-/* Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
+/* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "bridge.h"
-#include <assert.h>
#include <errno.h>
#include <inttypes.h>
#include <stdlib.h>
/* OVSDB IDL used to obtain configuration. */
static struct ovsdb_idl *idl;
+/* We want to complete daemonization, fully detaching from our parent process,
+ * only after we have completed our initial configuration, committed our state
+ * to the database, and received confirmation back from the database server
+ * that it applied the commit. This allows our parent process to know that,
+ * post-detach, ephemeral fields such as datapath-id and ofport are very likely
+ * to have already been filled in. (It is only "very likely" rather than
+ * certain because there is always a slim possibility that the transaction will
+ * fail or that some other client has added new bridges, ports, etc. while
+ * ovs-vswitchd was configuring using an old configuration.)
+ *
+ * We only need to do this once for our initial configuration at startup, so
+ * 'initial_config_done' tracks whether we've already done it. While we are
+ * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
+ * itself and is otherwise NULL. */
+static bool initial_config_done;
+static struct ovsdb_idl_txn *daemonize_txn;
+
/* Most recently processed IDL sequence number. */
static unsigned int idl_seqno;
static void bridge_configure_flow_eviction_threshold(struct bridge *);
static void bridge_configure_netflow(struct bridge *);
static void bridge_configure_forward_bpdu(struct bridge *);
-static void bridge_configure_mac_idle_time(struct bridge *);
+static void bridge_configure_mac_table(struct bridge *);
static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
static void bridge_configure_stp(struct bridge *);
static void bridge_configure_tables(struct bridge *);
+static void bridge_configure_dp_desc(struct bridge *);
static void bridge_configure_remotes(struct bridge *,
const struct sockaddr_in *managers,
size_t n_managers);
static void iface_refresh_stats(struct iface *);
static void iface_refresh_status(struct iface *);
static bool iface_is_synthetic(const struct iface *);
+static int64_t iface_pick_ofport(const struct ovsrec_interface *);
/* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
*
iface_hint = xmalloc(sizeof *iface_hint);
iface_hint->br_name = br_cfg->name;
iface_hint->br_type = br_cfg->datapath_type;
- iface_hint->ofp_port = if_cfg->n_ofport_request ?
- *if_cfg->ofport_request : OFPP_NONE;
+ iface_hint->ofp_port = iface_pick_ofport(if_cfg);
shash_add(&iface_hints, if_cfg->name, iface_hint);
}
bridge_init(const char *remote)
{
/* Create connection to database. */
- idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true);
+ idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
idl_seqno = ovsdb_idl_get_seqno(idl);
ovsdb_idl_set_lock(idl, "ovs_vswitchd");
ovsdb_idl_verify_write_only(idl);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
+ ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
COVERAGE_INC(bridge_reconfigure);
- assert(!reconfiguring);
+ ovs_assert(!reconfiguring);
reconfiguring = true;
/* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
struct bridge *br;
bool done;
- assert(reconfiguring);
+ ovs_assert(reconfiguring);
done = bridge_reconfigure_ofp();
/* Complete the configuration. */
bridge_configure_mirrors(br);
bridge_configure_flow_eviction_threshold(br);
bridge_configure_forward_bpdu(br);
- bridge_configure_mac_idle_time(br);
+ bridge_configure_mac_table(br);
bridge_configure_remotes(br, managers, n_managers);
bridge_configure_netflow(br);
bridge_configure_sflow(br, &sflow_bridge_number);
bridge_configure_stp(br);
bridge_configure_tables(br);
+ bridge_configure_dp_desc(br);
}
free(managers);
- if (done) {
- /* ovs-vswitchd has completed initialization, so allow the process that
- * forked us to exit successfully. */
- daemonize_complete();
- reconfiguring = false;
-
- VLOG_INFO("%s (Open vSwitch) %s", program_name, VERSION);
- }
-
return done;
}
{
struct bridge *br = iface->port->bridge;
