X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=ofproto%2Fin-band.c;h=42b3efbed43c367efa561599842dd8f22fc797f8;hb=HEAD;hp=d918d7134b8514a0ecfcd6ad88db765c7e22df9b;hpb=5dbdfff733b4833397feba64047dabac9bb36531;p=sliver-openvswitch.git diff --git a/ofproto/in-band.c b/ofproto/in-band.c index d918d7134..42b3efbed 100644 --- a/ofproto/in-band.c +++ b/ofproto/in-band.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010 Nicira Networks. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -19,174 +19,27 @@ #include #include #include +#include #include #include #include +#include "classifier.h" #include "dhcp.h" -#include "dpif.h" #include "flow.h" -#include "mac-learning.h" #include "netdev.h" +#include "netlink.h" #include "odp-util.h" -#include "ofp-print.h" +#include "ofp-actions.h" #include "ofproto.h" #include "ofpbuf.h" +#include "ofproto-provider.h" #include "openflow/openflow.h" -#include "openvswitch/datapath-protocol.h" #include "packets.h" #include "poll-loop.h" -#include "rconn.h" -#include "status.h" #include "timeval.h" -#include "vconn.h" - -#define THIS_MODULE VLM_in_band #include "vlog.h" -/* In-band control allows a single network to be used for OpenFlow - * traffic and other data traffic. Refer to ovs-vswitchd.conf(5) and - * secchan(8) for a description of configuring in-band control. - * - * This comment is an attempt to describe how in-band control works at a - * wire- and implementation-level. Correctly implementing in-band - * control has proven difficult due to its many subtleties, and has thus - * gone through many iterations. Please read through and understand the - * reasoning behind the chosen rules before making modifications. - * - * In Open vSwitch, in-band control is implemented as "hidden" flows (in - * that they are not visible through OpenFlow) and at a higher priority - * than wildcarded flows can be set up by the controller. This is done - * so that the controller cannot interfere with them and possibly break - * connectivity with its switches. It is possible to see all flows, - * including in-band ones, with the ovs-appctl "bridge/dump-flows" - * command. - * - * The following rules are always enabled with the "normal" action by a - * switch with in-band control: - * - * a. DHCP requests sent from the local port. - * b. ARP replies to the local port's MAC address. - * c. ARP requests from the local port's MAC address. - * d. ARP replies to the remote side's MAC address. Note that the - * remote side is either the controller or the gateway to reach - * the controller. - * e. ARP requests from the remote side's MAC address. Note that - * like (d), the MAC is either for the controller or gateway. - * f. ARP replies containing the controller's IP address as a target. - * g. ARP requests containing the controller's IP address as a source. - * h. OpenFlow (6633/tcp) traffic to the controller's IP. - * i. OpenFlow (6633/tcp) traffic from the controller's IP. - * - * The goal of these rules is to be as narrow as possible to allow a - * switch to join a network and be able to communicate with a - * controller. As mentioned earlier, these rules have higher priority - * than the controller's rules, so if they are too broad, they may - * prevent the controller from implementing its policy. As such, - * in-band actively monitors some aspects of flow and packet processing - * so that the rules can be made more precise. - * - * In-band control monitors attempts to add flows into the datapath that - * could interfere with its duties. The datapath only allows exact - * match entries, so in-band control is able to be very precise about - * the flows it prevents. Flows that miss in the datapath are sent to - * userspace to be processed, so preventing these flows from being - * cached in the "fast path" does not affect correctness. The only type - * of flow that is currently prevented is one that would prevent DHCP - * replies from being seen by the local port. For example, a rule that - * forwarded all DHCP traffic to the controller would not be allowed, - * but one that