datapath: Rename key_len to key_end
[sliver-openvswitch.git] / datapath / flow.c
index 2c11408..abce25e 100644 (file)
@@ -34,6 +34,7 @@
 #include <linux/if_arp.h>
 #include <linux/ip.h>
 #include <linux/ipv6.h>
+#include <linux/sctp.h>
 #include <linux/tcp.h>
 #include <linux/udp.h>
 #include <linux/icmp.h>
@@ -82,8 +83,9 @@ static void update_range__(struct sw_flow_match *match,
        do { \
                update_range__(match, offsetof(struct sw_flow_key, field),  \
                                     sizeof((match)->key->field), is_mask); \
-               if (is_mask && match->mask != NULL) {                       \
-                       (match)->mask->key.field = value;                   \
+               if (is_mask) {                                              \
+                       if ((match)->mask)                                  \
+                               (match)->mask->key.field = value;           \
                } else {                                                    \
                        (match)->key->field = value;                        \
                }                                                           \
@@ -93,8 +95,9 @@ static void update_range__(struct sw_flow_match *match,
        do { \
                update_range__(match, offsetof(struct sw_flow_key, field),  \
                                len, is_mask);                              \
-               if (is_mask && match->mask != NULL) {                       \
-                       memcpy(&(match)->mask->key.field, value_p, len);    \
+               if (is_mask) {                                              \
+                       if ((match)->mask)                                  \
+                               memcpy(&(match)->mask->key.field, value_p, len);\
                } else {                                                    \
                        memcpy(&(match)->key->field, value_p, len);         \
                }                                                           \
@@ -128,18 +131,16 @@ static bool ovs_match_validate(const struct sw_flow_match *match,
                        | (1ULL << OVS_KEY_ATTR_IPV6)
                        | (1ULL << OVS_KEY_ATTR_TCP)
                        | (1ULL << OVS_KEY_ATTR_UDP)
+                       | (1ULL << OVS_KEY_ATTR_SCTP)
                        | (1ULL << OVS_KEY_ATTR_ICMP)
                        | (1ULL << OVS_KEY_ATTR_ICMPV6)
                        | (1ULL << OVS_KEY_ATTR_ARP)
                        | (1ULL << OVS_KEY_ATTR_ND));
 
-       if (match->key->phy.in_port == DP_MAX_PORTS &&
-           match->mask && (match->mask->key.phy.in_port == 0xffff))
-               mask_allowed |= (1ULL << OVS_KEY_ATTR_IN_PORT);
-
-       if (match->key->eth.type == htons(ETH_P_802_2) &&
-           match->mask && (match->mask->key.eth.type == htons(0xffff)))
-               mask_allowed |= (1ULL << OVS_KEY_ATTR_ETHERTYPE);
+       /* Always allowed mask fields. */
+       mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
+                      | (1ULL << OVS_KEY_ATTR_IN_PORT)
+                      | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
 
        /* Check key attributes. */
        if (match->key->eth.type == htons(ETH_P_ARP)
@@ -161,6 +162,12 @@ static bool ovs_match_validate(const struct sw_flow_match *match,
                                        mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
                        }
 
+                       if (match->key->ip.proto == IPPROTO_SCTP) {
+                               key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
+                               if (match->mask && (match->mask->key.ip.proto == 0xff))
+                                       mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
+                       }
+
                        if (match->key->ip.proto == IPPROTO_TCP) {
                                key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
                                if (match->mask && (match->mask->key.ip.proto == 0xff))
@@ -187,6 +194,12 @@ static bool ovs_match_validate(const struct sw_flow_match *match,
                                        mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
                        }
 
+                       if (match->key->ip.proto == IPPROTO_SCTP) {
+                               key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
+                               if (match->mask && (match->mask->key.ip.proto == 0xff))
+                                       mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
+                       }
+
                        if (match->key->ip.proto == IPPROTO_TCP) {
                                key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
                                if (match->mask && (match->mask->key.ip.proto == 0xff))
@@ -282,6 +295,12 @@ static bool udphdr_ok(struct sk_buff *skb)
                                  sizeof(struct udphdr));
 }
 
