X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fpackets.c;h=57291674ac5db6ce38d6bb0379e93dc2560f994c;hb=41792464296d4ea9393adff3eea7bef514655cba;hp=60ee3903d9e34aeab3cec63479e42d39ed877899;hpb=07a6cf771dc049600033453d700ea93b9449bd33;p=openvswitch diff --git a/lib/packets.c b/lib/packets.c index 60ee3903..57291674 100644 --- a/lib/packets.c +++ b/lib/packets.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010 Nicira Networks. + * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -18,9 +18,12 @@ #include "packets.h" #include #include +#include #include #include #include "byte-order.h" +#include "csum.h" +#include "flow.h" #include "dynamic-string.h" #include "ofpbuf.h" @@ -40,6 +43,71 @@ dpid_from_string(const char *s, uint64_t *dpidp) return *dpidp != 0; } +/* Returns true if 'ea' is a reserved multicast address, that a bridge must + * never forward, false otherwise. Includes some proprietary vendor protocols + * that shouldn't be forwarded as well. + * + * If you change this function's behavior, please update corresponding + * documentation in vswitch.xml at the same time. */ +bool +eth_addr_is_reserved(const uint8_t ea[ETH_ADDR_LEN]) +{ + struct masked_eth_addr { + uint8_t ea[ETH_ADDR_LEN]; + uint8_t mask[ETH_ADDR_LEN]; + }; + + static struct masked_eth_addr mea[] = { + { /* STP, IEEE pause frames, and other reserved protocols. */ + {0x01, 0x08, 0xc2, 0x00, 0x00, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xff, 0xf0}}, + + { /* VRRP IPv4. */ + {0x00, 0x00, 0x5e, 0x00, 0x01, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}}, + + { /* VRRP IPv6. */ + {0x00, 0x00, 0x5e, 0x00, 0x02, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}}, + + { /* HSRPv1. */ + {0x00, 0x00, 0x0c, 0x07, 0xac, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xff, 0x00}}, + + { /* HSRPv2. */ + {0x00, 0x00, 0x0c, 0x9f, 0xf0, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xf0, 0x00}}, + + { /* GLBP. */ + {0x00, 0x07, 0xb4, 0x00, 0x00, 0x00}, + {0xff, 0xff, 0xff, 0x00, 0x00, 0x00}}, + + { /* Extreme Discovery Protocol. */ + {0x00, 0xE0, 0x2B, 0x00, 0x00, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xf0, 0x00}}, + + { /* Cisco Inter Switch Link. */ + {0x01, 0x00, 0x0c, 0x00, 0x00, 0x00}, + {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}}, + + { /* Cisco protocols plus others following the same pattern: + * + * CDP, VTP, DTP, PAgP (01-00-0c-cc-cc-cc) + * Spanning Tree PVSTP+ (01-00-0c-cc-cc-cd) + * STP Uplink Fast (01-00-0c-cd-cd-cd) */ + {0x01, 0x00, 0x0c, 0xcc, 0xcc, 0xcc}, + {0xff, 0xff, 0xff, 0xfe, 0xfe, 0xfe}}}; + + size_t i; + + for (i = 0; i < ARRAY_SIZE(mea); i++) { + if (eth_addr_equal_except(ea, mea[i].ea, mea[i].mask)) { + return true; + } + } + return false; +} + bool eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN]) { @@ -52,44 +120,156 @@ eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN]) } } -/* Fills 'b' with an 802.2 SNAP packet with Ethernet source address 'eth_src', - * the Nicira OUI as SNAP organization and 'snap_type' as SNAP type. The text - * string in 'tag' is enclosed as the packet payload. - * +/* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'. * This function is used by Open vSwitch to compose packets in cases where - * context is important but content doesn't (or shouldn't) matter. For this - * purpose, 'snap_type' should be a random number and 'tag' should be an - * English phrase that explains the purpose of the packet. (The