forwarded to all ports (including the local port) would. - * - * As mentioned earlier, packets that miss in the datapath are sent to - * the userspace for processing. The userspace has its own flow table, - * the "classifier", so in-band checks whether any special processing - * is needed before the classifier is consulted. If a packet is a DHCP - * response to a request from the local port, the packet is forwarded to - * the local port, regardless of the flow table. Note that this requires - * L7 processing of DHCP replies to determine whether the 'chaddr' field - * matches the MAC address of the local port. - * - * It is interesting to note that for an L3-based in-band control - * mechanism, the majority of rules are devoted to ARP traffic. At first - * glance, some of these rules appear redundant. However, each serves an - * important role. First, in order to determine the MAC address of the - * remote side (controller or gateway) for other ARP rules, we must allow - * ARP traffic for our local port with rules (b) and (c). If we are - * between a switch and its connection to the controller, we have to - * allow the other switch's ARP traffic to through. This is done with - * rules (d) and (e), since we do not know the addresses of the other - * switches a priori, but do know the controller's or gateway's. Finally, - * if the controller is running in a local guest VM that is not reached - * through the local port, the switch that is connected to the VM must - * allow ARP traffic based on the controller's IP address, since it will - * not know the MAC address of the local port that is sending the traffic - * or the MAC address of the controller in the guest VM. - * - * With a few notable exceptions below, in-band should work in most - * network setups. The following are considered "supported' in the - * current implementation: - * - * - Locally Connected. The switch and controller are on the same - * subnet. This uses rules (a), (b), (c), (h), and (i). - * - * - Reached through Gateway. The switch and controller are on - * different subnets and must go through a gateway. This uses - * rules (a), (b), (c), (h), and (i). - * - * - Between Switch and Controller. This switch is between another - * switch and the controller, and we want to allow the other - * switch's traffic through. This uses rules (d), (e), (h), and - * (i). It uses (b) and (c) indirectly in order to know the MAC - * address for rules (d) and (e). Note that DHCP for the other - * switch will not work unless the controller explicitly lets this - * switch pass the traffic. - * - * - Between Switch and Gateway. This switch is between another - * switch and the gateway, and we want to allow the other switch's - * traffic through. This uses the same rules and logic as the - * "Between Switch and Controller" configuration described earlier. - * - * - Controller on Local VM. The controller is a guest VM on the - * system running in-band control. This uses rules (a), (b), (c), - * (h), and (i). - * - * - Controller on Local VM with Different Networks. The controller - * is a guest VM on the system running in-band control, but the - * local port is not used to connect to the controller. For - * example, an IP address is configured on eth0 of the switch. The - * controller's VM is connected through eth1 of the switch, but an - * IP address has not been configured for that port on the switch. - * As such, the switch will use eth0 to connect to the controller, - * and eth1's rules about the local port will not work. In the - * example, the switch attached to eth0 would use rules (a), (b), - * (c), (h), and (i) on eth0. The switch attached to eth1 would use - * rules (f), (g), (h), and (i). - * - * The following are explicitly *not* supported by in-band control: - * - * - Specify Controller by Name. Currently, the controller must be - * identified by IP address. A naive approach would be to permit - * all DNS traffic. Unfortunately, this would prevent the - * controller from defining any policy over DNS. Since switches - * that are located behind