_slave(struct unixctl_conn *conn, const char *args,
+ void *aux UNUSED)
{
enable_slave(conn, args, true);
}
static void
-bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args)
+bond_unixctl_disable_slave(struct unixctl_conn *conn, const char *args,
+ void *aux UNUSED)
{
enable_slave(conn, args, false);
}
+static void
+bond_unixctl_hash(struct unixctl_conn *conn, const char *args,
+ void *aux UNUSED)
+{
+ uint8_t mac[ETH_ADDR_LEN];
+ uint8_t hash;
+ char *hash_cstr;
+
+ if (sscanf(args, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(mac))
+ == ETH_ADDR_SCAN_COUNT) {
+ hash = bond_hash(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
bond_init(void)
{
- unixctl_command_register("bond/list", bond_unixctl_list);
- unixctl_command_register("bond/show", bond_unixctl_show);
- unixctl_command_register("bond/migrate", bond_unixctl_migrate);
+ unixctl_command_register("bond/list", bond_unixctl_list, NULL);
+ unixctl_command_register("bond/show", bond_unixctl_show, NULL);
+ unixctl_command_register("bond/migrate", bond_unixctl_migrate, NULL);
unixctl_command_register("bond/set-active-slave",
- bond_unixctl_set_active_slave);
- unixctl_command_register("bond/enable-slave", bond_unixctl_enable_slave);
- unixctl_command_register("bond/disable-slave", bond_unixctl_disable_slave);
+ bond_unixctl_set_active_slave, NULL);
+ unixctl_command_register("bond/enable-slave", bond_unixctl_enable_slave,
+ NULL);
+ unixctl_command_register("bond/disable-slave", bond_unixctl_disable_slave,
+ NULL);
+ unixctl_command_register("bond/hash", bond_unixctl_hash, NULL);
}
/* Port functions. */
-static void
+static struct port *
port_create(struct bridge *br, const char *name)
{
struct port *port;
- port = xcalloc(1, sizeof *port);
+ port = xzalloc(sizeof *port);
port->bridge = br;
port->port_idx = br->n_ports;
port->vlan = -1;
port->trunks = NULL;
port->name = xstrdup(name);
port->active_iface = -1;
- port->stp_state = STP_DISABLED;
- port->stp_state_tag = 0;
if (br->n_ports >= br->allocated_ports) {
br->ports = x2nrealloc(br->ports, &br->allocated_ports,
@@ -2577,70 +2943,69 @@ port_create(struct bridge *br, const char *name)
VLOG_INFO("created port %s on bridge %s", port->name, br->name);
bridge_flush(br);
+
+ return port;
}
static void
-port_reconfigure(struct port *port)
+port_reconfigure(struct port *port, const struct ovsrec_port *cfg)
{
- bool bonded = cfg_has_section("bonding.%s", port->name);
- struct svec old_ifaces, new_ifaces;
+ struct shash old_ifaces, new_ifaces;
+ struct shash_node *node;
unsigned long *trunks;
int vlan;
size_t i;
+ port->cfg = cfg;
+
/* Collect old and new interfaces. */
- svec_init(&old_ifaces);
- svec_init(&new_ifaces);
+ shash_init(&old_ifaces);
+ shash_init(&new_ifaces);
for (i = 0; i < port->n_ifaces; i++) {
- svec_add(&old_ifaces, port->ifaces[i]->name);
+ shash_add(&old_ifaces, port->ifaces[i]->name, port->ifaces[i]);
}
- svec_sort(&old_ifaces);
- if (bonded) {
