X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=ofproto%2Fofproto-dpif-xlate.c;h=eb4931ec316eacc0105beb650b0b100a366007a8;hb=94887cf4caa74bfb5c2fd163cb99b145a6d6e2fb;hp=fba23cd5f7a7f57f5e1491a4c7081d5660bb7a09;hpb=ade6ad9cceeaac187ed7632824de413f25539241;p=sliver-openvswitch.git diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index fba23cd5f..eb4931ec3 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -1,4 +1,4 @@ -/* 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. @@ -16,6 +16,8 @@ #include "ofproto/ofproto-dpif-xlate.h" +#include + #include "bfd.h" #include "bitmap.h" #include "bond.h" @@ -40,12 +42,16 @@ #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); @@ -53,6 +59,12 @@ VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); * flow translation. */ #define MAX_RESUBMIT_RECURSION 64 +/* 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 ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ @@ -61,16 +73,30 @@ struct xbridge { 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_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 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 { @@ -112,7 +138,8 @@ struct xport { struct xport *peer; /* Patch port peer or null. */ enum ofputil_port_config config; /* OpenFlow port configuration. */ - int stp_port_no; /* STP port number or 0 if 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. */ @@ -148,13 +175,26 @@ struct xlate_ctx { /* The rule that we are currently translating, or NULL. */ struct rule_dpif *rule; - int recurse; /* Recursion level, via xlate_table_action. */ + /* 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 @@ -182,21 +222,21 @@ static struct hmap xports = HMAP_INITIALIZER(&xports); 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_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); +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); 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); @@ -206,11 +246,16 @@ static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, void xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, + 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) + const struct dpif_ipfix *ipfix, + const struct netflow *netflow, enum ofp_config_flags frag, + bool forward_bpdu, bool has_in_band, + bool variable_length_userdata, + size_t max_mpls_depth) { struct xbridge *xbridge = xbridge_lookup(ofproto); @@ -248,13 +293,22 @@ xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, 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->frag = frag; + xbridge->miss_rule = miss_rule; + xbridge->no_packet_in_rule = no_packet_in_rule; + xbridge->variable_length_userdata = variable_length_userdata; + xbridge->max_mpls_depth = max_mpls_depth; } void @@ -277,6 +331,12 @@ xlate_remove_ofproto(struct ofproto_dpif *ofproto) } 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); } @@ -352,7 +412,8 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, const struct cfm *cfm, const struct bfd *bfd, struct ofport_dpif *peer, int stp_port_no, const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, - enum ofputil_port_config config, bool is_tunnel, + enum ofputil_port_config config, + enum ofputil_port_state state, bool is_tunnel, bool may_enable) { struct xport *xport = xport_lookup(ofport); @@ -373,6 +434,7 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, ovs_assert(xport->ofp_port == ofp_port); xport->config = config; + xport->state = state; xport->stp_port_no = stp_port_no; xport->is_tunnel = is_tunnel; xport->may_enable = may_enable; @@ -396,7 +458,7 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, 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; } @@ -404,7 +466,7 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, 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); } @@ -414,9 +476,8 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, struct skb_priority_to_dscp *pdscp; uint32_t skb_priority; - if (ofproto_dpif_queue_to_priority(xport->xbridge->ofproto, - qdscp_list[i].queue, - &skb_priority)) { + if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, + &skb_priority)) { continue; } @@ -458,11 +519,99 @@ xlate_ofport_remove(struct ofport_dpif *ofport) 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) { @@ -477,6 +626,10 @@ xbundle_lookup(const struct ofbundle *ofbundle) { struct xbundle *xbundle; + if (!ofbundle) { + return NULL; + } + HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), &xbundles) { if (xbundle->ofbundle == ofbundle) { @@ -487,10 +640,14 @@ xbundle_lookup(const struct 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) { @@ -503,7 +660,7 @@ xport_lookup(struct ofport_dpif *ofport) static struct stp_port * xport_get_stp_port(const struct xport *xport) { - return xport->xbridge->stp && xport->stp_port_no + return xport->xbridge->stp && xport->stp_port_no != -1 ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) : NULL; } @@ -575,6 +732,106 @@ ofp_port_to_odp_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) { @@ -710,7 +967,7 @@ add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) 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); @@ -768,7 +1025,7 @@ input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) return vid ? vid : in_xbundle->vlan; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -821,7 +1078,7 @@ input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) return true; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -848,7 +1105,7 @@ output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) return vlan == out_xbundle->vlan ? 