us need to connect to the controller, - * in-band cannot simply add a rule that allows DNS traffic from - * the local port. The "correct" way to support this is to parse - * DNS requests to allow all traffic related to a request for the - * controller's name through. Due to the potential security - * problems and amount of processing, we decided to hold off for - * the time-being. - * - * - Differing Controllers for Switches. All switches must know - * the L3 addresses for all the controllers that other switches - * may use, since rules need to be set up to allow traffic related - * to those controllers through. See rules (f), (g), (h), and (i). - * - * - Differing Routes for Switches. In order for the switch to - * allow other switches to connect to a controller through a - * gateway, it allows the gateway's traffic through with rules (d) - * and (e). If the routes to the controller differ for the two - * switches, we will not know the MAC address of the alternate - * gateway. - */ +VLOG_DEFINE_THIS_MODULE(in_band); /* Priorities used in classifier for in-band rules. These values are higher * than any that may be set with OpenFlow, and "18" kind of looks like "IB". @@ -194,35 +47,49 @@ * by in-band control have the same action. The only reason to use more than * one priority is to make the kind of flow easier to see during debugging. */ enum { + /* One set per bridge. */ IBR_FROM_LOCAL_DHCP = 180000, /* (a) From local port, DHCP. */ IBR_TO_LOCAL_ARP, /* (b) To local port, ARP. */ IBR_FROM_LOCAL_ARP, /* (c) From local port, ARP. */ - IBR_TO_REMOTE_ARP, /* (d) To remote MAC, ARP. */ - IBR_FROM_REMOTE_ARP, /* (e) From remote MAC, ARP. */ - IBR_TO_CTL_ARP, /* (f) To controller IP, ARP. */ - IBR_FROM_CTL_ARP, /* (g) From controller IP, ARP. */ - IBR_TO_CTL_OFP, /* (h) To controller, OpenFlow port. */ - IBR_FROM_CTL_OFP /* (i) From controller, OpenFlow port. */ -}; -struct in_band_rule { - flow_t flow; - uint32_t wildcards; - unsigned int priority; + /* One set per unique next-hop MAC. */ + IBR_TO_NEXT_HOP_ARP, /* (d) To remote MAC, ARP. */ + IBR_FROM_NEXT_HOP_ARP, /* (e) From remote MAC, ARP. */ + + /* One set per unique remote IP address. */ + IBR_TO_REMOTE_ARP, /* (f) To remote IP, ARP. */ + IBR_FROM_REMOTE_ARP, /* (g) From remote IP, ARP. */ + + /* One set per unique remote (IP,port) pair. */ + IBR_TO_REMOTE_TCP, /* (h) To remote IP, TCP port. */ + IBR_FROM_REMOTE_TCP /* (i) From remote IP, TCP port. */ }; /* Track one remote IP and next hop information. */ struct in_band_remote { - struct rconn *rconn; /* Connection to remote. */ - uint32_t remote_ip; /* Remote IP, 0 if unknown. */ + struct sockaddr_in remote_addr; /* IP address, in network byte order. */ uint8_t remote_mac[ETH_ADDR_LEN]; /* Next-hop MAC, all-zeros if unknown. */ uint8_t last_remote_mac[ETH_ADDR_LEN]; /* Previous nonzero next-hop MAC. */ struct netdev *remote_netdev; /* Device to send to next-hop MAC. */ }; +/* What to do to an in_band_rule. */ +enum in_band_op { + ADD, /* Add the rule to ofproto's flow table. */ + DEL /* Delete the rule from ofproto's flow table. */ +}; + +/* A rule to add to or delete from ofproto's flow table. */ +struct in_band_rule { + struct hmap_node hmap_node; /* In struct in_band's "rules" hmap. */ + struct match match; + unsigned int priority; + enum in_band_op op; +}; + struct in_band { struct ofproto *ofproto; - struct status_category *ss_cat; + int queue_id; /* Remote information. */ time_t next_remote_refresh; /* Refresh timer. */ @@ -234,12 +101,8 @@ struct in_band { uint8_t local_mac[ETH_ADDR_LEN]; /* Current MAC. */ struct netdev *local_netdev; /* Local port's network device. */ - /* Local and remote addresses that are installed as flows. */ - uint8_t installed_local_mac[ETH_ADDR_LEN]; - uint32_t *remote_ips; - uint32_t n_remote_ips; - uint8_t *remote_macs; - size_t n_remote_macs; + /* Flow tracking. */ + struct hmap rules; /* Contains "struct in_band_rule"s. */ }; static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 60); @@ -247,47 +110,36 @@ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 