]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
8bfd0fda | 2 | * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc. |
064af421 | 3 | * |
a14bc59f BP |
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: | |
064af421 | 7 | * |
a14bc59f BP |
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. | |
064af421 BP |
15 | */ |
16 | #include <config.h> | |
17 | #include <sys/types.h> | |
18 | #include "flow.h" | |
d31f1109 | 19 | #include <errno.h> |
064af421 | 20 | #include <inttypes.h> |
5cb7a798 | 21 | #include <limits.h> |
064af421 | 22 | #include <netinet/in.h> |
d31f1109 JP |
23 | #include <netinet/icmp6.h> |
24 | #include <netinet/ip6.h> | |
5cb7a798 | 25 | #include <stdint.h> |
064af421 BP |
26 | #include <stdlib.h> |
27 | #include <string.h> | |
10a24935 | 28 | #include "byte-order.h" |
064af421 | 29 | #include "coverage.h" |
dc5a7ce7 | 30 | #include "csum.h" |
064af421 BP |
31 | #include "dynamic-string.h" |
32 | #include "hash.h" | |
c49d1dd1 | 33 | #include "jhash.h" |
aa6c9932 | 34 | #include "match.h" |
064af421 BP |
35 | #include "ofpbuf.h" |
36 | #include "openflow/openflow.h" | |
064af421 | 37 | #include "packets.h" |
b5e7e61a | 38 | #include "odp-util.h" |
94639963 | 39 | #include "random.h" |
176aaa65 | 40 | #include "unaligned.h" |
064af421 | 41 | |
d76f09ea | 42 | COVERAGE_DEFINE(flow_extract); |
5cb7a798 | 43 | COVERAGE_DEFINE(miniflow_malloc); |
d76f09ea | 44 | |
476f36e8 JR |
45 | /* U32 indices for segmented flow classification. */ |
46 | const uint8_t flow_segment_u32s[4] = { | |
47 | FLOW_SEGMENT_1_ENDS_AT / 4, | |
48 | FLOW_SEGMENT_2_ENDS_AT / 4, | |
49 | FLOW_SEGMENT_3_ENDS_AT / 4, | |
50 | FLOW_U32S | |
51 | }; | |
52 | ||
419681da JR |
53 | /* miniflow_extract() assumes the following to be true to optimize the |
54 | * extraction process. */ | |
55 | BUILD_ASSERT_DECL(offsetof(struct flow, dl_type) + 2 | |
56 | == offsetof(struct flow, vlan_tci) && | |
57 | offsetof(struct flow, dl_type) / 4 | |
58 | == offsetof(struct flow, vlan_tci) / 4 ); | |
59 | ||
60 | BUILD_ASSERT_DECL(offsetof(struct flow, nw_frag) + 3 | |
61 | == offsetof(struct flow, nw_proto) && | |
62 | offsetof(struct flow, nw_tos) + 2 | |
63 | == offsetof(struct flow, nw_proto) && | |
64 | offsetof(struct flow, nw_ttl) + 1 | |
65 | == offsetof(struct flow, nw_proto) && | |
66 | offsetof(struct flow, nw_frag) / 4 | |
67 | == offsetof(struct flow, nw_tos) / 4 && | |
68 | offsetof(struct flow, nw_ttl) / 4 | |
69 | == offsetof(struct flow, nw_tos) / 4 && | |
70 | offsetof(struct flow, nw_proto) / 4 | |
71 | == offsetof(struct flow, nw_tos) / 4); | |
72 | ||
73 | /* TCP flags in the first half of a BE32, zeroes in the other half. */ | |
74 | BUILD_ASSERT_DECL(offsetof(struct flow, tcp_flags) + 2 | |
c61f3870 | 75 | == offsetof(struct flow, pad2) && |
419681da | 76 | offsetof(struct flow, tcp_flags) / 4 |
c61f3870 | 77 | == offsetof(struct flow, pad2) / 4); |
419681da JR |
78 | #if WORDS_BIGENDIAN |
79 | #define TCP_FLAGS_BE32(tcp_ctl) ((OVS_FORCE ovs_be32)TCP_FLAGS_BE16(tcp_ctl) \ | |
80 | << 16) | |
81 | #else | |
82 | #define TCP_FLAGS_BE32(tcp_ctl) ((OVS_FORCE ovs_be32)TCP_FLAGS_BE16(tcp_ctl)) | |
83 | #endif | |
84 | ||
85 | BUILD_ASSERT_DECL(offsetof(struct flow, tp_src) + 2 | |
86 | == offsetof(struct flow, tp_dst) && | |
87 | offsetof(struct flow, tp_src) / 4 | |
88 | == offsetof(struct flow, tp_dst) / 4); | |
89 | ||
90 | /* Removes 'size' bytes from the head end of '*datap', of size '*sizep', which | |
91 | * must contain at least 'size' bytes of data. Returns the first byte of data | |
92 | * removed. */ | |
93 | static inline const void * | |
94 | data_pull(void **datap, size_t *sizep, size_t size) | |
a26ef517 | 95 | { |
419681da JR |
96 | char *data = (char *)*datap; |
97 | *datap = data + size; | |
98 | *sizep -= size; | |
99 | return data; | |
a26ef517 JP |
100 | } |
101 | ||
419681da JR |
102 | /* If '*datap' has at least 'size' bytes of data, removes that many bytes from |
103 | * the head end of '*datap' and returns the first byte removed. Otherwise, | |
104 | * returns a null pointer without modifying '*datap'. */ | |
105 | static inline const void * | |
106 | data_try_pull(void **datap, size_t *sizep, size_t size) | |
064af421 | 107 | { |
419681da | 108 | return OVS_LIKELY(*sizep >= size) ? data_pull(datap, sizep, size) : NULL; |
064af421 BP |
109 | } |
110 | ||
419681da JR |
111 | /* Context for pushing data to a miniflow. */ |
112 | struct mf_ctx { | |
113 | uint64_t map; | |
114 | uint32_t *data; | |
115 | uint32_t * const end; | |
116 | }; | |
064af421 | 117 | |
419681da JR |
118 | /* miniflow_push_* macros allow filling in a miniflow data values in order. |
119 | * Assertions are needed only when the layout of the struct flow is modified. | |
120 | * 'ofs' is a compile-time constant, which allows most of the code be optimized | |
694ffecc | 121 | * away. Some GCC versions gave warnings on ALWAYS_INLINE, so these are |
419681da JR |
122 | * defined as macros. */ |
123 | ||
c61f3870 | 124 | #if (FLOW_WC_SEQ != 28) |
419681da | 125 | #define MINIFLOW_ASSERT(X) ovs_assert(X) |
dce96af8 DDP |
126 | BUILD_MESSAGE("FLOW_WC_SEQ changed: miniflow_extract() will have runtime " |
127 | "assertions enabled. Consider updating FLOW_WC_SEQ after " | |
128 | "testing") | |
419681da JR |
129 | #else |
130 | #define MINIFLOW_ASSERT(X) | |
131 | #endif | |
132 | ||
133 | #define miniflow_push_uint32_(MF, OFS, VALUE) \ | |
134 | { \ | |
135 | MINIFLOW_ASSERT(MF.data < MF.end && (OFS) % 4 == 0 \ | |
136 | && !(MF.map & (UINT64_MAX << (OFS) / 4))); \ | |
137 | *MF.data++ = VALUE; \ | |
138 | MF.map |= UINT64_C(1) << (OFS) / 4; \ | |
d31f1109 JP |
139 | } |
140 | ||
419681da JR |
141 | #define miniflow_push_be32_(MF, OFS, VALUE) \ |
142 | miniflow_push_uint32_(MF, OFS, (OVS_FORCE uint32_t)(VALUE)) | |
143 | ||
144 | #define miniflow_push_uint16_(MF, OFS, VALUE) \ | |
145 | { \ | |
146 | MINIFLOW_ASSERT(MF.data < MF.end && \ | |
147 | (((OFS) % 4 == 0 && !(MF.map & (UINT64_MAX << (OFS) / 4))) \ | |
148 | || ((OFS) % 4 == 2 && MF.map & (UINT64_C(1) << (OFS) / 4) \ | |
149 | && !(MF.map & (UINT64_MAX << ((OFS) / 4 + 1)))))); \ | |
150 | \ | |
151 | if ((OFS) % 4 == 0) { \ | |
152 | *(uint16_t *)MF.data = VALUE; \ | |
153 | MF.map |= UINT64_C(1) << (OFS) / 4; \ | |
154 | } else if ((OFS) % 4 == 2) { \ | |
155 | *((uint16_t *)MF.data + 1) = VALUE; \ | |
156 | MF.data++; \ | |
157 | } \ | |
b02475c5 SH |
158 | } |
159 | ||
419681da JR |
160 | #define miniflow_push_be16_(MF, OFS, VALUE) \ |
161 | miniflow_push_uint16_(MF, OFS, (OVS_FORCE uint16_t)VALUE); | |
162 | ||
163 | /* Data at 'valuep' may be unaligned. */ | |
164 | #define miniflow_push_words_(MF, OFS, VALUEP, N_WORDS) \ | |
165 | { \ | |
166 | int ofs32 = (OFS) / 4; \ | |
167 | \ | |
168 | MINIFLOW_ASSERT(MF.data + (N_WORDS) <= MF.end && (OFS) % 4 == 0 \ | |
169 | && !(MF.map & (UINT64_MAX << ofs32))); \ | |
170 | \ | |
171 | memcpy(MF.data, (VALUEP), (N_WORDS) * sizeof *MF.data); \ | |
172 | MF.data += (N_WORDS); \ | |
173 | MF.map |= ((UINT64_MAX >> (64 - (N_WORDS))) << ofs32); \ | |
064af421 BP |
174 | } |
175 | ||
419681da JR |
176 | #define miniflow_push_uint32(MF, FIELD, VALUE) \ |
177 | miniflow_push_uint32_(MF, offsetof(struct flow, FIELD), VALUE) | |
50f06e16 | 178 | |
419681da JR |
179 | #define miniflow_push_be32(MF, FIELD, VALUE) \ |
180 | miniflow_push_be32_(MF, offsetof(struct flow, FIELD), VALUE) | |
50f06e16 | 181 | |
419681da JR |
182 | #define miniflow_push_uint32_check(MF, FIELD, VALUE) \ |
183 | { if (OVS_LIKELY(VALUE)) { \ | |
184 | miniflow_push_uint32_(MF, offsetof(struct flow, FIELD), VALUE); \ | |
185 | } \ | |
50f06e16 BP |
186 | } |
187 | ||
419681da JR |
188 | #define miniflow_push_be32_check(MF, FIELD, VALUE) \ |
189 | { if (OVS_LIKELY(VALUE)) { \ | |
190 | miniflow_push_be32_(MF, offsetof(struct flow, FIELD), VALUE); \ | |
191 | } \ | |
50f06e16 BP |
192 | } |
193 | ||
419681da JR |
194 | #define miniflow_push_uint16(MF, FIELD, VALUE) \ |
195 | miniflow_push_uint16_(MF, offsetof(struct flow, FIELD), VALUE) | |
9e69bc5f | 196 | |
419681da JR |
197 | #define miniflow_push_be16(MF, FIELD, VALUE) \ |
198 | miniflow_push_be16_(MF, offsetof(struct flow, FIELD), VALUE) | |
9e69bc5f | 199 | |
419681da JR |
200 | #define miniflow_push_words(MF, FIELD, VALUEP, N_WORDS) \ |
201 | miniflow_push_words_(MF, offsetof(struct flow, FIELD), VALUEP, N_WORDS) | |
064af421 | 202 | |
419681da JR |
203 | /* Pulls the MPLS headers at '*datap' and returns the count of them. */ |
204 | static inline int | |
205 | parse_mpls(void **datap, size_t *sizep) | |
d31f1109 | 206 | { |
419681da JR |
207 | const struct mpls_hdr *mh; |
208 | int count = 0; | |
d31f1109 | 209 | |
419681da JR |
210 | while ((mh = data_try_pull(datap, sizep, sizeof *mh))) { |
211 | count++; | |
212 | if (mh->mpls_lse.lo & htons(1 << MPLS_BOS_SHIFT)) { | |
d31f1109 JP |
213 | break; |
214 | } | |
419681da | 215 | } |
ba8561c6 | 216 | return MIN(count, FLOW_MAX_MPLS_LABELS); |
419681da | 217 | } |
d31f1109 | 218 | |
419681da JR |
219 | static inline ovs_be16 |
220 | parse_vlan(void **datap, size_t *sizep) | |
221 | { | |
222 | const struct eth_header *eth = *datap; | |
d31f1109 | 223 | |
419681da JR |
224 | struct qtag_prefix { |
225 | ovs_be16 eth_type; /* ETH_TYPE_VLAN */ | |
226 | ovs_be16 tci; | |
227 | }; | |
d31f1109 | 228 | |
419681da | 229 | data_pull(datap, sizep, ETH_ADDR_LEN * 2); |
d31f1109 | 230 | |
419681da JR |
231 | if (eth->eth_type == htons(ETH_TYPE_VLAN)) { |
232 | if (OVS_LIKELY(*sizep | |
233 | >= sizeof(struct qtag_prefix) + sizeof(ovs_be16))) { | |
234 | const struct qtag_prefix *qp = data_pull(datap, sizep, sizeof *qp); | |
235 | return qp->tci | htons(VLAN_CFI); | |
d31f1109 JP |
236 | } |
237 | } | |
88366484 | 238 | return 0; |
d31f1109 JP |
239 | } |
240 | ||
419681da JR |
241 | static inline ovs_be16 |
242 | parse_ethertype(void **datap, size_t *sizep) | |
88366484 | 243 | { |
419681da JR |
244 | const struct llc_snap_header *llc; |
245 | ovs_be16 proto; | |
5a51b2cd | 246 | |
419681da JR |
247 | proto = *(ovs_be16 *) data_pull(datap, sizep, sizeof proto); |
248 | if (OVS_LIKELY(ntohs(proto) >= ETH_TYPE_MIN)) { | |
249 | return proto; | |
88366484 | 250 | } |
5a51b2cd | 251 | |
419681da JR |
252 | if (OVS_UNLIKELY(*sizep < sizeof *llc)) { |
253 | return htons(FLOW_DL_TYPE_NONE); | |
88366484 | 254 | } |
5a51b2cd | 255 | |
419681da JR |
256 | llc = *datap; |
257 | if (OVS_UNLIKELY(llc->llc.llc_dsap != LLC_DSAP_SNAP | |
258 | || llc->llc.llc_ssap != LLC_SSAP_SNAP | |
259 | || llc->llc.llc_cntl != LLC_CNTL_SNAP | |
260 | || memcmp(llc->snap.snap_org, SNAP_ORG_ETHERNET, | |
261 | sizeof llc->snap.snap_org))) { | |
262 | return htons(FLOW_DL_TYPE_NONE); | |
c6bcb685 | 263 | } |
c6bcb685 | 264 | |
419681da | 265 | data_pull(datap, sizep, sizeof *llc); |