0 : vlan; default: - NOT_REACHED(); + OVS_NOT_REACHED(); } } @@ -873,12 +1130,17 @@ output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow, &ctx->xout->wc, vid); - xport = ofport ? xport_lookup(ofport) : NULL; + xport = xport_lookup(ofport); if (!xport) { /* No slaves enabled, so drop packet. */ return; } + + if (ctx->xin->resubmit_stats) { + bond_account(out_xbundle->bond, &ctx->xin->flow, vid, + ctx->xin->resubmit_stats->n_bytes); + } } old_tci = *flow_tci; @@ -923,21 +1185,67 @@ is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) } } -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; + } } - ovs_rwlock_wrlock(&xbridge->ml->rwlock); + 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)) { - goto out; + return; } mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); @@ -947,7 +1255,7 @@ update_learning_table(const struct xbridge *xbridge, if (!in_xbundle->bond) { mac_entry_set_grat_arp_lock(mac); } else if (mac_entry_is_grat_arp_locked(mac)) { - goto out; + return; } } @@ -955,6 +1263,7 @@ update_learning_table(const struct xbridge *xbridge, /* 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), @@ -963,8 +1272,32 @@ update_learning_table(const struct xbridge *xbridge, mac->port.p = in_xbundle->ofbundle; mac_learning_changed(xbridge->ml); } -out: +} + +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); + } } /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or @@ -1016,7 +1349,7 @@ is_admissible(struct xlate_ctx *ctx, struct xport *in_port, || 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); @@ -1035,6 +1368,7 @@ xlate_normal(struct xlate_ctx *ctx) struct xbundle *in_xbundle; struct xport *in_port; struct mac_entry *mac; + void *mac_port; uint16_t vlan; uint16_t vid; @@ -1097,8 +1431,11 @@ xlate_normal(struct xlate_ctx *ctx) /* Determine output bundle. */ ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); - if (mac) { - struct xbundle *mac_xbundle = xbundle_lookup(mac->port.p); + 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); @@ -1121,7 +1458,6 @@ xlate_normal(struct xlate_ctx *ctx) } ctx->xout->nf_output_iface = NF_OUT_FLOOD; } - ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); } /* Compose SAMPLE action for sFlow or IPFIX. The given probability is @@ -1137,15 +1473,19 @@ compose_sample_action(const struct xbridge *xbridge, 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); + 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); @@ -1304,6 +1644,14 @@ process_special(struct xlate_ctx *ctx, const struct flow *flow, } 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 @@ -1330,14 +1678,14 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, 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 == 24); if (!xport) { xlate_report(ctx, "Nonexistent output port"); @@ -1369,7 +1717,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, 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 (xport_stp_forward_state(peer)) { xlate_table_action(ctx, flow->in_port.ofp_port, 0, true); @@ -1387,18 +1735,21 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, } 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); + } } 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 (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { @@ -1430,26 +1781,25 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, &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); + ctx->xout->slow |= 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); @@ -1461,7 +1811,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, 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; } @@ -1471,70 +1821,227 @@ compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port) 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; + 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); + rule_actions_unref(actions); + 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) { + VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times", + MAX_RESUBMIT_RECURSION); + } else if (ctx->resubmits >= MAX_RESUBMITS) { + VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS); + } else if (ctx->xout->odp_actions.size > UINT16_MAX) { + VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions"); + } else if (ctx->stack.size >= 