]>
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 | |
75 | == offsetof(struct flow, pad) && | |
76 | offsetof(struct flow, tcp_flags) / 4 | |
77 | == offsetof(struct flow, pad) / 4); | |
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 | ||
694ffecc | 124 | #if (FLOW_WC_SEQ != 27) |
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; | |
426 | ||
427 | if (OVS_UNLIKELY(size < IP_HEADER_LEN)) { | |
428 | goto out; | |
429 | } | |
430 | ip_len = IP_IHL(nh->ip_ihl_ver) * 4; | |
431 | ||
432 | if (OVS_UNLIKELY(ip_len < IP_HEADER_LEN)) { | |
433 | goto out; | |
434 | } | |
435 | ||
436 | /* Push both source and destination address at once. */ | |
437 | miniflow_push_words(mf, nw_src, &nh->ip_src, 2); | |
438 | ||
439 | nw_tos = nh->ip_tos; | |
440 | nw_ttl = nh->ip_ttl; | |
441 | nw_proto = nh->ip_proto; | |
442 | if (OVS_UNLIKELY(IP_IS_FRAGMENT(nh->ip_frag_off))) { | |
443 | nw_frag = FLOW_NW_FRAG_ANY; | |
444 | if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) { | |
445 | nw_frag |= FLOW_NW_FRAG_LATER; | |
446 | } | |
447 | } | |
448 | if (OVS_UNLIKELY(size < ip_len)) { | |
449 | goto out; | |
450 | } | |
451 | data_pull(&data, &size, ip_len); | |
452 | ||
453 | } else if (dl_type == htons(ETH_TYPE_IPV6)) { | |
454 | const struct ovs_16aligned_ip6_hdr *nh; | |
455 | ovs_be32 tc_flow; | |
456 | ||
457 | if (OVS_UNLIKELY(size < sizeof *nh)) { | |
458 | goto out; | |
459 | } | |
460 | nh = data_pull(&data, &size, sizeof *nh); | |
461 | ||
462 | miniflow_push_words(mf, ipv6_src, &nh->ip6_src, | |
463 | sizeof nh->ip6_src / 4); | |
464 | miniflow_push_words(mf, ipv6_dst, &nh->ip6_dst, | |
465 | sizeof nh->ip6_dst / 4); | |
466 | ||
467 | tc_flow = get_16aligned_be32(&nh->ip6_flow); | |
468 | { | |
469 | ovs_be32 label = tc_flow & htonl(IPV6_LABEL_MASK); | |
470 | miniflow_push_be32_check(mf, ipv6_label, label); | |
471 | } | |
472 | ||
473 | nw_tos = ntohl(tc_flow) >> 20; | |
474 | nw_ttl = nh->ip6_hlim; | |
475 | nw_proto = nh->ip6_nxt; | |
476 | ||
477 | while (1) { | |
478 | if (OVS_LIKELY((nw_proto != IPPROTO_HOPOPTS) | |
479 | && (nw_proto != IPPROTO_ROUTING) | |
480 | && (nw_proto != IPPROTO_DSTOPTS) | |
481 | && (nw_proto != IPPROTO_AH) | |
482 | && (nw_proto != IPPROTO_FRAGMENT))) { | |
483 | /* It's either a terminal header (e.g., TCP, UDP) or one we | |
484 | * don't understand. In either case, we're done with the | |
485 | * packet, so use it to fill in 'nw_proto'. */ | |
486 | break; | |
487 | } | |
488 | ||
489 | /* We only verify that at least 8 bytes of the next header are | |
490 | * available, but many of these headers are longer. Ensure that | |
491 | * accesses within the extension header are within those first 8 | |
492 | * bytes. All extension headers are required to be at least 8 | |
493 | * bytes. */ | |
494 | if (OVS_UNLIKELY(size < 8)) { | |
495 | goto out; | |
7257b535 | 496 | } |
419681da JR |
497 | |
498 | if ((nw_proto == IPPROTO_HOPOPTS) | |
499 | || (nw_proto == IPPROTO_ROUTING) | |
500 | || (nw_proto == IPPROTO_DSTOPTS)) { | |
501 | /* These headers, while different, have the fields we care | |
