-/* Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
+/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 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 "ofproto/ofproto-dpif-xlate.h"
+#include <errno.h>
+
#include "bfd.h"
#include "bitmap.h"
#include "bond.h"
#include "ofp-actions.h"
#include "ofproto/ofproto-dpif-ipfix.h"
#include "ofproto/ofproto-dpif-mirror.h"
+#include "ofproto/ofproto-dpif-monitor.h"
#include "ofproto/ofproto-dpif-sflow.h"
#include "ofproto/ofproto-dpif.h"
+#include "ofproto/ofproto-provider.h"
#include "tunnel.h"
#include "vlog.h"
COVERAGE_DEFINE(xlate_actions);
+COVERAGE_DEFINE(xlate_actions_oversize);
+COVERAGE_DEFINE(xlate_actions_mpls_overflow);
VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate);
/* Maximum depth of flow table recursion (due to resubmit actions) in a
* flow translation. */
#define MAX_RESUBMIT_RECURSION 64
+#define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
+ internal table. */
+
+/* Maximum number of resubmit actions in a flow translation, whether they are
+ * recursive or not. */
+#define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION)
+
+struct ovs_rwlock xlate_rwlock = OVS_RWLOCK_INITIALIZER;
struct xbridge {
struct hmap_node hmap_node; /* Node in global 'xbridges' map. */
struct hmap xports; /* Indexed by ofp_port. */
char *name; /* Name used in log messages. */
+ struct dpif *dpif; /* Datapath interface. */
struct mac_learning *ml; /* Mac learning handle. */
struct mbridge *mbridge; /* Mirroring. */
struct dpif_sflow *sflow; /* SFlow handle, or null. */
struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */
+ struct netflow *netflow; /* Netflow handle, or null. */
+ struct stp *stp; /* STP or null if disabled. */
+
+ /* Special rules installed by ofproto-dpif. */
+ struct rule_dpif *miss_rule;
+ struct rule_dpif *no_packet_in_rule;
enum ofp_config_flags frag; /* Fragmentation handling. */
- bool has_stp; /* Bridge runs stp? */
- bool has_netflow; /* Bridge runs netflow? */
bool has_in_band; /* Bridge has in band control? */
bool forward_bpdu; /* Bridge forwards STP BPDUs? */
+
+ /* True if the datapath supports recirculation. */
+ bool enable_recirc;
+
+ /* 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;
+
+ /* Number of MPLS label stack entries that the datapath supports
+ * in matches. */
+ size_t max_mpls_depth;
};
struct xbundle {
struct xport *peer; /* Patch port peer or null. */
enum ofputil_port_config config; /* OpenFlow port configuration. */
- enum stp_state stp_state; /* STP_DISABLED if STP not in use. */
+ enum ofputil_port_state state; /* OpenFlow port state. */
+ int stp_port_no; /* STP port number or -1 if not in use. */
+
+ struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */
bool may_enable; /* May be enabled in bonds. */
bool is_tunnel; /* Is a tunnel port. */
/* The rule that we are currently translating, or NULL. */
struct rule_dpif *rule;
- int recurse; /* Recursion level, via xlate_table_action. */
- bool max_resubmit_trigger; /* Recursed too deeply during translation. */
+ /* Resubmit statistics, via xlate_table_action(). */
+ int recurse; /* Current resubmit nesting depth. */
+ int resubmits; /* Total number of resubmits. */
+ bool in_group; /* Currently translating ofgroup, if true. */
+
uint32_t orig_skb_priority; /* Priority when packet arrived. */
uint8_t table_id; /* OpenFlow table ID where flow was found. */
uint32_t sflow_n_outputs; /* Number of output ports. */
odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */
uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */
bool exit; /* No further actions should be processed. */
+
+ /* OpenFlow 1.1+ action set.
+ *
+ * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS.
+ * When translation is otherwise complete, ofpacts_execute_action_set()
+ * converts it to a set of "struct ofpact"s that can be translated into
+ * datapath actions. */
+ struct ofpbuf action_set; /* Action set. */
+ uint64_t action_set_stub[1024 / 8];
};
/* A controller may use OFPP_NONE as the ingress port to indicate that
* it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
* when an input bundle is needed for validation (e.g., mirroring or
* OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
- * any 'port' structs, so care must be taken when dealing with it.
- * The bundle's name and vlan mode are initialized in lookup_input_bundle() */
-static struct xbundle ofpp_none_bundle;
+ * any 'port' structs, so care must be taken when dealing with it. */
+static struct xbundle ofpp_none_bundle = {
+ .name = "OFPP_NONE",
+ .vlan_mode = PORT_VLAN_TRUNK
+};
+
+/* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
+ * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
+ * traffic egressing the 'ofport' with that priority should be marked with. */
+struct skb_priority_to_dscp {
+ struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */
+ uint32_t skb_priority; /* Priority of this queue (see struct flow). */
+
+ uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */
+};
+
+enum xc_type {
+ XC_RULE,
+ XC_BOND,
+ XC_NETDEV,
+ XC_NETFLOW,
+ XC_MIRROR,
+ XC_LEARN,
+ XC_NORMAL,
+ XC_FIN_TIMEOUT,
+};
+
+/* xlate_cache entries hold enough information to perform the side effects of
+ * xlate_actions() for a rule, without needing to perform rule translation
+ * from scratch. The primary usage of these is to submit statistics to objects
+ * that a flow relates to, although they may be used for other effects as well
+ * (for instance, refreshing hard timeouts for learned flows). */
+struct xc_entry {
+ enum xc_type type;
+ union {
+ struct rule_dpif *rule;
+ struct {
+ struct netdev *tx;
+ struct netdev *rx;
+ struct bfd *bfd;
+ } dev;
+ struct {
+ struct netflow *netflow;
+ struct flow *flow;
+ ofp_port_t iface;
+ } nf;
+ struct {
+ struct mbridge *mbridge;
+ mirror_mask_t mirrors;
+ } mirror;
+ struct {
+ struct bond *bond;
+ struct flow *flow;
+ uint16_t vid;
+ } bond;
+ struct {
+ struct ofproto_dpif *ofproto;
+ struct rule_dpif *rule;
+ } learn;
+ struct {
+ struct ofproto_dpif *ofproto;
+ struct flow *flow;
+ int vlan;
+ } normal;
+ struct {
+ struct rule_dpif *rule;
+ uint16_t idle;
+ uint16_t hard;
+ } fin;
+ } u;
+};
+
+#define XC_ENTRY_FOR_EACH(entry, entries, xcache) \
+ entries = xcache->entries; \
+ for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \
+ entry; \
+ entry = ofpbuf_try_pull(&entries, sizeof *entry))
+
+struct xlate_cache {
+ struct ofpbuf entries;
+};
static struct hmap xbridges = HMAP_INITIALIZER(&xbridges);
static struct hmap xbundles = HMAP_INITIALIZER(&xbundles);
static bool may_receive(const struct xport *, struct xlate_ctx *);
static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len,
struct xlate_ctx *);
+static void xlate_actions__(struct xlate_in *, struct xlate_out *)
+ OVS_REQ_RDLOCK(xlate_rwlock);
static void xlate_normal(struct xlate_ctx *);
static void xlate_report(struct xlate_ctx *, const char *);
static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
- uint8_t table_id, bool may_packet_in);
+ uint8_t table_id, bool may_packet_in,
+ bool honor_table_miss);
static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn);
static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid);
static void output_normal(struct xlate_ctx *, const struct xbundle *,
uint16_t vlan);
static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port);
-static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
-
static struct xbridge *xbridge_lookup(const struct ofproto_dpif *);
static struct xbundle *xbundle_lookup(const struct ofbundle *);
-static struct xport *xport_lookup(struct ofport_dpif *);
+static struct xport *xport_lookup(const struct ofport_dpif *);
static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port);
+static struct skb_priority_to_dscp *get_skb_priority(const struct xport *,
+ uint32_t skb_priority);
+static void clear_skb_priorities(struct xport *);
+static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority,
+ uint8_t *dscp);
+
+static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc,
+ enum xc_type type);
void
xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name,
- const struct mac_learning *ml, const struct mbridge *mbridge,
+ struct dpif *dpif, struct rule_dpif *miss_rule,
+ struct rule_dpif *no_packet_in_rule,
+ const struct mac_learning *ml, struct stp *stp,
+ const struct mbridge *mbridge,
const struct dpif_sflow *sflow,
- const struct dpif_ipfix *ipfix, enum ofp_config_flags frag,
- bool forward_bpdu, bool has_in_band, bool has_netflow,
- bool has_stp)
+ const struct dpif_ipfix *ipfix,
+ const struct netflow *netflow, enum ofp_config_flags frag,
+ bool forward_bpdu, bool has_in_band,
+ bool enable_recirc,
+ bool variable_length_userdata,
+ size_t max_mpls_depth)
{
struct xbridge *xbridge = xbridge_lookup(ofproto);
xbridge->ipfix = dpif_ipfix_ref(ipfix);
}
+ if (xbridge->stp != stp) {
+ stp_unref(xbridge->stp);
+ xbridge->stp = stp_ref(stp);
+ }
+
+ if (xbridge->netflow != netflow) {
+ netflow_unref(xbridge->netflow);
+ xbridge->netflow = netflow_ref(netflow);
+ }
+
free(xbridge->name);
xbridge->name = xstrdup(name);
+ xbridge->dpif = dpif;
xbridge->forward_bpdu = forward_bpdu;
xbridge->has_in_band = has_in_band;
- xbridge->has_netflow = has_netflow;
- xbridge->has_stp = has_stp;
xbridge->frag = frag;
+ xbridge->miss_rule = miss_rule;
+ xbridge->no_packet_in_rule = no_packet_in_rule;
+ xbridge->enable_recirc = enable_recirc;
+ xbridge->variable_length_userdata = variable_length_userdata;
+ xbridge->max_mpls_depth = max_mpls_depth;
}
void
}
hmap_remove(&xbridges, &xbridge->hmap_node);
+ mac_learning_unref(xbridge->ml);
+ mbridge_unref(xbridge->mbridge);
+ dpif_sflow_unref(xbridge->sflow);
+ dpif_ipfix_unref(xbridge->ipfix);
+ stp_unref(xbridge->stp);
+ hmap_destroy(&xbridge->xports);
free(xbridge->name);
free(xbridge);
}
struct ofport_dpif *ofport, ofp_port_t ofp_port,
odp_port_t odp_port, const struct netdev *netdev,
const struct cfm *cfm, const struct bfd *bfd,
- struct ofport_dpif *peer, enum ofputil_port_config config,
- enum stp_state stp_state, bool is_tunnel, bool may_enable)
+ struct ofport_dpif *peer, int stp_port_no,
+ const struct ofproto_port_queue *qdscp_list, size_t n_qdscp,
+ enum ofputil_port_config config,
+ enum ofputil_port_state state, bool is_tunnel,
+ bool may_enable)
{
struct xport *xport = xport_lookup(ofport);
+ size_t i;
if (!xport) {
xport = xzalloc(sizeof *xport);
xport->xbridge = xbridge_lookup(ofproto);
xport->ofp_port = ofp_port;
+ hmap_init(&xport->skb_priorities);
hmap_insert(&xports, &xport->hmap_node, hash_pointer(ofport, 0));
hmap_insert(&xport->xbridge->xports, &xport->ofp_node,
hash_ofp_port(xport->ofp_port));
ovs_assert(xport->ofp_port == ofp_port);
xport->config = config;
- xport->stp_state = stp_state;
+ xport->state = state;
+ xport->stp_port_no = stp_port_no;
xport->is_tunnel = is_tunnel;
xport->may_enable = may_enable;
xport->odp_port = odp_port;
if (xport->peer) {
xport->peer->peer = NULL;
}
- xport->peer = peer ? xport_lookup(peer) : NULL;
+ xport->peer = xport_lookup(peer);
if (xport->peer) {
xport->peer->peer = xport;
}
if (xport->xbundle) {
list_remove(&xport->bundle_node);
}
- xport->xbundle = ofbundle ? xbundle_lookup(ofbundle) : NULL;
+ xport->xbundle = xbundle_lookup(ofbundle);
if (xport->xbundle) {
list_insert(&xport->xbundle->xports, &xport->bundle_node);
}
+
+ clear_skb_priorities(xport);
+ for (i = 0; i < n_qdscp; i++) {
+ struct skb_priority_to_dscp *pdscp;
+ uint32_t skb_priority;
+
+ if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue,
+ &skb_priority)) {
+ continue;
+ }
+
+ pdscp = xmalloc(sizeof *pdscp);
+ pdscp->skb_priority = skb_priority;
+ pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK;
+ hmap_insert(&xport->skb_priorities, &pdscp->hmap_node,
+ hash_int(pdscp->skb_priority, 0));
+ }
}
void
xport->peer = NULL;
}
- list_remove(&xport->bundle_node);
+ if (xport->xbundle) {
+ list_remove(&xport->bundle_node);
+ }
+
+ clear_skb_priorities(xport);
+ hmap_destroy(&xport->skb_priorities);
+
hmap_remove(&xports, &xport->hmap_node);
hmap_remove(&xport->xbridge->xports, &xport->ofp_node);
free(xport);
}
+/* Given a datpath, packet, and flow metadata ('backer', 'packet', and 'key'
+ * respectively), populates 'flow' with the result of odp_flow_key_to_flow().
