treewide: Remove trailing whitespace
[sliver-openvswitch.git] / lib / flow.c
index c1f6240..462df08 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2008, 2009 Nicira Networks.
+ * Copyright (c) 2008, 2009, 2010 Nicira Networks.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
 #include "openflow/openflow.h"
 #include "openvswitch/datapath-protocol.h"
 #include "packets.h"
-
+#include "unaligned.h"
 #include "vlog.h"
-#define THIS_MODULE VLM_flow
+#include "xtoxll.h"
+
+VLOG_DEFINE_THIS_MODULE(flow)
 
 static struct arp_eth_header *
 pull_arp(struct ofpbuf *packet)
@@ -51,7 +53,7 @@ pull_ip(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;
@@ -64,32 +66,80 @@ pull_tcp(struct ofpbuf *packet)
 }
 
 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, flow_t *flow)
 {
-    return ofpbuf_try_pull(packet, ETH_HEADER_LEN);
+    struct qtag_prefix {
+        uint16_t eth_type;      /* ETH_TYPE_VLAN */
+        uint16_t tci;
+    };
+
+    if (b->size >= sizeof(struct qtag_prefix) + sizeof(uint16_t)) {
+        struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
+        flow->dl_vlan = qp->tci & htons(VLAN_VID_MASK);
+        flow->dl_vlan_pcp = (ntohs(qp->tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT;
+    }
 }
 
-static struct vlan_header *
-pull_vlan(struct ofpbuf *packet)
+static uint16_t
+parse_ethertype(struct ofpbuf *b)
 {
-    return ofpbuf_try_pull(packet, VLAN_HEADER_LEN);
+    struct llc_snap_header *llc;
+    uint16_t proto;
+
+    proto = *(uint16_t *) 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 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, uint16_t in_port, flow_t *flow)
+flow_extract(struct ofpbuf *packet, uint32_t tun_id, uint16_t in_port,
+             flow_t *flow)
 {
     struct ofpbuf b = *packet;
     struct eth_header *eth;
@@ -98,115 +148,83 @@ flow_extract(struct ofpbuf *packet, uint16_t in_port, flow_t *flow)
     COVERAGE_INC(flow_extract);
 
     memset(flow, 0, sizeof *flow);
-    flow->dl_vlan = htons(OFP_VLAN_NONE);
+    flow->tun_id = tun_id;
     flow->in_port = in_port;
+    flow->dl_vlan = htons(OFP_VLAN_NONE);
 
     packet->l2 = b.data;
     packet->l3 = NULL;
     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);
-            }
-        }
-        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 = nh->ip_src;
-                flow->nw_dst = nh->ip_dst;
-                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;
             }
         }
     }
@@ -217,14 +235,12 @@ flow_extract(struct ofpbuf *packet, uint16_t in_port, flow_t *flow)
  * 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 flow_t *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);
@@ -235,75 +251,64 @@ flow_extract_stats(const flow_t *flow, struct ofpbuf *packet,
     stats->n_packets = 1;
 }
 
-/* The Open vSwitch datapath supports matching on ARP payloads, which 
- * OpenFlow does not.  This function is identical to 'flow_to_match',
- * but does not hide the datapath's ability to match on ARP. */
+/* Extract 'flow' with 'wildcards' into the OpenFlow match structure
+ * 'match'. */
 void
-flow_to_ovs_match(const flow_t *flow, uint32_t wildcards, 
-                  struct ofp_match *match)
+flow_to_match(const flow_t *flow, uint32_t wildcards, bool tun_id_from_cookie,
+              struct ofp_match *match)
 {
+    if (!tun_id_from_cookie) {
+        wildcards &= OFPFW_ALL;
+    }
     match->wildcards = htonl(wildcards);
+
     match->in_port = htons(flow->in_port == ODPP_LOCAL ? OFPP_LOCAL
                            : flow->in_port);
     match->dl_vlan = flow->dl_vlan;
+    match->dl_vlan_pcp = flow->dl_vlan_pcp;
     memcpy(match->dl_src, flow->dl_src, ETH_ADDR_LEN);
     memcpy(match->dl_dst, flow->dl_dst, ETH_ADDR_LEN);
     match->dl_type = flow->dl_type;
     match->nw_src = flow->nw_src;
     match->nw_dst = flow->nw_dst;
+    match->nw_tos = flow->nw_tos;
     match->nw_proto = flow->nw_proto;
     match->tp_src = flow->tp_src;
     match->tp_dst = flow->tp_dst;
-    match->pad = 0;
+    memset(match->pad1, '\0', sizeof match->pad1);
+    memset(match->pad2, '\0', sizeof match->pad2);
 }
 
