X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=ofproto%2Fofproto-dpif-xlate.c;h=f3b56f1bc91c7583a2348293a83a187e635627b8;hb=4f20179ddc79843ef89f68d2c8342df0eb69a814;hp=3430a57d25457ec3c1247483d6e0b57c6d768b3b;hpb=9d0b5cd889374dd50a3b5228598dfd59f7119ca0;p=cascardo%2Fovs.git diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index 3430a57d2..f3b56f1bc 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -205,8 +205,9 @@ struct xlate_ctx { uint32_t orig_skb_priority; /* Priority when packet arrived. */ uint32_t sflow_n_outputs; /* Number of output ports. */ odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ - uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */ + ofp_port_t nf_output_iface; /* Output interface index for NetFlow. */ bool exit; /* No further actions should be processed. */ + mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ /* These are used for non-bond recirculation. The recirculation IDs are * stored in xout and must be associated with a datapath flow (ukey), @@ -299,6 +300,14 @@ struct xlate_ctx { * the MPLS label stack that was originally present. */ bool was_mpls; + /* True if conntrack has been performed on this packet during processing + * on the current bridge. This is used to determine whether conntrack + * state from the datapath should be honored after recirculation. */ + bool conntracked; + + /* Pointer to an embedded NAT action in a conntrack action, or NULL. */ + struct ofpact_nat *ct_nat_action; + /* OpenFlow 1.1+ action set. * * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. @@ -307,9 +316,35 @@ struct xlate_ctx { * datapath actions. */ bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ struct ofpbuf action_set; /* Action set. */ + + enum xlate_error error; /* Translation failed. */ }; +const char *xlate_strerror(enum xlate_error error) +{ + switch (error) { + case XLATE_OK: + return "OK"; + case XLATE_BRIDGE_NOT_FOUND: + return "Bridge not found"; + case XLATE_RECURSION_TOO_DEEP: + return "Recursion too deep"; + case XLATE_TOO_MANY_RESUBMITS: + return "Too many resubmits"; + case XLATE_STACK_TOO_DEEP: + return "Stack too deep"; + case XLATE_NO_RECIRCULATION_CONTEXT: + return "No recirculation context"; + case XLATE_RECIRCULATION_CONFLICT: + return "Recirculation conflict"; + case XLATE_TOO_MANY_MPLS_LABELS: + return "Too many MPLS labels"; + } + return "Unknown error"; +} + static void xlate_action_set(struct xlate_ctx *ctx); +static void xlate_commit_actions(struct xlate_ctx *ctx); static void ctx_trigger_recirculation(struct xlate_ctx *ctx) @@ -529,6 +564,29 @@ xlate_report(struct xlate_ctx *ctx, const char *format, ...) } } +static struct vlog_rate_limit error_report_rl = VLOG_RATE_LIMIT_INIT(1, 5); + +#define XLATE_REPORT_ERROR(CTX, ...) \ + do { \ + if (OVS_UNLIKELY((CTX)->xin->report_hook)) { \ + xlate_report(CTX, __VA_ARGS__); \ + } else { \ + VLOG_ERR_RL(&error_report_rl, __VA_ARGS__); \ + } \ + } while (0) + +static inline void +xlate_report_actions(struct xlate_ctx *ctx, const char *title, + const struct ofpact *ofpacts, size_t ofpacts_len) +{ + if (OVS_UNLIKELY(ctx->xin->report_hook)) { + struct ds s = DS_EMPTY_INITIALIZER; + ofpacts_format(ofpacts, ofpacts_len, &s); + xlate_report(ctx, "%s: %s", title, ds_cstr(&s)); + ds_destroy(&s); + } +} + static void xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) { @@ -1529,74 +1587,61 @@ lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, } static void -add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) +mirror_packet(struct xlate_ctx *ctx, struct xbundle *xbundle, + mirror_mask_t mirrors) { - const struct xbridge *xbridge = ctx->xbridge; - mirror_mask_t mirrors; - struct xbundle *in_xbundle; - uint16_t vlan; - uint16_t vid; - - mirrors = ctx->xout->mirrors; - ctx->xout->mirrors = 0; - - in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port, - ctx->xin->packet != NULL, NULL); - if (!in_xbundle) { + bool warn = ctx->xin->packet != NULL; + uint16_t vid = vlan_tci_to_vid(ctx->xin->flow.vlan_tci); + if (!input_vid_is_valid(vid, xbundle, warn)) { return; } - mirrors |= xbundle_mirror_src(xbridge, in_xbundle); + uint16_t vlan = input_vid_to_vlan(xbundle, vid); - /* Drop frames on bundles reserved for mirroring. */ - if (xbundle_mirror_out(xbridge, in_xbundle)) { - 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 received on port " - "%s, which is reserved exclusively for mirroring", - ctx->xbridge->name, in_xbundle->name); - } - ofpbuf_clear(ctx->odp_actions); - return; - } + const struct xbridge *xbridge = ctx->xbridge; - /* Check VLAN. */ - vid = vlan_tci_to_vid(orig_flow->vlan_tci); - if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { + /* Don't mirror to destinations that we've already mirrored to. */ + mirrors &= ~ctx->mirrors; + if (!mirrors) { return; } - vlan = input_vid_to_vlan(in_xbundle, vid); - if (!mirrors) { - return; + /* Record these mirrors so that we don't mirror to them again. */ + ctx->mirrors |= mirrors; + + if (ctx->xin->resubmit_stats) { + mirror_update_stats(xbridge->mbridge, mirrors, + ctx->xin->resubmit_stats->n_packets, + ctx->xin->resubmit_stats->n_bytes); } + if (ctx->xin->xcache) { + struct xc_entry *entry; - /* Restore the original packet before adding the mirror actions. */ - ctx->xin->flow = *orig_flow; + entry = xlate_cache_add_entry(ctx->xin->xcache, XC_MIRROR); + entry->u.mirror.mbridge = mbridge_ref(xbridge->mbridge); + entry->u.mirror.mirrors = mirrors; + } while (mirrors) { + const unsigned long *vlans; mirror_mask_t dup_mirrors; struct ofbundle *out; - const unsigned long *vlans; - bool vlan_mirrored; - bool has_mirror; int out_vlan; - has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), - &vlans, &dup_mirrors, &out, &out_vlan); + bool has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), + &vlans, &dup_mirrors, &out, &out_vlan); ovs_assert(has_mirror); if (vlans) { ctx->wc->masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); } - vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan); - if (!vlan_mirrored) { + if (vlans && !bitmap_is_set(vlans, vlan)) { mirrors = zero_rightmost_1bit(mirrors); continue; } mirrors &= ~dup_mirrors; - ctx->xout->mirrors |= dup_mirrors; + ctx->mirrors |= dup_mirrors; if (out) { struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); @@ -1604,7 +1649,7 @@ add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) output_normal(ctx, out_xbundle, vlan); } } else if (vlan != out_vlan - && !eth_addr_is_reserved(orig_flow->dl_dst)) { + && !eth_addr_is_reserved(ctx->xin->flow.dl_dst)) { struct xbundle *xbundle; LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { @@ -1617,6 +1662,20 @@ add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) } } +static void +mirror_ingress_packet(struct xlate_ctx *ctx) +{ + if (mbridge_has_mirrors(ctx->xbridge->mbridge)) { + bool warn = ctx->xin->packet != NULL; + struct xbundle *xbundle = lookup_input_bundle( + ctx->xbridge, ctx->xin->flow.in_port.ofp_port, warn, NULL); + if (xbundle) { + mirror_packet(ctx, xbundle, + xbundle_mirror_src(ctx->xbridge, xbundle)); + } + } +} + /* Given 'vid', the VID obtained from the 802.1Q header that was received as * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', * the bundle on which the packet was received, returns the VLAN to which the @@ -2245,7 +2304,7 @@ xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, output_normal(ctx, xbundle, vlan); } } - ctx->xout->nf_output_iface = NF_OUT_FLOOD; + ctx->nf_output_iface = NF_OUT_FLOOD; } static void @@ -2260,8 +2319,6 @@ xlate_normal(struct xlate_ctx *ctx) uint16_t vlan; uint16_t vid; - ctx->xout->has_normal = true; - memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); @@ -2440,211 +2497,147 @@ xlate_normal(struct xlate_ctx *ctx) } } -/* Compose SAMPLE action for sFlow or IPFIX. The given probability is - * the number of packets out of UINT32_MAX to sample. The given - * cookie is passed back in the callback for each sampled packet. +/* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The + * 'probability' is the number of packets out of UINT32_MAX to sample. The + * 'cookie' (of length 'cookie_size' bytes) is passed back in the callback for + * each sampled packet. 