/*
- * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
+ * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 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 "lacp.h"
#include "learn.h"
#include "mac-learning.h"
+#include "mcast-snooping.h"
#include "meta-flow.h"
#include "multipath.h"
#include "netdev-vport.h"
#include "ofproto-dpif-upcall.h"
#include "ofproto-dpif-xlate.h"
#include "poll-loop.h"
+#include "ovs-rcu.h"
+#include "ovs-router.h"
#include "seq.h"
#include "simap.h"
#include "smap.h"
#include "unaligned.h"
#include "unixctl.h"
#include "vlan-bitmap.h"
-#include "vlog.h"
+#include "openvswitch/vlog.h"
VLOG_DEFINE_THIS_MODULE(ofproto_dpif);
char *name; /* Identifier for log messages. */
/* Configuration. */
- struct list ports; /* Contains "struct ofport"s. */
+ struct ovs_list ports; /* Contains "struct ofport"s. */
enum port_vlan_mode vlan_mode; /* VLAN mode */
int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
static void bundle_del_port(struct ofport_dpif *);
static void bundle_run(struct ofbundle *);
static void bundle_wait(struct ofbundle *);
+static void bundle_flush_macs(struct ofbundle *, bool);
+static void bundle_move(struct ofbundle *, struct ofbundle *);
static void stp_run(struct ofproto_dpif *ofproto);
static void stp_wait(struct ofproto_dpif *ofproto);
static int set_stp_port(struct ofport *,
const struct ofproto_port_stp_settings *);
+static void rstp_run(struct ofproto_dpif *ofproto);
+static void set_rstp_port(struct ofport *,
+ const struct ofproto_port_rstp_settings *);
+
struct ofport_dpif {
struct hmap_node odp_port_node; /* In dpif_backer's "odp_to_ofport_map". */
struct ofport up;
odp_port_t odp_port;
struct ofbundle *bundle; /* Bundle that contains this port, if any. */
- struct list bundle_node; /* In struct ofbundle's "ports" list. */
+ struct ovs_list bundle_node;/* In struct ofbundle's "ports" list. */
struct cfm *cfm; /* Connectivity Fault Management, if any. */
struct bfd *bfd; /* BFD, if any. */
bool may_enable; /* May be enabled in bonds. */
enum stp_state stp_state; /* Always STP_DISABLED if STP not in use. */
long long int stp_state_entered;
+ /* Rapid Spanning Tree. */
+ struct rstp_port *rstp_port; /* Rapid Spanning Tree Protocol, if any. */
+ enum rstp_state rstp_state; /* Always RSTP_DISABLED if RSTP not in use. */
+
/* Queue to DSCP mapping. */
struct ofproto_port_queue *qdscp;
size_t n_qdscp;
enum revalidate_reason {
REV_RECONFIGURE = 1, /* Switch configuration changed. */
REV_STP, /* Spanning tree protocol port status change. */
+ REV_RSTP, /* RSTP port status change. */
REV_BOND, /* Bonding changed. */
REV_PORT_TOGGLED, /* Port enabled or disabled by CFM, LACP, ...*/
REV_FLOW_TABLE, /* Flow table changed. */
REV_MAC_LEARNING, /* Mac learning changed. */
+ REV_MCAST_SNOOPING, /* Multicast snooping changed. */
};
COVERAGE_DEFINE(rev_reconfigure);
COVERAGE_DEFINE(rev_stp);
+COVERAGE_DEFINE(rev_rstp);
COVERAGE_DEFINE(rev_bond);
COVERAGE_DEFINE(rev_port_toggled);
COVERAGE_DEFINE(rev_flow_table);
COVERAGE_DEFINE(rev_mac_learning);
+COVERAGE_DEFINE(rev_mcast_snooping);
+
+/* Stores mapping between 'recirc_id' and 'ofproto-dpif'. */
+struct dpif_backer_recirc_node {
+ struct cmap_node cmap_node;
+ struct ofproto_dpif *ofproto;
+ uint32_t recirc_id;
+};
/* All datapaths of a given type share a single dpif backer instance. */
struct dpif_backer {
/* Recirculation. */
struct recirc_id_pool *rid_pool; /* Recirculation ID pool. */
+ struct cmap recirc_map; /* Map of 'recirc_id's to 'ofproto's. */
+ struct ovs_mutex recirc_mutex; /* Protects 'recirc_map'. */
bool enable_recirc; /* True if the datapath supports recirculation */
+ /* True if the datapath supports unique flow identifiers */
+ bool enable_ufid;
+
/* True if the datapath supports variable-length
* OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
* False if the datapath supports only 8-byte (or shorter) userdata. */
bool variable_length_userdata;
+ /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET
+ * actions. */
+ bool masked_set_action;
+
/* Maximum number of MPLS label stack entries that the datapath supports
* in a match */
size_t max_mpls_depth;
+
+ /* Version string of the datapath stored in OVSDB. */
+ char *dp_version_string;
+
+ /* True if the datapath supports tnl_push and pop actions. */
+ bool enable_tnl_push_pop;
+ struct atomic_count tnl_count;
};
/* All existing ofproto_backer instances, indexed by ofproto->up.type. */
struct dpif_ipfix *ipfix;
struct hmap bundles; /* Contains "struct ofbundle"s. */
struct mac_learning *ml;
+ struct mcast_snooping *ms;
bool has_bonded_bundles;
bool lacp_enabled;
struct mbridge *mbridge;
struct stp *stp;
long long int stp_last_tick;
+ /* Rapid Spanning Tree. */
+ struct rstp *rstp;
+ long long int rstp_last_tick;
+
/* VLAN splinters. */
struct ovs_mutex vsp_mutex;
struct hmap realdev_vid_map OVS_GUARDED; /* (realdev,vid) -> vlandev. */
/* All existing ofproto_dpif instances, indexed by ->up.name. */
static struct hmap all_ofproto_dpifs = HMAP_INITIALIZER(&all_ofproto_dpifs);
-static void ofproto_dpif_unixctl_init(void);
+static bool ofproto_use_tnl_push_pop = true;
+static void ofproto_unixctl_init(void);
static inline struct ofproto_dpif *
ofproto_dpif_cast(const struct ofproto *ofproto)
return ofproto->backer->enable_recirc;
}
+bool
+ofproto_dpif_get_enable_ufid(struct dpif_backer *backer)
+{
+ return backer->enable_ufid;
+}
+
static struct ofport_dpif *get_ofp_port(const struct ofproto_dpif *ofproto,
ofp_port_t ofp_port);
static void ofproto_trace(struct ofproto_dpif *, struct flow *,
return 0;
}
- dpif_run(backer->dpif);
+
+ if (dpif_run(backer->dpif)) {
+ backer->need_revalidate = REV_RECONFIGURE;
+ }
+
+ udpif_run(backer->udpif);
/* If vswitchd started with other_config:flow_restore_wait set as "true",
* and the configuration has now changed to "false", enable receiving
udpif_set_threads(backer->udpif, n_handlers, n_revalidators);
}
+ dpif_poll_threads_set(backer->dpif, n_dpdk_rxqs, pmd_cpu_mask);
+
if (backer->need_revalidate) {
struct ofproto_dpif *ofproto;
struct simap_node *node;
iter->odp_port = node ? u32_to_odp(node->data) : ODPP_NONE;
if (tnl_port_reconfigure(iter, iter->up.netdev,
- iter->odp_port)) {
+ iter->odp_port,
+ ovs_native_tunneling_is_on(ofproto), dp_port)) {
backer->need_revalidate = REV_RECONFIGURE;
}
}
switch (backer->need_revalidate) {
case REV_RECONFIGURE: COVERAGE_INC(rev_reconfigure); break;
case REV_STP: COVERAGE_INC(rev_stp); break;
+ case REV_RSTP: COVERAGE_INC(rev_rstp); break;
case REV_BOND: COVERAGE_INC(rev_bond); break;
case REV_PORT_TOGGLED: COVERAGE_INC(rev_port_toggled); break;
case REV_FLOW_TABLE: COVERAGE_INC(rev_flow_table); break;
case REV_MAC_LEARNING: COVERAGE_INC(rev_mac_learning); break;
+ case REV_MCAST_SNOOPING: COVERAGE_INC(rev_mcast_snooping); break;
}
backer->need_revalidate = 0;
xlate_txn_start();
xlate_ofproto_set(ofproto, ofproto->up.name,
- ofproto->backer->dpif, ofproto->miss_rule,
- ofproto->no_packet_in_rule, ofproto->ml,
- ofproto->stp, ofproto->mbridge,
- ofproto->sflow, ofproto->ipfix,
- ofproto->netflow, ofproto->up.frag_handling,
+ ofproto->backer->dpif, ofproto->ml,
+ ofproto->stp, ofproto->rstp, ofproto->ms,
+ ofproto->mbridge, ofproto->sflow, ofproto->ipfix,
+ ofproto->netflow,
ofproto->up.forward_bpdu,
connmgr_has_in_band(ofproto->up.connmgr),
ofproto->backer->enable_recirc,
ofproto->backer->variable_length_userdata,
- ofproto->backer->max_mpls_depth);
+ ofproto->backer->max_mpls_depth,
+ ofproto->backer->masked_set_action);
HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
xlate_bundle_set(ofproto, bundle, bundle->name,
ofport->up.ofp_port, ofport->odp_port,
ofport->up.netdev, ofport->cfm,
ofport->bfd, ofport->peer, stp_port,
- ofport->qdscp, ofport->n_qdscp,
- ofport->up.pp.config, ofport->up.pp.state,
- ofport->is_tunnel, ofport->may_enable);
+ ofport->rstp_port, ofport->qdscp,
+ ofport->n_qdscp, ofport->up.pp.config,
+ ofport->up.pp.state, ofport->is_tunnel,
+ ofport->may_enable);
}
xlate_txn_commit();
}
free(ofproto);
}
+/* Called when 'ofproto' is destructed. Checks for and clears any
+ * recirc_id leak. */
+static void
+dpif_backer_recirc_clear_ofproto(struct dpif_backer *backer,
+ struct ofproto_dpif *ofproto)
+{
+ struct dpif_backer_recirc_node *node;
+
+ ovs_mutex_lock(&backer->recirc_mutex);
+ CMAP_FOR_EACH (node, cmap_node, &backer->recirc_map) {
+ if (node->ofproto == ofproto) {
+ VLOG_ERR("recirc_id %"PRIu32", not freed when ofproto (%s) "
+ "is destructed", node->recirc_id, ofproto->up.name);
+ cmap_remove(&backer->recirc_map, &node->cmap_node,
+ node->recirc_id);
+ ovsrcu_postpone(free, node);
+ }
+ }
+ ovs_mutex_unlock(&backer->recirc_mutex);
+}
+
static void
close_dpif_backer(struct dpif_backer *backer)
{
hmap_destroy(&backer->odp_to_ofport_map);
shash_find_and_delete(&all_dpif_backers, backer->type);
recirc_id_pool_destroy(backer->rid_pool);
+ cmap_destroy(&backer->recirc_map);
+ ovs_mutex_destroy(&backer->recirc_mutex);
free(backer->type);
+ free(backer->dp_version_string);
dpif_close(backer->dpif);
free(backer);
}
/* Datapath port slated for removal from datapath. */
struct odp_garbage {
- struct list list_node;
+ struct ovs_list list_node;
odp_port_t odp_port;
};
static bool check_variable_length_userdata(struct dpif_backer *backer);
static size_t check_max_mpls_depth(struct dpif_backer *backer);
static bool check_recirc(struct dpif_backer *backer);
+static bool check_ufid(struct dpif_backer *backer);
+static bool check_masked_set_action(struct dpif_backer *backer);
static int
open_dpif_backer(const char *type, struct dpif_backer **backerp)
struct dpif_port_dump port_dump;
struct dpif_port port;
struct shash_node *node;
- struct list garbage_list;
+ struct ovs_list garbage_list;
struct odp_garbage *garbage, *next;
struct sset names;
shash_add(&all_dpif_backers, type, backer);
+ backer->enable_recirc = check_recirc(backer);
+ backer->max_mpls_depth = check_max_mpls_depth(backer);
+ backer->masked_set_action = check_masked_set_action(backer);
+ backer->enable_ufid = check_ufid(backer);
+ backer->rid_pool = recirc_id_pool_create();
+ ovs_mutex_init(&backer->recirc_mutex);
+ cmap_init(&backer->recirc_map);
+
+ backer->enable_tnl_push_pop = dpif_supports_tnl_push_pop(backer->dpif);
+ atomic_count_init(&backer->tnl_count, 0);
+
error = dpif_recv_set(backer->dpif, backer->recv_set_enable);
if (error) {
VLOG_ERR("failed to listen on datapath of type %s: %s",
close_dpif_backer(backer);
return error;
}
- backer->enable_recirc = check_recirc(backer);
- backer->variable_length_userdata = check_variable_length_userdata(backer);
- backer->max_mpls_depth = check_max_mpls_depth(backer);
- backer->rid_pool = recirc_id_pool_create();
if (backer->recv_set_enable) {
udpif_set_threads(backer->udpif, n_handlers, n_revalidators);
}
+ /* This check fails if performed before udpif threads have been set,
+ * as the kernel module checks that the 'pid' in userspace action
+ * is non-zero. */
+ backer->variable_length_userdata = check_variable_length_userdata(backer);
+ backer->dp_version_string = dpif_get_dp_version(backer->dpif);
+
return error;
}
+bool
+ovs_native_tunneling_is_on(struct ofproto_dpif *ofproto)
+{
+ return ofproto_use_tnl_push_pop && ofproto->backer->enable_tnl_push_pop &&
+ atomic_count_get(&ofproto->backer->tnl_count);
+}
+
/* Tests whether 'backer''s datapath supports recirculation. Only newer
* datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
* features on older datapaths that don't support this feature.