685a51a5 | 266 | |
419681da JR |
267 | if (OVS_LIKELY(ntohs(llc->snap.snap_type) >= ETH_TYPE_MIN)) { |
268 | return llc->snap.snap_type; | |
685a51a5 JP |
269 | } |
270 | ||
419681da JR |
271 | return htons(FLOW_DL_TYPE_NONE); |
272 | } | |
685a51a5 | 273 | |
419681da JR |
274 | static inline bool |
275 | parse_icmpv6(void **datap, size_t *sizep, const struct icmp6_hdr *icmp, | |
276 | const struct in6_addr **nd_target, | |
277 | uint8_t arp_buf[2][ETH_ADDR_LEN]) | |
278 | { | |
88366484 JG |
279 | if (icmp->icmp6_code == 0 && |
280 | (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || | |
281 | icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { | |
685a51a5 | 282 | |
b0e2ec32 | 283 | *nd_target = data_try_pull(datap, sizep, sizeof **nd_target); |
419681da JR |
284 | if (OVS_UNLIKELY(!*nd_target)) { |
285 | return false; | |
685a51a5 | 286 | } |
685a51a5 | 287 | |
419681da | 288 | while (*sizep >= 8) { |
685a51a5 JP |
289 | /* The minimum size of an option is 8 bytes, which also is |
290 | * the size of Ethernet link-layer options. */ | |
419681da | 291 | const struct nd_opt_hdr *nd_opt = *datap; |
88366484 JG |
292 | int opt_len = nd_opt->nd_opt_len * 8; |
293 | ||
419681da | 294 | if (!opt_len || opt_len > *sizep) { |
685a51a5 JP |
295 | goto invalid; |
296 | } | |
685a51a5 JP |
297 | |
298 | /* Store the link layer address if the appropriate option is | |
299 | * provided. It is considered an error if the same link | |
300 | * layer option is specified twice. */ | |
301 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LINKADDR | |
302 | && opt_len == 8) { | |
419681da JR |
303 | if (OVS_LIKELY(eth_addr_is_zero(arp_buf[0]))) { |
304 | memcpy(arp_buf[0], nd_opt + 1, ETH_ADDR_LEN); | |
685a51a5 JP |
305 | } else { |
306 | goto invalid; | |
307 | } | |
308 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LINKADDR | |
309 | && opt_len == 8) { | |
419681da JR |
310 | if (OVS_LIKELY(eth_addr_is_zero(arp_buf[1]))) { |
311 | memcpy(arp_buf[1], nd_opt + 1, ETH_ADDR_LEN); | |
685a51a5 JP |
312 | } else { |
313 | goto invalid; | |
314 | } | |
315 | } | |
316 | ||
419681da | 317 | if (OVS_UNLIKELY(!data_try_pull(datap, sizep, opt_len))) { |
685a51a5 JP |
318 | goto invalid; |
319 | } | |
685a51a5 JP |
320 | } |
321 | } | |
322 | ||
419681da | 323 | return true; |
685a51a5 JP |
324 | |
325 | invalid: | |
419681da | 326 | return false; |
685a51a5 JP |
327 | } |
328 | ||
b5e7e61a | 329 | /* Initializes 'flow' members from 'packet' and 'md' |
deedf7e7 | 330 | * |
437d0d22 JR |
331 | * Initializes 'packet' header l2 pointer to the start of the Ethernet |
332 | * header, and the layer offsets as follows: | |
ca78c6b6 | 333 | * |
437d0d22 JR |
334 | * - packet->l2_5_ofs to the start of the MPLS shim header, or UINT16_MAX |
335 | * when there is no MPLS shim header. | |
ca78c6b6 | 336 | * |
437d0d22 | 337 | * - packet->l3_ofs to just past the Ethernet header, or just past the |
ca78c6b6 | 338 | * vlan_header if one is present, to the first byte of the payload of the |
437d0d22 JR |
339 | * Ethernet frame. UINT16_MAX if the frame is too short to contain an |
340 | * Ethernet header. | |
ca78c6b6 | 341 | * |
437d0d22 JR |
342 | * - packet->l4_ofs to just past the IPv4 header, if one is present and |
343 | * has at least the content used for the fields of interest for the flow, | |
344 | * otherwise UINT16_MAX. | |
ca78c6b6 | 345 | */ |
7257b535 | 346 | void |
b5e7e61a | 347 | flow_extract(struct ofpbuf *packet, const struct pkt_metadata *md, |
296e07ac | 348 | struct flow *flow) |
064af421 | 349 | { |
27bbe15d JR |
350 | struct { |
351 | struct miniflow mf; | |
352 | uint32_t buf[FLOW_U32S]; | |
353 | } m; | |
064af421 BP |
354 | |
355 | COVERAGE_INC(flow_extract); | |
356 | ||
27bbe15d JR |
357 | miniflow_initialize(&m.mf, m.buf); |
358 | miniflow_extract(packet, md, &m.mf); | |
359 | miniflow_expand(&m.mf, flow); | |
419681da | 360 | } |
296e07ac | 361 | |
27bbe15d JR |
362 | /* Caller is responsible for initializing 'dst' with enough storage for |
363 | * FLOW_U32S * 4 bytes. */ | |
419681da JR |
364 | void |
365 | miniflow_extract(struct ofpbuf *packet, const struct pkt_metadata *md, | |
366 | struct miniflow *dst) | |
367 | { | |
368 | void *data = ofpbuf_data(packet); | |
369 | size_t size = ofpbuf_size(packet); | |
27bbe15d JR |
370 | uint32_t *values = miniflow_values(dst); |
371 | struct mf_ctx mf = { 0, values, values + FLOW_U32S }; | |
419681da | 372 | char *l2; |
419681da JR |
373 | ovs_be16 dl_type; |
374 | uint8_t nw_frag, nw_tos, nw_ttl, nw_proto; | |
375 | ||
376 | /* Metadata. */ | |
b5e7e61a | 377 | if (md) { |
419681da JR |
378 | if (md->tunnel.ip_dst) { |
379 | miniflow_push_words(mf, tunnel, &md->tunnel, | |
380 | sizeof md->tunnel / 4); | |
381 | } | |
382 | miniflow_push_uint32_check(mf, skb_priority, md->skb_priority); | |
383 | miniflow_push_uint32_check(mf, pkt_mark, md->pkt_mark); | |
384 | miniflow_push_uint32_check(mf, recirc_id, md->recirc_id); | |
385 | miniflow_push_uint32(mf, in_port, odp_to_u32(md->in_port.odp_port)); | |
296e07ac | 386 | } |
064af421 | 387 | |
419681da JR |
388 | /* Initialize packet's layer pointer and offsets. */ |
389 | l2 = data; | |
390 | ofpbuf_set_frame(packet, data); | |
064af421 | 391 | |
419681da JR |
392 | /* Must have full Ethernet header to proceed. */ |
393 | if (OVS_UNLIKELY(size < sizeof(struct eth_header))) { | |
394 | goto out; | |
395 | } else { | |
396 | ovs_be16 vlan_tci; | |
50f06e16 | 397 | |
419681da JR |
398 | /* Link layer. */ |
399 | BUILD_ASSERT(offsetof(struct flow, dl_dst) + 6 | |
400 | == offsetof(struct flow, dl_src)); | |
401 | miniflow_push_words(mf, dl_dst, data, ETH_ADDR_LEN * 2 / 4); | |
402 | /* dl_type, vlan_tci. */ | |
403 | vlan_tci = parse_vlan(&data, &size); | |
404 | dl_type = parse_ethertype(&data, &size); | |
405 | miniflow_push_be16(mf, dl_type, dl_type); | |
406 | miniflow_push_be16(mf, vlan_tci, vlan_tci); | |
50f06e16 | 407 | } |
50f06e16 | 408 | |
419681da JR |
409 | /* Parse mpls. */ |
410 | if (OVS_UNLIKELY(eth_type_mpls(dl_type))) { | |
411 | int count; | |
412 | const void *mpls = data; | |
413 | ||
414 | packet->l2_5_ofs = (char *)data - l2; | |
415 | count = parse_mpls(&data, &size); | |
416 | miniflow_push_words(mf, mpls_lse, mpls, count); | |
b02475c5 SH |
417 | } |
418 | ||
ad128cc1 | 419 | /* Network layer. */ |
419681da JR |
420 | packet->l3_ofs = (char *)data - l2; |
421 | ||
422 | nw_frag = 0; | |
423 | if (OVS_LIKELY(dl_type == htons(ETH_TYPE_IP))) { | |
424 | const struct ip_header *nh = data; | |
425 | int ip_len; | |
fa8d9001 | 426 | uint16_t tot_len; |
419681da JR |
427 | |
428 | if (OVS_UNLIKELY(size < IP_HEADER_LEN)) { | |
429 | goto out; | |
430 | } | |
431 | ip_len = IP_IHL(nh->ip_ihl_ver) * 4; | |
432 | ||
433 | if (OVS_UNLIKELY(ip_len < IP_HEADER_LEN)) { | |
434 | goto out; | |
435 | } | |
fa8d9001 JR |
436 | if (OVS_UNLIKELY(size < ip_len)) { |
437 | goto out; | |
438 | } | |
439 | tot_len = ntohs(nh->ip_tot_len); | |
440 | if (OVS_UNLIKELY(tot_len > size)) { | |
441 | goto out; | |
442 | } | |
443 | if (OVS_UNLIKELY(size - tot_len > UINT8_MAX)) { | |
444 | goto out; | |
445 | } | |
446 | ofpbuf_set_l2_pad_size(packet, size - tot_len); | |
447 | size = tot_len; /* Never pull padding. */ | |
419681da JR |
448 | |
449 | /* Push both source and destination address at once. */ | |
450 | miniflow_push_words(mf, nw_src, &nh->ip_src, 2); | |
451 | ||
452 | nw_tos = nh->ip_tos; | |
453 | nw_ttl = nh->ip_ttl; | |
454 | nw_proto = nh->ip_proto; | |
455 | if (OVS_UNLIKELY(IP_IS_FRAGMENT(nh->ip_frag_off))) { | |
456 | nw_frag = FLOW_NW_FRAG_ANY; | |
457 | if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) { | |
458 | nw_frag |= FLOW_NW_FRAG_LATER; | |
459 | } | |
460 | } | |
419681da | 461 | data_pull(&data, &size, ip_len); |
419681da JR |
462 | } else if (dl_type == htons(ETH_TYPE_IPV6)) { |
463 | const struct ovs_16aligned_ip6_hdr *nh; | |
464 | ovs_be32 tc_flow; | |
fa8d9001 | 465 | uint16_t plen; |
419681da JR |
466 | |
467 | if (OVS_UNLIKELY(size < sizeof *nh)) { | |
468 | goto out; | |
469 | } | |
470 | nh = data_pull(&data, &size, sizeof *nh); | |
471 | ||
fa8d9001 JR |
472 | plen = ntohs(nh->ip6_plen); |
473 | if (OVS_UNLIKELY(plen > size)) { | |
474 | goto out; | |
475 | } | |
476 | /* Jumbo Payload option not supported yet. */ | |
477 | if (OVS_UNLIKELY(size - plen > UINT8_MAX)) { | |
478 | goto out; | |
479 | } | |
480 | ofpbuf_set_l2_pad_size(packet, size - plen); | |
481 | size = plen; /* Never pull padding. */ | |
482 | ||
419681da JR |
483 | miniflow_push_words(mf, ipv6_src, &nh->ip6_src, |
484 | sizeof nh->ip6_src / 4); | |
485 | miniflow_push_words(mf, ipv6_dst, &nh->ip6_dst, | |
486 | sizeof nh->ip6_dst / 4); | |
487 | ||
488 | tc_flow = get_16aligned_be32(&nh->ip6_flow); | |
489 | { | |
490 | ovs_be32 label = tc_flow & htonl(IPV6_LABEL_MASK); | |
491 | miniflow_push_be32_check(mf, ipv6_label, label); | |
492 | } | |
493 | ||
494 | nw_tos = ntohl(tc_flow) >> 20; | |
495 | nw_ttl = nh->ip6_hlim; | |
496 | nw_proto = nh->ip6_nxt; | |
497 | ||
498 | while (1) { | |
499 | if (OVS_LIKELY((nw_proto != IPPROTO_HOPOPTS) | |
500 | && (nw_proto != IPPROTO_ROUTING) | |
501 | && (nw_proto != IPPROTO_DSTOPTS) | |
502 | && (nw_proto != IPPROTO_AH) | |
503 | && (nw_proto != IPPROTO_FRAGMENT))) { | |
504 | /* It's either a terminal header (e.g., TCP, UDP) or one we | |
505 | * don't understand. In either case, we're done with the | |
506 | * packet, so use it to fill in 'nw_proto'. */ | |
507 | break; | |
508 | } | |
509 | ||
510 | /* We only verify that at least 8 bytes of the next header are | |
511 | * available, but many of these headers are longer. Ensure that | |
512 | * accesses within the extension header are within those first 8 | |
513 | * bytes. All extension headers are required to be at least 8 | |
514 | * bytes. */ | |
515 | if (OVS_UNLIKELY(size < 8)) { | |
516 | goto out; | |
7257b535 | 517 | } |
419681da JR |
518 | |
519 | if ((nw_proto == IPPROTO_HOPOPTS) | |
520 | || (nw_proto == IPPROTO_ROUTING) | |
521 | || (nw_proto == IPPROTO_DSTOPTS)) { | |
522 | /* These headers, while different, have the fields we care | |
523 | * about in the same location and with the same | |
524 | * interpretation. */ | |
525 | const struct ip6_ext *ext_hdr = data; | |
526 | nw_proto = ext_hdr->ip6e_nxt; | |
527 | if (OVS_UNLIKELY(!data_try_pull(&data, &size, | |
528 | (ext_hdr->ip6e_len + 1) * 8))) { | |
529 | goto out; | |
530 | } | |
531 | } else if (nw_proto == IPPROTO_AH) { | |
532 | /* A standard AH definition isn't available, but the fields | |
533 | * we care about are in the same location as the generic | |
534 | * option header--only the header length is calculated | |
535 | * differently. */ | |
536 | const struct ip6_ext *ext_hdr = data; | |