'tunnel_out_port', if not ODPP_NONE, is added as the + * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', an + * OVS_USERSPACE_ATTR_ACTIONS attribute is added. */ static size_t -compose_sample_action(const struct xbridge *xbridge, - struct ofpbuf *odp_actions, - const struct flow *flow, +compose_sample_action(struct xlate_ctx *ctx, const uint32_t probability, const union user_action_cookie *cookie, const size_t cookie_size, const odp_port_t tunnel_out_port, bool include_actions) { - size_t sample_offset, actions_offset; - odp_port_t odp_port; - int cookie_offset; - uint32_t pid; + size_t sample_offset = nl_msg_start_nested(ctx->odp_actions, + OVS_ACTION_ATTR_SAMPLE); - sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE); + nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); - nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); + size_t actions_offset = nl_msg_start_nested(ctx->odp_actions, + OVS_SAMPLE_ATTR_ACTIONS); - actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS); + odp_port_t odp_port = ofp_port_to_odp_port( + ctx->xbridge, ctx->xin->flow.in_port.ofp_port); + uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port, + flow_hash_5tuple(&ctx->xin->flow, 0)); + int cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, + tunnel_out_port, + include_actions, + ctx->odp_actions); - odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port); - pid = dpif_port_get_pid(xbridge->dpif, odp_port, - flow_hash_5tuple(flow, 0)); - cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, - tunnel_out_port, - include_actions, - odp_actions); + nl_msg_end_nested(ctx->odp_actions, actions_offset); + nl_msg_end_nested(ctx->odp_actions, sample_offset); - nl_msg_end_nested(odp_actions, actions_offset); - nl_msg_end_nested(odp_actions, sample_offset); return cookie_offset; } -static void -compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci, - odp_port_t odp_port, unsigned int n_outputs, - union user_action_cookie *cookie) -{ - int ifindex; - - cookie->type = USER_ACTION_COOKIE_SFLOW; - cookie->sflow.vlan_tci = vlan_tci; - - /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output - * port information") for the interpretation of cookie->output. */ - switch (n_outputs) { - case 0: - /* 0x40000000 | 256 means "packet dropped for unknown reason". */ - cookie->sflow.output = 0x40000000 | 256; - break; - - case 1: - ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port); - if (ifindex) { - cookie->sflow.output = ifindex; - break; - } - /* Fall through. */ - default: - /* 0x80000000 means "multiple output ports. */ - cookie->sflow.output = 0x80000000 | n_outputs; - break; - } -} - -/* Compose SAMPLE action for sFlow bridge sampling. */ +/* If sFLow is not enabled, returns 0 without doing anything. + * + * If sFlow is enabled, appends a template "sample" action to the ODP actions + * in 'ctx'. This action is a template because some of the information needed + * to fill it out is not available until flow translation is complete. In this + * case, this functions returns an offset, which is always nonzero, to pass + * later to fix_sflow_action() to fill in the rest of the template. */ static size_t -compose_sflow_action(const struct xbridge *xbridge, - struct ofpbuf *odp_actions, - const struct flow *flow, - odp_port_t odp_port) +compose_sflow_action(struct xlate_ctx *ctx) { - uint32_t probability; - union user_action_cookie cookie; - - if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) { + struct dpif_sflow *sflow = ctx->xbridge->sflow; + if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { return 0; } - probability = dpif_sflow_get_probability(xbridge->sflow); - compose_sflow_cookie(xbridge, htons(0), odp_port, - odp_port == ODPP_NONE ? 0 : 1, &cookie); - - return compose_sample_action(xbridge, odp_actions, flow, probability, + union user_action_cookie cookie = { .type = USER_ACTION_COOKIE_SFLOW }; + return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), &cookie, sizeof cookie.sflow, ODPP_NONE, true); } +/* If IPFIX is enabled, this appends a "sample" action to implement IPFIX to + * 'ctx->odp_actions'. */ static void -compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id, - uint32_t obs_domain_id, uint32_t obs_point_id, - union user_action_cookie *cookie) -{ - cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE; - cookie->flow_sample.probability = probability; - cookie->flow_sample.collector_set_id = collector_set_id; - cookie->flow_sample.obs_domain_id = obs_domain_id; - cookie->flow_sample.obs_point_id = obs_point_id; -} - -static void -compose_ipfix_cookie(union user_action_cookie *cookie, - odp_port_t output_odp_port) +compose_ipfix_action(struct xlate_ctx *ctx, 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, - odp_port_t output_odp_port) -{ - uint32_t probability; - union user_action_cookie cookie; + struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; odp_port_t tunnel_out_port = ODPP_NONE; - if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) { + if (!ipfix || ctx->xin->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)) { + !dpif_ipfix_get_bridge_exporter_input_sampling(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)) { + if (!dpif_ipfix_get_bridge_exporter_output_sampling(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) ) { + if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && + dpif_ipfix_get_tunnel_port(ipfix, output_odp_port) ) { tunnel_out_port = output_odp_port; } } - probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix); - compose_ipfix_cookie(&cookie, output_odp_port); - - compose_sample_action(xbridge, odp_actions, flow, probability, + union user_action_cookie cookie = { + .ipfix = { + .type = USER_ACTION_COOKIE_IPFIX, + .output_odp_port = output_odp_port, + } + }; + compose_sample_action(ctx, + dpif_ipfix_get_bridge_exporter_probability(ipfix), &cookie, sizeof cookie.ipfix, tunnel_out_port, false); } -/* SAMPLE action for sFlow must be first action in any given list of - * actions. At this point we do not have all information required to - * build it. So try to build sample action as complete as possible. */ -static void -add_sflow_action(struct xlate_ctx *ctx) -{ - ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge, - ctx->odp_actions, - &ctx->xin->flow, ODPP_NONE); - ctx->sflow_odp_port = 0; - ctx->sflow_n_outputs = 0; -} - -/* SAMPLE action for IPFIX must be 1st or 2nd action in any given list - * of actions, eventually after the SAMPLE action for sFlow. */ -static void -add_ipfix_action(struct xlate_ctx *ctx) -{ - compose_ipfix_action(ctx->xbridge, ctx->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->odp_actions, - &ctx->xin->flow, port); -} - -/* Fix SAMPLE action according to data collected while composing ODP actions. - * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested - * USERSPACE