#include <netinet/in.h>
#include <stdlib.h>
#include <string.h>
+#include "byte-order.h"
#include "coverage.h"
#include "dynamic-string.h"
#include "hash.h"
#include "packets.h"
#include "unaligned.h"
#include "vlog.h"
-#include "xtoxll.h"
-VLOG_DEFINE_THIS_MODULE(flow)
+VLOG_DEFINE_THIS_MODULE(flow);
static struct arp_eth_header *
pull_arp(struct ofpbuf *packet)
}
static struct tcp_header *
-pull_tcp(struct ofpbuf *packet)
+pull_tcp(struct ofpbuf *packet)
{
if (packet->size >= TCP_HEADER_LEN) {
struct tcp_header *tcp = packet->data;
}
static struct udp_header *
-pull_udp(struct ofpbuf *packet)
+pull_udp(struct ofpbuf *packet)
{
return ofpbuf_try_pull(packet, UDP_HEADER_LEN);
}
static struct icmp_header *
-pull_icmp(struct ofpbuf *packet)
+pull_icmp(struct ofpbuf *packet)
{
return ofpbuf_try_pull(packet, ICMP_HEADER_LEN);
}
-static struct eth_header *
-pull_eth(struct ofpbuf *packet)
+static void
+parse_vlan(struct ofpbuf *b, struct flow *flow)
{
- return ofpbuf_try_pull(packet, ETH_HEADER_LEN);
+ struct qtag_prefix {
+ ovs_be16 eth_type; /* ETH_TYPE_VLAN */
+ ovs_be16 tci;
+ };
+
+ if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) {
+ struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
+ flow->dl_vlan = qp->tci & htons(VLAN_VID_MASK);
+ flow->dl_vlan_pcp = vlan_tci_to_pcp(qp->tci);
+ }
}
-static struct vlan_header *
-pull_vlan(struct ofpbuf *packet)
+static ovs_be16
+parse_ethertype(struct ofpbuf *b)
{
- return ofpbuf_try_pull(packet, VLAN_HEADER_LEN);
+ struct llc_snap_header *llc;
+ ovs_be16 proto;
+
+ proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto);
+ if (ntohs(proto) >= ODP_DL_TYPE_ETH2_CUTOFF) {
+ return proto;
+ }
+
+ if (b->size < sizeof *llc) {
+ return htons(ODP_DL_TYPE_NOT_ETH_TYPE);
+ }
+
+ llc = b->data;
+ if (llc->llc.llc_dsap != LLC_DSAP_SNAP
+ || llc->llc.llc_ssap != LLC_SSAP_SNAP
+ || llc->llc.llc_cntl != LLC_CNTL_SNAP
+ || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET,
+ sizeof llc->snap.snap_org)) {
+ return htons(ODP_DL_TYPE_NOT_ETH_TYPE);
+ }
+
+ ofpbuf_pull(b, sizeof *llc);
+ return llc->snap.snap_type;
}
-/* Returns 1 if 'packet' is an IP fragment, 0 otherwise.
- * 'tun_id' is in network byte order, while 'in_port' is in host byte order.
- * These byte orders are the same as they are in struct odp_flow_key. */
+/* Initializes 'flow' members from 'packet', 'tun_id', and 'in_port.
+ * Initializes 'packet' header pointers as follows:
+ *
+ * - packet->l2 to the start of the Ethernet header.
+ *
+ * - packet->l3 to just past the Ethernet header, or just past the
+ * vlan_header if one is present, to the first byte of the payload of the
+ * Ethernet frame.
+ *
+ * - packet->l4 to just past the IPv4 header, if one is present and has a
+ * correct length, and otherwise NULL.
+ *
+ * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
+ * present and has a correct length, and otherwise NULL.