struct flow flow;
struct odputil_keybuf keybuf;
struct ofpbuf key;
- int error;
- bool enable_recirc = false;
+ bool enable_recirc;
memset(&flow, 0, sizeof flow);
flow.recirc_id = 1;
ofpbuf_use_stack(&key, &keybuf, sizeof keybuf);
odp_flow_key_from_flow(&key, &flow, NULL, 0, true);
+ enable_recirc = dpif_probe_feature(backer->dpif, "recirculation", &key,
+ NULL);
- error = dpif_flow_put(backer->dpif, DPIF_FP_CREATE | DPIF_FP_MODIFY,
- ofpbuf_data(&key), ofpbuf_size(&key), NULL, 0, NULL,
- 0, NULL);
- if (error && error != EEXIST) {
- if (error != EINVAL) {
- VLOG_WARN("%s: Reciculation flow probe failed (%s)",
- dpif_name(backer->dpif), ovs_strerror(error));
- }
- goto done;
- }
-
- error = dpif_flow_del(backer->dpif, ofpbuf_data(&key), ofpbuf_size(&key),
- NULL);
- if (error) {
- VLOG_WARN("%s: failed to delete recirculation feature probe flow",
- dpif_name(backer->dpif));
- }
-
- enable_recirc = true;
-
-done:
if (enable_recirc) {
VLOG_INFO("%s: Datapath supports recirculation",
dpif_name(backer->dpif));
return enable_recirc;
}
+/* Tests whether 'dpif' supports userspace flow ids. We can skip serializing
+ * some flow attributes for datapaths that support this feature.
+ *
+ * Returns true if 'dpif' supports UFID for flow operations.
+ * Returns false if 'dpif' does not support UFID. */
+static bool
+check_ufid(struct dpif_backer *backer)
+{
+ struct flow flow;
+ struct odputil_keybuf keybuf;
+ struct ofpbuf key;
+ ovs_u128 ufid;
+ bool enable_ufid;
+
+ memset(&flow, 0, sizeof flow);
+ flow.dl_type = htons(0x1234);
+
+ ofpbuf_use_stack(&key, &keybuf, sizeof keybuf);
+ odp_flow_key_from_flow(&key, &flow, NULL, 0, true);
+ dpif_flow_hash(backer->dpif, ofpbuf_data(&key), ofpbuf_size(&key), &ufid);
+
+ enable_ufid = dpif_probe_feature(backer->dpif, "UFID", &key, &ufid);
+
+ if (enable_ufid) {
+ VLOG_INFO("%s: Datapath supports userspace flow ids",
+ dpif_name(backer->dpif));
+ } else {
+ VLOG_INFO("%s: Datapath does not support userspace flow ids",
+ dpif_name(backer->dpif));
+ }
+ return enable_ufid;
+}
+
/* Tests whether 'backer''s datapath supports variable-length
* OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
* to disable some features on older datapaths that don't support this
execute.packet = &packet;
execute.md = PKT_METADATA_INITIALIZER(0);
execute.needs_help = false;
+ execute.probe = true;
error = dpif_execute(backer->dpif, &execute);
switch (error) {
case 0:
- /* Variable-length userdata is supported.
- *
- * Purge received packets to avoid processing the nonsense packet we
- * sent to userspace, then report success. */
- dpif_recv_purge(backer->dpif);
return true;
case ERANGE:
for (n = 0; n < FLOW_MAX_MPLS_LABELS; n++) {
struct odputil_keybuf keybuf;
struct ofpbuf key;
- int error;
memset(&flow, 0, sizeof flow);
flow.dl_type = htons(ETH_TYPE_MPLS);
ofpbuf_use_stack(&key, &keybuf, sizeof keybuf);
odp_flow_key_from_flow(&key, &flow, NULL, 0, false);
-
- error = dpif_flow_put(backer->dpif, DPIF_FP_CREATE | DPIF_FP_MODIFY,
- ofpbuf_data(&key), ofpbuf_size(&key), NULL, 0, NULL, 0, NULL);
- if (error && error != EEXIST) {
- if (error != EINVAL) {
- VLOG_WARN("%s: MPLS stack length feature probe failed (%s)",
- dpif_name(backer->dpif), ovs_strerror(error));
- }
+ if (!dpif_probe_feature(backer->dpif, "MPLS", &key, NULL)) {
break;
}
-
- error = dpif_flow_del(backer->dpif, ofpbuf_data(&key), ofpbuf_size(&key), NULL);
- if (error) {
- VLOG_WARN("%s: failed to delete MPLS feature probe flow",
- dpif_name(backer->dpif));
- }
}
VLOG_INFO("%s: MPLS label stack length probed as %d",
return n;
}
+/* Tests whether 'backer''s datapath supports masked data in
+ * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
+ * datapaths that don't support this feature. */
+static bool
+check_masked_set_action(struct dpif_backer *backer)
+{
+ struct eth_header *eth;
+ struct ofpbuf actions;
+ struct dpif_execute execute;
+ struct ofpbuf packet;
+ int error;
+ struct ovs_key_ethernet key, mask;
+
+ /* Compose a set action that will cause an EINVAL error on older
+ * datapaths that don't support masked set actions.
+ * Avoid using a full mask, as it could be translated to a non-masked
+ * set action instead. */
+ ofpbuf_init(&actions, 64);
+ memset(&key, 0x53, sizeof key);
+ memset(&mask, 0x7f, sizeof mask);
+ commit_masked_set_action(&actions, OVS_KEY_ATTR_ETHERNET, &key, &mask,
+ sizeof key);
+
+ /* Compose a dummy ethernet packet. */
+ ofpbuf_init(&packet, ETH_HEADER_LEN);
+ eth = ofpbuf_put_zeros(&packet, ETH_HEADER_LEN);
+ eth->eth_type = htons(0x1234);
+
+ /* Execute the actions. On older datapaths this fails with EINVAL, on
+ * newer datapaths it succeeds. */
+ execute.actions = ofpbuf_data(&actions);
+ execute.actions_len = ofpbuf_size(&actions);
+ execute.packet = &packet;
+ execute.md = PKT_METADATA_INITIALIZER(0);
+ execute.needs_help = false;
+ execute.probe = true;
+
+ error = dpif_execute(backer->dpif, &execute);
+
+ ofpbuf_uninit(&packet);
+ ofpbuf_uninit(&actions);
+
+ if (error) {
+ /* Masked set action is not supported. */
+ VLOG_INFO("%s: datapath does not support masked set action feature.",
+ dpif_name(backer->dpif));
+ }
+ return !error;
+}
+
static int
construct(struct ofproto *ofproto_)
{
struct shash_node *node, *next;
int error;
+ /* Tunnel module can get used right after the udpif threads are running. */
+ ofproto_tunnel_init();
+
error = open_dpif_backer(ofproto->up.type, &ofproto->backer);
if (error) {
return error;
ofproto->sflow = NULL;
ofproto->ipfix = NULL;
ofproto->stp = NULL;
+ ofproto->rstp = NULL;
ofproto->dump_seq = 0;
hmap_init(&ofproto->bundles);
ofproto->ml = mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME);
+ ofproto->ms = NULL;
ofproto->mbridge = mbridge_create();
ofproto->has_bonded_bundles = false;
ofproto->lacp_enabled = false;
guarded_list_init(&ofproto->pins);
- ofproto_dpif_unixctl_init();
+ ofproto_unixctl_init();
hmap_init(&ofproto->vlandev_map);
hmap_init(&ofproto->realdev_vid_map);
return error;
}
- /* Continue non-recirculation rule lookups from table 0.
+ /* Drop any run away non-recirc rule lookups. Recirc_id has to be
+ * zero when reaching this rule.