action's user-cookie which is required for sflow. */ +/* Fix "sample" action according to data collected while composing ODP actions, + * as described in compose_sflow_action(). + * + * 'user_cookie_offset' must be the offset returned by add_sflow_action(). */ static void -fix_sflow_action(struct xlate_ctx *ctx) +fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) { const struct flow *base = &ctx->base_flow; union user_action_cookie *cookie; - if (!ctx->user_cookie_offset) { - return; - } - - cookie = ofpbuf_at(ctx->odp_actions, ctx->user_cookie_offset, + cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, sizeof cookie->sflow); ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); - compose_sflow_cookie(ctx->xbridge, base->vlan_tci, - ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie); + cookie->type = USER_ACTION_COOKIE_SFLOW; + cookie->sflow.vlan_tci = base->vlan_tci; + + /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output + * port information") for the interpretation of cookie->output. */ + switch (ctx->sflow_n_outputs) { + case 0: + /* 0x40000000 | 256 means "packet dropped for unknown reason". */ + cookie->sflow.output = 0x40000000 | 256; + break; + + case 1: + cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( + ctx->xbridge->sflow, ctx->sflow_odp_port); + if (cookie->sflow.output) { + break; + } + /* Fall through. */ + default: + /* 0x80000000 means "multiple output ports. */ + cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; + break; + } } static bool @@ -2712,7 +2705,7 @@ tnl_route_lookup_flow(const struct flow *oflow, struct xlate_cfg *xcfg; ovs_be32 gw; - if (!ovs_router_lookup(oflow->tunnel.ip_dst, out_dev, &gw)) { + if (!ovs_router_lookup4(oflow->tunnel.ip_dst, out_dev, &gw)) { return -ENOENT; } @@ -2741,35 +2734,36 @@ tnl_route_lookup_flow(const struct flow *oflow, } static int -xlate_flood_packet(struct xbridge *xbridge, struct dp_packet *packet) +compose_table_xlate(struct xlate_ctx *ctx, const struct xport *out_dev, + struct dp_packet *packet) { + struct xbridge *xbridge = out_dev->xbridge; 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, &flow); - flow.in_port.ofp_port = OFPP_NONE; - output.port = OFPP_FLOOD; + flow.in_port.ofp_port = out_dev->ofp_port; + output.port = OFPP_TABLE; output.max_len = 0; - return ofproto_dpif_execute_actions(xbridge->ofproto, &flow, NULL, - &output.ofpact, sizeof output, - packet); + return ofproto_dpif_execute_actions__(xbridge->ofproto, &flow, NULL, + &output.ofpact, sizeof output, + ctx->recurse, ctx->resubmits, packet); } static void -tnl_send_arp_request(const struct xport *out_dev, const uint8_t eth_src[ETH_ADDR_LEN], +tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev, + const struct eth_addr eth_src, ovs_be32 ip_src, ovs_be32 ip_dst) { - struct xbridge *xbridge = out_dev->xbridge; struct dp_packet packet; dp_packet_init(&packet, 0); compose_arp(&packet, ARP_OP_REQUEST, eth_src, eth_addr_zero, true, ip_src, ip_dst); - xlate_flood_packet(xbridge, &packet); + compose_table_xlate(ctx, out_dev, &packet); dp_packet_uninit(&packet); } @@ -2780,8 +2774,8 @@ build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, 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]; + struct eth_addr smac; + struct eth_addr dmac; int err; err = tnl_route_lookup_flow(flow, &d_ip, &out_dev); @@ -2793,7 +2787,7 @@ build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, IP_ARGS(d_ip), netdev_get_name(out_dev->netdev)); /* Use mac addr of bridge port of the peer. */ - err = netdev_get_etheraddr(out_dev->netdev, smac); + err = netdev_get_etheraddr(out_dev->netdev, &smac); if (err) { xlate_report(ctx, "tunnel output device lacks Ethernet address"); return err; @@ -2805,12 +2799,12 @@ build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, return err; } - err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, dmac); + err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, &dmac); if (err) { xlate_report(ctx, "ARP cache miss for "IP_FMT" on bridge %s, " "sending ARP request", IP_ARGS(d_ip), out_dev->xbridge->name); - tnl_send_arp_request(out_dev, smac, s_ip, d_ip); + tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip); return err; } if (ctx->xin->xcache) { @@ -2837,6 +2831,25 @@ build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, return 0; } +static void +xlate_commit_actions(struct xlate_ctx *ctx) +{ + bool use_masked = ctx->xbridge->support.masked_set_action; + + ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, + ctx->odp_actions, ctx->wc, + use_masked); +} + +static void +clear_conntrack(struct flow *flow) +{ + flow->ct_state = 0; + flow->ct_zone = 0; + flow->ct_mark = 0; + memset(&flow->ct_label, 0, sizeof flow->ct_label); +} + static void compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, const struct xlate_bond_recirc *xr, bool check_stp) @@ -2854,7 +2867,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, /* If 'struct flow' gets additional metadata, we'll need to zero it out * before traversing a patch port. */ - BUILD_ASSERT_DECL(FLOW_WC_SEQ == 33); + BUILD_ASSERT_DECL(FLOW_WC_SEQ == 34); memset(&flow_tnl, 0, sizeof flow_tnl); if (!xport) { @@ -2889,14 +2902,10 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, } } - if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { - ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge, - xport->xbundle); - } - if (xport->peer) { const struct xport *peer = xport->peer; struct flow old_flow = ctx->xin->flow; + bool old_conntrack = ctx->conntracked; bool old_was_mpls = ctx->was_mpls; cls_version_t old_version = ctx->tables_version; struct ofpbuf old_stack = ctx->stack; @@ -2912,6 +2921,8 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, memset(&flow->tunnel, 0, sizeof flow->tunnel); memset(flow->regs, 0, sizeof flow->regs); flow->actset_output = OFPP_UNSET; + ctx->conntracked = false; + clear_conntrack(flow); /* The bridge is now known so obtain its table version. */ ctx->tables_version @@ -2936,10 +2947,10 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, * the learning action look at the packet, then drop it. */ struct flow old_base_flow = ctx->base_flow; size_t old_size = ctx->odp_actions->size; - mirror_mask_t old_mirrors = ctx->xout->mirrors; + mirror_mask_t old_mirrors = ctx->mirrors; xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); - ctx->xout->mirrors = old_mirrors; + ctx->mirrors = old_mirrors; ctx->base_flow = old_base_flow; ctx->odp_actions->size = old_size; @@ -2966,6 +2977,10 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, * bridge. */ ctx->was_mpls = old_was_mpls; + /* The peer bridge's conntrack execution should have no effect on the + * original bridge. */ + ctx->conntracked = old_conntrack; + /* The fact that the peer bridge exits (for any reason) does not mean * that the original bridge should exit. Specifically, if the peer * bridge recirculates (which typically modifies the packet), the @@ -2973,6 +2988,9 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, * recirculated packet! */ ctx->exit = false; + /* Peer bridge errors do not propagate back. */ + ctx->error = XLATE_OK; + if (ctx->xin->resubmit_stats) { netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); @@ -3054,11 +3072,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, } if (out_port != ODPP_NONE) { - bool use_masked = ctx->xbridge->support.masked_set_action; - - ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, - ctx->odp_actions, - wc, use_masked); + xlate_commit_actions(ctx); if (xr) { struct ovs_action_hash *act_hash; @@ -3096,7 +3110,7 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, } else { /* Tunnel push-pop action is not compatible with * IPFIX action. */ - add_ipfix_output_action(ctx, out_port); + compose_ipfix_action(ctx, out_port); nl_msg_put_odp_port(ctx->odp_actions, OVS_ACTION_ATTR_OUTPUT, out_port); @@ -3106,7 +3120,13 @@ compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, ctx->sflow_odp_port = odp_port; ctx->sflow_n_outputs++; - ctx->xout->nf_output_iface = ofp_port; + ctx->nf_output_iface = ofp_port; + } + + if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { + mirror_packet(ctx, xport->xbundle, + xbundle_mirror_dst(xport->xbundle->xbridge, + xport->xbundle)); } out: @@ -3148,17 +3168,20 @@ xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) static bool xlate_resubmit_resource_check(struct xlate_ctx *ctx) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); - if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) { - VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times", - MAX_RESUBMIT_RECURSION); + XLATE_REPORT_ERROR(ctx, "resubmit actions recursed over %d times", + MAX_RESUBMIT_RECURSION); + ctx->error = XLATE_RECURSION_TOO_DEEP; } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) { - VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS); + XLATE_REPORT_ERROR(ctx, "over %d resubmit actions", MAX_RESUBMITS); + ctx->error = XLATE_TOO_MANY_RESUBMITS; } else if (ctx->odp_actions->size > UINT16_MAX) { - VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions"); + XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of actions"); + /* NOT an error, as we'll be slow-pathing the flow in this case? */ + ctx->exit = true; /* XXX: translation still terminated! */ } else if (ctx->stack.size >= 65536) { - VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack"); + XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of stack"); + ctx->error = XLATE_STACK_TOO_DEEP; } else { return true; } @@ -3184,7 +3207,6 @@ xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, ctx->tables_version, &ctx->xin->flow, ctx->xin->wc, - ctx->xin->xcache != NULL, ctx->xin->resubmit_stats, &ctx->table_id, in_port, may_packet_in, honor_table_miss); @@ -3203,6 +3225,7 @@ xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); entry->u.rule = rule; + rule_dpif_ref(rule); } xlate_recursively(ctx, rule); } @@ -3210,8 +3233,6 @@ xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, ctx->table_id = old_table_id; return; } - - ctx->exit = true; } static void @@ -3377,6 +3398,11 @@ xlate_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) } basis = hash_bytes(&value, mf->n_bytes, basis); + /* For tunnels, hash in whether the field is present. */ + if (mf_is_tun_metadata(mf)) { + basis = hash_boolean(mf_is_set(mf, &ctx->xin->flow), basis); + } + mf_mask_field(mf, &ctx->wc->masks); } } @@ -3493,7 +3519,7 @@ flood_packets(struct xlate_ctx *ctx, bool all) } } - ctx->xout->nf_output_iface = NF_OUT_FLOOD; + ctx->nf_output_iface = NF_OUT_FLOOD; } static void @@ -3503,7 +3529,6 @@ execute_controller_action(struct xlate_ctx *ctx, int len, { struct ofproto_packet_in *pin; struct dp_packet *packet; - bool use_masked; ctx->xout->slow |= SLOW_CONTROLLER; if (!ctx->xin->packet) { @@ -3512,10 +3537,7 @@ execute_controller_action(struct xlate_ctx *ctx, int len, packet = dp_packet_clone(ctx->xin->packet); - use_masked = ctx->xbridge->support.masked_set_action; - ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, - ctx->odp_actions, - ctx->wc, use_masked); + xlate_commit_actions(ctx); odp_execute_actions(NULL, &packet, 1, false, ctx->odp_actions->data, ctx->odp_actions->size, NULL); @@ -3551,47 +3573,40 @@ execute_controller_action(struct xlate_ctx *ctx, int len, dp_packet_delete(packet); } -/* Called only when ctx->recirc_action_offset is set. */ static void -compose_recirculate_action(struct xlate_ctx *ctx) +compose_recirculate_action__(struct xlate_ctx *ctx, uint8_t table) { struct recirc_metadata md; - bool use_masked; uint32_t id; - use_masked = ctx->xbridge->support.masked_set_action; - ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, - ctx->odp_actions, - ctx->wc, use_masked); - recirc_metadata_from_flow(&md, &ctx->xin->flow); ovs_assert(ctx->recirc_action_offset >= 0); - /* Only allocate recirculation ID if we have a packet. */ - if (ctx->xin->packet) { - /* Allocate a unique recirc id for the given metadata state in the - * flow. The life-cycle of this recirc id is managed by associating it - * with the udpif key ('ukey') created for each new datapath flow. */ - id = recirc_alloc_id_ctx(ctx->xbridge->ofproto, 0, &md, &ctx->stack, - ctx->recirc_action_offset, - ctx->action_set.size, ctx->action_set.data); - if (!id) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - VLOG_ERR_RL(&rl, "Failed to allocate recirculation id"); - return; - } - xlate_out_add_recirc(ctx->xout, id); - } else { - /* Look up an existing recirc id for the given metadata state in the - * flow. No new reference is taken, as the ID is RCU protected and is - * only required temporarily for verification. */ - id = recirc_find_id(ctx->xbridge->ofproto, 0, &md, &ctx->stack, - ctx->recirc_action_offset, - ctx->action_set.size, ctx->action_set.data); - /* We let zero 'id' to be used in the RECIRC action below, which will - * fail all revalidations as zero is not a valid recirculation ID. */ + struct recirc_state state = { + .table_id = table, + .ofproto = ctx->xbridge->ofproto, + .metadata = md, + .stack = &ctx->stack, + .mirrors = ctx->mirrors, + .conntracked = ctx->conntracked, + .action_set_len = ctx->recirc_action_offset, + .ofpacts_len = ctx->action_set.size, + .ofpacts = ctx->action_set.data, + }; + + /* Allocate a unique recirc id for the given metadata state in the + * flow. An existing id, with a new reference to the corresponding + * recirculation context, will be returned if possible. + * The life-cycle of this recirc id is managed by associating it + * with the udpif key ('ukey') created for each new datapath flow. */ + id = recirc_alloc_id_ctx(&state); + if (!id) { + XLATE_REPORT_ERROR(ctx, "Failed to allocate recirculation id"); + ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; + return; } + recirc_refs_add(&ctx->xout->recircs, id); nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, id); @@ -3601,6 +3616,25 @@ compose_recirculate_action(struct xlate_ctx *ctx) ctx->last_unroll_offset = -1; } +/* Called only when ctx->recirc_action_offset is set. */ +static void +compose_recirculate_action(struct xlate_ctx *ctx) +{ + xlate_commit_actions(ctx); + compose_recirculate_action__(ctx, 0); +} + +/* Fork the pipeline here. The current packet will continue processing the + * current action list. A clone of the current packet will recirculate, skip + * the remainder of the current action list and asynchronously resume pipeline + * processing in 'table' with the current metadata and action set. */ +static void +compose_recirculate_and_fork(struct xlate_ctx *ctx, uint8_t table) +{ + ctx->recirc_action_offset = ctx->action_set.size; + compose_recirculate_action__(ctx, table); +} + static void compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) { @@ -3611,20 +3645,15 @@ compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) n = flow_count_mpls_labels(flow, ctx->wc); if (!n) { - bool use_masked = ctx->xbridge->support.masked_set_action; - - ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, - ctx->odp_actions, - ctx->wc, use_masked); + xlate_commit_actions(ctx); } 