+ * Optionally populates 'ofproto' with the ofproto_dpif, 'odp_in_port' with
+ * the datapath in_port, that 'packet' ingressed, and 'ipfix', 'sflow', and
+ * 'netflow' with the appropriate handles for those protocols if they're
+ * enabled. Caller is responsible for unrefing them.
+ *
+ * If 'ofproto' is nonnull, requires 'flow''s in_port to exist. Otherwise sets
+ * 'flow''s in_port to OFPP_NONE.
+ *
+ * This function does post-processing on data returned from
+ * odp_flow_key_to_flow() to help make VLAN splinters transparent to the rest
+ * of the upcall processing logic. In particular, if the extracted in_port is
+ * a VLAN splinter port, it replaces flow->in_port by the "real" port, sets
+ * flow->vlan_tci correctly for the VLAN of the VLAN splinter port, and pushes
+ * a VLAN header onto 'packet' (if it is nonnull).
+ *
+ * Similarly, this function also includes some logic to help with tunnels. It
+ * may modify 'flow' as necessary to make the tunneling implementation
+ * transparent to the upcall processing logic.
+ *
+ * Returns 0 if successful, ENODEV if the parsed flow has no associated ofport,
+ * or some other positive errno if there are other problems. */
+int
+xlate_receive(const struct dpif_backer *backer, struct ofpbuf *packet,
+ const struct nlattr *key, size_t key_len, struct flow *flow,
+ struct ofproto_dpif **ofproto, struct dpif_ipfix **ipfix,
+ struct dpif_sflow **sflow, struct netflow **netflow,
+ odp_port_t *odp_in_port)
+{
+ const struct xport *xport;
+ int error = ENODEV;
+
+ ovs_rwlock_rdlock(&xlate_rwlock);
+ if (odp_flow_key_to_flow(key, key_len, flow) == ODP_FIT_ERROR) {
+ error = EINVAL;
+ goto exit;
+ }
+
+ if (odp_in_port) {
+ *odp_in_port = flow->in_port.odp_port;
+ }
+
+ xport = xport_lookup(tnl_port_should_receive(flow)
+ ? tnl_port_receive(flow)
+ : odp_port_to_ofport(backer, flow->in_port.odp_port));
+
+ flow->in_port.ofp_port = xport ? xport->ofp_port : OFPP_NONE;
+ if (!xport) {
+ goto exit;
+ }
+
+ if (vsp_adjust_flow(xport->xbridge->ofproto, flow)) {
+ 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);
+ }
+ }
+ error = 0;
+
+ if (ofproto) {
+ *ofproto = xport->xbridge->ofproto;
+ }
+
+ if (ipfix) {
+ *ipfix = dpif_ipfix_ref(xport->xbridge->ipfix);
+ }
+
+ if (sflow) {
+ *sflow = dpif_sflow_ref(xport->xbridge->sflow);
+ }
+
+ if (netflow) {
+ *netflow = netflow_ref(xport->xbridge->netflow);
+ }
+
+exit:
+ ovs_rwlock_unlock(&xlate_rwlock);
+ return error;
+}
+
static struct xbridge *
xbridge_lookup(const struct ofproto_dpif *ofproto)
{
struct xbridge *xbridge;
+ if (!ofproto) {
+ return NULL;
+ }
+
HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0),
&xbridges) {
if (xbridge->ofproto == ofproto) {
{
struct xbundle *xbundle;
+ if (!ofbundle) {
+ return NULL;
+ }
+
HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0),
&xbundles) {
if (xbundle->ofbundle == ofbundle) {
}
static struct xport *
-xport_lookup(struct ofport_dpif *ofport)
+xport_lookup(const struct ofport_dpif *ofport)
{
struct xport *xport;
+ if (!ofport) {
+ return NULL;
+ }
+
HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0),
&xports) {
if (xport->ofport == ofport) {
return NULL;
}
+static struct stp_port *
+xport_get_stp_port(const struct xport *xport)
+{
+ return xport->xbridge->stp && xport->stp_port_no != -1
+ ? stp_get_port(xport->xbridge->stp, xport->stp_port_no)
+ : NULL;
+}
+
+static bool
+xport_stp_learn_state(const struct xport *xport)
+{
+ struct stp_port *sp = xport_get_stp_port(xport);
+ return stp_learn_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
+}
+
+static bool
+xport_stp_forward_state(const struct xport *xport)
+{
+ struct stp_port *sp = xport_get_stp_port(xport);
+ return stp_forward_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
+}
+
+static bool
+xport_stp_listen_state(const struct xport *xport)
+{
+ struct stp_port *sp = xport_get_stp_port(xport);
+ return stp_listen_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
+}
+
+/* Returns true if STP should process 'flow'. Sets fields in 'wc' that
+ * were used to make the determination.*/
+static bool
+stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc)
+{
+ /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */
+ memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
+ return is_stp(flow);
+}
+
+static void
+stp_process_packet(const struct xport *xport, const struct ofpbuf *packet)
+{
+ struct stp_port *sp = xport_get_stp_port(xport);
+ struct ofpbuf payload = *packet;
+ struct eth_header *eth = ofpbuf_data(&payload);
+
+ /* Sink packets on ports that have STP disabled when the bridge has
+ * STP enabled. */
+ if (!sp || stp_port_get_state(sp) == STP_DISABLED) {
+ return;
+ }
+
+ /* Trim off padding on payload. */
+ if (ofpbuf_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
+ ofpbuf_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN);
+ }
+
+ if (ofpbuf_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) {
+ stp_received_bpdu(sp, ofpbuf_data(&payload), ofpbuf_size(&payload));
+ }
+}
+
static struct xport *
get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
{
return xport ? xport->odp_port : ODPP_NONE;
}
+static bool
+odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port)
+{
+ struct xport *xport;
+
+ xport = get_ofp_port(ctx->xbridge, ofp_port);
+ if (!xport || xport->config & OFPUTIL_PC_PORT_DOWN ||
+ xport->state & OFPUTIL_PS_LINK_DOWN) {
+ return false;
+ }
+
+ return true;
+}
+
+static const struct ofputil_bucket *
+group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *,
+ int depth);
+
+static bool
+group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth)
+{
+ struct group_dpif *group;
+ bool hit;
+
+ hit = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group);
+ if (!hit) {
+ return false;
+ }
+
+ hit = group_first_live_bucket(ctx, group, depth) != NULL;
+
+ group_dpif_release(group);
+ return hit;
+}
+
+#define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */
+
+static bool
+bucket_is_alive(const struct xlate_ctx *ctx,
+ const struct ofputil_bucket *bucket, int depth)
+{
+ if (depth >= MAX_LIVENESS_RECURSION) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+
+ VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links",
+ MAX_LIVENESS_RECURSION);
+ return false;
+ }
+
+ return !ofputil_bucket_has_liveness(bucket) ||
+ (bucket->watch_port != OFPP_ANY &&
+ odp_port_is_alive(ctx, bucket->watch_port)) ||
+ (bucket->watch_group != OFPG_ANY &&
+ group_is_alive(ctx, bucket->watch_group, depth + 1));
+}
+
+static const struct ofputil_bucket *
+group_first_live_bucket(const struct xlate_ctx *ctx,
+ const struct group_dpif *group, int depth)
+{
+ struct ofputil_bucket *bucket;
+ const struct list *buckets;
+
+ group_dpif_get_buckets(group, &buckets);
+ LIST_FOR_EACH (bucket, list_node, buckets) {
+ if (bucket_is_alive(ctx, bucket, depth)) {
+ return bucket;
+ }
+ }
+
+ return NULL;
+}
+
+static const struct ofputil_bucket *
+group_best_live_bucket(const struct xlate_ctx *ctx,
+ const struct group_dpif *group,
+ uint32_t basis)
+{
+ const struct ofputil_bucket *best_bucket = NULL;
+ uint32_t best_score = 0;
+ int i = 0;
+
+ const struct ofputil_bucket *bucket;
+ const struct list *buckets;
+
+ group_dpif_get_buckets(group, &buckets);
+ LIST_FOR_EACH (bucket, list_node, buckets) {
+ if (bucket_is_alive(ctx, bucket, 0)) {
+ uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight;
+ if (score >= best_score) {
+ best_bucket = bucket;
+ best_score = score;
+ }
+ }
+ i++;
+ }
+
+ return best_bucket;
+}
+
static bool
xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan)
{
/* Special-case OFPP_NONE, which a controller may use as the ingress
* port for traffic that it is sourcing. */
if (in_port == OFPP_NONE) {
- ofpp_none_bundle.name = "OFPP_NONE";
- ofpp_none_bundle.vlan_mode = PORT_VLAN_TRUNK;
return &ofpp_none_bundle;
}
bool has_mirror;
int out_vlan;
- has_mirror = mirror_get(xbridge->mbridge, mirror_mask_ffs(mirrors) - 1,
+ has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors),
&vlans, &dup_mirrors, &out, &out_vlan);
ovs_assert(has_mirror);
return vid ? vid : in_xbundle->vlan;
default:
- NOT_REACHED();
+ OVS_NOT_REACHED();
}
}
return true;
default:
- NOT_REACHED();
+ OVS_NOT_REACHED();
}
}
return vlan == out_xbundle->vlan ? 0 : vlan;
default:
- NOT_REACHED();
+ OVS_NOT_REACHED();
}
}
/* Partially configured bundle with no slaves. Drop the packet. */
return;
} else if (!out_xbundle->bond) {
+ ctx->xout->use_recirc = false;
xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport,
bundle_node);
} else {
struct ofport_dpif *ofport;
+ struct xlate_recirc *xr = &ctx->xout->recirc;
+ struct flow_wildcards *wc = &ctx->xout->wc;
+
+ if (ctx->xbridge->enable_recirc) {
+ ctx->xout->use_recirc = bond_may_recirc(
+ out_xbundle->bond, &xr->recirc_id, &xr->hash_basis);
+
+ if (ctx->xout->use_recirc) {
+ /* Only TCP mode uses recirculation. */
+ xr->hash_alg = OVS_HASH_ALG_L4;
+ bond_update_post_recirc_rules(out_xbundle->bond, false);
- ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow,
- &ctx->xout->wc, vid,
- &ctx->xout->tags);
- xport = ofport ? xport_lookup(ofport) : NULL;
+ /* Recirculation does not require unmasking hash fields. */
+ wc = NULL;
+ }
+ }
+
+ ofport = bond_choose_output_slave(out_xbundle->bond,
+ &ctx->xin->flow, wc, vid);
+ xport = xport_lookup(ofport);
if (!xport) {
/* No slaves enabled, so drop packet. */
return;
}
+
+ /* If ctx->xout->use_recirc is set, the main thread will handle stats
+ * accounting for this bond. */
+ if (!ctx->xout->use_recirc) {
+ if (ctx->xin->resubmit_stats) {
+ bond_account(out_xbundle->bond, &ctx->xin->flow, vid,
+ ctx->xin->resubmit_stats->n_bytes);
+ }
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+ struct flow *flow;
+
+ flow = &ctx->xin->flow;
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND);
+ entry->u.bond.bond = bond_ref(out_xbundle->bond);
+ entry->u.bond.flow = xmemdup(flow, sizeof *flow);
+ entry->u.bond.vid = vid;
+ }
+ }
}
old_tci = *flow_tci;
}
}
-static void
-update_learning_table(const struct xbridge *xbridge,
- const struct flow *flow, struct flow_wildcards *wc,
- int vlan, struct xbundle *in_xbundle)
+/* Checks whether a MAC learning update is necessary for MAC learning table
+ * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in
+ * 'vlan'.
