2 * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include <sys/types.h>
20 #include <netinet/in.h>
24 #include "byte-order.h"
25 #include "openflow/nicira-ext.h"
26 #include "openflow/openflow.h"
30 struct dpif_flow_stats;
32 struct flow_wildcards;
37 /* This sequence number should be incremented whenever anything involving flows
38 * or the wildcarding of flows changes. This will cause build assertion
39 * failures in places which likely need to be updated. */
40 #define FLOW_WC_SEQ 21
43 BUILD_ASSERT_DECL(FLOW_N_REGS <= NXM_NX_MAX_REGS);
45 /* Used for struct flow's dl_type member for frames that have no Ethernet
46 * type, that is, pure 802.2 frames. */
47 #define FLOW_DL_TYPE_NONE 0x5ff
49 /* Fragment bits, used for IPv4 and IPv6, always zero for non-IP flows. */
50 #define FLOW_NW_FRAG_ANY (1 << 0) /* Set for any IP frag. */
51 #define FLOW_NW_FRAG_LATER (1 << 1) /* Set for IP frag with nonzero offset. */
52 #define FLOW_NW_FRAG_MASK (FLOW_NW_FRAG_ANY | FLOW_NW_FRAG_LATER)
54 BUILD_ASSERT_DECL(FLOW_NW_FRAG_ANY == NX_IP_FRAG_ANY);
55 BUILD_ASSERT_DECL(FLOW_NW_FRAG_LATER == NX_IP_FRAG_LATER);
57 #define FLOW_TNL_F_DONT_FRAGMENT (1 << 0)
58 #define FLOW_TNL_F_CSUM (1 << 1)
59 #define FLOW_TNL_F_KEY (1 << 2)
61 const char *flow_tun_flag_to_string(uint32_t flags);
72 /* Unfortunately, a "struct flow" sometimes has to handle OpenFlow port
73 * numbers and other times datapath (dpif) port numbers. This union allows
81 * A flow in the network.
83 * The meaning of 'in_port' is context-dependent. In most cases, it is a
84 * 16-bit OpenFlow 1.0 port number. In the software datapath interface (dpif)
85 * layer and its implementations (e.g. dpif-linux, dpif-netdev), it is instead
86 * a 32-bit datapath port number.
89 struct flow_tnl tunnel; /* Encapsulating tunnel parameters. */
90 ovs_be64 metadata; /* OpenFlow Metadata. */
91 struct in6_addr ipv6_src; /* IPv6 source address. */
92 struct in6_addr ipv6_dst; /* IPv6 destination address. */
93 struct in6_addr nd_target; /* IPv6 neighbor discovery (ND) target. */
94 uint32_t skb_priority; /* Packet priority for QoS. */
95 uint32_t regs[FLOW_N_REGS]; /* Registers. */
96 ovs_be32 nw_src; /* IPv4 source address. */
97 ovs_be32 nw_dst; /* IPv4 destination address. */
98 ovs_be32 ipv6_label; /* IPv6 flow label. */
99 union flow_in_port in_port; /* Input port.*/
100 uint32_t pkt_mark; /* Packet mark. */
101 ovs_be32 mpls_lse; /* MPLS label stack entry. */
102 ovs_be16 vlan_tci; /* If 802.1Q, TCI | VLAN_CFI; otherwise 0. */
103 ovs_be16 dl_type; /* Ethernet frame type. */
104 ovs_be16 tp_src; /* TCP/UDP/SCTP source port. */
105 ovs_be16 tp_dst; /* TCP/UDP/SCTP destination port. */
106 uint8_t dl_src[6]; /* Ethernet source address. */
107 uint8_t dl_dst[6]; /* Ethernet destination address. */
108 uint8_t nw_proto; /* IP protocol or low 8 bits of ARP opcode. */
109 uint8_t nw_tos; /* IP ToS (including DSCP and ECN). */
110 uint8_t arp_sha[6]; /* ARP/ND source hardware address. */
111 uint8_t arp_tha[6]; /* ARP/ND target hardware address. */
112 uint8_t nw_ttl; /* IP TTL/Hop Limit. */
113 uint8_t nw_frag; /* FLOW_FRAG_* flags. */
115 BUILD_ASSERT_DECL(sizeof(struct flow) % 4 == 0);
117 #define FLOW_U32S (sizeof(struct flow) / 4)
119 /* Remember to update FLOW_WC_SEQ when changing 'struct flow'. */
120 BUILD_ASSERT_DECL(sizeof(struct flow) == sizeof(struct flow_tnl) + 152 &&
123 /* Represents the metadata fields of struct flow. */
124 struct flow_metadata {
125 ovs_be64 tun_id; /* Encapsulating tunnel ID. */
126 ovs_be32 tun_src; /* Tunnel outer IPv4 src addr */
127 ovs_be32 tun_dst; /* Tunnel outer IPv4 dst addr */
128 ovs_be64 metadata; /* OpenFlow 1.1+ metadata field. */
129 uint32_t regs[FLOW_N_REGS]; /* Registers. */
130 uint32_t pkt_mark; /* Packet mark. */
131 ofp_port_t in_port; /* OpenFlow port or zero. */
134 void flow_extract(struct ofpbuf *, uint32_t priority, uint32_t mark,
135 const struct flow_tnl *, const union flow_in_port *in_port,
138 void flow_zero_wildcards(struct flow *, const struct flow_wildcards *);
139 void flow_get_metadata(const struct flow *, struct flow_metadata *);