*
- * (priority=2), recirc=0, actions=resubmit(, 0)
+ * (priority=2), recirc_id=0, actions=drop
*/
- resubmit = ofpact_put_RESUBMIT(&ofpacts);
- resubmit->ofpact.compat = 0;
- resubmit->in_port = OFPP_IN_PORT;
- resubmit->table_id = 0;
-
+ ofpbuf_clear(&ofpacts);
match_init_catchall(&match);
match_set_recirc_id(&match, 0);
-
error = ofproto_dpif_add_internal_flow(ofproto, &match, 2, 0, &ofpacts,
&unused_rulep);
if (error) {
return error;
}
- /* Drop any run away recirc rule lookups. Recirc_id has to be
- * non-zero when reaching this rule.
+ /* Continue rule lookups for not-matched recirc rules from table 0.
*
- * (priority=1), *, actions=drop
+ * (priority=1), actions=resubmit(, 0)
*/
- ofpbuf_clear(&ofpacts);
+ resubmit = ofpact_put_RESUBMIT(&ofpacts);
+ resubmit->in_port = OFPP_IN_PORT;
+ resubmit->table_id = 0;
+
match_init_catchall(&match);
error = ofproto_dpif_add_internal_flow(ofproto, &match, 1, 0, &ofpacts,
&unused_rulep);
destruct(struct ofproto *ofproto_)
{
struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
- struct rule_dpif *rule, *next_rule;
struct ofproto_packet_in *pin, *next_pin;
+ struct rule_dpif *rule;
struct oftable *table;
- struct list pins;
+ struct ovs_list pins;
ofproto->backer->need_revalidate = REV_RECONFIGURE;
xlate_txn_start();
hmap_remove(&all_ofproto_dpifs, &ofproto->all_ofproto_dpifs_node);
OFPROTO_FOR_EACH_TABLE (table, &ofproto->up) {
- struct cls_cursor cursor;
-
- fat_rwlock_rdlock(&table->cls.rwlock);
- cls_cursor_init(&cursor, &table->cls, NULL);
- fat_rwlock_unlock(&table->cls.rwlock);
- CLS_CURSOR_FOR_EACH_SAFE (rule, next_rule, up.cr, &cursor) {
+ CLS_FOR_EACH (rule, up.cr, &table->cls) {
ofproto_rule_delete(&ofproto->up, &rule->up);
}
}
}
guarded_list_destroy(&ofproto->pins);
+ dpif_backer_recirc_clear_ofproto(ofproto->backer, ofproto);
+
mbridge_unref(ofproto->mbridge);
netflow_unref(ofproto->netflow);
dpif_sflow_unref(ofproto->sflow);
+ dpif_ipfix_unref(ofproto->ipfix);
hmap_destroy(&ofproto->bundles);
mac_learning_unref(ofproto->ml);
+ mcast_snooping_unref(ofproto->ms);
hmap_destroy(&ofproto->vlandev_map);
hmap_destroy(&ofproto->realdev_vid_map);
ovs_rwlock_wrlock(&ofproto->ml->rwlock);
mac_learning_flush(ofproto->ml);
ovs_rwlock_unlock(&ofproto->ml->rwlock);
+ mcast_snooping_mdb_flush(ofproto->ms);
}
/* Always updates the ofproto->pins_seqno to avoid frequent wakeup during
* waiting for flow restore to complete. */
if (!ofproto_get_flow_restore_wait()) {
struct ofproto_packet_in *pin, *next_pin;
- struct list pins;
+ struct ovs_list pins;
guarded_list_pop_all(&ofproto->pins, &pins);
LIST_FOR_EACH_SAFE (pin, next_pin, list_node, &pins) {
}
stp_run(ofproto);
+ rstp_run(ofproto);
ovs_rwlock_wrlock(&ofproto->ml->rwlock);
if (mac_learning_run(ofproto->ml)) {
ofproto->backer->need_revalidate = REV_MAC_LEARNING;
}
ovs_rwlock_unlock(&ofproto->ml->rwlock);
+ if (mcast_snooping_run(ofproto->ms)) {
+ ofproto->backer->need_revalidate = REV_MCAST_SNOOPING;
+ }
+
new_dump_seq = seq_read(udpif_dump_seq(ofproto->backer->udpif));
if (ofproto->dump_seq != new_dump_seq) {
struct rule *rule, *next_rule;
}
}
}
-
return 0;
}
ovs_rwlock_rdlock(&ofproto->ml->rwlock);
mac_learning_wait(ofproto->ml);
ovs_rwlock_unlock(&ofproto->ml->rwlock);
+ mcast_snooping_wait(ofproto->ms);
stp_wait(ofproto);
if (ofproto->backer->need_revalidate) {
/* Shouldn't happen, but if it does just go around again. */
}
static void
-get_features(struct ofproto *ofproto_ OVS_UNUSED,
- bool *arp_match_ip, enum ofputil_action_bitmap *actions)
-{
- *arp_match_ip = true;
- *actions = (OFPUTIL_A_OUTPUT |
- OFPUTIL_A_SET_VLAN_VID |
- OFPUTIL_A_SET_VLAN_PCP |
- OFPUTIL_A_STRIP_VLAN |
- OFPUTIL_A_SET_DL_SRC |
- OFPUTIL_A_SET_DL_DST |
- OFPUTIL_A_SET_NW_SRC |
- OFPUTIL_A_SET_NW_DST |
- OFPUTIL_A_SET_NW_TOS |
- OFPUTIL_A_SET_TP_SRC |
- OFPUTIL_A_SET_TP_DST |
- OFPUTIL_A_ENQUEUE);
-}
-
-static void
-get_tables(struct ofproto *ofproto, struct ofp12_table_stats *ots)
+query_tables(struct ofproto *ofproto,
+ struct ofputil_table_features *features,
+ struct ofputil_table_stats *stats)
{
- int i;
+ strcpy(features->name, "classifier");
- strcpy(ots->name, "classifier");
+ if (stats) {
+ int i;
- for (i = 0; i < ofproto->n_tables; i++) {
- unsigned long missed, matched;
+ for (i = 0; i < ofproto->n_tables; i++) {
+ unsigned long missed, matched;
- atomic_read(&ofproto->tables[i].n_matched, &matched);
- ots[i].matched_count = htonll(matched);
- atomic_read(&ofproto->tables[i].n_missed, &missed);
- ots[i].lookup_count = htonll(matched + missed);
+ atomic_read_relaxed(&ofproto->tables[i].n_matched, &matched);
+ atomic_read_relaxed(&ofproto->tables[i].n_missed, &missed);
+
+ stats[i].matched_count = matched;
+ stats[i].lookup_count = matched + missed;
+ }
}
}
port->bundle = NULL;
port->cfm = NULL;
port->bfd = NULL;
- port->may_enable = true;
+ port->may_enable = false;
port->stp_port = NULL;
port->stp_state = STP_DISABLED;
+ port->rstp_port = NULL;
+ port->rstp_state = RSTP_DISABLED;
port->is_tunnel = false;
port->peer = NULL;
port->qdscp = NULL;
port->odp_port = dpif_port.port_no;
if (netdev_get_tunnel_config(netdev)) {
- tnl_port_add(port, port->up.netdev, port->odp_port);
+ atomic_count_inc(&ofproto->backer->tnl_count);
+ tnl_port_add(port, port->up.netdev, port->odp_port,
+ ovs_native_tunneling_is_on(ofproto), namebuf);
port->is_tunnel = true;
+ if (ofproto->ipfix) {
+ dpif_ipfix_add_tunnel_port(ofproto->ipfix, port_, port->odp_port);
+ }
} else {
/* Sanity-check that a mapping doesn't already exist. This
* shouldn't happen for non-tunnel ports. */
ovs_rwlock_unlock(&ofproto->backer->odp_to_ofport_lock);
}
+ if (port->is_tunnel) {
+ atomic_count_dec(&ofproto->backer->tnl_count);
+ }
+
+ if (port->is_tunnel && ofproto->ipfix) {
+ dpif_ipfix_del_tunnel_port(ofproto->ipfix, port->odp_port);
+ }
+
tnl_port_del(port);
sset_find_and_delete(&ofproto->ports, devname);
sset_find_and_delete(&ofproto->ghost_ports, devname);
if (port->stp_port) {
stp_port_disable(port->stp_port);
}
+ set_rstp_port(port_, NULL);
if (ofproto->sflow) {
dpif_sflow_del_port(ofproto->sflow, port->odp_port);
}
port_modified(struct ofport *port_)
{
struct ofport_dpif *port = ofport_dpif_cast(port_);
+ char namebuf[NETDEV_VPORT_NAME_BUFSIZE];
+ struct netdev *netdev = port->up.netdev;
if (port->bundle && port->bundle->bond) {
- bond_slave_set_netdev(port->bundle->bond, port, port->up.netdev);
+ bond_slave_set_netdev(port->bundle->bond, port, netdev);
}
if (port->cfm) {
- cfm_set_netdev(port->cfm, port->up.netdev);
+ cfm_set_netdev(port->cfm, netdev);
}
if (port->bfd) {
- bfd_set_netdev(port->bfd, port->up.netdev);
+ bfd_set_netdev(port->bfd, netdev);
}
ofproto_dpif_monitor_port_update(port, port->bfd, port->cfm,
port->up.pp.hw_addr);
- if (port->is_tunnel && tnl_port_reconfigure(port, port->up.netdev,
- port->odp_port)) {
- ofproto_dpif_cast(port->up.ofproto)->backer->need_revalidate =
- REV_RECONFIGURE;
+ netdev_vport_get_dpif_port(netdev, namebuf, sizeof namebuf);
+
+ if (port->is_tunnel) {
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(port->up.ofproto);
+
+ if (tnl_port_reconfigure(port, netdev, port->odp_port,
+ ovs_native_tunneling_is_on(ofproto), namebuf)) {
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ }
}
ofport_update_peer(port);
struct dpif_sflow *ds = ofproto->sflow;
if (sflow_options) {
+ uint32_t old_probability = ds ? dpif_sflow_get_probability(ds) : 0;
if (!ds) {
struct ofport_dpif *ofport;
HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
dpif_sflow_add_port(ds, &ofport->up, ofport->odp_port);
}
- ofproto->backer->need_revalidate = REV_RECONFIGURE;
}
dpif_sflow_set_options(ds, sflow_options);
+ if (dpif_sflow_get_probability(ds) != old_probability) {
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ }
} else {
if (ds) {
dpif_sflow_unref(ds);
struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
struct dpif_ipfix *di = ofproto->ipfix;
bool has_options = bridge_exporter_options || flow_exporters_options;
+ bool new_di = false;
if (has_options && !di) {
di = ofproto->ipfix = dpif_ipfix_create();
+ new_di = true;
}
if (di) {
di, bridge_exporter_options, flow_exporters_options,
n_flow_exporters_options);
+ /* Add tunnel ports only when a new ipfix created */
+ if (new_di == true) {
+ struct ofport_dpif *ofport;
+ HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
+ if (ofport->is_tunnel == true) {
+ dpif_ipfix_add_tunnel_port(di, &ofport->up, ofport->odp_port);
+ }
+ }
+ }
+
if (!has_options) {
dpif_ipfix_unref(di);
ofproto->ipfix = NULL;
\f
/* Spanning Tree. */
+/* Called while rstp_mutex is held. */
+static void