+ *
+ * Most packets processed through the MAC learning table do not actually
+ * change it in any way. This function requires only a read lock on the MAC
+ * learning table, so it is much cheaper in this common case.
+ *
+ * Keep the code here synchronized with that in update_learning_table__()
+ * below. */
+static bool
+is_mac_learning_update_needed(const struct mac_learning *ml,
+ const struct flow *flow,
+ struct flow_wildcards *wc,
+ int vlan, struct xbundle *in_xbundle)
+OVS_REQ_RDLOCK(ml->rwlock)
{
struct mac_entry *mac;
- /* Don't learn the OFPP_NONE port. */
- if (in_xbundle == &ofpp_none_bundle) {
- return;
+ if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) {
+ return false;
+ }
+
+ mac = mac_learning_lookup(ml, flow->dl_src, vlan);
+ if (!mac || mac_entry_age(ml, mac)) {
+ return true;
+ }
+
+ if (is_gratuitous_arp(flow, wc)) {
+ /* We don't want to learn from gratuitous ARP packets that are
+ * reflected back over bond slaves so we lock the learning table. */
+ if (!in_xbundle->bond) {
+ return true;
+ } else if (mac_entry_is_grat_arp_locked(mac)) {
+ return false;
+ }
}
+ return mac->port.p != in_xbundle->ofbundle;
+}
+
+
+/* Updates MAC learning table 'ml' given that a packet matching 'flow' was
+ * received on 'in_xbundle' in 'vlan'.
+ *
+ * This code repeats all the checks in is_mac_learning_update_needed() because
+ * the lock was released between there and here and thus the MAC learning state
+ * could have changed.
+ *
+ * Keep the code here synchronized with that in is_mac_learning_update_needed()
+ * above. */
+static void
+update_learning_table__(const struct xbridge *xbridge,
+ const struct flow *flow, struct flow_wildcards *wc,
+ int vlan, struct xbundle *in_xbundle)
+OVS_REQ_WRLOCK(xbridge->ml->rwlock)
+{
+ struct mac_entry *mac;
+
if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) {
return;
}
}
}
- if (mac_entry_is_new(mac) || mac->port.p != in_xbundle->ofbundle) {
+ if (mac->port.p != in_xbundle->ofbundle) {
/* 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",
xbridge->name, ETH_ADDR_ARGS(flow->dl_src),
in_xbundle->name, vlan);
mac->port.p = in_xbundle->ofbundle;
- mac_learning_changed(xbridge->ml, mac);
+ mac_learning_changed(xbridge->ml);
+ }
+}
+
+static void
+update_learning_table(const struct xbridge *xbridge,
+ const struct flow *flow, struct flow_wildcards *wc,
+ int vlan, struct xbundle *in_xbundle)
+{
+ bool need_update;
+
+ /* Don't learn the OFPP_NONE port. */
+ if (in_xbundle == &ofpp_none_bundle) {
+ return;
+ }
+
+ /* First try the common case: no change to MAC learning table. */
+ ovs_rwlock_rdlock(&xbridge->ml->rwlock);
+ need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan,
+ in_xbundle);
+ ovs_rwlock_unlock(&xbridge->ml->rwlock);
+
+ if (need_update) {
+ /* Slow path: MAC learning table might need an update. */
+ ovs_rwlock_wrlock(&xbridge->ml->rwlock);
+ update_learning_table__(xbridge, flow, wc, vlan, in_xbundle);
+ ovs_rwlock_unlock(&xbridge->ml->rwlock);
}
}
struct mac_entry *mac;
switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport,
- flow->dl_dst, &ctx->xout->tags)) {
+ flow->dl_dst)) {
case BV_ACCEPT:
break;
return false;
case BV_DROP_IF_MOVED:
- mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan, NULL);
+ ovs_rwlock_rdlock(&xbridge->ml->rwlock);
+ mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan);
if (mac && mac->port.p != in_xbundle->ofbundle &&
(!is_gratuitous_arp(flow, &ctx->xout->wc)
|| mac_entry_is_grat_arp_locked(mac))) {
+ ovs_rwlock_unlock(&xbridge->ml->rwlock);
xlate_report(ctx, "SLB bond thinks this packet looped back, "
- "dropping");
+ "dropping");
return false;
}
+ ovs_rwlock_unlock(&xbridge->ml->rwlock);
break;
}
}
struct xbundle *in_xbundle;
struct xport *in_port;
struct mac_entry *mac;
+ void *mac_port;
uint16_t vlan;
uint16_t vid;
if (ctx->xin->may_learn) {
update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle);
}
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+
+ /* Save enough info to update mac learning table later. */
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL);
+ entry->u.normal.ofproto = ctx->xin->ofproto;
+ entry->u.normal.flow = xmemdup(flow, sizeof *flow);
+ entry->u.normal.vlan = vlan;
+ }
/* Determine output bundle. */
- mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan,
- &ctx->xout->tags);
- if (mac) {
- struct xbundle *mac_xbundle = xbundle_lookup(mac->port.p);
+ ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
+ mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
+ mac_port = mac ? mac->port.p : NULL;
+ ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
+
+ if (mac_port) {
+ struct xbundle *mac_xbundle = xbundle_lookup(mac_port);
if (mac_xbundle && mac_xbundle != in_xbundle) {
xlate_report(ctx, "forwarding to learned port");
output_normal(ctx, mac_xbundle, vlan);
const size_t cookie_size)
{
size_t sample_offset, actions_offset;
+ odp_port_t odp_port;
int cookie_offset;
+ uint32_t pid;
sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE);
nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability);
actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS);
- cookie_offset = put_userspace_action(xbridge->ofproto, odp_actions, flow,
- cookie, cookie_size);
+
+ odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port);
+ pid = dpif_port_get_pid(xbridge->dpif, odp_port,
+ flow_hash_5tuple(flow, 0));
+ cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size,
+ odp_actions);
nl_msg_end_nested(odp_actions, actions_offset);
nl_msg_end_nested(odp_actions, sample_offset);
} else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) {
if (packet) {
bfd_process_packet(xport->bfd, flow, packet);
+ /* If POLL received, immediately sends FINAL back. */
+ if (bfd_should_send_packet(xport->bfd)) {
+ if (xport->peer) {
+ ofproto_dpif_monitor_port_send_soon(xport->ofport);
+ } else {
+ ofproto_dpif_monitor_port_send_soon_safe(xport->ofport);
+ }
+ }
}
return SLOW_BFD;
} else if (xport->xbundle && xport->xbundle->lacp
lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet);
}
return SLOW_LACP;
- } else if (xbridge->has_stp && stp_should_process_flow(flow, wc)) {
+ } else if (xbridge->stp && stp_should_process_flow(flow, wc)) {
if (packet) {
- stp_process_packet(xport->ofport, packet);
+ stp_process_packet(xport, packet);
}
return SLOW_STP;
} else {
struct flow_wildcards *wc = &ctx->xout->wc;
struct flow *flow = &ctx->xin->flow;
ovs_be16 flow_vlan_tci;
- uint32_t flow_skb_mark;
+ uint32_t flow_pkt_mark;
uint8_t flow_nw_tos;
odp_port_t out_port, odp_port;
uint8_t dscp;
/* If 'struct flow' gets additional metadata, we'll need to zero it out
* before traversing a patch port. */
- BUILD_ASSERT_DECL(FLOW_WC_SEQ == 20);
+ BUILD_ASSERT_DECL(FLOW_WC_SEQ == 26);
if (!xport) {
xlate_report(ctx, "Nonexistent output port");
} else if (xport->config & OFPUTIL_PC_NO_FWD) {
xlate_report(ctx, "OFPPC_NO_FWD set, skipping output");
return;
- } else if (check_stp && !stp_forward_in_state(xport->stp_state)) {
- xlate_report(ctx, "STP not in forwarding state, skipping output");
- return;
+ } else if (check_stp) {
+ if (is_stp(&ctx->base_flow)) {
+ if (!xport_stp_listen_state(xport)) {
+ xlate_report(ctx, "STP not in listening state, "
+ "skipping bpdu output");
+ return;
+ }
+ } else if (!xport_stp_forward_state(xport)) {
+ xlate_report(ctx, "STP not in forwarding state, "
+ "skipping output");
+ return;
+ }
}
if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) {
special = process_special(ctx, &ctx->xin->flow, peer,
ctx->xin->packet);
if (special) {
- ctx->xout->slow = special;
+ ctx->xout->slow |= special;
} else if (may_receive(peer, ctx)) {
- if (stp_forward_in_state(peer->stp_state)) {
- xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
+ if (xport_stp_forward_state(peer)) {
+ xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
} else {
/* Forwarding is disabled by STP. Let OFPP_NORMAL and the
* learning action look at the packet, then drop it. */
struct flow old_base_flow = ctx->base_flow;
- size_t old_size = ctx->xout->odp_actions.size;
+ size_t old_size = ofpbuf_size(&ctx->xout->odp_actions);
mirror_mask_t old_mirrors = ctx->xout->mirrors;
- xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
+ xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
ctx->xout->mirrors = old_mirrors;
ctx->base_flow = old_base_flow;
- ctx->xout->odp_actions.size = old_size;
+ ofpbuf_set_size(&ctx->xout->odp_actions, old_size);
}
}
ctx->xin->flow = old_flow;
- ctx->xbridge = xport->xbundle->xbridge;
+ ctx->xbridge = xport->xbridge;
