X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=ofproto%2Fofproto-dpif-xlate.c;h=a0a10b86bf9fe0a62266dd27c00888bae5190446;hb=9d078ec2f17adc7cceb2793687f8faa1f1c6f4f3;hp=a6bfc8702888416ea193f3e7bdcc50ce573d36b5;hpb=1f317cb5c2aa446c4b0252634a4a70dcc3682f93;p=cascardo%2Fovs.git diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index a6bfc8702..a0a10b86b 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc. +/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,7 +17,12 @@ #include "ofproto/ofproto-dpif-xlate.h" #include +#include +#include +#include +#include +#include "tnl-arp-cache.h" #include "bfd.h" #include "bitmap.h" #include "bond.h" @@ -33,6 +38,7 @@ #include "learn.h" #include "list.h" #include "mac-learning.h" +#include "mcast-snooping.h" #include "meta-flow.h" #include "multipath.h" #include "netdev-vport.h" @@ -46,49 +52,55 @@ #include "ofproto/ofproto-dpif-sflow.h" #include "ofproto/ofproto-dpif.h" #include "ofproto/ofproto-provider.h" +#include "packet-dpif.h" +#include "ovs-router.h" +#include "tnl-ports.h" #include "tunnel.h" -#include "vlog.h" +#include "openvswitch/vlog.h" COVERAGE_DEFINE(xlate_actions); COVERAGE_DEFINE(xlate_actions_oversize); -COVERAGE_DEFINE(xlate_actions_mpls_overflow); +COVERAGE_DEFINE(xlate_actions_too_many_output); 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. */ + +/* Timeout for internal rules created to handle recirculation */ +#define RECIRC_TIMEOUT 60 /* 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. */ - struct list xbundles; /* Owned xbundles. */ + struct ovs_list xbundles; /* Owned xbundles. */ 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 mcast_snooping *ms; /* Multicast Snooping 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. */ + struct rstp *rstp; /* RSTP 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_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. */ @@ -97,16 +109,20 @@ struct xbridge { /* Number of MPLS label stack entries that the datapath supports * in matches. */ size_t max_mpls_depth; + + /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET + * actions. */ + bool masked_set_action; }; struct xbundle { struct hmap_node hmap_node; /* In global 'xbundles' map. */ struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ - struct list list_node; /* In parent 'xbridges' list. */ + struct ovs_list list_node; /* In parent 'xbridges' list. */ struct xbridge *xbridge; /* Parent xbridge. */ - struct list xports; /* Contains "struct xport"s. */ + struct ovs_list xports; /* Contains "struct xport"s. */ char *name; /* Name used in log messages. */ struct bond *bond; /* Nonnull iff more than one port. */ @@ -129,7 +145,7 @@ struct xport { odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ - struct list bundle_node; /* In parent xbundle (if it exists). */ + struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ struct xbundle *xbundle; /* Parent xbundle or null. */ struct netdev *netdev; /* 'ofport''s netdev. */ @@ -140,6 +156,7 @@ struct xport { enum ofputil_port_config config; /* OpenFlow port configuration. */ enum ofputil_port_state state; /* OpenFlow port state. */ int stp_port_no; /* STP port number or -1 if not in use. */ + struct rstp_port *rstp_port; /* RSTP port or null. */ struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ @@ -179,6 +196,7 @@ struct xlate_ctx { int recurse; /* Current resubmit nesting depth. */ int resubmits; /* Total number of resubmits. */ bool in_group; /* Currently translating ofgroup, if true. */ + bool in_action_set; /* Currently translating action_set, if true. */ uint32_t orig_skb_priority; /* Priority when packet arrived. */ uint8_t table_id; /* OpenFlow table ID where flow was found. */ @@ -187,12 +205,23 @@ struct xlate_ctx { uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */ bool exit; /* No further actions should be processed. */ + bool use_recirc; /* Should generate recirc? */ + struct xlate_recirc recirc; /* Information used for generating + * recirculation actions */ + + /* True if a packet was but is no longer MPLS (due to an MPLS pop action). + * This is a trigger for recirculation in cases where translating an action + * or looking up a flow requires access to the fields of the packet after + * the MPLS label stack that was originally present. */ + bool was_mpls; + /* 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. */ + bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ struct ofpbuf action_set; /* Action set. */ uint64_t action_set_stub[1024 / 8]; }; @@ -217,17 +246,101 @@ struct skb_priority_to_dscp { uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ }; -static struct hmap xbridges = HMAP_INITIALIZER(&xbridges); -static struct hmap xbundles = HMAP_INITIALIZER(&xbundles); -static struct hmap xports = HMAP_INITIALIZER(&xports); +enum xc_type { + XC_RULE, + XC_BOND, + XC_NETDEV, + XC_NETFLOW, + XC_MIRROR, + XC_LEARN, + XC_NORMAL, + XC_FIN_TIMEOUT, + XC_GROUP, + XC_TNL_ARP, +}; + +/* 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 ofputil_flow_mod *fm; + struct ofpbuf *ofpacts; + } learn; + struct { + struct ofproto_dpif *ofproto; + struct flow *flow; + int vlan; + } normal; + struct { + struct rule_dpif *rule; + uint16_t idle; + uint16_t hard; + } fin; + struct { + struct group_dpif *group; + struct ofputil_bucket *bucket; + } group; + struct { + char br_name[IFNAMSIZ]; + ovs_be32 d_ip; + } tnl_arp_cache; + } 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; +}; + +/* Xlate config contains hash maps of all bridges, bundles and ports. + * Xcfgp contains the pointer to the current xlate configuration. + * When the main thread needs to change the configuration, it copies xcfgp to + * new_xcfg and edits new_xcfg. This enables the use of RCU locking which + * does not block handler and revalidator threads. */ +struct xlate_cfg { + struct hmap xbridges; + struct hmap xbundles; + struct hmap xports; +}; +static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); +static struct xlate_cfg *new_xcfg = NULL; 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_normal(struct xlate_ctx *); +static inline 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, bool honor_table_miss); @@ -237,45 +350,119 @@ 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 xbridge *xbridge_lookup(const struct ofproto_dpif *); -static struct xbundle *xbundle_lookup(const struct ofbundle *); -static struct xport *xport_lookup(const struct ofport_dpif *); +static struct xbridge *xbridge_lookup(struct xlate_cfg *, + const struct ofproto_dpif *); +static struct xbundle *xbundle_lookup(struct xlate_cfg *, + const struct ofbundle *); +static struct xport *xport_lookup(struct xlate_cfg *, + 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 size_t count_skb_priorities(const struct xport *); static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, uint8_t *dscp); -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, +static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, + enum xc_type type); +static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); +static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); +static void xlate_xport_init(struct xlate_cfg *, struct xport *); +static void xlate_xbridge_set(struct xbridge *, struct dpif *, + const struct mac_learning *, struct stp *, + struct rstp *, const struct mcast_snooping *, + const struct mbridge *, + const struct dpif_sflow *, + const struct dpif_ipfix *, + const struct netflow *, + bool forward_bpdu, bool has_in_band, + bool enable_recirc, + bool variable_length_userdata, + size_t max_mpls_depth, + bool masked_set_action); +static void xlate_xbundle_set(struct xbundle *xbundle, + enum port_vlan_mode vlan_mode, int vlan, + unsigned long *trunks, bool use_priority_tags, + const struct bond *bond, const struct lacp *lacp, + bool floodable); +static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, + const struct netdev *netdev, const struct cfm *cfm, + const struct bfd *bfd, int stp_port_no, + const struct rstp_port *rstp_port, + enum ofputil_port_config config, + enum ofputil_port_state state, bool is_tunnel, + bool may_enable); +static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); +static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); +static void xlate_xport_remove(struct xlate_cfg *, struct xport *); +static void xlate_xbridge_copy(struct xbridge *); +static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); +static void xlate_xport_copy(struct xbridge *, struct xbundle *, + struct xport *); +static void xlate_xcfg_free(struct xlate_cfg *); + +static inline void +xlate_report(struct xlate_ctx *ctx, const char *s) +{ + if (OVS_UNLIKELY(ctx->xin->report_hook)) { + ctx->xin->report_hook(ctx->xin, s, ctx->recurse); + } +} + +static void +xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) +{ + list_init(&xbridge->xbundles); + hmap_init(&xbridge->xports); + hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, + hash_pointer(xbridge->ofproto, 0)); +} + +static void +xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) +{ + list_init(&xbundle->xports); + list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); + hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, + hash_pointer(xbundle->ofbundle, 0)); +} + +static void +xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) +{ + hmap_init(&xport->skb_priorities); + hmap_insert(&xcfg->xports, &xport->hmap_node, + hash_pointer(xport->ofport, 0)); + hmap_insert(&xport->xbridge->xports, &xport->ofp_node, + hash_ofp_port(xport->ofp_port)); +} + +static void +xlate_xbridge_set(struct xbridge *xbridge, + struct dpif *dpif, const struct mac_learning *ml, struct stp *stp, + struct rstp *rstp, const struct mcast_snooping *ms, const struct mbridge *mbridge, const struct dpif_sflow *sflow, const struct dpif_ipfix *ipfix, - const struct netflow *netflow, enum ofp_config_flags frag, + const struct netflow *netflow, bool forward_bpdu, bool has_in_band, + bool enable_recirc, bool variable_length_userdata, - size_t max_mpls_depth) + size_t max_mpls_depth, + bool masked_set_action) { - struct xbridge *xbridge = xbridge_lookup(ofproto); - - if (!xbridge) { - xbridge = xzalloc(sizeof *xbridge); - xbridge->ofproto = ofproto; - - hmap_insert(&xbridges, &xbridge->hmap_node, hash_pointer(ofproto, 0)); - hmap_init(&xbridge->xports); - list_init(&xbridge->xbundles); - } - if (xbridge->ml != ml) { mac_learning_unref(xbridge->ml); xbridge->ml = mac_learning_ref(ml); } + if (xbridge->ms != ms) { + mcast_snooping_unref(xbridge->ms); + xbridge->ms = mcast_snooping_ref(ms); + } + if (xbridge->mbridge != mbridge) { mbridge_unref(xbridge->mbridge); xbridge->mbridge = mbridge_ref(mbridge); @@ -296,28 +483,283 @@ xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, xbridge->stp = stp_ref(stp); } + if (xbridge->rstp != rstp) { + rstp_unref(xbridge->rstp); + xbridge->rstp = rstp_ref(rstp); + } + 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->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; + xbridge->masked_set_action = masked_set_action; +} + +static void +xlate_xbundle_set(struct xbundle *xbundle, + enum port_vlan_mode vlan_mode, int vlan, + unsigned long *trunks, bool use_priority_tags, + const struct bond *bond, const struct lacp *lacp, + bool floodable) +{ + ovs_assert(xbundle->xbridge); + + xbundle->vlan_mode = vlan_mode; + xbundle->vlan = vlan; + xbundle->trunks = trunks; + xbundle->use_priority_tags = use_priority_tags; + xbundle->floodable = floodable; + + if (xbundle->bond != bond) { + bond_unref(xbundle->bond); + xbundle->bond = bond_ref(bond); + } + + if (xbundle->lacp != lacp) { + lacp_unref(xbundle->lacp); + xbundle->lacp = lacp_ref(lacp); + } +} + +static void +xlate_xport_set(struct xport *xport, odp_port_t odp_port, + const struct netdev *netdev, const struct cfm *cfm, + const