]> git.proxmox.com Git - ovs.git/blob - lib/flow.c
ovs-ofctl: Fix small typo about nw_tos in man page.
[ovs.git] / lib / flow.c
1 /*
2 * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks.
3 *
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:
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
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.
15 */
16 #include <config.h>
17 #include <sys/types.h>
18 #include "flow.h"
19 #include <inttypes.h>
20 #include <netinet/in.h>
21 #include <stdlib.h>
22 #include <string.h>
23 #include "byte-order.h"
24 #include "coverage.h"
25 #include "dpif.h"
26 #include "dynamic-string.h"
27 #include "hash.h"
28 #include "ofpbuf.h"
29 #include "openflow/openflow.h"
30 #include "openvswitch/datapath-protocol.h"
31 #include "packets.h"
32 #include "unaligned.h"
33 #include "vlog.h"
34
35 VLOG_DEFINE_THIS_MODULE(flow);
36
37 COVERAGE_DEFINE(flow_extract);
38
39 static struct arp_eth_header *
40 pull_arp(struct ofpbuf *packet)
41 {
42 return ofpbuf_try_pull(packet, ARP_ETH_HEADER_LEN);
43 }
44
45 static struct ip_header *
46 pull_ip(struct ofpbuf *packet)
47 {
48 if (packet->size >= IP_HEADER_LEN) {
49 struct ip_header *ip = packet->data;
50 int ip_len = IP_IHL(ip->ip_ihl_ver) * 4;
51 if (ip_len >= IP_HEADER_LEN && packet->size >= ip_len) {
52 return ofpbuf_pull(packet, ip_len);
53 }
54 }
55 return NULL;
56 }
57
58 static struct tcp_header *
59 pull_tcp(struct ofpbuf *packet)
60 {
61 if (packet->size >= TCP_HEADER_LEN) {
62 struct tcp_header *tcp = packet->data;
63 int tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4;
64 if (tcp_len >= TCP_HEADER_LEN && packet->size >= tcp_len) {
65 return ofpbuf_pull(packet, tcp_len);
66 }
67 }
68 return NULL;
69 }
70
71 static struct udp_header *
72 pull_udp(struct ofpbuf *packet)
73 {
74 return ofpbuf_try_pull(packet, UDP_HEADER_LEN);
75 }
76
77 static struct icmp_header *
78 pull_icmp(struct ofpbuf *packet)
79 {
80 return ofpbuf_try_pull(packet, ICMP_HEADER_LEN);
81 }
82
83 static void
84 parse_vlan(struct ofpbuf *b, struct flow *flow)
85 {
86 struct qtag_prefix {
87 ovs_be16 eth_type; /* ETH_TYPE_VLAN */
88 ovs_be16 tci;
89 };
90
91 if (b->size >= sizeof(struct qtag_prefix) + sizeof(ovs_be16)) {
92 struct qtag_prefix *qp = ofpbuf_pull(b, sizeof *qp);
93 flow->vlan_tci = qp->tci | htons(VLAN_CFI);
94 }
95 }
96
97 static ovs_be16
98 parse_ethertype(struct ofpbuf *b)
99 {
100 struct llc_snap_header *llc;
101 ovs_be16 proto;
102
103 proto = *(ovs_be16 *) ofpbuf_pull(b, sizeof proto);
104 if (ntohs(proto) >= ETH_TYPE_MIN) {
105 return proto;
106 }
107
108 if (b->size < sizeof *llc) {
109 return htons(FLOW_DL_TYPE_NONE);
110 }
111
112 llc = b->data;
113 if (llc->llc.llc_dsap != LLC_DSAP_SNAP
114 || llc->llc.llc_ssap != LLC_SSAP_SNAP
115 || llc->llc.llc_cntl != LLC_CNTL_SNAP
116 || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET,
117 sizeof llc->snap.snap_org)) {
118 return htons(FLOW_DL_TYPE_NONE);
119 }
120
121 ofpbuf_pull(b, sizeof *llc);
122 return llc->snap.snap_type;
123 }
124
125 /* Initializes 'flow' members from 'packet', 'tun_id', and 'in_port.