60); static int refresh_remote(struct in_band *ib, struct in_band_remote *r) { - struct in_addr remote_inaddr; struct in_addr next_hop_inaddr; char *next_hop_dev; int retval; - memset(r->remote_mac, 0, sizeof r->remote_mac); - - /* Get remote IP address. */ - r->remote_ip = rconn_get_remote_ip(r->rconn); - if (!r->remote_ip) { - /* No remote IP address means that this rconn is probably either - * configured for a non-IP based protocol (e.g. "unix:") or - * misconfigured entirely. No point in refreshing quickly. */ - return 10; - } - /* Find the next-hop IP address. */ - remote_inaddr.s_addr = r->remote_ip; - retval = netdev_get_next_hop(ib->local_netdev, &remote_inaddr, + memset(r->remote_mac, 0, sizeof r->remote_mac); + retval = netdev_get_next_hop(ib->local_netdev, &r->remote_addr.sin_addr, &next_hop_inaddr, &next_hop_dev); if (retval) { VLOG_WARN("cannot find route for controller ("IP_FMT"): %s", - IP_ARGS(&r->remote_ip), strerror(retval)); + IP_ARGS(r->remote_addr.sin_addr.s_addr), + ovs_strerror(retval)); return 1; } if (!next_hop_inaddr.s_addr) { - next_hop_inaddr.s_addr = remote_inaddr.s_addr; + next_hop_inaddr = r->remote_addr.sin_addr; } - /* Get the next-hop IP and network device. */ + /* Open the next-hop network device. */ if (!r->remote_netdev || strcmp(netdev_get_name(r->remote_netdev), next_hop_dev)) { netdev_close(r->remote_netdev); - retval = netdev_open_default(next_hop_dev, &r->remote_netdev); + retval = netdev_open(next_hop_dev, "system", &r->remote_netdev); if (retval) { VLOG_WARN_RL(&rl, "cannot open netdev %s (next hop " "to controller "IP_FMT"): %s", - next_hop_dev, IP_ARGS(&r->remote_ip), - strerror(retval)); + next_hop_dev, IP_ARGS(r->remote_addr.sin_addr.s_addr), + ovs_strerror(retval)); free(next_hop_dev); return 1; } @@ -299,7 +151,7 @@ refresh_remote(struct in_band *ib, struct in_band_remote *r) r->remote_mac); if (retval) { VLOG_DBG_RL(&rl, "cannot look up remote MAC address ("IP_FMT"): %s", - IP_ARGS(&next_hop_inaddr.s_addr), strerror(retval)); + IP_ARGS(next_hop_inaddr.s_addr), ovs_strerror(retval)); } /* If we don't have a MAC address, then refresh quickly, since we probably @@ -371,413 +223,217 @@ refresh_local(struct in_band *ib) return true; } -static void -in_band_status_cb(struct status_reply *sr, void *in_band_) -{ - struct in_band *in_band = in_band_; - - if (!eth_addr_is_zero(in_band->local_mac)) { - status_reply_put(sr, "local-mac="ETH_ADDR_FMT, - ETH_ADDR_ARGS(in_band->local_mac)); - } - - if (in_band->n_remotes - && !eth_addr_is_zero(in_band->remotes[0].remote_mac)) { - status_reply_put(sr, "remote-mac="ETH_ADDR_FMT, - ETH_ADDR_ARGS(in_band->remotes[0].remote_mac)); - } -} - -/* Returns true if 'packet' should be sent to the local port regardless - * of the flow table. */ +/* Returns true if packets in 'flow' should be directed to the local port. + * (This keeps the flow table from preventing DHCP replies from being seen by + * the local port.) */ bool -in_band_msg_in_hook(struct in_band *in_band, const flow_t *flow, - const struct ofpbuf *packet) +in_band_must_output_to_local_port(const struct flow *flow) { - if (!in_band) { - return false; - } - - /* Regardless of how the flow table is configured, we want to be - * able to see replies to our DHCP requests. */ - if (flow->dl_type == htons(ETH_TYPE_IP) - && flow->nw_proto == IP_TYPE_UDP + return (flow->dl_type == htons(ETH_TYPE_IP) + && flow->nw_proto == IPPROTO_UDP && flow->tp_src == htons(DHCP_SERVER_PORT) - && flow->tp_dst == htons(DHCP_CLIENT_PORT) - && packet->l7) { - struct dhcp_header *dhcp; - - dhcp = ofpbuf_at(packet, (char *)packet->l7 - (char *)packet->data, - sizeof *dhcp); - if (!dhcp) { - return false; - } - - refresh_local(in_band); - if (!eth_addr_is_zero(in_band->local_mac) - && eth_addr_equals(dhcp->chaddr, in_band->local_mac)) { - return true; - } - } - - return false; + && flow->tp_dst == htons(DHCP_CLIENT_PORT)); } -/* Returns true if the rule that would match 'flow' with 'actions' is - * allowed to be set up in the datapath. */ -bool -in_band_rule_check(struct in_band *in_band, const flow_t *flow, - const struct odp_actions *actions) +static void +add_rule(struct in_band *ib, const struct match *match, unsigned int priority) { - if (!in_band) { - return true; - } + uint32_t hash = match_hash(match, 0); + struct in_band_rule *rule; - /* Don't allow flows that would prevent DHCP replies from being seen - * by the local port. */ - if (flow->dl_type == htons(ETH_TYPE_IP) - && flow->nw_proto == IP_TYPE_UDP - && flow->tp_src == htons(DHCP_SERVER_PORT) - && flow->tp_dst == htons(DHCP_CLIENT_PORT)) { - int i; - - for (i=0; in_actions; i++) { - if (actions->actions[i].output.type == ODPAT_OUTPUT - && actions->actions[i].output.port == ODPP_LOCAL) { - return true; - } + HMAP_FOR_EACH_WITH_HASH (rule, hmap_node, hash, &ib->rules) { + if (match_equal(&rule->match, match)) { + rule->op = ADD; + return; } - return false; } - return true; -} - -static void -init_rule(struct in_band_rule *rule, unsigned int priority) -{ - rule->wildcards = OVSFW_ALL; + rule = xmalloc(sizeof *rule); + rule->match = *match; rule->priority = priority; - - /* Not strictly necessary but seems cleaner. */ - memset(&rule->flow, 0, sizeof rule->flow); -} - -static void -set_in_port(struct in_band_rule *rule, uint16_t odp_port) -{ - rule->wildcards &= ~OFPFW_IN_PORT; - rule->flow.in_port = odp_port; -} - -static void -set_dl_type(struct in_band_rule *rule, uint16_t dl_type) -{ - rule->wildcards &= ~OFPFW_DL_TYPE; - rule->flow.dl_type = htons(dl_type); -} - -static void -set_dl_src(struct in_band_rule *rule, const uint8_t dl_src[ETH_ADDR_LEN]) -{ - rule->wildcards &= ~OFPFW_DL_SRC; - memcpy(rule->flow.dl_src, dl_src, ETH_ADDR_LEN); -} - -static void -set_dl_dst(struct in_band_rule *rule, const uint8_t dl_dst[ETH_ADDR_LEN]) -{ - rule->wildcards &= ~OFPFW_DL_DST; - memcpy(rule->flow.dl_dst, dl_dst, ETH_ADDR_LEN); -} - -static void -set_tp_src(struct in_band_rule *rule, uint16_t tp_src) -{ - rule->wildcards &= ~OFPFW_TP_SRC; - rule->flow.tp_src = htons(tp_src); -} - -static void -set_tp_dst(struct in_band_rule *rule, uint16_t tp_dst) -{ - rule->wildcards &= ~OFPFW_TP_DST; - rule->flow.tp_dst = htons(tp_dst); -} - -static void -set_nw_proto(struct in_band_rule *rule, uint8_t nw_proto) -{ - rule->wildcards &= ~OFPFW_NW_PROTO; - rule->flow.nw_proto = nw_proto; + rule->op = ADD; + hmap_insert(&ib->rules, &rule->hmap_node, hash); } static void -set_nw_src(struct in_band_rule *rule, uint32_t nw_src) +update_rules(struct in_band *ib) { - rule->wildcards &= ~OFPFW_NW_SRC_MASK; - rule->flow.nw_src = nw_src; -} - -static void -set_nw_dst(struct in_band_rule *rule, uint32_t nw_dst) -{ - rule->wildcards &= ~OFPFW_NW_DST_MASK; - rule->flow.nw_dst = nw_dst; -} - -static void -make_rules(struct in_band *ib, - void (*cb)(struct in_band *, const struct in_band_rule *)) -{ - struct in_band_rule rule; - size_t i; + struct in_band_rule *ib_rule; + struct in_band_remote *r; + struct match match; - if (!eth_addr_is_zero(ib->installed_local_mac)) { - /* Allow DHCP requests to be sent from the local port. */ - init_rule(&rule, IBR_FROM_LOCAL_DHCP); - set_in_port(&rule, ODPP_LOCAL); - set_dl_type(&rule, ETH_TYPE_IP); - set_dl_src(&rule, ib->installed_local_mac); - set_nw_proto(&rule, IP_TYPE_UDP); - set_tp_src(&rule, DHCP_CLIENT_PORT); - set_tp_dst(&rule, DHCP_SERVER_PORT); - cb(ib, &rule); - - /* Allow the connection's interface to receive directed ARP traffic. */ - init_rule(&rule, IBR_TO_LOCAL_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_dl_dst(&rule, ib->installed_local_mac); - set_nw_proto(&rule, ARP_OP_REPLY); - cb(ib, &rule); - - /* Allow the connection's interface to be the source of ARP traffic. */ - init_rule(&rule, IBR_FROM_LOCAL_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_dl_src(&rule, ib->installed_local_mac); - set_nw_proto(&rule, ARP_OP_REQUEST); - cb(ib, &rule); + /* Mark all the existing rules for deletion. (Afterward we will re-add any + * rules that are still valid.) */ + HMAP_FOR_EACH (ib_rule, hmap_node, &ib->rules) { + ib_rule->op = DEL; } - for (i = 0; i < ib->n_remote_macs; i++) { - const