struct bfd *bfd, int stp_port_no, + const struct rstp_port* rstp_port, + enum ofputil_port_config config, enum ofputil_port_state state, + bool is_tunnel, bool may_enable) +{ + xport->config = config; + 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->rstp_port != rstp_port) { + rstp_port_unref(xport->rstp_port); + xport->rstp_port = rstp_port_ref(rstp_port); + } + + if (xport->cfm != cfm) { + cfm_unref(xport->cfm); + xport->cfm = cfm_ref(cfm); + } + + if (xport->bfd != bfd) { + bfd_unref(xport->bfd); + xport->bfd = bfd_ref(bfd); + } + + if (xport->netdev != netdev) { + netdev_close(xport->netdev); + xport->netdev = netdev_ref(netdev); + } +} + +static void +xlate_xbridge_copy(struct xbridge *xbridge) +{ + struct xbundle *xbundle; + struct xport *xport; + struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); + new_xbridge->ofproto = xbridge->ofproto; + new_xbridge->name = xstrdup(xbridge->name); + xlate_xbridge_init(new_xcfg, new_xbridge); + + xlate_xbridge_set(new_xbridge, + xbridge->dpif, xbridge->ml, xbridge->stp, + xbridge->rstp, xbridge->ms, xbridge->mbridge, + xbridge->sflow, xbridge->ipfix, xbridge->netflow, + xbridge->forward_bpdu, + xbridge->has_in_band, xbridge->enable_recirc, + xbridge->variable_length_userdata, + xbridge->max_mpls_depth, xbridge->masked_set_action); + LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { + xlate_xbundle_copy(new_xbridge, xbundle); + } + + /* Copy xports which are not part of a xbundle */ + HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { + if (!xport->xbundle) { + xlate_xport_copy(new_xbridge, NULL, xport); + } + } +} + +static void +xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) +{ + struct xport *xport; + struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); + new_xbundle->ofbundle = xbundle->ofbundle; + new_xbundle->xbridge = xbridge; + new_xbundle->name = xstrdup(xbundle->name); + xlate_xbundle_init(new_xcfg, new_xbundle); + + xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, + xbundle->vlan, xbundle->trunks, + xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, + xbundle->floodable); + LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { + xlate_xport_copy(xbridge, new_xbundle, xport); + } +} + +static void +xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, + struct xport *xport) +{ + struct skb_priority_to_dscp *pdscp, *new_pdscp; + struct xport *new_xport = xzalloc(sizeof *xport); + new_xport->ofport = xport->ofport; + new_xport->ofp_port = xport->ofp_port; + new_xport->xbridge = xbridge; + xlate_xport_init(new_xcfg, new_xport); + + xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, + xport->bfd, xport->stp_port_no, xport->rstp_port, + xport->config, xport->state, xport->is_tunnel, + xport->may_enable); + + if (xport->peer) { + struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); + if (peer) { + new_xport->peer = peer; + new_xport->peer->peer = new_xport; + } + } + + if (xbundle) { + new_xport->xbundle = xbundle; + list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); + } + + HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { + new_pdscp = xmalloc(sizeof *pdscp); + new_pdscp->skb_priority = pdscp->skb_priority; + new_pdscp->dscp = pdscp->dscp; + hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, + hash_int(new_pdscp->skb_priority, 0)); + } } +/* Sets the current xlate configuration to new_xcfg and frees the old xlate + * configuration in xcfgp. + * + * This needs to be called after editing the xlate configuration. + * + * Functions that edit the new xlate configuration are + * xlate__set and xlate__remove. + * + * A sample workflow: + * + * xlate_txn_start(); + * ... + * edit_xlate_configuration(); + * ... + * xlate_txn_commit(); */ void -xlate_remove_ofproto(struct ofproto_dpif *ofproto) +xlate_txn_commit(void) +{ + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + + ovsrcu_set(&xcfgp, new_xcfg); + ovsrcu_synchronize(); + xlate_xcfg_free(xcfg); + new_xcfg = NULL; +} + +/* Copies the current xlate configuration in xcfgp to new_xcfg. + * + * This needs to be called prior to editing the xlate configuration. */ +void +xlate_txn_start(void) +{ + struct xbridge *xbridge; + struct xlate_cfg *xcfg; + + ovs_assert(!new_xcfg); + + new_xcfg = xmalloc(sizeof *new_xcfg); + hmap_init(&new_xcfg->xbridges); + hmap_init(&new_xcfg->xbundles); + hmap_init(&new_xcfg->xports); + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + if (!xcfg) { + return; + } + + HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { + xlate_xbridge_copy(xbridge); + } +} + + +static void +xlate_xcfg_free(struct xlate_cfg *xcfg) +{ + struct xbridge *xbridge, *next_xbridge; + + if (!xcfg) { + return; + } + + HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { + xlate_xbridge_remove(xcfg, xbridge); + } + + hmap_destroy(&xcfg->xbridges); + hmap_destroy(&xcfg->xbundles); + hmap_destroy(&xcfg->xports); + free(xcfg); +} + +void +xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, + struct dpif *dpif, + const struct mac_learning *ml, struct stp *stp, + struct rstp *rstp, const struct mcast_snooping *ms, + const struct mbridge *mbridge, + const struct dpif_sflow *sflow, + const struct dpif_ipfix *ipfix, + const struct netflow *netflow, + bool forward_bpdu, bool has_in_band, bool enable_recirc, + bool variable_length_userdata, size_t max_mpls_depth, + bool masked_set_action) +{ + struct xbridge *xbridge; + + ovs_assert(new_xcfg); + + xbridge = xbridge_lookup(new_xcfg, ofproto); + if (!xbridge) { + xbridge = xzalloc(sizeof *xbridge); + xbridge->ofproto = ofproto; + + xlate_xbridge_init(new_xcfg, xbridge); + } + + free(xbridge->name); + xbridge->name = xstrdup(name); + + xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, + netflow, forward_bpdu, has_in_band, enable_recirc, + variable_length_userdata, max_mpls_depth, + masked_set_action); +} + +static void +xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) { - struct xbridge *xbridge = xbridge_lookup(ofproto); struct xbundle *xbundle, *next_xbundle; struct xport *xport, *next_xport; @@ -326,24 +768,37 @@ xlate_remove_ofproto(struct ofproto_dpif *ofproto) } HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { - xlate_ofport_remove(xport->ofport); + xlate_xport_remove(xcfg, xport); } LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { - xlate_bundle_remove(xbundle->ofbundle); + xlate_xbundle_remove(xcfg, xbundle); } - hmap_remove(&xbridges, &xbridge->hmap_node); + hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); mac_learning_unref(xbridge->ml); + mcast_snooping_unref(xbridge->ms); mbridge_unref(xbridge->mbridge); dpif_sflow_unref(xbridge->sflow); dpif_ipfix_unref(xbridge->ipfix); stp_unref(xbridge->stp); + rstp_unref(xbridge->rstp); hmap_destroy(&xbridge->xports); free(xbridge->name); free(xbridge); } +void +xlate_remove_ofproto(struct ofproto_dpif *ofproto) +{ + struct xbridge *xbridge; + + ovs_assert(new_xcfg); + + xbridge = xbridge_lookup(new_xcfg, ofproto); + xlate_xbridge_remove(new_xcfg, xbridge); +} + void xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, const char *name, enum port_vlan_mode vlan_mode, int vlan, @@ -351,44 +806,29 @@ xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, const struct bond *bond, const struct lacp *lacp, bool floodable) { - struct xbundle *xbundle = xbundle_lookup(ofbundle); + struct xbundle *xbundle; + ovs_assert(new_xcfg); + + xbundle = xbundle_lookup(new_xcfg, ofbundle); if (!xbundle) { xbundle = xzalloc(sizeof *xbundle); xbundle->ofbundle = ofbundle; - xbundle->xbridge = xbridge_lookup(ofproto); + xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); - hmap_insert(&xbundles, &xbundle->hmap_node, hash_pointer(ofbundle, 0)); - list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); - list_init(&xbundle->xports); + xlate_xbundle_init(new_xcfg, xbundle); } - ovs_assert(xbundle->xbridge); - free(xbundle->name); xbundle->name = xstrdup(name); - xbundle->vlan_mode = vlan_mode; - xbundle->vlan = vlan; - xbundle->trunks = trunks; - xbundle->use_priority_tags = use_priority_tags; - xbundle->floodable = floodable; - - if (xbundle->bond != bond) { - bond_unref(xbundle->bond); - xbundle->bond = bond_ref(bond); - } - - if (xbundle->lacp != lacp) { - lacp_unref(xbundle->lacp); - xbundle->lacp = lacp_ref(lacp); - } + xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, + use_priority_tags, bond, lacp, floodable); } -void -xlate_bundle_remove(struct ofbundle *ofbundle) +static void +xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) { - struct xbundle *xbundle = xbundle_lookup(ofbundle); struct xport *xport, *next; if (!xbundle) { @@ -400,7 +840,7 @@ xlate_bundle_remove(struct ofbundle *ofbundle) xport->xbundle = NULL; } - hmap_remove(&xbundles, &xbundle->hmap_node); + hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); list_remove(&xbundle->list_node); bond_unref(xbundle->bond); lacp_unref(xbundle->lacp); @@ -408,60 +848,53 @@ xlate_bundle_remove(struct ofbundle *ofbundle) free(xbundle); } +void +xlate_bundle_remove(struct ofbundle *ofbundle) +{ + struct xbundle *xbundle; + + ovs_assert(new_xcfg); + + xbundle = xbundle_lookup(new_xcfg, ofbundle); + xlate_xbundle_remove(new_xcfg, xbundle); +} + void xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, 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, int stp_port_no, + const struct rstp_port *rstp_port, 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; + struct xport *xport; + ovs_assert(new_xcfg); + + xport = xport_lookup(new_xcfg, ofport); if (!xport) { xport = xzalloc(sizeof *xport); xport->ofport = ofport; - xport->xbridge = xbridge_lookup(ofproto); + xport->xbridge = xbridge_lookup(new_xcfg, 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)); + xlate_xport_init(new_xcfg, xport); } 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; - xport->odp_port = odp_port; - - if (xport->netdev != netdev) { - netdev_close(xport->netdev); - xport->netdev = netdev_ref(netdev); - } - - if (xport->cfm != cfm) { - cfm_unref(xport->cfm); - xport->cfm = cfm_ref(cfm); - } - - if (xport->bfd != bfd) { - bfd_unref(xport->bfd); - xport->bfd = bfd_ref(bfd); - } + xlate_xport_set(xport, odp_port, netdev, cfm, bfd, stp_port_no, + rstp_port, config, state, is_tunnel, may_enable); if (xport->peer) { xport->peer->peer = NULL; } - xport->peer = xport_lookup(peer); + xport->peer = xport_lookup(new_xcfg, peer); if (xport->peer) { xport->peer->peer = xport; } @@ -469,7 +902,7 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, if (xport->xbundle) { list_remove(&xport->bundle_node); } - xport->xbundle = xbundle_lookup(ofbundle); + xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); if (xport->xbundle) { list_insert(&xport->xbundle->xports, &xport->bundle_node); } @@ -492,11 +925,9 @@ xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, } } -void -xlate_ofport_remove(struct ofport_dpif *ofport) +static void +xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) { - struct xport *xport = xport_lookup(ofport); - if (!xport) { return; } @@ -513,110 +944,155 @@ xlate_ofport_remove(struct ofport_dpif *ofport) clear_skb_priorities(xport); hmap_destroy(&xport->skb_priorities); - hmap_remove(&xports, &xport->hmap_node); + hmap_remove(&xcfg->xports, &xport->hmap_node); hmap_remove(&xport->xbridge->xports, &xport->ofp_node); netdev_close(xport->netdev); + rstp_port_unref(xport->rstp_port); cfm_unref(xport->cfm); bfd_unref(xport->bfd); 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) +void +xlate_ofport_remove(struct ofport_dpif *ofport) { - const struct xport *xport; - int error = ENODEV; + struct xport *xport; - ovs_rwlock_rdlock(&xlate_rwlock); - if (odp_flow_key_to_flow(key, key_len, flow) == ODP_FIT_ERROR) { - error = EINVAL; - goto exit; - } + ovs_assert(new_xcfg); - if (odp_in_port) { - *odp_in_port = flow->in_port.odp_port; - } + xport = xport_lookup(new_xcfg, ofport); + xlate_xport_remove(new_xcfg, xport); +} + +/* Given a datapath and flow metadata ('backer', and 'flow' respectively) + * returns the corresponding struct xport, or NULL if none is found. */ +static struct xport * +xlate_lookup_xport(const struct dpif_backer *backer, const struct flow *flow) +{ + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); - xport = xport_lookup(tnl_port_should_receive(flow) + return xport_lookup(xcfg, 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; +static struct ofproto_dpif * +xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, + ofp_port_t *ofp_in_port, const struct xport **xportp) +{ + struct ofproto_dpif *recv_ofproto = NULL; + struct ofproto_dpif *recirc_ofproto = NULL; + const struct xport *xport; + ofp_port_t in_port = OFPP_NONE; + + *xportp = xport = xlate_lookup_xport(backer, flow); + + if (xport) { + recv_ofproto = xport->xbridge->ofproto; + in_port = xport->ofp_port; } - 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); + /* When recirc_id is set in 'flow', checks whether the ofproto_dpif that + * corresponds to the recirc_id is same as the receiving bridge. If they + * are the same, uses the 'recv_ofproto' and keeps the 'ofp_in_port' as + * assigned. Otherwise, uses the 'recirc_ofproto' that owns recirc_id and + * assigns OFPP_NONE to 'ofp_in_port'. Doing this is in that, the + * recirculated flow must be processced by the ofproto which originates + * the recirculation, and as bridges can only see their own ports, the + * in_port of the 'recv_ofproto' should not be passed to the + * 'recirc_ofproto'. + * + * Admittedly, setting the 'ofp_in_port' to OFPP_NONE limits the + * 'recirc_ofproto' from meaningfully matching on in_port of recirculated + * flow, and should be fixed in the near future. + * + * TODO: Restore the original patch port. + */ + if (recv_ofproto && flow->recirc_id) { + recirc_ofproto = ofproto_dpif_recirc_get_ofproto(backer, + flow->recirc_id); + if (recv_ofproto != recirc_ofproto) { + *xportp = xport = NULL; + in_port = OFPP_NONE; } } - error = 0; - if (ofproto) { - *ofproto = xport->xbridge->ofproto; + if (ofp_in_port) { + *ofp_in_port = in_port; + } + + return xport ? recv_ofproto : recirc_ofproto; +} + +/* Given a datapath and flow metadata ('backer', and 'flow' respectively) + * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ +struct ofproto_dpif * +xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, + ofp_port_t *ofp_in_port) +{ + const struct xport *xport; + + return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); +} + +/* Given a datapath and flow metadata ('backer', and 'flow' respectively), + * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the + * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate + * handles for those protocols if they're enabled. Caller may use the returned + * pointers until quiescing, for longer term use additional references must + * be taken. + * + * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. + */ +int +xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, + struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, + struct dpif_sflow **sflow, struct netflow **netflow, + ofp_port_t *ofp_in_port) +{ + struct ofproto_dpif *ofproto; + const struct xport *xport; + + ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); + + if (!ofproto) { + return ENODEV; + } + + if (ofprotop) { + *ofprotop = ofproto; } if (ipfix) { - *ipfix = dpif_ipfix_ref(xport->xbridge->ipfix); + *ipfix = xport ? xport->xbridge->ipfix : NULL; } if (sflow) { - *sflow = dpif_sflow_ref(xport->xbridge->sflow); + *sflow = xport ? xport->xbridge->sflow : NULL; } if (netflow) { - *netflow = netflow_ref(xport->xbridge->netflow); + *netflow = xport ? xport->xbridge->netflow : NULL; } -exit: - ovs_rwlock_unlock(&xlate_rwlock); - return error; + return 0; } static struct xbridge * -xbridge_lookup(const struct ofproto_dpif *ofproto) +xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) { + struct hmap *xbridges; struct xbridge *xbridge; - if (!ofproto) { + if (!ofproto || !xcfg) { return NULL; } + xbridges = &xcfg->xbridges; + HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), - &xbridges) { + xbridges) { if (xbridge->ofproto == ofproto) { return xbridge; } @@ -625,16 +1101,19 @@ xbridge_lookup(const struct ofproto_dpif *ofproto) } static struct xbundle * -xbundle_lookup(const struct ofbundle *ofbundle) +xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) { + struct hmap *xbundles; struct xbundle *xbundle; - if (!ofbundle) { + if (!ofbundle || !xcfg) { return NULL; } + xbundles = &xcfg->xbundles; + HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), - &xbundles) { + xbundles) { if (xbundle->ofbundle == ofbundle) { return xbundle; } @@ -643,16 +1122,19 @@ xbundle_lookup(const struct ofbundle *ofbundle) } static struct xport * -xport_lookup(const struct ofport_dpif *ofport) +xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) { + struct hmap *xports; struct xport *xport; - if (!ofport) { + if (!ofport || !xcfg) { return NULL; } + xports = &xcfg->xports; + HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), - &xports) { + xports) { if (xport->ofport == ofport) { return xport; } @@ -672,21 +1154,25 @@ 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); + return sp + ? stp_learn_in_state(stp_port_get_state(sp)) + : true; } 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); + return sp + ? stp_forward_in_state(stp_port_get_state(sp)) + : true; } static bool -xport_stp_listen_state(const struct xport *xport) +xport_stp_should_forward_bpdu(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); + return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); } /* Returns true if STP should process 'flow'. Sets fields in 'wc' that @@ -694,8 +1180,9 @@ xport_stp_listen_state(const struct xport *xport) 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 eth_addr_equals(flow->dl_dst, eth_addr_stp); + return is_stp(flow); } static void @@ -721,6 +1208,58 @@ stp_process_packet(const struct xport *xport, const struct ofpbuf *packet) } } +static enum rstp_state +xport_get_rstp_port_state(const struct xport *xport) +{ + return xport->rstp_port + ? rstp_port_get_state(xport->rstp_port) + : RSTP_DISABLED; +} + +static bool +xport_rstp_learn_state(const struct xport *xport) +{ + return xport->xbridge->rstp && xport->rstp_port + ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) + : true; +} + +static bool +xport_rstp_forward_state(const struct xport *xport) +{ + return xport->xbridge->rstp && xport->rstp_port + ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) + : true; +} + +static bool +xport_rstp_should_manage_bpdu(const struct xport *xport) +{ + return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); +} + +static void +rstp_process_packet(const struct xport *xport, const struct ofpbuf *packet) +{ + struct ofpbuf payload = *packet; + struct eth_header *eth = ofpbuf_data(&payload); + + /* Sink packets on ports that have no RSTP. */ + if (!xport->rstp_port) { + 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)) { + rstp_port_received_bpdu(xport->rstp_port, ofpbuf_data(&payload), + ofpbuf_size(&payload)); + } +} + static struct xport * get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) { @@ -745,18 +1284,11 @@ ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) 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; + struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); + return xport && xport->may_enable; } -static const struct ofputil_bucket * +static struct ofputil_bucket * group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, int depth); @@ -764,24 +1296,23 @@ 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; - } + if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { + struct ofputil_bucket *bucket; - hit = group_first_live_bucket(ctx, group, depth) != NULL; + bucket = group_first_live_bucket(ctx, group, depth); + group_dpif_unref(group); + return bucket == NULL; + } - group_dpif_release(group); - return hit; + return false; } #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ static bool bucket_is_alive(const struct xlate_ctx *ctx, - const struct ofputil_bucket *bucket, int depth) + struct ofputil_bucket *bucket, int depth) { if (depth >= MAX_LIVENESS_RECURSION) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); @@ -791,19 +1322,19 @@ bucket_is_alive(const struct xlate_ctx *ctx, 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)); + 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 * +static 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; + const struct ovs_list *buckets; group_dpif_get_buckets(group, &buckets); LIST_FOR_EACH (bucket, list_node, buckets) { @@ -815,17 +1346,17 @@ group_first_live_bucket(const struct xlate_ctx *ctx, return NULL; } -static const struct ofputil_bucket * +static 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; + struct ofputil_bucket *best_bucket = NULL; uint32_t best_score = 0; int i = 0; - const struct ofputil_bucket *bucket; - const struct list *buckets; + struct ofputil_bucket *bucket; + const struct ovs_list *buckets; group_dpif_get_buckets(group, &buckets); LIST_FOR_EACH (bucket, list_node, buckets) { @@ -894,9 +1425,10 @@ lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, return xport->xbundle; } - /* Special-case OFPP_NONE, which a controller may use as the ingress - * port for traffic that it is sourcing. */ - if (in_port == OFPP_NONE) { + /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), + * which a controller may use as the ingress port for traffic that + * it is sourcing. */ + if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { return &ofpp_none_bundle; } @@ -949,7 +1481,7 @@ add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) "%s, which is reserved exclusively for mirroring", ctx->xbridge->name, in_xbundle->name); } - ofpbuf_clear(&ctx->xout->odp_actions); + ofpbuf_clear(ctx->xout->odp_actions); return; } @@ -993,7 +1525,8 @@ add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) mirrors &= ~dup_mirrors; ctx->xout->mirrors |= dup_mirrors; if (out) { - struct xbundle *out_xbundle = xbundle_lookup(out); + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); if (out_xbundle) { output_normal(ctx, out_xbundle, vlan); } @@ -1131,23 +1664,55 @@ output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, /* Partially configured bundle with no slaves. Drop the packet. */ return; } else if (!out_xbundle->bond) { + ctx->use_recirc = false; xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport, bundle_node); } else { + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + struct flow_wildcards *wc = &ctx->xout->wc; + struct xlate_recirc *xr = &ctx->recirc; struct ofport_dpif *ofport; - ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow, - &ctx->xout->wc, vid); - xport = xport_lookup(ofport); + if (ctx->xbridge->enable_recirc) { + ctx->use_recirc = bond_may_recirc( + out_xbundle->bond, &xr->recirc_id, &xr->hash_basis); + + if (ctx->use_recirc) { + /* Only TCP mode uses recirculation. */ + xr->hash_alg = OVS_HASH_ALG_L4; + bond_update_post_recirc_rules(out_xbundle->bond, false); + + /* 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(xcfg, 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); + /* If ctx->xout->use_recirc is set, the main thread will handle stats + * accounting for this bond. */ + if (!ctx->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; + } } } @@ -1180,17 +1745,78 @@ is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) return false; } - memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - if (flow->nw_proto == ARP_OP_REPLY) { - return true; - } else if (flow->nw_proto == ARP_OP_REQUEST) { - memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); - memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); + memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); + if (flow->nw_proto == ARP_OP_REPLY) { + return true; + } else if (flow->nw_proto == ARP_OP_REQUEST) { + memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); + memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); + + return flow->nw_src == flow->nw_dst; + } else { + return false; + } +} + +/* Determines whether packets in 'flow' within 'xbridge' should be forwarded or + * dropped. Returns true if they may be forwarded, false if they should be + * dropped. + * + * 'in_port' must be the xport that corresponds to flow->in_port. + * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). + * + * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as + * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as + * checked by input_vid_is_valid(). + * + * May also add tags to '*tags', although the current implementation only does + * so in one special case. + */ +static bool +is_admissible(struct xlate_ctx *ctx, struct xport *in_port, + uint16_t vlan) +{ + struct xbundle *in_xbundle = in_port->xbundle; + const struct xbridge *xbridge = ctx->xbridge; + struct flow *flow = &ctx->xin->flow; + + /* Drop frames for reserved multicast addresses + * only if forward_bpdu option is absent. */ + if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { + xlate_report(ctx, "packet has reserved destination MAC, dropping"); + return false; + } + + if (in_xbundle->bond) { + struct mac_entry *mac; - return flow->nw_src == flow->nw_dst; - } else { - return false; + switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, + flow->dl_dst)) { + case BV_ACCEPT: + break; + + case BV_DROP: + xlate_report(ctx, "bonding refused admissibility, dropping"); + return false; + + case BV_DROP_IF_MOVED: + ovs_rwlock_rdlock(&xbridge->ml->rwlock); + mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); + if (mac + && mac_entry_get_port(xbridge->ml, mac) != 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"); + return false; + } + ovs_rwlock_unlock(&xbridge->ml->rwlock); + break; + } } + + return true; } /* Checks whether a MAC learning update is necessary for MAC learning table @@ -1231,7 +1857,7 @@ OVS_REQ_RDLOCK(ml->rwlock) } } - return mac->port.p != in_xbundle->ofbundle; + return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; } @@ -1267,7 +1893,7 @@ OVS_REQ_WRLOCK(xbridge->ml->rwlock) } } - if (mac->port.p != in_xbundle->ofbundle) { + if (mac_entry_get_port(xbridge->ml, mac) != 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); @@ -1277,8 +1903,7 @@ OVS_REQ_WRLOCK(xbridge->ml->rwlock) xbridge->name, ETH_ADDR_ARGS(flow->dl_src), in_xbundle->name, vlan); - mac->port.p = in_xbundle->ofbundle; - mac_learning_changed(xbridge->ml); + mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); } } @@ -1308,64 +1933,197 @@ update_learning_table(const struct xbridge *xbridge, } } -/* Determines whether packets in 'flow' within 'xbridge' should be forwarded or - * dropped. Returns true if they may be forwarded, false if they should be - * dropped. - * - * 'in_port' must be the xport that corresponds to flow->in_port. - * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). - * - * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as - * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as - * checked by input_vid_is_valid(). - * - * May also add tags to '*tags', although the current implementation only does - * so in one special case. - */ -static bool -is_admissible(struct xlate_ctx *ctx, struct xport *in_port, - uint16_t vlan) +/* Updates multicast snooping table 'ms' given that a packet matching 'flow' + * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ +static void +update_mcast_snooping_table__(const struct xbridge *xbridge, + const struct flow *flow, + struct mcast_snooping *ms, + ovs_be32 ip4, int vlan, + struct xbundle *in_xbundle) + OVS_REQ_WRLOCK(ms->rwlock) { - struct xbundle *in_xbundle = in_port->xbundle; - const struct xbridge *xbridge = ctx->xbridge; - struct flow *flow = &ctx->xin->flow; - - /* Drop frames for reserved multicast addresses - * only if forward_bpdu option is absent. */ - if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { - xlate_report(ctx, "packet has reserved destination MAC, dropping"); - return false; + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); + + switch (ntohs(flow->tp_src)) { + case IGMP_HOST_MEMBERSHIP_REPORT: + case IGMPV2_HOST_MEMBERSHIP_REPORT: + if (mcast_snooping_add_group(ms, ip4, vlan, in_xbundle->ofbundle)) { + VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " + IP_FMT" is on port %s in VLAN %d", + xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); + } + break; + case IGMP_HOST_LEAVE_MESSAGE: + if (mcast_snooping_leave_group(ms, ip4, vlan, in_xbundle->ofbundle)) { + VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " + IP_FMT" is on port %s in VLAN %d", + xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); + } + break; + case IGMP_HOST_MEMBERSHIP_QUERY: + if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, + in_xbundle->ofbundle)) { + VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " + IP_FMT" is on port %s in VLAN %d", + xbridge->name, IP_ARGS(flow->nw_src), + in_xbundle->name, vlan); + } + break; } +} - if (in_xbundle->bond) { - struct mac_entry *mac; +/* Updates multicast snooping table 'ms' given that a packet matching 'flow' + * was received on 'in_xbundle' in 'vlan'. */ +static void +update_mcast_snooping_table(const struct xbridge *xbridge, + const struct flow *flow, int vlan, + struct xbundle *in_xbundle) +{ + struct mcast_snooping *ms = xbridge->ms; + struct xlate_cfg *xcfg; + struct xbundle *mcast_xbundle; + struct mcast_port_bundle *fport; - switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, - flow->dl_dst)) { - case BV_ACCEPT: + /* Don't learn the OFPP_NONE port. */ + if (in_xbundle == &ofpp_none_bundle) { + return; + } + + /* Don't learn from flood ports */ + mcast_xbundle = NULL; + ovs_rwlock_wrlock(&ms->rwlock); + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + LIST_FOR_EACH(fport, node, &ms->fport_list) { + mcast_xbundle = xbundle_lookup(xcfg, fport->port); + if (mcast_xbundle == in_xbundle) { break; + } + } - case BV_DROP: - xlate_report(ctx, "bonding refused admissibility, dropping"); - return false; + if (!mcast_xbundle || mcast_xbundle != in_xbundle) { + update_mcast_snooping_table__(xbridge, flow, ms, flow->igmp_group_ip4, + vlan, in_xbundle); + } + ovs_rwlock_unlock(&ms->rwlock); +} - case BV_DROP_IF_MOVED: - 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"); - return false; - } - ovs_rwlock_unlock(&xbridge->ml->rwlock); - break; +/* send the packet to ports having the multicast group learned */ +static void +xlate_normal_mcast_send_group(struct xlate_ctx *ctx, + struct mcast_snooping *ms OVS_UNUSED, + struct mcast_group *grp, + struct xbundle *in_xbundle, uint16_t vlan) + OVS_REQ_RDLOCK(ms->rwlock) +{ + struct xlate_cfg *xcfg; + struct mcast_group_bundle *b; + struct xbundle *mcast_xbundle; + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { + mcast_xbundle = xbundle_lookup(xcfg, b->port); + if (mcast_xbundle && mcast_xbundle != in_xbundle) { + xlate_report(ctx, "forwarding to mcast group port"); + output_normal(ctx, mcast_xbundle, vlan); + } else if (!mcast_xbundle) { + xlate_report(ctx, "mcast group port is unknown, dropping"); + } else { + xlate_report(ctx, "mcast group port is input port, dropping"); } } +} - return true; +/* send the packet to ports connected to multicast routers */ +static void +xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, + struct mcast_snooping *ms, + struct xbundle *in_xbundle, uint16_t vlan) + OVS_REQ_RDLOCK(ms->rwlock) +{ + struct xlate_cfg *xcfg; + struct mcast_mrouter_bundle *mrouter; + struct xbundle *mcast_xbundle; + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { + mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); + if (mcast_xbundle && mcast_xbundle != in_xbundle) { + xlate_report(ctx, "forwarding to mcast router port"); + output_normal(ctx, mcast_xbundle, vlan); + } else if (!mcast_xbundle) { + xlate_report(ctx, "mcast router port is unknown, dropping"); + } else { + xlate_report(ctx, "mcast router port is input port, dropping"); + } + } +} + +/* send the packet to ports flagged to be flooded */ +static void +xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, + struct mcast_snooping *ms, + struct xbundle *in_xbundle, uint16_t vlan) + OVS_REQ_RDLOCK(ms->rwlock) +{ + struct xlate_cfg *xcfg; + struct mcast_port_bundle *fport; + struct xbundle *mcast_xbundle; + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + LIST_FOR_EACH(fport, node, &ms->fport_list) { + mcast_xbundle = xbundle_lookup(xcfg, fport->port); + if (mcast_xbundle && mcast_xbundle != in_xbundle) { + xlate_report(ctx, "forwarding to mcast flood port"); + output_normal(ctx, mcast_xbundle, vlan); + } else if (!mcast_xbundle) { + xlate_report(ctx, "mcast flood port is unknown, dropping"); + } else { + xlate_report(ctx, "mcast flood port is input port, dropping"); + } + } +} + +/* forward the Reports to configured ports */ +static void +xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, + struct mcast_snooping *ms, + struct xbundle *in_xbundle, uint16_t vlan) + OVS_REQ_RDLOCK(ms->rwlock) +{ + struct xlate_cfg *xcfg; + struct mcast_port_bundle *rport; + struct xbundle *mcast_xbundle; + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + LIST_FOR_EACH(rport, node, &ms->rport_list) { + mcast_xbundle = xbundle_lookup(xcfg, rport->port); + if (mcast_xbundle && mcast_xbundle != in_xbundle) { + xlate_report(ctx, "forwarding Report to mcast flagged port"); + output_normal(ctx, mcast_xbundle, vlan); + } else if (!mcast_xbundle) { + xlate_report(ctx, "mcast port is unknown, dropping the Report"); + } else { + xlate_report(ctx, "mcast port is input port, dropping the Report"); + } + } +} + +static void +xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, + uint16_t vlan) +{ + struct xbundle *xbundle; + + LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { + if (xbundle != in_xbundle + && xbundle_includes_vlan(xbundle, vlan) + && xbundle->floodable + && !xbundle_mirror_out(ctx->xbridge, xbundle)) { + output_normal(ctx, xbundle, vlan); + } + } + ctx->xout->nf_output_iface = NF_OUT_FLOOD; } static void @@ -1435,36 +2193,100 @@ xlate_normal(struct xlate_ctx *ctx) 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->xbridge->ofproto; + entry->u.normal.flow = xmemdup(flow, sizeof *flow); + entry->u.normal.vlan = vlan; + } /* Determine output bundle. */ - 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); - } else if (!mac_xbundle) { - xlate_report(ctx, "learned port is unknown, dropping"); + if (mcast_snooping_enabled(ctx->xbridge->ms) + && !eth_addr_is_broadcast(flow->dl_dst) + && eth_addr_is_multicast(flow->dl_dst) + && flow->dl_type == htons(ETH_TYPE_IP)) { + struct mcast_snooping *ms = ctx->xbridge->ms; + struct mcast_group *grp; + + if (flow->nw_proto == IPPROTO_IGMP) { + if (ctx->xin->may_learn) { + if (mcast_snooping_is_membership(flow->tp_src) || + mcast_snooping_is_query(flow->tp_src)) { + update_mcast_snooping_table(ctx->xbridge, flow, vlan, + in_xbundle); + } + } + + if (mcast_snooping_is_membership(flow->tp_src)) { + ovs_rwlock_rdlock(&ms->rwlock); + xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); + /* RFC4541: section 2.1.1, item 1: A snooping switch should + * forward IGMP Membership Reports only to those ports where + * multicast routers are attached. Alternatively stated: a + * snooping switch should not forward IGMP Membership Reports + * to ports on which only hosts are attached. + * An administrative control may be provided to override this + * restriction, allowing the report messages to be