- cfg_get_all_keys(&new_ifaces, "bonding.%s.slave", port->name);
- if (!new_ifaces.n) {
- VLOG_ERR("port %s: no interfaces specified for bonded port",
- port->name);
- } else if (new_ifaces.n == 1) {
- VLOG_WARN("port %s: only 1 interface specified for bonded port",
- port->name);
- }
-
- port->updelay = cfg_get_int(0, "bonding.%s.updelay", port->name);
- if (port->updelay < 0) {
- port->updelay = 0;
- }
- port->downdelay = cfg_get_int(0, "bonding.%s.downdelay", port->name);
- if (port->downdelay < 0) {
- port->downdelay = 0;
+ for (i = 0; i < cfg->n_interfaces; i++) {
+ const char *name = cfg->interfaces[i]->name;
+ if (!shash_add_once(&new_ifaces, name, cfg->interfaces[i])) {
+ VLOG_WARN("port %s: %s specified twice as port interface",
+ port->name, name);
}
- } else {
- svec_init(&new_ifaces);
- svec_add(&new_ifaces, port->name);
+ }
+ port->updelay = cfg->bond_updelay;
+ if (port->updelay < 0) {
+ port->updelay = 0;
+ }
+ port->updelay = cfg->bond_downdelay;
+ if (port->downdelay < 0) {
+ port->downdelay = 0;
}
/* Get rid of deleted interfaces and add new interfaces. */
- for (i = 0; i < port->n_ifaces; i++) {
- struct iface *iface = port->ifaces[i];
- if (!svec_contains(&new_ifaces, iface->name)) {
- iface_destroy(iface);
- } else {
- i++;
+ SHASH_FOR_EACH (node, &old_ifaces) {
+ if (!shash_find(&new_ifaces, node->name)) {
+ iface_destroy(node->data);
}
}
- for (i = 0; i < new_ifaces.n; i++) {
- const char *name = new_ifaces.names[i];
- if (!svec_contains(&old_ifaces, name)) {
- iface_create(port, name);
+ SHASH_FOR_EACH (node, &new_ifaces) {
+ const struct ovsrec_interface *if_cfg = node->data;
+ struct iface *iface;
+
+ iface = shash_find_data(&old_ifaces, if_cfg->name);
+ if (!iface) {
+ iface = iface_create(port, if_cfg);
}
+ iface->cfg = if_cfg;
}
/* Get VLAN tag. */
vlan = -1;
- if (cfg_has("vlan.%s.tag", port->name)) {
- if (!bonded) {
- vlan = cfg_get_vlan(0, "vlan.%s.tag", port->name);
+ if (cfg->tag) {
+ if (port->n_ifaces < 2) {
+ vlan = *cfg->tag;
if (vlan >= 0 && vlan <= 4095) {
VLOG_DBG("port %s: assigning VLAN tag %d", port->name, vlan);
+ } else {
+ vlan = -1;
}
} else {
/* It's possible that bonded, VLAN-tagged ports make sense. Maybe
@@ -2657,14 +3022,13 @@ port_reconfigure(struct port *port)
/* Get trunked VLANs. */
trunks = NULL;
if (vlan < 0) {
- size_t n_trunks, n_errors;
+ size_t n_errors;
size_t i;
trunks = bitmap_allocate(4096);
- n_trunks = cfg_count("vlan.%s.trunks", port->name);
n_errors = 0;
- for (i = 0; i < n_trunks; i++) {
- int trunk = cfg_get_vlan(i, "vlan.%s.trunks", port->name);
+ for (i = 0; i < cfg->n_trunks; i++) {
+ int trunk = cfg->trunks[i];
if (trunk >= 0) {
bitmap_set1(trunks, trunk);
} else {
@@ -2673,9 +3037,9 @@ port_reconfigure(struct port *port)
}
if (n_errors) {
VLOG_ERR("port %s: invalid values for %zu trunk VLANs",
- port->name, n_trunks);
+ port->name, cfg->n_trunks);
}
- if (n_errors == n_trunks) {
+ if (n_errors == cfg->n_trunks) {
if (n_errors) {
