Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / openvswitch / flow.h
diff --git a/kernel/net/openvswitch/flow.h b/kernel/net/openvswitch/flow.h
new file mode 100644 (file)
index 0000000..a076e44
--- /dev/null
@@ -0,0 +1,284 @@
+/*
+ * Copyright (c) 2007-2014 Nicira, Inc.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA
+ */
+
+#ifndef FLOW_H
+#define FLOW_H 1
+
+#include <linux/cache.h>
+#include <linux/kernel.h>
+#include <linux/netlink.h>
+#include <linux/openvswitch.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/rcupdate.h>
+#include <linux/if_ether.h>
+#include <linux/in6.h>
+#include <linux/jiffies.h>
+#include <linux/time.h>
+#include <linux/flex_array.h>
+#include <net/inet_ecn.h>
+
+struct sk_buff;
+
+/* Used to memset ovs_key_ipv4_tunnel padding. */
+#define OVS_TUNNEL_KEY_SIZE                                    \
+       (offsetof(struct ovs_key_ipv4_tunnel, tp_dst) +         \
+        FIELD_SIZEOF(struct ovs_key_ipv4_tunnel, tp_dst))
+
+struct ovs_key_ipv4_tunnel {
+       __be64 tun_id;
+       __be32 ipv4_src;
+       __be32 ipv4_dst;
+       __be16 tun_flags;
+       u8   ipv4_tos;
+       u8   ipv4_ttl;
+       __be16 tp_src;
+       __be16 tp_dst;
+} __packed __aligned(4); /* Minimize padding. */
+
+struct ovs_tunnel_info {
+       struct ovs_key_ipv4_tunnel tunnel;
+       const void *options;
+       u8 options_len;
+};
+
+/* Store options at the end of the array if they are less than the
+ * maximum size. This allows us to get the benefits of variable length
+ * matching for small options.
+ */
+#define TUN_METADATA_OFFSET(opt_len) \
+       (FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
+#define TUN_METADATA_OPTS(flow_key, opt_len) \
+       ((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
+
+static inline void __ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
+                                           __be32 saddr, __be32 daddr,
+                                           u8 tos, u8 ttl,
+                                           __be16 tp_src,
+                                           __be16 tp_dst,
+                                           __be64 tun_id,
+                                           __be16 tun_flags,
+                                           const void *opts,
+                                           u8 opts_len)
+{
+       tun_info->tunnel.tun_id = tun_id;
+       tun_info->tunnel.ipv4_src = saddr;
+       tun_info->tunnel.ipv4_dst = daddr;
+       tun_info->tunnel.ipv4_tos = tos;
+       tun_info->tunnel.ipv4_ttl = ttl;
+       tun_info->tunnel.tun_flags = tun_flags;
+
+       /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
+        * the upper tunnel are used.
+        * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
+        */
+       tun_info->tunnel.tp_src = tp_src;
+       tun_info->tunnel.tp_dst = tp_dst;
+
+       /* Clear struct padding. */
+       if (sizeof(tun_info->tunnel) != OVS_TUNNEL_KEY_SIZE)
+               memset((unsigned char *)&tun_info->tunnel + OVS_TUNNEL_KEY_SIZE,
+                      0, sizeof(tun_info->tunnel) - OVS_TUNNEL_KEY_SIZE);
+
+       tun_info->options = opts;
+       tun_info->options_len = opts_len;
+}
+
+static inline void ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
+                                         const struct iphdr *iph,
+                                         __be16 tp_src,
+                                         __be16 tp_dst,
+                                         __be64 tun_id,
+                                         __be16 tun_flags,
+                                         const void *opts,
+                                         u8 opts_len)
+{
+       __ovs_flow_tun_info_init(tun_info, iph->saddr, iph->daddr,
+                                iph->tos, iph->ttl,
+                                tp_src, tp_dst,
+                                tun_id, tun_flags,
+                                opts, opts_len);
+}
+
+#define OVS_SW_FLOW_KEY_METADATA_SIZE                  \
+       (offsetof(struct sw_flow_key, recirc_id) +      \
+       FIELD_SIZEOF(struct sw_flow_key, recirc_id))
+
+struct sw_flow_key {
+       u8 tun_opts[255];
+       u8 tun_opts_len;
+       struct ovs_key_ipv4_tunnel tun_key;  /* Encapsulating tunnel key. */
+       struct {
+               u32     priority;       /* Packet QoS priority. */
+               u32     skb_mark;       /* SKB mark. */
+               u16     in_port;        /* Input switch port (or DP_MAX_PORTS). */
+       } __packed phy; /* Safe when right after 'tun_key'. */
+       u32 ovs_flow_hash;              /* Datapath computed hash value.  */
+       u32 recirc_id;                  /* Recirculation ID.  */
+       struct {
+               u8     src[ETH_ALEN];   /* Ethernet source address. */
+               u8     dst[ETH_ALEN];   /* Ethernet destination address. */
+               __be16 tci;             /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
+               __be16 type;            /* Ethernet frame type. */
+       } eth;
+       union {
+               struct {
+                       __be32 top_lse; /* top label stack entry */
+               } mpls;
+               struct {
+                       u8     proto;   /* IP protocol or lower 8 bits of ARP opcode. */
+                       u8     tos;         /* IP ToS. */
+                       u8     ttl;         /* IP TTL/hop limit. */