uint8_t *remote_mac = &ib->remote_macs[i * ETH_ADDR_LEN]; - - if (i > 0) { - const uint8_t *prev_mac = &ib->remote_macs[(i - 1) * ETH_ADDR_LEN]; - if (eth_addr_equals(remote_mac, prev_mac)) { - /* Skip duplicates. */ - continue; - } - } - - /* Allow ARP replies to the remote side's MAC. */ - init_rule(&rule, IBR_TO_REMOTE_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_dl_dst(&rule, remote_mac); - set_nw_proto(&rule, ARP_OP_REPLY); - cb(ib, &rule); - - /* Allow ARP requests from the remote side's MAC. */ - init_rule(&rule, IBR_FROM_REMOTE_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_dl_src(&rule, remote_mac); - set_nw_proto(&rule, ARP_OP_REQUEST); - cb(ib, &rule); + if (ib->n_remotes && !eth_addr_is_zero(ib->local_mac)) { + /* (a) Allow DHCP requests sent from the local port. */ + match_init_catchall(&match); + match_set_in_port(&match, OFPP_LOCAL); + match_set_dl_type(&match, htons(ETH_TYPE_IP)); + match_set_dl_src(&match, ib->local_mac); + match_set_nw_proto(&match, IPPROTO_UDP); + match_set_tp_src(&match, htons(DHCP_CLIENT_PORT)); + match_set_tp_dst(&match, htons(DHCP_SERVER_PORT)); + add_rule(ib, &match, IBR_FROM_LOCAL_DHCP); + + /* (b) Allow ARP replies to the local port's MAC address. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_dl_dst(&match, ib->local_mac); + match_set_nw_proto(&match, ARP_OP_REPLY); + add_rule(ib, &match, IBR_TO_LOCAL_ARP); + + /* (c) Allow ARP requests from the local port's MAC address. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_dl_src(&match, ib->local_mac); + match_set_nw_proto(&match, ARP_OP_REQUEST); + add_rule(ib, &match, IBR_FROM_LOCAL_ARP); } - for (i = 0; i < ib->n_remote_ips; i++) { - uint32_t remote_ip = ib->remote_ips[i]; + for (r = ib->remotes; r < &ib->remotes[ib->n_remotes]; r++) { + const uint8_t *remote_mac = r->remote_mac; - if (i > 0 && ib->remote_ips[i - 1] == remote_ip) { - /* Skip duplicates. */ + if (eth_addr_is_zero(remote_mac)) { continue; } - /* Allow ARP replies to the controller's IP. */ - init_rule(&rule, IBR_TO_CTL_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_nw_proto(&rule, ARP_OP_REPLY); - set_nw_dst(&rule, remote_ip); - cb(ib, &rule); - - /* Allow ARP requests from the controller's IP. */ - init_rule(&rule, IBR_FROM_CTL_ARP); - set_dl_type(&rule, ETH_TYPE_ARP); - set_nw_proto(&rule, ARP_OP_REQUEST); - set_nw_src(&rule, remote_ip); - cb(ib, &rule); - - /* OpenFlow traffic to the controller. */ - init_rule(&rule, IBR_TO_CTL_OFP); - set_dl_type(&rule, ETH_TYPE_IP); - set_nw_proto(&rule, IP_TYPE_TCP); - set_nw_dst(&rule, remote_ip); - set_tp_dst(&rule, OFP_TCP_PORT); - cb(ib, &rule); - - /* OpenFlow traffic from the controller. */ - init_rule(&rule, IBR_FROM_CTL_OFP); - set_dl_type(&rule, ETH_TYPE_IP); - set_nw_proto(&rule, IP_TYPE_TCP); - set_nw_src(&rule, remote_ip); - set_tp_src(&rule, OFP_TCP_PORT); - cb(ib, &rule); + /* (d) Allow ARP replies to the next hop's MAC address. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_dl_dst(&match, remote_mac); + match_set_nw_proto(&match, ARP_OP_REPLY); + add_rule(ib, &match, IBR_TO_NEXT_HOP_ARP); + + /* (e) Allow ARP requests from the next hop's MAC address. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_dl_src(&match, remote_mac); + match_set_nw_proto(&match, ARP_OP_REQUEST); + add_rule(ib, &match, IBR_FROM_NEXT_HOP_ARP); } -} - -static void -clear_rules(struct in_band *ib) -{ - memset(ib->installed_local_mac, 0, sizeof ib->installed_local_mac); - - free(ib->remote_ips); - ib->remote_ips = NULL; - ib->n_remote_ips = 0; - - free(ib->remote_macs); - ib->remote_macs = NULL; - ib->n_remote_macs = 0; -} -static void -drop_rule(struct in_band *ib, const struct in_band_rule *rule) -{ - ofproto_delete_flow(ib->ofproto, &rule->flow, - rule->wildcards, rule->priority); -} - -static void -drop_rules(struct in_band *ib) -{ - make_rules(ib, drop_rule); - clear_rules(ib); -} - -static void -add_rule(struct in_band *ib, const struct in_band_rule *rule) -{ - union ofp_action action; - - action.type = htons(OFPAT_OUTPUT); - action.output.len = htons(sizeof