+ */
int
-flow_extract(struct ofpbuf *packet, uint32_t tun_id, uint16_t in_port,
- flow_t *flow)
+flow_extract(struct ofpbuf *packet, ovs_be32 tun_id, uint16_t in_port,
+ struct flow *flow)
{
struct ofpbuf b = *packet;
struct eth_header *eth;
packet->l4 = NULL;
packet->l7 = NULL;
- eth = pull_eth(&b);
- if (eth) {
- if (ntohs(eth->eth_type) >= OFP_DL_TYPE_ETH2_CUTOFF) {
- /* This is an Ethernet II frame */
- flow->dl_type = eth->eth_type;
- } else {
- /* This is an 802.2 frame */
- struct llc_header *llc = ofpbuf_at(&b, 0, sizeof *llc);
- struct snap_header *snap = ofpbuf_at(&b, sizeof *llc,
- sizeof *snap);
- if (llc == NULL) {
- return 0;
- }
- if (snap
- && llc->llc_dsap == LLC_DSAP_SNAP
- && llc->llc_ssap == LLC_SSAP_SNAP
- && llc->llc_cntl == LLC_CNTL_SNAP
- && !memcmp(snap->snap_org, SNAP_ORG_ETHERNET,
- sizeof snap->snap_org)) {
- flow->dl_type = snap->snap_type;
- ofpbuf_pull(&b, LLC_SNAP_HEADER_LEN);
- } else {
- flow->dl_type = htons(OFP_DL_TYPE_NOT_ETH_TYPE);
- ofpbuf_pull(&b, sizeof(struct llc_header));
- }
- }
+ if (b.size < sizeof *eth) {
+ return 0;
+ }
- /* Check for a VLAN tag */
- if (flow->dl_type == htons(ETH_TYPE_VLAN)) {
- struct vlan_header *vh = pull_vlan(&b);
- if (vh) {
- flow->dl_type = vh->vlan_next_type;
- flow->dl_vlan = vh->vlan_tci & htons(VLAN_VID_MASK);
- flow->dl_vlan_pcp = (ntohs(vh->vlan_tci) & 0xe000) >> 13;
- }
- }
- memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
- memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
-
- packet->l3 = b.data;
- if (flow->dl_type == htons(ETH_TYPE_IP)) {
- const struct ip_header *nh = pull_ip(&b);
- if (nh) {
- flow->nw_src = get_unaligned_u32(&nh->ip_src);
- flow->nw_dst = get_unaligned_u32(&nh->ip_dst);
- flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
- flow->nw_proto = nh->ip_proto;
- packet->l4 = b.data;
- if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
- if (flow->nw_proto == IP_TYPE_TCP) {
- const struct tcp_header *tcp = pull_tcp(&b);
- if (tcp) {
- flow->tp_src = tcp->tcp_src;
- flow->tp_dst = tcp->tcp_dst;
- packet->l7 = b.data;
- } else {
- /* Avoid tricking other code into thinking that
- * this packet has an L4 header. */
- flow->nw_proto = 0;
- }
- } else if (flow->nw_proto == IP_TYPE_UDP) {
- const struct udp_header *udp = pull_udp(&b);
- if (udp) {
- flow->tp_src = udp->udp_src;
- flow->tp_dst = udp->udp_dst;
- packet->l7 = b.data;
- } else {
- /* Avoid tricking other code into thinking that
- * this packet has an L4 header. */
- flow->nw_proto = 0;
- }
- } else if (flow->nw_proto == IP_TYPE_ICMP) {
- const struct icmp_header *icmp = pull_icmp(&b);
- if (icmp) {
- flow->icmp_type = htons(icmp->icmp_type);
- flow->icmp_code = htons(icmp->icmp_code);