+rstp_send_bpdu_cb(struct ofpbuf *pkt, void *ofport_, void *ofproto_)
+{
+ struct ofproto_dpif *ofproto = ofproto_;
+ struct ofport_dpif *ofport = ofport_;
+ struct eth_header *eth = ofpbuf_l2(pkt);
+
+ netdev_get_etheraddr(ofport->up.netdev, eth->eth_src);
+ if (eth_addr_is_zero(eth->eth_src)) {
+ VLOG_WARN_RL(&rl, "%s port %d: cannot send RSTP BPDU on a port which "
+ "does not have a configured source MAC address.",
+ ofproto->up.name, ofp_to_u16(ofport->up.ofp_port));
+ } else {
+ ofproto_dpif_send_packet(ofport, pkt);
+ }
+ ofpbuf_delete(pkt);
+}
+
static void
send_bpdu_cb(struct ofpbuf *pkt, int port_num, void *ofproto_)
{
ofpbuf_delete(pkt);
}
+/* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
+static void
+set_rstp(struct ofproto *ofproto_, const struct ofproto_rstp_settings *s)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ /* Only revalidate flows if the configuration changed. */
+ if (!s != !ofproto->rstp) {
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ }
+
+ if (s) {
+ if (!ofproto->rstp) {
+ ofproto->rstp = rstp_create(ofproto_->name, s->address,
+ rstp_send_bpdu_cb, ofproto);
+ ofproto->rstp_last_tick = time_msec();
+ }
+ rstp_set_bridge_address(ofproto->rstp, s->address);
+ rstp_set_bridge_priority(ofproto->rstp, s->priority);
+ rstp_set_bridge_ageing_time(ofproto->rstp, s->ageing_time);
+ rstp_set_bridge_force_protocol_version(ofproto->rstp,
+ s->force_protocol_version);
+ rstp_set_bridge_max_age(ofproto->rstp, s->bridge_max_age);
+ rstp_set_bridge_forward_delay(ofproto->rstp, s->bridge_forward_delay);
+ rstp_set_bridge_transmit_hold_count(ofproto->rstp,
+ s->transmit_hold_count);
+ } else {
+ struct ofport *ofport;
+ HMAP_FOR_EACH (ofport, hmap_node, &ofproto->up.ports) {
+ set_rstp_port(ofport, NULL);
+ }
+ rstp_unref(ofproto->rstp);
+ ofproto->rstp = NULL;
+ }
+}
+
+static void
+get_rstp_status(struct ofproto *ofproto_, struct ofproto_rstp_status *s)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ if (ofproto->rstp) {
+ s->enabled = true;
+ s->root_id = rstp_get_root_id(ofproto->rstp);
+ s->bridge_id = rstp_get_bridge_id(ofproto->rstp);
+ s->designated_id = rstp_get_designated_id(ofproto->rstp);
+ s->root_path_cost = rstp_get_root_path_cost(ofproto->rstp);
+ s->designated_port_id = rstp_get_designated_port_id(ofproto->rstp);
+ s->bridge_port_id = rstp_get_bridge_port_id(ofproto->rstp);
+ } else {
+ s->enabled = false;
+ }
+}
+
+static void
+update_rstp_port_state(struct ofport_dpif *ofport)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
+ enum rstp_state state;
+
+ /* Figure out new state. */
+ state = ofport->rstp_port ? rstp_port_get_state(ofport->rstp_port)
+ : RSTP_DISABLED;
+
+ /* Update state. */
+ if (ofport->rstp_state != state) {
+ enum ofputil_port_state of_state;
+ bool fwd_change;
+
+ VLOG_DBG("port %s: RSTP state changed from %s to %s",
+ netdev_get_name(ofport->up.netdev),
+ rstp_state_name(ofport->rstp_state),
+ rstp_state_name(state));
+
+ if (rstp_learn_in_state(ofport->rstp_state)
+ != rstp_learn_in_state(state)) {
+ /* XXX: Learning action flows should also be flushed. */
+ if (ofport->bundle) {
+ if (!rstp_shift_root_learned_address(ofproto->rstp)
+ || rstp_get_old_root_aux(ofproto->rstp) != ofport) {
+ bundle_flush_macs(ofport->bundle, false);
+ }
+ }
+ }
+ fwd_change = rstp_forward_in_state(ofport->rstp_state)
+ != rstp_forward_in_state(state);
+
+ ofproto->backer->need_revalidate = REV_RSTP;
+ ofport->rstp_state = state;
+
+ if (fwd_change && ofport->bundle) {
+ bundle_update(ofport->bundle);
+ }
+
+ /* Update the RSTP state bits in the OpenFlow port description. */
+ of_state = ofport->up.pp.state & ~OFPUTIL_PS_STP_MASK;
+ of_state |= (state == RSTP_LEARNING ? OFPUTIL_PS_STP_LEARN
+ : state == RSTP_FORWARDING ? OFPUTIL_PS_STP_FORWARD
+ : state == RSTP_DISCARDING ? OFPUTIL_PS_STP_LISTEN
+ : 0);
+ ofproto_port_set_state(&ofport->up, of_state);
+ }
+}
+
+static void
+rstp_run(struct ofproto_dpif *ofproto)
+{
+ if (ofproto->rstp) {
+ long long int now = time_msec();
+ long long int elapsed = now - ofproto->rstp_last_tick;
+ struct rstp_port *rp;
+ struct ofport_dpif *ofport;
+
+ /* Every second, decrease the values of the timers. */
+ if (elapsed >= 1000) {
+ rstp_tick_timers(ofproto->rstp);
+ ofproto->rstp_last_tick = now;
+ }
+ rp = NULL;
+ while ((ofport = rstp_get_next_changed_port_aux(ofproto->rstp, &rp))) {
+ update_rstp_port_state(ofport);
+ }
+ rp = NULL;
+ ofport = NULL;
+ /* FIXME: This check should be done on-event (i.e., when setting
+ * p->fdb_flush) and not periodically.
+ */
+ while ((ofport = rstp_check_and_reset_fdb_flush(ofproto->rstp, &rp))) {
+ if (!rstp_shift_root_learned_address(ofproto->rstp)
+ || rstp_get_old_root_aux(ofproto->rstp) != ofport) {
+ bundle_flush_macs(ofport->bundle, false);
+ }
+ }
+
+ if (rstp_shift_root_learned_address(ofproto->rstp)) {
+ bundle_move(((struct ofport_dpif *)rstp_get_old_root_aux(ofproto->rstp))->bundle,
+ ((struct ofport_dpif *)rstp_get_new_root_aux(ofproto->rstp))->bundle);
+ rstp_reset_root_changed(ofproto->rstp);
+ }
+ }
+}
+
/* Configures STP on 'ofproto_' using the settings defined in 's'. */
static int
set_stp(struct ofproto *ofproto_, const struct ofproto_stp_settings *s)
enum ofputil_port_state of_state;
bool fwd_change;
- VLOG_DBG_RL(&rl, "port %s: STP state changed from %s to %s",
- netdev_get_name(ofport->up.netdev),
- stp_state_name(ofport->stp_state),
- stp_state_name(state));
+ VLOG_DBG("port %s: STP state changed from %s to %s",
+ netdev_get_name(ofport->up.netdev),
+ stp_state_name(ofport->stp_state),
+ stp_state_name(state));
if (stp_learn_in_state(ofport->stp_state)
!= stp_learn_in_state(state)) {
/* xxx Learning action flows should also be flushed. */
ovs_rwlock_wrlock(&ofproto->ml->rwlock);
mac_learning_flush(ofproto->ml);
ovs_rwlock_unlock(&ofproto->ml->rwlock);
+ mcast_snooping_mdb_flush(ofproto->ms);
}
fwd_change = stp_forward_in_state(ofport->stp_state)
!= stp_forward_in_state(state);
}
return 0;
} else if (sp && stp_port_no(sp) != s->port_num
- && ofport == stp_port_get_aux(sp)) {
+ && ofport == stp_port_get_aux(sp)) {
/* The port-id changed, so disable the old one if it's not
* already in use by another port. */
stp_port_disable(sp);
}
sp = ofport->stp_port = stp_get_port(ofproto->stp, s->port_num);
+
+ /* Set name before enabling the port so that debugging messages can print
+ * the name. */
+ stp_port_set_name(sp, netdev_get_name(ofport->up.netdev));
stp_port_enable(sp);
stp_port_set_aux(sp, ofport);
ovs_rwlock_wrlock(&ofproto->ml->rwlock);
mac_learning_flush(ofproto->ml);
ovs_rwlock_unlock(&ofproto->ml->rwlock);
+ mcast_snooping_mdb_flush(ofproto->ms);
}
}
}
poll_timer_wait(1000);
}
}
+
+/* Configures RSTP on 'ofport_' using the settings defined in 's'. The
+ * caller is responsible for assigning RSTP port numbers and ensuring
+ * there are no duplicates. */
+static void
+set_rstp_port(struct ofport *ofport_,
+ const struct ofproto_port_rstp_settings *s)
+{
+ struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
+ struct rstp_port *rp = ofport->rstp_port;
+
+ if (!s || !s->enable) {
+ if (rp) {
+ rstp_port_unref(rp);
+ ofport->rstp_port = NULL;
+ update_rstp_port_state(ofport);
+ }
+ return;
+ }
+
+ /* Check if need to add a new port. */
+ if (!rp) {
+ rp = ofport->rstp_port = rstp_add_port(ofproto->rstp);
+ }
+
+ rstp_port_set(rp, s->port_num, s->priority, s->path_cost,
+ s->admin_edge_port, s->auto_edge,
+ s->admin_p2p_mac_state, s->admin_port_state, s->mcheck,
+ ofport);
+ update_rstp_port_state(ofport);
+ /* Synchronize operational status. */
+ rstp_port_set_mac_operational(rp, ofport->may_enable);
+}
+
+static void
+get_rstp_port_status(struct ofport *ofport_,
+ struct ofproto_port_rstp_status *s)
+{
+ struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
+ struct rstp_port *rp = ofport->rstp_port;
+
+ if (!ofproto->rstp || !rp) {
+ s->enabled = false;
+ return;
+ }
+
+ s->enabled = true;
+ rstp_port_get_status(rp, &s->port_id, &s->state, &s->role,
+ &s->designated_bridge_id, &s->designated_port_id,
+ &s->designated_path_cost, &s->tx_count,
+ &s->rx_count, &s->error_count, &s->uptime);
+}
+
\f
static int
set_queues(struct ofport *ofport_, const struct ofproto_port_queue *qdscp,
ofproto->backer->need_revalidate = REV_RECONFIGURE;
ovs_rwlock_wrlock(&ml->rwlock);
LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
- if (mac->port.p == bundle) {
+ if (mac_entry_get_port(ml, mac) == bundle) {
if (all_ofprotos) {
struct ofproto_dpif *o;
ovs_rwlock_unlock(&ml->rwlock);
}
+static void
+bundle_move(struct ofbundle *old, struct ofbundle *new)
+{
+ struct ofproto_dpif *ofproto = old->ofproto;