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 " + XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " "MPLS push action can't be performed as it would " "have more MPLS LSEs than the %d supported.", ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); } - ctx->exit = true; + ctx->error = XLATE_TOO_MANY_MPLS_LABELS; return; } @@ -3643,13 +3672,12 @@ compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) } } 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 " + XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " "MPLS pop action can't be performed as it has " "more MPLS LSEs than the %d supported.", ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); } - ctx->exit = true; + ctx->error = XLATE_TOO_MANY_MPLS_LABELS; ofpbuf_clear(ctx->odp_actions); } } @@ -3733,9 +3761,9 @@ static void xlate_output_action(struct xlate_ctx *ctx, ofp_port_t port, uint16_t max_len, bool may_packet_in) { - ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface; + ofp_port_t prev_nf_output_iface = ctx->nf_output_iface; - ctx->xout->nf_output_iface = NF_OUT_DROP; + ctx->nf_output_iface = NF_OUT_DROP; switch (port) { case OFPP_IN_PORT: @@ -3774,12 +3802,12 @@ xlate_output_action(struct xlate_ctx *ctx, } if (prev_nf_output_iface == NF_OUT_FLOOD) { - ctx->xout->nf_output_iface = NF_OUT_FLOOD; - } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) { - ctx->xout->nf_output_iface = prev_nf_output_iface; + ctx->nf_output_iface = NF_OUT_FLOOD; + } else if (ctx->nf_output_iface == NF_OUT_DROP) { + ctx->nf_output_iface = prev_nf_output_iface; } else if (prev_nf_output_iface != NF_OUT_DROP && - ctx->xout->nf_output_iface != NF_OUT_FLOOD) { - ctx->xout->nf_output_iface = NF_OUT_MULTI; + ctx->nf_output_iface != NF_OUT_FLOOD) { + ctx->nf_output_iface = NF_OUT_MULTI; } } @@ -3829,10 +3857,10 @@ xlate_enqueue_action(struct xlate_ctx *ctx, ctx->xin->flow.skb_priority = flow_priority; /* Update NetFlow output port. */ - if (ctx->xout->nf_output_iface == NF_OUT_DROP) { - ctx->xout->nf_output_iface = ofp_port; - } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) { - ctx->xout->nf_output_iface = NF_OUT_MULTI; + if (ctx->nf_output_iface == NF_OUT_DROP) { + ctx->nf_output_iface = ofp_port; + } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { + ctx->nf_output_iface = NF_OUT_MULTI; } } @@ -3899,7 +3927,6 @@ xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, static void xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) { - ctx->xout->has_learn = true; learn_mask(learn, ctx->wc); if (ctx->xin->xcache) { @@ -3955,11 +3982,9 @@ static void xlate_sample_action(struct xlate_ctx *ctx, const struct ofpact_sample *os) { - union user_action_cookie cookie; /* Scale the probability from 16-bit to 32-bit while representing * the same percentage. */ uint32_t probability = (os->probability << 16) | os->probability; - bool use_masked; if (!ctx->xbridge->support.variable_length_userdata) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); @@ -3970,17 +3995,19 @@ xlate_sample_action(struct xlate_ctx *ctx, return; } - use_masked = ctx->xbridge->support.masked_set_action; - ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, - ctx->odp_actions, - ctx->wc, use_masked); - - 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->odp_actions, - &ctx->xin->flow, probability, &cookie, - sizeof cookie.flow_sample, ODPP_NONE, - false); + xlate_commit_actions(ctx); + + union user_action_cookie cookie = { + .flow_sample = { + .type = USER_ACTION_COOKIE_FLOW_SAMPLE, + .probability = os->probability, + .collector_set_id = os->collector_set_id, + .obs_domain_id = os->obs_domain_id, + .obs_point_id = os->obs_point_id, + } + }; + compose_sample_action(ctx, probability, &cookie, sizeof cookie.flow_sample, + ODPP_NONE, false); } static bool @@ -4138,6 +4165,9 @@ recirc_unroll_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_WRITE_ACTIONS: case OFPACT_METER: case OFPACT_SAMPLE: + case OFPACT_DEBUG_RECIRC: + case OFPACT_CT: + case OFPACT_NAT: break; /* These need not be copied for restoration. */ @@ -4160,6 +4190,161 @@ recirc_unroll_actions(const struct ofpact *ofpacts, size_t ofpacts_len, CHECK_MPLS_RECIRCULATION(); \ } +static void +put_ct_mark(const struct flow *flow, struct flow *base_flow, + struct ofpbuf *odp_actions, struct flow_wildcards *wc) +{ + struct { + uint32_t key; + uint32_t mask; + } odp_attr; + + odp_attr.key = flow->ct_mark; + odp_attr.mask = wc->masks.ct_mark; + + if (odp_attr.mask && odp_attr.key != base_flow->ct_mark) { + nl_msg_put_unspec(odp_actions, OVS_CT_ATTR_MARK, &odp_attr, + sizeof(odp_attr)); + } +} + +static void +put_ct_label(const struct flow *flow, struct flow *base_flow, + struct ofpbuf *odp_actions, struct flow_wildcards *wc) +{ + if (!ovs_u128_is_zero(&wc->masks.ct_label) + && !ovs_u128_equals(&flow->ct_label, &base_flow->ct_label)) { + struct { + ovs_u128 key; + ovs_u128 mask; + } *odp_ct_label; + + odp_ct_label = nl_msg_put_unspec_uninit(odp_actions, + OVS_CT_ATTR_LABELS, + sizeof(*odp_ct_label)); + odp_ct_label->key = flow->ct_label; + odp_ct_label->mask = wc->masks.ct_label; + } +} + +static void +put_ct_helper(struct ofpbuf *odp_actions, struct ofpact_conntrack *ofc) +{ + if (ofc->alg) { + if (ofc->alg == IPPORT_FTP) { + nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "ftp"); + } else { + VLOG_WARN("Cannot serialize ct_helper %d\n", ofc->alg); + } + } +} + +static void +put_ct_nat(struct xlate_ctx *ctx) +{ + struct ofpact_nat *ofn = ctx->ct_nat_action; + size_t nat_offset; + + if (!ofn) { + return; + } + + nat_offset = nl_msg_start_nested(ctx->odp_actions, OVS_CT_ATTR_NAT); + if (ofn->flags & NX_NAT_F_SRC || ofn->flags & NX_NAT_F_DST) { + nl_msg_put_flag(ctx->odp_actions, ofn->flags & NX_NAT_F_SRC + ? OVS_NAT_ATTR_SRC : OVS_NAT_ATTR_DST); + if (ofn->flags & NX_NAT_F_PERSISTENT) { + nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PERSISTENT); + } + if (ofn->flags & NX_NAT_F_PROTO_HASH) { + nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_HASH); + } else if (ofn->flags & NX_NAT_F_PROTO_RANDOM) { + nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_RANDOM); + } + if (ofn->range_af == AF_INET) { + nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, + ofn->range.addr.ipv4.min); + if (ofn->range.addr.ipv4.max && + (ntohl(ofn->range.addr.ipv4.max) + > ntohl(ofn->range.addr.ipv4.min))) { + nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, + ofn->range.addr.ipv4.max); + } + } else if (ofn->range_af == AF_INET6) { + nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, + &ofn->range.addr.ipv6.min, + sizeof ofn->range.addr.ipv6.min); + if (!ipv6_mask_is_any(&ofn->range.addr.ipv6.max) && + memcmp(&ofn->range.addr.ipv6.max, &ofn->range.addr.ipv6.min, + sizeof ofn->range.addr.ipv6.max) > 0) { + nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, + &ofn->range.addr.ipv6.max, + sizeof ofn->range.addr.ipv6.max); + } + } + if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { + nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MIN, + ofn->range.proto.min); + if (ofn->range.proto.max && + ofn->range.proto.max > ofn->range.proto.min) { + nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MAX, + ofn->range.proto.max); + } + } + } + nl_msg_end_nested(ctx->odp_actions, nat_offset); +} + +static void +compose_conntrack_action(struct xlate_ctx *ctx, struct ofpact_conntrack *ofc) +{ + ovs_u128 old_ct_label = ctx->base_flow.ct_label; + uint32_t old_ct_mark = ctx->base_flow.ct_mark; + size_t ct_offset; + uint16_t