English phrase - * gives hapless admins running Wireshark the opportunity to figure out what's - * going on.) */ + * context is important but content doesn't (or shouldn't) matter. + * + * The returned packet has enough headroom to insert an 802.1Q VLAN header if + * desired. */ void -compose_benign_packet(struct ofpbuf *b, const char *tag, uint16_t snap_type, - const uint8_t eth_src[ETH_ADDR_LEN]) +compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN]) { struct eth_header *eth; - struct llc_snap_header *llc_snap; + struct rarp_header *rarp; - /* Compose basic packet structure. (We need the payload size to stick into - * the 802.2 header.) */ ofpbuf_clear(b); - eth = ofpbuf_put_zeros(b, ETH_HEADER_LEN); - llc_snap = ofpbuf_put_zeros(b, LLC_SNAP_HEADER_LEN); - ofpbuf_put(b, tag, strlen(tag) + 1); /* Includes null byte. */ - ofpbuf_put(b, eth_src, ETH_ADDR_LEN); - - /* Compose 802.2 header. */ + ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + + RARP_HEADER_LEN); + ofpbuf_reserve(b, VLAN_HEADER_LEN); + eth = ofpbuf_put_uninit(b, sizeof *eth); memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN); memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN); - eth->eth_type = htons(b->size - ETH_HEADER_LEN); - - /* Compose LLC, SNAP headers. */ - llc_snap->llc.llc_dsap = LLC_DSAP_SNAP; - llc_snap->llc.llc_ssap = LLC_SSAP_SNAP; - llc_snap->llc.llc_cntl = LLC_CNTL_SNAP; - memcpy(llc_snap->snap.snap_org, "\x00\x23\x20", 3); - llc_snap->snap.snap_type = htons(snap_type); + eth->eth_type = htons(ETH_TYPE_RARP); + + rarp = ofpbuf_put_uninit(b, sizeof *rarp); + rarp->hw_addr_space = htons(ARP_HTYPE_ETH); + rarp->proto_addr_space = htons(ETH_TYPE_IP); + rarp->hw_addr_length = ETH_ADDR_LEN; + rarp->proto_addr_length = sizeof rarp->src_proto_addr; + rarp->opcode = htons(RARP_REQUEST_REVERSE); + memcpy(rarp->src_hw_addr, eth_src, ETH_ADDR_LEN); + rarp->src_proto_addr = htonl(0); + memcpy(rarp->target_hw_addr, eth_src, ETH_ADDR_LEN); + rarp->target_proto_addr = htonl(0); +} + +/* Insert VLAN header according to given TCI. Packet passed must be Ethernet + * packet. Ignores the CFI bit of 'tci' using 0 instead. + * + * Also sets 'packet->l2' to point to the new Ethernet header. */ +void +eth_push_vlan(struct ofpbuf *packet, ovs_be16 tci) +{ + struct eth_header *eh = packet->data; + struct vlan_eth_header *veh; + + /* Insert new 802.1Q header. */ + struct vlan_eth_header tmp; + memcpy(tmp.veth_dst, eh->eth_dst, ETH_ADDR_LEN); + memcpy(tmp.veth_src, eh->eth_src, ETH_ADDR_LEN); + tmp.veth_type = htons(ETH_TYPE_VLAN); + tmp.veth_tci = tci & htons(~VLAN_CFI); + tmp.veth_next_type = eh->eth_type; + + veh = ofpbuf_push_uninit(packet, VLAN_HEADER_LEN); + memcpy(veh, &tmp, sizeof tmp); + + packet->l2 = packet->data; +} + +/* Removes outermost VLAN header (if any is present) from 'packet'. + * + * 'packet->l2' must initially point to 'packet''s Ethernet header. */ +void +eth_pop_vlan(struct ofpbuf *packet) +{ + struct vlan_eth_header *veh = packet->l2; + if (packet->size >= sizeof *veh + && veh->veth_type == htons(ETH_TYPE_VLAN)) { + struct eth_header tmp; + + memcpy(tmp.eth_dst, veh->veth_dst, ETH_ADDR_LEN); + memcpy(tmp.eth_src, veh->veth_src, ETH_ADDR_LEN); + tmp.eth_type = veh->veth_next_type; + + ofpbuf_pull(packet, VLAN_HEADER_LEN); + packet->l2 = (char*)packet->l2 + VLAN_HEADER_LEN; + memcpy(packet->data, &tmp, sizeof tmp); + } +} + +/* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The + * caller