+ struct mac_learning *ml = ofproto->ml;
+ struct mac_entry *mac, *next_mac;
+
+ ovs_assert(new->ofproto == old->ofproto);
+
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ ovs_rwlock_wrlock(&ml->rwlock);
+ LIST_FOR_EACH_SAFE (mac, next_mac, lru_node, &ml->lrus) {
+ if (mac_entry_get_port(ml, mac) == old) {
+ mac_entry_set_port(ml, mac, new);
+ }
+ }
+ ovs_rwlock_unlock(&ml->rwlock);
+}
+
static struct ofbundle *
bundle_lookup(const struct ofproto_dpif *ofproto, void *aux)
{
LIST_FOR_EACH (port, bundle_node, &bundle->ports) {
if (port->up.pp.config & OFPUTIL_PC_NO_FLOOD
|| port->is_layer3
- || !stp_forward_in_state(port->stp_state)) {
+ || (bundle->ofproto->stp && !stp_forward_in_state(port->stp_state))
+ || (bundle->ofproto->rstp && !rstp_forward_in_state(port->rstp_state))) {
bundle->floodable = false;
break;
}
list_push_back(&bundle->ports, &port->bundle_node);
if (port->up.pp.config & OFPUTIL_PC_NO_FLOOD
|| port->is_layer3
- || !stp_forward_in_state(port->stp_state)) {
+ || (bundle->ofproto->stp && !stp_forward_in_state(port->stp_state))
+ || (bundle->ofproto->rstp && !rstp_forward_in_state(port->rstp_state))) {
bundle->floodable = false;
}
}
}
ofproto = bundle->ofproto;
- mbridge_unregister_bundle(ofproto->mbridge, bundle->aux);
+ mbridge_unregister_bundle(ofproto->mbridge, bundle);
xlate_txn_start();
xlate_bundle_remove(bundle);
struct ofpbuf *learning_packet;
int error, n_packets, n_errors;
struct mac_entry *e;
- struct list packets;
+ struct ovs_list packets;
list_init(&packets);
ovs_rwlock_rdlock(&ofproto->ml->rwlock);
LIST_FOR_EACH (e, lru_node, &ofproto->ml->lrus) {
- if (e->port.p != bundle) {
+ if (mac_entry_get_port(ofproto->ml, e) != bundle) {
void *port_void;
learning_packet = bond_compose_learning_packet(bundle->bond,
mac_learning_set_max_entries(ofproto->ml, max_entries);
ovs_rwlock_unlock(&ofproto->ml->rwlock);
}
+
+/* Configures multicast snooping on 'ofport' using the settings
+ * defined in 's'. */
+static int
+set_mcast_snooping(struct ofproto *ofproto_,
+ const struct ofproto_mcast_snooping_settings *s)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ /* Only revalidate flows if the configuration changed. */
+ if (!s != !ofproto->ms) {
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ }
+
+ if (s) {
+ if (!ofproto->ms) {
+ ofproto->ms = mcast_snooping_create();
+ }
+
+ ovs_rwlock_wrlock(&ofproto->ms->rwlock);
+ mcast_snooping_set_idle_time(ofproto->ms, s->idle_time);
+ mcast_snooping_set_max_entries(ofproto->ms, s->max_entries);
+ if (mcast_snooping_set_flood_unreg(ofproto->ms, s->flood_unreg)) {
+ ofproto->backer->need_revalidate = REV_RECONFIGURE;
+ }
+ ovs_rwlock_unlock(&ofproto->ms->rwlock);
+ } else {
+ mcast_snooping_unref(ofproto->ms);
+ ofproto->ms = NULL;
+ }
+
+ return 0;
+}
+
+/* Configures multicast snooping port's flood settings on 'ofproto'. */
+static int
+set_mcast_snooping_port(struct ofproto *ofproto_, void *aux,
+ const struct ofproto_mcast_snooping_port_settings *s)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+ struct ofbundle *bundle = bundle_lookup(ofproto, aux);
+
+ if (ofproto->ms && s) {
+ ovs_rwlock_wrlock(&ofproto->ms->rwlock);
+ mcast_snooping_set_port_flood(ofproto->ms, bundle, s->flood);
+ mcast_snooping_set_port_flood_reports(ofproto->ms, bundle,
+ s->flood_reports);
+ ovs_rwlock_unlock(&ofproto->ms->rwlock);
+ }
+ return 0;
+}
+
\f
/* Ports. */
if (ofport->may_enable != enable) {
struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofport->up.ofproto);
+
ofproto->backer->need_revalidate = REV_PORT_TOGGLED;
+
+ if (ofport->rstp_port) {
+ rstp_port_set_mac_operational(ofport->rstp_port, enable);
+ }
}
ofport->may_enable = enable;
return error;
}
+static int
+port_get_lacp_stats(const struct ofport *ofport_, struct lacp_slave_stats *stats)
+{
+ struct ofport_dpif *ofport = ofport_dpif_cast(ofport_);
+ if (ofport->bundle && ofport->bundle->lacp) {
+ if (lacp_get_slave_stats(ofport->bundle->lacp, ofport, stats)) {
+ return 0;
+ }
+ }
+ return -1;
+}
+
struct port_dump_state {
uint32_t bucket;
uint32_t offset;
rule_dpif_credit_stats(rule, &stats);
}
- xlate_in_init(&xin, ofproto, flow, rule, stats.tcp_flags, packet);
+ xlate_in_init(&xin, ofproto, flow, flow->in_port.ofp_port, rule,
+ stats.tcp_flags, packet);
xin.ofpacts = ofpacts;
xin.ofpacts_len = ofpacts_len;
xin.resubmit_stats = &stats;
xlate_actions(&xin, &xout);
- execute.actions = ofpbuf_data(&xout.odp_actions);
- execute.actions_len = ofpbuf_size(&xout.odp_actions);
+ execute.actions = ofpbuf_data(xout.odp_actions);
+ execute.actions_len = ofpbuf_size(xout.odp_actions);
+
execute.packet = packet;
execute.md = pkt_metadata_from_flow(flow);
execute.needs_help = (xout.slow & SLOW_ACTION) != 0;
+ execute.probe = false;
/* Fix up in_port. */
in_port = flow->in_port.ofp_port;
ofproto_rule_reduce_timeouts(&rule->up, idle_timeout, hard_timeout);
}
-/* Returns 'rule''s actions. The caller owns a reference on the returned
- * actions and must eventually release it (with rule_actions_unref()) to avoid
- * a memory leak. */
+/* Returns 'rule''s actions. The returned actions are RCU-protected, and can
+ * be read until the calling thread quiesces. */
const struct rule_actions *
rule_dpif_get_actions(const struct rule_dpif *rule)
{
/* Lookup 'flow' in table 0 of 'ofproto''s classifier.
* If 'wc' is non-null, sets the fields that were relevant as part of
- * the lookup. Returns the table_id where a match or miss occurred.
- *
- * The return value will be zero unless there was a miss and
+ * the lookup. Returns the table id where a match or miss occurred via
+ * 'table_id'. This will be zero unless there was a miss and
* OFPTC11_TABLE_MISS_CONTINUE is in effect for the sequence of tables
- * where misses occur.
+ * where misses occur, or TBL_INTERNAL if the rule has a non-zero
+ * recirculation ID, and a match was found in the internal table, or if
+ * there was no match and one of the special rules (drop_frags_rule,
+ * miss_rule, or no_packet_in_rule) was returned.
*
- * The rule is returned in '*rule', which is valid at least until the next
- * RCU quiescent period. If the '*rule' needs to stay around longer,
+ * The return value is the found rule, which is valid at least until the next
+ * RCU quiescent period. If the rule needs to stay around longer,
* a non-zero 'take_ref' must be passed in to cause a reference to be taken
* on it before this returns. */
-uint8_t
+struct rule_dpif *
rule_dpif_lookup(struct ofproto_dpif *ofproto, struct flow *flow,
- struct flow_wildcards *wc, struct rule_dpif **rule,
- bool take_ref, const struct dpif_flow_stats *stats)
+ struct flow_wildcards *wc, bool take_ref,
+ const struct dpif_flow_stats *stats, uint8_t *table_id)
{
- enum rule_dpif_lookup_verdict verdict;
- enum ofputil_port_config config = 0;
- uint8_t table_id;
+ *table_id = 0;
if (ofproto_dpif_get_enable_recirc(ofproto)) {
/* Always exactly match recirc_id since datapath supports
if (wc) {
wc->masks.recirc_id = UINT32_MAX;
}
-
- /* Start looking up from internal table for post recirculation flows
- * or packets. We can also simply send all, including normal flows
- * or packets to the internal table. They will not match any post
- * recirculation rules except the 'catch all' rule that resubmit
- * them to table 0.
- *
- * As an optimization, we send normal flows and packets to table 0
- * directly, saving one table lookup. */
- table_id = flow->recirc_id ? TBL_INTERNAL : 0;
- } else {
- table_id = 0;
- }
-
- verdict = rule_dpif_lookup_from_table(ofproto, flow, wc, true,
- &table_id, rule, take_ref, stats);
-
- switch (verdict) {
- case RULE_DPIF_LOOKUP_VERDICT_MATCH:
- return table_id;
- case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER: {
- struct ofport_dpif *port;
-
- port = get_ofp_port(ofproto, flow->in_port.ofp_port);
- if (!port) {
- VLOG_WARN_RL(&rl, "packet-in on unknown OpenFlow port %"PRIu16,
- flow->in_port.ofp_port);
- }
- config = port ? port->up.pp.config : 0;
- break;
- }
- case RULE_DPIF_LOOKUP_VERDICT_DROP:
- config = OFPUTIL_PC_NO_PACKET_IN;
- break;
- case RULE_DPIF_LOOKUP_VERDICT_DEFAULT:
- if (!connmgr_wants_packet_in_on_miss(ofproto->up.connmgr)) {
- config = OFPUTIL_PC_NO_PACKET_IN;
- }
- break;
- default:
- OVS_NOT_REACHED();
+ *table_id = rule_dpif_lookup_get_init_table_id(flow);
}
- choose_miss_rule(config, ofproto->miss_rule,
- ofproto->no_packet_in_rule, rule, take_ref);
- return table_id;
+ return rule_dpif_lookup_from_table(ofproto, flow, wc, take_ref, stats,
+ table_id, flow->in_port.ofp_port, true,
+ true);
}
-/* The returned rule is valid at least until the next RCU quiescent period.