if (ctx->xin->resubmit_stats) {
netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats);
+ if (peer->bfd) {
+ bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats);
+ }
+ }
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV);
+ entry->u.dev.tx = netdev_ref(xport->netdev);
+ entry->u.dev.rx = netdev_ref(peer->netdev);
+ entry->u.dev.bfd = bfd_ref(peer->bfd);
}
return;
}
flow_vlan_tci = flow->vlan_tci;
- flow_skb_mark = flow->skb_mark;
+ flow_pkt_mark = flow->pkt_mark;
flow_nw_tos = flow->nw_tos;
- if (ofproto_dpif_dscp_from_priority(xport->ofport, flow->skb_priority,
- &dscp)) {
- wc->masks.nw_tos |= IP_ECN_MASK;
+ if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
+ wc->masks.nw_tos |= IP_DSCP_MASK;
flow->nw_tos &= ~IP_DSCP_MASK;
flow->nw_tos |= dscp;
}
if (ctx->xin->resubmit_stats) {
netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
}
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV);
+ entry->u.dev.tx = netdev_ref(xport->netdev);
+ }
out_port = odp_port;
commit_odp_tunnel_action(flow, &ctx->base_flow,
&ctx->xout->odp_actions);
flow->tunnel = flow_tnl; /* Restore tunnel metadata */
} else {
- ofp_port_t vlandev_port;
-
odp_port = xport->odp_port;
+ out_port = odp_port;
if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) {
+ ofp_port_t vlandev_port;
+
wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
+ vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto,
+ ofp_port, flow->vlan_tci);
+ if (vlandev_port != ofp_port) {
+ out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port);
+ flow->vlan_tci = htons(0);
+ }
}
- vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, ofp_port,
- flow->vlan_tci);
- if (vlandev_port == ofp_port) {
- out_port = odp_port;
- } else {
- out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port);
- flow->vlan_tci = htons(0);
- }
- flow->skb_mark &= ~IPSEC_MARK;
}
if (out_port != ODPP_NONE) {
- commit_odp_actions(flow, &ctx->base_flow,
- &ctx->xout->odp_actions, &ctx->xout->wc);
- nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
- out_port);
+ ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
+ &ctx->xout->odp_actions,
+ &ctx->xout->wc);
+
+ if (ctx->xout->use_recirc) {
+ struct ovs_action_hash *act_hash;
+ struct xlate_recirc *xr = &ctx->xout->recirc;
+
+ /* Hash action. */
+ act_hash = nl_msg_put_unspec_uninit(&ctx->xout->odp_actions,
+ OVS_ACTION_ATTR_HASH,
+ sizeof *act_hash);
+ act_hash->hash_alg = xr->hash_alg;
+ act_hash->hash_basis = xr->hash_basis;
+
+ /* Recirc action. */
+ nl_msg_put_u32(&ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC,
+ xr->recirc_id);
+ } else {
+ nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
+ out_port);
+ }
ctx->sflow_odp_port = odp_port;
ctx->sflow_n_outputs++;
out:
/* Restore flow */
flow->vlan_tci = flow_vlan_tci;
- flow->skb_mark = flow_skb_mark;
+ flow->pkt_mark = flow_pkt_mark;
flow->nw_tos = flow_nw_tos;
}
compose_output_action__(ctx, ofp_port, true);
}
-/* Common rule processing in one place to avoid duplicating code. */
-static struct rule_dpif *
-ctx_rule_hooks(struct xlate_ctx *ctx, struct rule_dpif *rule,
- bool may_packet_in)
+static void
+xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule)
{
- if (ctx->xin->resubmit_hook) {
- ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
- }
- if (rule == NULL && may_packet_in) {
- /* XXX
- * check if table configuration flags
- * OFPTC_TABLE_MISS_CONTROLLER, default.
- * OFPTC_TABLE_MISS_CONTINUE,
- * OFPTC_TABLE_MISS_DROP
- * When OF1.0, OFPTC_TABLE_MISS_CONTINUE is used. What to do?
- */
- rule = rule_dpif_miss_rule(ctx->xbridge->ofproto, &ctx->xin->flow);
+ struct rule_dpif *old_rule = ctx->rule;
+ const struct rule_actions *actions;
+
+ if (ctx->xin->resubmit_stats) {
+ rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats);
}
- if (rule && ctx->xin->resubmit_stats) {
- rule_credit_stats(rule, ctx->xin->resubmit_stats);
+
+ ctx->resubmits++;
+ ctx->recurse++;
+ ctx->rule = rule;
+ actions = rule_dpif_get_actions(rule);
+ do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx);
+ ctx->rule = old_rule;
+ ctx->recurse--;
+}
+
+static bool
+xlate_resubmit_resource_check(struct xlate_ctx *ctx)
+{
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+
+ if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) {
+ VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times",
+ MAX_RESUBMIT_RECURSION);
+ } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) {
+ VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS);
+ } else if (ofpbuf_size(&ctx->xout->odp_actions) > UINT16_MAX) {
+ VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions");
+ } else if (ofpbuf_size(&ctx->stack) >= 65536) {
+ VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack");
+ } else {
+ return true;
}
- return rule;
+
+ return false;
}
static void
-xlate_table_action(struct xlate_ctx *ctx,
- ofp_port_t in_port, uint8_t table_id, bool may_packet_in)
+xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id,
+ bool may_packet_in, bool honor_table_miss)
{
- if (ctx->recurse < MAX_RESUBMIT_RECURSION) {
- struct rule_dpif *rule;
+ if (xlate_resubmit_resource_check(ctx)) {
ofp_port_t old_in_port = ctx->xin->flow.in_port.ofp_port;
+ bool skip_wildcards = ctx->xin->skip_wildcards;
uint8_t old_table_id = ctx->table_id;
+ struct rule_dpif *rule;
+ enum rule_dpif_lookup_verdict verdict;
+ enum ofputil_port_config config = 0;
ctx->table_id = table_id;
- /* Look up a flow with 'in_port' as the input port. */
+ /* 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). */
ctx->xin->flow.in_port.ofp_port = in_port;
- rule = rule_dpif_lookup_in_table(ctx->xbridge->ofproto,
- &ctx->xin->flow, &ctx->xout->wc,
- table_id);
-
- ctx->xout->tags |= calculate_flow_tag(ctx->xbridge->ofproto,
- &ctx->xin->flow, ctx->table_id,
- rule);
-
- /* Restore the original input port. Otherwise OFPP_NORMAL and
- * OFPP_IN_PORT will have surprising behavior. */
+ verdict = rule_dpif_lookup_from_table(ctx->xbridge->ofproto,
+ &ctx->xin->flow,
+ !skip_wildcards
+ ? &ctx->xout->wc : NULL,
+ honor_table_miss,
+ &ctx->table_id, &rule,
+ ctx->xin->xcache != NULL);
ctx->xin->flow.in_port.ofp_port = old_in_port;
- rule = ctx_rule_hooks(ctx, rule, may_packet_in);
+ if (ctx->xin->resubmit_hook) {
+ ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
+ }
- if (rule) {
- struct rule_dpif *old_rule = ctx->rule;
+ switch (verdict) {
+ case RULE_DPIF_LOOKUP_VERDICT_MATCH:
+ goto match;
+ case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER:
+ if (may_packet_in) {
+ struct xport *xport;
+
+ xport = get_ofp_port(ctx->xbridge,
+ ctx->xin->flow.in_port.ofp_port);
+ config = xport ? xport->config : 0;
+ break;
+ }
+ /* Fall through to drop */
+ case RULE_DPIF_LOOKUP_VERDICT_DROP:
+ config = OFPUTIL_PC_NO_PACKET_IN;
+ break;
+ case RULE_DPIF_LOOKUP_VERDICT_DEFAULT:
+ if (!ofproto_dpif_wants_packet_in_on_miss(ctx->xbridge->ofproto)) {
+ config = OFPUTIL_PC_NO_PACKET_IN;
+ }
+ break;
+ default:
+ OVS_NOT_REACHED();
+ }
- ctx->recurse++;
- ctx->rule = rule;
- do_xlate_actions(rule->up.ofpacts, rule->up.ofpacts_len, ctx);
- ctx->rule = old_rule;
- ctx->recurse--;
+ choose_miss_rule(config, ctx->xbridge->miss_rule,
+ ctx->xbridge->no_packet_in_rule, &rule,
+ ctx->xin->xcache != NULL);
+
+match:
+ if (rule) {
+ /* Fill in the cache entry here instead of xlate_recursively
+ * to make the reference counting more explicit. We take a
+ * reference in the lookups above if we are going to cache the
+ * rule. */
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE);
+ entry->u.rule = rule;
+ }
+ xlate_recursively(ctx, rule);
}
ctx->table_id = old_table_id;
+ return;
+ }
+
+ ctx->exit = true;
+}
+
+static void
+xlate_group_bucket(struct xlate_ctx *ctx, const struct ofputil_bucket *bucket)
+{
+ uint64_t action_list_stub[1024 / 8];
+ struct ofpbuf action_list, action_set;
+
+ ofpbuf_use_const(&action_set, bucket->ofpacts, bucket->ofpacts_len);
+ ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub);
+
+ ofpacts_execute_action_set(&action_list, &action_set);
+ ctx->recurse++;
+ do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
+ ctx->recurse--;
+
+ ofpbuf_uninit(&action_set);
+ ofpbuf_uninit(&action_list);
+}
+
+static void
+xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group)
+{
+ const struct ofputil_bucket *bucket;
+ const struct list *buckets;
+ struct flow old_flow = ctx->xin->flow;
+
+ group_dpif_get_buckets(group, &buckets);
+
+ LIST_FOR_EACH (bucket, list_node, buckets) {
+ xlate_group_bucket(ctx, bucket);
+ /* Roll back flow to previous state.