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) { - 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; 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); - - /* Restore the original input port. Otherwise OFPP_NORMAL and - * OFPP_IN_PORT will have surprising behavior. */ + rule_dpif_lookup_in_table(ctx->xbridge->ofproto, &ctx->xin->flow, + !skip_wildcards ? &ctx->xout->wc : NULL, + table_id, &rule); 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; + if (!rule && may_packet_in) { + struct xport *xport; - ctx->recurse++; - ctx->rule = rule; - do_xlate_actions(rule->up.ofpacts, rule->up.ofpacts_len, ctx); - ctx->rule = old_rule; - ctx->recurse--; + /* XXX + * check if table configuration flags + * OFPTC11_TABLE_MISS_CONTROLLER, default. + * OFPTC11_TABLE_MISS_CONTINUE, + * OFPTC11_TABLE_MISS_DROP + * When OF1.0, OFPTC11_TABLE_MISS_CONTINUE is used. What to do? */ + xport = get_ofp_port(ctx->xbridge, ctx->xin->flow.in_port.ofp_port); + choose_miss_rule(xport ? xport->config : 0, + ctx->xbridge->miss_rule, + ctx->xbridge->no_packet_in_rule, &rule); + } + if (rule) { + xlate_recursively(ctx, rule); + rule_dpif_unref(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(action_list.data, action_list.size, 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_bytes(ctx->xin->flow.dl_dst, sizeof ctx->xin->flow.dl_dst, 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); +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 @@ -1582,73 +2089,73 @@ execute_controller_action(struct xlate_ctx *ctx, int len, 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, &md, ctx->xout->odp_actions.data, + ctx->xout->odp_actions.size, NULL); + + pin = xmalloc(sizeof *pin); + pin->up.packet_len = packet->size; + 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; + pin->generated_by_table_miss = (ctx->rule + && rule_dpif_is_table_miss(ctx->rule)); + 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)); - - 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; + ovs_assert(eth_type_mpls(mpls->ethertype)); - 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 @@ -1656,19 +2163,18 @@ compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) { 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); } } @@ -1698,39 +2204,53 @@ compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) } } -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); } +} - ctx->xout->wc.masks.mpls_lse |= htonl(MPLS_TTL_MASK); - 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; } } @@ -1810,8 +2330,7 @@ xlate_enqueue_action(struct xlate_ctx *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); @@ -1844,8 +2363,7 @@ xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) { 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 @@ -1896,7 +2414,8 @@ static void xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) { - struct ofputil_flow_mod *fm; + uint64_t ofpacts_stub[1024 / 8]; + struct ofputil_flow_mod fm; struct ofpbuf ofpacts; ctx->xout->has_learn = true; @@ -1907,21 +2426,10 @@ xlate_learn_action(struct xlate_ctx *ctx, return; } - fm = xmalloc(sizeof *fm); - ofpbuf_init(&ofpacts, 0); - learn_execute(learn, &ctx->xin->flow, fm, &ofpacts); - - ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); -} - -/* 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) -{ - if (max && (!*timeout || *timeout > max)) { - *timeout = max; - } + 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); } static void @@ -1929,14 +2437,8 @@ 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); + rule_dpif_reduce_timeouts(ctx->rule, oft->fin_idle_timeout, + oft->fin_hard_timeout); } } @@ -1949,8 +2451,18 @@ xlate_sample_action(struct xlate_ctx *ctx, * 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); @@ -1978,22 +2490,24 @@ may_receive(const struct xport *xport, struct xlate_ctx *ctx) 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(action_list.data, action_list.size, ctx); + ofpbuf_uninit(&action_list); } static void @@ -2002,18 +2516,15 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, { 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. */ 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; @@ -2025,6 +2536,12 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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, @@ -2038,17 +2555,22 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_SET_VLAN_VID: wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); - flow->vlan_tci &= ~htons(VLAN_VID_MASK); - flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) - | htons(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: wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); - flow->vlan_tci &= ~htons(VLAN_PCP_MASK); - flow->vlan_tci |= - htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp << VLAN_PCP_SHIFT) - | 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: @@ -2073,40 +2595,54 