VLOG_ERR("port %s: no valid trunks, trunking all VLANs",
port->name);
@@ -2683,9 +3047,9 @@ port_reconfigure(struct port *port)
bitmap_set_multiple(trunks, 0, 4096, 1);
}
} else {
- if (cfg_has("vlan.%s.trunks", port->name)) {
- VLOG_ERR("ignoring vlan.%s.trunks in favor of vlan.%s.vlan",
- port->name, port->name);
+ if (cfg->n_trunks) {
+ VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
+ port->name);
}
}
if (trunks == NULL
@@ -2696,8 +3060,8 @@ port_reconfigure(struct port *port)
bitmap_free(port->trunks);
port->trunks = trunks;
- svec_destroy(&old_ifaces);
- svec_destroy(&new_ifaces);
+ shash_destroy(&old_ifaces);
+ shash_destroy(&new_ifaces);
}
static void
@@ -2706,16 +3070,18 @@ port_destroy(struct port *port)
if (port) {
struct bridge *br = port->bridge;
struct port *del;
- size_t i;
proc_net_compat_update_vlan(port->name, NULL, 0);
+ proc_net_compat_update_bond(port->name, NULL);
+#if 0
for (i = 0; i < MAX_MIRRORS; i++) {
struct mirror *m = br->mirrors[i];
if (m && m->out_port == port) {
mirror_destroy(m);
}
}
+#endif
while (port->n_ifaces > 0) {
iface_destroy(port->ifaces[port->n_ifaces - 1]);
@@ -2775,7 +3141,8 @@ port_update_bonding(struct port *port)
if (port->bond_hash) {
free(port->bond_hash);
port->bond_hash = NULL;
- proc_net_compat_update_bond(port->name, NULL);
+ port->bond_compat_is_stale = true;
+ port->bond_fake_iface = false;
}
} else {
if (!port->bond_hash) {
@@ -2790,36 +3157,80 @@ port_update_bonding(struct port *port)
port->no_ifaces_tag = tag_create_random();
bond_choose_active_iface(port);
}
- port_update_bond_compat(port);
+ port->bond_compat_is_stale = true;
+ port->bond_fake_iface = port->cfg->bond_fake_iface;
}
}
static void
port_update_bond_compat(struct port *port)
{
+ struct compat_bond_hash compat_hashes[BOND_MASK + 1];
struct compat_bond bond;
size_t i;
if (port->n_ifaces < 2) {
+ proc_net_compat_update_bond(port->name, NULL);
return;
}
bond.up = false;
bond.updelay = port->updelay;
bond.downdelay = port->downdelay;
+
+ bond.n_hashes = 0;
+ bond.hashes = compat_hashes;
+ if (port->bond_hash) {
+ const struct bond_entry *e;
+ for (e = port->bond_hash; e <= &port->bond_hash[BOND_MASK]; e++) {
+ if (e->iface_idx >= 0 && e->iface_idx < port->n_ifaces) {
+ struct compat_bond_hash *cbh = &bond.hashes[bond.n_hashes++];
+ cbh->hash = e - port->bond_hash;
+ cbh->netdev_name = port->ifaces[e->iface_idx]->name;
+ }
+ }
+ }
+
bond.n_slaves = port->n_ifaces;
bond.slaves = xmalloc(port->n_ifaces * sizeof *bond.slaves);
for (i = 0; i < port->n_ifaces; i++) {
struct iface *iface = port->ifaces[i];
struct compat_bond_slave *slave = &bond.slaves[i];
slave->name = iface->name;
- slave->up = ((iface->enabled && iface->delay_expires == LLONG_MAX) ||
- (!iface->enabled && iface->delay_expires != LLONG_MAX));
+
+ /* 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
+ * 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
+ * to be "up" and anything else to be "down".