+ * This is equivalent to cloning the packet for each bucket.
+ *
+ * As a side effect any subsequently applied actions will
+ * also effectively be applied to a clone of the packet taken
+ * just before applying the all or indirect group. */
+ ctx->xin->flow = old_flow;
+ }
+}
+
+static void
+xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group)
+{
+ const struct ofputil_bucket *bucket;
+
+ bucket = group_first_live_bucket(ctx, group, 0);
+ if (bucket) {
+ xlate_group_bucket(ctx, bucket);
+ }
+}
+
+static void
+xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group)
+{
+ struct flow_wildcards *wc = &ctx->xout->wc;
+ const struct ofputil_bucket *bucket;
+ uint32_t basis;
+
+ basis = hash_mac(ctx->xin->flow.dl_dst, 0, 0);
+ bucket = group_best_live_bucket(ctx, group, basis);
+ if (bucket) {
+ memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
+ xlate_group_bucket(ctx, bucket);
+ }
+}
+
+static void
+xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group)
+{
+ ctx->in_group = true;
+
+ switch (group_dpif_get_type(group)) {
+ case OFPGT11_ALL:
+ case OFPGT11_INDIRECT:
+ xlate_all_group(ctx, group);
+ break;
+ case OFPGT11_SELECT:
+ xlate_select_group(ctx, group);
+ break;
+ case OFPGT11_FF:
+ xlate_ff_group(ctx, group);
+ break;
+ default:
+ OVS_NOT_REACHED();
+ }
+ group_dpif_release(group);
+
+ ctx->in_group = false;
+}
+
+static bool
+xlate_group_resource_check(struct xlate_ctx *ctx)
+{
+ if (!xlate_resubmit_resource_check(ctx)) {
+ return false;
+ } else if (ctx->in_group) {
+ /* Prevent nested translation of OpenFlow groups.
+ *
+ * OpenFlow allows this restriction. We enforce this restriction only
+ * because, with the current architecture, we would otherwise have to
+ * take a possibly recursive read lock on the ofgroup rwlock, which is
+ * unsafe given that POSIX allows taking a read lock to block if there
+ * is a thread blocked on taking the write lock. Other solutions
+ * without this restriction are also possible, but seem unwarranted
+ * given the current limited use of groups. */
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+
+ VLOG_ERR_RL(&rl, "cannot recursively translate OpenFlow group");
+ return false;
} else {
- static struct vlog_rate_limit recurse_rl = VLOG_RATE_LIMIT_INIT(1, 1);
+ return true;
+ }
+}
- VLOG_ERR_RL(&recurse_rl, "resubmit actions recursed over %d times",
- MAX_RESUBMIT_RECURSION);
- ctx->max_resubmit_trigger = true;
+static bool
+xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id)
+{
+ if (xlate_group_resource_check(ctx)) {
+ struct group_dpif *group;
+ bool got_group;
+
+ got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group);
+ if (got_group) {
+ xlate_group_action__(ctx, group);
+ } else {
+ return true;
+ }
}
+
+ return false;
}
static void
{
ofp_port_t in_port;
uint8_t table_id;
+ bool may_packet_in = false;
+ bool honor_table_miss = false;
+
+ if (ctx->rule && rule_dpif_is_internal(ctx->rule)) {
+ /* Still allow missed packets to be sent to the controller
+ * if resubmitting from an internal table. */
+ may_packet_in = true;
+ honor_table_miss = true;
+ }
in_port = resubmit->in_port;
if (in_port == OFPP_IN_PORT) {
table_id = ctx->table_id;
}
- xlate_table_action(ctx, in_port, table_id, false);
+ xlate_table_action(ctx, in_port, table_id, may_packet_in,
+ honor_table_miss);
}
static void
enum ofp_packet_in_reason reason,
uint16_t controller_id)
{
- struct ofputil_packet_in pin;
+ struct ofproto_packet_in *pin;
struct ofpbuf *packet;
- struct flow key;
+ struct pkt_metadata md = PKT_METADATA_INITIALIZER(0);
- ovs_assert(!ctx->xout->slow || ctx->xout->slow == SLOW_CONTROLLER);
- ctx->xout->slow = SLOW_CONTROLLER;
+ ctx->xout->slow |= SLOW_CONTROLLER;
if (!ctx->xin->packet) {
return;
}
packet = ofpbuf_clone(ctx->xin->packet);
- key.skb_priority = 0;
- key.skb_mark = 0;
- memset(&key.tunnel, 0, sizeof key.tunnel);
-
- commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
- &ctx->xout->odp_actions, &ctx->xout->wc);
-
- odp_execute_actions(NULL, packet, &key, ctx->xout->odp_actions.data,
- ctx->xout->odp_actions.size, NULL, NULL);
-
- pin.packet = packet->data;
- pin.packet_len = packet->size;
- pin.reason = reason;
- pin.controller_id = controller_id;
- pin.table_id = ctx->table_id;
- pin.cookie = ctx->rule ? ctx->rule->up.flow_cookie : 0;
-
- pin.send_len = len;
- flow_get_metadata(&ctx->xin->flow, &pin.fmd);
-
- ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, &pin);
+ ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
+ &ctx->xout->odp_actions,
+ &ctx->xout->wc);
+
+ odp_execute_actions(NULL, packet, false, &md,
+ ofpbuf_data(&ctx->xout->odp_actions),
+ ofpbuf_size(&ctx->xout->odp_actions), NULL);
+
+ pin = xmalloc(sizeof *pin);
+ pin->up.packet_len = ofpbuf_size(packet);
+ pin->up.packet = ofpbuf_steal_data(packet);
+ pin->up.reason = reason;
+ pin->up.table_id = ctx->table_id;
+ pin->up.cookie = (ctx->rule
+ ? rule_dpif_get_flow_cookie(ctx->rule)
+ : OVS_BE64_MAX);
+
+ flow_get_metadata(&ctx->xin->flow, &pin->up.fmd);
+
+ pin->controller_id = controller_id;
+ pin->send_len = len;
+ /* If a rule is a table-miss rule then this is
+ * a table-miss handled by a table-miss rule.
+ *
+ * Else, if rule is internal and has a controller action,
+ * the later being implied by the rule being processed here,
+ * then this is a table-miss handled without a table-miss rule.
+ *
+ * Otherwise this is not a table-miss. */
+ pin->miss_type = OFPROTO_PACKET_IN_NO_MISS;
+ if (ctx->rule) {
+ if (rule_dpif_is_table_miss(ctx->rule)) {
+ pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW;
+ } else if (rule_dpif_is_internal(ctx->rule)) {
+ pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW;
+ }
+ }
+ ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin);
ofpbuf_delete(packet);
}
static void
-compose_mpls_push_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
+compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls)
{
struct flow_wildcards *wc = &ctx->xout->wc;
struct flow *flow = &ctx->xin->flow;
+ int n;
- ovs_assert(eth_type_mpls(eth_type));
+ ovs_assert(eth_type_mpls(mpls->ethertype));
- memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
- memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
-
- if (flow->mpls_depth) {
- flow->mpls_lse &= ~htonl(MPLS_BOS_MASK);
- flow->mpls_depth++;
- } else {
- ovs_be32 label;
- uint8_t tc, ttl;
-
- if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
- label = htonl(0x2); /* IPV6 Explicit Null. */
- } else {
- label = htonl(0x0); /* IPV4 Explicit Null. */
+ n = flow_count_mpls_labels(flow, wc);
+ if (!n) {
+ ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
+ &ctx->xout->odp_actions,
+ &ctx->xout->wc);
+ } else if (n >= FLOW_MAX_MPLS_LABELS) {
+ if (ctx->xin->packet != NULL) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an "
+ "MPLS push action can't be performed as it would "
+ "have more MPLS LSEs than the %d supported.",
+ ctx->xbridge->name, FLOW_MAX_MPLS_LABELS);
}
- wc->masks.nw_tos |= IP_DSCP_MASK;
- wc->masks.nw_ttl = 0xff;
- tc = (flow->nw_tos & IP_DSCP_MASK) >> 2;
- ttl = flow->nw_ttl ? flow->nw_ttl : 0x40;
- flow->mpls_lse = set_mpls_lse_values(ttl, tc, 1, label);
- flow->mpls_depth = 1;
+ ctx->exit = true;
+ return;
+ } else if (n >= ctx->xbridge->max_mpls_depth) {
+ COVERAGE_INC(xlate_actions_mpls_overflow);
+ ctx->xout->slow |= SLOW_ACTION;
}
- flow->dl_type = eth_type;
+
+ flow_push_mpls(flow, n, mpls->ethertype, wc);
}
static void
{
struct flow_wildcards *wc = &ctx->xout->wc;
struct flow *flow = &ctx->xin->flow;
+ int n = flow_count_mpls_labels(flow, wc);
- ovs_assert(eth_type_mpls(ctx->xin->flow.dl_type));
- ovs_assert(!eth_type_mpls(eth_type));
-
- memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
- memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
-
- if (flow->mpls_depth) {
- flow->mpls_depth--;
- flow->mpls_lse = htonl(0);
- if (!flow->mpls_depth) {
- flow->dl_type = eth_type;
+ if (!flow_pop_mpls(flow, n, eth_type, wc) && n >= FLOW_MAX_MPLS_LABELS) {
+ if (ctx->xin->packet != NULL) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an "
+ "MPLS pop action can't be performed as it has "
+ "more MPLS LSEs than the %d supported.",
+ ctx->xbridge->name, FLOW_MAX_MPLS_LABELS);
}
+ ctx->exit = true;
+ ofpbuf_clear(&ctx->xout->odp_actions);
}
}
}
}
-static bool
-compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl)
+static void
+compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label)
{
- if (!eth_type_mpls(ctx->xin->flow.dl_type)) {
- return true;
+ if (eth_type_mpls(ctx->xin->flow.dl_type)) {
+ ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK);
+ set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label);
}
+}
- set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse, ttl);
- return false;
+static void
+compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc)
+{
+ if (eth_type_mpls(ctx->xin->flow.dl_type)) {
+ ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TC_MASK);
+ set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc);
+ }
+}
+
+static void
+compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl)
+{
+ if (eth_type_mpls(ctx->xin->flow.dl_type)) {
+ ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK);
+ set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl);
+ }
}
static bool
compose_dec_mpls_ttl_action(struct xlate_ctx *ctx)
{
struct flow *flow = &ctx->xin->flow;
- uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse);
+ uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]);
struct flow_wildcards *wc = &ctx->xout->wc;
memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
+ if (eth_type_mpls(flow->dl_type)) {
+ if (ttl > 1) {
+ ttl--;
+ set_mpls_lse_ttl(&flow->mpls_lse[0], ttl);
+ return false;
+ } else {
+ execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
- if (!eth_type_mpls(flow->dl_type)) {
- return false;
- }
-
- if (ttl > 1) {
- ttl--;
- set_mpls_lse_ttl(&flow->mpls_lse, ttl);
- return false;
+ /* Stop processing for current table. */
+ return true;
+ }
} else {
- execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
-
- /* Stop processing for current table. */
return true;
}
}
break;
case OFPP_TABLE:
xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
- 0, may_packet_in);
+ 0, may_packet_in, true);
break;
case OFPP_NORMAL:
xlate_normal(ctx);
int error;
/* Translate queue to priority. */
- error = ofproto_dpif_queue_to_priority(ctx->xbridge->ofproto, queue_id,
- &priority);
+ error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority);
if (error) {
/* Fall back to ordinary output action. */
xlate_output_action(ctx, enqueue->port, 0, false);
{
uint32_t skb_priority;
- if (!ofproto_dpif_queue_to_priority(ctx->xbridge->ofproto, queue_id,
- &skb_priority)) {
+ if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) {
ctx->xin->flow.skb_priority = skb_priority;
} else {
/* Couldn't translate queue to a priority. Nothing to do. A warning
