X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fpackets.c;h=d87aa8eb34019f9ed00cf9c4b9fc47baded420fe;hb=d578065e7d9dd70ca6a81ddbc5e4ecb763d1152e;hp=9c30b95c9b5465f75b507e15831aa317bbe54da9;hpb=8c45d00f567159f9b7bf31fcc0862eabbd190b3a;p=sliver-openvswitch.git diff --git a/lib/packets.c b/lib/packets.c index 9c30b95c9..d87aa8eb3 100644 --- a/lib/packets.c +++ b/lib/packets.c @@ -23,11 +23,13 @@ #include #include "byte-order.h" #include "csum.h" +#include "crc32c.h" #include "flow.h" #include "hmap.h" #include "dynamic-string.h" #include "ofpbuf.h" #include "ovs-thread.h" +#include "unaligned.h" const struct in6_addr in6addr_exact = IN6ADDR_EXACT_INIT; @@ -128,8 +130,7 @@ eth_addr_is_reserved(const uint8_t ea[ETH_ADDR_LEN]) bool eth_addr_from_string(const char *s, uint8_t ea[ETH_ADDR_LEN]) { - if (sscanf(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea)) - == ETH_ADDR_SCAN_COUNT) { + if (ovs_scan(s, ETH_ADDR_SCAN_FMT, ETH_ADDR_SCAN_ARGS(ea))) { return true; } else { memset(ea, 0, ETH_ADDR_LEN); @@ -150,9 +151,9 @@ compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN]) struct arp_eth_header *arp; ofpbuf_clear(b); - ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + ARP_ETH_HEADER_LEN); - ofpbuf_reserve(b, VLAN_HEADER_LEN); + ofpbuf_reserve(b, 2 + 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); @@ -165,9 +166,9 @@ compose_rarp(struct ofpbuf *b, const uint8_t eth_src[ETH_ADDR_LEN]) arp->ar_pln = sizeof arp->ar_spa; arp->ar_op = htons(ARP_OP_RARP); memcpy(arp->ar_sha, eth_src, ETH_ADDR_LEN); - arp->ar_spa = htonl(0); + put_16aligned_be32(&arp->ar_spa, htonl(0)); memcpy(arp->ar_tha, eth_src, ETH_ADDR_LEN); - arp->ar_tpa = htonl(0); + put_16aligned_be32(&arp->ar_tpa, htonl(0)); } /* Insert VLAN header according to given TCI. Packet passed must be Ethernet @@ -216,32 +217,6 @@ eth_pop_vlan(struct ofpbuf *packet) } } -/* Return depth of mpls stack. - * - * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header - * or may be NULL if there are no MPLS headers. */ -uint16_t -eth_mpls_depth(const struct ofpbuf *packet) -{ - struct mpls_hdr *mh = packet->l2_5; - uint16_t depth; - - if (!mh) { - return 0; - } - - depth = 0; - while (packet->size >= ((char *)mh - (char *)packet->data) + sizeof *mh) { - depth++; - if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) { - break; - } - mh++; - } - - return depth; -} - /* Set ethertype of the packet. */ void set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type) @@ -250,7 +225,8 @@ set_ethertype(struct ofpbuf *packet, ovs_be16 eth_type) if (eh->eth_type == htons(ETH_TYPE_VLAN)) { ovs_be16 *p; - p = (ovs_be16 *)((char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2); + p = ALIGNED_CAST(ovs_be16 *, + (char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2); *p = eth_type; } else { eh->eth_type = eth_type; @@ -393,13 +369,16 @@ pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype) /* 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'. */ + * an error message and stores NULL in '*packetp'. + * + * Aligns the L3 header of '*packetp' on a 32-bit boundary. */ const char * eth_from_hex(const char *hex, struct ofpbuf **packetp) { struct ofpbuf *packet; - packet = *packetp = ofpbuf_new(strlen(hex) / 2); + /* Use 2 bytes of headroom to 32-bit align the L3 header. */ + packet = *packetp = ofpbuf_new_with_headroom(strlen(hex) / 2, 2); if (ofpbuf_put_hex(packet, hex, NULL)[0] != '\0') { ofpbuf_delete(packet); @@ -446,14 +425,14 @@ eth_addr_bitand(const uint8_t src[ETH_ADDR_LEN], int ip_count_cidr_bits(ovs_be32 netmask) { - return 32 - ctz(ntohl(netmask)); + return 32 - ctz32(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 (mask != OVS_BE32_MAX) { if (ip_is_cidr(mask)) { ds_put_format(s, "/%d", ip_count_cidr_bits(mask)); } else { @@ -600,7 +579,8 @@ ipv6_is_cidr(const struct in6_addr *netmask) * '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. + * Ethernet header and payload respectively. Aligns b->l3 on a 32-bit + * boundary. * * The returned packet has enough headroom to insert an 802.1Q VLAN header if * desired. */ @@ -614,8 +594,10 @@ eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN], ofpbuf_clear(b); - ofpbuf_prealloc_tailroom(b, ETH_HEADER_LEN + VLAN_HEADER_LEN + size); - ofpbuf_reserve(b, VLAN_HEADER_LEN); + /* The magic 2 here ensures that the L3 header (when it is added later) + * will be 32-bit aligned. */ + ofpbuf_prealloc_tailroom(b, 2 + ETH_HEADER_LEN + VLAN_HEADER_LEN + size); + ofpbuf_reserve(b, 2 + VLAN_HEADER_LEN); eth = ofpbuf_put_uninit(b, ETH_HEADER_LEN); data = ofpbuf_put_uninit(b, size); @@ -630,26 +612,28 @@ eth_compose(struct ofpbuf *b, const uint8_t eth_dst[ETH_ADDR_LEN], } static void -packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr) +packet_set_ipv4_addr(struct ofpbuf *packet, + ovs_16aligned_be32 *addr, ovs_be32 new_addr) { struct ip_header *nh = packet->l3; + ovs_be32 old_addr = get_16aligned_be32(addr); 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); + th->tcp_csum = recalc_csum32(th->tcp_csum, old_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); + uh->udp_csum = recalc_csum32(uh->udp_csum, old_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; + nh->ip_csum = recalc_csum32(nh->ip_csum, old_addr, new_addr); + put_16aligned_be32(addr, new_addr); } /* Returns true, if packet contains at least one routing header where @@ -659,7 +643,7 @@ packet_set_ipv4_addr(struct ofpbuf *packet, ovs_be32 *addr, ovs_be32 new_addr) static bool packet_rh_present(struct ofpbuf *packet) { - const struct ip6_hdr *nh; + const struct ovs_16aligned_ip6_hdr *nh; int nexthdr; size_t len; size_t remaining; @@ -670,7 +654,7 @@ packet_rh_present(struct ofpbuf *packet) if (remaining < sizeof *nh) { return false; } - nh = (struct ip6_hdr *)data; + nh = ALIGNED_CAST(struct ovs_16aligned_ip6_hdr *, data); data += sizeof *nh; remaining -= sizeof *nh; nexthdr = nh->ip6_nxt; @@ -706,7 +690,8 @@ packet_rh_present(struct ofpbuf *packet) nexthdr = ext_hdr->ip6e_nxt; len = (ext_hdr->ip6e_len + 2) * 4; } else if (nexthdr == IPPROTO_FRAGMENT) { - const struct ip6_frag *frag_hdr = (struct ip6_frag *)data; + const struct ovs_16aligned_ip6_frag *frag_hdr + = ALIGNED_CAST(struct ovs_16aligned_ip6_frag *, data); nexthdr = frag_hdr->ip6f_nxt; len = sizeof *frag_hdr; @@ -738,7 +723,7 @@ packet_rh_present(struct ofpbuf *packet) static void packet_update_csum128(struct ofpbuf *packet, uint8_t proto, - ovs_be32 addr[4], const ovs_be32 new_addr[4]) + ovs_16aligned_be32 addr[4], const ovs_be32 new_addr[4]) { if (proto == IPPROTO_TCP && packet->l7) { struct tcp_header *th = packet->l4; @@ -758,25 +743,29 @@ packet_update_csum128(struct ofpbuf *packet, uint8_t proto, static void packet_set_ipv6_addr(struct ofpbuf *packet, uint8_t proto, - struct in6_addr *addr, const ovs_be32 new_addr[4], + ovs_16aligned_be32 *addr, const ovs_be32 new_addr[4], bool recalculate_csum) { if (recalculate_csum) { - packet_update_csum128(packet, proto, (ovs_be32 *)addr, new_addr); + packet_update_csum128(packet, proto, addr, new_addr); } memcpy(addr, new_addr, sizeof(*addr)); } static void -packet_set_ipv6_flow_label(ovs_be32 *flow_label, ovs_be32 flow_key) +packet_set_ipv6_flow_label(ovs_16aligned_be32 *flow_label, ovs_be32 flow_key) { - *flow_label = (*flow_label & htonl(~IPV6_LABEL_MASK)) | flow_key; + ovs_be32 old_label = get_16aligned_be32(flow_label); + ovs_be32 new_label = (old_label & htonl(~IPV6_LABEL_MASK)) | flow_key; + put_16aligned_be32(flow_label, new_label); } static