flooded to + * other ports. */ + xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); + ovs_rwlock_unlock(&ms->rwlock); + } else { + xlate_report(ctx, "multicast traffic, flooding"); + xlate_normal_flood(ctx, in_xbundle, vlan); + } + return; + } else { + if (ip_is_local_multicast(flow->nw_dst)) { + /* RFC4541: section 2.1.2, item 2: Packets with a dst IP + * address in the 224.0.0.x range which are not IGMP must + * be forwarded on all ports */ + xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); + xlate_normal_flood(ctx, in_xbundle, vlan); + return; + } + } + + /* forwarding to group base ports */ + ovs_rwlock_rdlock(&ms->rwlock); + grp = mcast_snooping_lookup(ms, flow->nw_dst, vlan); + if (grp) { + xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); + xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); + xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); } else { - xlate_report(ctx, "learned port is input port, dropping"); + if (mcast_snooping_flood_unreg(ms)) { + xlate_report(ctx, "unregistered multicast, flooding"); + xlate_normal_flood(ctx, in_xbundle, vlan); + } else { + xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); + xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); + } } + ovs_rwlock_unlock(&ms->rwlock); } else { - struct xbundle *xbundle; - - xlate_report(ctx, "no learned MAC for destination, flooding"); - LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { - if (xbundle != in_xbundle - && xbundle_includes_vlan(xbundle, vlan) - && xbundle->floodable - && !xbundle_mirror_out(ctx->xbridge, xbundle)) { - output_normal(ctx, xbundle, vlan); + ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); + mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); + mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; + ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); + + if (mac_port) { + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); + if (mac_xbundle && mac_xbundle != in_xbundle) { + xlate_report(ctx, "forwarding to learned port"); + output_normal(ctx, mac_xbundle, vlan); + } else if (!mac_xbundle) { + xlate_report(ctx, "learned port is unknown, dropping"); + } else { + xlate_report(ctx, "learned port is input port, dropping"); } + } else { + xlate_report(ctx, "no learned MAC for destination, flooding"); + xlate_normal_flood(ctx, in_xbundle, vlan); } - ctx->xout->nf_output_iface = NF_OUT_FLOOD; } } @@ -1478,7 +2300,8 @@ compose_sample_action(const struct xbridge *xbridge, const struct flow *flow, const uint32_t probability, const union user_action_cookie *cookie, - const size_t cookie_size) + const size_t cookie_size, + const odp_port_t tunnel_out_port) { size_t sample_offset, actions_offset; odp_port_t odp_port; @@ -1492,8 +2315,10 @@ compose_sample_action(const struct xbridge *xbridge, actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS); odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port); - pid = dpif_port_get_pid(xbridge->dpif, odp_port, 0); - cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, odp_actions); + pid = dpif_port_get_pid(xbridge->dpif, odp_port, + flow_hash_5tuple(flow, 0)); + cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, + tunnel_out_port, odp_actions); nl_msg_end_nested(odp_actions, actions_offset); nl_msg_end_nested(odp_actions, sample_offset); @@ -1551,7 +2376,7 @@ compose_sflow_action(const struct xbridge *xbridge, odp_port == ODPP_NONE ? 0 : 1, &cookie); return compose_sample_action(xbridge, odp_actions, flow, probability, - &cookie, sizeof cookie.sflow); + &cookie, sizeof cookie.sflow, ODPP_NONE); } static void @@ -1567,29 +2392,54 @@ compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id, } static void -compose_ipfix_cookie(union user_action_cookie *cookie) +compose_ipfix_cookie(union user_action_cookie *cookie, + odp_port_t output_odp_port) { cookie->type = USER_ACTION_COOKIE_IPFIX; + cookie->ipfix.output_odp_port = output_odp_port; } /* Compose SAMPLE action for IPFIX bridge sampling. */ static void compose_ipfix_action(const struct xbridge *xbridge, struct ofpbuf *odp_actions, - const struct flow *flow) + const struct flow *flow, + odp_port_t output_odp_port) { uint32_t probability; union user_action_cookie cookie; + odp_port_t tunnel_out_port = ODPP_NONE; if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) { return; } + /* For input case, output_odp_port is ODPP_NONE, which is an invalid port + * number. */ + if (output_odp_port == ODPP_NONE && + !dpif_ipfix_get_bridge_exporter_input_sampling(xbridge->ipfix)) { + return; + } + + /* For output case, output_odp_port is valid*/ + if (output_odp_port != ODPP_NONE) { + if (!dpif_ipfix_get_bridge_exporter_output_sampling(xbridge->ipfix)) { + return; + } + /* If tunnel sampling is enabled, put an additional option attribute: + * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT + */ + if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(xbridge->ipfix) && + dpif_ipfix_get_tunnel_port(xbridge->ipfix, output_odp_port) ) { + tunnel_out_port = output_odp_port; + } + } + probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix); - compose_ipfix_cookie(&cookie); + compose_ipfix_cookie(&cookie, output_odp_port); compose_sample_action(xbridge, odp_actions, flow, probability, - &cookie, sizeof cookie.ipfix); + &cookie, sizeof cookie.ipfix, tunnel_out_port); } /* SAMPLE action for sFlow must be first action in any given list of @@ -1599,7 +2449,7 @@ static void add_sflow_action(struct xlate_ctx *ctx) { ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge, - &ctx->xout->odp_actions, + ctx->xout->odp_actions, &ctx->xin->flow, ODPP_NONE); ctx->sflow_odp_port = 0; ctx->sflow_n_outputs = 0; @@ -1610,8 +2460,15 @@ add_sflow_action(struct xlate_ctx *ctx) static void add_ipfix_action(struct xlate_ctx *ctx) { - compose_ipfix_action(ctx->xbridge, &ctx->xout->odp_actions, - &ctx->xin->flow); + compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, + &ctx->xin->flow, ODPP_NONE); +} + +static void +add_ipfix_output_action(struct xlate_ctx *ctx, odp_port_t port) +{ + compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, + &ctx->xin->flow, port); } /* Fix SAMPLE action according to data collected while composing ODP actions. @@ -1627,7 +2484,7 @@ fix_sflow_action(struct xlate_ctx *ctx) return; } - cookie = ofpbuf_at(&ctx->xout->odp_actions, ctx->user_cookie_offset, + cookie = ofpbuf_at(ctx->xout->odp_actions, ctx->user_cookie_offset, sizeof cookie->sflow); ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); @@ -1654,11 +2511,7 @@ process_special(struct xlate_ctx *ctx, const struct flow *flow, 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); - } + ofproto_dpif_monitor_port_send_soon(xport->ofport); } } return SLOW_BFD; @@ -1668,9 +2521,12 @@ process_special(struct xlate_ctx *ctx, const struct flow *flow, lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); } return SLOW_LACP; - } else if (xbridge->stp && stp_should_process_flow(flow, wc)) { + } else if ((xbridge->stp || xbridge->rstp) && + stp_should_process_flow(flow, wc)) { if (packet) { - stp_process_packet(xport, packet); + xbridge->stp + ? stp_process_packet(xport, packet) + : rstp_process_packet(xport, packet); } return SLOW_STP; } else { @@ -1678,6 +2534,125 @@ process_special(struct xlate_ctx *ctx, const struct flow *flow, } } +static int +tnl_route_lookup_flow(const struct flow *oflow, + ovs_be32 *ip, struct xport **out_port) +{ + char out_dev[IFNAMSIZ]; + struct xbridge *xbridge; + struct xlate_cfg *xcfg; + ovs_be32 gw; + + if (!ovs_router_lookup(oflow->tunnel.ip_dst, out_dev, &gw)) { + return -ENOENT; + } + + if (gw) { + *ip = gw; + } else { + *ip = oflow->tunnel.ip_dst; + } + + xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + ovs_assert(xcfg); + + HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { + if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { + struct xport *port; + + HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { + if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { + *out_port = port; + return 0; + } + } + } + } + return -ENOENT; +} + +static int +xlate_flood_packet(struct xbridge *xbridge, struct ofpbuf *packet) +{ + 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; + output.port = OFPP_FLOOD; + output.max_len = 0; + + return ofproto_dpif_execute_actions(xbridge->ofproto, &flow, NULL, + &output.ofpact, sizeof output, + packet); +} + +static void +tnl_send_arp_request(const struct xport *out_dev, const uint8_t eth_src[ETH_ADDR_LEN], + ovs_be32 ip_src, ovs_be32 ip_dst) +{ + struct xbridge *xbridge = out_dev->xbridge; + struct ofpbuf packet; + + ofpbuf_init(&packet, 0); + compose_arp(&packet, eth_src, ip_src, ip_dst); + + xlate_flood_packet(xbridge, &packet); + ofpbuf_uninit(&packet); +} + +static int +build_tunnel_send(const struct xlate_ctx *ctx, const struct xport *xport, + const struct flow *flow, odp_port_t tunnel_odp_port) +{ + struct ovs_action_push_tnl tnl_push_data; + struct xport *out_dev = NULL; + ovs_be32 s_ip, d_ip = 0; + uint8_t smac[ETH_ADDR_LEN]; + uint8_t dmac[ETH_ADDR_LEN]; + int err; + + err = tnl_route_lookup_flow(flow, &d_ip, &out_dev); + if (err) { + return err; + } + + /* Use mac addr of bridge port of the peer. */ + err = netdev_get_etheraddr(out_dev->netdev, smac); + if (err) { + return err; + } + + err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL); + if (err) { + return err; + } + + err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, dmac); + if (err) { + tnl_send_arp_request(out_dev, smac, s_ip, d_ip); + return err; + } + if (ctx->xin->xcache) { + struct xc_entry *entry; + + entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_ARP); + strncpy(entry->u.tnl_arp_cache.br_name, out_dev->xbridge->name, IFNAMSIZ); + entry->u.tnl_arp_cache.d_ip = d_ip; + } + err = tnl_port_build_header(xport->ofport, flow, + dmac, smac, s_ip, &tnl_push_data); + if (err) { + return err; + } + tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); + tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); + odp_put_tnl_push_action(ctx->xout->odp_actions, &tnl_push_data); + return 0; +} + static void compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, bool check_stp) @@ -1685,15 +2660,18 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); struct flow_wildcards *wc = &ctx->xout->wc; struct flow *flow = &ctx->xin->flow; + struct flow_tnl flow_tnl; ovs_be16 flow_vlan_tci; uint32_t flow_pkt_mark; uint8_t flow_nw_tos; odp_port_t out_port, odp_port; + bool tnl_push_pop_send = false; 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 == 25); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 30); + memset(&flow_tnl, 0, sizeof flow_tnl); if (!xport) { xlate_report(ctx, "Nonexistent output port"); @@ -1702,15 +2680,27 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); return; } else if (check_stp) { - if (eth_addr_equals(ctx->base_flow.dl_dst, eth_addr_stp)) { - if (!xport_stp_listen_state(xport)) { - xlate_report(ctx, "STP not in listening state, " - "skipping bpdu output"); + if (is_stp(&ctx->base_flow)) { + if (!xport_stp_should_forward_bpdu(xport) && + !xport_rstp_should_manage_bpdu(xport)) { + if (ctx->xbridge->stp != NULL) { + xlate_report(ctx, "STP not in listening state, " + "skipping bpdu output"); + } else if (ctx->xbridge->rstp != NULL) { + xlate_report(ctx, "RSTP not managing BPDU in this state, " + "skipping bpdu output"); + } return; } - } else if (!xport_stp_forward_state(xport)) { - xlate_report(ctx, "STP not in forwarding state, " - "skipping output"); + } else if (!xport_stp_forward_state(xport) || + !xport_rstp_forward_state(xport)) { + if (ctx->xbridge->stp != NULL) { + xlate_report(ctx, "STP not in forwarding state, " + "skipping output"); + } else if (ctx->xbridge->rstp != NULL) { + xlate_report(ctx, "RSTP not in forwarding state, " + "skipping output"); + } return; } } @@ -1724,30 +2714,34 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, const struct xport *peer = xport->peer; struct flow old_flow = ctx->xin->flow; enum slow_path_reason special; + uint8_t table_id = rule_dpif_lookup_get_init_table_id(&ctx->xin->flow); ctx->xbridge = peer->xbridge; flow->in_port.ofp_port = peer->ofp_port; flow->metadata = htonll(0); memset(&flow->tunnel, 0, sizeof flow->tunnel); memset(flow->regs, 0, sizeof flow->regs); + flow->actset_output = OFPP_UNSET; special = process_special(ctx, &ctx->xin->flow, peer, ctx->xin->packet); if (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, true); + if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { + xlate_table_action(ctx, flow->in_port.ofp_port, table_id, + true, true); } else { - /* Forwarding is disabled by STP. Let OFPP_NORMAL and the - * learning action look at the packet, then drop it. */ + /* Forwarding is disabled by STP and RSTP. 