- assert(iface->ofp_port < 0 && ofp_port >= 0);
+ ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
iface->ofp_port = ofp_port;
hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
iface_set_ofport(iface->cfg, ofp_port);
* internal datastructures may not be consistent. Eventually, when port
* additions and deletions are cheaper, these calls should be removed. */
bridge_run_fast();
- assert(!iface_lookup(br, iface_cfg->name));
+ ovs_assert(!iface_lookup(br, iface_cfg->name));
error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
bridge_run_fast();
if (error) {
+ iface_set_ofport(iface_cfg, -1);
iface_clear_db_record(iface_cfg);
ok = false;
goto done;
error = netdev_open(port->name, "internal", &netdev);
if (!error) {
- uint16_t ofp_port = if_cfg->ofport;
+ uint16_t fake_ofp_port = if_cfg->ofport;
- ofproto_port_add(br->ofproto, netdev, &ofp_port);
+ ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
netdev_close(netdev);
} else {
VLOG_WARN("could not open network device %s (%s)",
false));
}
-/* Set MAC aging time for 'br'. */
+/* Set MAC learning table configuration for 'br'. */
static void
-bridge_configure_mac_idle_time(struct bridge *br)
+bridge_configure_mac_table(struct bridge *br)
{
const char *idle_time_str;
int idle_time;
+ const char *mac_table_size_str;
+ int mac_table_size;
+
idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
idle_time = (idle_time_str && atoi(idle_time_str)
? atoi(idle_time_str)
: MAC_ENTRY_DEFAULT_IDLE_TIME);
- ofproto_set_mac_idle_time(br->ofproto, idle_time);
+
+ mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
+ mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
+ ? atoi(mac_table_size_str)
+ : MAC_DEFAULT_MAX);
+
+ ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
}
static void
found_addr = true;
}
}
- if (found_addr) {
- VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT,
- br->name, ETH_ADDR_ARGS(ea));
- } else {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 10);
+
+ if (!found_addr) {
memcpy(ea, br->default_ea, ETH_ADDR_LEN);
*hw_addr_iface = NULL;
- VLOG_WARN_RL(&rl, "bridge %s: using default bridge Ethernet "
- "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
}
hmapx_destroy(&mirror_output_ports);
int64_t bps;
int mtu;
int64_t mtu_64;
+ uint8_t mac[ETH_ADDR_LEN];
int error;
if (iface_is_synthetic(iface)) {
smap_init(&smap);
- if (!netdev_get_drv_info(iface->netdev, &smap)) {
+ if (!netdev_get_status(iface->netdev, &smap)) {
ovsrec_interface_set_status(iface->cfg, &smap);
} else {
ovsrec_interface_set_status(iface->cfg, NULL);
else {
ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
}
+
+ error = netdev_get_etheraddr(iface->netdev, mac);
+ if (!error) {
+ char mac_string[32];
+
+ sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
+ ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
+ } else {
+ ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
+ }
}
/* Writes 'iface''s CFM statistics to the database. 'iface' must not be
#define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
IFACE_STATS;
#undef IFACE_STAT
- assert(i == ARRAY_SIZE(keys));
+ ovs_assert(i == ARRAY_SIZE(keys));
ovsrec_interface_set_statistics(iface->cfg, keys, values,
ARRAY_SIZE(keys));
ofproto_free_ofproto_controller_info(&info);
}
+\f
+/* "Instant" stats.
+ *
+ * Some information in the database must be kept as up-to-date as possible to
+ * allow controllers to respond rapidly to network outages. We call these
+ * statistics "instant" stats.
+ *
+ * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
+ * assuming that they've changed. The only means we have to determine whether
+ * they have changed are:
+ *
+ * - Try to commit changes to the database. If nothing changed, then
+ * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
+ * value.
+ *
+ * - instant_stats_run() is called late in the run loop, after anything that
+ * might change any of the instant stats.
+ *
+ * We use these two facts together to avoid waking the process up every
+ * INSTANT_INTERVAL_MSEC whether there is any change or not.