537 | nw_proto = ext_hdr->ip6e_nxt; | |
538 | if (OVS_UNLIKELY(!data_try_pull(&data, &size, | |
539 | (ext_hdr->ip6e_len + 2) * 4))) { | |
540 | goto out; | |
541 | } | |
542 | } else if (nw_proto == IPPROTO_FRAGMENT) { | |
543 | const struct ovs_16aligned_ip6_frag *frag_hdr = data; | |
544 | ||
545 | nw_proto = frag_hdr->ip6f_nxt; | |
546 | if (!data_try_pull(&data, &size, sizeof *frag_hdr)) { | |
547 | goto out; | |
548 | } | |
549 | ||
550 | /* We only process the first fragment. */ | |
551 | if (frag_hdr->ip6f_offlg != htons(0)) { | |
552 | nw_frag = FLOW_NW_FRAG_ANY; | |
553 | if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) { | |
554 | nw_frag |= FLOW_NW_FRAG_LATER; | |
555 | nw_proto = IPPROTO_FRAGMENT; | |
556 | break; | |
064af421 | 557 | } |
064af421 | 558 | } |
50f06e16 BP |
559 | } |
560 | } | |
419681da JR |
561 | } else { |
562 | if (dl_type == htons(ETH_TYPE_ARP) || | |
563 | dl_type == htons(ETH_TYPE_RARP)) { | |
564 | uint8_t arp_buf[2][ETH_ADDR_LEN]; | |
565 | const struct arp_eth_header *arp = (const struct arp_eth_header *) | |
566 | data_try_pull(&data, &size, ARP_ETH_HEADER_LEN); | |
567 | ||
568 | if (OVS_LIKELY(arp) && OVS_LIKELY(arp->ar_hrd == htons(1)) | |
569 | && OVS_LIKELY(arp->ar_pro == htons(ETH_TYPE_IP)) | |
570 | && OVS_LIKELY(arp->ar_hln == ETH_ADDR_LEN) | |
571 | && OVS_LIKELY(arp->ar_pln == 4)) { | |
572 | miniflow_push_words(mf, nw_src, &arp->ar_spa, 1); | |
573 | miniflow_push_words(mf, nw_dst, &arp->ar_tpa, 1); | |
574 | ||
575 | /* We only match on the lower 8 bits of the opcode. */ | |
576 | if (OVS_LIKELY(ntohs(arp->ar_op) <= 0xff)) { | |
577 | miniflow_push_be32(mf, nw_frag, htonl(ntohs(arp->ar_op))); | |
578 | } | |
d31f1109 | 579 | |
419681da JR |
580 | /* Must be adjacent. */ |
581 | BUILD_ASSERT(offsetof(struct flow, arp_sha) + 6 | |
582 | == offsetof(struct flow, arp_tha)); | |
583 | ||
584 | memcpy(arp_buf[0], arp->ar_sha, ETH_ADDR_LEN); | |
585 | memcpy(arp_buf[1], arp->ar_tha, ETH_ADDR_LEN); | |
586 | miniflow_push_words(mf, arp_sha, arp_buf, | |
587 | ETH_ADDR_LEN * 2 / 4); | |
588 | } | |
d31f1109 | 589 | } |
419681da JR |
590 | goto out; |
591 | } | |
592 | ||
593 | packet->l4_ofs = (char *)data - l2; | |
594 | miniflow_push_be32(mf, nw_frag, | |
595 | BYTES_TO_BE32(nw_frag, nw_tos, nw_ttl, nw_proto)); | |
596 | ||
597 | if (OVS_LIKELY(!(nw_frag & FLOW_NW_FRAG_LATER))) { | |
598 | if (OVS_LIKELY(nw_proto == IPPROTO_TCP)) { | |
599 | if (OVS_LIKELY(size >= TCP_HEADER_LEN)) { | |
600 | const struct tcp_header *tcp = data; | |
601 | ||
602 | miniflow_push_be32(mf, tcp_flags, | |
603 | TCP_FLAGS_BE32(tcp->tcp_ctl)); | |
604 | miniflow_push_words(mf, tp_src, &tcp->tcp_src, 1); | |
605 | } | |
606 | } else if (OVS_LIKELY(nw_proto == IPPROTO_UDP)) { | |
607 | if (OVS_LIKELY(size >= UDP_HEADER_LEN)) { | |
608 | const struct udp_header *udp = data; | |
609 | ||
610 | miniflow_push_words(mf, tp_src, &udp->udp_src, 1); | |
064af421 | 611 | } |
419681da JR |
612 | } else if (OVS_LIKELY(nw_proto == IPPROTO_SCTP)) { |
613 | if (OVS_LIKELY(size >= SCTP_HEADER_LEN)) { | |
614 | const struct sctp_header *sctp = data; | |
a26ef517 | 615 | |
419681da JR |
616 | miniflow_push_words(mf, tp_src, &sctp->sctp_src, 1); |
617 | } | |
618 | } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMP)) { | |
619 | if (OVS_LIKELY(size >= ICMP_HEADER_LEN)) { | |
620 | const struct icmp_header *icmp = data; | |
621 | ||
622 | miniflow_push_be16(mf, tp_src, htons(icmp->icmp_type)); | |
623 | miniflow_push_be16(mf, tp_dst, htons(icmp->icmp_code)); | |
624 | } | |
0e612675 FL |
625 | } else if (OVS_LIKELY(nw_proto == IPPROTO_IGMP)) { |
626 | if (OVS_LIKELY(size >= IGMP_HEADER_LEN)) { | |
627 | const struct igmp_header *igmp = data; | |
628 | ||
629 | miniflow_push_be16(mf, tp_src, htons(igmp->igmp_type)); | |
630 | miniflow_push_be16(mf, tp_dst, htons(igmp->igmp_code)); | |
631 | miniflow_push_be32(mf, igmp_group_ip4, | |
632 | get_16aligned_be32(&igmp->group)); | |
633 | } | |
419681da JR |
634 | } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMPV6)) { |
635 | if (OVS_LIKELY(size >= sizeof(struct icmp6_hdr))) { | |
636 | const struct in6_addr *nd_target = NULL; | |
637 | uint8_t arp_buf[2][ETH_ADDR_LEN]; | |
638 | const struct icmp6_hdr *icmp = data_pull(&data, &size, | |
639 | sizeof *icmp); | |
640 | memset(arp_buf, 0, sizeof arp_buf); | |
641 | if (OVS_LIKELY(parse_icmpv6(&data, &size, icmp, &nd_target, | |
642 | arp_buf))) { | |
b0e2ec32 JR |
643 | miniflow_push_words(mf, arp_sha, arp_buf, |
644 | ETH_ADDR_LEN * 2 / 4); | |
419681da JR |
645 | if (nd_target) { |
646 | miniflow_push_words(mf, nd_target, nd_target, | |
647 | sizeof *nd_target / 4); | |
648 | } | |
419681da JR |
649 | miniflow_push_be16(mf, tp_src, htons(icmp->icmp6_type)); |
650 | miniflow_push_be16(mf, tp_dst, htons(icmp->icmp6_code)); | |
651 | } | |
652 | } | |
064af421 BP |
653 | } |
654 | } | |
419681da JR |
655 | if (md) { |
656 | miniflow_push_uint32_check(mf, dp_hash, md->dp_hash); | |
657 | } | |
658 | out: | |
659 | dst->map = mf.map; | |
064af421 BP |
660 | } |
661 | ||
993410fb BP |
662 | /* For every bit of a field that is wildcarded in 'wildcards', sets the |
663 | * corresponding bit in 'flow' to zero. */ | |
664 | void | |
665 | flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) | |
666 | { | |
659c2346 BP |
667 | uint32_t *flow_u32 = (uint32_t *) flow; |
668 | const uint32_t *wc_u32 = (const uint32_t *) &wildcards->masks; | |
669 | size_t i; | |
993410fb | 670 | |
659c2346 BP |
671 | for (i = 0; i < FLOW_U32S; i++) { |
672 | flow_u32[i] &= wc_u32[i]; | |
26720e24 | 673 | } |
993410fb BP |
674 | } |
675 | ||
d8d9c698 EJ |
676 | void |
677 | flow_unwildcard_tp_ports(const struct flow *flow, struct flow_wildcards *wc) | |
678 | { | |
679 | if (flow->nw_proto != IPPROTO_ICMP) { | |
680 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
681 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
682 | } else { | |
683 | wc->masks.tp_src = htons(0xff); | |
684 | wc->masks.tp_dst = htons(0xff); | |
685 | } | |
686 | } | |
687 | ||
5d6c3af0 EJ |
688 | /* Initializes 'fmd' with the metadata found in 'flow'. */ |
689 | void | |
690 | flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) | |
691 | { | |
c61f3870 | 692 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 28); |
e9358af6 | 693 | |
a79f29f2 AZ |
694 | fmd->dp_hash = flow->dp_hash; |
695 | fmd->recirc_id = flow->recirc_id; | |
296e07ac | 696 | fmd->tun_id = flow->tunnel.tun_id; |
0ad90c84 JR |
697 | fmd->tun_src = flow->tunnel.ip_src; |
698 | fmd->tun_dst = flow->tunnel.ip_dst; | |
969fc56c | 699 | fmd->metadata = flow->metadata; |
5d6c3af0 | 700 | memcpy(fmd->regs, flow->regs, sizeof fmd->regs); |
ac923e91 | 701 | fmd->pkt_mark = flow->pkt_mark; |
4e022ec0 | 702 | fmd->in_port = flow->in_port.ofp_port; |
5d6c3af0 EJ |
703 | } |
704 | ||
064af421 | 705 | char * |
ae412e7d | 706 | flow_to_string(const struct flow *flow) |
064af421 BP |
707 | { |
708 | struct ds ds = DS_EMPTY_INITIALIZER; | |
709 | flow_format(&ds, flow); | |
710 | return ds_cstr(&ds); | |
711 | } | |
712 | ||
4fe3445a PS |
713 | const char * |
714 | flow_tun_flag_to_string(uint32_t flags) | |
715 | { | |
716 | switch (flags) { | |
717 | case FLOW_TNL_F_DONT_FRAGMENT: | |
718 | return "df"; | |
719 | case FLOW_TNL_F_CSUM: | |
720 | return "csum"; | |
721 | case FLOW_TNL_F_KEY: | |
722 | return "key"; | |
94872594 JG |
723 | case FLOW_TNL_F_OAM: |
724 | return "oam"; | |
4fe3445a PS |
725 | default: |
726 | return NULL; | |
727 | } | |
728 | } | |
729 | ||
730 | void | |
731 | format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), | |
732 | uint32_t flags, char del) | |
733 | { | |
734 | uint32_t bad = 0; | |
735 | ||
736 | if (!flags) { | |
737 | return; | |
738 | } | |
739 | while (flags) { | |
740 | uint32_t bit = rightmost_1bit(flags); | |
741 | const char *s; | |
742 | ||
743 | s = bit_to_string(bit); | |
744 | if (s) { | |
745 | ds_put_format(ds, "%s%c", s, del); | |
746 | } else { | |
747 | bad |= bit; | |
748 | } | |
749 | ||
750 | flags &= ~bit; | |
751 | } | |
752 | ||
753 | if (bad) { | |
754 | ds_put_format(ds, "0x%"PRIx32"%c", bad, del); | |
755 | } | |
756 | ds_chomp(ds, del); | |
757 | } | |
758 | ||
61bf6666 JR |
759 | void |
760 | format_flags_masked(struct ds *ds, const char *name, | |
761 | const char *(*bit_to_string)(uint32_t), uint32_t flags, | |
762 | uint32_t mask) | |
763 | { | |
764 | if (name) { | |
765 | ds_put_format(ds, "%s=", name); | |
766 | } | |
767 | while (mask) { | |
768 | uint32_t bit = rightmost_1bit(mask); | |
769 | const char *s = bit_to_string(bit); | |
770 | ||
771 | ds_put_format(ds, "%s%s", (flags & bit) ? "+" : "-", | |
772 | s ? s : "[Unknown]"); | |
773 | mask &= ~bit; | |
774 | } | |
775 | } | |
776 | ||
064af421 | 777 | void |
ae412e7d | 778 | flow_format(struct ds *ds, const struct flow *flow) |
064af421 | 779 | { |
aa6c9932 | 780 | struct match match; |
78c9486d | 781 | struct flow_wildcards *wc = &match.wc; |
296e07ac | 782 | |
aa6c9932 | 783 | match_wc_init(&match, flow); |
78c9486d JR |
784 | |
785 | /* As this function is most often used for formatting a packet in a | |
786 | * packet-in message, skip formatting the packet context fields that are | |
787 | * all-zeroes (Openflow spec encourages leaving out all-zeroes context | |
788 | * fields from the packet-in messages). We make an exception with the | |
789 | * 'in_port' field, which we always format, as packets usually have an | |
790 | * in_port, and 0 is a port just like any other port. */ | |
791 | if (!flow->skb_priority) { | |
792 | WC_UNMASK_FIELD(wc, skb_priority); | |
793 | } | |
794 | if (!flow->pkt_mark) { | |
795 | WC_UNMASK_FIELD(wc, pkt_mark); | |
796 | } | |
797 | if (!flow->recirc_id) { | |
798 | WC_UNMASK_FIELD(wc, recirc_id); | |
799 | } | |
800 | for (int i = 0; i < FLOW_N_REGS; i++) { | |
801 | if (!flow->regs[i]) { | |
802 | WC_UNMASK_FIELD(wc, regs[i]); | |
803 | } | |
804 | } | |
805 | if (!flow->metadata) { | |
806 | WC_UNMASK_FIELD(wc, metadata); | |
807 | } | |
808 | ||
3f78c3cc | 809 | match_format(&match, ds, OFP_DEFAULT_PRIORITY); |
064af421 BP |
810 | } |
811 | ||
812 | void | |
ae412e7d | 813 | flow_print(FILE *stream, const struct flow *flow) |
064af421 BP |
814 | { |
815 | char *s = flow_to_string(flow); | |
816 | fputs(s, stream); | |
817 | free(s); | |
818 | } | |
54363004 BP |
819 | \f |
820 | /* flow_wildcards functions. */ | |
821 | ||
d8ae4d67 | 822 | /* Initializes 'wc' as a set of wildcards that matches every packet. */ |
54363004 | 823 | void |
d8ae4d67 | 824 | flow_wildcards_init_catchall(struct flow_wildcards *wc) |
54363004 | 825 | { |
659c2346 | 826 | memset(&wc->masks, 0, sizeof wc->masks); |
54363004 BP |
827 | } |
828 | ||
78c9486d JR |
829 | /* Converts a flow into flow wildcards. It sets the wildcard masks based on |