+ */
+ slave->up = iface->enabled && iface->delay_expires == LLONG_MAX;
if (slave->up) {
bond.up = true;
}
- memcpy(slave->mac, iface->mac, ETH_ADDR_LEN);
+ netdev_get_etheraddr(iface->netdev, slave->mac);
+ }
+
+ if (port->bond_fake_iface) {
+ struct netdev *bond_netdev;
+
+ if (!netdev_open(port->name, NETDEV_ETH_TYPE_NONE, &bond_netdev)) {
+ if (bond.up) {
+ netdev_turn_flags_on(bond_netdev, NETDEV_UP, true);
+ } else {
+ netdev_turn_flags_off(bond_netdev, NETDEV_UP, true);
+ }
+ netdev_close(bond_netdev);
+ }
}
+
proc_net_compat_update_bond(port->name, &bond);
free(bond.slaves);
}
@@ -2848,7 +3259,8 @@ port_update_vlan_compat(struct port *port)
&& p->n_ifaces
&& (!vlandev_name || strcmp(p->name, vlandev_name) <= 0))
{
- const uint8_t *ea = p->ifaces[0]->mac;
+ uint8_t ea[ETH_ADDR_LEN];
+ netdev_get_etheraddr(p->ifaces[0]->netdev, ea);
if (!eth_addr_is_multicast(ea) &&
!eth_addr_is_reserved(ea) &&
!eth_addr_is_zero(ea)) {
@@ -2862,21 +3274,21 @@ port_update_vlan_compat(struct port *port)
/* Interface functions. */
-static void
-iface_create(struct port *port, const char *name)
+static struct iface *
+iface_create(struct port *port, const struct ovsrec_interface *if_cfg)
{
struct iface *iface;
+ char *name = if_cfg->name;
+ int error;
- iface = xcalloc(1, sizeof *iface);
+ iface = xzalloc(sizeof *iface);
iface->port = port;
iface->port_ifidx = port->n_ifaces;
iface->name = xstrdup(name);
iface->dp_ifidx = -1;
iface->tag = tag_create_random();
iface->delay_expires = LLONG_MAX;
-
- netdev_nodev_get_etheraddr(name, iface->mac);
- netdev_nodev_get_carrier(name, &iface->enabled);
+ iface->netdev = NULL;
if (port->n_ifaces >= port->allocated_ifaces) {
port->ifaces = x2nrealloc(port->ifaces, &port->allocated_ifaces,
@@ -2887,10 +3299,19 @@ iface_create(struct port *port, const char *name)
port->bridge->has_bonded_ports = true;
}
+ /* Attempt to create the network interface in case it
+ * doesn't exist yet. */
+ error = set_up_iface(if_cfg, true);
+ if (error) {
+ VLOG_WARN("could not create iface %s: %s\n", iface->name,
+ strerror(error));
+ }
+
VLOG_DBG("attached network device %s to port %s", iface->name, port->name);
- port_update_bonding(port);
bridge_flush(port->bridge);
+
+ return iface;
}
static void
@@ -2909,8 +3330,7 @@ iface_destroy(struct iface *iface)
del = port->ifaces[iface->port_ifidx] = port->ifaces[--port->n_ifaces];
del->port_ifidx = iface->port_ifidx;
- free(iface->name);
- free(iface);
+ netdev_close(iface->netdev);
if (del_active) {
ofproto_revalidate(port->bridge->ofproto, port->active_iface_tag);
@@ -2918,7 +3338,10 @@ iface_destroy(struct iface *iface)
bond_send_learning_packets(port);
}
- port_update_bonding(port);
+ netdev_destroy(iface->name);
+ free(iface->name);
+ free(iface);
+
bridge_flush(port->bridge);
}
}
@@ -2945,14 +3368,73 @@ 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..internal = true".