action); - action.output.port = htons(OFPP_NORMAL); - action.output.max_len = htons(0); - ofproto_add_flow(ib->ofproto, &rule->flow, rule->wildcards, - rule->priority, &action, 1, 0); -} - -static void -add_rules(struct in_band *ib) -{ - make_rules(ib, add_rule); -} - -static int -compare_ips(const void *a, const void *b) -{ - return memcmp(a, b, sizeof(uint32_t)); -} - -static int -compare_macs(const void *a, const void *b) -{ - return memcmp(a, b, ETH_ADDR_LEN); + for (r = ib->remotes; r < &ib->remotes[ib->n_remotes]; r++) { + const struct sockaddr_in *a = &r->remote_addr; + + /* (f) Allow ARP replies containing the remote's IP address as a + * target. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_nw_proto(&match, ARP_OP_REPLY); + match_set_nw_dst(&match, a->sin_addr.s_addr); + add_rule(ib, &match, IBR_TO_REMOTE_ARP); + + /* (g) Allow ARP requests containing the remote's IP address as a + * source. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_ARP)); + match_set_nw_proto(&match, ARP_OP_REQUEST); + match_set_nw_src(&match, a->sin_addr.s_addr); + add_rule(ib, &match, IBR_FROM_REMOTE_ARP); + + /* (h) Allow TCP traffic to the remote's IP and port. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_IP)); + match_set_nw_proto(&match, IPPROTO_TCP); + match_set_nw_dst(&match, a->sin_addr.s_addr); + match_set_tp_dst(&match, a->sin_port); + add_rule(ib, &match, IBR_TO_REMOTE_TCP); + + /* (i) Allow TCP traffic from the remote's IP and port. */ + match_init_catchall(&match); + match_set_dl_type(&match, htons(ETH_TYPE_IP)); + match_set_nw_proto(&match, IPPROTO_TCP); + match_set_nw_src(&match, a->sin_addr.s_addr); + match_set_tp_src(&match, a->sin_port); + add_rule(ib, &match, IBR_FROM_REMOTE_TCP); + } } -void +/* Updates the OpenFlow flow table for the current state of in-band control. + * Returns true ordinarily. Returns false if no remotes are configured on 'ib' + * and 'ib' doesn't have any rules left to remove from the OpenFlow flow + * table. Thus, a false return value means that the caller can destroy 'ib' + * without leaving extra flows hanging around in the flow table. */ +bool in_band_run(struct in_band *ib) { - struct in_band_remote *r; - bool local_change, remote_change; + uint64_t ofpacts_stub[128 / 8]; + struct ofpbuf ofpacts; - local_change = refresh_local(ib); - remote_change = refresh_remotes(ib); - if (!local_change && !remote_change) { - /* Nothing changed, nothing to do. */ - return; - } + struct in_band_rule *rule, *next; - /* Drop old rules. */ - drop_rules(ib); + ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); - /* Figure out new rules. */ - memcpy(ib->installed_local_mac, ib->local_mac, ETH_ADDR_LEN); - ib->remote_ips = xmalloc(ib->n_remotes * sizeof *ib->remote_ips); - ib->n_remote_ips = 0; - ib->remote_macs = xmalloc(ib->n_remotes * ETH_ADDR_LEN); - ib->n_remote_macs = 0; - for (r = ib->remotes; r < &ib->remotes[ib->n_remotes]; r++) { - if (r->remote_ip) { - ib->remote_ips[ib->n_remote_ips++] = r->remote_ip; - } - if (!eth_addr_is_zero(r->remote_mac)) { - memcpy(&ib->remote_macs[ib->n_remote_macs * ETH_ADDR_LEN], - r->remote_mac, ETH_ADDR_LEN); - ib->n_remote_macs++; + if (ib->queue_id >= 0) { + ofpact_put_SET_QUEUE(&ofpacts)->queue_id = ib->queue_id; + } + ofpact_put_OUTPUT(&ofpacts)->port = OFPP_NORMAL; + + refresh_local(ib); + refresh_remotes(ib); + + update_rules(ib); + + HMAP_FOR_EACH_SAFE (rule, next, hmap_node, &ib->rules) { + switch (rule->op) { + case ADD: + ofproto_add_flow(ib->ofproto, &rule->match, rule->priority, + ofpbuf_data(&ofpacts), ofpbuf_size(&ofpacts)); + break; + + case DEL: + if (ofproto_delete_flow(ib->ofproto, + &rule->match, rule->priority)) { + /* ofproto doesn't have the rule anymore so there's no reason + * for us to track it any longer. */ + hmap_remove(&ib->rules, &rule->hmap_node); + free(rule); + } + break; } } - /* Sort, to allow make_rules() to easily skip duplicates. */ - qsort(ib->remote_ips, ib->n_remote_ips, sizeof *ib->remote_ips, - compare_ips); - qsort(ib->remote_macs, ib->n_remote_macs, ETH_ADDR_LEN, compare_macs); + ofpbuf_uninit(&ofpacts); - /* Add