126 * Initializes 'packet' header pointers as follows:
127 *
128 * - packet->l2 to the start of the Ethernet header.
129 *
130 * - packet->l3 to just past the Ethernet header, or just past the
131 * vlan_header if one is present, to the first byte of the payload of the
132 * Ethernet frame.
133 *
134 * - packet->l4 to just past the IPv4 header, if one is present and has a
135 * correct length, and otherwise NULL.
136 *
137 * - packet->l7 to just past the TCP or UDP or ICMP header, if one is
138 * present and has a correct length, and otherwise NULL.
139 */
140 int
141 flow_extract(struct ofpbuf *packet, ovs_be64 tun_id, uint16_t in_port,
142 struct flow *flow)
143 {
144 struct ofpbuf b = *packet;
145 struct eth_header *eth;
146 int retval = 0;
147
148 COVERAGE_INC(flow_extract);
149
150 memset(flow, 0, sizeof *flow);
151 flow->tun_id = tun_id;
152 flow->in_port = in_port;
153
154 packet->l2 = b.data;
155 packet->l3 = NULL;
156 packet->l4 = NULL;
157 packet->l7 = NULL;
158
159 if (b.size < sizeof *eth) {
160 return 0;
161 }
162
163 /* Link layer. */
164 eth = b.data;
165 memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
166 memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
167
168 /* dl_type, vlan_tci. */
169 ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
170 if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
171 parse_vlan(&b, flow);
172 }
173 flow->dl_type = parse_ethertype(&b);
174
175 /* Network layer. */
176 packet->l3 = b.data;
177 if (flow->dl_type == htons(ETH_TYPE_IP)) {
178 const struct ip_header *nh = pull_ip(&b);
179 if (nh) {
180 flow->nw_src = get_unaligned_be32(&nh->ip_src);
181 flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
182 flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
183 flow->nw_proto = nh->ip_proto;
184 packet->l4 = b.data;
185 if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
186 if (flow->nw_proto == IP_TYPE_TCP) {
187 const struct tcp_header *tcp = pull_tcp(&b);
188 if (tcp) {
189 flow->tp_src = tcp->tcp_src;
190 flow->tp_dst = tcp->tcp_dst;
191 packet->l7 = b.data;
192 }
193 } else if (flow->nw_proto == IP_TYPE_UDP) {
194 const struct udp_header *udp = pull_udp(&b);
195 if (udp) {
196 flow->tp_src = udp->udp_src;
197 flow->tp_dst = udp->udp_dst;
198 packet->l7 = b.data;
199 }
200 } else if (flow->nw_proto == IP_TYPE_ICMP) {
201 const struct icmp_header *icmp = pull_icmp(&b);
202 if (icmp) {
203 flow->icmp_type = htons(icmp->icmp_type);
204 flow->icmp_code = htons(icmp->icmp_code);
205 packet->l7 = b.data;
206 }
207 }
208 } else {
209 retval = 1;
210 }
211 }
212 } else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
213 const struct arp_eth_header *arp = pull_arp(&b);
214 if (arp && arp->ar_hrd == htons(1)
215 && arp->ar_pro == htons(ETH_TYPE_IP)
216 && arp->ar_hln == ETH_ADDR_LEN
217 && arp->ar_pln == 4) {
218 /* We only match on the lower 8 bits of the opcode. */
219 if (ntohs(arp->ar_op) <= 0xff) {
220 flow->nw_proto = ntohs(arp->ar_op);
221 }
222
223 if ((flow->nw_proto == ARP_OP_REQUEST)
224 || (flow->nw_proto == ARP_OP_REPLY)) {
225 flow->nw_src = arp->ar_spa;
226 flow->nw_dst = arp->ar_tpa;
227 }
228 }
229 }
230 return retval;
231 }
232
233 /* Extracts the flow stats for a packet. The 'flow' and 'packet'
234 * arguments must have been initialized through a call to flow_extract().