+                       u8     frag;    /* One of OVS_FRAG_TYPE_*. */
+               } ip;
+       };
+       struct {
+               __be16 src;             /* TCP/UDP/SCTP source port. */
+               __be16 dst;             /* TCP/UDP/SCTP destination port. */
+               __be16 flags;           /* TCP flags. */
+       } tp;
+       union {
+               struct {
+                       struct {
+                               __be32 src;     /* IP source address. */
+                               __be32 dst;     /* IP destination address. */
+                       } addr;
+                       struct {
+                               u8 sha[ETH_ALEN];       /* ARP source hardware address. */
+                               u8 tha[ETH_ALEN];       /* ARP target hardware address. */
+                       } arp;
+               } ipv4;
+               struct {
+                       struct {
+                               struct in6_addr src;    /* IPv6 source address. */
+                               struct in6_addr dst;    /* IPv6 destination address. */
+                       } addr;
+                       __be32 label;                   /* IPv6 flow label. */
+                       struct {
+                               struct in6_addr target; /* ND target address. */
+                               u8 sll[ETH_ALEN];       /* ND source link layer address. */
+                               u8 tll[ETH_ALEN];       /* ND target link layer address. */
+                       } nd;
+               } ipv6;
+       };
+} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
+
+struct sw_flow_key_range {
+       unsigned short int start;
+       unsigned short int end;
+};
+
+struct sw_flow_mask {
+       int ref_count;
+       struct rcu_head rcu;
+       struct list_head list;
+       struct sw_flow_key_range range;
+       struct sw_flow_key key;
+};
+
+struct sw_flow_match {
+       struct sw_flow_key *key;
+       struct sw_flow_key_range range;
+       struct sw_flow_mask *mask;
+};
+
+#define MAX_UFID_LENGTH 16 /* 128 bits */
+
+struct sw_flow_id {
+       u32 ufid_len;
+       union {
+               u32 ufid[MAX_UFID_LENGTH / 4];
+               struct sw_flow_key *unmasked_key;
+       };
+};
+
+struct sw_flow_actions {
+       struct rcu_head rcu;
+       u32 actions_len;
+       struct nlattr actions[];
+};
+
+struct flow_stats {
+       u64 packet_count;               /* Number of packets matched. */
+       u64 byte_count;                 /* Number of bytes matched. */
+       unsigned long used;             /* Last used time (in jiffies). */
+       spinlock_t lock;                /* Lock for atomic stats update. */
+       __be16 tcp_flags;               /* Union of seen TCP flags. */
+};
+
+struct sw_flow {
+       struct rcu_head rcu;
+       struct {
+               struct hlist_node node[2];
+               u32 hash;
+       } flow_table, ufid_table;
+       int stats_last_writer;          /* NUMA-node id of the last writer on
+                                        * 'stats[0]'.
+                                        */
+       struct sw_flow_key key;
+       struct sw_flow_id id;
+       struct sw_flow_mask *mask;
+       struct sw_flow_actions __rcu *sf_acts;
+       struct flow_stats __rcu *stats[]; /* One for each NUMA node.  First one
+                                          * is allocated at flow creation time,
+                                          * the rest are allocated on demand
+                                          * while holding the 'stats[0].lock'.
+                                          */
+};
+
+struct arp_eth_header {
+       __be16      ar_hrd;     /* format of hardware address   */
+       __be16      ar_pro;     /* format of protocol address   */
+       unsigned char   ar_hln; /* length of hardware address   */
+       unsigned char   ar_pln; /* length of protocol address   */
+       __be16      ar_op;      /* ARP opcode (command)     */
+
+       /* Ethernet+IPv4 specific members. */
+       unsigned char       ar_sha[ETH_ALEN];   /* sender hardware address  */
+       unsigned char       ar_sip[4];          /* sender IP address        */
+       unsigned char       ar_tha[ETH_ALEN];   /* target hardware address  */
+       unsigned char       ar_tip[4];          /* target IP address        */
+} __packed;
+
+static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
+{
+       return sfid->ufid_len;
+}
+
+static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
+{
+       return !ovs_identifier_is_ufid(sfid);
+}
+
+void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
+                          const struct sk_buff *);
+void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
+                       unsigned long *used, __be16 *tcp_flags);
+void ovs_flow_stats_clear(struct sw_flow *);
+u64 ovs_flow_used_time(unsigned long flow_jiffies);
+
+int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
+int ovs_flow_key_extract(const struct ovs_tunnel_info *tun_info,
+                        struct sk_buff *skb,
+                        struct sw_flow_key *key);
+/* Extract key from packet coming from userspace. */
+int ovs_flow_key_extract_userspace(const struct nlattr *attr,
+                                  struct sk_buff *skb,
+                                  struct sw_flow_key *key, bool log);
+
+#endif /* flow.h */