+ */
+
+/* Minimum interval between writing updates to the instant stats to the
+ * database. */
+#define INSTANT_INTERVAL_MSEC 100
+
+/* Current instant stats database transaction, NULL if there is no ongoing
+ * transaction. */
+static struct ovsdb_idl_txn *instant_txn;
+
+/* Next time (in msec on monotonic clock) at which we will update the instant
+ * stats. */
+static long long int instant_next_txn = LLONG_MIN;
+
+/* True if the run loop has run since we last saw that the instant stats were
+ * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
+ * to refresh the instant stats. */
+static bool instant_stats_could_have_changed;
static void
-refresh_instant_stats(void)
+instant_stats_run(void)
{
- static struct ovsdb_idl_txn *txn = NULL;
+ enum ovsdb_idl_txn_status status;
- if (!txn) {
+ instant_stats_could_have_changed = true;
+
+ if (!instant_txn) {
struct bridge *br;
- txn = ovsdb_idl_txn_create(idl);
+ if (time_msec() < instant_next_txn) {
+ return;
+ }
+ instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
+ instant_txn = ovsdb_idl_txn_create(idl);
HMAP_FOR_EACH (br, node, &all_bridges) {
struct iface *iface;
struct port *port;
}
}
- if (ovsdb_idl_txn_commit(txn) != TXN_INCOMPLETE) {
- ovsdb_idl_txn_destroy(txn);
- txn = NULL;
+ status = ovsdb_idl_txn_commit(instant_txn);
+ if (status != TXN_INCOMPLETE) {
+ ovsdb_idl_txn_destroy(instant_txn);
+ instant_txn = NULL;
+ }
+ if (status == TXN_UNCHANGED) {
+ instant_stats_could_have_changed = false;
}
}
+static void
+instant_stats_wait(void)
+{
+ if (instant_txn) {
+ ovsdb_idl_txn_wait(instant_txn);
+ } else if (instant_stats_could_have_changed) {
+ poll_timer_wait_until(instant_next_txn);
+ }
+}
+\f
/* Performs periodic activity required by bridges that needs to be done with
* the least possible latency.
*
void
bridge_run(void)
{
- static const struct ovsrec_open_vswitch null_cfg;
+ static struct ovsrec_open_vswitch null_cfg;
const struct ovsrec_open_vswitch *cfg;
struct ovsdb_idl_txn *reconf_txn = NULL;
struct sset types;
bool vlan_splinters_changed;
struct bridge *br;
- ovsrec_open_vswitch_init((struct ovsrec_open_vswitch *) &null_cfg);
+ ovsrec_open_vswitch_init(&null_cfg);
/* (Re)configure if necessary. */
if (!reconfiguring) {
HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
bridge_destroy(br);
}
+ /* Since we will not be running system_stats_run() in this process
+ * with the current situation of multiple ovs-vswitchd daemons,
+ * disable system stats collection. */
+ system_stats_enable(false);
return;
} else if (!ovsdb_idl_has_lock(idl)) {
return;
}
if (reconfiguring) {
- if (cfg) {
- if (!reconf_txn) {
- reconf_txn = ovsdb_idl_txn_create(idl);
- }
- if (bridge_reconfigure_continue(cfg)) {
+ if (!reconf_txn) {
+ reconf_txn = ovsdb_idl_txn_create(idl);
+ }
+
+ if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
+ reconfiguring = false;
+
+ if (cfg) {
ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
}
- } else {
- bridge_reconfigure_continue(&null_cfg);
+
+ /* If we are completing our initial configuration for this run
+ * of ovs-vswitchd, then keep the transaction around to monitor
+ * it for completion. */
+ if (!initial_config_done) {
+ initial_config_done = true;
+ daemonize_txn = reconf_txn;
+ reconf_txn = NULL;
+ }
}
}
reconf_txn = NULL;
}
+ if (daemonize_txn) {
+ enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
+ if (status != TXN_INCOMPLETE) {
+ ovsdb_idl_txn_destroy(daemonize_txn);
+ daemonize_txn = NULL;
+
+ /* ovs-vswitchd has completed initialization, so allow the
+ * process that forked us to exit successfully. */
+ daemonize_complete();
+
+ VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
+ }
+ }
+
/* Refresh interface and mirror stats if necessary. */
if (time_msec() >= iface_stats_timer) {
if (cfg) {
}
run_system_stats();
- refresh_instant_stats();
+ instant_stats_run();
}
void
const char *type;
ovsdb_idl_wait(idl);
+ if (daemonize_txn) {
+ ovsdb_idl_txn_wait(daemonize_txn);
+ }
if (reconfiguring) {
poll_immediate_wake();
}
system_stats_wait();
+ instant_stats_wait();
}
/* Adds some memory usage statistics for bridges into 'usage', for use with
{
struct bridge *br;
- assert(!bridge_lookup(br_cfg->name));
+ ovs_assert(!bridge_lookup(br_cfg->name));
br = xzalloc(sizeof *br);
br->name = xstrdup(br_cfg->name);
if_cfg->cfg = cfg;
if_cfg->parent = parent;
- if_cfg->ofport = cfg->n_ofport_request ? *cfg->ofport_request : OFPP_NONE;
+ if_cfg->ofport = iface_pick_ofport(cfg);
hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
hash_string(if_cfg->cfg->name, 0));
}
struct shash new_ports;
size_t i;
- assert(hmap_is_empty(&br->if_cfg_todo));
+ ovs_assert(hmap_is_empty(&br->if_cfg_todo));
/* Collect new ports. */
shash_init(&new_ports);
}
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));
+ br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
}
/* Configure the default gateway. */
&& gateway.s_addr) {
if (!netdev_add_router(netdev, gateway)) {
VLOG_INFO("bridge %s: configured gateway "IP_FMT,
- br->name, IP_ARGS(&gateway.s_addr));
+ br->name, IP_ARGS(gateway.s_addr));
}
}
}
/* Returns true if 'a' and 'b' are the same except that any number of slashes
* in either string are treated as equal to any number of slashes in the other,
- * e.g. "x///y" is equal to "x/y". */
+ * e.g. "x///y" is equal to "x/y".