-/* Extract 'flow' with 'wildcards' into the OpenFlow match structure
- * 'match'. */
 void
-flow_to_match(const flow_t *flow, uint32_t wildcards, struct ofp_match *match)
+flow_from_match(const struct ofp_match *match, bool tun_id_from_cookie,
+                uint64_t cookie, flow_t *flow, uint32_t *flow_wildcards)
 {
-    flow_to_ovs_match(flow, wildcards, match);
-
-    /* The datapath supports matching on an ARP's opcode and IP addresses, 
-     * but OpenFlow does not.  We wildcard and zero out the appropriate
-     * fields so that OpenFlow is unaware of our trickery. */
-    if (flow->dl_type == htons(ETH_TYPE_ARP)) {
-        wildcards |= (OFPFW_NW_PROTO | OFPFW_NW_SRC_ALL | OFPFW_NW_DST_ALL);
-        match->nw_src = 0;
-        match->nw_dst = 0;
-        match->nw_proto = 0;
-    }
-    match->wildcards = htonl(wildcards);
-}
-
-
-void
-flow_from_match(flow_t *flow, uint32_t *wildcards,
-                const struct ofp_match *match)
-{
-    if (wildcards) {
-        *wildcards = ntohl(match->wildcards);
-    }
-    /* The datapath supports matching on an ARP's opcode and IP addresses, 
-     * but OpenFlow does not.  In case the controller hasn't, we need to 
-     * set the appropriate wildcard bits so that we're externally 
-     * OpenFlow-compliant. */
-    if (match->dl_type == htons(ETH_TYPE_ARP)) {
-        *wildcards |= (OFPFW_NW_PROTO | OFPFW_NW_SRC_ALL | OFPFW_NW_DST_ALL);
-    }
+       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)) {
+        flow->tun_id = htonl(ntohll(cookie) >> 32);
+    } else {
+        wildcards |= NXFW_TUN_ID;
+        flow->tun_id = 0;
+    }
     flow->in_port = (match->in_port == htons(OFPP_LOCAL) ? ODPP_LOCAL
                      : ntohs(match->in_port));
     flow->dl_vlan = match->dl_vlan;
+    flow->dl_vlan_pcp = match->dl_vlan_pcp;
     flow->dl_type = match->dl_type;
     flow->tp_src = match->tp_src;
     flow->tp_dst = match->tp_dst;
     memcpy(flow->dl_src, match->dl_src, ETH_ADDR_LEN);
     memcpy(flow->dl_dst, match->dl_dst, ETH_ADDR_LEN);
+    flow->nw_tos = match->nw_tos;
     flow->nw_proto = match->nw_proto;
-    flow->reserved = 0;
+    memset(flow->reserved, 0, sizeof flow->reserved);
+
+    if (flow_wildcards) {
+        *flow_wildcards = wildcards;
+    }
 }
 
 char *
@@ -317,17 +322,31 @@ flow_to_string(const flow_t *flow)
 void
 flow_format(struct ds *ds, const flow_t *flow)
 {
-    ds_put_format(ds, "port%04x:vlan%d mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT" "
-                  "type%04x proto%"PRId8" ip"IP_FMT"->"IP_FMT" port%d->%d",
-                  flow->in_port, ntohs(flow->dl_vlan),
-                  ETH_ADDR_ARGS(flow->dl_src), ETH_ADDR_ARGS(flow->dl_dst),
-                  ntohs(flow->dl_type), flow->nw_proto,
-                  IP_ARGS(&flow->nw_src), IP_ARGS(&flow->nw_dst),
-                  ntohs(flow->tp_src), ntohs(flow->tp_dst));
+    ds_put_format(ds, "tunnel%08"PRIx32":in_port%04"PRIx16
+                      ":vlan%"PRIu16":pcp%"PRIu8
+                      " mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
+                      " type%04"PRIx16
+                      " proto%"PRIu8
+                      " tos%"PRIu8
+                      " ip"IP_FMT"->"IP_FMT
+                      " port%"PRIu16"->%"PRIu16,
+                  ntohl(flow->tun_id),
+                  flow->in_port,
+                  ntohs(flow->dl_vlan),
+                  flow->dl_vlan_pcp,
+                  ETH_ADDR_ARGS(flow->dl_src),
+                  ETH_ADDR_ARGS(flow->dl_dst),
+                  ntohs(flow->dl_type),
+                  flow->nw_proto,
+                  flow->nw_tos,
+                  IP_ARGS(&flow->nw_src),
+                  IP_ARGS(&flow->nw_dst),
+                  ntohs(flow->tp_src),
+                  ntohs(flow->tp_dst));
 }
 
 void
-flow_print(FILE *stream, const flow_t *flow) 
+flow_print(FILE *stream, const flow_t *flow)
 {
     char *s = flow_to_string(flow);
     fputs(s, stream);