- packet->l7 = b.data;
- } else {
- /* Avoid tricking other code into thinking that
- * this packet has an L4 header. */
- flow->nw_proto = 0;
- }
+ /* Link layer. */
+ eth = b.data;
+ memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
+ memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
+
+ /* dl_type, dl_vlan, dl_vlan_pcp. */
+ ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
+ if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
+ parse_vlan(&b, flow);
+ }
+ flow->dl_type = parse_ethertype(&b);
+
+ /* Network layer. */
+ packet->l3 = b.data;
+ if (flow->dl_type == htons(ETH_TYPE_IP)) {
+ const struct ip_header *nh = pull_ip(&b);
+ if (nh) {
+ flow->nw_src = get_unaligned_u32(&nh->ip_src);
+ flow->nw_dst = get_unaligned_u32(&nh->ip_dst);
+ flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
+ flow->nw_proto = nh->ip_proto;
+ packet->l4 = b.data;
+ if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
+ if (flow->nw_proto == IP_TYPE_TCP) {
+ const struct tcp_header *tcp = pull_tcp(&b);
+ if (tcp) {
+ flow->tp_src = tcp->tcp_src;
+ flow->tp_dst = tcp->tcp_dst;
+ packet->l7 = b.data;
+ }
+ } else if (flow->nw_proto == IP_TYPE_UDP) {
+ const struct udp_header *udp = pull_udp(&b);
+ if (udp) {
+ flow->tp_src = udp->udp_src;
+ flow->tp_dst = udp->udp_dst;
+ packet->l7 = b.data;
+ }
+ } else if (flow->nw_proto == IP_TYPE_ICMP) {
+ const struct icmp_header *icmp = pull_icmp(&b);
+ if (icmp) {
+ flow->icmp_type = htons(icmp->icmp_type);
+ flow->icmp_code = htons(icmp->icmp_code);
+ packet->l7 = b.data;
}
- } else {
- retval = 1;
}
+ } else {
+ retval = 1;
+ }
+ }
+ } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
+ const struct arp_eth_header *arp = pull_arp(&b);
+ if (arp && arp->ar_hrd == htons(1)
+ && arp->ar_pro == htons(ETH_TYPE_IP)
+ && arp->ar_hln == ETH_ADDR_LEN
+ && arp->ar_pln == 4) {
+ /* We only match on the lower 8 bits of the opcode. */
+ if (ntohs(arp->ar_op) <= 0xff) {
+ flow->nw_proto = ntohs(arp->ar_op);
}
- } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
- const struct arp_eth_header *arp = pull_arp(&b);
- if (arp && arp->ar_hrd == htons(1)
- && arp->ar_pro == htons(ETH_TYPE_IP)
- && arp->ar_hln == ETH_ADDR_LEN
- && arp->ar_pln == 4) {
- /* We only match on the lower 8 bits of the opcode. */
- if (ntohs(arp->ar_op) <= 0xff) {
- flow->nw_proto = ntohs(arp->ar_op);
- }
- if ((flow->nw_proto == ARP_OP_REQUEST)
- || (flow->nw_proto == ARP_OP_REPLY)) {
- flow->nw_src = arp->ar_spa;
- flow->nw_dst = arp->ar_tpa;
- }
+ if ((flow->nw_proto == ARP_OP_REQUEST)
+ || (flow->nw_proto == ARP_OP_REPLY)) {
+ flow->nw_src = arp->ar_spa;
+ flow->nw_dst = arp->ar_tpa;
}
}
}
* arguments must have been initialized through a call to flow_extract().