+ *
+ * In addition, we have a kluge-y feature that creates an internal port with
+ * the name of a bonded port if "bonding..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)
+{
+ /* XXX wastes time */
+ 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->n_ifaces > 1 && port->cfg->bond_fake_iface) {
+ return true;
+ }
+ return false;
+}
+
+/* Set Ethernet address of 'iface', if one is specified in the configuration
+ * file. */
+static void
+iface_set_mac(struct iface *iface)
+{
+ uint8_t ea[ETH_ADDR_LEN];
+
+ if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
+ if (eth_addr_is_multicast(ea)) {
+ VLOG_ERR("interface %s: cannot set MAC to multicast address",
+ iface->name);
+ } else if (iface->dp_ifidx == ODPP_LOCAL) {
+ VLOG_ERR("ignoring iface.%s.mac; use bridge.%s.mac instead",
+ iface->name, iface->name);
+ } else {
+ int error = netdev_set_etheraddr(iface->netdev, ea);
+ if (error) {
+ VLOG_ERR("interface %s: setting MAC failed (%s)",
+ iface->name, strerror(error));
+ }
+ }
+ }
+}
/* Port mirroring. */
+#if 0
static void
-mirror_reconfigure(struct bridge *br)
+mirror_reconfigure(struct bridge *br UNUSED)
{
struct svec old_mirrors, new_mirrors;
- size_t i;
+ size_t i, n_rspan_vlans;
+ unsigned long *rspan_vlans;
/* Collect old and new mirrors. */
svec_init(&old_mirrors);
@@ -3001,6 +3483,29 @@ mirror_reconfigure(struct bridge *br)
m->out_port->is_mirror_output_port = true;
}
}
+
+ /* Update learning disabled vlans (for RSPAN). */
+ rspan_vlans = NULL;
+ n_rspan_vlans = cfg_count("vlan.%s.disable-learning", br->name);
+ if (n_rspan_vlans) {
+ rspan_vlans = bitmap_allocate(4096);
+
+ for (i = 0; i < n_rspan_vlans; i++) {
+ int vlan = cfg_get_vlan(i, "vlan.%s.disable-learning", br->name);
+ if (vlan >= 0) {
+ bitmap_set1(rspan_vlans, vlan);
+ VLOG_INFO("bridge %s: disabling learning on vlan %d\n",
+ br->name, vlan);
+ } else {
+ VLOG_ERR("bridge %s: invalid value '%s' for learning disabled "
+ "VLAN", br->name,
+ cfg_get_string(i, "vlan.%s.disable-learning", br->name));
+ }
+ }
+ }
+ if (mac_learning_set_disabled_vlans(br->ml, rspan_vlans)) {
+ bridge_flush(br);
+ }
}
static void
@@ -3023,7 +3528,7 @@ mirror_create(struct bridge *br, const char *name)
VLOG_INFO("created port mirror %s on bridge %s", name, br->name);
bridge_flush(br);
- br->mirrors[i] = m = xcalloc(1, sizeof *m);
+ br->mirrors[i] = m = xzalloc(sizeof *m);
m->bridge = br;
m->idx = i;
m->name = xstrdup(name);
@@ -3132,7 +3637,7 @@ port_trunks_any_mirrored_vlan(const struct mirror *m, const struct port *p)
}
static void
-mirror_reconfigure_one(struct mirror *m)
+mirror_reconfigure_one(struct mirror *m UNUSED)
{
char *pfx = xasprintf("mirror.%s.%s", m->bridge->name, m->name);
struct svec src_ports, dst_ports, ports;
@@ -3145,6 +3650,7 @@ mirror_reconfigure_one(struct mirror *m)
int *vlans;
size_t i;
bool mirror_all_ports;
+ bool any_ports_specified;
/* Get output port. */
out_port_name = cfg_get_key(0, "mirror.%s.%s.output.port",
@@ -3183,11 +3689,18 @@ mirror_reconfigure_one(struct mirror *m)