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 = ofpbuf_size(&ctx->xout->odp_actions); + 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, true); + xlate_table_action(ctx, flow->in_port.ofp_port, table_id, + true, true); ctx->xout->mirrors = old_mirrors; ctx->base_flow = old_base_flow; - ofpbuf_set_size(&ctx->xout->odp_actions, old_size); + ofpbuf_set_size(ctx->xout->odp_actions, old_size); } } @@ -1761,7 +2755,14 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, 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; } @@ -1769,10 +2770,13 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, flow_pkt_mark = flow->pkt_mark; flow_nw_tos = flow->nw_tos; - if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { - wc->masks.nw_tos |= IP_ECN_MASK; - flow->nw_tos &= ~IP_DSCP_MASK; - flow->nw_tos |= dscp; + if (count_skb_priorities(xport)) { + memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); + 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 (xport->is_tunnel) { @@ -1780,7 +2784,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, * the Logical (tunnel) Port are not visible for any further * matches, while explicit set actions on tunnel metadata are. */ - struct flow_tnl flow_tnl = flow->tunnel; + flow_tnl = flow->tunnel; odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc); if (odp_port == ODPP_NONE) { xlate_report(ctx, "Tunneling decided against output"); @@ -1793,10 +2797,20 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, 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 */ + if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { + tnl_push_pop_send = true; + } else { + commit_odp_tunnel_action(flow, &ctx->base_flow, + ctx->xout->odp_actions); + flow->tunnel = flow_tnl; /* Restore tunnel metadata */ + } } else { odp_port = xport->odp_port; out_port = odp_port; @@ -1815,10 +2829,54 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, if (out_port != ODPP_NONE) { 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); + ctx->xout->odp_actions, + wc, + ctx->xbridge->masked_set_action); + + if (ctx->use_recirc) { + struct ovs_action_hash *act_hash; + struct xlate_recirc *xr = &ctx->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 { + + if (tnl_push_pop_send) { + build_tunnel_send(ctx, xport, flow, odp_port); + flow->tunnel = flow_tnl; /* Restore tunnel metadata */ + } else { + odp_port_t odp_tnl_port = ODPP_NONE; + + /* XXX: Write better Filter for tunnel port. We can use inport + * int tunnel-port flow to avoid these checks completely. */ + if (ofp_port == OFPP_LOCAL && + ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { + + odp_tnl_port = tnl_port_map_lookup(flow, wc); + } + + if (odp_tnl_port != ODPP_NONE) { + nl_msg_put_odp_port(ctx->xout->odp_actions, + OVS_ACTION_ATTR_TUNNEL_POP, + odp_tnl_port); + } else { + /* Tunnel push-pop action is not compatible with + * IPFIX action. */ + add_ipfix_output_action(ctx, out_port); + 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++; @@ -1842,7 +2900,7 @@ static void xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) { struct rule_dpif *old_rule = ctx->rule; - struct rule_actions *actions; + const struct rule_actions *actions; if (ctx->xin->resubmit_stats) { rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); @@ -1862,12 +2920,12 @@ 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) { + 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) { + } 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) { + } 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"); @@ -1883,58 +2941,36 @@ 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 (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; + struct flow_wildcards *wc; 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; + wc = (ctx->xin->skip_wildcards) ? NULL : &ctx->xout->wc; - /* 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; - 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->flow.in_port.ofp_port = old_in_port; - - if (ctx->xin->resubmit_hook) { - ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse); - } - - 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; - default: - OVS_NOT_REACHED(); - } + rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, + &ctx->xin->flow, wc, + ctx->xin->xcache != NULL, + ctx->xin->resubmit_stats, + &ctx->table_id, in_port, + may_packet_in, honor_table_miss); - choose_miss_rule(config, ctx->xbridge->miss_rule, - ctx->xbridge->no_packet_in_rule, &rule); + if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { + ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse + 1); + } -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); - rule_dpif_unref(rule); } ctx->table_id = old_table_id; @@ -1945,7 +2981,23 @@ match: } static void -xlate_group_bucket(struct xlate_ctx *ctx, const struct ofputil_bucket *bucket) +xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, + struct ofputil_bucket *bucket) +{ + if (ctx->xin->resubmit_stats) { + group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); + } + if (ctx->xin->xcache) { + struct xc_entry *entry; + + entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); + entry->u.group.group = group_dpif_ref(group); + entry->u.group.bucket = bucket; + } +} + +static void +xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) { uint64_t action_list_stub[1024 / 8]; struct ofpbuf action_list, action_set; @@ -1965,8 +3017,8 @@ xlate_group_bucket(struct xlate_ctx *ctx, const struct ofputil_bucket *bucket) static void xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) { - const struct ofputil_bucket *bucket; - const struct list *buckets; + struct ofputil_bucket *bucket; + const struct ovs_list *buckets; struct flow old_flow = ctx->xin->flow; group_dpif_get_buckets(group, &buckets); @@ -1981,16 +3033,18 @@ xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) * just before applying the all or indirect group. */ ctx->xin->flow = old_flow; } + xlate_group_stats(ctx, group, NULL); } static void xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) { - const struct ofputil_bucket *bucket; + struct ofputil_bucket *bucket; bucket = group_first_live_bucket(ctx, group, 0); if (bucket) { xlate_group_bucket(ctx, bucket); + xlate_group_stats(ctx, group, bucket); } } @@ -1998,14 +3052,24 @@ 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; + struct ofputil_bucket *bucket; uint32_t basis; - basis = hash_mac(ctx->xin->flow.dl_dst, 0, 0); + basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); bucket = group_best_live_bucket(ctx, group, basis); if (bucket) { memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); + memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); + memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type); + memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); + memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); + memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); + memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); + memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); + memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); + xlate_group_bucket(ctx, bucket); + xlate_group_stats(ctx, group, bucket); } } @@ -2028,7 +3092,7 @@ xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) default: OVS_NOT_REACHED(); } - group_dpif_release(group); + group_dpif_unref(group); ctx->in_group = false; } @@ -2081,6 +3145,15 @@ xlate_ofpact_resubmit(struct xlate_ctx *ctx, { 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) { @@ -2092,7 +3165,8 @@ xlate_ofpact_resubmit(struct xlate_ctx *ctx, table_id = ctx->table_id; } - xlate_table_action(ctx, in_port, table_id, false, false); + xlate_table_action(ctx, in_port, table_id, may_packet_in, + honor_table_miss); } static void @@ -2121,27 +3195,27 @@ execute_controller_action(struct xlate_ctx *ctx, int len, uint16_t controller_id) { struct ofproto_packet_in *pin; - struct ofpbuf *packet; - struct pkt_metadata md = PKT_METADATA_INITIALIZER(0); + struct dpif_packet *packet; ctx->xout->slow |= SLOW_CONTROLLER; if (!ctx->xin->packet) { return; } - packet = ofpbuf_clone(ctx->xin->packet); + packet = dpif_packet_clone_from_ofpbuf(ctx->xin->packet); ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, - &ctx->xout->odp_actions, - &ctx->xout->wc); + ctx->xout->odp_actions, + &ctx->xout->wc, + ctx->xbridge->masked_set_action); - odp_execute_actions(NULL, packet, false, &md, - ofpbuf_data(&ctx->xout->odp_actions), - ofpbuf_size(&ctx->xout->odp_actions), NULL); + odp_execute_actions(NULL, &packet, 1, false, + 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.packet_len = ofpbuf_size(&packet->ofpbuf); + pin->up.packet = ofpbuf_steal_data(&packet->ofpbuf); pin->up.reason = reason; pin->up.table_id = ctx->table_id; pin->up.cookie = (ctx->rule @@ -2169,7 +3243,69 @@ execute_controller_action(struct xlate_ctx *ctx, int len, } } ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin); - ofpbuf_delete(packet); + dpif_packet_delete(packet); +} + +static void +compose_recirculate_action(struct xlate_ctx *ctx, + const struct ofpact *ofpacts_base, + const struct ofpact *ofpact_current, + size_t ofpacts_base_len) +{ + uint32_t id; + int error; + unsigned ofpacts_len; + struct match match; + struct rule *rule; + struct ofpbuf ofpacts; + + ctx->exit = true; + + ofpacts_len = ofpacts_base_len - + ((uint8_t *)ofpact_current - (uint8_t *)ofpacts_base); + + if (ctx->rule) { + id = rule_dpif_get_recirc_id(ctx->rule); + } else { + /* In the case where ctx has no rule then allocate a recirc id. + * The life-cycle of this recirc id is managed by associating it + * with the internal rule that is created to to handle + * recirculation below. + * + * The known use-case of this is packet_out which + * translates actions without a rule */ + id = ofproto_dpif_alloc_recirc_id(ctx->xbridge->ofproto); + } + if (!id) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + VLOG_ERR_RL(&rl, "Failed to allocate recirculation id"); + return; + } + + match_init_catchall(&match); + match_set_recirc_id(&match, id); + ofpbuf_use_const(&ofpacts, ofpact_current, ofpacts_len); + error = ofproto_dpif_add_internal_flow(ctx->xbridge->ofproto, &match, + RECIRC_RULE_PRIORITY, + RECIRC_TIMEOUT, &ofpacts, &rule); + if (error) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + VLOG_ERR_RL(&rl, "Failed to add post recirculation flow %s", + match_to_string(&match, 0)); + return; + } + /* If ctx has no rule then associate the recirc id, which + * was allocated above, with the internal rule. This allows + * the recirc id to be released when the internal rule times out. */ + if (!ctx->rule) { + rule_set_recirc_id(rule, id); + } + + ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, + ctx->xout->odp_actions, + &ctx->xout->wc, + ctx->xbridge->masked_set_action); + nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, id); } static void @@ -2184,8 +3320,9 @@ compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) 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); + ctx->xout->odp_actions, + &ctx->xout->wc, + ctx->xbridge->masked_set_action); } else if (n >= FLOW_MAX_MPLS_LABELS) { if (ctx->xin->packet != NULL) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); @@ -2196,9 +3333,6 @@ compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) } ctx->exit = true; return; - } else if (n >= ctx->xbridge->max_mpls_depth) { - COVERAGE_INC(xlate_actions_mpls_overflow); - ctx->xout->slow |= SLOW_ACTION; } flow_push_mpls(flow, n, mpls->ethertype, wc); @@ -2211,7 +3345,11 @@ compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) struct flow *flow = &ctx->xin->flow; int n = flow_count_mpls_labels(flow, wc); - if (!flow_pop_mpls(flow, n, eth_type, wc) && n >= FLOW_MAX_MPLS_LABELS) { + if (flow_pop_mpls(flow, n, eth_type, wc)) { + if (ctx->xbridge->enable_recirc) { + ctx->was_mpls = true; + } + } 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 " @@ -2220,7 +3358,7 @@ compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); } ctx->exit = true; - ofpbuf_clear(&ctx->xout->odp_actions); + ofpbuf_clear(ctx->xout->odp_actions); } } @@ -2281,24 +3419,23 @@ 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[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)) { + uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); + + wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); 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); - - /* Stop processing for current table. */ - return true; } - } else { - return true; } + + /* Stop processing for current table. */ + return true; } static void @@ -2327,7 +3464,11 @@ xlate_output_action(struct xlate_ctx *ctx, flood_packets(ctx, true); break; case OFPP_CONTROLLER: - execute_controller_action(ctx, max_len, OFPR_ACTION, 0); + execute_controller_action(ctx, max_len, + (ctx->in_group ? OFPR_GROUP + : ctx->in_action_set ? OFPR_ACTION_SET + : OFPR_ACTION), + 0); break; case OFPP_NONE: break; @@ -2457,34 +3598,67 @@ xlate_bundle_action(struct xlate_ctx *ctx, } static void -xlate_learn_action(struct xlate_ctx *ctx, - const struct ofpact_learn *learn) +xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, + struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) { - uint64_t ofpacts_stub[1024 / 8]; - struct ofputil_flow_mod fm; - struct ofpbuf ofpacts; + learn_execute(learn, &ctx->xin->flow, fm, ofpacts); + if (ctx->xin->may_learn) { + ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); + } +} +static void +xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) +{ ctx->xout->has_learn = true; - learn_mask(learn, &ctx->xout->wc); - if (!ctx->xin->may_learn) { - return; + if (ctx->xin->xcache) { + struct xc_entry *entry; + + entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); + entry->u.learn.ofproto = ctx->xbridge->ofproto; + entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); + entry->u.learn.ofpacts = ofpbuf_new(64); + xlate_learn_action__(ctx, learn, entry->u.learn.fm, + entry->u.learn.ofpacts); + } else if (ctx->xin->may_learn) { + uint64_t ofpacts_stub[1024 / 8]; + struct ofputil_flow_mod fm; + struct ofpbuf ofpacts; + + ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); + xlate_learn_action__(ctx, learn, &fm, &ofpacts); + ofpbuf_uninit(&ofpacts); } +} - 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 +xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, + uint16_t idle_timeout, uint16_t hard_timeout) +{ + if (tcp_flags & (TCP_FIN | TCP_RST)) { + rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); + } } static void xlate_fin_timeout(struct xlate_ctx *ctx, const struct ofpact_fin_timeout *oft) { - if (ctx->xin->tcp_flags & (TCP_FIN | TCP_RST) && ctx->rule) { - rule_dpif_reduce_timeouts(ctx->rule, oft->fin_idle_timeout, - oft->fin_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; + } } } @@ -2507,19 +3681,21 @@ xlate_sample_action(struct xlate_ctx *ctx, } ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, - &ctx->xout->odp_actions, - &ctx->xout->wc); + ctx->xout->odp_actions, + &ctx->xout->wc, + ctx->xbridge->masked_set_action); compose_flow_sample_cookie(os->probability, os->collector_set_id, os->obs_domain_id, os->obs_point_id, &cookie); - compose_sample_action(ctx->xbridge, &ctx->xout->odp_actions, &ctx->xin->flow, - probability, &cookie, sizeof cookie.flow_sample); + compose_sample_action(ctx->xbridge, ctx->xout->odp_actions, &ctx->xin->flow, + probability, &cookie, sizeof cookie.flow_sample, + ODPP_NONE); } 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; @@ -2529,7 +3705,8 @@ may_receive(const struct xport *xport, struct xlate_ctx *ctx) * 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 (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) { + if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || + (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { return false; } @@ -2539,8 +3716,31 @@ may_receive(const struct xport *xport, struct xlate_ctx *ctx) static void xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a) { - struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a); - ofpbuf_put(&ctx->action_set, on->actions, ofpact_nest_get_action_len(on)); + const struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a); + size_t on_len = ofpact_nest_get_action_len(on); + const struct ofpact *inner; + + /* Maintain actset_output depending on the contents of the action set: + * + * - OFPP_UNSET, if there is no "output" action. + * + * - The output port, if there is an "output" action and no "group" + * action. + * + * - OFPP_UNSET, if there is a "group" action. + */ + if (!ctx->action_set_has_group) { + OFPACT_FOR_EACH (inner, on->actions, on_len) { + if (inner->type == OFPACT_OUTPUT) { + ctx->xin->flow.actset_output = ofpact_get_OUTPUT(inner)->port; + } else if (inner->type == OFPACT_GROUP) { + ctx->xin->flow.actset_output = OFPP_UNSET; + ctx->action_set_has_group = true; + } + } + } + + ofpbuf_put(&ctx->action_set, on->actions, on_len); ofpact_pad(&ctx->action_set); } @@ -2550,12 +3750,92 @@ xlate_action_set(struct xlate_ctx *ctx) uint64_t action_list_stub[1024 / 64]; struct ofpbuf action_list; + ctx->in_action_set = true; 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); + ctx->in_action_set = false; ofpbuf_uninit(&action_list); } +static bool +ofpact_needs_recirculation_after_mpls(const struct ofpact *a, struct xlate_ctx *ctx) +{ + struct flow_wildcards *wc = &ctx->xout->wc; + struct flow *flow = &ctx->xin->flow; + + if (!ctx->was_mpls) { + return false; + } + + switch (a->type) { + case OFPACT_OUTPUT: + case OFPACT_GROUP: + case OFPACT_CONTROLLER: + case OFPACT_STRIP_VLAN: + case OFPACT_SET_VLAN_PCP: + case OFPACT_SET_VLAN_VID: + case OFPACT_ENQUEUE: + case OFPACT_PUSH_VLAN: + case OFPACT_SET_ETH_SRC: + case OFPACT_SET_ETH_DST: + case OFPACT_SET_TUNNEL: + case OFPACT_SET_QUEUE: + case OFPACT_POP_QUEUE: + case OFPACT_CONJUNCTION: + case OFPACT_NOTE: + case OFPACT_OUTPUT_REG: + case OFPACT_EXIT: + case OFPACT_METER: + case OFPACT_WRITE_METADATA: + case OFPACT_WRITE_ACTIONS: + case OFPACT_CLEAR_ACTIONS: + case OFPACT_SAMPLE: + return false; + + case OFPACT_POP_MPLS: + case OFPACT_DEC_MPLS_TTL: + case OFPACT_SET_MPLS_TTL: + case OFPACT_SET_MPLS_TC: + case OFPACT_SET_MPLS_LABEL: + case OFPACT_SET_IPV4_SRC: + case OFPACT_SET_IPV4_DST: + case OFPACT_SET_IP_DSCP: + case OFPACT_SET_IP_ECN: + case OFPACT_SET_IP_TTL: + case OFPACT_SET_L4_SRC_PORT: + case OFPACT_SET_L4_DST_PORT: + case OFPACT_RESUBMIT: + case OFPACT_STACK_PUSH: + case OFPACT_STACK_POP: + case OFPACT_DEC_TTL: + case OFPACT_MULTIPATH: + case OFPACT_BUNDLE: + case OFPACT_LEARN: + case OFPACT_FIN_TIMEOUT: + case OFPACT_GOTO_TABLE: + return true; + + case OFPACT_REG_MOVE: + return (mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->dst.field) || + mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->src.field)); + + case OFPACT_SET_FIELD: + return mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field); + + case OFPACT_PUSH_MPLS: + /* Recirculate if it is an IP packet with a zero ttl. This may + * indicate that the packet was previously MPLS and an MPLS pop action + * converted it to IP. In this case recirculating should reveal the IP + * TTL which is used as the basis for a new MPLS LSE. */ + return (!flow_count_mpls_labels(flow, wc) + && flow->nw_ttl == 0 + && is_ip_any(flow)); + } + + OVS_NOT_REACHED(); +} + static void do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, struct xlate_ctx *ctx) @@ -2564,6 +3844,9 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, struct flow *flow = &ctx->xin->flow; const struct ofpact *a; + if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { + tnl_arp_snoop(flow, wc, ctx->xbridge->name); + } /* dl_type already in the mask, not set below. */ OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { @@ -2576,6 +3859,11 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; } + if (ofpact_needs_recirculation_after_mpls(a, ctx)) { + compose_recirculate_action(ctx, ofpacts, a, ofpacts_len); + return; + } + switch (a->type) { case OFPACT_OUTPUT: xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, @@ -2596,6 +3884,8 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_ENQUEUE: + memset(&wc->masks.skb_priority, 0xff, + sizeof wc->masks.skb_priority); xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); break; @@ -2678,7 +3968,7 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_SET_L4_SRC_PORT: - if (is_ip_any(flow)) { + if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { 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); @@ -2686,7 +3976,7 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_SET_L4_DST_PORT: - if (is_ip_any(flow)) { + if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { 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); @@ -2702,10 +3992,14 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_SET_QUEUE: + memset(&wc->masks.skb_priority, 0xff, + sizeof wc->masks.skb_priority); xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); break; case OFPACT_POP_QUEUE: + memset(&wc->masks.skb_priority, 0xff, + sizeof wc->masks.skb_priority); flow->skb_priority = ctx->orig_skb_priority; break; @@ -2713,21 +4007,28 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); break; - case OFPACT_REG_LOAD: - 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); + 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; + } + /* A flow may wildcard nw_frag. Do nothing if setting a trasport + * header field on a packet that does not have them. */ + mf_mask_field_and_prereqs(mf, &wc->masks); + if (mf_are_prereqs_ok(mf, flow)) { + mf_set_flow_value_masked(mf, &set_field->value, + &set_field->mask, flow); } break; @@ -2795,6 +4096,16 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, xlate_learn_action(ctx, ofpact_get_LEARN(a)); break; + case OFPACT_CONJUNCTION: { + /* A flow with a "conjunction" action represents part of a special + * kind of "set membership match". Such a flow should not actually + * get executed, but it could via, say, a "packet-out", even though + * that wouldn't be useful. Log it to help debugging. */ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + VLOG_INFO_RL(&rl, "executing no-op conjunction action"); + break; + } + case OFPACT_EXIT: ctx->exit = true; break; @@ -2807,6 +4118,8 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_CLEAR_ACTIONS: ofpbuf_clear(&ctx->action_set); + ctx->xin->flow.actset_output = OFPP_UNSET; + ctx->action_set_has_group = false; break; case OFPACT_WRITE_ACTIONS: @@ -2826,7 +4139,12 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_GOTO_TABLE: { struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); - ovs_assert(ctx->table_id < ogt->table_id); + /* Allow ctx->table_id == TBL_INTERNAL, which will be greater + * than