new rules. */ - add_rules(ib); + return ib->n_remotes || !hmap_is_empty(&ib->rules); } void in_band_wait(struct in_band *in_band) { - time_t now = time_now(); - time_t wakeup + long long int wakeup = MIN(in_band->next_remote_refresh, in_band->next_local_refresh); - if (wakeup > now) { - poll_timer_wait((wakeup - now) * 1000); - } else { - poll_immediate_wake(); - } -} - -/* ofproto has flushed all flows from the flow table and it is calling us back - * to allow us to reinstall the ones that are important to us. */ -void -in_band_flushed(struct in_band *in_band) -{ - add_rules(in_band); + poll_timer_wait_until(wakeup * 1000); } int -in_band_create(struct ofproto *ofproto, struct dpif *dpif, - struct switch_status *ss, struct in_band **in_bandp) +in_band_create(struct ofproto *ofproto, const char *local_name, + struct in_band **in_bandp) { struct in_band *in_band; - char local_name[IF_NAMESIZE]; struct netdev *local_netdev; int error; - error = dpif_port_get_name(dpif, ODPP_LOCAL, - local_name, sizeof local_name); - if (error) { - VLOG_ERR("failed to initialize in-band control: cannot get name " - "of datapath local port (%s)", strerror(error)); - return error; - } - - error = netdev_open_default(local_name, &local_netdev); + *in_bandp = NULL; + error = netdev_open(local_name, "internal", &local_netdev); if (error) { VLOG_ERR("failed to initialize in-band control: cannot open " - "datapath local port %s (%s)", local_name, strerror(error)); + "datapath local port %s (%s)", + local_name, ovs_strerror(error)); return error; } in_band = xzalloc(sizeof *in_band); in_band->ofproto = ofproto; - in_band->ss_cat = switch_status_register(ss, "in-band", - in_band_status_cb, in_band); + in_band->queue_id = -1; in_band->next_remote_refresh = TIME_MIN; in_band->next_local_refresh = TIME_MIN; in_band->local_netdev = local_netdev; + hmap_init(&in_band->rules); *in_bandp = in_band; @@ -788,41 +444,75 @@ void in_band_destroy(struct in_band *ib) { if (ib) { - drop_rules(ib); + struct in_band_rule *rule, *next; + + HMAP_FOR_EACH_SAFE (rule, next, hmap_node, &ib->rules) { + hmap_remove(&ib->rules, &rule->hmap_node); + free(rule); + } + hmap_destroy(&ib->rules); in_band_set_remotes(ib, NULL, 0); - switch_status_unregister(ib->ss_cat); netdev_close(ib->local_netdev); free(ib); } } -void -in_band_set_remotes(struct in_band *ib, struct rconn **remotes, size_t n) +static bool +any_addresses_changed(struct in_band *ib, + const struct sockaddr_in *addresses, size_t n) { size_t i; - /* Optimize the case where the rconns are the same as last time. */ - if (n == ib->n_remotes) { - for (i = 0; i < n; i++) { - if (ib->remotes[i].rconn != remotes[i]) { - goto different; - } + if (n != ib->n_remotes) { + return true; + } + + for (i = 0; i < n; i++) { + const struct sockaddr_in *old = &ib->remotes[i].remote_addr; + const struct sockaddr_in *new = &addresses[i]; + + if (old->sin_addr.s_addr != new->sin_addr.s_addr || + old->sin_port != new->sin_port) { + return true; } - return; + } + + return false; +} - different:; +void +in_band_set_remotes(struct in_band *ib, + const struct sockaddr_in *addresses, size_t n) +{ + size_t i; + + if (!any_addresses_changed(ib, addresses, n)) { + return; } + /* Clear old remotes. */ for (i = 0; i < ib->n_remotes; i++) { - /* We don't own the rconn. */ netdev_close(ib->remotes[i].remote_netdev); } free(ib->remotes); - ib->next_remote_refresh = TIME_MIN; - ib->remotes = n ? xzalloc(n * sizeof *ib->remotes) : 0; + /* Set up new remotes. */ + ib->remotes = n ? xzalloc(n * sizeof *ib->remotes) : NULL; ib->n_remotes = n; for (i = 0; i < n; i++) { - ib->remotes[i].rconn = remotes[i]; + ib->remotes[i].remote_addr = addresses[i]; } + + /* Force refresh in next call to in_band_run(). */ + ib->next_remote_refresh = TIME_MIN; } + +/* Sets the OpenFlow queue used by flows set up by 'ib' to 'queue_id'. If + * 'queue_id' is negative, 'ib' will not set any queue (which is also the + * default). */ +void +in_band_set_queue(struct in_band *ib, int queue_id) +{ + ib->queue_id = queue_id; +} +