void -packet_set_ipv6_tc(ovs_be32 *flow_label, uint8_t tc) +packet_set_ipv6_tc(ovs_16aligned_be32 *flow_label, uint8_t tc) { - *flow_label = (*flow_label & htonl(0xF00FFFFF)) | htonl(tc << 20); + ovs_be32 old_label = get_16aligned_be32(flow_label); + ovs_be32 new_label = (old_label & htonl(0xF00FFFFF)) | htonl(tc << 20); + put_16aligned_be32(flow_label, new_label); } /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src', @@ -789,11 +778,11 @@ packet_set_ipv4(struct ofpbuf *packet, ovs_be32 src, ovs_be32 dst, { struct ip_header *nh = packet->l3; - if (nh->ip_src != src) { + if (get_16aligned_be32(&nh->ip_src) != src) { packet_set_ipv4_addr(packet, &nh->ip_src, src); } - if (nh->ip_dst != dst) { + if (get_16aligned_be32(&nh->ip_dst) != dst) { packet_set_ipv4_addr(packet, &nh->ip_dst, dst); } @@ -823,14 +812,14 @@ packet_set_ipv6(struct ofpbuf *packet, uint8_t proto, const ovs_be32 src[4], const ovs_be32 dst[4], uint8_t key_tc, ovs_be32 key_fl, uint8_t key_hl) { - struct ip6_hdr *nh = packet->l3; + struct ovs_16aligned_ip6_hdr *nh = packet->l3; if (memcmp(&nh->ip6_src, src, sizeof(ovs_be32[4]))) { - packet_set_ipv6_addr(packet, proto, &nh->ip6_src, src, true); + packet_set_ipv6_addr(packet, proto, nh->ip6_src.be32, src, true); } if (memcmp(&nh->ip6_dst, dst, sizeof(ovs_be32[4]))) { - packet_set_ipv6_addr(packet, proto, &nh->ip6_dst, dst, + packet_set_ipv6_addr(packet, proto, nh->ip6_dst.be32, dst, !packet_rh_present(packet)); } @@ -883,11 +872,32 @@ packet_set_udp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst) } } +/* Sets the SCTP source and destination port ('src' and 'dst' respectively) of + * the SCTP header contained in 'packet'. 'packet' must be a valid SCTP packet + * with its l4 marker properly populated. */ +void +packet_set_sctp_port(struct ofpbuf *packet, ovs_be16 src, ovs_be16 dst) +{ + struct sctp_header *sh = packet->l4; + ovs_be32 old_csum, old_correct_csum, new_csum; + uint16_t tp_len = packet->size - ((uint8_t*)sh - (uint8_t*)packet->data); + + old_csum = sh->sctp_csum; + sh->sctp_csum = 0; + old_correct_csum = crc32c(packet->l4, tp_len); + + sh->sctp_src = src; + sh->sctp_dst = dst; + + new_csum = crc32c(packet->l4, tp_len); + sh->sctp_csum = old_csum ^ old_correct_csum ^ new_csum; +} + /* 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 +uint16_t packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow) { if (dl_type_is_ip_any(flow->dl_type) && @@ -899,11 +909,44 @@ packet_get_tcp_flags(const struct ofpbuf *packet, const struct flow *flow) } } +const char * +packet_tcp_flag_to_string(uint32_t flag) +{ + switch (flag) { + case TCP_FIN: + return "fin"; + case TCP_SYN: + return "syn"; + case TCP_RST: + return "rst"; + case TCP_PSH: + return "psh"; + case TCP_ACK: + return "ack"; + case TCP_URG: + return "urg"; + case TCP_ECE: + return "ece"; + case TCP_CWR: + return "cwr"; + case TCP_NS: + return "ns"; + case 0x200: + return "[200]"; + case 0x400: + return "[400]"; + case 0x800: + return "[800]"; + default: + return NULL; + } +} + /* 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) +packet_format_tcp_flags(struct ds *s, uint16_t tcp_flags) { if (!tcp_flags) { ds_put_cstr(s, "none"); @@ -928,10 +971,22 @@ packet_format_tcp_flags(struct ds *s, uint8_t tcp_flags) if (tcp_flags & TCP_ACK) { ds_put_char(s, '.'); } - if (tcp_flags & 0x40) { - ds_put_cstr(s, "[40]"); + if (tcp_flags & TCP_ECE) { + ds_put_cstr(s, "E"); + } + if (tcp_flags & TCP_CWR) { + ds_put_cstr(s, "C"); + } + if (tcp_flags & TCP_NS) { + ds_put_cstr(s, "N"); + } + if (tcp_flags & 0x200) { + ds_put_cstr(s, "[200]"); + } + if (tcp_flags & 0x400) { + ds_put_cstr(s, "[400]"); } - if (tcp_flags & 0x80) { - ds_put_cstr(s, "[80]"); + if (tcp_flags & 0x800) { + ds_put_cstr(s, "[800]"); } }