zone; + + /* Ensure that any prior actions are applied before composing the new + * conntrack action. */ + xlate_commit_actions(ctx); + + /* Process nested actions first, to populate the key. */ + ctx->ct_nat_action = NULL; + do_xlate_actions(ofc->actions, ofpact_ct_get_action_len(ofc), ctx); + + if (ofc->zone_src.field) { + zone = mf_get_subfield(&ofc->zone_src, &ctx->xin->flow); + } else { + zone = ofc->zone_imm; + } + + ct_offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CT); + if (ofc->flags & NX_CT_F_COMMIT) { + nl_msg_put_flag(ctx->odp_actions, OVS_CT_ATTR_COMMIT); + } + nl_msg_put_u16(ctx->odp_actions, OVS_CT_ATTR_ZONE, zone); + put_ct_mark(&ctx->xin->flow, &ctx->base_flow, ctx->odp_actions, ctx->wc); + put_ct_label(&ctx->xin->flow, &ctx->base_flow, ctx->odp_actions, ctx->wc); + put_ct_helper(ctx->odp_actions, ofc); + put_ct_nat(ctx); + ctx->ct_nat_action = NULL; + nl_msg_end_nested(ctx->odp_actions, ct_offset); + + /* Restore the original ct fields in the key. These should only be exposed + * after recirculation to another table. */ + ctx->base_flow.ct_mark = old_ct_mark; + ctx->base_flow.ct_label = old_ct_label; + + if (ofc->recirc_table == NX_CT_RECIRC_NONE) { + /* If we do not recirculate as part of this action, hide the results of + * connection tracking from subsequent recirculations. */ + ctx->conntracked = false; + } else { + /* Use ct_* fields from datapath during recirculation upcall. */ + ctx->conntracked = true; + compose_recirculate_and_fork(ctx, ofc->recirc_table); + } +} + static void do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, struct xlate_ctx *ctx) @@ -4170,6 +4355,7 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { tnl_arp_snoop(flow, wc, ctx->xbridge->name); + tnl_nd_snoop(flow, wc, ctx->xbridge->name); } /* dl_type already in the mask, not set below. */ @@ -4179,6 +4365,10 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, const struct ofpact_set_field *set_field; const struct mf_field *mf; + if (ctx->error) { + break; + } + if (ctx->exit) { /* Check if need to store the remaining actions for later * execution. */ @@ -4249,13 +4439,13 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_SET_ETH_SRC: - memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); - memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN); + WC_MASK_FIELD(wc, dl_src); + flow->dl_src = ofpact_get_SET_ETH_SRC(a)->mac; break; case OFPACT_SET_ETH_DST: - memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); - memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN); + WC_MASK_FIELD(wc, dl_dst); + flow->dl_dst = ofpact_get_SET_ETH_DST(a)->mac; break; case OFPACT_SET_IPV4_SRC: @@ -4363,9 +4553,9 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, && !eth_type_mpls(flow->dl_type)) { break; } - /* A flow may wildcard nw_frag. Do nothing if setting a trasport + /* A flow may wildcard nw_frag. Do nothing if setting a transport * header field on a packet that does not have them. */ - mf_mask_field_and_prereqs(mf, &wc->masks); + mf_mask_field_and_prereqs(mf, wc); if (mf_are_prereqs_ok(mf, flow)) { mf_set_flow_value_masked(mf, &set_field->value, &set_field->mask, flow); @@ -4483,7 +4673,6 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_FIN_TIMEOUT: CHECK_MPLS_RECIRCULATION(); memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); - ctx->xout->has_fin_timeout = true; xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); break; @@ -4524,11 +4713,26 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, case OFPACT_SAMPLE: xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); break; + + case OFPACT_CT: + CHECK_MPLS_RECIRCULATION(); + compose_conntrack_action(ctx, ofpact_get_CT(a)); + break; + + case OFPACT_NAT: + /* This will be processed by compose_conntrack_action(). */ + ctx->ct_nat_action = ofpact_get_NAT(a); + break; + + case OFPACT_DEBUG_RECIRC: + ctx_trigger_recirculation(ctx); + a = ofpact_next(a); + break; } /* Check if need to store this and the remaining actions for later * execution. */ - if (ctx->exit && ctx_first_recirculation_action(ctx)) { + if (!ctx->error && ctx->exit && ctx_first_recirculation_action(ctx)) { recirc_unroll_actions(a, OFPACT_ALIGN(ofpacts_len - ((uint8_t *)a - (uint8_t *)ofpacts)), @@ -4559,6 +4763,8 @@ xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, xin->resubmit_hook = NULL; xin->report_hook = NULL; xin->resubmit_stats = NULL; + xin->recurse = 0; + xin->resubmits = 0; xin->wc = wc; xin->odp_actions = odp_actions; @@ -4572,7 +4778,7 @@ void xlate_out_uninit(struct xlate_out *xout) { if (xout) { - xlate_out_free_recircs(xout); + recirc_refs_unref(&xout->recircs); } } @@ -4582,8 +4788,15 @@ void xlate_actions_for_side_effects(struct xlate_in *xin) { struct xlate_out xout; + enum xlate_error error; + + error = xlate_actions(xin, &xout); + if (error) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + VLOG_WARN_RL(&rl, "xlate_actions failed (%s)!", xlate_strerror(error)); + } - xlate_actions(xin, &xout); xlate_out_uninit(&xout); } @@ -4662,7 +4875,7 @@ netdev_max_backlog(void) stream = fopen(filename, "r"); if (!stream) { - VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno)); + VLOG_INFO("%s: open failed (%s)", filename, ovs_strerror(errno)); } else { if (fscanf(stream, "%d", &n) != 1) { VLOG_WARN("%s: read error", filename); @@ -4717,28 +4930,79 @@ too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) #endif } +static void +xlate_wc_init(struct xlate_ctx *ctx) +{ + flow_wildcards_init_catchall(ctx->wc); + + /* Some fields we consider to always be examined. */ + WC_MASK_FIELD(ctx->wc, in_port); + WC_MASK_FIELD(ctx->wc, dl_type); + if (is_ip_any(&ctx->xin->flow)) { + WC_MASK_FIELD_MASK(ctx->wc, nw_frag, FLOW_NW_FRAG_MASK); + } + + if (ctx->xbridge->support.odp.recirc) { + /* Always exactly match recirc_id when datapath supports + * recirculation. */ + WC_MASK_FIELD(ctx->wc, recirc_id); + } + + if (ctx->xbridge->netflow) { + netflow_mask_wc(&ctx->xin->flow, ctx->wc); + } + + tnl_wc_init(&ctx->xin->flow, ctx->wc); +} + +static void +xlate_wc_finish(struct xlate_ctx *ctx) +{ + /* Clear the metadata and register wildcard masks, because we won't + * use non-header fields as part of the cache. */ + flow_wildcards_clear_non_packet_fields(ctx->wc); + + /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow + * uses the low 8 bits of the 16-bit tp_src and tp_dst members to + * represent these fields. The datapath interface, on the other hand, + * represents them with just 8 bits each. This means that if the high + * 8 bits of the masks for these fields somehow become set, then they + * will get chopped off by a round trip through the datapath, and + * revalidation will spot that as an inconsistency and delete the flow. + * Avoid the problem here by making sure that only the low 8 bits of + * either field can be unwildcarded for ICMP. + */ + if (is_icmpv4(&ctx->xin->flow) || is_icmpv6(&ctx->xin->flow)) { + ctx->wc->masks.tp_src &= htons(UINT8_MAX); + ctx->wc->masks.tp_dst &= htons(UINT8_MAX); + } + /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ + if (ctx->wc->masks.vlan_tci) { + ctx->wc->masks.vlan_tci |= htons(VLAN_CFI); + } +} + /* Translates the flow, actions, or rule in 'xin' into datapath actions in * 'xout'. * The caller must take responsibility for eventually freeing 'xout', with - * xlate_out_uninit(). */ -void + * xlate_out_uninit(). + * Returns 'XLATE_OK' if translation was successful. In case of an error an + * empty set of actions will be returned in 'xin->odp_actions' (if non-NULL), + * so that most callers may ignore the return value and transparently install a + * drop flow when the translation fails. */ +enum xlate_error xlate_actions(struct xlate_in *xin, struct xlate_out *xout) { *xout = (struct xlate_out) { .slow = 0, .fail_open = false, - .has_learn = false, - .has_normal = false, - .has_fin_timeout = false, - .nf_output_iface = NF_OUT_DROP, - .mirrors = 0, - .n_recircs = 0, + .recircs = RECIRC_REFS_EMPTY_INITIALIZER, }; struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); if (!xbridge) { - return; + return XLATE_BRIDGE_NOT_FOUND; } struct flow *flow = &xin->flow; @@ -4759,8 +5023,8 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) .wc = xin->wc ? xin->wc : &scratch_wc, .