+static bool sctphdr_ok(struct sk_buff *skb)
+{
+       return pskb_may_pull(skb, skb_transport_offset(skb) +
+                                 sizeof(struct sctphdr));
+}
+
 static bool icmphdr_ok(struct sk_buff *skb)
 {
        return pskb_may_pull(skb, skb_transport_offset(skb) +
@@ -800,7 +819,6 @@ invalid:
  * Ethernet header
  * @in_port: port number on which @skb was received.
  * @key: output flow key
- * @key_lenp: length of output flow key
  *
  * The caller must ensure that skb->len >= ETH_HLEN.
  *
@@ -901,6 +919,12 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
                                key->ipv4.tp.src = udp->source;
                                key->ipv4.tp.dst = udp->dest;
                        }
+               } else if (key->ip.proto == IPPROTO_SCTP) {
+                       if (sctphdr_ok(skb)) {
+                               struct sctphdr *sctp = sctp_hdr(skb);
+                               key->ipv4.tp.src = sctp->source;
+                               key->ipv4.tp.dst = sctp->dest;
+                       }
                } else if (key->ip.proto == IPPROTO_ICMP) {
                        if (icmphdr_ok(skb)) {
                                struct icmphdr *icmp = icmp_hdr(skb);
@@ -963,6 +987,12 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
                                key->ipv6.tp.src = udp->source;
                                key->ipv6.tp.dst = udp->dest;
                        }
+               } else if (key->ip.proto == NEXTHDR_SCTP) {
+                       if (sctphdr_ok(skb)) {
+                               struct sctphdr *sctp = sctp_hdr(skb);
+                               key->ipv6.tp.src = sctp->source;
+                               key->ipv6.tp.dst = sctp->dest;
+                       }
                } else if (key->ip.proto == NEXTHDR_ICMP) {
                        if (icmp6hdr_ok(skb)) {
                                error = parse_icmpv6(skb, key, nh_len);
@@ -975,10 +1005,11 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
        return 0;
 }
 
-static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start, int key_len)
+static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start,
+                        int key_end)
 {
        return jhash2((u32 *)((u8 *)key + key_start),
-                     DIV_ROUND_UP(key_len - key_start, sizeof(u32)), 0);
+                     DIV_ROUND_UP(key_end - key_start, sizeof(u32)), 0);
 }
 
 static int flow_key_start(const struct sw_flow_key *key)
@@ -990,31 +1021,31 @@ static int flow_key_start(const struct sw_flow_key *key)
 }
 
 static bool __cmp_key(const struct sw_flow_key *key1,
-               const struct sw_flow_key *key2,  int key_start, int key_len)
+               const struct sw_flow_key *key2,  int key_start, int key_end)
 {
        return !memcmp((u8 *)key1 + key_start,
-                       (u8 *)key2 + key_start, (key_len - key_start));
+                       (u8 *)key2 + key_start, (key_end - key_start));
 }
 
 static bool __flow_cmp_key(const struct sw_flow *flow,
-               const struct sw_flow_key *key, int key_start, int key_len)
+               const struct sw_flow_key *key, int key_start, int key_end)
 {
-       return __cmp_key(&flow->key, key, key_start, key_len);
+       return __cmp_key(&flow->key, key, key_start, key_end);
 }
 
 static bool __flow_cmp_unmasked_key(const struct sw_flow *flow,
-                 const struct sw_flow_key *key, int key_start, int key_len)
+                 const struct sw_flow_key *key, int key_start, int key_end)
 {
-       return __cmp_key(&flow->unmasked_key, key, key_start, key_len);
+       return __cmp_key(&flow->unmasked_key, key, key_start, key_end);
 }
 
 bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
-               const struct sw_flow_key *key, int key_len)
+               const struct sw_flow_key *key, int key_end)
 {
        int key_start;
        key_start = flow_key_start(key);
 
-       return __flow_cmp_unmasked_key(flow, key, key_start, key_len);
+       return __flow_cmp_unmasked_key(flow, key, key_start, key_end);
 
 }
 
@@ -1022,11 +1053,11 @@ struct sw_flow *ovs_flow_lookup_unmasked_key(struct flow_table *table,
                                       struct sw_flow_match *match)
 {
        struct sw_flow_key *unmasked = match->key;
-       int key_len = match->range.end;
+       int key_end = match->range.end;
        struct sw_flow *flow;
 
        flow = ovs_flow_lookup(table, unmasked);
-       if (flow && (!ovs_flow_cmp_unmasked_key(flow, unmasked, key_len)))
+       if (flow && (!ovs_flow_cmp_unmasked_key(flow, unmasked, key_end)))
                flow = NULL;
 