ogt->table_id. This is to allow goto_table actions that + * triggered recirculation: ctx->table_id will be TBL_INTERNAL + * after recirculation. */ + ovs_assert(ctx->table_id == TBL_INTERNAL + || ctx->table_id < ogt->table_id); xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, ogt->table_id, true, true); break; @@ -2841,14 +4159,18 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, void xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, - const struct flow *flow, struct rule_dpif *rule, - uint16_t tcp_flags, const struct ofpbuf *packet) + const struct flow *flow, ofp_port_t in_port, + struct rule_dpif *rule, uint16_t tcp_flags, + const struct ofpbuf *packet) { xin->ofproto = ofproto; xin->flow = *flow; + xin->flow.in_port.ofp_port = in_port; + xin->flow.actset_output = OFPP_UNSET; 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; @@ -2856,13 +4178,14 @@ xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, xin->report_hook = NULL; xin->resubmit_stats = NULL; xin->skip_wildcards = false; + xin->odp_actions = NULL; } void xlate_out_uninit(struct xlate_out *xout) { - if (xout) { - ofpbuf_uninit(&xout->odp_actions); + if (xout && xout->odp_actions == &xout->odp_actions_buf) { + ofpbuf_uninit(xout->odp_actions); } } @@ -2877,14 +4200,6 @@ xlate_actions_for_side_effects(struct xlate_in *xin) xlate_out_uninit(&xout); } -static void -xlate_report(struct xlate_ctx *ctx, const char *s) -{ - if (ctx->xin->report_hook) { - ctx->xin->report_hook(ctx->xin, s, ctx->recurse); - } -} - void xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src) { @@ -2896,10 +4211,11 @@ xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src) dst->nf_output_iface = src->nf_output_iface; dst->mirrors = src->mirrors; - ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub, + dst->odp_actions = &dst->odp_actions_buf; + ofpbuf_use_stub(dst->odp_actions, dst->odp_actions_stub, sizeof dst->odp_actions_stub); - ofpbuf_put(&dst->odp_actions, ofpbuf_data(&src->odp_actions), - ofpbuf_size(&src->odp_actions)); + ofpbuf_put(dst->odp_actions, ofpbuf_data(src->odp_actions), + ofpbuf_size(src->odp_actions)); } static struct skb_priority_to_dscp * @@ -2926,6 +4242,12 @@ dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, return pdscp != NULL; } +static size_t +count_skb_priorities(const struct xport *xport) +{ + return hmap_count(&xport->skb_priorities); +} + static void clear_skb_priorities(struct xport *xport) { @@ -2944,8 +4266,8 @@ actions_output_to_local_port(const struct xlate_ctx *ctx) const struct nlattr *a; unsigned int left; - NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(&ctx->xout->odp_actions), - ofpbuf_size(&ctx->xout->odp_actions)) { + 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; @@ -2954,14 +4276,77 @@ actions_output_to_local_port(const struct xlate_ctx *ctx) return false; } -/* Thread safe call to xlate_actions__(). */ -void -xlate_actions(struct xlate_in *xin, struct xlate_out *xout) - OVS_EXCLUDED(xlate_rwlock) +#if defined(__linux__) +/* Returns the maximum number of packets that the Linux kernel is willing to + * queue up internally to certain kinds of software-implemented ports, or the + * default (and rarely modified) value if it cannot be determined. */ +static int +netdev_max_backlog(void) +{ + static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; + static int max_backlog = 1000; /* The normal default value. */ + + if (ovsthread_once_start(&once)) { + static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; + FILE *stream; + int n; + + stream = fopen(filename, "r"); + if (!stream) { + VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno)); + } else { + if (fscanf(stream, "%d", &n) != 1) { + VLOG_WARN("%s: read error", filename); + } else if (n <= 100) { + VLOG_WARN("%s: unexpectedly small value %d", filename, n); + } else { + max_backlog = n; + } + fclose(stream); + } + ovsthread_once_done(&once); + + VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); + } + + return max_backlog; +} + +/* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in + * 'odp_actions'. */ +static int +count_output_actions(const struct ofpbuf *odp_actions) +{ + const struct nlattr *a; + size_t left; + int n = 0; + + NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(odp_actions), + ofpbuf_size(odp_actions)) { + if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { + n++; + } + } + return n; +} +#endif /* defined(__linux__) */ + +/* Returns true if 'odp_actions' contains more output actions than the datapath + * can reliably handle in one go. On Linux, this is the value of the + * net.core.netdev_max_backlog sysctl, which limits the maximum number of + * packets that the kernel is willing to queue up for processing while the + * datapath is processing a set of actions. */ +static bool +too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) { - ovs_rwlock_rdlock(&xlate_rwlock); - xlate_actions__(xin, xout); - ovs_rwlock_unlock(&xlate_rwlock); +#ifdef __linux__ + return (ofpbuf_size(odp_actions) / NL_A_U32_SIZE > netdev_max_backlog() + && count_output_actions(odp_actions) > netdev_max_backlog()); +#else + /* OSes other than Linux might have similar limits, but we don't know how + * to determine them.*/ + return false; +#endif } /* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts' @@ -2969,15 +4354,14 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) * * 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) +void +xlate_actions(struct xlate_in *xin, struct xlate_out *xout) { + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); 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; @@ -3018,13 +4402,18 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) ctx.xout->has_fin_timeout = false; ctx.xout->nf_output_iface = NF_OUT_DROP; ctx.xout->mirrors = 0; - ofpbuf_use_stub(&ctx.xout->odp_actions, ctx.xout->odp_actions_stub, - sizeof ctx.xout->odp_actions_stub); - ofpbuf_reserve(&ctx.xout->odp_actions, NL_A_U32_SIZE); - ctx.xbridge = xbridge_lookup(xin->ofproto); + xout->odp_actions = xin->odp_actions; + if (!xout->odp_actions) { + xout->odp_actions = &xout->odp_actions_buf; + ofpbuf_use_stub(xout->odp_actions, xout->odp_actions_stub, + sizeof xout->odp_actions_stub); + } + ofpbuf_reserve(xout->odp_actions, NL_A_U32_SIZE); + + ctx.xbridge = xbridge_lookup(xcfg, xin->ofproto); if (!ctx.xbridge) { - goto out; + return; } ctx.rule = xin->rule; @@ -3035,7 +4424,6 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) flow_wildcards_init_catchall(wc); 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); if (is_ip_any(flow)) { wc->masks.nw_frag |= FLOW_NW_FRAG_MASK; @@ -3050,18 +4438,32 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) ctx.recurse = 0; ctx.resubmits = 0; ctx.in_group = false; + ctx.in_action_set = false; ctx.orig_skb_priority = flow->skb_priority; ctx.table_id = 0; ctx.exit = false; + ctx.use_recirc = false; + ctx.was_mpls = false; if (!xin->ofpacts && !ctx.rule) { - ctx.table_id = rule_dpif_lookup(ctx.xbridge->ofproto, flow, - !xin->skip_wildcards ? wc : NULL, - &rule); + rule = rule_dpif_lookup(ctx.xbridge->ofproto, flow, + xin->skip_wildcards ? NULL : wc, + ctx.xin->xcache != NULL, + ctx.xin->resubmit_stats, &ctx.table_id); 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; + + if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { + ctx.xin->resubmit_hook(ctx.xin, rule, 0); + } } xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); @@ -3069,7 +4471,8 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) ofpacts = xin->ofpacts; ofpacts_len = xin->ofpacts_len; } else if (ctx.rule) { - actions = rule_dpif_get_actions(ctx.rule); + const struct rule_actions *actions = rule_dpif_get_actions(ctx.rule); + ofpacts = actions->ofpacts; ofpacts_len = actions->ofpacts_len; } else { @@ -3077,6 +4480,8 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) } ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack); + + ctx.action_set_has_group = false; ofpbuf_use_stub(&ctx.action_set, ctx.action_set_stub, sizeof ctx.action_set_stub); @@ -3086,34 +4491,20 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) orig_flow = *flow; } - if (flow->nw_frag & FLOW_NW_FRAG_ANY) { - switch (ctx.xbridge->frag) { - case OFPC_FRAG_NORMAL: - /* We must pretend that transport ports are unavailable. */ - flow->tp_src = ctx.base_flow.tp_src = htons(0); - flow->tp_dst = ctx.base_flow.tp_dst = htons(0); - break; - - case OFPC_FRAG_DROP: - goto out; - - case OFPC_FRAG_REASM: - OVS_NOT_REACHED(); - - case OFPC_FRAG_NX_MATCH: - /* Nothing to do. */ - break; - - case OFPC_INVALID_TTL_TO_CONTROLLER: - 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; - 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); + 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); } } @@ -3132,15 +4523,16 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) add_sflow_action(&ctx); add_ipfix_action(&ctx); - sample_actions_len = ofpbuf_size(&ctx.xout->odp_actions); + sample_actions_len = ofpbuf_size(ctx.xout->odp_actions); 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 && !xport_stp_forward_state(in_port)) { - ofpbuf_set_size(&ctx.xout->odp_actions, sample_actions_len); + if (in_port && (!xport_stp_forward_state(in_port) || + !xport_rstp_forward_state(in_port))) { + ofpbuf_set_size(ctx.xout->odp_actions, sample_actions_len); } } @@ -3161,37 +4553,54 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) } } - if (nl_attr_oversized(ofpbuf_size(&ctx.xout->odp_actions))) { + 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; + } else if (too_many_output_actions(ctx.xout->odp_actions)) { + COVERAGE_INC(xlate_actions_too_many_output); + 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.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. */ + if (ctx.xbridge->netflow) { + /* 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; + } } } @@ -3216,9 +4625,6 @@ xlate_actions__(struct xlate_in *xin, struct xlate_out *xout) wc->masks.tp_src &= htons(UINT8_MAX); wc->masks.tp_dst &= htons(UINT8_MAX); } - -out: - rule_dpif_unref(rule); } /* Sends 'packet' out 'ofport'. @@ -3227,6 +4633,7 @@ out: int xlate_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet) { + struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); struct xport *xport; struct ofpact_output output; struct flow flow; @@ -3236,17 +4643,211 @@ xlate_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet) flow_extract(packet, NULL, &flow); flow.in_port.ofp_port = OFPP_NONE; - ovs_rwlock_rdlock(&xlate_rwlock); - xport = xport_lookup(ofport); + xport = xport_lookup(xcfg, 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 xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); + struct xbridge *xbridge; + struct xbundle *xbundle; + struct flow_wildcards wc; + + xbridge = xbridge_lookup(xcfg, 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, + const struct dpif_flow_stats *stats) +{ + struct xc_entry *entry; + struct ofpbuf entries = xcache->entries; + uint8_t dmac[ETH_ADDR_LEN]; + + if (!stats->n_packets) { + return; + } + + 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: + ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); + 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; + case XC_GROUP: + group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, + stats); + break; + case XC_TNL_ARP: + /* Lookup arp to avoid arp timeout. */ + tnl_arp_lookup(entry->u.tnl_arp_cache.br_name, entry->u.tnl_arp_cache.d_ip, dmac); + 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_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: + free(entry->u.learn.fm); + ofpbuf_delete(entry->u.learn.ofpacts); + 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; + case XC_GROUP: + group_dpif_unref(entry->u.group.group); + break; + case XC_TNL_ARP: + 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); +}