141 char *flow_to_string(const struct flow *);
142 void format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t),
143 uint32_t flags, char del);
145 void flow_format(struct ds *, const struct flow *);
146 void flow_print(FILE *, const struct flow *);
147 static inline int flow_compare_3way(const struct flow *, const struct flow *);
148 static inline bool flow_equal(const struct flow *, const struct flow *);
149 static inline size_t flow_hash(const struct flow *, uint32_t basis);
151 void flow_set_dl_vlan(struct flow *, ovs_be16 vid);
152 void flow_set_vlan_vid(struct flow *, ovs_be16 vid);
153 void flow_set_vlan_pcp(struct flow *, uint8_t pcp);
155 void flow_set_mpls_label(struct flow *flow, ovs_be32 label);
156 void flow_set_mpls_ttl(struct flow *flow, uint8_t ttl);
157 void flow_set_mpls_tc(struct flow *flow, uint8_t tc);
158 void flow_set_mpls_bos(struct flow *flow, uint8_t stack);
160 void flow_compose(struct ofpbuf *, const struct flow *);
163 flow_compare_3way(const struct flow *a, const struct flow *b)
165 return memcmp(a, b, sizeof *a);
169 flow_equal(const struct flow *a, const struct flow *b)
171 return !flow_compare_3way(a, b);
175 flow_hash(const struct flow *flow, uint32_t basis)
177 return hash_words((const uint32_t *) flow, sizeof *flow / 4, basis);
180 static inline uint16_t
181 ofp_to_u16(ofp_port_t ofp_port)
183 return (OVS_FORCE uint16_t) ofp_port;
186 static inline uint32_t
187 odp_to_u32(odp_port_t odp_port)
189 return (OVS_FORCE uint32_t) odp_port;
192 static inline uint32_t
193 ofp11_to_u32(ofp11_port_t ofp11_port)
195 return (OVS_FORCE uint32_t) ofp11_port;
198 static inline ofp_port_t
199 u16_to_ofp(uint16_t port)
201 return OFP_PORT_C(port);
204 static inline odp_port_t
205 u32_to_odp(uint32_t port)
207 return ODP_PORT_C(port);
210 static inline ofp11_port_t
211 u32_to_ofp11(uint32_t port)
213 return OFP11_PORT_C(port);
216 static inline uint32_t
217 hash_ofp_port(ofp_port_t ofp_port)
219 return hash_int(ofp_to_u16(ofp_port), 0);
222 static inline uint32_t
223 hash_odp_port(odp_port_t odp_port)
225 return hash_int(odp_to_u32(odp_port), 0);
228 uint32_t flow_hash_in_minimask(const struct flow *, const struct minimask *,
231 /* Wildcards for a flow.
233 * A 1-bit in each bit in 'masks' indicates that the corresponding bit of
234 * the flow is significant (must match). A 0-bit indicates that the
235 * corresponding bit of the flow is wildcarded (need not match). */
236 struct flow_wildcards {
240 void flow_wildcards_init_catchall(struct flow_wildcards *);
241 void flow_wildcards_init_exact(struct flow_wildcards *);
243 bool flow_wildcards_is_catchall(const struct flow_wildcards *);
245 void flow_wildcards_set_reg_mask(struct flow_wildcards *,
246 int idx, uint32_t mask);
248 void flow_wildcards_and(struct flow_wildcards *dst,
249 const struct flow_wildcards *src1,
250 const struct flow_wildcards *src2);
251 void flow_wildcards_or(struct flow_wildcards *dst,
252 const struct flow_wildcards *src1,
253 const struct flow_wildcards *src2);
254 bool flow_wildcards_has_extra(const struct flow_wildcards *,
255 const struct flow_wildcards *);
257 void flow_wildcards_fold_minimask(struct flow_wildcards *,
258 const struct minimask *);
260 uint32_t flow_wildcards_hash(const struct flow_wildcards *, uint32_t basis);
261 bool flow_wildcards_equal(const struct flow_wildcards *,
262 const struct flow_wildcards *);
263 uint32_t flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis);
265 void flow_mask_hash_fields(const struct flow *, struct flow_wildcards *,
266 enum nx_hash_fields);
267 uint32_t flow_hash_fields(const struct flow *, enum nx_hash_fields,
269 const char *flow_hash_fields_to_str(enum nx_hash_fields);
270 bool flow_hash_fields_valid(enum nx_hash_fields);
272 uint32_t flow_hash_in_wildcards(const struct flow *,
273 const struct flow_wildcards *,
276 bool flow_equal_except(const struct flow *a, const struct flow *b,
277 const struct flow_wildcards *);
279 /* Compressed flow. */
281 #define MINI_N_INLINE (sizeof(void *) == 4 ? 7 : 8)
282 #define MINI_N_MAPS DIV_ROUND_UP(FLOW_U32S, 32)
284 /* A sparse representation of a "struct flow".