502 | * about in the same location and with the same | |
503 | * interpretation. */ | |
504 | const struct ip6_ext *ext_hdr = data; | |
505 | nw_proto = ext_hdr->ip6e_nxt; | |
506 | if (OVS_UNLIKELY(!data_try_pull(&data, &size, | |
507 | (ext_hdr->ip6e_len + 1) * 8))) { | |
508 | goto out; | |
509 | } | |
510 | } else if (nw_proto == IPPROTO_AH) { | |
511 | /* A standard AH definition isn't available, but the fields | |
512 | * we care about are in the same location as the generic | |
513 | * option header--only the header length is calculated | |
514 | * differently. */ | |
515 | const struct ip6_ext *ext_hdr = data; | |
516 | nw_proto = ext_hdr->ip6e_nxt; | |
517 | if (OVS_UNLIKELY(!data_try_pull(&data, &size, | |
518 | (ext_hdr->ip6e_len + 2) * 4))) { | |
519 | goto out; | |
520 | } | |
521 | } else if (nw_proto == IPPROTO_FRAGMENT) { | |
522 | const struct ovs_16aligned_ip6_frag *frag_hdr = data; | |
523 | ||
524 | nw_proto = frag_hdr->ip6f_nxt; | |
525 | if (!data_try_pull(&data, &size, sizeof *frag_hdr)) { | |
526 | goto out; | |
527 | } | |
528 | ||
529 | /* We only process the first fragment. */ | |
530 | if (frag_hdr->ip6f_offlg != htons(0)) { | |
531 | nw_frag = FLOW_NW_FRAG_ANY; | |
532 | if ((frag_hdr->ip6f_offlg & IP6F_OFF_MASK) != htons(0)) { | |
533 | nw_frag |= FLOW_NW_FRAG_LATER; | |
534 | nw_proto = IPPROTO_FRAGMENT; | |
535 | break; | |
064af421 | 536 | } |
064af421 | 537 | } |
50f06e16 BP |
538 | } |
539 | } | |
419681da JR |
540 | } else { |
541 | if (dl_type == htons(ETH_TYPE_ARP) || | |
542 | dl_type == htons(ETH_TYPE_RARP)) { | |
543 | uint8_t arp_buf[2][ETH_ADDR_LEN]; | |
544 | const struct arp_eth_header *arp = (const struct arp_eth_header *) | |
545 | data_try_pull(&data, &size, ARP_ETH_HEADER_LEN); | |
546 | ||
547 | if (OVS_LIKELY(arp) && OVS_LIKELY(arp->ar_hrd == htons(1)) | |
548 | && OVS_LIKELY(arp->ar_pro == htons(ETH_TYPE_IP)) | |
549 | && OVS_LIKELY(arp->ar_hln == ETH_ADDR_LEN) | |
550 | && OVS_LIKELY(arp->ar_pln == 4)) { | |
551 | miniflow_push_words(mf, nw_src, &arp->ar_spa, 1); | |
552 | miniflow_push_words(mf, nw_dst, &arp->ar_tpa, 1); | |
553 | ||
554 | /* We only match on the lower 8 bits of the opcode. */ | |
555 | if (OVS_LIKELY(ntohs(arp->ar_op) <= 0xff)) { | |
556 | miniflow_push_be32(mf, nw_frag, htonl(ntohs(arp->ar_op))); | |
557 | } | |
d31f1109 | 558 | |
419681da JR |
559 | /* Must be adjacent. */ |
560 | BUILD_ASSERT(offsetof(struct flow, arp_sha) + 6 | |
561 | == offsetof(struct flow, arp_tha)); | |
562 | ||
563 | memcpy(arp_buf[0], arp->ar_sha, ETH_ADDR_LEN); | |
564 | memcpy(arp_buf[1], arp->ar_tha, ETH_ADDR_LEN); | |
565 | miniflow_push_words(mf, arp_sha, arp_buf, | |
566 | ETH_ADDR_LEN * 2 / 4); | |
567 | } | |
d31f1109 | 568 | } |
419681da JR |
569 | goto out; |
570 | } | |
571 | ||
572 | packet->l4_ofs = (char *)data - l2; | |
573 | miniflow_push_be32(mf, nw_frag, | |
574 | BYTES_TO_BE32(nw_frag, nw_tos, nw_ttl, nw_proto)); | |
575 | ||
576 | if (OVS_LIKELY(!(nw_frag & FLOW_NW_FRAG_LATER))) { | |
577 | if (OVS_LIKELY(nw_proto == IPPROTO_TCP)) { | |
578 | if (OVS_LIKELY(size >= TCP_HEADER_LEN)) { | |