+ *
+ * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
+ * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
+ * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
+ * 'b' against a prefix of 'a'.
+ */
static bool
-equal_pathnames(const char *a, const char *b)
+equal_pathnames(const char *a, const char *b, size_t b_stoplen)
{
- while (*a == *b) {
- if (*a == '/') {
+ const char *b_start = b;
+ for (;;) {
+ if (b - b_start >= b_stoplen) {
+ return true;
+ } else if (*a != *b) {
+ return false;
+ } else if (*a == '/') {
a += strspn(a, "/");
b += strspn(b, "/");
} else if (*a == '\0') {
b++;
}
}
- return false;
}
static void
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
char *whitelist;
- whitelist = xasprintf("unix:%s/%s.controller",
+ if (!strncmp(c->target, "unix:", 5)) {
+ /* Connect to a listening socket */
+ whitelist = xasprintf("unix:%s/", ovs_rundir());
+ if (strchr(c->target, '/') &&
+ !equal_pathnames(c->target, whitelist,
+ strlen(whitelist))) {
+ /* Absolute path specified, but not in ovs_rundir */
+ VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
+ "controller \"%s\" due to possibility for "
+ "remote exploit. Instead, specify socket "
+ "in whitelisted \"%s\" or connect to "
+ "\"unix:%s/%s.mgmt\" (which is always "
+ "available without special configuration).",
+ br->name, c->target, whitelist,
ovs_rundir(), br->name);
- if (!equal_pathnames(c->target, whitelist)) {
- /* Prevent remote ovsdb-server users from accessing arbitrary
- * Unix domain sockets and overwriting arbitrary local
- * files. */
- VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
- "controller \"%s\" due to possibility for remote "
- "exploit. Instead, specify whitelisted \"%s\" or "
- "connect to \"unix:%s/%s.mgmt\" (which is always "
- "available without special configuration).",
- br->name, c->target, whitelist,
- ovs_rundir(), br->name);
- free(whitelist);
- continue;
+ free(whitelist);
+ continue;
+ }
+ } else {
+ whitelist = xasprintf("punix:%s/%s.controller",
+ ovs_rundir(), br->name);
+ if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
+ /* Prevent remote ovsdb-server users from accessing
+ * arbitrary Unix domain sockets and overwriting arbitrary
+ * local files. */
+ VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
+ "controller \"%s\" due to possibility of "
+ "overwriting local files. Instead, specify "
+ "whitelisted \"%s\" or connect to "
+ "\"unix:%s/%s.mgmt\" (which is always "
+ "available without special configuration).",
+ br->name, c->target, whitelist,
+ ovs_rundir(), br->name);
+ free(whitelist);
+ continue;
+ }
}
free(whitelist);
br->cfg->key_flow_tables[j]);
}
}
+
+static void
+bridge_configure_dp_desc(struct bridge *br)
+{
+ ofproto_set_dp_desc(br->ofproto,
+ smap_get(&br->cfg->other_config, "dp-desc"));
+}
\f
/* Port functions. */
iface_clear_db_record(const struct ovsrec_interface *if_cfg)
{
if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
- iface_set_ofport(if_cfg, -1);
ovsrec_interface_set_status(if_cfg, NULL);
ovsrec_interface_set_admin_state(if_cfg, NULL);
ovsrec_interface_set_duplex(if_cfg, NULL);
ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
ovsrec_interface_set_link_state(if_cfg, NULL);
+ ovsrec_interface_set_mac_in_use(if_cfg, NULL);
ovsrec_interface_set_mtu(if_cfg, NULL, 0);
ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
}
+static int64_t
+iface_pick_ofport(const struct ovsrec_interface *cfg)
+{
+ int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
+ return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
+}
+
\f
/* Port mirroring. */