235 */
236 void
237 flow_extract_stats(const struct flow *flow, struct ofpbuf *packet,
238 struct dpif_flow_stats *stats)
239 {
240 memset(stats, 0, sizeof(*stats));
241
242 if ((flow->dl_type == htons(ETH_TYPE_IP)) && packet->l4) {
243 if ((flow->nw_proto == IP_TYPE_TCP) && packet->l7) {
244 struct tcp_header *tcp = packet->l4;
245 stats->tcp_flags = TCP_FLAGS(tcp->tcp_ctl);
246 }
247 }
248
249 stats->n_bytes = packet->size;
250 stats->n_packets = 1;
251 }
252
253 char *
254 flow_to_string(const struct flow *flow)
255 {
256 struct ds ds = DS_EMPTY_INITIALIZER;
257 flow_format(&ds, flow);
258 return ds_cstr(&ds);
259 }
260
261 void
262 flow_format(struct ds *ds, const struct flow *flow)
263 {
264 ds_put_format(ds, "tunnel%#"PRIx64":in_port%04"PRIx16":tci(",
265 flow->tun_id, flow->in_port);
266 if (flow->vlan_tci) {
267 ds_put_format(ds, "vlan%"PRIu16",pcp%d",
268 vlan_tci_to_vid(flow->vlan_tci),
269 vlan_tci_to_pcp(flow->vlan_tci));
270 } else {
271 ds_put_char(ds, '0');
272 }
273 ds_put_format(ds, ") mac"ETH_ADDR_FMT"->"ETH_ADDR_FMT
274 " type%04"PRIx16
275 " proto%"PRIu8
276 " tos%"PRIu8
277 " ip"IP_FMT"->"IP_FMT
278 " port%"PRIu16"->%"PRIu16,
279 ETH_ADDR_ARGS(flow->dl_src),
280 ETH_ADDR_ARGS(flow->dl_dst),
281 ntohs(flow->dl_type),
282 flow->nw_proto,
283 flow->nw_tos,
284 IP_ARGS(&flow->nw_src),
285 IP_ARGS(&flow->nw_dst),
286 ntohs(flow->tp_src),
287 ntohs(flow->tp_dst));
288 }
289
290 void
291 flow_print(FILE *stream, const struct flow *flow)
292 {
293 char *s = flow_to_string(flow);
294 fputs(s, stream);
295 free(s);
296 }
297 \f
298 /* flow_wildcards functions. */
299
300 /* Initializes 'wc' as a set of wildcards that matches every packet. */
301 void
302 flow_wildcards_init_catchall(struct flow_wildcards *wc)
303 {
304 wc->wildcards = FWW_ALL;
305 wc->tun_id_mask = htonll(0);
306 wc->nw_src_mask = htonl(0);
307 wc->nw_dst_mask = htonl(0);
308 memset(wc->reg_masks, 0, sizeof wc->reg_masks);
309 wc->vlan_tci_mask = htons(0);
310 wc->zero = 0;
311 }
312
313 /* Initializes 'wc' as an exact-match set of wildcards; that is, 'wc' does not
314 * wildcard any bits or fields. */
315 void
316 flow_wildcards_init_exact(struct flow_wildcards *wc)
317 {
318 wc->wildcards = 0;
319 wc->tun_id_mask = htonll(UINT64_MAX);
320 wc->nw_src_mask = htonl(UINT32_MAX);
321 wc->nw_dst_mask = htonl(UINT32_MAX);
322 memset(wc->reg_masks, 0xff, sizeof wc->reg_masks);
323 wc->vlan_tci_mask = htons(UINT16_MAX);
324 wc->zero = 0;
325 }
326
327 /* Returns true if 'wc' is exact-match, false if 'wc' wildcards any bits or
328 * fields. */
329 bool
330 flow_wildcards_is_exact(const struct flow_wildcards *wc)
331 {
332 int i;
333
334 if (wc->wildcards
335 || wc->tun_id_mask != htonll(UINT64_MAX)
336 || wc->nw_src_mask != htonl(UINT32_MAX)
337 || wc->nw_dst_mask != htonl(UINT32_MAX)
338 || wc->vlan_tci_mask != htons(UINT16_MAX)) {
339 return false;
340 }
341
342 for (i = 0; i < FLOW_N_REGS; i++) {
343 if (wc->reg_masks[i] != htonl(UINT32_MAX)) {
344 return false;
345 }
346 }
347
348 return true;
349 }
350
351 /* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
352 * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
353 * 'src1' or 'src2' or both. */
354 void
355 flow_wildcards_combine(struct flow_wildcards *dst,
356 const struct flow_wildcards *src1,