- * If the '*rule' needs to stay around longer, a non-zero 'take_ref' must be
- * passed in to cause a reference to be taken on it before this returns. */
+/* The returned rule (if any) is valid at least until the next RCU quiescent
+ * period. If the rule needs to stay around longer, a non-zero 'take_ref'
+ * must be passed in to cause a reference to be taken on it.
+ *
+ * 'flow' is non-const to allow for temporary modifications during the lookup.
+ * Any changes are restored before returning. */
static struct rule_dpif *
rule_dpif_lookup_in_table(struct ofproto_dpif *ofproto, uint8_t table_id,
- const struct flow *flow, struct flow_wildcards *wc,
+ struct flow *flow, struct flow_wildcards *wc,
bool take_ref)
{
struct classifier *cls = &ofproto->up.tables[table_id].cls;
const struct cls_rule *cls_rule;
struct rule_dpif *rule;
- fat_rwlock_rdlock(&cls->rwlock);
- if (ofproto->up.frag_handling != OFPC_FRAG_NX_MATCH) {
- if (wc) {
- memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
- if (is_ip_any(flow)) {
- wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
- }
- }
-
- if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
- if (ofproto->up.frag_handling == OFPC_FRAG_NORMAL) {
- /* We must pretend that transport ports are unavailable. */
- struct flow ofpc_normal_flow = *flow;
- ofpc_normal_flow.tp_src = htons(0);
- ofpc_normal_flow.tp_dst = htons(0);
- cls_rule = classifier_lookup(cls, &ofpc_normal_flow, wc);
- } else {
- /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM). */
- cls_rule = &ofproto->drop_frags_rule->up.cr;
- }
- } else {
- cls_rule = classifier_lookup(cls, flow, wc);
- }
- } else {
+ do {
cls_rule = classifier_lookup(cls, flow, wc);
- }
- rule = rule_dpif_cast(rule_from_cls_rule(cls_rule));
- if (take_ref) {
- rule_dpif_ref(rule);
- }
- fat_rwlock_unlock(&cls->rwlock);
+ rule = rule_dpif_cast(rule_from_cls_rule(cls_rule));
+
+ /* Try again if the rule was released before we get the reference. */
+ } while (rule && take_ref && !rule_dpif_try_ref(rule));
return rule;
}
/* Look up 'flow' in 'ofproto''s classifier starting from table '*table_id'.
- * Stores the rule that was found in '*rule', or NULL if none was found.
+ * Returns the rule that was found, which may be one of the special rules
+ * according to packet miss hadling. If 'may_packet_in' is false, returning of
+ * the miss_rule (which issues packet ins for the controller) is avoided.
* Updates 'wc', if nonnull, to reflect the fields that were used during the
* lookup.
*
* If 'honor_table_miss' is false, then only one table lookup occurs, in
* '*table_id'.
*
- * Returns:
- *
- * - RULE_DPIF_LOOKUP_VERDICT_MATCH if a rule (in '*rule') was found.
- *
- * - RULE_OFPTC_TABLE_MISS_CONTROLLER if no rule was found and either:
- * + 'honor_table_miss' is false
- * + a table miss configuration specified that the packet should be
- * sent to the controller in this case.
- *
- * - RULE_DPIF_LOOKUP_VERDICT_DROP if no rule was found, 'honor_table_miss'
- * is true and a table miss configuration specified that the packet
- * should be dropped in this case.
- *
- * - RULE_DPIF_LOOKUP_VERDICT_DEFAULT if no rule was found,
- * 'honor_table_miss' is true and a table miss configuration has
- * not been specified in this case.
- *
* The rule is returned in '*rule', which is valid at least until the next
* RCU quiescent period. If the '*rule' needs to stay around longer,
* a non-zero 'take_ref' must be passed in to cause a reference to be taken
- * on it before this returns. */
-enum rule_dpif_lookup_verdict
-rule_dpif_lookup_from_table(struct ofproto_dpif *ofproto,
- const struct flow *flow,
- struct flow_wildcards *wc,
- bool honor_table_miss,
- uint8_t *table_id, struct rule_dpif **rule,
- bool take_ref, const struct dpif_flow_stats *stats)
-{
+ * on it before this returns.
+ *
+ * 'in_port' allows the lookup to take place as if the in port had the value
+ * 'in_port'. This is needed for resubmit action support.
+ *
+ * 'flow' is non-const to allow for temporary modifications during the lookup.
+ * Any changes are restored before returning. */
+struct rule_dpif *
+rule_dpif_lookup_from_table(struct ofproto_dpif *ofproto, struct flow *flow,
+ struct flow_wildcards *wc, bool take_ref,
+ const struct dpif_flow_stats *stats,
+ uint8_t *table_id, ofp_port_t in_port,
+ bool may_packet_in, bool honor_table_miss)
+{
+ ovs_be16 old_tp_src = flow->tp_src, old_tp_dst = flow->tp_dst;
+ ofp_port_t old_in_port = flow->in_port.ofp_port;
+ enum ofputil_table_miss miss_config;
+ struct rule_dpif *rule;
uint8_t next_id;
+ /* We always unwildcard nw_frag (for IP), so they
+ * need not be unwildcarded here. */
+ if (flow->nw_frag & FLOW_NW_FRAG_ANY
+ && ofproto->up.frag_handling != OFPC_FRAG_NX_MATCH) {
+ if (ofproto->up.frag_handling == OFPC_FRAG_NORMAL) {
+ /* We must pretend that transport ports are unavailable. */
+ flow->tp_src = htons(0);
+ flow->tp_dst = htons(0);
+ } else {
+ /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM).
+ * Use the drop_frags_rule (which cannot disappear). */
+ rule = ofproto->drop_frags_rule;
+ if (take_ref) {
+ rule_dpif_ref(rule);
+ }
+ if (stats) {
+ struct oftable *tbl = &ofproto->up.tables[*table_id];
+ unsigned long orig;
+
+ atomic_add_relaxed(&tbl->n_matched, stats->n_packets, &orig);
+ }
+ return rule;
+ }
+ }
+
+ /* Look up a flow with 'in_port' as the input port. Then restore the
+ * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
+ * have surprising behavior). */
+ flow->in_port.ofp_port = in_port;
+
+ /* Our current implementation depends on n_tables == N_TABLES, and
+ * TBL_INTERNAL being the last table. */
+ BUILD_ASSERT_DECL(N_TABLES == TBL_INTERNAL + 1);
+
+ miss_config = OFPUTIL_TABLE_MISS_CONTINUE;
+
for (next_id = *table_id;
next_id < ofproto->up.n_tables;
next_id++, next_id += (next_id == TBL_INTERNAL))
{
*table_id = next_id;
- *rule = rule_dpif_lookup_in_table(ofproto, *table_id, flow, wc,
- take_ref);
+ rule = rule_dpif_lookup_in_table(ofproto, next_id, flow, wc, take_ref);
if (stats) {
struct oftable *tbl = &ofproto->up.tables[next_id];
- atomic_ulong *stat = *rule ? &tbl->n_matched : &tbl->n_missed;
unsigned long orig;
- atomic_add(stat, stats->n_packets, &orig);
- }
- if (*rule) {
- return RULE_DPIF_LOOKUP_VERDICT_MATCH;
- } else if (!honor_table_miss) {
- return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER;
- } else {
- switch (ofproto_table_get_config(&ofproto->up, *table_id)) {
- case OFPROTO_TABLE_MISS_CONTINUE:
- break;
-
- case OFPROTO_TABLE_MISS_CONTROLLER:
- return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER;
- case OFPROTO_TABLE_MISS_DROP:
- return RULE_DPIF_LOOKUP_VERDICT_DROP;
-
- case OFPROTO_TABLE_MISS_DEFAULT:
- return RULE_DPIF_LOOKUP_VERDICT_DEFAULT;
+ atomic_add_relaxed(rule ? &tbl->n_matched : &tbl->n_missed,
+ stats->n_packets, &orig);
+ }
+ if (rule) {
+ goto out; /* Match. */
+ }
+ if (honor_table_miss) {
+ miss_config = ofproto_table_get_miss_config(&ofproto->up,
+ *table_id);
+ if (miss_config == OFPUTIL_TABLE_MISS_CONTINUE) {
+ continue;
}
}
+ break;
+ }
+ /* Miss. */
+ rule = ofproto->no_packet_in_rule;
+ if (may_packet_in) {
+ if (miss_config == OFPUTIL_TABLE_MISS_CONTINUE
+ || miss_config == OFPUTIL_TABLE_MISS_CONTROLLER) {
+ struct ofport_dpif *port;
+
+ port = get_ofp_port(ofproto, old_in_port);
+ if (!port) {
+ VLOG_WARN_RL(&rl, "packet-in on unknown OpenFlow port %"PRIu16,
+ old_in_port);
+ } else if (!(port->up.pp.config & OFPUTIL_PC_NO_PACKET_IN)) {
+ rule = ofproto->miss_rule;
+ }
+ } else if (miss_config == OFPUTIL_TABLE_MISS_DEFAULT &&
+ connmgr_wants_packet_in_on_miss(ofproto->up.connmgr)) {
+ rule = ofproto->miss_rule;
+ }
}
-
- return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER;
-}
-
-/* Given a port configuration (specified as zero if there's no port), chooses
- * which of 'miss_rule' and 'no_packet_in_rule' should be used in case of a
- * flow table miss.
- *
- * The rule is returned in '*rule', which is valid at least until the next
- * RCU quiescent period. If the '*rule' needs to stay around longer,
- * a reference must be taken on it (rule_dpif_ref()).
- */
-void
-choose_miss_rule(enum ofputil_port_config config, struct rule_dpif *miss_rule,
- struct rule_dpif *no_packet_in_rule, struct rule_dpif **rule,
- bool take_ref)
-{
- *rule = config & OFPUTIL_PC_NO_PACKET_IN ? no_packet_in_rule : miss_rule;
if (take_ref) {
- rule_dpif_ref(*rule);
+ rule_dpif_ref(rule);
}
+out:
+ /* Restore port numbers, as they may have been modified above. */
+ flow->tp_src = old_tp_src;
+ flow->tp_dst = old_tp_dst;
+ /* Restore the old in port. */
+ flow->in_port.ofp_port = old_in_port;
+
+ return rule;
}
static void
OVS_REQUIRES(group->stats_mutex)
{
struct ofputil_bucket *bucket;
- const struct list *buckets;
+ const struct ovs_list *buckets;
group->packet_count = 0;
group->byte_count = 0;
bucket->stats.packet_count += stats->n_packets;
bucket->stats.byte_count += stats->n_bytes;
} else { /* Credit to all buckets */
- const struct list *buckets;
+ const struct ovs_list *buckets;
group_dpif_get_buckets(group, &buckets);
LIST_FOR_EACH (bucket, list_node, buckets) {
{
struct group_dpif *group = group_dpif_cast(group_);
struct ofputil_bucket *bucket;
- const struct list *buckets;
+ const struct ovs_list *buckets;
struct bucket_counter *bucket_stats;
ovs_mutex_lock(&group->stats_mutex);
void
group_dpif_get_buckets(const struct group_dpif *group,
- const struct list **buckets)
+ const struct ovs_list **buckets)
{
*buckets = &group->up.buckets;
}
return error;
}
\f
+/* Return the version string of the datapath that backs up
+ * this 'ofproto'.
