/*
- * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc.
+ * Copyright (c) 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.
#include <stdlib.h>
#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;
{
struct eth_addr_node {
struct hmap_node hmap_node;
- uint64_t ea64;
+ const uint64_t ea64;
};
static struct eth_addr_node nodes[] = {
/* STP, IEEE pause frames, and other reserved protocols. */
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000000ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000001ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000002ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000003ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000004ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000005ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000006ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000007ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000008ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c2000009ULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000aULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000bULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000cULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000dULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000eULL },
- { HMAP_NODE_NULL_INITIALIZER, 0x0108c200000fULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000000ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000001ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000002ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000003ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000004ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000005ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000006ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000007ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000008ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c2000009ULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000aULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000bULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000cULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000dULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000eULL },
+ { HMAP_NODE_NULL_INITIALIZER, 0x0180c200000fULL },
/* Extreme protocols. */
{ HMAP_NODE_NULL_INITIALIZER, 0x00e02b000000ULL }, /* EDP. */
{ HMAP_NODE_NULL_INITIALIZER, 0x01000cccccc7ULL },
};
- static struct hmap addrs = HMAP_INITIALIZER(&addrs);
+ static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
struct eth_addr_node *node;
+ static struct hmap addrs;
uint64_t ea64;
- if (hmap_is_empty(&addrs)) {
+ if (ovsthread_once_start(&once)) {
+ hmap_init(&addrs);
for (node = nodes; node < &nodes[ARRAY_SIZE(nodes)]; node++) {
hmap_insert(&addrs, &node->hmap_node,
hash_2words(node->ea64, node->ea64 >> 32));
}
+ ovsthread_once_done(&once);
}
ea64 = eth_addr_to_uint64(ea);
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);
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
/* Removes outermost VLAN header (if any is present) from 'packet'.
*
- * 'packet->l2' must initially point to 'packet''s Ethernet header. */
+ * 'packet->l2_5' should initially point to 'packet''s outer-most MPLS header
+ * or may be NULL if there are no MPLS headers. */
void
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)
+{
+ struct eth_header *eh = packet->data;
+
+ if (eh->eth_type == htons(ETH_TYPE_VLAN)) {
+ ovs_be16 *p;
+ p = ALIGNED_CAST(ovs_be16 *,
+ (char *)(packet->l2_5 ? packet->l2_5 : packet->l3) - 2);
+ *p = eth_type;
+ } else {
+ eh->eth_type = eth_type;
+ }
+}
+
+static bool is_mpls(struct ofpbuf *packet)
+{
+ return packet->l2_5 != NULL;
+}
+
+/* Set time to live (TTL) of an MPLS label stack entry (LSE). */
+void
+set_mpls_lse_ttl(ovs_be32 *lse, uint8_t ttl)
+{
+ *lse &= ~htonl(MPLS_TTL_MASK);
+ *lse |= htonl((ttl << MPLS_TTL_SHIFT) & MPLS_TTL_MASK);
+}
+
+/* Set traffic class (TC) of an MPLS label stack entry (LSE). */
+void
+set_mpls_lse_tc(ovs_be32 *lse, uint8_t tc)
+{
+ *lse &= ~htonl(MPLS_TC_MASK);
+ *lse |= htonl((tc << MPLS_TC_SHIFT) & MPLS_TC_MASK);
+}
+
+/* Set label of an MPLS label stack entry (LSE). */
+void
+set_mpls_lse_label(ovs_be32 *lse, ovs_be32 label)
+{
+ *lse &= ~htonl(MPLS_LABEL_MASK);
+ *lse |= htonl((ntohl(label) << MPLS_LABEL_SHIFT) & MPLS_LABEL_MASK);
+}
+
+/* Set bottom of stack (BoS) bit of an MPLS label stack entry (LSE). */
+void
+set_mpls_lse_bos(ovs_be32 *lse, uint8_t bos)
+{
+ *lse &= ~htonl(MPLS_BOS_MASK);
+ *lse |= htonl((bos << MPLS_BOS_SHIFT) & MPLS_BOS_MASK);
+}
+
+/* Compose an MPLS label stack entry (LSE) from its components:
+ * label, traffic class (TC), time to live (TTL) and
+ * bottom of stack (BoS) bit. */
+ovs_be32
+set_mpls_lse_values(uint8_t ttl, uint8_t tc, uint8_t bos, ovs_be32 label)
+{
+ ovs_be32 lse = htonl(0);
+ set_mpls_lse_ttl(&lse, ttl);
+ set_mpls_lse_tc(&lse, tc);
+ set_mpls_lse_bos(&lse, bos);
+ set_mpls_lse_label(&lse, label);
+ return lse;
+}
+
+/* Push an new MPLS stack entry onto the MPLS stack and adjust 'packet->l2' and
+ * 'packet->l2_5' accordingly. The new entry will be the outermost entry on
+ * the stack.