must free '*packetp'. On success, returns NULL. On failure, returns + * an error message and stores NULL in '*packetp'. */ +const char * +eth_from_hex(const char *hex, struct ofpbuf **packetp) +{ + struct ofpbuf *packet; + + packet = *packetp = ofpbuf_new(strlen(hex) / 2); + + if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') { + ofpbuf_delete(packet); + *packetp = NULL; + return "Trailing garbage in packet data"; + } + + if (packet->size < ETH_HEADER_LEN) { + ofpbuf_delete(packet); + *packetp = NULL; + return "Packet data too short for Ethernet"; + } + + return NULL; +} + +void +eth_format_masked(const uint8_t eth[ETH_ADDR_LEN], + const uint8_t mask[ETH_ADDR_LEN], struct ds *s) +{ + ds_put_format(s, ETH_ADDR_FMT, ETH_ADDR_ARGS(eth)); + if (mask && !eth_mask_is_exact(mask)) { + ds_put_format(s, "/"ETH_ADDR_FMT, ETH_ADDR_ARGS(mask)); + } +} + +void +eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN], + const uint8_t mask[ETH_ADDR_LEN], + uint8_t dst[ETH_ADDR_LEN]) +{ + int i; + + for (i = 0; i < ETH_ADDR_LEN; i++) { + dst[i] = src[i] & mask[i]; + } } +/* Given the IP netmask 'netmask', returns the number of bits of the IP address + * that it specifies, that is, the number of 1-bits in 'netmask'. + * + * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will + * still be in the valid range but isn't otherwise meaningful. */ +int +ip_count_cidr_bits(ovs_be32 netmask) +{ + return 32 - ctz(ntohl(netmask)); +} + +void +ip_format_masked(ovs_be32 ip, ovs_be32 mask, struct ds *s) +{ + ds_put_format(s, IP_FMT, IP_ARGS(&ip)); + if (mask != htonl(UINT32_MAX)) { + if (ip_is_cidr(mask)) { + ds_put_format(s, "/%d", ip_count_cidr_bits(mask)); + } else { + ds_put_format(s, "/"IP_FMT, IP_ARGS(&mask)); + } + } +} + + /* Stores the string representation of the IPv6 address 'addr' into the * character array 'addr_str', which must be at least INET6_ADDRSTRLEN * bytes long. */ @@ -102,10 +282,29 @@ format_ipv6_addr(char *addr_str, const struct in6_addr *addr) void print_ipv6_addr(struct ds *string, const struct in6_addr *addr) { - char addr_str[INET6_ADDRSTRLEN]; + char *dst; + + ds_reserve(string, string->length + INET6_ADDRSTRLEN); - format_ipv6_addr(addr_str, addr); - ds_put_format(string, "%s", addr_str); + dst = string->string + string->length; + format_ipv6_addr(dst, addr); + string->length += strlen(dst); +} + +void +print_ipv6_masked(struct ds *s, const struct in6_addr *addr, + const struct in6_addr *mask) +{ + print_ipv6_addr(s, addr); + if (mask && !ipv6_mask_is_exact(mask)) { + if (ipv6_is_cidr(mask)) { + int cidr_bits = ipv6_count_cidr_bits(mask); + ds_put_format(s, "/%d", cidr_bits); + } else { + ds_put_char(s, '/'); + print_ipv6_addr(s, mask); + } + } } struct in6_addr ipv6_addr_bitand(const struct in6_addr *a, @@ -149,9 +348,12 @@ ipv6_create_mask(int mask) return netmask; } -/* Given the IPv6 netmask 'netmask', returns the number of bits of the - * IPv6 address that it wildcards. 