slave_enabled_cb,
CONST_CAST(struct xbridge *, ctx->xbridge));
if (bundle->dst.field) {
- nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow);
+ nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow,
+ &ctx->xout->wc);
} else {
xlate_output_action(ctx, port, 0, false);
}
xlate_learn_action(struct xlate_ctx *ctx,
const struct ofpact_learn *learn)
{
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 1);
- struct ofputil_flow_mod fm;
uint64_t ofpacts_stub[1024 / 8];
+ struct ofputil_flow_mod fm;
struct ofpbuf ofpacts;
- int error;
ctx->xout->has_learn = true;
return;
}
- ofpbuf_use_stack(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
+ ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
learn_execute(learn, &ctx->xin->flow, &fm, &ofpacts);
+ ofproto_dpif_flow_mod(ctx->xbridge->ofproto, &fm);
+ ofpbuf_uninit(&ofpacts);
- error = ofproto_dpif_flow_mod(ctx->xbridge->ofproto, &fm);
- if (error && !VLOG_DROP_WARN(&rl)) {
- VLOG_WARN("learning action failed to modify flow table (%s)",
- ofperr_get_name(error));
- }
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
- ofpbuf_uninit(&ofpacts);
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN);
+ entry->u.learn.ofproto = ctx->xin->ofproto;
+ /* Lookup the learned rule, taking a reference on it. The reference
+ * is released when this cache entry is deleted. */
+ rule_dpif_lookup(ctx->xbridge->ofproto, &ctx->xin->flow, NULL,
+ &entry->u.learn.rule, true);
+ }
}
-/* Reduces '*timeout' to no more than 'max'. A value of zero in either case
- * means "infinite". */
static void
-reduce_timeout(uint16_t max, uint16_t *timeout)
+xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags,
+ uint16_t idle_timeout, uint16_t hard_timeout)
{
- if (max && (!*timeout || *timeout > max)) {
- *timeout = max;
+ if (tcp_flags & (TCP_FIN | TCP_RST)) {
+ rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout);
}
}
xlate_fin_timeout(struct xlate_ctx *ctx,
const struct ofpact_fin_timeout *oft)
{
- if (ctx->xin->tcp_flags & (TCP_FIN | TCP_RST) && ctx->rule) {
- struct rule_dpif *rule = ctx->rule;
-
- if (list_is_empty(&rule->up.expirable)) {
- list_insert(&rule->up.ofproto->expirable, &rule->up.expirable);
- }
-
- reduce_timeout(oft->fin_idle_timeout, &rule->up.idle_timeout);
- reduce_timeout(oft->fin_hard_timeout, &rule->up.hard_timeout);
+ if (ctx->rule) {
+ xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags,
+ oft->fin_idle_timeout, oft->fin_hard_timeout);
+ if (ctx->xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT);
+ /* XC_RULE already holds a reference on the rule, none is taken
+ * here. */
+ entry->u.fin.rule = ctx->rule;
+ entry->u.fin.idle = oft->fin_idle_timeout;
+ entry->u.fin.hard = oft->fin_hard_timeout;
+ }
}
}
* the same percentage. */
uint32_t probability = (os->probability << 16) | os->probability;
- commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
- &ctx->xout->odp_actions, &ctx->xout->wc);
+ if (!ctx->xbridge->variable_length_userdata) {
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+
+ VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath "
+ "lacks support (needs Linux 3.10+ or kernel module from "
+ "OVS 1.11+)");
+ return;
+ }
+
+ ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
+ &ctx->xout->odp_actions,
+ &ctx->xout->wc);
compose_flow_sample_cookie(os->probability, os->collector_set_id,
os->obs_domain_id, os->obs_point_id, &cookie);
static bool
may_receive(const struct xport *xport, struct xlate_ctx *ctx)
{
- if (xport->config & (eth_addr_equals(ctx->xin->flow.dl_dst, eth_addr_stp)
+ if (xport->config & (is_stp(&ctx->xin->flow)
? OFPUTIL_PC_NO_RECV_STP
: OFPUTIL_PC_NO_RECV)) {
return false;
* disabled. If just learning is enabled, we need to have
* OFPP_NORMAL and the learning action have a look at the packet
* before we can drop it. */
- if (!stp_forward_in_state(xport->stp_state)
- && !stp_learn_in_state(xport->stp_state)) {
+ if (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) {
return false;
}
return true;
}
-static bool
-tunnel_ecn_ok(struct xlate_ctx *ctx)
+static void
+xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a)
{
- if (is_ip_any(&ctx->base_flow)
- && (ctx->xin->flow.tunnel.ip_tos & IP_ECN_MASK) == IP_ECN_CE) {
- if ((ctx->base_flow.nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
- VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
- " but is not ECN capable");
- return false;
- } else {
- /* Set the ECN CE value in the tunneled packet. */
- ctx->xin->flow.nw_tos |= IP_ECN_CE;
- }
- }
+ struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a);
+ ofpbuf_put(&ctx->action_set, on->actions, ofpact_nest_get_action_len(on));
+ ofpact_pad(&ctx->action_set);
+}
- return true;
+static void
+xlate_action_set(struct xlate_ctx *ctx)
+{
+ uint64_t action_list_stub[1024 / 64];
+ struct ofpbuf action_list;
+
+ ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub);
+ ofpacts_execute_action_set(&action_list, &ctx->action_set);
+ do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
+ ofpbuf_uninit(&action_list);
}
static void
{
struct flow_wildcards *wc = &ctx->xout->wc;
struct flow *flow = &ctx->xin->flow;
- bool was_evictable = true;
const struct ofpact *a;
- if (ctx->rule) {
- /* Don't let the rule we're working on get evicted underneath us. */
- was_evictable = ctx->rule->up.evictable;
- ctx->rule->up.evictable = false;
- }
+ /* dl_type already in the mask, not set below. */
- do_xlate_actions_again:
OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
struct ofpact_controller *controller;
const struct ofpact_metadata *metadata;
+ const struct ofpact_set_field *set_field;
+ const struct mf_field *mf;
if (ctx->exit) {
break;
ofpact_get_OUTPUT(a)->max_len, true);
break;
+ case OFPACT_GROUP:
+ if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) {
+ return;
+ }
+ break;
+
case OFPACT_CONTROLLER:
controller = ofpact_get_CONTROLLER(a);
execute_controller_action(ctx, controller->max_len,
break;
case OFPACT_SET_VLAN_VID:
- flow->vlan_tci &= ~htons(VLAN_VID_MASK);
- flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid)
- | htons(VLAN_CFI));
+ wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
+ if (flow->vlan_tci & htons(VLAN_CFI) ||
+ ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) {
+ flow->vlan_tci &= ~htons(VLAN_VID_MASK);
+ flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid)
+ | htons(VLAN_CFI));
+ }
break;
case OFPACT_SET_VLAN_PCP:
- flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
- flow->vlan_tci |=
- htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp << VLAN_PCP_SHIFT)
- | VLAN_CFI);
+ wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI);
+ if (flow->vlan_tci & htons(VLAN_CFI) ||
+ ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) {
+ flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
+ flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp
+ << VLAN_PCP_SHIFT) | VLAN_CFI);
+ }
break;
case OFPACT_STRIP_VLAN:
+ memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
flow->vlan_tci = htons(0);
break;
case OFPACT_PUSH_VLAN:
/* XXX 802.1AD(QinQ) */
+ memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
flow->vlan_tci = htons(VLAN_CFI);
break;
case OFPACT_SET_ETH_SRC:
+ memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN);
break;
case OFPACT_SET_ETH_DST:
+ memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN);
break;
case OFPACT_SET_IPV4_SRC:
if (flow->dl_type == htons(ETH_TYPE_IP)) {
+ memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4;
}
break;
case OFPACT_SET_IPV4_DST:
if (flow->dl_type == htons(ETH_TYPE_IP)) {
+ memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4;
}
break;
- case OFPACT_SET_IPV4_DSCP:
- /* OpenFlow 1.0 only supports IPv4. */
- if (flow->dl_type == htons(ETH_TYPE_IP)) {
+ case OFPACT_SET_IP_DSCP:
+ if (is_ip_any(flow)) {
+ wc->masks.nw_tos |= IP_DSCP_MASK;
flow->nw_tos &= ~IP_DSCP_MASK;
- flow->nw_tos |= ofpact_get_SET_IPV4_DSCP(a)->dscp;
+ flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp;
+ }
+ break;
+
+ case OFPACT_SET_IP_ECN:
+ if (is_ip_any(flow)) {
+ wc->masks.nw_tos |= IP_ECN_MASK;
+ flow->nw_tos &= ~IP_ECN_MASK;
+ flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn;
+ }
+ break;
+
+ case OFPACT_SET_IP_TTL:
+ if (is_ip_any(flow)) {
+ wc->masks.nw_ttl = 0xff;
+ flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl;
}
break;
case OFPACT_SET_L4_SRC_PORT:
- memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
if (is_ip_any(flow)) {
+ memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
+ memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src);
flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port);
}
break;
case OFPACT_SET_L4_DST_PORT:
- memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
if (is_ip_any(flow)) {
+ memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
+ memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port);
}
break;
break;
case OFPACT_REG_LOAD:
- nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow);
+ nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow, wc);
+ break;
+
+ case OFPACT_SET_FIELD:
+ set_field = ofpact_get_SET_FIELD(a);
+ mf = set_field->field;
+
+ /* Set field action only ever overwrites packet's outermost
+ * applicable header fields. Do nothing if no header exists. */
+ if (mf->id == MFF_VLAN_VID) {
+ wc->masks.vlan_tci |= htons(VLAN_CFI);
+ if (!(flow->vlan_tci & htons(VLAN_CFI))) {
+ break;
+ }
+ } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC)
+ /* 'dl_type' is already unwildcarded. */
+ && !eth_type_mpls(flow->dl_type)) {
+ break;
+ }
+
+ mf_mask_field_and_prereqs(mf, &wc->masks);
+ mf_set_flow_value(mf, &set_field->value, flow);
break;
case OFPACT_STACK_PUSH:
break;
case OFPACT_STACK_POP:
- nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, &ctx->stack);
+ nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc,
+ &ctx->stack);
break;
case OFPACT_PUSH_MPLS:
- compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)->ethertype);
+ compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a));
break;
case OFPACT_POP_MPLS:
compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype);
break;
+ case OFPACT_SET_MPLS_LABEL:
+ compose_set_mpls_label_action(
+ ctx, ofpact_get_SET_MPLS_LABEL(a)->label);
+ break;
+
+ case OFPACT_SET_MPLS_TC:
+ compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc);
+ break;
+
case OFPACT_SET_MPLS_TTL:
- if (compose_set_mpls_ttl_action(ctx,
- ofpact_get_SET_MPLS_TTL(a)->ttl)) {
- goto out;
- }
+ compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl);
break;
case OFPACT_DEC_MPLS_TTL:
if (compose_dec_mpls_ttl_action(ctx)) {
- goto out;
+ return;
}
break;
case OFPACT_DEC_TTL:
+ wc->masks.nw_ttl = 0xff;
if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) {
- goto out;
+ return;
}
break;
break;
case OFPACT_CLEAR_ACTIONS:
- /* XXX
- * Nothing to do because writa-actions is not supported for now.