+ */
+static const char *
+get_datapath_version(const struct ofproto *ofproto_)
+{
+ struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
+
+ return ofproto->backer->dp_version_string;
+}
+
static bool
set_frag_handling(struct ofproto *ofproto_,
enum ofp_config_flags frag_handling)
unixctl_command_reply(conn, "table successfully flushed");
}
+static void
+ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn *conn, int argc,
+ const char *argv[], void *aux OVS_UNUSED)
+{
+ struct ofproto_dpif *ofproto;
+
+ if (argc > 1) {
+ ofproto = ofproto_dpif_lookup(argv[1]);
+ if (!ofproto) {
+ unixctl_command_reply_error(conn, "no such bridge");
+ return;
+ }
+
+ if (!mcast_snooping_enabled(ofproto->ms)) {
+ unixctl_command_reply_error(conn, "multicast snooping is disabled");
+ return;
+ }
+ mcast_snooping_mdb_flush(ofproto->ms);
+ } else {
+ HMAP_FOR_EACH (ofproto, all_ofproto_dpifs_node, &all_ofproto_dpifs) {
+ if (!mcast_snooping_enabled(ofproto->ms)) {
+ continue;
+ }
+ mcast_snooping_mdb_flush(ofproto->ms);
+ }
+ }
+
+ unixctl_command_reply(conn, "table successfully flushed");
+}
+
static struct ofport_dpif *
ofbundle_get_a_port(const struct ofbundle *bundle)
{
ds_put_cstr(&ds, " port VLAN MAC Age\n");
ovs_rwlock_rdlock(&ofproto->ml->rwlock);
LIST_FOR_EACH (e, lru_node, &ofproto->ml->lrus) {
- struct ofbundle *bundle = e->port.p;
+ struct ofbundle *bundle = mac_entry_get_port(ofproto->ml, e);
char name[OFP_MAX_PORT_NAME_LEN];
ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
ds_destroy(&ds);
}
+static void
+ofproto_unixctl_mcast_snooping_show(struct unixctl_conn *conn,
+ int argc OVS_UNUSED,
+ const char *argv[],
+ void *aux OVS_UNUSED)
+{
+ struct ds ds = DS_EMPTY_INITIALIZER;
+ const struct ofproto_dpif *ofproto;
+ const struct ofbundle *bundle;
+ const struct mcast_group *grp;
+ struct mcast_group_bundle *b;
+ struct mcast_mrouter_bundle *mrouter;
+
+ ofproto = ofproto_dpif_lookup(argv[1]);
+ if (!ofproto) {
+ unixctl_command_reply_error(conn, "no such bridge");
+ return;
+ }
+
+ if (!mcast_snooping_enabled(ofproto->ms)) {
+ unixctl_command_reply_error(conn, "multicast snooping is disabled");
+ return;
+ }
+
+ ds_put_cstr(&ds, " port VLAN GROUP Age\n");
+ ovs_rwlock_rdlock(&ofproto->ms->rwlock);
+ LIST_FOR_EACH (grp, group_node, &ofproto->ms->group_lru) {
+ LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) {
+ char name[OFP_MAX_PORT_NAME_LEN];
+
+ bundle = b->port;
+ ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
+ name, sizeof name);
+ ds_put_format(&ds, "%5s %4d "IP_FMT" %3d\n",
+ name, grp->vlan, IP_ARGS(grp->ip4),
+ mcast_bundle_age(ofproto->ms, b));
+ }
+ }
+
+ /* ports connected to multicast routers */
+ LIST_FOR_EACH(mrouter, mrouter_node, &ofproto->ms->mrouter_lru) {
+ char name[OFP_MAX_PORT_NAME_LEN];
+
+ bundle = mrouter->port;
+ ofputil_port_to_string(ofbundle_get_a_port(bundle)->up.ofp_port,
+ name, sizeof name);
+ ds_put_format(&ds, "%5s %4d querier %3d\n",
+ name, mrouter->vlan,
+ mcast_mrouter_age(ofproto->ms, mrouter));
+ }
+ ovs_rwlock_unlock(&ofproto->ms->rwlock);
+ unixctl_command_reply(conn, ds_cstr(&ds));
+ ds_destroy(&ds);
+}
+
struct trace_ctx {
struct xlate_out xout;
struct xlate_in xin;
trace_format_odp(struct ds *result, int level, const char *title,
struct trace_ctx *trace)
{
- struct ofpbuf *odp_actions = &trace->xout.odp_actions;
+ struct ofpbuf *odp_actions = trace->xout.odp_actions;
ds_put_char_multiple(result, '\t', level);
ds_put_format(result, "%s: ", title);
ds_put_char(result, '\n');
}
+static void trace_report(struct xlate_in *xin, const char *s, int recurse);
+
static void
trace_resubmit(struct xlate_in *xin, struct rule_dpif *rule, int recurse)
{
struct trace_ctx *trace = CONTAINER_OF(xin, struct trace_ctx, xin);
struct ds *result = trace->result;
+ if (!recurse) {
+ if (rule == xin->ofproto->miss_rule) {
+ trace_report(xin, "No match, flow generates \"packet in\"s.",
+ recurse);
+ } else if (rule == xin->ofproto->no_packet_in_rule) {
+ trace_report(xin, "No match, packets dropped because "
+ "OFPPC_NO_PACKET_IN is set on in_port.", recurse);
+ } else if (rule == xin->ofproto->drop_frags_rule) {
+ trace_report(xin, "Packets dropped because they are IP "
+ "fragments and the fragment handling mode is "
+ "\"drop\".", recurse);
+ }
+ }
+
ds_put_char(result, '\n');
- trace_format_flow(result, recurse + 1, "Resubmitted flow", trace);
- trace_format_regs(result, recurse + 1, "Resubmitted regs", trace);
- trace_format_odp(result, recurse + 1, "Resubmitted odp", trace);
- trace_format_megaflow(result, recurse + 1, "Resubmitted megaflow", trace);
- trace_format_rule(result, recurse + 1, rule);
+ if (recurse) {
+ trace_format_flow(result, recurse, "Resubmitted flow", trace);
+ trace_format_regs(result, recurse, "Resubmitted regs", trace);
+ trace_format_odp(result, recurse, "Resubmitted odp", trace);
+ trace_format_megaflow(result, recurse, "Resubmitted megaflow", trace);
+ }
+ trace_format_rule(result, recurse, rule);
}
static void
*
* On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
* returns a nonnull malloced error message. */
-static char * WARN_UNUSED_RESULT
+static char * OVS_WARN_UNUSED_RESULT
parse_flow_and_packet(int argc, const char *argv[],
struct ofproto_dpif **ofprotop, struct flow *flow,
struct ofpbuf **packetp)
goto exit;
}
- if (xlate_receive(backer, NULL, ofpbuf_data(&odp_key),
- ofpbuf_size(&odp_key), flow,
- ofprotop, NULL, NULL, NULL, NULL)) {
+ if (odp_flow_key_to_flow(ofpbuf_data(&odp_key), ofpbuf_size(&odp_key),
+ flow) == ODP_FIT_ERROR) {
+ error = "Failed to parse flow key";
+ goto exit;
+ }
+
+ *ofprotop = xlate_lookup_ofproto(backer, flow,
+ &flow->in_port.ofp_port);
+ if (*ofprotop == NULL) {
error = "Invalid datapath flow";
goto exit;
}
+
+ vsp_adjust_flow(*ofprotop, flow, NULL);
+
} else {
char *err = parse_ofp_exact_flow(flow, NULL, argv[argc - 1], NULL);
ofpbuf_init(&ofpacts, 0);
/* Parse actions. */
- error = parse_ofpacts(argv[--argc], &ofpacts, &usable_protocols);
+ error = ofpacts_parse_actions(argv[--argc], &ofpacts, &usable_protocols);
if (error) {
unixctl_command_reply_error(conn, error);
free(error);
/* Do the same checks as handle_packet_out() in ofproto.c.