830 | * the packet headers extracted to 'flow'. It will not set the mask for fields | |
831 | * that do not make sense for the packet type. OpenFlow-only metadata is | |
832 | * wildcarded, but other metadata is unconditionally exact-matched. */ | |
833 | void flow_wildcards_init_for_packet(struct flow_wildcards *wc, | |
834 | const struct flow *flow) | |
835 | { | |
836 | memset(&wc->masks, 0x0, sizeof wc->masks); | |
837 | ||
0de8783a | 838 | /* Update this function whenever struct flow changes. */ |
c61f3870 | 839 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 28); |
0de8783a | 840 | |
78c9486d JR |
841 | if (flow->tunnel.ip_dst) { |
842 | if (flow->tunnel.flags & FLOW_TNL_F_KEY) { | |
843 | WC_MASK_FIELD(wc, tunnel.tun_id); | |
844 | } | |
845 | WC_MASK_FIELD(wc, tunnel.ip_src); | |
846 | WC_MASK_FIELD(wc, tunnel.ip_dst); | |
847 | WC_MASK_FIELD(wc, tunnel.flags); | |
848 | WC_MASK_FIELD(wc, tunnel.ip_tos); | |
849 | WC_MASK_FIELD(wc, tunnel.ip_ttl); | |
850 | WC_MASK_FIELD(wc, tunnel.tp_src); | |
851 | WC_MASK_FIELD(wc, tunnel.tp_dst); | |
852 | } else if (flow->tunnel.tun_id) { | |
853 | WC_MASK_FIELD(wc, tunnel.tun_id); | |
854 | } | |
855 | ||
856 | /* metadata and regs wildcarded. */ | |
857 | ||
858 | WC_MASK_FIELD(wc, skb_priority); | |
859 | WC_MASK_FIELD(wc, pkt_mark); | |
860 | WC_MASK_FIELD(wc, recirc_id); | |
861 | WC_MASK_FIELD(wc, dp_hash); | |
862 | WC_MASK_FIELD(wc, in_port); | |
863 | ||
c61f3870 BP |
864 | /* actset_output wildcarded. */ |
865 | ||
78c9486d JR |
866 | WC_MASK_FIELD(wc, dl_dst); |
867 | WC_MASK_FIELD(wc, dl_src); | |
868 | WC_MASK_FIELD(wc, dl_type); | |
869 | WC_MASK_FIELD(wc, vlan_tci); | |
870 | ||
871 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
872 | WC_MASK_FIELD(wc, nw_src); | |
873 | WC_MASK_FIELD(wc, nw_dst); | |
874 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
875 | WC_MASK_FIELD(wc, ipv6_src); | |
876 | WC_MASK_FIELD(wc, ipv6_dst); | |
877 | WC_MASK_FIELD(wc, ipv6_label); | |
878 | } else if (flow->dl_type == htons(ETH_TYPE_ARP) || | |
879 | flow->dl_type == htons(ETH_TYPE_RARP)) { | |
880 | WC_MASK_FIELD(wc, nw_src); | |
881 | WC_MASK_FIELD(wc, nw_dst); | |
882 | WC_MASK_FIELD(wc, nw_proto); | |
883 | WC_MASK_FIELD(wc, arp_sha); | |
884 | WC_MASK_FIELD(wc, arp_tha); | |
885 | return; | |
886 | } else if (eth_type_mpls(flow->dl_type)) { | |
887 | for (int i = 0; i < FLOW_MAX_MPLS_LABELS; i++) { | |
888 | WC_MASK_FIELD(wc, mpls_lse[i]); | |
889 | if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { | |
890 | break; | |
891 | } | |
892 | } | |
893 | return; | |
894 | } else { | |
895 | return; /* Unknown ethertype. */ | |
896 | } | |
897 | ||
898 | /* IPv4 or IPv6. */ | |
899 | WC_MASK_FIELD(wc, nw_frag); | |
900 | WC_MASK_FIELD(wc, nw_tos); | |
901 | WC_MASK_FIELD(wc, nw_ttl); | |
902 | WC_MASK_FIELD(wc, nw_proto); | |
903 | ||
904 | /* No transport layer header in later fragments. */ | |
905 | if (!(flow->nw_frag & FLOW_NW_FRAG_LATER) && | |
906 | (flow->nw_proto == IPPROTO_ICMP || | |
907 | flow->nw_proto == IPPROTO_ICMPV6 || | |
908 | flow->nw_proto == IPPROTO_TCP || | |
909 | flow->nw_proto == IPPROTO_UDP || | |
910 | flow->nw_proto == IPPROTO_SCTP || | |
911 | flow->nw_proto == IPPROTO_IGMP)) { | |
912 | WC_MASK_FIELD(wc, tp_src); | |
913 | WC_MASK_FIELD(wc, tp_dst); | |
914 | ||
915 | if (flow->nw_proto == IPPROTO_TCP) { | |
916 | WC_MASK_FIELD(wc, tcp_flags); | |
917 | } else if (flow->nw_proto == IPPROTO_ICMPV6) { | |
918 | WC_MASK_FIELD(wc, arp_sha); | |
919 | WC_MASK_FIELD(wc, arp_tha); | |
920 | WC_MASK_FIELD(wc, nd_target); | |
921 | } else if (flow->nw_proto == IPPROTO_IGMP) { | |
922 | WC_MASK_FIELD(wc, igmp_group_ip4); | |
923 | } | |
924 | } | |
925 | } | |
926 | ||
0de8783a JR |
927 | /* Return a map of possible fields for a packet of the same type as 'flow'. |
928 | * Including extra bits in the returned mask is not wrong, it is just less | |
929 | * optimal. | |
930 | * | |
931 | * This is a less precise version of flow_wildcards_init_for_packet() above. */ | |
932 | uint64_t | |
933 | flow_wc_map(const struct flow *flow) | |
934 | { | |
935 | /* Update this function whenever struct flow changes. */ | |
c61f3870 | 936 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 28); |
0de8783a JR |
937 | |
938 | uint64_t map = (flow->tunnel.ip_dst) ? MINIFLOW_MAP(tunnel) : 0; | |
939 | ||
940 | /* Metadata fields that can appear on packet input. */ | |
941 | map |= MINIFLOW_MAP(skb_priority) | MINIFLOW_MAP(pkt_mark) | |
942 | | MINIFLOW_MAP(recirc_id) | MINIFLOW_MAP(dp_hash) | |
943 | | MINIFLOW_MAP(in_port) | |
944 | | MINIFLOW_MAP(dl_dst) | MINIFLOW_MAP(dl_src) | |
945 | | MINIFLOW_MAP(dl_type) | MINIFLOW_MAP(vlan_tci); | |
946 | ||
947 | /* Ethertype-dependent fields. */ | |
948 | if (OVS_LIKELY(flow->dl_type == htons(ETH_TYPE_IP))) { | |
949 | map |= MINIFLOW_MAP(nw_src) | MINIFLOW_MAP(nw_dst) | |
950 | | MINIFLOW_MAP(nw_proto) | MINIFLOW_MAP(nw_frag) | |
951 | | MINIFLOW_MAP(nw_tos) | MINIFLOW_MAP(nw_ttl); | |
952 | if (OVS_UNLIKELY(flow->nw_proto == IPPROTO_IGMP)) { | |
953 | map |= MINIFLOW_MAP(igmp_group_ip4); | |
954 | } else { | |
955 | map |= MINIFLOW_MAP(tcp_flags) | |
956 | | MINIFLOW_MAP(tp_src) | MINIFLOW_MAP(tp_dst); | |
957 | } | |
958 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
959 | map |= MINIFLOW_MAP(ipv6_src) | MINIFLOW_MAP(ipv6_dst) | |
960 | | MINIFLOW_MAP(ipv6_label) | |
961 | | MINIFLOW_MAP(nw_proto) | MINIFLOW_MAP(nw_frag) | |
962 | | MINIFLOW_MAP(nw_tos) | MINIFLOW_MAP(nw_ttl); | |
963 | if (OVS_UNLIKELY(flow->nw_proto == IPPROTO_ICMPV6)) { | |
964 | map |= MINIFLOW_MAP(nd_target) | |
965 | | MINIFLOW_MAP(arp_sha) | MINIFLOW_MAP(arp_tha); | |
966 | } else { | |
967 | map |= MINIFLOW_MAP(tcp_flags) | |
968 | | MINIFLOW_MAP(tp_src) | MINIFLOW_MAP(tp_dst); | |
969 | } | |
970 | } else if (eth_type_mpls(flow->dl_type)) { | |
971 | map |= MINIFLOW_MAP(mpls_lse); | |
972 | } else if (flow->dl_type == htons(ETH_TYPE_ARP) || | |
973 | flow->dl_type == htons(ETH_TYPE_RARP)) { | |
974 | map |= MINIFLOW_MAP(nw_src) | MINIFLOW_MAP(nw_dst) | |
975 | | MINIFLOW_MAP(nw_proto) | |
976 | | MINIFLOW_MAP(arp_sha) | MINIFLOW_MAP(arp_tha); | |
977 | } | |
978 | ||
979 | return map; | |
980 | } | |
981 | ||
c11c6faa AZ |
982 | /* Clear the metadata and register wildcard masks. They are not packet |
983 | * header fields. */ | |
984 | void | |
985 | flow_wildcards_clear_non_packet_fields(struct flow_wildcards *wc) | |
986 | { | |
0de8783a | 987 | /* Update this function whenever struct flow changes. */ |
c61f3870 | 988 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 28); |
0de8783a | 989 | |
c11c6faa AZ |
990 | memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata); |
991 | memset(&wc->masks.regs, 0, sizeof wc->masks.regs); | |
c61f3870 | 992 | wc->masks.actset_output = 0; |
c11c6faa AZ |
993 | } |
994 | ||
ecf1e7ac BP |
995 | /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or |
996 | * fields. */ | |
997 | bool | |
998 | flow_wildcards_is_catchall(const struct flow_wildcards *wc) | |
999 | { | |
659c2346 BP |
1000 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; |
1001 | size_t i; | |
ecf1e7ac | 1002 | |
659c2346 BP |
1003 | for (i = 0; i < FLOW_U32S; i++) { |
1004 | if (wc_u32[i]) { | |
ecf1e7ac BP |
1005 | return false; |
1006 | } | |
1007 | } | |
ecf1e7ac BP |
1008 | return true; |
1009 | } | |
1010 | ||
368eefac EJ |
1011 | /* Sets 'dst' as the bitwise AND of wildcards in 'src1' and 'src2'. |
1012 | * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded | |
1013 | * in 'src1' or 'src2' or both. */ | |
b5d97350 | 1014 | void |
368eefac EJ |
1015 | flow_wildcards_and(struct flow_wildcards *dst, |
1016 | const struct flow_wildcards *src1, | |
1017 | const struct flow_wildcards *src2) | |
b5d97350 | 1018 | { |
659c2346 BP |
1019 | uint32_t *dst_u32 = (uint32_t *) &dst->masks; |
1020 | const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; | |
1021 | const uint32_t *src2_u32 = (const uint32_t *) &src2->masks; | |
1022 | size_t i; | |
a79c50f3 | 1023 | |
659c2346 BP |
1024 | for (i = 0; i < FLOW_U32S; i++) { |
1025 | dst_u32[i] = src1_u32[i] & src2_u32[i]; | |
26720e24 | 1026 | } |
b5d97350 BP |
1027 | } |
1028 | ||
368eefac EJ |
1029 | /* Sets 'dst' as the bitwise OR of wildcards in 'src1' and 'src2'. That |
1030 | * is, a bit or a field is wildcarded in 'dst' if it is neither | |
1031 | * wildcarded in 'src1' nor 'src2'. */ | |
1032 | void | |
1033 | flow_wildcards_or(struct flow_wildcards *dst, | |
1034 | const struct flow_wildcards *src1, | |
1035 | const struct flow_wildcards *src2) | |
1036 | { | |
1037 | uint32_t *dst_u32 = (uint32_t *) &dst->masks; | |
1038 | const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; | |
1039 | const uint32_t *src2_u32 = (const uint32_t *) &src2->masks; | |
1040 | size_t i; | |
1041 | ||
1042 | for (i = 0; i < FLOW_U32S; i++) { | |
1043 | dst_u32[i] = src1_u32[i] | src2_u32[i]; | |
1044 | } | |
1045 | } | |
1046 | ||
b5d97350 BP |
1047 | /* Returns a hash of the wildcards in 'wc'. */ |
1048 | uint32_t | |
1006cda6 | 1049 | flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis) |
b5d97350 | 1050 | { |
ac31c5af | 1051 | return flow_hash(&wc->masks, basis); |
b5d97350 BP |
1052 | } |
1053 | ||
1054 | /* Returns true if 'a' and 'b' represent the same wildcards, false if they are | |
1055 | * different. */ | |
1056 | bool | |
1057 | flow_wildcards_equal(const struct flow_wildcards *a, | |
1058 | const struct flow_wildcards *b) | |
1059 | { | |
659c2346 | 1060 | return flow_equal(&a->masks, &b->masks); |
b5d97350 BP |
1061 | } |
1062 | ||
1063 | /* Returns true if at least one bit or field is wildcarded in 'a' but not in | |
1064 | * 'b', false otherwise. */ | |
1065 | bool | |
1066 | flow_wildcards_has_extra(const struct flow_wildcards *a, | |
1067 | const struct flow_wildcards *b) | |
1068 | { | |
659c2346 BP |
1069 | const uint32_t *a_u32 = (const uint32_t *) &a->masks; |
1070 | const uint32_t *b_u32 = (const uint32_t *) &b->masks; | |
1071 | size_t i; | |
a79c50f3 | 1072 | |
659c2346 BP |
1073 | for (i = 0; i < FLOW_U32S; i++) { |
1074 | if ((a_u32[i] & b_u32[i]) != b_u32[i]) { | |
b6c9e612 BP |
1075 | return true; |
1076 | } | |
1077 | } | |
659c2346 BP |
1078 | return false; |
1079 | } | |
b6c9e612 | 1080 | |
659c2346 BP |
1081 | /* Returns true if 'a' and 'b' are equal, except that 0-bits (wildcarded bits) |
1082 | * in 'wc' do not need to be equal in 'a' and 'b'. */ | |
1083 | bool | |
1084 | flow_equal_except(const struct flow *a, const struct flow *b, | |
1085 | const struct flow_wildcards *wc) | |
1086 | { | |
1087 | const uint32_t *a_u32 = (const uint32_t *) a; | |
1088 | const uint32_t *b_u32 = (const uint32_t *) b; | |
1089 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; | |
1090 | size_t i; | |