- * When writa-actions is supported, clear-actions also must
- * be supported at the same time.
- */
+ ofpbuf_clear(&ctx->action_set);
+ break;
+
+ case OFPACT_WRITE_ACTIONS:
+ xlate_write_actions(ctx, a);
break;
case OFPACT_WRITE_METADATA:
break;
case OFPACT_GOTO_TABLE: {
- /* It is assumed that goto-table is the last action. */
struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
- struct rule_dpif *rule;
ovs_assert(ctx->table_id < ogt->table_id);
-
- ctx->table_id = ogt->table_id;
-
- /* Look up a flow from the new table. */
- rule = rule_dpif_lookup_in_table(ctx->xbridge->ofproto, flow, wc,
- ctx->table_id);
-
- ctx->xout->tags |= calculate_flow_tag(ctx->xbridge->ofproto,
- &ctx->xin->flow,
- ctx->table_id, rule);
-
- rule = ctx_rule_hooks(ctx, rule, true);
-
- if (rule) {
- if (ctx->rule) {
- ctx->rule->up.evictable = was_evictable;
- }
- ctx->rule = rule;
- was_evictable = rule->up.evictable;
- rule->up.evictable = false;
-
- /* Tail recursion removal. */
- ofpacts = rule->up.ofpacts;
- ofpacts_len = rule->up.ofpacts_len;
- goto do_xlate_actions_again;
- }
+ xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
+ ogt->table_id, true, true);
break;
}
break;
}
}
-
-out:
- if (ctx->rule) {
- ctx->rule->up.evictable = was_evictable;
- }
}
void
xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto,
const struct flow *flow, struct rule_dpif *rule,
- uint8_t tcp_flags, const struct ofpbuf *packet)
+ uint16_t tcp_flags, const struct ofpbuf *packet)
{
xin->ofproto = ofproto;
xin->flow = *flow;
xin->packet = packet;
xin->may_learn = packet != NULL;
xin->rule = rule;
+ xin->xcache = NULL;
xin->ofpacts = NULL;
xin->ofpacts_len = 0;
xin->tcp_flags = tcp_flags;
xin->resubmit_hook = NULL;
xin->report_hook = NULL;
xin->resubmit_stats = NULL;
+ xin->skip_wildcards = false;
}
void
xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src)
{
dst->wc = src->wc;
- dst->tags = src->tags;
dst->slow = src->slow;
dst->has_learn = src->has_learn;
dst->has_normal = src->has_normal;
ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
sizeof dst->odp_actions_stub);
- ofpbuf_put(&dst->odp_actions, src->odp_actions.data,
- src->odp_actions.size);
+ ofpbuf_put(&dst->odp_actions, ofpbuf_data(&src->odp_actions),
+ ofpbuf_size(&src->odp_actions));
}
\f
+static struct skb_priority_to_dscp *
+get_skb_priority(const struct xport *xport, uint32_t skb_priority)
+{
+ struct skb_priority_to_dscp *pdscp;
+ uint32_t hash;
+
+ hash = hash_int(skb_priority, 0);
+ HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) {
+ if (pdscp->skb_priority == skb_priority) {
+ return pdscp;
+ }
+ }
+ return NULL;
+}
+
+static bool
+dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority,
+ uint8_t *dscp)
+{
+ struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority);
+ *dscp = pdscp ? pdscp->dscp : 0;
+ return pdscp != NULL;
+}
+
+static void
+clear_skb_priorities(struct xport *xport)
+{
+ struct skb_priority_to_dscp *pdscp, *next;
+
+ HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) {
+ hmap_remove(&xport->skb_priorities, &pdscp->hmap_node);
+ free(pdscp);
+ }
+}
+
static bool
actions_output_to_local_port(const struct xlate_ctx *ctx)
{
const struct nlattr *a;
unsigned int left;
- NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions.data,
- ctx->xout->odp_actions.size) {
+ NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(&ctx->xout->odp_actions),
+ ofpbuf_size(&ctx->xout->odp_actions)) {
if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT
&& nl_attr_get_odp_port(a) == local_odp_port) {
return true;
return false;
}
-/* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
- * into datapath actions in 'odp_actions', using 'ctx'. */
+/* Thread safe call to xlate_actions__(). */
void
xlate_actions(struct xlate_in *xin, struct xlate_out *xout)
+ OVS_EXCLUDED(xlate_rwlock)
{
- /* Normally false. Set to true if we ever hit MAX_RESUBMIT_RECURSION, so
- * that in the future we always keep a copy of the original flow for
- * tracing purposes. */
- static bool hit_resubmit_limit;
+ ovs_rwlock_rdlock(&xlate_rwlock);
+ xlate_actions__(xin, xout);
+ ovs_rwlock_unlock(&xlate_rwlock);
+}
+/* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
+ * into datapath actions in 'odp_actions', using 'ctx'.
+ *
+ * The caller must take responsibility for eventually freeing 'xout', with
+ * xlate_out_uninit(). */
+static void
+xlate_actions__(struct xlate_in *xin, struct xlate_out *xout)
+ OVS_REQ_RDLOCK(xlate_rwlock)
+{
struct flow_wildcards *wc = &xout->wc;
struct flow *flow = &xin->flow;
+ struct rule_dpif *rule = NULL;
+ const struct rule_actions *actions = NULL;
enum slow_path_reason special;
const struct ofpact *ofpacts;
struct xport *in_port;
struct flow orig_flow;
struct xlate_ctx ctx;
size_t ofpacts_len;
+ bool tnl_may_send;
+ bool is_icmp;
COVERAGE_INC(xlate_actions);
ctx.xin = xin;
ctx.xout = xout;
- ctx.xout->tags = 0;
ctx.xout->slow = 0;
ctx.xout->has_learn = false;
ctx.xout->has_normal = false;
memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port);
memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
- wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
-
- if (tnl_port_should_receive(&ctx.xin->flow)) {
- memset(&wc->masks.tunnel, 0xff, sizeof wc->masks.tunnel);
+ if (is_ip_any(flow)) {
+ wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
}
- if (ctx.xbridge->has_netflow) {
+ is_icmp = is_icmpv4(flow) || is_icmpv6(flow);
+
+ tnl_may_send = tnl_xlate_init(&ctx.base_flow, flow, wc);
+ if (ctx.xbridge->netflow) {
netflow_mask_wc(flow, wc);
}
ctx.recurse = 0;
- ctx.max_resubmit_trigger = false;
+ ctx.resubmits = 0;
+ ctx.in_group = false;
ctx.orig_skb_priority = flow->skb_priority;
ctx.table_id = 0;
ctx.exit = false;
+ if (!xin->ofpacts && !ctx.rule) {
+ ctx.table_id = rule_dpif_lookup(ctx.xbridge->ofproto, flow,
+ !xin->skip_wildcards ? wc : NULL,
+ &rule, ctx.xin->xcache != NULL);
+ if (ctx.xin->resubmit_stats) {
+ rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats);
+ }
+ if (ctx.xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE);
+ entry->u.rule = rule;
+ }
+ ctx.rule = rule;
+ }
+ xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule);
+ xout->use_recirc = false;
+
if (xin->ofpacts) {
ofpacts = xin->ofpacts;
ofpacts_len = xin->ofpacts_len;
- } else if (xin->rule) {
- ofpacts = xin->rule->up.ofpacts;
- ofpacts_len = xin->rule->up.ofpacts_len;
+ } else if (ctx.rule) {
+ actions = rule_dpif_get_actions(ctx.rule);
+ ofpacts = actions->ofpacts;
+ ofpacts_len = actions->ofpacts_len;
} else {
- NOT_REACHED();
+ OVS_NOT_REACHED();
}
ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack);
+ ofpbuf_use_stub(&ctx.action_set,
+ ctx.action_set_stub, sizeof ctx.action_set_stub);
- if (mbridge_has_mirrors(ctx.xbridge->mbridge) || hit_resubmit_limit) {
+ if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
/* Do this conditionally because the copy is expensive enough that it
* shows up in profiles. */
orig_flow = *flow;
return;
case OFPC_FRAG_REASM:
- NOT_REACHED();
+ OVS_NOT_REACHED();
case OFPC_FRAG_NX_MATCH:
/* Nothing to do. */
break;
case OFPC_INVALID_TTL_TO_CONTROLLER:
- NOT_REACHED();
+ OVS_NOT_REACHED();
}
}
in_port = get_ofp_port(ctx.xbridge, flow->in_port.ofp_port);
+ if (in_port && in_port->is_tunnel) {
+ if (ctx.xin->resubmit_stats) {
+ netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats);
+ if (in_port->bfd) {
+ bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats);
+ }
+ }
+ if (ctx.xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV);
+ entry->u.dev.rx = netdev_ref(in_port->netdev);
+ entry->u.dev.bfd = bfd_ref(in_port->bfd);
+ }
+ }
+
special = process_special(&ctx, flow, in_port, ctx.xin->packet);
if (special) {
- ctx.xout->slow = special;
+ ctx.xout->slow |= special;
} else {
- static struct vlog_rate_limit trace_rl = VLOG_RATE_LIMIT_INIT(1, 1);
size_t sample_actions_len;
if (flow->in_port.ofp_port
add_sflow_action(&ctx);
add_ipfix_action(&ctx);
- sample_actions_len = ctx.xout->odp_actions.size;
+ sample_actions_len = ofpbuf_size(&ctx.xout->odp_actions);
- if (tunnel_ecn_ok(&ctx) && (!in_port || may_receive(in_port, &ctx))) {
+ if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) {
do_xlate_actions(ofpacts, ofpacts_len, &ctx);
/* We've let OFPP_NORMAL and the learning action look at the
* packet, so drop it now if forwarding is disabled. */
- if (in_port && !stp_forward_in_state(in_port->stp_state)) {