*/
void
-flow_extract_stats(const flow_t *flow, struct ofpbuf *packet,
+flow_extract_stats(const struct flow *flow, struct ofpbuf *packet,
struct odp_flow_stats *stats)
{
memset(stats, '\0', sizeof(*stats));
if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) {
- struct ip_header *ip = packet->l3;
- stats->ip_tos = ip->ip_tos;
if ((flow->nw_proto == IP_TYPE_TCP) && packet->l7) {
struct tcp_header *tcp = packet->l4;
stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl);
}
/* Extract 'flow' with 'wildcards' into the OpenFlow match structure
- * 'match'. */
+ * 'match'. 'flow_format' should be one of NXFF_*. */
void
-flow_to_match(const flow_t *flow, uint32_t wildcards, bool tun_id_from_cookie,
- struct ofp_match *match)
+flow_to_match(const struct flow *flow, uint32_t wildcards,
+ int flow_format, struct ofp_match *match)
{
- if (!tun_id_from_cookie) {
+ if (flow_format != NXFF_TUN_ID_FROM_COOKIE) {
wildcards &= OFPFW_ALL;
}
match->wildcards = htonl(wildcards);
}
void
-flow_from_match(const struct ofp_match *match, bool tun_id_from_cookie,
- uint64_t cookie, flow_t *flow, uint32_t *flow_wildcards)
+flow_from_match(const struct ofp_match *match, int flow_format,
+ ovs_be64 cookie, struct flow *flow, uint32_t *flow_wildcards)
{
uint32_t wildcards = ntohl(match->wildcards);
flow->nw_src = match->nw_src;
flow->nw_dst = match->nw_dst;
- if (tun_id_from_cookie && !(wildcards & NXFW_TUN_ID)) {
+ if (flow_format == NXFF_TUN_ID_FROM_COOKIE && !(wildcards & NXFW_TUN_ID)) {
flow->tun_id = htonl(ntohll(cookie) >> 32);
} else {
wildcards |= NXFW_TUN_ID;
memcpy(flow->dl_dst, match->dl_dst, ETH_ADDR_LEN);
flow->nw_tos = match->nw_tos;
flow->nw_proto = match->nw_proto;
- memset(flow->reserved, 0, sizeof flow->reserved);
-
if (flow_wildcards) {
*flow_wildcards = wildcards;
}
}
char *
-flow_to_string(const flow_t *flow)
+flow_to_string(const struct flow *flow)
{
struct ds ds = DS_EMPTY_INITIALIZER;
flow_format(&ds, flow);
}
void
-flow_format(struct ds *ds, const flow_t *flow)
+flow_format(struct ds *ds, const struct flow *flow)
{
ds_put_format(ds, "tunnel%08"PRIx32":in_port%04"PRIx16
":vlan%"PRIu16":pcp%"PRIu8
}
void
-flow_print(FILE *stream, const flow_t *flow)
+flow_print(FILE *stream, const struct flow *flow)
{
char *s = flow_to_string(flow);
fputs(s, stream);
free(s);
}
+\f
+/* flow_wildcards functions. */
+
+/* Given the wildcard bit count in bits 'shift' through 'shift + 5' (inclusive)
+ * of 'wildcards', returns a 32-bit bit mask with a 1 in each bit that must
+ * match and a 0 in each bit that is wildcarded.
+ *
+ * The bits in 'wildcards' are in the format used in enum ofp_flow_wildcards: 0
+ * is exact match, 1 ignores the LSB, 2 ignores the 2 least-significant bits,
+ * ..., 32 and higher wildcard the entire field. This is the *opposite* of the
+ * usual convention where e.g. /24 indicates that 8 bits (not 24 bits) are
+ * wildcarded. */
+ovs_be32
+flow_nw_bits_to_mask(uint32_t wildcards, int shift)
+{
+ wildcards = (wildcards >> shift) & 0x3f;
+ return wildcards < 32 ? htonl(~((1u << wildcards) - 1)) : 0;
+}
+
+/* Return 'wildcards' in "normal form":
+ *
+ * - Forces unknown bits to 0.
+ *
+ * - Forces nw_src and nw_dst masks greater than 32 to exactly 32.