d31f1109 | 1091 | |
659c2346 BP |
1092 | for (i = 0; i < FLOW_U32S; i++) { |
1093 | if ((a_u32[i] ^ b_u32[i]) & wc_u32[i]) { | |
1094 | return false; | |
1095 | } | |
47284b1f | 1096 | } |
659c2346 | 1097 | return true; |
b5d97350 BP |
1098 | } |
1099 | ||
b6c9e612 BP |
1100 | /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. |
1101 | * (A 0-bit indicates a wildcard bit.) */ | |
1102 | void | |
1103 | flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) | |
1104 | { | |
26720e24 | 1105 | wc->masks.regs[idx] = mask; |
b6c9e612 | 1106 | } |
ff55ea1f | 1107 | |
79fe0f46 BP |
1108 | /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. |
1109 | * (A 0-bit indicates a wildcard bit.) */ | |
1110 | void | |
1111 | flow_wildcards_set_xreg_mask(struct flow_wildcards *wc, int idx, uint64_t mask) | |
1112 | { | |
1113 | flow_set_xreg(&wc->masks, idx, mask); | |
1114 | } | |
1115 | ||
28a560d9 JR |
1116 | /* Calculates the 5-tuple hash from the given miniflow. |
1117 | * This returns the same value as flow_hash_5tuple for the corresponding | |
1118 | * flow. */ | |
4f150744 JR |
1119 | uint32_t |
1120 | miniflow_hash_5tuple(const struct miniflow *flow, uint32_t basis) | |
1121 | { | |
28a560d9 | 1122 | uint32_t hash = basis; |
4f150744 | 1123 | |
28a560d9 JR |
1124 | if (flow) { |
1125 | ovs_be16 dl_type = MINIFLOW_GET_BE16(flow, dl_type); | |
1126 | ||
33c6a1b9 | 1127 | hash = hash_add(hash, MINIFLOW_GET_U8(flow, nw_proto)); |
28a560d9 JR |
1128 | |
1129 | /* Separate loops for better optimization. */ | |
1130 | if (dl_type == htons(ETH_TYPE_IPV6)) { | |
1131 | uint64_t map = MINIFLOW_MAP(ipv6_src) | MINIFLOW_MAP(ipv6_dst) | |
1132 | | MINIFLOW_MAP(tp_src); /* Covers both ports */ | |
1133 | uint32_t value; | |
4f150744 | 1134 | |
28a560d9 | 1135 | MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { |
33c6a1b9 | 1136 | hash = hash_add(hash, value); |
28a560d9 JR |
1137 | } |
1138 | } else { | |
1139 | uint64_t map = MINIFLOW_MAP(nw_src) | MINIFLOW_MAP(nw_dst) | |
1140 | | MINIFLOW_MAP(tp_src); /* Covers both ports */ | |
1141 | uint32_t value; | |
4f150744 | 1142 | |
28a560d9 | 1143 | MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { |
33c6a1b9 | 1144 | hash = hash_add(hash, value); |
28a560d9 JR |
1145 | } |
1146 | } | |
33c6a1b9 | 1147 | hash = hash_finish(hash, 42); /* Arbitrary number. */ |
28a560d9 JR |
1148 | } |
1149 | return hash; | |
4f150744 JR |
1150 | } |
1151 | ||
1152 | BUILD_ASSERT_DECL(offsetof(struct flow, tp_src) + 2 | |
1153 | == offsetof(struct flow, tp_dst) && | |
1154 | offsetof(struct flow, tp_src) / 4 | |
1155 | == offsetof(struct flow, tp_dst) / 4); | |
28a560d9 JR |
1156 | BUILD_ASSERT_DECL(offsetof(struct flow, ipv6_src) + 16 |
1157 | == offsetof(struct flow, ipv6_dst)); | |
4f150744 | 1158 | |
63be20be AW |
1159 | /* Calculates the 5-tuple hash from the given flow. */ |
1160 | uint32_t | |
1161 | flow_hash_5tuple(const struct flow *flow, uint32_t basis) | |
1162 | { | |
28a560d9 | 1163 | uint32_t hash = basis; |
63be20be | 1164 | |
28a560d9 JR |
1165 | if (flow) { |
1166 | const uint32_t *flow_u32 = (const uint32_t *)flow; | |
1167 | ||
33c6a1b9 | 1168 | hash = hash_add(hash, flow->nw_proto); |
28a560d9 JR |
1169 | |
1170 | if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
1171 | int ofs = offsetof(struct flow, ipv6_src) / 4; | |
1172 | int end = ofs + 2 * sizeof flow->ipv6_src / 4; | |
63be20be | 1173 | |
28a560d9 | 1174 | while (ofs < end) { |
33c6a1b9 | 1175 | hash = hash_add(hash, flow_u32[ofs++]); |
28a560d9 JR |
1176 | } |
1177 | } else { | |
33c6a1b9 JR |
1178 | hash = hash_add(hash, (OVS_FORCE uint32_t) flow->nw_src); |
1179 | hash = hash_add(hash, (OVS_FORCE uint32_t) flow->nw_dst); | |
28a560d9 | 1180 | } |
33c6a1b9 | 1181 | hash = hash_add(hash, flow_u32[offsetof(struct flow, tp_src) / 4]); |
63be20be | 1182 | |
33c6a1b9 | 1183 | hash = hash_finish(hash, 42); /* Arbitrary number. */ |
28a560d9 JR |
1184 | } |
1185 | return hash; | |
63be20be AW |
1186 | } |
1187 | ||
ff55ea1f EJ |
1188 | /* Hashes 'flow' based on its L2 through L4 protocol information. */ |
1189 | uint32_t | |
1190 | flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) | |
1191 | { | |
1192 | struct { | |
d31f1109 JP |
1193 | union { |
1194 | ovs_be32 ipv4_addr; | |
1195 | struct in6_addr ipv6_addr; | |
1196 | }; | |
ff55ea1f EJ |
1197 | ovs_be16 eth_type; |
1198 | ovs_be16 vlan_tci; | |
5b909cbb | 1199 | ovs_be16 tp_port; |
ff55ea1f EJ |
1200 | uint8_t eth_addr[ETH_ADDR_LEN]; |
1201 | uint8_t ip_proto; | |
1202 | } fields; | |
1203 | ||
1204 | int i; | |
1205 | ||
1206 | memset(&fields, 0, sizeof fields); | |
1207 | for (i = 0; i < ETH_ADDR_LEN; i++) { | |
1208 | fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i]; | |
1209 | } | |
1210 | fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK); | |
1211 | fields.eth_type = flow->dl_type; | |
3e3eda95 EJ |
1212 | |
1213 | /* UDP source and destination port are not taken into account because they | |
1214 | * will not necessarily be symmetric in a bidirectional flow. */ | |
ff55ea1f | 1215 | if (fields.eth_type == htons(ETH_TYPE_IP)) { |
d31f1109 JP |
1216 | fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; |
1217 | fields.ip_proto = flow->nw_proto; | |
c6bcb685 | 1218 | if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { |
5b909cbb | 1219 | fields.tp_port = flow->tp_src ^ flow->tp_dst; |
d31f1109 JP |
1220 | } |
1221 | } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) { | |
1222 | const uint8_t *a = &flow->ipv6_src.s6_addr[0]; | |
1223 | const uint8_t *b = &flow->ipv6_dst.s6_addr[0]; | |
1224 | uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0]; | |
1225 | ||
1226 | for (i=0; i<16; i++) { | |
1227 | ipv6_addr[i] = a[i] ^ b[i]; | |
1228 | } | |
ff55ea1f | 1229 | fields.ip_proto = flow->nw_proto; |
c6bcb685 | 1230 | if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { |
5b909cbb | 1231 | fields.tp_port = flow->tp_src ^ flow->tp_dst; |
ff55ea1f | 1232 | } |
ff55ea1f | 1233 | } |
c49d1dd1 | 1234 | return jhash_bytes(&fields, sizeof fields, basis); |
ff55ea1f | 1235 | } |
520e9a2a | 1236 | |
94639963 JR |
1237 | /* Initialize a flow with random fields that matter for nx_hash_fields. */ |
1238 | void | |
1239 | flow_random_hash_fields(struct flow *flow) | |
1240 | { | |
1241 | uint16_t rnd = random_uint16(); | |
1242 | ||
1243 | /* Initialize to all zeros. */ | |
1244 | memset(flow, 0, sizeof *flow); | |
1245 | ||
1246 | eth_addr_random(flow->dl_src); | |
1247 | eth_addr_random(flow->dl_dst); | |
1248 | ||
1249 | flow->vlan_tci = (OVS_FORCE ovs_be16) (random_uint16() & VLAN_VID_MASK); | |
1250 | ||
1251 | /* Make most of the random flows IPv4, some IPv6, and rest random. */ | |
1252 | flow->dl_type = rnd < 0x8000 ? htons(ETH_TYPE_IP) : | |
1253 | rnd < 0xc000 ? htons(ETH_TYPE_IPV6) : (OVS_FORCE ovs_be16)rnd; | |
1254 | ||
1255 | if (dl_type_is_ip_any(flow->dl_type)) { | |
1256 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
1257 | flow->nw_src = (OVS_FORCE ovs_be32)random_uint32(); | |
1258 | flow->nw_dst = (OVS_FORCE ovs_be32)random_uint32(); | |
1259 | } else { | |
1260 | random_bytes(&flow->ipv6_src, sizeof flow->ipv6_src); | |
1261 | random_bytes(&flow->ipv6_dst, sizeof flow->ipv6_dst); | |
1262 | } | |
1263 | /* Make most of IP flows TCP, some UDP or SCTP, and rest random. */ | |
1264 | rnd = random_uint16(); | |
1265 | flow->nw_proto = rnd < 0x8000 ? IPPROTO_TCP : | |
1266 | rnd < 0xc000 ? IPPROTO_UDP : | |
1267 | rnd < 0xd000 ? IPPROTO_SCTP : (uint8_t)rnd; | |
1268 | if (flow->nw_proto == IPPROTO_TCP || | |
1269 | flow->nw_proto == IPPROTO_UDP || | |
1270 | flow->nw_proto == IPPROTO_SCTP) { | |
1271 | flow->tp_src = (OVS_FORCE ovs_be16)random_uint16(); | |
1272 | flow->tp_dst = (OVS_FORCE ovs_be16)random_uint16(); | |
1273 | } | |
1274 | } | |
1275 | } | |
1276 | ||
bcd2633a JP |
1277 | /* Masks the fields in 'wc' that are used by the flow hash 'fields'. */ |
1278 | void | |
6cdd5145 JP |
1279 | flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, |
1280 | enum nx_hash_fields fields) | |
bcd2633a JP |
1281 | { |
1282 | switch (fields) { | |
1283 | case NX_HASH_FIELDS_ETH_SRC: | |
1284 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); | |
1285 | break; | |
1286 | ||
1287 | case NX_HASH_FIELDS_SYMMETRIC_L4: | |
1288 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); | |
1289 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
6cdd5145 JP |
1290 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
1291 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1292 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
7f8a65ca | 1293 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { |
6cdd5145 JP |
1294 | memset(&wc->masks.ipv6_src, 0xff, sizeof wc->masks.ipv6_src); |
1295 | memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); | |
1296 | } | |
1297 | if (is_ip_any(flow)) { | |
1298 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
d8d9c698 | 1299 | flow_unwildcard_tp_ports(flow, wc); |
6cdd5145 | 1300 | } |
1dd35f8a | 1301 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
bcd2633a JP |
1302 | break; |
1303 | ||
1304 | default: | |
428b2edd | 1305 | OVS_NOT_REACHED(); |
bcd2633a JP |
1306 | } |
1307 | } | |
1308 | ||
520e9a2a EJ |
1309 | /* Hashes the portions of 'flow' designated by 'fields'. */ |
1310 | uint32_t | |
1311 | flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, | |
1312 | uint16_t basis) | |
1313 | { | |
1314 | switch (fields) { | |
1315 | ||
1316 | case NX_HASH_FIELDS_ETH_SRC: | |
c49d1dd1 | 1317 | return jhash_bytes(flow->dl_src, sizeof flow->dl_src, basis); |
520e9a2a EJ |
1318 | |
1319 | case NX_HASH_FIELDS_SYMMETRIC_L4: | |
1320 | return flow_hash_symmetric_l4(flow, basis); | |
1321 | } | |
1322 | ||
428b2edd | 1323 | OVS_NOT_REACHED(); |
520e9a2a EJ |
1324 | } |
1325 | ||
1326 | /* Returns a string representation of 'fields'. */ | |
1327 | const char * | |
1328 | flow_hash_fields_to_str(enum nx_hash_fields fields) | |
1329 | { | |
1330 | switch (fields) { | |
1331 | case NX_HASH_FIELDS_ETH_SRC: return "eth_src"; | |
1332 | case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4"; | |
1333 | default: return "<unknown>"; | |
1334 | } | |
1335 | } | |
1336 | ||
1337 | /* Returns true if the value of 'fields' is supported. Otherwise false. */ | |
1338 | bool | |
1339 | flow_hash_fields_valid(enum nx_hash_fields fields) | |
1340 | { | |
1341 | return fields == NX_HASH_FIELDS_ETH_SRC | |
1342 | || fields == NX_HASH_FIELDS_SYMMETRIC_L4; | |
1343 | } | |
8b3b8dd1 | 1344 | |
368eefac EJ |
1345 | /* Returns a hash value for the bits of 'flow' that are active based on |
1346 | * 'wc', given 'basis'. */ | |
1347 | uint32_t | |
1348 | flow_hash_in_wildcards(const struct flow *flow, | |
1349 | const struct flow_wildcards *wc, uint32_t basis) | |