'netmask' must be a CIDR netmask (see - * ipv6_is_cidr()). */ +/* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6 + * address that it specifies, that is, the number of 1-bits in 'netmask'. + * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()). + * + * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value + * will still be in the valid range but isn't otherwise meaningful. */ int ipv6_count_cidr_bits(const struct in6_addr *netmask) { @@ -159,8 +361,6 @@ ipv6_count_cidr_bits(const struct in6_addr *netmask) int count = 0; const uint8_t *netmaskp = &netmask->s6_addr[0]; - assert(ipv6_is_cidr(netmask)); - for (i=0; i<16; i++) { if (netmaskp[i] == 0xff) { count += 8; @@ -178,7 +378,6 @@ ipv6_count_cidr_bits(const struct in6_addr *netmask) return count; } - /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N * high-order 1-bits and 128-N low-order 0-bits. */ bool @@ -204,65 +403,189 @@ ipv6_is_cidr(const struct in6_addr *netmask) return true; } -/* Populates 'b' with a LACP packet containing 'pdu' with source address - * 'eth_src'. */ -void -compose_lacp_packet(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN], - const struct lacp_pdu *pdu) +/* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst', + * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated + * in 'b' and returned. This payload may be populated with appropriate + * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the + * Ethernet header and payload respectively. + * + * The returned packet has enough headroom to insert an 802.1Q VLAN header if + * desired. */ +void * +eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN], + const uint8_t eth_src[ETH_ADDR_LEN], uint16_t eth_type, + size_t size) { + void *data; struct eth_header *eth; - struct lacp_pdu *eth_pdu; ofpbuf_clear(b); - ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + LACP_PDU_LEN); - eth = ofpbuf_put_zeros(b, ETH_HEADER_LEN); - eth_pdu = ofpbuf_put(b, pdu, LACP_PDU_LEN); + ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + size); + ofpbuf_reserve(b, VLAN_HEADER_LEN); + eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN); + data = ofpbuf_put_uninit(b, size); - memcpy(eth->eth_dst, eth_addr_lacp, ETH_ADDR_LEN); + memcpy(eth->eth_dst, eth_dst, ETH_ADDR_LEN); memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN); - eth->eth_type = htons(ETH_TYPE_LACP); + eth->eth_type = htons(eth_type); + + b->l2 = eth; + b->l3 = data; + + return data; +} + +static void +packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr) +{ + struct ip_header *nh = packet->l3; + + if (nh->ip_proto == IPPROTO_TCP && packet->l7) { + struct tcp_header *th = packet->l4; + + th->tcp_csum = recalc_csum32(th->tcp_csum, *addr, new_addr); + } else if (nh->ip_proto == IPPROTO_UDP && packet->l7) { + struct udp_header *uh = packet->l4; + + if (uh->udp_csum) { + uh->udp_csum = recalc_csum32(uh->udp_csum, *addr, new_addr); + if (!uh->udp_csum) { + uh->udp_csum = htons(0xffff); + } + } + } + nh->ip_csum = recalc_csum32(nh->ip_csum, *addr, new_addr); + *addr = new_addr; } -/* Populates 'pdu' with a LACP PDU comprised of 'actor' and 'partner'. */ +/* Modifies the IPv4 header fields of 'packet' to be consistent with 'src', + * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate. + * 'packet' must contain a valid IPv4 packet with correctly populated l[347] + * markers. */ void -compose_lacp_pdu(const struct lacp_info *actor, - const struct lacp_info *partner, struct lacp_pdu *pdu) +packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst, + uint8_t tos, uint8_t ttl) { - memset(pdu, 0, sizeof *pdu); + struct ip_header *nh = packet->l3; - pdu->subtype = 1; - pdu->version = 1; + if (nh->ip_src != src) { + packet_set_ipv4_addr(packet, &nh->ip_src, src); + } - pdu->actor_type = 1; - pdu->actor_len = 20; - pdu->actor = *actor; + if (nh->ip_dst != dst) { + packet_set_ipv4_addr(packet, &nh->ip_dst, dst); + } + + if (nh->ip_tos != tos) { + uint8_t *field = &nh->ip_tos; + + nh->ip_csum = recalc_csum16(nh->ip_csum, htons((uint16_t) *field), + htons((uint16_t) tos)); + *field = tos; + } - pdu->partner_type = 2; - pdu->partner_len = 20; - pdu->partner = *partner; + if (nh->ip_ttl != ttl) { + uint8_t *field = &nh->ip_ttl; - pdu->collector_type = 3; - pdu->collector_len = 16; - pdu->collector_delay = htons(0); + nh->ip_csum = recalc_csum16(nh->ip_csum, htons(*field << 8), + htons(ttl << 8)); + *field = ttl; + } } -/* Parses 'b' which represents a packet containing a LACP PDU. This function - * returns NULL if 'b' is malformed, or does not represent a LACP PDU format - * supported by OVS. Otherwise, it returns a pointer to the lacp_pdu contained - * within 'b'. */ -const struct lacp_pdu * -parse_lacp_packet(const struct ofpbuf *b) +static void +packet_set_port(ovs_be16 *port, ovs_be16 new_port, ovs_be16 *csum) { - const struct lacp_pdu *pdu; + if (*port != new_port) { + *csum = recalc_csum16(*csum, *port, new_port); + *port = new_port; + } +} - pdu = ofpbuf_at(b, (uint8_t *)b->l3 - (uint8_t *)b->data, LACP_PDU_LEN); +/* Sets the TCP source and destination port ('src' and 'dst' respectively) of + * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet + * with its l4 marker properly populated. */ +void +packet_set_tcp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst) +{ + struct tcp_header *th = packet->l4; + + packet_set_port(&th->tcp_src, src, &th->tcp_csum); + packet_set_port(&th->tcp_dst, dst, &th->tcp_csum); +} + +/* Sets the UDP source and destination port ('src' and 'dst' respectively) of + * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet + * with its l4 marker properly populated. */ +void +packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst) +{ + struct udp_header *uh = packet->l4; + + if (uh->udp_csum) { + packet_set_port(&uh->udp_src, src, &uh->udp_csum); + packet_set_port(&uh->udp_dst, dst, &uh->udp_csum); - if (pdu && pdu->subtype == 1 - && pdu->actor_type == 1 && pdu->actor_len == 20 - && pdu->partner_type == 2 && pdu->partner_len == 20) { - return pdu; + if (!uh->udp_csum) { + uh->udp_csum = htons(0xffff); + } + } else { + uh->udp_src = src; + uh->udp_dst = dst; + } +} + +/* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0. + * + * 'flow' must be the flow corresponding to 'packet' and 'packet''s header + * pointers must be properly initialized (e.g. with flow_extract()). */ +uint8_t +packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow) +{ + if ((flow->dl_type == htons(ETH_TYPE_IP) || + flow->dl_type == htons(ETH_TYPE_IPV6)) && + flow->nw_proto == IPPROTO_TCP && packet->l7) { + const struct tcp_header *tcp = packet->l4; + return TCP_FLAGS(tcp->tcp_ctl); } else { - return NULL; + return 0; + } +} + +/* Appends a string representation of the TCP flags value 'tcp_flags' + * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the + * format used by tcpdump. */ +void +packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags) +{ + if (!tcp_flags) { + ds_put_cstr(s, "none"); + return; + } + + if (tcp_flags & TCP_SYN) { + ds_put_char(s, 'S'); + } + if (tcp_flags & TCP_FIN) { + ds_put_char(s, 'F'); + } + if (tcp_flags & TCP_PSH) { + ds_put_char(s, 'P'); + } + if (tcp_flags & TCP_RST) { + ds_put_char(s, 'R'); + } + if (tcp_flags & TCP_URG) { + ds_put_char(s, 'U'); + } + if (tcp_flags & TCP_ACK) { + ds_put_char(s, '.'); + } + if (tcp_flags & 0x40) { + ds_put_cstr(s, "[40]"); + } + if (tcp_flags & 0x80) { + ds_put_cstr(s, "[80]"); } }