- ctx.xout->odp_actions.size = sample_actions_len;
+ if (in_port && !xport_stp_forward_state(in_port)) {
+ ofpbuf_set_size(&ctx.xout->odp_actions, sample_actions_len);
}
}
- if (ctx.max_resubmit_trigger && !ctx.xin->resubmit_hook) {
- if (!hit_resubmit_limit) {
- /* We didn't record the original flow. Make sure we do from
- * now on. */
- hit_resubmit_limit = true;
- } else if (!VLOG_DROP_ERR(&trace_rl)) {
- struct ds ds = DS_EMPTY_INITIALIZER;
-
- ofproto_trace(ctx.xbridge->ofproto, &orig_flow,
- ctx.xin->packet, &ds);
- VLOG_ERR("Trace triggered by excessive resubmit "
- "recursion:\n%s", ds_cstr(&ds));
- ds_destroy(&ds);
- }
+ if (ofpbuf_size(&ctx.action_set)) {
+ xlate_action_set(&ctx);
}
if (ctx.xbridge->has_in_band
}
}
+ if (nl_attr_oversized(ofpbuf_size(&ctx.xout->odp_actions))) {
+ /* These datapath actions are too big for a Netlink attribute, so we
+ * can't hand them to the kernel directly. dpif_execute() can execute
+ * them one by one with help, so just mark the result as SLOW_ACTION to
+ * prevent the flow from being installed. */
+ COVERAGE_INC(xlate_actions_oversize);
+ ctx.xout->slow |= SLOW_ACTION;
+ }
+
+ if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
+ if (ctx.xin->resubmit_stats) {
+ mirror_update_stats(ctx.xbridge->mbridge, xout->mirrors,
+ ctx.xin->resubmit_stats->n_packets,
+ ctx.xin->resubmit_stats->n_bytes);
+ }
+ if (ctx.xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx.xin->xcache, XC_MIRROR);
+ entry->u.mirror.mbridge = mbridge_ref(ctx.xbridge->mbridge);
+ entry->u.mirror.mirrors = xout->mirrors;
+ }
+ }
+
+ if (ctx.xbridge->netflow) {
+ const struct ofpact *ofpacts = actions->ofpacts;
+ size_t ofpacts_len = actions->ofpacts_len;
+
+ /* Only update netflow if we don't have controller flow. We don't
+ * report NetFlow expiration messages for such facets because they
+ * are just part of the control logic for the network, not real
+ * traffic. */
+ if (ofpacts_len == 0
+ || ofpacts->type != OFPACT_CONTROLLER
+ || ofpact_next(ofpacts) < ofpact_end(ofpacts, ofpacts_len)) {
+ if (ctx.xin->resubmit_stats) {
+ netflow_flow_update(ctx.xbridge->netflow, flow,
+ xout->nf_output_iface,
+ ctx.xin->resubmit_stats);
+ }
+ if (ctx.xin->xcache) {
+ struct xc_entry *entry;
+
+ entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW);
+ entry->u.nf.netflow = netflow_ref(ctx.xbridge->netflow);
+ entry->u.nf.flow = xmemdup(flow, sizeof *flow);
+ entry->u.nf.iface = xout->nf_output_iface;
+ }
+ }
+ }
+
ofpbuf_uninit(&ctx.stack);
+ ofpbuf_uninit(&ctx.action_set);
/* Clear the metadata and register wildcard masks, because we won't
* use non-header fields as part of the cache. */
- memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata);
- memset(&wc->masks.regs, 0, sizeof wc->masks.regs);
+ flow_wildcards_clear_non_packet_fields(wc);
+
+ /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow uses
+ * the low 8 bits of the 16-bit tp_src and tp_dst members to represent
+ * these fields. The datapath interface, on the other hand, represents
+ * them with just 8 bits each. This means that if the high 8 bits of the
+ * masks for these fields somehow become set, then they will get chopped
+ * off by a round trip through the datapath, and revalidation will spot
+ * that as an inconsistency and delete the flow. Avoid the problem here by
+ * making sure that only the low 8 bits of either field can be unwildcarded
+ * for ICMP.
+ */
+ if (is_icmp) {
+ wc->masks.tp_src &= htons(UINT8_MAX);
+ wc->masks.tp_dst &= htons(UINT8_MAX);
+ }
+}
+
+/* Sends 'packet' out 'ofport'.
+ * May modify 'packet'.
+ * Returns 0 if successful, otherwise a positive errno value. */
+int
+xlate_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet)
+{
+ struct xport *xport;
+ struct ofpact_output output;
+ struct flow flow;
+
+ ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output);
+ /* Use OFPP_NONE as the in_port to avoid special packet processing. */
+ flow_extract(packet, NULL, &flow);
+ flow.in_port.ofp_port = OFPP_NONE;
+
+ ovs_rwlock_rdlock(&xlate_rwlock);
+ xport = xport_lookup(ofport);
+ if (!xport) {
+ ovs_rwlock_unlock(&xlate_rwlock);
+ return EINVAL;
+ }
+ output.port = xport->ofp_port;
+ output.max_len = 0;
+ ovs_rwlock_unlock(&xlate_rwlock);
+
+ return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL,
+ &output.ofpact, sizeof output,
+ packet);
+}
+
+struct xlate_cache *
+xlate_cache_new(void)
+{
+ struct xlate_cache *xcache = xmalloc(sizeof *xcache);
+
+ ofpbuf_init(&xcache->entries, 512);
+ return xcache;
+}
+
+static struct xc_entry *
+xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type)
+{
+ struct xc_entry *entry;
+
+ entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry);
+ entry->type = type;
+
+ return entry;
+}
+
+static void
+xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats)
+{
+ if (entry->u.dev.tx) {
+ netdev_vport_inc_tx(entry->u.dev.tx, stats);
+ }
+ if (entry->u.dev.rx) {
+ netdev_vport_inc_rx(entry->u.dev.rx, stats);
+ }
+ if (entry->u.dev.bfd) {
+ bfd_account_rx(entry->u.dev.bfd, stats);
+ }
+}
+
+static void
+xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan)
+{
+ struct xbridge *xbridge;
+ struct xbundle *xbundle;
+ struct flow_wildcards wc;
+
+ xbridge = xbridge_lookup(ofproto);
+ if (!xbridge) {
+ return;
+ }
+
+ xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false,
+ NULL);
+ if (!xbundle) {
+ return;
+ }
+
+ update_learning_table(xbridge, flow, &wc, vlan, xbundle);
+}
+
+/* Push stats and perform side effects of flow translation. */
+void
+xlate_push_stats(struct xlate_cache *xcache, bool may_learn,
+ const struct dpif_flow_stats *stats)
+{
+ struct xc_entry *entry;
+ struct ofpbuf entries = xcache->entries;
+
+ XC_ENTRY_FOR_EACH (entry, entries, xcache) {
+ switch (entry->type) {
+ case XC_RULE:
+ rule_dpif_credit_stats(entry->u.rule, stats);
+ break;
+ case XC_BOND:
+ bond_account(entry->u.bond.bond, entry->u.bond.flow,
+ entry->u.bond.vid, stats->n_bytes);
+ break;
+ case XC_NETDEV:
+ xlate_cache_netdev(entry, stats);
+ break;
+ case XC_NETFLOW:
+ netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow,
+ entry->u.nf.iface, stats);
+ break;
+ case XC_MIRROR:
+ mirror_update_stats(entry->u.mirror.mbridge,
+ entry->u.mirror.mirrors,
+ stats->n_packets, stats->n_bytes);
+ break;
+ case XC_LEARN:
+ if (may_learn) {
+ struct rule_dpif *rule = entry->u.learn.rule;
+
+ /* Reset the modified time for a rule that is equivalent to
+ * the currently cached rule. If the rule is not the exact
+ * rule we have cached, update the reference that we have. */
+ entry->u.learn.rule = ofproto_dpif_refresh_rule(rule);
+ }
+ break;
+ case XC_NORMAL:
+ xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow,
+ entry->u.normal.vlan);
+ break;
+ case XC_FIN_TIMEOUT:
+ xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags,
+ entry->u.fin.idle, entry->u.fin.hard);
+ break;
+ default:
+ OVS_NOT_REACHED();
+ }
+ }
+}
+
+static void
+xlate_dev_unref(struct xc_entry *entry)
+{
+ if (entry->u.dev.tx) {
+ netdev_close(entry->u.dev.tx);
+ }
+ if (entry->u.dev.rx) {
+ netdev_close(entry->u.dev.rx);
+ }
+ if (entry->u.dev.bfd) {
+ bfd_unref(entry->u.dev.bfd);
+ }
+}
+
+static void
+xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow)
+{
+ netflow_expire(netflow, flow);
+ netflow_flow_clear(netflow, flow);
+ netflow_unref(netflow);
+ free(flow);
+}
+
+void
+xlate_cache_clear(struct xlate_cache *xcache)
+{
+ struct xc_entry *entry;
+ struct ofpbuf entries;
+
+ if (!xcache) {
+ return;
+ }
+
+ XC_ENTRY_FOR_EACH (entry, entries, xcache) {
+ switch (entry->type) {
+ case XC_RULE:
+ rule_dpif_unref(entry->u.rule);
+ break;
+ case XC_BOND:
+ free(entry->u.bond.flow);
+ bond_unref(entry->u.bond.bond);
+ break;
+ case XC_NETDEV:
+ xlate_dev_unref(entry);
+ break;
+ case XC_NETFLOW:
+ xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow);
+ break;
+ case XC_MIRROR:
+ mbridge_unref(entry->u.mirror.mbridge);
+ break;
+ case XC_LEARN:
+ /* 'u.learn.rule' is the learned rule. */
+ rule_dpif_unref(entry->u.learn.rule);
+ break;
+ case XC_NORMAL:
+ free(entry->u.normal.flow);
+ break;
+ case XC_FIN_TIMEOUT:
+ /* 'u.fin.rule' is always already held as a XC_RULE, which
+ * has already released it's reference above. */
+ break;
+ default:
+ OVS_NOT_REACHED();
+ }
+ }
+
+ ofpbuf_clear(&xcache->entries);
+}
+
+void
+xlate_cache_delete(struct xlate_cache *xcache)
+{
+ xlate_cache_clear(xcache);
+ ofpbuf_uninit(&xcache->entries);
+ free(xcache);
}