1350 | { | |
1351 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; | |
1352 | const uint32_t *flow_u32 = (const uint32_t *) flow; | |
1353 | uint32_t hash; | |
1354 | size_t i; | |
1355 | ||
1356 | hash = basis; | |
1357 | for (i = 0; i < FLOW_U32S; i++) { | |
33c6a1b9 | 1358 | hash = hash_add(hash, flow_u32[i] & wc_u32[i]); |
368eefac | 1359 | } |
33c6a1b9 | 1360 | return hash_finish(hash, 4 * FLOW_U32S); |
368eefac EJ |
1361 | } |
1362 | ||
3719455c BP |
1363 | /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an |
1364 | * OpenFlow 1.0 "dl_vlan" value: | |
1365 | * | |
1366 | * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match | |
1367 | * that VLAN. Any existing PCP match is unchanged (it becomes 0 if | |
1368 | * 'flow' previously matched packets without a VLAN header). | |
1369 | * | |
1370 | * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet | |
1371 | * without a VLAN tag. | |
1372 | * | |
1373 | * - Other values of 'vid' should not be used. */ | |
1374 | void | |
fb0451d9 | 1375 | flow_set_dl_vlan(struct flow *flow, ovs_be16 vid) |
3719455c | 1376 | { |
0c436519 | 1377 | if (vid == htons(OFP10_VLAN_NONE)) { |
3719455c BP |
1378 | flow->vlan_tci = htons(0); |
1379 | } else { | |
1380 | vid &= htons(VLAN_VID_MASK); | |
1381 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
1382 | flow->vlan_tci |= htons(VLAN_CFI) | vid; | |
1383 | } | |
1384 | } | |
1385 | ||
cc34bc8c BP |
1386 | /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an |
1387 | * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID | |
1388 | * plus CFI). */ | |
1389 | void | |
1390 | flow_set_vlan_vid(struct flow *flow, ovs_be16 vid) | |
1391 | { | |
1392 | ovs_be16 mask = htons(VLAN_VID_MASK | VLAN_CFI); | |
1393 | flow->vlan_tci &= ~mask; | |
1394 | flow->vlan_tci |= vid & mask; | |
1395 | } | |
1396 | ||
3719455c BP |
1397 | /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the |
1398 | * range 0...7. | |
1399 | * | |
1400 | * This function has no effect on the VLAN ID that 'flow' matches. | |
1401 | * | |
1402 | * After calling this function, 'flow' will not match packets without a VLAN | |
1403 | * header. */ | |
1404 | void | |
1405 | flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) | |
1406 | { | |
1407 | pcp &= 0x07; | |
1408 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
1409 | flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); | |
1410 | } | |
1411 | ||
8bfd0fda BP |
1412 | /* Returns the number of MPLS LSEs present in 'flow' |
1413 | * | |
1414 | * Returns 0 if the 'dl_type' of 'flow' is not an MPLS ethernet type. | |
1415 | * Otherwise traverses 'flow''s MPLS label stack stopping at the | |
1416 | * first entry that has the BoS bit set. If no such entry exists then | |
1417 | * the maximum number of LSEs that can be stored in 'flow' is returned. | |
1418 | */ | |
1419 | int | |
1420 | flow_count_mpls_labels(const struct flow *flow, struct flow_wildcards *wc) | |
1421 | { | |
22d38fca | 1422 | /* dl_type is always masked. */ |
8bfd0fda BP |
1423 | if (eth_type_mpls(flow->dl_type)) { |
1424 | int i; | |
1425 | int len = FLOW_MAX_MPLS_LABELS; | |
1426 | ||
1427 | for (i = 0; i < len; i++) { | |
1428 | if (wc) { | |
1429 | wc->masks.mpls_lse[i] |= htonl(MPLS_BOS_MASK); | |
1430 | } | |
1431 | if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { | |
1432 | return i + 1; | |
1433 | } | |
1434 | } | |
1435 | ||
1436 | return len; | |
1437 | } else { | |
1438 | return 0; | |
1439 | } | |
1440 | } | |
1441 | ||
1442 | /* Returns the number consecutive of MPLS LSEs, starting at the | |
1443 | * innermost LSE, that are common in 'a' and 'b'. | |
1444 | * | |
1445 | * 'an' must be flow_count_mpls_labels(a). | |
1446 | * 'bn' must be flow_count_mpls_labels(b). | |
1447 | */ | |
1448 | int | |
1449 | flow_count_common_mpls_labels(const struct flow *a, int an, | |
1450 | const struct flow *b, int bn, | |
1451 | struct flow_wildcards *wc) | |
1452 | { | |
1453 | int min_n = MIN(an, bn); | |
1454 | if (min_n == 0) { | |
1455 | return 0; | |
1456 | } else { | |
1457 | int common_n = 0; | |
1458 | int a_last = an - 1; | |
1459 | int b_last = bn - 1; | |
1460 | int i; | |
1461 | ||
1462 | for (i = 0; i < min_n; i++) { | |
1463 | if (wc) { | |
1464 | wc->masks.mpls_lse[a_last - i] = OVS_BE32_MAX; | |
1465 | wc->masks.mpls_lse[b_last - i] = OVS_BE32_MAX; | |
1466 | } | |
1467 | if (a->mpls_lse[a_last - i] != b->mpls_lse[b_last - i]) { | |
1468 | break; | |
1469 | } else { | |
1470 | common_n++; | |
1471 | } | |
1472 | } | |
1473 | ||
1474 | return common_n; | |
1475 | } | |
1476 | } | |
1477 | ||
1478 | /* Adds a new outermost MPLS label to 'flow' and changes 'flow''s Ethernet type | |
1479 | * to 'mpls_eth_type', which must be an MPLS Ethertype. | |
1480 | * | |
1481 | * If the new label is the first MPLS label in 'flow', it is generated as; | |
1482 | * | |
1483 | * - label: 2, if 'flow' is IPv6, otherwise 0. | |
1484 | * | |
1485 | * - TTL: IPv4 or IPv6 TTL, if present and nonzero, otherwise 64. | |
1486 | * | |
1487 | * - TC: IPv4 or IPv6 TOS, if present, otherwise 0. | |
1488 | * | |
1489 | * - BoS: 1. | |
1490 | * | |
22d38fca | 1491 | * If the new label is the second or later label MPLS label in 'flow', it is |
8bfd0fda BP |
1492 | * generated as; |
1493 | * | |
368fb7e6 | 1494 | * - label: Copied from outer label. |
8bfd0fda BP |
1495 | * |
1496 | * - TTL: Copied from outer label. | |
1497 | * | |
1498 | * - TC: Copied from outer label. | |
1499 | * | |
1500 | * - BoS: 0. | |
1501 | * | |
1502 | * 'n' must be flow_count_mpls_labels(flow). 'n' must be less than | |
1503 | * FLOW_MAX_MPLS_LABELS (because otherwise flow->mpls_lse[] would overflow). | |
1504 | */ | |
1505 | void | |
1506 | flow_push_mpls(struct flow *flow, int n, ovs_be16 mpls_eth_type, | |
1507 | struct flow_wildcards *wc) | |
1508 | { | |
1509 | ovs_assert(eth_type_mpls(mpls_eth_type)); | |
1510 | ovs_assert(n < FLOW_MAX_MPLS_LABELS); | |
1511 | ||
8bfd0fda BP |
1512 | if (n) { |
1513 | int i; | |
1514 | ||
22d38fca JR |
1515 | if (wc) { |
1516 | memset(&wc->masks.mpls_lse, 0xff, sizeof *wc->masks.mpls_lse * n); | |
1517 | } | |
8bfd0fda BP |
1518 | for (i = n; i >= 1; i--) { |
1519 | flow->mpls_lse[i] = flow->mpls_lse[i - 1]; | |
1520 | } | |
22d38fca | 1521 | flow->mpls_lse[0] = (flow->mpls_lse[1] & htonl(~MPLS_BOS_MASK)); |
8bfd0fda BP |
1522 | } else { |
1523 | int label = 0; /* IPv4 Explicit Null. */ | |
1524 | int tc = 0; | |
1525 | int ttl = 64; | |
1526 | ||
1527 | if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
1528 | label = 2; | |
1529 | } | |
1530 | ||
1531 | if (is_ip_any(flow)) { | |
1532 | tc = (flow->nw_tos & IP_DSCP_MASK) >> 2; | |
22d38fca JR |
1533 | if (wc) { |
1534 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
1535 | wc->masks.nw_ttl = 0xff; | |
1536 | } | |
8bfd0fda BP |
1537 | |
1538 | if (flow->nw_ttl) { | |
1539 | ttl = flow->nw_ttl; | |
1540 | } | |
8bfd0fda BP |
1541 | } |
1542 | ||
1543 | flow->mpls_lse[0] = set_mpls_lse_values(ttl, tc, 1, htonl(label)); | |
1544 | ||
1545 | /* Clear all L3 and L4 fields. */ | |
c61f3870 | 1546 | BUILD_ASSERT(FLOW_WC_SEQ == 28); |
8bfd0fda BP |
1547 | memset((char *) flow + FLOW_SEGMENT_2_ENDS_AT, 0, |
1548 | sizeof(struct flow) - FLOW_SEGMENT_2_ENDS_AT); | |
1549 | } | |
1550 | flow->dl_type = mpls_eth_type; | |
1551 | } | |
1552 | ||
1553 | /* Tries to remove the outermost MPLS label from 'flow'. Returns true if | |
1554 | * successful, false otherwise. On success, sets 'flow''s Ethernet type to | |
1555 | * 'eth_type'. | |
1556 | * | |
1557 | * 'n' must be flow_count_mpls_labels(flow). */ | |
1558 | bool | |
1559 | flow_pop_mpls(struct flow *flow, int n, ovs_be16 eth_type, | |
1560 | struct flow_wildcards *wc) | |
1561 | { | |
1562 | int i; | |
1563 | ||
1564 | if (n == 0) { | |
1565 | /* Nothing to pop. */ | |
1566 | return false; | |
22d38fca JR |
1567 | } else if (n == FLOW_MAX_MPLS_LABELS) { |
1568 | if (wc) { | |
1569 | wc->masks.mpls_lse[n - 1] |= htonl(MPLS_BOS_MASK); | |
1570 | } | |
1571 | if (!(flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK))) { | |
1572 | /* Can't pop because don't know what to fill in mpls_lse[n - 1]. */ | |
1573 | return false; | |
1574 | } | |
8bfd0fda BP |
1575 | } |
1576 | ||
22d38fca JR |
1577 | if (wc) { |
1578 | memset(&wc->masks.mpls_lse[1], 0xff, | |
1579 | sizeof *wc->masks.mpls_lse * (n - 1)); | |
1580 | } | |
8bfd0fda BP |
1581 | for (i = 1; i < n; i++) { |
1582 | flow->mpls_lse[i - 1] = flow->mpls_lse[i]; | |
1583 | } | |
1584 | flow->mpls_lse[n - 1] = 0; | |
1585 | flow->dl_type = eth_type; | |
1586 | return true; | |
1587 | } | |
1588 | ||
b02475c5 SH |
1589 | /* Sets the MPLS Label that 'flow' matches to 'label', which is interpreted |
1590 | * as an OpenFlow 1.1 "mpls_label" value. */ | |
1591 | void | |
8bfd0fda | 1592 | flow_set_mpls_label(struct flow *flow, int idx, ovs_be32 label) |
b02475c5 | 1593 | { |
8bfd0fda | 1594 | set_mpls_lse_label(&flow->mpls_lse[idx], label); |
b02475c5 SH |
1595 | } |
1596 | ||
b676167a SH |
1597 | /* Sets the MPLS TTL that 'flow' matches to 'ttl', which should be in the |
1598 | * range 0...255. */ | |
1599 | void | |
8bfd0fda | 1600 | flow_set_mpls_ttl(struct flow *flow, int idx, uint8_t ttl) |
b676167a | 1601 | { |
8bfd0fda | 1602 | set_mpls_lse_ttl(&flow->mpls_lse[idx], ttl); |
b676167a SH |
1603 | } |
1604 | ||
b02475c5 SH |
1605 | /* Sets the MPLS TC that 'flow' matches to 'tc', which should be in the |
1606 | * range 0...7. */ | |
1607 | void | |
8bfd0fda | 1608 | flow_set_mpls_tc(struct flow *flow, int idx, uint8_t tc) |
b02475c5 | 1609 | { |
8bfd0fda | 1610 | set_mpls_lse_tc(&flow->mpls_lse[idx], tc); |
b02475c5 SH |
1611 | } |
1612 | ||
1613 | /* Sets the MPLS BOS bit that 'flow' matches to which should be 0 or 1. */ | |
1614 | void | |
8bfd0fda | 1615 | flow_set_mpls_bos(struct flow *flow, int idx, uint8_t bos) |
b02475c5 | 1616 | { |
8bfd0fda | 1617 | set_mpls_lse_bos(&flow->mpls_lse[idx], bos); |
b02475c5 SH |
1618 | } |
1619 | ||
8bfd0fda BP |
1620 | /* Sets the entire MPLS LSE. */ |
1621 | void | |
1622 | flow_set_mpls_lse(struct flow *flow, int idx, ovs_be32 lse) | |
1623 | { | |
1624 | flow->mpls_lse[idx] = lse; | |
1625 | } | |
52105b67 | 1626 | |
437d0d22 | 1627 | static size_t |
52105b67 JR |
1628 | flow_compose_l4(struct ofpbuf *b, const struct flow *flow) |
1629 | { | |
437d0d22 JR |
1630 | size_t l4_len = 0; |
1631 | ||
52105b67 JR |
1632 | if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) |
1633 | || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { | |
1634 | if (flow->nw_proto == IPPROTO_TCP) { | |
1635 | struct tcp_header *tcp; | |
1636 | ||
437d0d22 JR |
1637 | l4_len = sizeof *tcp; |
1638 | tcp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1639 | tcp->tcp_src = flow->tp_src; |