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, - .recurse = 0, - .resubmits = 0, + .recurse = xin->recurse, + .resubmits = xin->resubmits, .in_group = false, .in_action_set = false, @@ -4769,92 +5033,79 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) .orig_skb_priority = flow->skb_priority, .sflow_n_outputs = 0, .sflow_odp_port = 0, - .user_cookie_offset = 0, + .nf_output_iface = NF_OUT_DROP, .exit = false, + .error = XLATE_OK, + .mirrors = 0, .recirc_action_offset = -1, .last_unroll_offset = -1, .was_mpls = false, + .conntracked = false, + + .ct_nat_action = NULL, .action_set_has_group = false, .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), }; - memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); - ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); - - struct xport *in_port; - struct flow orig_flow; - bool tnl_may_send; - - COVERAGE_INC(xlate_actions); - /* Flow initialization rules: - * - 'base_flow' must match the kernel's view of the packet at the - * time that action processing starts. 'flow' represents any - * transformations we wish to make through actions. - * - By default 'base_flow' and 'flow' are the same since the input - * packet matches the output before any actions are applied. - * - When using VLAN splinters, 'base_flow''s VLAN is set to the value - * of the received packet as seen by the kernel. If we later output - * to another device without any modifications this will cause us to - * insert a new tag since the original one was stripped off by the - * VLAN device. - * - Tunnel metadata as received is retained in 'flow'. This allows - * tunnel metadata matching also in later tables. - * Since a kernel action for setting the tunnel metadata will only be - * generated with actual tunnel output, changing the tunnel metadata - * values in 'flow' (such as tun_id) will only have effect with a later - * tunnel output action. - * - Tunnel 'base_flow' is completely cleared since that is what the - * kernel does. If we wish to maintain the original values an action - * needs to be generated. */ + /* 'base_flow' reflects the packet as it came in, but we need it to reflect + * the packet as the datapath will treat it for output actions: + * + * - Our datapath doesn't retain tunneling information without us + * re-setting it, so clear the tunnel data. + * + * - For VLAN splinters, a higher layer may pretend that the packet + * came in on 'flow->in_port.ofp_port' with 'flow->vlan_tci' + * attached, because that's how we want to treat it from an OpenFlow + * perspective. But from the datapath's perspective it actually came + * in on a VLAN device without any VLAN attached. So here we put the + * datapath's view of the VLAN information in 'base_flow' to ensure + * correct treatment. + */ + memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); + if (flow->in_port.ofp_port + != vsp_realdev_to_vlandev(xbridge->ofproto, + flow->in_port.ofp_port, + flow->vlan_tci)) { + ctx.base_flow.vlan_tci = 0; + } + ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); if (xin->wc) { - flow_wildcards_init_catchall(ctx.wc); - memset(&ctx.wc->masks.in_port, 0xff, sizeof ctx.wc->masks.in_port); - memset(&ctx.wc->masks.dl_type, 0xff, sizeof ctx.wc->masks.dl_type); - if (is_ip_any(flow)) { - ctx.wc->masks.nw_frag |= FLOW_NW_FRAG_MASK; - } - if (xbridge->support.odp.recirc) { - /* Always exactly match recirc_id when datapath supports - * recirculation. */ - ctx.wc->masks.recirc_id = UINT32_MAX; - } - if (xbridge->netflow) { - netflow_mask_wc(flow, ctx.wc); - } + xlate_wc_init(&ctx); } - tnl_may_send = tnl_xlate_init(flow, xin->wc); - - /* The in_port of the original packet before recirculation. */ - in_port = get_ofp_port(xbridge, flow->in_port.ofp_port); + COVERAGE_INC(xlate_actions); if (xin->recirc) { - const struct recirc_id_node *recirc = xin->recirc; + const struct recirc_state *state = &xin->recirc->state; + + xlate_report(&ctx, "Restoring state post-recirculation:"); if (xin->ofpacts_len > 0 || ctx.rule) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); + const char *conflict = xin->ofpacts_len ? "actions" : "rule"; - VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", - xin->ofpacts_len > 0 - ? "actions" - : "rule"); + VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", conflict); + xlate_report(&ctx, "- Recirculation conflict (%s)!", conflict); + ctx.error = XLATE_RECIRCULATION_CONFLICT; goto exit; } /* Set the bridge for post-recirculation processing if needed. */ - if (ctx.xbridge->ofproto != recirc->ofproto) { + if (ctx.xbridge->ofproto != state->ofproto) { struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); - const struct xbridge *new_bridge = xbridge_lookup(xcfg, - recirc->ofproto); + const struct xbridge *new_bridge + = xbridge_lookup(xcfg, state->ofproto); if (OVS_UNLIKELY(!new_bridge)) { /* Drop the packet if the bridge cannot be found. */ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); VLOG_WARN_RL(&rl, "Recirculation bridge no longer exists."); + xlate_report(&ctx, "- Recirculation bridge no longer exists."); + ctx.error = XLATE_BRIDGE_NOT_FOUND; goto exit; } ctx.xbridge = new_bridge; @@ -4862,26 +5113,36 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) /* Set the post-recirculation table id. Note: A table lookup is done * only if there are no post-recirculation actions. */ - ctx.table_id = recirc->table_id; + ctx.table_id = state->table_id; + xlate_report(&ctx, "- Resuming from table %"PRIu8, ctx.table_id); + + if (!state->conntracked) { + clear_conntrack(flow); + } /* Restore pipeline metadata. May change flow's in_port and other * metadata to the values that existed when recirculation was * triggered. */ - recirc_metadata_to_flow(&recirc->metadata, flow); + recirc_metadata_to_flow(&state->metadata, flow); /* Restore stack, if any. */ - if (recirc->stack) { - ofpbuf_put(&ctx.stack, recirc->stack->data, recirc->stack->size); + if (state->stack) { + ofpbuf_put(&ctx.stack, state->stack->data, state->stack->size); } + /* Restore mirror state. */ + ctx.mirrors = state->mirrors; + /* Restore action set, if any. */ - if (recirc->action_set_len) { + if (state->action_set_len) { const struct ofpact *a; - ofpbuf_put(&ctx.action_set, recirc->ofpacts, - recirc->action_set_len); + xlate_report_actions(&ctx, "- Restoring action set", + state->ofpacts, state->action_set_len); - OFPACT_FOR_EACH(a, recirc->ofpacts, recirc->action_set_len) { + ofpbuf_put(&ctx.action_set, state->ofpacts, state->action_set_len); + + OFPACT_FOR_EACH(a, state->ofpacts, state->action_set_len) { if (a->type == OFPACT_GROUP) { ctx.action_set_has_group = true; break; @@ -4891,16 +5152,20 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) /* Restore recirculation actions. If there are no actions, processing * will start with a