+ */
+static inline uint32_t
+flow_wildcards_normalize(uint32_t wildcards)
+{
+ wildcards &= wildcards & OVSFW_ALL;
+ if (wildcards & (0x20 << OFPFW_NW_SRC_SHIFT)) {
+ wildcards &= ~(0x1f << OFPFW_NW_SRC_SHIFT);
+ }
+ if (wildcards & (0x20 << OFPFW_NW_DST_SHIFT)) {
+ wildcards &= ~(0x1f << OFPFW_NW_DST_SHIFT);
+ }
+ return wildcards;
+}
+
+/* Initializes 'wc' from 'wildcards', which may be any combination of the
+ * OFPFW_* and OVSFW_* wildcard bits. */
+void
+flow_wildcards_init(struct flow_wildcards *wc, uint32_t wildcards)
+{
+ wc->wildcards = flow_wildcards_normalize(wildcards);
+ wc->nw_src_mask = flow_nw_bits_to_mask(wc->wildcards, OFPFW_NW_SRC_SHIFT);
+ wc->nw_dst_mask = flow_nw_bits_to_mask(wc->wildcards, OFPFW_NW_DST_SHIFT);
+}
+
+/* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
+ * wildcard any bits or fields. */
+void
+flow_wildcards_init_exact(struct flow_wildcards *wc)
+{
+ flow_wildcards_init(wc, 0);
+}
+
+static inline uint32_t
+combine_nw_bits(uint32_t wb1, uint32_t wb2, int shift)
+{
+ uint32_t sb1 = (wb1 >> shift) & 0x3f;
+ uint32_t sb2 = (wb2 >> shift) & 0x3f;
+ return MAX(sb1, sb2) << shift;
+}
+
+/* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
+ * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
+ * 'src1' or 'src2' or both. */
+void
+flow_wildcards_combine(struct flow_wildcards *dst,
+ const struct flow_wildcards *src1,
+ const struct flow_wildcards *src2)
+{
+ uint32_t wb1 = src1->wildcards;
+ uint32_t wb2 = src2->wildcards;
+
+ dst->wildcards = (wb1 | wb2) & ~(OFPFW_NW_SRC_MASK | OFPFW_NW_DST_MASK);
+ dst->wildcards |= combine_nw_bits(wb1, wb2, OFPFW_NW_SRC_SHIFT);
+ dst->wildcards |= combine_nw_bits(wb1, wb2, OFPFW_NW_DST_SHIFT);
+ dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
+ dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
+}
+
+/* Returns a hash of the wildcards in 'wc'. */
+uint32_t
+flow_wildcards_hash(const struct flow_wildcards *wc)
+{
+ /* There is no need to include nw_src_mask or nw_dst_mask because they do
+ * not add any information (they can be computed from wc->wildcards). */
+ return hash_int(wc->wildcards, 0);
+}
+
+/* Returns true if 'a' and 'b' represent the same wildcards, false if they are
+ * different. */
+bool
+flow_wildcards_equal(const struct flow_wildcards *a,
+ const struct flow_wildcards *b)
+{
+ return a->wildcards == b->wildcards;
+}
+
+/* Returns true if at least one bit or field is wildcarded in 'a' but not in
+ * 'b', false otherwise. */
+bool
+flow_wildcards_has_extra(const struct flow_wildcards *a,
+ const struct flow_wildcards *b)
+{
+#define OFPFW_NW_MASK (OFPFW_NW_SRC_MASK | OFPFW_NW_DST_MASK)
+ return ((a->wildcards & ~(b->wildcards | OFPFW_NW_MASK))
+ || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
+ || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask);
+}
+
+static int
+count_ones(ovs_be32 mask)
+{
+#if __GNUC__ >= 4
+ return __builtin_popcount(mask);
+#else
+ int bits;
+
+ for (bits = 0; mask; bits++) {
+ mask &= mask - 1;
+ }
+
+ return bits;
+#endif
+}
+
+static bool
+set_nw_mask(struct flow_wildcards *wc, ovs_be32 mask,
+ ovs_be32 *maskp, int shift)
+{
+ int wcbits = 32 - count_ones(mask);
+ if (flow_nw_bits_to_mask(wcbits, 0) == mask) {
+ wc->wildcards &= ~(0x3f << shift);
+ wc->wildcards |= wcbits << shift;
+ *maskp = mask;
+ return true;
+ } else {
+ return false;
+ }
+}
+
+/* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N
+ * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
+ * false if 'mask' is not a CIDR mask. */
+bool
+flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask)
+{
+ return set_nw_mask(wc, mask, &wc->nw_src_mask, OFPFW_NW_SRC_SHIFT);
+}
+
+/* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of
+ * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
+ * false if 'mask' is not a CIDR mask. */
+bool
+flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask)
+{
+ return set_nw_mask(wc, mask, &wc->nw_dst_mask, OFPFW_NW_DST_SHIFT);
+}