1640 | tcp->tcp_dst = flow->tp_dst; | |
1641 | tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); | |
52105b67 JR |
1642 | } else if (flow->nw_proto == IPPROTO_UDP) { |
1643 | struct udp_header *udp; | |
1644 | ||
437d0d22 JR |
1645 | l4_len = sizeof *udp; |
1646 | udp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1647 | udp->udp_src = flow->tp_src; |
1648 | udp->udp_dst = flow->tp_dst; | |
52105b67 JR |
1649 | } else if (flow->nw_proto == IPPROTO_SCTP) { |
1650 | struct sctp_header *sctp; | |
1651 | ||
437d0d22 JR |
1652 | l4_len = sizeof *sctp; |
1653 | sctp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1654 | sctp->sctp_src = flow->tp_src; |
1655 | sctp->sctp_dst = flow->tp_dst; | |
52105b67 JR |
1656 | } else if (flow->nw_proto == IPPROTO_ICMP) { |
1657 | struct icmp_header *icmp; | |
1658 | ||
437d0d22 JR |
1659 | l4_len = sizeof *icmp; |
1660 | icmp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1661 | icmp->icmp_type = ntohs(flow->tp_src); |
1662 | icmp->icmp_code = ntohs(flow->tp_dst); | |
1663 | icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); | |
0e612675 FL |
1664 | } else if (flow->nw_proto == IPPROTO_IGMP) { |
1665 | struct igmp_header *igmp; | |
1666 | ||
1667 | l4_len = sizeof *igmp; | |
1668 | igmp = ofpbuf_put_zeros(b, l4_len); | |
1669 | igmp->igmp_type = ntohs(flow->tp_src); | |
1670 | igmp->igmp_code = ntohs(flow->tp_dst); | |
1671 | put_16aligned_be32(&igmp->group, flow->igmp_group_ip4); | |
1672 | igmp->igmp_csum = csum(igmp, IGMP_HEADER_LEN); | |
52105b67 JR |
1673 | } else if (flow->nw_proto == IPPROTO_ICMPV6) { |
1674 | struct icmp6_hdr *icmp; | |
1675 | ||
437d0d22 JR |
1676 | l4_len = sizeof *icmp; |
1677 | icmp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1678 | icmp->icmp6_type = ntohs(flow->tp_src); |
1679 | icmp->icmp6_code = ntohs(flow->tp_dst); | |
1680 | ||
1681 | if (icmp->icmp6_code == 0 && | |
1682 | (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || | |
1683 | icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { | |
1684 | struct in6_addr *nd_target; | |
1685 | struct nd_opt_hdr *nd_opt; | |
1686 | ||
437d0d22 | 1687 | l4_len += sizeof *nd_target; |
52105b67 JR |
1688 | nd_target = ofpbuf_put_zeros(b, sizeof *nd_target); |
1689 | *nd_target = flow->nd_target; | |
1690 | ||
1691 | if (!eth_addr_is_zero(flow->arp_sha)) { | |
437d0d22 | 1692 | l4_len += 8; |
52105b67 JR |
1693 | nd_opt = ofpbuf_put_zeros(b, 8); |
1694 | nd_opt->nd_opt_len = 1; | |
1695 | nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; | |
1696 | memcpy(nd_opt + 1, flow->arp_sha, ETH_ADDR_LEN); | |
1697 | } | |
1698 | if (!eth_addr_is_zero(flow->arp_tha)) { | |
437d0d22 | 1699 | l4_len += 8; |
52105b67 JR |
1700 | nd_opt = ofpbuf_put_zeros(b, 8); |
1701 | nd_opt->nd_opt_len = 1; | |
1702 | nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; | |
1703 | memcpy(nd_opt + 1, flow->arp_tha, ETH_ADDR_LEN); | |
1704 | } | |
1705 | } | |
1706 | icmp->icmp6_cksum = (OVS_FORCE uint16_t) | |
1707 | csum(icmp, (char *)ofpbuf_tail(b) - (char *)icmp); | |
52105b67 JR |
1708 | } |
1709 | } | |
437d0d22 | 1710 | return l4_len; |
52105b67 JR |
1711 | } |
1712 | ||
8b3b8dd1 BP |
1713 | /* Puts into 'b' a packet that flow_extract() would parse as having the given |
1714 | * 'flow'. | |
1715 | * | |
1716 | * (This is useful only for testing, obviously, and the packet isn't really | |
dc5a7ce7 | 1717 | * valid. It hasn't got some checksums filled in, for one, and lots of fields |
8b3b8dd1 BP |
1718 | * are just zeroed.) */ |
1719 | void | |
1720 | flow_compose(struct ofpbuf *b, const struct flow *flow) | |
1721 | { | |
437d0d22 JR |
1722 | size_t l4_len; |
1723 | ||
52105b67 | 1724 | /* eth_compose() sets l3 pointer and makes sure it is 32-bit aligned. */ |
8b3b8dd1 BP |
1725 | eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0); |
1726 | if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) { | |
cf3b7538 | 1727 | struct eth_header *eth = ofpbuf_l2(b); |
1f317cb5 | 1728 | eth->eth_type = htons(ofpbuf_size(b)); |
8b3b8dd1 BP |
1729 | return; |
1730 | } | |
1731 | ||
1732 | if (flow->vlan_tci & htons(VLAN_CFI)) { | |
1bf02876 | 1733 | eth_push_vlan(b, htons(ETH_TYPE_VLAN), flow->vlan_tci); |
8b3b8dd1 BP |
1734 | } |
1735 | ||
cff78c88 | 1736 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
8b3b8dd1 BP |
1737 | struct ip_header *ip; |
1738 | ||
52105b67 | 1739 | ip = ofpbuf_put_zeros(b, sizeof *ip); |
8b3b8dd1 | 1740 | ip->ip_ihl_ver = IP_IHL_VER(5, 4); |
eadef313 | 1741 | ip->ip_tos = flow->nw_tos; |
aabf5352 | 1742 | ip->ip_ttl = flow->nw_ttl; |
8b3b8dd1 | 1743 | ip->ip_proto = flow->nw_proto; |
7c457c33 BP |
1744 | put_16aligned_be32(&ip->ip_src, flow->nw_src); |
1745 | put_16aligned_be32(&ip->ip_dst, flow->nw_dst); | |
8b3b8dd1 | 1746 | |
eadef313 | 1747 | if (flow->nw_frag & FLOW_NW_FRAG_ANY) { |
7257b535 | 1748 | ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS); |
eadef313 | 1749 | if (flow->nw_frag & FLOW_NW_FRAG_LATER) { |
7257b535 BP |
1750 | ip->ip_frag_off |= htons(100); |
1751 | } | |
1752 | } | |
df9b6612 | 1753 | |
437d0d22 | 1754 | ofpbuf_set_l4(b, ofpbuf_tail(b)); |
52105b67 | 1755 | |
437d0d22 | 1756 | l4_len = flow_compose_l4(b, flow); |
52105b67 | 1757 | |
1c98d0ad | 1758 | ip = ofpbuf_l3(b); |
437d0d22 | 1759 | ip->ip_tot_len = htons(b->l4_ofs - b->l3_ofs + l4_len); |
dc5a7ce7 | 1760 | ip->ip_csum = csum(ip, sizeof *ip); |
cff78c88 | 1761 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { |
52105b67 JR |
1762 | struct ovs_16aligned_ip6_hdr *nh; |
1763 | ||
1764 | nh = ofpbuf_put_zeros(b, sizeof *nh); | |
1765 | put_16aligned_be32(&nh->ip6_flow, htonl(6 << 28) | | |
1766 | htonl(flow->nw_tos << 20) | flow->ipv6_label); | |
1767 | nh->ip6_hlim = flow->nw_ttl; | |
1768 | nh->ip6_nxt = flow->nw_proto; | |
1769 | ||
1770 | memcpy(&nh->ip6_src, &flow->ipv6_src, sizeof(nh->ip6_src)); | |
1771 | memcpy(&nh->ip6_dst, &flow->ipv6_dst, sizeof(nh->ip6_dst)); | |
1772 | ||
437d0d22 | 1773 | ofpbuf_set_l4(b, ofpbuf_tail(b)); |
52105b67 | 1774 | |
437d0d22 | 1775 | l4_len = flow_compose_l4(b, flow); |
52105b67 | 1776 | |
1c98d0ad | 1777 | nh = ofpbuf_l3(b); |
437d0d22 | 1778 | nh->ip6_plen = htons(l4_len); |
cff78c88 SH |
1779 | } else if (flow->dl_type == htons(ETH_TYPE_ARP) || |
1780 | flow->dl_type == htons(ETH_TYPE_RARP)) { | |
8b3b8dd1 BP |
1781 | struct arp_eth_header *arp; |
1782 | ||
437d0d22 JR |
1783 | arp = ofpbuf_put_zeros(b, sizeof *arp); |
1784 | ofpbuf_set_l3(b, arp); | |
8b3b8dd1 BP |
1785 | arp->ar_hrd = htons(1); |
1786 | arp->ar_pro = htons(ETH_TYPE_IP); | |
1787 | arp->ar_hln = ETH_ADDR_LEN; | |
1788 | arp->ar_pln = 4; | |
1789 | arp->ar_op = htons(flow->nw_proto); | |
1790 | ||
1791 | if (flow->nw_proto == ARP_OP_REQUEST || | |
1792 | flow->nw_proto == ARP_OP_REPLY) { | |
7c457c33 BP |
1793 | put_16aligned_be32(&arp->ar_spa, flow->nw_src); |
1794 | put_16aligned_be32(&arp->ar_tpa, flow->nw_dst); | |
8b3b8dd1 BP |
1795 | memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN); |
1796 | memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN); | |
1797 | } | |
1798 | } | |
b02475c5 SH |
1799 | |
1800 | if (eth_type_mpls(flow->dl_type)) { | |
8bfd0fda BP |
1801 | int n; |
1802 | ||
437d0d22 | 1803 | b->l2_5_ofs = b->l3_ofs; |
8bfd0fda BP |
1804 | for (n = 1; n < FLOW_MAX_MPLS_LABELS; n++) { |
1805 | if (flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK)) { | |
1806 | break; | |
1807 | } | |
1808 | } | |
1809 | while (n > 0) { | |
1810 | push_mpls(b, flow->dl_type, flow->mpls_lse[--n]); | |
1811 | } | |
b02475c5 | 1812 | } |
8b3b8dd1 | 1813 | } |
5cb7a798 BP |
1814 | \f |
1815 | /* Compressed flow. */ | |
1816 | ||
1817 | static int | |
1818 | miniflow_n_values(const struct miniflow *flow) | |
1819 | { | |
fb9aefa3 | 1820 | return count_1bits(flow->map); |
5cb7a798 BP |
1821 | } |
1822 | ||
1823 | static uint32_t * | |
1824 | miniflow_alloc_values(struct miniflow *flow, int n) | |
1825 | { | |
3016f3e4 JR |
1826 | int size = MINIFLOW_VALUES_SIZE(n); |
1827 | ||
1828 | if (size <= sizeof flow->inline_values) { | |
27bbe15d | 1829 | flow->values_inline = true; |
5cb7a798 BP |
1830 | return flow->inline_values; |
1831 | } else { | |
1832 | COVERAGE_INC(miniflow_malloc); | |
27bbe15d | 1833 | flow->values_inline = false; |
3016f3e4 | 1834 | flow->offline_values = xmalloc(size); |
27bbe15d | 1835 | return flow->offline_values; |
5cb7a798 BP |
1836 | } |
1837 | } | |
1838 | ||
df40c152 BP |
1839 | /* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by |
1840 | * the caller. The caller must have already initialized 'dst->map' properly | |
13751fd8 JR |
1841 | * to indicate the significant uint32_t elements of 'src'. 'n' must be the |
1842 | * number of 1-bits in 'dst->map'. | |
1843 | * | |
1844 | * Normally the significant elements are the ones that are non-zero. However, | |
1845 | * when a miniflow is initialized from a (mini)mask, the values can be zeroes, | |
1846 | * so that the flow and mask always have the same maps. | |
df40c152 | 1847 | * |
27bbe15d | 1848 | * This function initializes values (either inline if possible or with |
13751fd8 JR |
1849 | * malloc() otherwise) and copies the uint32_t elements of 'src' indicated by |
1850 | * 'dst->map' into it. */ | |
df40c152 BP |
1851 | static void |
1852 | miniflow_init__(struct miniflow *dst, const struct flow *src, int n) | |
1853 | { | |
1854 | const uint32_t *src_u32 = (const uint32_t *) src; | |
27bbe15d | 1855 | uint32_t *dst_u32 = miniflow_alloc_values(dst, n); |
080e28d0 | 1856 | uint64_t map; |
df40c152 | 1857 | |
080e28d0 | 1858 | for (map = dst->map; map; map = zero_rightmost_1bit(map)) { |
27bbe15d | 1859 | *dst_u32++ = src_u32[raw_ctz(map)]; |
df40c152 BP |
1860 | } |
1861 | } | |
1862 | ||
5cb7a798 | 1863 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' |
27bbe15d JR |
1864 | * with miniflow_destroy(). |
1865 | * Always allocates offline storage. */ | |
5cb7a798 BP |
1866 | void |
1867 | miniflow_init(struct miniflow *dst, const struct flow *src) | |
1868 | { | |
1869 | const uint32_t *src_u32 = (const uint32_t *) src; | |
5cb7a798 BP |
1870 | unsigned int i; |
1871 | int n; | |
1872 | ||
1873 | /* Initialize dst->map, counting the number of nonzero elements. */ | |
1874 | n = 0; | |
080e28d0 JR |
1875 | dst->map = 0; |
1876 | ||
5cb7a798 BP |
1877 | for (i = 0; i < FLOW_U32S; i++) { |
1878 | if (src_u32[i]) { | |
080e28d0 | 1879 | dst->map |= UINT64_C(1) << i; |
5cb7a798 BP |
1880 | n++; |
1881 | } | |
1882 | } | |
1883 | ||
df40c152 BP |
1884 | miniflow_init__(dst, src, n); |
1885 | } | |
5cb7a798 | 1886 | |
df40c152 BP |
1887 | /* Initializes 'dst' as a copy of 'src', using 'mask->map' as 'dst''s map. The |
1888 | * caller must eventually free 'dst' with miniflow_destroy(). */ | |
1889 | void | |