cfg_get_all_keys(&src_ports, "%s.select.src-port", pfx);
cfg_get_all_keys(&dst_ports, "%s.select.dst-port", pfx);
cfg_get_all_keys(&ports, "%s.select.port", pfx);
+ any_ports_specified = src_ports.n || dst_ports.n || ports.n;
svec_append(&src_ports, &ports);
svec_append(&dst_ports, &ports);
svec_destroy(&ports);
prune_ports(m, &src_ports);
prune_ports(m, &dst_ports);
+ if (any_ports_specified && !src_ports.n && !dst_ports.n) {
+ VLOG_ERR("%s: none of the specified ports exist; "
+ "disabling port mirror %s", pfx, pfx);
+ mirror_destroy(m);
+ goto exit;
+ }
/* Get all the vlans, and drop duplicate and invalid vlans. */
svec_init(&vlan_strings);
@@ -3239,214 +3752,9 @@ mirror_reconfigure_one(struct mirror *m)
}
/* Clean up. */
+exit:
svec_destroy(&src_ports);
svec_destroy(&dst_ports);
free(pfx);
}
-
-/* Spanning tree protocol. */
-
-static void brstp_update_port_state(struct port *);
-
-static void
-brstp_send_bpdu(struct ofpbuf *pkt, int port_no, void *br_)
-{
- struct bridge *br = br_;
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
- struct iface *iface = iface_from_dp_ifidx(br, port_no);
- if (!iface) {
- VLOG_WARN_RL(&rl, "%s: cannot send BPDU on unknown port %d",
- br->name, port_no);
- } else if (eth_addr_is_zero(iface->mac)) {
- VLOG_WARN_RL(&rl, "%s: cannot send BPDU on port %d with unknown MAC",
- br->name, port_no);
- } else {
- union ofp_action action;
- struct eth_header *eth = pkt->l2;
- flow_t flow;
-
- memcpy(eth->eth_src, iface->mac, ETH_ADDR_LEN);
-
- memset(&action, 0, sizeof action);
- action.type = htons(OFPAT_OUTPUT);
- action.output.len = htons(sizeof action);
- action.output.port = htons(port_no);
-
- flow_extract(pkt, ODPP_NONE, &flow);
- ofproto_send_packet(br->ofproto, &flow, &action, 1, pkt);
- }
- ofpbuf_delete(pkt);
-}
-
-static void
-brstp_reconfigure(struct bridge *br)
-{
- size_t i;
-
- if (!cfg_get_bool(0, "stp.%s.enabled", br->name)) {
- if (br->stp) {
- stp_destroy(br->stp);
- br->stp = NULL;
-
- bridge_flush(br);
- }
- } else {
- uint64_t bridge_address, bridge_id;
- int bridge_priority;
-
- bridge_address = cfg_get_mac(0, "stp.%s.address", br->name);
- if (!bridge_address) {
- if (br->stp) {
- bridge_address = (stp_get_bridge_id(br->stp)
- & ((UINT64_C(1) << 48) - 1));
- } else {
- uint8_t mac[ETH_ADDR_LEN];
- eth_addr_random(mac);
- bridge_address = eth_addr_to_uint64(mac);
- }
- }
-
- if (cfg_is_valid(CFG_INT | CFG_REQUIRED, "stp.%s.priority",
- br->name)) {
- bridge_priority = cfg_get_int(0, "stp.%s.priority", br->name);
- } else {
- bridge_priority = STP_DEFAULT_BRIDGE_PRIORITY;
- }
-
- bridge_id = bridge_address | ((uint64_t) bridge_priority << 48);
- if (!br->stp) {
- br->stp = stp_create(br->name, bridge_id, brstp_send_bpdu, br);
- br->stp_last_tick = time_msec();
- bridge_flush(br);
- } else {
- if (bridge_id != stp_get_bridge_id(br->stp)) {
- stp_set_bridge_id(br->stp, bridge_id);
- bridge_flush(br);
- }
- }
-
- for (i = 0; i < br->n_ports; i++) {
- struct port *p = br->ports[i];
- int dp_ifidx;