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_SET_IPV4_SRC: - memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_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: - memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_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: - wc->masks.nw_tos |= IP_DSCP_MASK; - /* 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); - memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); 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); - memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); 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; @@ -2132,7 +2668,21 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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; + mf_mask_field_and_prereqs(mf, &wc->masks); + + /* Set field action only ever overwrites packet's outermost + * applicable header fields. Do nothing if no header exists. */ + if ((mf->id != MFF_VLAN_VID || flow->vlan_tci & htons(VLAN_CFI)) + && ((mf->id != MFF_MPLS_LABEL && mf->id != MFF_MPLS_TC) + || eth_type_mpls(flow->dl_type))) { + mf_set_flow_value(mf, &set_field->value, flow); + } break; case OFPACT_STACK_PUSH: @@ -2146,30 +2696,36 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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; @@ -2204,11 +2760,11 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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: @@ -2222,7 +2778,6 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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); ovs_assert(ctx->table_id < ogt->table_id); @@ -2236,17 +2791,12 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, 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; @@ -2259,6 +2809,7 @@ xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, xin->resubmit_hook = NULL; xin->report_hook = NULL; xin->resubmit_stats = NULL; + xin->skip_wildcards = false; } void @@ -2357,20 +2908,38 @@ actions_output_to_local_port(const struct xlate_ctx *ctx) 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) +{ + 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; + 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); @@ -2409,7 +2978,7 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) ctx.xbridge = xbridge_lookup(xin->ofproto); if (!ctx.xbridge) { - return; + goto out; } ctx.rule = xin->rule; @@ -2422,34 +2991,47 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) 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); - /* skb_mark is currently used only by tunnels but that will likely - * change in the future. */ - memset(&wc->masks.skb_mark, 0xff, sizeof wc->masks.skb_mark); + 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.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) { + rule_dpif_lookup(ctx.xbridge->ofproto, flow, + !xin->skip_wildcards ? wc : NULL, &rule); + if (ctx.xin->resubmit_stats) { + rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats); + } + ctx.rule = rule; + } + xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); + 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)) { /* Do this conditionally because the copy is expensive enough that it @@ -2466,24 +3048,31 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) break; case OFPC_FRAG_DROP: - return; + goto out; 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 && 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); + } + } + special = process_special(&ctx, flow, in_port, ctx.xin->packet); if (special) { - ctx.xout->slow = special; + ctx.xout->slow |= special; } else { size_t sample_actions_len; @@ -2498,7 +3087,7 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) add_ipfix_action(&ctx); sample_actions_len = ctx.xout->odp_actions.size; - 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 @@ -2508,6 +3097,10 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) } } + if (ctx.action_set.size) { + xlate_action_set(&ctx); + } + if (ctx.xbridge->has_in_band && in_band_must_output_to_local_port(flow) && !actions_output_to_local_port(&ctx)) { @@ -2521,10 +3114,93 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) } } + if (nl_attr_oversized(ctx.xout->odp_actions.size)) { + /* 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 (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.xbridge->netflow) { + const struct ofpact *ofpacts; + size_t ofpacts_len; + + ofpacts_len = actions->ofpacts_len; + ofpacts = actions->ofpacts; + if (ofpacts_len == 0 + || ofpacts->type != OFPACT_CONTROLLER + || ofpact_next(ofpacts) < ofpact_end(ofpacts, 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. */ + netflow_flow_update(ctx.xbridge->netflow, flow, + xout->nf_output_iface, + ctx.xin->resubmit_stats); + } + } + } + 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); + } + +out: + rule_actions_unref(actions); + rule_dpif_unref(rule); +} + +/* 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); }