        return flow;
@@ -1039,16 +1070,16 @@ static struct sw_flow *ovs_masked_flow_lookup(struct flow_table *table,
        struct sw_flow *flow;
        struct hlist_head *head;
        int key_start = mask->range.start;
-       int key_len = mask->range.end;
+       int key_end = mask->range.end;
        u32 hash;
        struct sw_flow_key masked_key;
 
        ovs_flow_key_mask(&masked_key, flow_key, mask);
-       hash = ovs_flow_hash(&masked_key, key_start, key_len);
+       hash = ovs_flow_hash(&masked_key, key_start, key_end);
        head = find_bucket(table, hash);
        hlist_for_each_entry_rcu(flow, head, hash_node[table->node_ver]) {
                if (flow->mask == mask &&
-                   __flow_cmp_key(flow, &masked_key, key_start, key_len))
+                   __flow_cmp_key(flow, &masked_key, key_start, key_end))
                        return flow;
        }
        return NULL;
@@ -1097,6 +1128,7 @@ const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
        [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
        [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
        [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
+       [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
        [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
        [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
        [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
@@ -1175,8 +1207,8 @@ static int parse_flow_nlattrs(const struct nlattr *attr,
        return __parse_flow_nlattrs(attr, a, attrsp, false);
 }
 
-int ipv4_tun_from_nlattr(const struct nlattr *attr,
-                        struct sw_flow_match *match, bool is_mask)
+int ovs_ipv4_tun_from_nlattr(const struct nlattr *attr,
+                            struct sw_flow_match *match, bool is_mask)
 {
        struct nlattr *a;
        int rem;
@@ -1203,7 +1235,7 @@ int ipv4_tun_from_nlattr(const struct nlattr *attr,
 
                if (ovs_tunnel_key_lens[type] != nla_len(a)) {
                        OVS_NLERR("IPv4 tunnel attribute type has unexpected "
-                                 " legnth (type=%d, length=%d, expected=%d).\n",
+                                 " length (type=%d, length=%d, expected=%d).\n",
                                  type, nla_len(a), ovs_tunnel_key_lens[type]);
                        return -EINVAL;
                }
@@ -1249,22 +1281,24 @@ int ipv4_tun_from_nlattr(const struct nlattr *attr,
                return -EINVAL;
        }
 
-       if (!match->key->tun_key.ipv4_dst) {
-               OVS_NLERR("IPv4 tunnel destination address is zero.\n");
-               return -EINVAL;
-       }
+       if (!is_mask) {
+               if (!match->key->tun_key.ipv4_dst) {
+                       OVS_NLERR("IPv4 tunnel destination address is zero.\n");
+                       return -EINVAL;
+               }
 
-       if (!ttl) {
-               OVS_NLERR("IPv4 tunnel TTL not specified.\n");
-               return -EINVAL;
+               if (!ttl) {
+                       OVS_NLERR("IPv4 tunnel TTL not specified.\n");
+                       return -EINVAL;
+               }
        }
 
        return 0;
 }
 
-int ipv4_tun_to_nlattr(struct sk_buff *skb,
-                       const struct ovs_key_ipv4_tunnel *tun_key,
-                       const struct ovs_key_ipv4_tunnel *output)
+int ovs_ipv4_tun_to_nlattr(struct sk_buff *skb,
+                          const struct ovs_key_ipv4_tunnel *tun_key,
+                          const struct ovs_key_ipv4_tunnel *output)
 {
        struct nlattr *nla;
 
@@ -1310,8 +1344,11 @@ static int metadata_from_nlattrs(struct sw_flow_match *match,  u64 *attrs,
        if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
                u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
 
-               if (!is_mask && in_port >= DP_MAX_PORTS)
+               if (is_mask)
+                       in_port = 0xffffffff; /* Always exact match in_port. */
+               else if (in_port >= DP_MAX_PORTS)
                        return -EINVAL;
+
                SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
                *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
        } else if (!is_mask) {
@@ -1330,7 +1367,7 @@ static int metadata_from_nlattrs(struct sw_flow_match *match,  u64 *attrs,
                *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
        }
        if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
-               if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
+               if (ovs_ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
                                        is_mask))
                        return -EINVAL;
                *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
@@ -1363,22 +1400,29 @@ static int ovs_key_from_nlattrs(struct sw_flow_match *match,  u64 attrs,
                __be16 tci;
 
                tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
-               if (!is_mask)
-                       if (!(tci & htons(VLAN_TAG_PRESENT))) {
+               if (!(tci & htons(VLAN_TAG_PRESENT))) {
+                       if (is_mask)
+                               OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
+                       else
                                OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
-                               return -EINVAL;
-                       }
+
+                       return -EINVAL;
+               }
 
                SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
                attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
-       }
+       } else if (!is_mask)
+               SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
 
        if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
                __be16 eth_type;
 
                eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
-               if (!is_mask && ntohs(eth_type) < ETH_P_802_3_MIN) {
-                       OVS_NLERR("EtherType is less than mimimum (type=%x, min=%x).\n",
+               if (is_mask) {
+                       /* Always exact match EtherType. */
+                       eth_type = htons(0xffff);
+               } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
+                       OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
                                        ntohs(eth_type), ETH_P_802_3_MIN);
                        return -EINVAL;
                }
@@ -1504,6 +1548,24 @@ static int ovs_key_from_nlattrs(struct sw_flow_match *match,  u64 attrs,
                attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
        }
 
+       if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
+               const struct ovs_key_sctp *sctp_key;
+
+               sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
+               if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
+                       SW_FLOW_KEY_PUT(match, ipv4.tp.src,
+                                       sctp_key->sctp_src, is_mask);
+                       SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
+                                       sctp_key->sctp_dst, is_mask);
+               } else {
+                       SW_FLOW_KEY_PUT(match, ipv6.tp.src,
+                                       sctp_key->sctp_src, is_mask);
+                       SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
+                                       sctp_key->sctp_dst, is_mask);
+               }
+               attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
+       }
+
        if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
                const struct ovs_key_icmp *icmp_key;
 
@@ -1687,30 +1749,29 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey,
 {
        struct ovs_key_ethernet *eth_key;
        struct nlattr *nla, *encap;
+       bool is_mask = (swkey != output);
 
-       if (output->phy.priority &&
-               nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
+       if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
                goto nla_put_failure;
 
-       if (swkey->tun_key.ipv4_dst &&
-           ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
+       if ((swkey->tun_key.ipv4_dst || is_mask) &&
+           ovs_ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
                goto nla_put_failure;
 
        if (swkey->phy.in_port == DP_MAX_PORTS) {
-               if ((swkey != output) && (output->phy.in_port == 0xffff))
+               if (is_mask && (output->phy.in_port == 0xffff))
                        if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
                                goto nla_put_failure;
        } else {
                u16 upper_u16;
-               upper_u16 = (swkey == output) ? 0 : 0xffff;
+               upper_u16 = !is_mask ? 0 : 0xffff;
 
                if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
                                (upper_u16 << 16) | output->phy.in_port))
                        goto nla_put_failure;
        }
 
-       if (output->phy.skb_mark &&
-               nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
+       if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
                goto nla_put_failure;
 
        nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
@@ -1723,7 +1784,7 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey,
 
        if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
                __be16 eth_type;
-               eth_type = (swkey == output) ? htons(ETH_P_8021Q) : htons(0xffff) ;
+               eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
                if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
                    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
                        goto nla_put_failure;
@@ -1740,7 +1801,7 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey,
                 * 0xffff in the mask attribute.  Ethertype can also
                 * be wildcarded.
                 */
-               if (swkey != output && output->eth.type)
+               if (is_mask && output->eth.type)
                        if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
                                                output->eth.type))
                                goto nla_put_failure;
@@ -1827,6 +1888,20 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey,
                                udp_key->udp_src = output->ipv6.tp.src;
                                udp_key->udp_dst = output->ipv6.tp.dst;
                        }
+               } else if (swkey->ip.proto == IPPROTO_SCTP) {
+                       struct ovs_key_sctp *sctp_key;
+
+                       nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
+                       if (!nla)
+                               goto nla_put_failure;
+                       sctp_key = nla_data(nla);
+                       if (swkey->eth.type == htons(ETH_P_IP)) {
+                               sctp_key->sctp_src = swkey->ipv4.tp.src;
+                               sctp_key->sctp_dst = swkey->ipv4.tp.dst;
+                       } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
+                               sctp_key->sctp_src = swkey->ipv6.tp.src;
+                               sctp_key->sctp_dst = swkey->ipv6.tp.dst;
+                       }
                } else if (swkey->eth.type == htons(ETH_P_IP) &&
                           swkey->ip.proto == IPPROTO_ICMP) {
                        struct ovs_key_icmp *icmp_key;