- struct stp_port *sp;
- int path_cost, priority;
- bool enable;
-
- if (!p->n_ifaces) {
- continue;
- }
- dp_ifidx = p->ifaces[0]->dp_ifidx;
- if (dp_ifidx < 0 || dp_ifidx >= STP_MAX_PORTS) {
- continue;
- }
-
- sp = stp_get_port(br->stp, dp_ifidx);
- enable = (!cfg_is_valid(CFG_BOOL | CFG_REQUIRED,
- "stp.%s.port.%s.enabled",
- br->name, p->name)
- || cfg_get_bool(0, "stp.%s.port.%s.enabled",
- br->name, p->name));
- if (p->is_mirror_output_port) {
- enable = false;
- }
- if (enable != (stp_port_get_state(sp) != STP_DISABLED)) {
- bridge_flush(br); /* Might not be necessary. */
- if (enable) {
- stp_port_enable(sp);
- } else {
- stp_port_disable(sp);
- }
- }
-
- path_cost = cfg_get_int(0, "stp.%s.port.%s.path-cost",
- br->name, p->name);
- stp_port_set_path_cost(sp, path_cost ? path_cost : 19 /* XXX */);
-
- priority = (cfg_is_valid(CFG_INT | CFG_REQUIRED,
- "stp.%s.port.%s.priority",
- br->name, p->name)
- ? cfg_get_int(0, "stp.%s.port.%s.priority",
- br->name, p->name)
- : STP_DEFAULT_PORT_PRIORITY);
- stp_port_set_priority(sp, priority);
- }
-
- brstp_adjust_timers(br);
- }
- for (i = 0; i < br->n_ports; i++) {
- brstp_update_port_state(br->ports[i]);
- }
-}
-
-static void
-brstp_update_port_state(struct port *p)
-{
- struct bridge *br = p->bridge;
- enum stp_state state;
-
- /* Figure out new state. */
- state = STP_DISABLED;
- if (br->stp && p->n_ifaces > 0) {
- int dp_ifidx = p->ifaces[0]->dp_ifidx;
- if (dp_ifidx >= 0 && dp_ifidx < STP_MAX_PORTS) {
- state = stp_port_get_state(stp_get_port(br->stp, dp_ifidx));
- }
- }
-
- /* Update state. */
- if (p->stp_state != state) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
- VLOG_INFO_RL(&rl, "port %s: STP state changed from %s to %s",
- p->name, stp_state_name(p->stp_state),
- stp_state_name(state));
- if (p->stp_state == STP_DISABLED) {
- bridge_flush(br);
- } else {
- ofproto_revalidate(p->bridge->ofproto, p->stp_state_tag);
- }
- p->stp_state = state;
- p->stp_state_tag = (p->stp_state == STP_DISABLED ? 0
- : tag_create_random());
- }
-}
-
-static void
-brstp_adjust_timers(struct bridge *br)
-{
- int hello_time = cfg_get_int(0, "stp.%s.hello-time", br->name);
- int max_age = cfg_get_int(0, "stp.%s.max-age", br->name);
- int forward_delay = cfg_get_int(0, "stp.%s.forward-delay", br->name);
-
- stp_set_hello_time(br->stp, hello_time ? hello_time : 2000);
- stp_set_max_age(br->stp, max_age ? max_age : 20000);
- stp_set_forward_delay(br->stp, forward_delay ? forward_delay : 15000);
-}
-
-static void
-brstp_run(struct bridge *br)
-{
- if (br->stp) {
- long long int now = time_msec();
- long long int elapsed = now - br->stp_last_tick;
- struct stp_port *sp;
-
- if (elapsed > 0) {
- stp_tick(br->stp, MIN(INT_MAX, elapsed));
- br->stp_last_tick = now;
- }
- while (stp_get_changed_port(br->stp, &sp)) {
- struct port *p = port_from_dp_ifidx(br, stp_port_no(sp));
- if (p) {
- brstp_update_port_state(p);
- }
- }
- }
-}
-
-static void
-brstp_wait(struct bridge *br)
-{
- if (br->stp) {
- poll_timer_wait(1000);
- }
-}
+#endif