*
- * We pass a 'table_id' of 0 to ofproto_check_ofpacts(), which isn't
+ * We pass a 'table_id' of 0 to ofpacts_check(), which isn't
* strictly correct because these actions aren't in any table, but it's OK
* because it 'table_id' is used only to check goto_table instructions, but
* packet-outs take a list of actions and therefore it can't include
const struct ofpact ofpacts[], size_t ofpacts_len,
struct ds *ds)
{
- struct rule_dpif *rule;
struct trace_ctx trace;
ds_put_format(ds, "Bridge: %s\n", ofproto->up.name);
ds_put_char(ds, '\n');
flow_wildcards_init_catchall(&trace.wc);
- if (ofpacts) {
- rule = NULL;
- } else {
- rule_dpif_lookup(ofproto, flow, &trace.wc, &rule, false, NULL);
- trace_format_rule(ds, 0, rule);
- if (rule == ofproto->miss_rule) {
- ds_put_cstr(ds, "\nNo match, flow generates \"packet in\"s.\n");
- } else if (rule == ofproto->no_packet_in_rule) {
- ds_put_cstr(ds, "\nNo match, packets dropped because "
- "OFPPC_NO_PACKET_IN is set on in_port.\n");
- } else if (rule == ofproto->drop_frags_rule) {
- ds_put_cstr(ds, "\nPackets dropped because they are IP fragments "
- "and the fragment handling mode is \"drop\".\n");
- }
- }
-
- if (rule || ofpacts) {
- trace.result = ds;
- trace.key = flow; /* Original flow key, used for megaflow. */
- trace.flow = *flow; /* May be modified by actions. */
- xlate_in_init(&trace.xin, ofproto, flow, rule, ntohs(flow->tcp_flags),
- packet);
- if (ofpacts) {
- trace.xin.ofpacts = ofpacts;
- trace.xin.ofpacts_len = ofpacts_len;
- }
- trace.xin.resubmit_hook = trace_resubmit;
- trace.xin.report_hook = trace_report;
+ trace.result = ds;
+ trace.key = flow; /* Original flow key, used for megaflow. */
+ trace.flow = *flow; /* May be modified by actions. */
+ xlate_in_init(&trace.xin, ofproto, flow, flow->in_port.ofp_port, NULL,
+ ntohs(flow->tcp_flags), packet);
+ trace.xin.ofpacts = ofpacts;
+ trace.xin.ofpacts_len = ofpacts_len;
+ trace.xin.resubmit_hook = trace_resubmit;
+ trace.xin.report_hook = trace_report;
- xlate_actions(&trace.xin, &trace.xout);
+ xlate_actions(&trace.xin, &trace.xout);
- ds_put_char(ds, '\n');
- trace_format_flow(ds, 0, "Final flow", &trace);
- trace_format_megaflow(ds, 0, "Megaflow", &trace);
+ ds_put_char(ds, '\n');
+ trace_format_flow(ds, 0, "Final flow", &trace);
+ trace_format_megaflow(ds, 0, "Megaflow", &trace);
- ds_put_cstr(ds, "Datapath actions: ");
- format_odp_actions(ds, ofpbuf_data(&trace.xout.odp_actions),
- ofpbuf_size(&trace.xout.odp_actions));
+ ds_put_cstr(ds, "Datapath actions: ");
+ format_odp_actions(ds, ofpbuf_data(trace.xout.odp_actions),
+ ofpbuf_size(trace.xout.odp_actions));
- if (trace.xout.slow) {
- enum slow_path_reason slow;
+ if (trace.xout.slow) {
+ enum slow_path_reason slow;
- ds_put_cstr(ds, "\nThis flow is handled by the userspace "
- "slow path because it:");
+ ds_put_cstr(ds, "\nThis flow is handled by the userspace "
+ "slow path because it:");
- slow = trace.xout.slow;
- while (slow) {
- enum slow_path_reason bit = rightmost_1bit(slow);
+ slow = trace.xout.slow;
+ while (slow) {
+ enum slow_path_reason bit = rightmost_1bit(slow);
- ds_put_format(ds, "\n\t- %s.",
- slow_path_reason_to_explanation(bit));
+ ds_put_format(ds, "\n\t- %s.",
+ slow_path_reason_to_explanation(bit));
- slow &= ~bit;
- }
+ slow &= ~bit;
}
-
- xlate_out_uninit(&trace.xout);
}
+
+ xlate_out_uninit(&trace.xout);
}
/* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
ds_destroy(&ds);
}
-static bool
-ofproto_dpif_contains_flow(const struct ofproto_dpif *ofproto,
- const struct nlattr *key, size_t key_len)
-{
- struct ofproto_dpif *ofp;
- struct flow flow;
-
- xlate_receive(ofproto->backer, NULL, key, key_len, &flow, &ofp,
- NULL, NULL, NULL, NULL);
- return ofp == ofproto;
-}
-
static void
ofproto_unixctl_dpif_dump_flows(struct unixctl_conn *conn,
int argc OVS_UNUSED, const char *argv[],
}
ds_init(&ds);
- flow_dump = dpif_flow_dump_create(ofproto->backer->dpif);
+ flow_dump = dpif_flow_dump_create(ofproto->backer->dpif, false);
flow_dump_thread = dpif_flow_dump_thread_create(flow_dump);
while (dpif_flow_dump_next(flow_dump_thread, &f, 1)) {
- if (!ofproto_dpif_contains_flow(ofproto, f.key, f.key_len)) {
+ struct flow flow;
+
+ if (odp_flow_key_to_flow(f.key, f.key_len, &flow) == ODP_FIT_ERROR
+ || xlate_lookup_ofproto(ofproto->backer, &flow, NULL) != ofproto) {
continue;
}
+ if (verbosity) {
+ odp_format_ufid(&f.ufid, &ds);
+ ds_put_cstr(&ds, " ");
+ }
odp_flow_format(f.key, f.key_len, f.mask, f.mask_len,
&portno_names, &ds, verbosity);
ds_put_cstr(&ds, ", ");
}
static void
-ofproto_dpif_unixctl_init(void)
+ofproto_revalidate_all_backers(void)
+{
+ const struct shash_node **backers;
+ int i;
+
+ backers = shash_sort(&all_dpif_backers);
+ for (i = 0; i < shash_count(&all_dpif_backers); i++) {
+ struct dpif_backer *backer = backers[i]->data;
+ backer->need_revalidate = REV_RECONFIGURE;
+ }
+ free(backers);
+}
+
+static void
+disable_tnl_push_pop(struct unixctl_conn *conn OVS_UNUSED, int argc OVS_UNUSED,
+ const char *argv[], void *aux OVS_UNUSED)
+{
+ if (!strcasecmp(argv[1], "off")) {
+ ofproto_use_tnl_push_pop = false;
+ unixctl_command_reply(conn, "Tunnel push-pop off");
+ ofproto_revalidate_all_backers();
+ } else if (!strcasecmp(argv[1], "on")) {
+ ofproto_use_tnl_push_pop = true;
+ unixctl_command_reply(conn, "Tunnel push-pop on");
+ ofproto_revalidate_all_backers();
+ }
+}
+
+static void
+ofproto_unixctl_init(void)
{
static bool registered;
if (registered) {
ofproto_unixctl_fdb_flush, NULL);
unixctl_command_register("fdb/show", "bridge", 1, 1,
ofproto_unixctl_fdb_show, NULL);
+ unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
+ ofproto_unixctl_mcast_snooping_flush, NULL);
+ unixctl_command_register("mdb/show", "bridge", 1, 1,
+ ofproto_unixctl_mcast_snooping_show, NULL);
unixctl_command_register("dpif/dump-dps", "", 0, 0,
ofproto_unixctl_dpif_dump_dps, NULL);
unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show,
NULL);
unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
ofproto_unixctl_dpif_dump_flows, NULL);
+
+ unixctl_command_register("ofproto/tnl-push-pop", "[on]|[off]", 1, 1,
+ disable_tnl_push_pop, NULL);
}
/* Returns true if 'table' is the table used for internal rules,
/* Given 'flow', a flow representing a packet received on 'ofproto', checks
* whether 'flow->in_port' represents a Linux VLAN device. If so, changes
* 'flow->in_port' to the "real" device backing the VLAN device, sets
- * 'flow->vlan_tci' to the VLAN VID, and returns true. Otherwise (which is
- * always the case unless VLAN splinters are enabled), returns false without
- * making any changes. */
+ * 'flow->vlan_tci' to the VLAN VID, and returns true. Optionally pushes the
+ * appropriate VLAN on 'packet' if provided. Otherwise (which is always the
+ * case unless VLAN splinters are enabled), returns false without making any
+ * changes. */
bool
-vsp_adjust_flow(const struct ofproto_dpif *ofproto, struct flow *flow)
+vsp_adjust_flow(const struct ofproto_dpif *ofproto, struct flow *flow,
+ struct ofpbuf *packet)
OVS_EXCLUDED(ofproto->vsp_mutex)
{
ofp_port_t realdev;
* the VLAN device's VLAN ID. */
flow->in_port.ofp_port = realdev;
flow->vlan_tci = htons((vid & VLAN_VID_MASK) | VLAN_CFI);
+
+ if (packet) {
+ /* Make the packet resemble the flow, so that it gets sent to an
+ * OpenFlow controller properly, so that it looks correct for sFlow,
+ * and so that flow_extract() will get the correct vlan_tci if it is
+ * called on 'packet'. */
+ eth_push_vlan(packet, htons(ETH_TYPE_VLAN), flow->vlan_tci);
+ }
+
return true;
}
}
}
+struct ofproto_dpif *
+ofproto_dpif_recirc_get_ofproto(const struct dpif_backer *backer,
+ uint32_t recirc_id)
+{
+ struct dpif_backer_recirc_node *node;
+
+ node = CONTAINER_OF(cmap_find(&backer->recirc_map, recirc_id),
+ struct dpif_backer_recirc_node, cmap_node);
+
+ return node ? node->ofproto : NULL;
+}
+
uint32_t
ofproto_dpif_alloc_recirc_id(struct ofproto_dpif *ofproto)
{
struct dpif_backer *backer = ofproto->backer;
+ uint32_t recirc_id = recirc_id_alloc(backer->rid_pool);
- return recirc_id_alloc(backer->rid_pool);
+ if (recirc_id) {
+ struct dpif_backer_recirc_node *node = xmalloc(sizeof *node);
+
+ node->recirc_id = recirc_id;
+ node->ofproto = ofproto;
+
+ ovs_mutex_lock(&backer->recirc_mutex);
+ cmap_insert(&backer->recirc_map, &node->cmap_node, node->recirc_id);
+ ovs_mutex_unlock(&backer->recirc_mutex);
+ }
+
+ return recirc_id;
}
void
ofproto_dpif_free_recirc_id(struct ofproto_dpif *ofproto, uint32_t recirc_id)
{
struct dpif_backer *backer = ofproto->backer;
+ struct dpif_backer_recirc_node *node;
- recirc_id_free(backer->rid_pool, recirc_id);
+ node = CONTAINER_OF(cmap_find(&backer->recirc_map, recirc_id),
+ struct dpif_backer_recirc_node, cmap_node);
+ if (node) {
+ ovs_mutex_lock(&backer->recirc_mutex);
+ cmap_remove(&backer->recirc_map, &node->cmap_node, node->recirc_id);
+ ovs_mutex_unlock(&backer->recirc_mutex);
+ recirc_id_free(backer->rid_pool, node->recirc_id);
+
+ /* 'recirc_id' should never be freed by non-owning 'ofproto'. */
+ ovs_assert(node->ofproto == ofproto);
+
+ /* RCU postpone the free, since other threads may be referring
+ * to 'node' at same time. */
+ ovsrcu_postpone(free, node);
+ }
}
int
fm.command = OFPFC_ADD;
fm.idle_timeout = idle_timeout;
fm.hard_timeout = 0;
+ fm.importance = 0;
fm.buffer_id = 0;
fm.out_port = 0;
fm.flags = OFPUTIL_FF_HIDDEN_FIELDS | OFPUTIL_FF_NO_READONLY;
NULL, /* get_memory_usage. */
type_get_memory_usage,
flush,
- get_features,
- get_tables,
+ query_tables,
port_alloc,
port_construct,
port_destruct,
port_poll,
port_poll_wait,
port_is_lacp_current,
+ port_get_lacp_stats,
NULL, /* rule_choose_table */
rule_alloc,
rule_construct,
set_stp_port,
get_stp_port_status,
get_stp_port_stats,
+ set_rstp,
+ get_rstp_status,
+ set_rstp_port,
+ get_rstp_port_status,
set_queues,
bundle_set,
bundle_remove,
is_mirror_output_bundle,
forward_bpdu_changed,
set_mac_table_config,
+ set_mcast_snooping,
+ set_mcast_snooping_port,
set_realdev,
NULL, /* meter_get_features */
NULL, /* meter_set */
group_dealloc, /* group_dealloc */
group_modify, /* group_modify */
group_get_stats, /* group_get_stats */
+ get_datapath_version, /* get_datapath_version */
};