1890 | miniflow_init_with_minimask(struct miniflow *dst, const struct flow *src, | |
1891 | const struct minimask *mask) | |
1892 | { | |
080e28d0 | 1893 | dst->map = mask->masks.map; |
df40c152 | 1894 | miniflow_init__(dst, src, miniflow_n_values(dst)); |
5cb7a798 BP |
1895 | } |
1896 | ||
1897 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
1898 | * with miniflow_destroy(). */ | |
1899 | void | |
1900 | miniflow_clone(struct miniflow *dst, const struct miniflow *src) | |
1901 | { | |
3016f3e4 JR |
1902 | int size = MINIFLOW_VALUES_SIZE(miniflow_n_values(src)); |
1903 | uint32_t *values; | |
1904 | ||
080e28d0 | 1905 | dst->map = src->map; |
3016f3e4 JR |
1906 | if (size <= sizeof dst->inline_values) { |
1907 | dst->values_inline = true; | |
1908 | values = dst->inline_values; | |
1909 | } else { | |
1910 | dst->values_inline = false; | |
1911 | COVERAGE_INC(miniflow_malloc); | |
1912 | dst->offline_values = xmalloc(size); | |
1913 | values = dst->offline_values; | |
1914 | } | |
1915 | memcpy(values, miniflow_get_values(src), size); | |
1916 | } | |
1917 | ||
1918 | /* Initializes 'dst' as a copy of 'src'. The caller must have allocated | |
1919 | * 'dst' to have inline space all data in 'src'. */ | |
1920 | void | |
1921 | miniflow_clone_inline(struct miniflow *dst, const struct miniflow *src, | |
1922 | size_t n_values) | |
1923 | { | |
1924 | dst->values_inline = true; | |
1925 | dst->map = src->map; | |
1926 | memcpy(dst->inline_values, miniflow_get_values(src), | |
1927 | MINIFLOW_VALUES_SIZE(n_values)); | |
5cb7a798 BP |
1928 | } |
1929 | ||
b2c1f00b | 1930 | /* Initializes 'dst' with the data in 'src', destroying 'src'. |
3016f3e4 JR |
1931 | * The caller must eventually free 'dst' with miniflow_destroy(). |
1932 | * 'dst' must be regularly sized miniflow, but 'src' can have | |
3c30d111 JR |
1933 | * storage for more than the default MINI_N_INLINE inline |
1934 | * values. */ | |
b2c1f00b BP |
1935 | void |
1936 | miniflow_move(struct miniflow *dst, struct miniflow *src) | |
1937 | { | |
3016f3e4 JR |
1938 | int size = MINIFLOW_VALUES_SIZE(miniflow_n_values(src)); |
1939 | ||
1940 | dst->map = src->map; | |
1941 | if (size <= sizeof dst->inline_values) { | |
1942 | dst->values_inline = true; | |
1943 | memcpy(dst->inline_values, miniflow_get_values(src), size); | |
1944 | miniflow_destroy(src); | |
097673a6 JR |
1945 | } else if (src->values_inline) { |
1946 | dst->values_inline = false; | |
1947 | COVERAGE_INC(miniflow_malloc); | |
1948 | dst->offline_values = xmalloc(size); | |
1949 | memcpy(dst->offline_values, src->inline_values, size); | |
3016f3e4 JR |
1950 | } else { |
1951 | dst->values_inline = false; | |
1952 | dst->offline_values = src->offline_values; | |
1953 | } | |
b2c1f00b BP |
1954 | } |
1955 | ||
5cb7a798 BP |
1956 | /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' |
1957 | * itself resides; the caller is responsible for that. */ | |
1958 | void | |
1959 | miniflow_destroy(struct miniflow *flow) | |
1960 | { | |
27bbe15d JR |
1961 | if (!flow->values_inline) { |
1962 | free(flow->offline_values); | |
5cb7a798 BP |
1963 | } |
1964 | } | |
1965 | ||
1966 | /* Initializes 'dst' as a copy of 'src'. */ | |
1967 | void | |
1968 | miniflow_expand(const struct miniflow *src, struct flow *dst) | |
1969 | { | |
ad77e3c5 EJ |
1970 | memset(dst, 0, sizeof *dst); |
1971 | flow_union_with_miniflow(dst, src); | |
5cb7a798 BP |
1972 | } |
1973 | ||
5cb7a798 BP |
1974 | /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow' |
1975 | * were expanded into a "struct flow". */ | |
b7807e4f | 1976 | static uint32_t |
5cb7a798 BP |
1977 | miniflow_get(const struct miniflow *flow, unsigned int u32_ofs) |
1978 | { | |
b7807e4f | 1979 | return (flow->map & UINT64_C(1) << u32_ofs) |
27bbe15d | 1980 | ? *(miniflow_get_u32_values(flow) + |
b7807e4f JR |
1981 | count_1bits(flow->map & ((UINT64_C(1) << u32_ofs) - 1))) |
1982 | : 0; | |
5cb7a798 BP |
1983 | } |
1984 | ||
8e86f090 | 1985 | /* Returns true if 'a' and 'b' are the equal miniflow, false otherwise. */ |
5cb7a798 BP |
1986 | bool |
1987 | miniflow_equal(const struct miniflow *a, const struct miniflow *b) | |
1988 | { | |
27bbe15d JR |
1989 | const uint32_t *ap = miniflow_get_u32_values(a); |
1990 | const uint32_t *bp = miniflow_get_u32_values(b); | |
080e28d0 JR |
1991 | const uint64_t a_map = a->map; |
1992 | const uint64_t b_map = b->map; | |
5cb7a798 | 1993 | |
27bbe15d JR |
1994 | if (OVS_LIKELY(a_map == b_map)) { |
1995 | int count = miniflow_n_values(a); | |
1996 | ||
8cc86801 | 1997 | return !memcmp(ap, bp, count * sizeof *ap); |
080e28d0 | 1998 | } else { |
27bbe15d JR |
1999 | uint64_t map; |
2000 | ||
080e28d0 JR |
2001 | for (map = a_map | b_map; map; map = zero_rightmost_1bit(map)) { |
2002 | uint64_t bit = rightmost_1bit(map); | |
2003 | uint64_t a_value = a_map & bit ? *ap++ : 0; | |
2004 | uint64_t b_value = b_map & bit ? *bp++ : 0; | |
df40c152 | 2005 | |
080e28d0 JR |
2006 | if (a_value != b_value) { |
2007 | return false; | |
df40c152 | 2008 | } |
5cb7a798 BP |
2009 | } |
2010 | } | |
2011 | ||
df40c152 | 2012 | return true; |
5cb7a798 BP |
2013 | } |
2014 | ||
de4ad4a2 JR |
2015 | /* Returns false if 'a' and 'b' differ at the places where there are 1-bits |
2016 | * in 'mask', true otherwise. */ | |
5cb7a798 BP |
2017 | bool |
2018 | miniflow_equal_in_minimask(const struct miniflow *a, const struct miniflow *b, | |
2019 | const struct minimask *mask) | |
2020 | { | |
27bbe15d | 2021 | const uint32_t *p = miniflow_get_u32_values(&mask->masks); |
080e28d0 | 2022 | uint64_t map; |
5cb7a798 | 2023 | |
080e28d0 | 2024 | for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { |
d43d314e | 2025 | int ofs = raw_ctz(map); |
5cb7a798 | 2026 | |
27bbe15d | 2027 | if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p++) { |
080e28d0 | 2028 | return false; |
5cb7a798 BP |
2029 | } |
2030 | } | |
2031 | ||
2032 | return true; | |
2033 | } | |
2034 | ||
2035 | /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits | |
2036 | * in 'mask', false if they differ. */ | |
2037 | bool | |
2038 | miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, | |
2039 | const struct minimask *mask) | |
2040 | { | |
2041 | const uint32_t *b_u32 = (const uint32_t *) b; | |
27bbe15d | 2042 | const uint32_t *p = miniflow_get_u32_values(&mask->masks); |
080e28d0 | 2043 | uint64_t map; |
5cb7a798 | 2044 | |
080e28d0 | 2045 | for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { |
d43d314e | 2046 | int ofs = raw_ctz(map); |
5cb7a798 | 2047 | |
27bbe15d | 2048 | if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p++) { |
080e28d0 | 2049 | return false; |
5cb7a798 BP |
2050 | } |
2051 | } | |
2052 | ||
2053 | return true; | |
2054 | } | |
2055 | ||
5cb7a798 BP |
2056 | \f |
2057 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
2058 | * with minimask_destroy(). */ | |
2059 | void | |
2060 | minimask_init(struct minimask *mask, const struct flow_wildcards *wc) | |
2061 | { | |
2062 | miniflow_init(&mask->masks, &wc->masks); | |
2063 | } | |
2064 | ||
2065 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
2066 | * with minimask_destroy(). */ | |
2067 | void | |
2068 | minimask_clone(struct minimask *dst, const struct minimask *src) | |
2069 | { | |
2070 | miniflow_clone(&dst->masks, &src->masks); | |
2071 | } | |
2072 | ||
b2c1f00b BP |
2073 | /* Initializes 'dst' with the data in 'src', destroying 'src'. |
2074 | * The caller must eventually free 'dst' with minimask_destroy(). */ | |
2075 | void | |
2076 | minimask_move(struct minimask *dst, struct minimask *src) | |
2077 | { | |
a24de7ee | 2078 | miniflow_move(&dst->masks, &src->masks); |
b2c1f00b BP |
2079 | } |
2080 | ||
5cb7a798 BP |
2081 | /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'. |
2082 | * | |
2083 | * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use | |
2084 | * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */ | |
2085 | void | |
2086 | minimask_combine(struct minimask *dst_, | |
2087 | const struct minimask *a_, const struct minimask *b_, | |
2088 | uint32_t storage[FLOW_U32S]) | |
2089 | { | |
2090 | struct miniflow *dst = &dst_->masks; | |
27bbe15d | 2091 | uint32_t *dst_values = storage; |
5cb7a798 BP |
2092 | const struct miniflow *a = &a_->masks; |
2093 | const struct miniflow *b = &b_->masks; | |
080e28d0 JR |
2094 | uint64_t map; |
2095 | int n = 0; | |
5cb7a798 | 2096 | |
27bbe15d JR |
2097 | dst->values_inline = false; |
2098 | dst->offline_values = storage; | |
080e28d0 JR |
2099 | |
2100 | dst->map = 0; | |
2101 | for (map = a->map & b->map; map; map = zero_rightmost_1bit(map)) { | |
d43d314e | 2102 | int ofs = raw_ctz(map); |
080e28d0 JR |
2103 | uint32_t mask = miniflow_get(a, ofs) & miniflow_get(b, ofs); |
2104 | ||
2105 | if (mask) { | |
2106 | dst->map |= rightmost_1bit(map); | |
27bbe15d | 2107 | dst_values[n++] = mask; |
5cb7a798 BP |
2108 | } |
2109 | } | |
2110 | } | |
2111 | ||
2112 | /* Frees any memory owned by 'mask'. Does not free the storage in which 'mask' | |
2113 | * itself resides; the caller is responsible for that. */ | |
2114 | void | |
2115 | minimask_destroy(struct minimask *mask) | |
2116 | { | |
2117 | miniflow_destroy(&mask->masks); | |
2118 | } | |
2119 | ||
2120 | /* Initializes 'dst' as a copy of 'src'. */ | |
2121 | void | |
2122 | minimask_expand(const struct minimask *mask, struct flow_wildcards *wc) | |
2123 | { | |
2124 | miniflow_expand(&mask->masks, &wc->masks); | |
2125 | } | |
2126 | ||
2127 | /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask' | |
2128 | * were expanded into a "struct flow_wildcards". */ | |
2129 | uint32_t | |
2130 | minimask_get(const struct minimask *mask, unsigned int u32_ofs) | |
2131 | { | |
2132 | return miniflow_get(&mask->masks, u32_ofs); | |
2133 | } | |
2134 | ||
5cb7a798 BP |
2135 | /* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */ |
2136 | bool | |
2137 | minimask_equal(const struct minimask *a, const struct minimask *b) | |
2138 | { | |
2139 | return miniflow_equal(&a->masks, &b->masks); | |
2140 | } | |
2141 | ||
d4570fd8 | 2142 | /* Returns true if at least one bit matched by 'b' is wildcarded by 'a', |
5cb7a798 BP |
2143 | * false otherwise. */ |
2144 | bool | |
d4570fd8 | 2145 | minimask_has_extra(const struct minimask *a, const struct minimask *b) |
5cb7a798 | 2146 | { |
27bbe15d | 2147 | const uint32_t *p = miniflow_get_u32_values(&b->masks); |
080e28d0 | 2148 | uint64_t map; |
5cb7a798 | 2149 | |
d4570fd8 JR |
2150 | for (map = b->masks.map; map; map = zero_rightmost_1bit(map)) { |
2151 | uint32_t a_u32 = minimask_get(a, raw_ctz(map)); | |
2152 | uint32_t b_u32 = *p++; | |
5cb7a798 | 2153 | |
080e28d0 JR |
2154 | if ((a_u32 & b_u32) != b_u32) { |
2155 | return true; | |
5cb7a798 BP |
2156 | } |
2157 | } | |
2158 | ||
2159 | return false; | |
2160 | } |