286 * A "struct flow" is fairly large and tends to be mostly zeros. Sparse
287 * representation has two advantages. First, it saves memory. Second, it
288 * saves time when the goal is to iterate over only the nonzero parts of the
291 * The 'map' member holds one bit for each uint32_t in a "struct flow". Each
292 * 0-bit indicates that the corresponding uint32_t is zero, each 1-bit that it
295 * 'values' points to the start of an array that has one element for each 1-bit
296 * in 'map'. The least-numbered 1-bit is in values[0], the next 1-bit is in
297 * values[1], and so on.
299 * 'values' may point to a few different locations:
301 * - If 'map' has MINI_N_INLINE or fewer 1-bits, it may point to
302 * 'inline_values'. One hopes that this is the common case.
304 * - If 'map' has more than MINI_N_INLINE 1-bits, it may point to memory
305 * allocated with malloc().
307 * - The caller could provide storage on the stack for situations where
308 * that makes sense. So far that's only proved useful for
309 * minimask_combine(), but the principle works elsewhere.
311 * Elements in 'values' are allowed to be zero. This is useful for "struct
312 * minimatch", for which ensuring that the miniflow and minimask members have
313 * same 'map' allows optimization .
317 uint32_t inline_values[MINI_N_INLINE];
318 uint32_t map[MINI_N_MAPS];
321 void miniflow_init(struct miniflow *, const struct flow *);
322 void miniflow_init_with_minimask(struct miniflow *, const struct flow *,
323 const struct minimask *);
324 void miniflow_clone(struct miniflow *, const struct miniflow *);
325 void miniflow_move(struct miniflow *dst, struct miniflow *);
326 void miniflow_destroy(struct miniflow *);
328 void miniflow_expand(const struct miniflow *, struct flow *);
330 uint32_t miniflow_get(const struct miniflow *, unsigned int u32_ofs);
331 uint16_t miniflow_get_vid(const struct miniflow *);
332 static inline ovs_be64 miniflow_get_metadata(const struct miniflow *);
334 bool miniflow_equal(const struct miniflow *a, const struct miniflow *b);
335 bool miniflow_equal_in_minimask(const struct miniflow *a,
336 const struct miniflow *b,
337 const struct minimask *);
338 bool miniflow_equal_flow_in_minimask(const struct miniflow *a,
339 const struct flow *b,
340 const struct minimask *);
341 uint32_t miniflow_hash(const struct miniflow *, uint32_t basis);
342 uint32_t miniflow_hash_in_minimask(const struct miniflow *,
343 const struct minimask *, uint32_t basis);
345 /* Compressed flow wildcards. */
347 /* A sparse representation of a "struct flow_wildcards".
349 * See the large comment on struct miniflow for details. */
351 struct miniflow masks;
354 void minimask_init(struct minimask *, const struct flow_wildcards *);
355 void minimask_clone(struct minimask *, const struct minimask *);
356 void minimask_move(struct minimask *dst, struct minimask *src);
357 void minimask_combine(struct minimask *dst,
358 const struct minimask *a, const struct minimask *b,
359 uint32_t storage[FLOW_U32S]);
360 void minimask_destroy(struct minimask *);
362 void minimask_expand(const struct minimask *, struct flow_wildcards *);
364 uint32_t minimask_get(const struct minimask *, unsigned int u32_ofs);
365 uint16_t minimask_get_vid_mask(const struct minimask *);
366 static inline ovs_be64 minimask_get_metadata_mask(const struct minimask *);
368 bool minimask_equal(const struct minimask *a, const struct minimask *b);
369 uint32_t minimask_hash(const struct minimask *, uint32_t basis);
371 bool minimask_has_extra(const struct minimask *, const struct minimask *);
372 bool minimask_is_catchall(const struct minimask *);
374 /* Returns the value of the OpenFlow 1.1+ "metadata" field in 'flow'. */
375 static inline ovs_be64
376 miniflow_get_metadata(const struct miniflow *flow)
378 enum { MD_OFS = offsetof(struct flow, metadata) };
379 BUILD_ASSERT_DECL(MD_OFS % sizeof(uint32_t) == 0);
380 ovs_be32 hi = (OVS_FORCE ovs_be32) miniflow_get(flow, MD_OFS / 4);
381 ovs_be32 lo = (OVS_FORCE ovs_be32) miniflow_get(flow, MD_OFS / 4 + 1);
383 return htonll(((uint64_t) ntohl(hi) << 32) | ntohl(lo));
386 /* Returns the mask for the OpenFlow 1.1+ "metadata" field in 'mask'.
388 * The return value is all-1-bits if 'mask' matches on the whole value of the
389 * metadata field, all-0-bits if 'mask' entirely wildcards the metadata field,
390 * or some other value if the metadata field is partially matched, partially
392 static inline ovs_be64
393 minimask_get_metadata_mask(const struct minimask *mask)
395 return miniflow_get_metadata(&mask->masks);