+ *
+ * Previous to calling this function, 'packet->l2_5' must be set; if the MPLS
+ * label to be pushed will be the first label in 'packet', then it should be
+ * the same as 'packet->l3'. */
+static void
+push_mpls_lse(struct ofpbuf *packet, struct mpls_hdr *mh)
+{
+ char * header;
+ size_t len;
+ header = ofpbuf_push_uninit(packet, MPLS_HLEN);
+ len = (char *)packet->l2_5 - (char *)packet->l2;
+ memmove(header, packet->l2, len);
+ memcpy(header + len, mh, sizeof *mh);
+ packet->l2 = (char*)packet->l2 - MPLS_HLEN;
+ packet->l2_5 = (char*)packet->l2_5 - MPLS_HLEN;
+}
+
+/* Set MPLS label stack entry to outermost MPLS header.*/
+void
+set_mpls_lse(struct ofpbuf *packet, ovs_be32 mpls_lse)
+{
+ struct mpls_hdr *mh = packet->l2_5;
+
+ /* Packet type should be MPLS to set label stack entry. */
+ if (is_mpls(packet)) {
+ /* Update mpls label stack entry. */
+ mh->mpls_lse = mpls_lse;
+ }
+}
+
+/* Push MPLS label stack entry 'lse' onto 'packet' as the the outermost MPLS
+ * header. If 'packet' does not already have any MPLS labels, then its
+ * Ethertype is changed to 'ethtype' (which must be an MPLS Ethertype). */
+void
+push_mpls(struct ofpbuf *packet, ovs_be16 ethtype, ovs_be32 lse)
+{
+ struct mpls_hdr mh;
+
+ if (!eth_type_mpls(ethtype)) {
+ return;
+ }
+
+ if (!is_mpls(packet)) {
+ /* Set ethtype and MPLS label stack entry. */
+ set_ethertype(packet, ethtype);
+ packet->l2_5 = packet->l3;
+ }
+
+ /* Push new MPLS shim header onto packet. */
+ mh.mpls_lse = lse;
+ push_mpls_lse(packet, &mh);
+}
+
+/* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
+ * If the label that was removed was the only MPLS label, changes 'packet''s
+ * Ethertype to 'ethtype' (which ordinarily should not be an MPLS
+ * Ethertype). */
+void
+pop_mpls(struct ofpbuf *packet, ovs_be16 ethtype)
+{
+ struct mpls_hdr *mh = NULL;
+
+ if (is_mpls(packet)) {
+ size_t len;
+ mh = packet->l2_5;
+ len = (char*)packet->l2_5 - (char*)packet->l2;
+ set_ethertype(packet, ethtype);
+ if (mh->mpls_lse & htonl(MPLS_BOS_MASK)) {
+ packet->l2_5 = NULL;
+ } else {
+ packet->l2_5 = (char*)packet->l2_5 + MPLS_HLEN;
+ }
+ /* Shift the l2 header forward. */
+ memmove((char*)packet->data + MPLS_HLEN, packet->data, len);
+ packet->size -= MPLS_HLEN;
+ packet->data = (char*)packet->data + MPLS_HLEN;
+ packet->l2 = (char*)packet->l2 + MPLS_HLEN;
+ }
+}
+
/* 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);
* '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. */
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);
}
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
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;
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;
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;
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;
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',
{
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);
}
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));
}
}
}
+/* 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
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)) &&
+ if (dl_type_is_ip_any(flow->dl_type) &&
flow->nw_proto == IPPROTO_TCP && packet->l7) {
const struct tcp_header *tcp = packet->l4;
return TCP_FLAGS(tcp->tcp_ctl);