357 const struct flow_wildcards *src2)
358 {
359 int i;
360
361 dst->wildcards = src1->wildcards | src2->wildcards;
362 dst->tun_id_mask = src1->tun_id_mask & src2->tun_id_mask;
363 dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
364 dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
365 for (i = 0; i < FLOW_N_REGS; i++) {
366 dst->reg_masks[i] = src1->reg_masks[i] & src2->reg_masks[i];
367 }
368 dst->vlan_tci_mask = src1->vlan_tci_mask & src2->vlan_tci_mask;
369 }
370
371 /* Returns a hash of the wildcards in 'wc'. */
372 uint32_t
373 flow_wildcards_hash(const struct flow_wildcards *wc)
374 {
375 /* If you change struct flow_wildcards and thereby trigger this
376 * assertion, please check that the new struct flow_wildcards has no holes
377 * in it before you update the assertion. */
378 BUILD_ASSERT_DECL(sizeof *wc == 24 + FLOW_N_REGS * 4);
379 return hash_bytes(wc, sizeof *wc, 0);
380 }
381
382 /* Returns true if 'a' and 'b' represent the same wildcards, false if they are
383 * different. */
384 bool
385 flow_wildcards_equal(const struct flow_wildcards *a,
386 const struct flow_wildcards *b)
387 {
388 int i;
389
390 if (a->wildcards != b->wildcards
391 || a->tun_id_mask != b->tun_id_mask
392 || a->nw_src_mask != b->nw_src_mask
393 || a->nw_dst_mask != b->nw_dst_mask
394 || a->vlan_tci_mask != b->vlan_tci_mask) {
395 return false;
396 }
397
398 for (i = 0; i < FLOW_N_REGS; i++) {
399 if (a->reg_masks[i] != b->reg_masks[i]) {
400 return false;
401 }
402 }
403
404 return true;
405 }
406
407 /* Returns true if at least one bit or field is wildcarded in 'a' but not in
408 * 'b', false otherwise. */
409 bool
410 flow_wildcards_has_extra(const struct flow_wildcards *a,
411 const struct flow_wildcards *b)
412 {
413 int i;
414
415 for (i = 0; i < FLOW_N_REGS; i++) {
416 if ((a->reg_masks[i] & b->reg_masks[i]) != b->reg_masks[i]) {
417 return true;
418 }
419 }
420
421 return (a->wildcards & ~b->wildcards
422 || (a->tun_id_mask & b->tun_id_mask) != b->tun_id_mask
423 || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
424 || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask
425 || (a->vlan_tci_mask & b->vlan_tci_mask) != b->vlan_tci_mask);
426 }
427
428 static bool
429 set_nw_mask(ovs_be32 *maskp, ovs_be32 mask)
430 {
431 if (ip_is_cidr(mask)) {
432 *maskp = mask;
433 return true;
434 } else {
435 return false;
436 }
437 }
438
439 /* Sets the IP (or ARP) source wildcard mask to CIDR 'mask' (consisting of N
440 * high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
441 * false if 'mask' is not a CIDR mask. */
442 bool
443 flow_wildcards_set_nw_src_mask(struct flow_wildcards *wc, ovs_be32 mask)
444 {
445 return set_nw_mask(&wc->nw_src_mask, mask);
446 }
447
448 /* Sets the IP (or ARP) destination wildcard mask to CIDR 'mask' (consisting of
449 * N high-order 1-bit and 32-N low-order 0-bits). Returns true if successful,
450 * false if 'mask' is not a CIDR mask. */
451 bool
452 flow_wildcards_set_nw_dst_mask(struct flow_wildcards *wc, ovs_be32 mask)
453 {
454 return set_nw_mask(&wc->nw_dst_mask, mask);
455 }
456
457 /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'.
458 * (A 0-bit indicates a wildcard bit.) */
459 void
460 flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask)
461 {
462 wc->reg_masks[idx] = mask;
463 }