lookup in the table set above. */ - if (recirc->ofpacts_len > recirc->action_set_len) { - xin->ofpacts_len = recirc->ofpacts_len - recirc->action_set_len; - xin->ofpacts = recirc->ofpacts + - recirc->action_set_len / sizeof *recirc->ofpacts; + if (state->ofpacts_len > state->action_set_len) { + xin->ofpacts_len = state->ofpacts_len - state->action_set_len; + xin->ofpacts = state->ofpacts + + state->action_set_len / sizeof *state->ofpacts; + + xlate_report_actions(&ctx, "- Restoring actions", + xin->ofpacts, xin->ofpacts_len); } } else if (OVS_UNLIKELY(flow->recirc_id)) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); VLOG_WARN_RL(&rl, "Recirculation context not found for ID %"PRIx32, flow->recirc_id); + ctx.error = XLATE_NO_RECIRCULATION_CONTEXT; goto exit; } /* The bridge is now known so obtain its table version. */ @@ -4909,7 +5174,7 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) if (!xin->ofpacts && !ctx.rule) { ctx.rule = rule_dpif_lookup_from_table( ctx.xbridge->ofproto, ctx.tables_version, flow, xin->wc, - ctx.xin->xcache != NULL, ctx.xin->resubmit_stats, &ctx.table_id, + ctx.xin->resubmit_stats, &ctx.table_id, flow->in_port.ofp_port, true, true); if (ctx.xin->resubmit_stats) { rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats); @@ -4919,6 +5184,7 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); entry->u.rule = ctx.rule; + rule_dpif_ref(ctx.rule); } if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { @@ -4927,11 +5193,10 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) } xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); - if (mbridge_has_mirrors(xbridge->mbridge)) { - /* Do this conditionally because the copy is expensive enough that it - * shows up in profiles. */ - orig_flow = *flow; - } + /* Get the proximate input port of the packet. (If xin->recirc, + * flow->in_port is the ultimate input port of the packet.) */ + struct xport *in_port = get_ofp_port(xbridge, + ctx.base_flow.in_port.ofp_port); /* Tunnel stats only for non-recirculated packets. */ if (!xin->recirc && in_port && in_port->is_tunnel) { @@ -4950,28 +5215,30 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) } } - /* Do not perform special processing on recirculated packets, - * as recirculated packets are not really received by the bridge. */ - if (xin->recirc || !process_special(&ctx, in_port)) { - size_t sample_actions_len; - - if (flow->in_port.ofp_port - != vsp_realdev_to_vlandev(xbridge->ofproto, - flow->in_port.ofp_port, - flow->vlan_tci)) { - ctx.base_flow.vlan_tci = 0; + if (!xin->recirc && process_special(&ctx, in_port)) { + /* process_special() did all the processing for this packet. + * + * We do not perform special processing on recirculated packets, as + * recirculated packets are not really received by the bridge.*/ + } else if (in_port && in_port->xbundle + && xbundle_mirror_out(xbridge, in_port->xbundle)) { + 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 received on port " + "%s, which is reserved exclusively for mirroring", + ctx.xbridge->name, in_port->xbundle->name); } - + } else { /* Sampling is done only for packets really received by the bridge. */ + unsigned int user_cookie_offset = 0; if (!xin->recirc) { - add_sflow_action(&ctx); - add_ipfix_action(&ctx); - sample_actions_len = ctx.odp_actions->size; - } else { - sample_actions_len = 0; + user_cookie_offset = compose_sflow_action(&ctx); + compose_ipfix_action(&ctx, ODPP_NONE); } + size_t sample_actions_len = ctx.odp_actions->size; - if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) { + if (tnl_process_ecn(flow) + && (!in_port || may_receive(in_port, &ctx))) { const struct ofpact *ofpacts; size_t ofpacts_len; @@ -4988,7 +5255,11 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) OVS_NOT_REACHED(); } + mirror_ingress_packet(&ctx); do_xlate_actions(ofpacts, ofpacts_len, &ctx); + if (ctx.error) { + goto exit; + } /* We've let OFPP_NORMAL and the learning action look at the * packet, so drop it now if forwarding is disabled. */ @@ -5024,13 +5295,8 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) compose_output_action(&ctx, OFPP_LOCAL, NULL); } - if (!xin->recirc) { - fix_sflow_action(&ctx); - } - /* Only mirror fully processed packets. */ - if (!exit_recirculates(&ctx) - && mbridge_has_mirrors(xbridge->mbridge)) { - add_mirror_actions(&ctx, &orig_flow); + if (user_cookie_offset) { + fix_sflow_action(&ctx, user_cookie_offset); } } @@ -5046,29 +5312,13 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) ctx.xout->slow |= SLOW_ACTION; } - /* Update mirror stats only for packets really received by the bridge. */ - if (!xin->recirc && mbridge_has_mirrors(xbridge->mbridge)) { - if (ctx.xin->resubmit_stats) { - mirror_update_stats(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(xbridge->mbridge); - entry->u.mirror.mirrors = xout->mirrors; - } - } - /* Do netflow only for packets really received by the bridge and not sent * to the controller. We consider packets sent to the controller to be * part of the control plane rather than the data plane. */ if (!xin->recirc && xbridge->netflow && !(xout->slow & SLOW_CONTROLLER)) { if (ctx.xin->resubmit_stats) { netflow_flow_update(xbridge->netflow, flow, - xout->nf_output_iface, + ctx.nf_output_iface, ctx.xin->resubmit_stats); } if (ctx.xin->xcache) { @@ -5077,39 +5327,27 @@ xlate_actions(struct xlate_in *xin, struct xlate_out *xout) entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); entry->u.nf.netflow = netflow_ref(xbridge->netflow); entry->u.nf.flow = xmemdup(flow, sizeof *flow); - entry->u.nf.iface = xout->nf_output_iface; + entry->u.nf.iface = ctx.nf_output_iface; } } if (xin->wc) { - /* Clear the metadata and register wildcard masks, because we won't - * use non-header fields as part of the cache. */ - flow_wildcards_clear_non_packet_fields(ctx.wc); - - /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow - * uses the low 8 bits of the 16-bit tp_src and tp_dst members to - * represent these fields. The datapath interface, on the other hand, - * represents them with just 8 bits each. This means that if the high - * 8 bits of the masks for these fields somehow become set, then they - * will get chopped off by a round trip through the datapath, and - * revalidation will spot that as an inconsistency and delete the flow. - * Avoid the problem here by making sure that only the low 8 bits of - * either field can be unwildcarded for ICMP. - */ - if (is_icmpv4(flow) || is_icmpv6(flow)) { - ctx.wc->masks.tp_src &= htons(UINT8_MAX); - ctx.wc->masks.tp_dst &= htons(UINT8_MAX); - } - /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ - if (ctx.wc->masks.vlan_tci) { - ctx.wc->masks.vlan_tci |= htons(VLAN_CFI); - } + xlate_wc_finish(&ctx); } exit: ofpbuf_uninit(&ctx.stack); ofpbuf_uninit(&ctx.action_set); ofpbuf_uninit(&scratch_actions); + + /* Make sure we return a "drop flow" in case of an error. */ + if (ctx.error) { + xout->slow = 0; + if (xin->odp_actions) { + ofpbuf_clear(xin->odp_actions); + } + } + return ctx.error; } /* Sends 'packet' out 'ofport'. @@ -5203,7 +5441,7 @@ xlate_push_stats(struct xlate_cache *xcache, { struct xc_entry *entry; struct ofpbuf entries = xcache->entries; - uint8_t dmac[ETH_ADDR_LEN]; + struct eth_addr dmac; if (!stats->n_packets) { return; @@ -5247,7 +5485,8 @@ xlate_push_stats(struct xlate_cache *xcache, 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); + tnl_arp_lookup(entry->u.tnl_arp_cache.br_name, + entry->u.tnl_arp_cache.d_ip, &dmac); break; default: OVS_NOT_REACHED();