579 | const struct tcp_header *tcp = data; | |
580 | ||
581 | miniflow_push_be32(mf, tcp_flags, | |
582 | TCP_FLAGS_BE32(tcp->tcp_ctl)); | |
583 | miniflow_push_words(mf, tp_src, &tcp->tcp_src, 1); | |
584 | } | |
585 | } else if (OVS_LIKELY(nw_proto == IPPROTO_UDP)) { | |
586 | if (OVS_LIKELY(size >= UDP_HEADER_LEN)) { | |
587 | const struct udp_header *udp = data; | |
588 | ||
589 | miniflow_push_words(mf, tp_src, &udp->udp_src, 1); | |
064af421 | 590 | } |
419681da JR |
591 | } else if (OVS_LIKELY(nw_proto == IPPROTO_SCTP)) { |
592 | if (OVS_LIKELY(size >= SCTP_HEADER_LEN)) { | |
593 | const struct sctp_header *sctp = data; | |
a26ef517 | 594 | |
419681da JR |
595 | miniflow_push_words(mf, tp_src, &sctp->sctp_src, 1); |
596 | } | |
597 | } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMP)) { | |
598 | if (OVS_LIKELY(size >= ICMP_HEADER_LEN)) { | |
599 | const struct icmp_header *icmp = data; | |
600 | ||
601 | miniflow_push_be16(mf, tp_src, htons(icmp->icmp_type)); | |
602 | miniflow_push_be16(mf, tp_dst, htons(icmp->icmp_code)); | |
603 | } | |
0e612675 FL |
604 | } else if (OVS_LIKELY(nw_proto == IPPROTO_IGMP)) { |
605 | if (OVS_LIKELY(size >= IGMP_HEADER_LEN)) { | |
606 | const struct igmp_header *igmp = data; | |
607 | ||
608 | miniflow_push_be16(mf, tp_src, htons(igmp->igmp_type)); | |
609 | miniflow_push_be16(mf, tp_dst, htons(igmp->igmp_code)); | |
610 | miniflow_push_be32(mf, igmp_group_ip4, | |
611 | get_16aligned_be32(&igmp->group)); | |
612 | } | |
419681da JR |
613 | } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMPV6)) { |
614 | if (OVS_LIKELY(size >= sizeof(struct icmp6_hdr))) { | |
615 | const struct in6_addr *nd_target = NULL; | |
616 | uint8_t arp_buf[2][ETH_ADDR_LEN]; | |
617 | const struct icmp6_hdr *icmp = data_pull(&data, &size, | |
618 | sizeof *icmp); | |
619 | memset(arp_buf, 0, sizeof arp_buf); | |
620 | if (OVS_LIKELY(parse_icmpv6(&data, &size, icmp, &nd_target, | |
621 | arp_buf))) { | |
b0e2ec32 JR |
622 | miniflow_push_words(mf, arp_sha, arp_buf, |
623 | ETH_ADDR_LEN * 2 / 4); | |
419681da JR |
624 | if (nd_target) { |
625 | miniflow_push_words(mf, nd_target, nd_target, | |
626 | sizeof *nd_target / 4); | |
627 | } | |
419681da JR |
628 | miniflow_push_be16(mf, tp_src, htons(icmp->icmp6_type)); |
629 | miniflow_push_be16(mf, tp_dst, htons(icmp->icmp6_code)); | |
630 | } | |
631 | } | |
064af421 BP |
632 | } |
633 | } | |
419681da JR |
634 | if (md) { |
635 | miniflow_push_uint32_check(mf, dp_hash, md->dp_hash); | |
636 | } | |
637 | out: | |
638 | dst->map = mf.map; | |
064af421 BP |
639 | } |
640 | ||
993410fb BP |
641 | /* For every bit of a field that is wildcarded in 'wildcards', sets the |
642 | * corresponding bit in 'flow' to zero. */ | |
643 | void | |
644 | flow_zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards) | |
645 | { | |
659c2346 BP |
646 | uint32_t *flow_u32 = (uint32_t *) flow; |
647 | const uint32_t *wc_u32 = (const uint32_t *) &wildcards->masks; | |
648 | size_t i; | |
993410fb | 649 | |
659c2346 BP |
650 | for (i = 0; i < FLOW_U32S; i++) { |
651 | flow_u32[i] &= wc_u32[i]; | |
26720e24 | 652 | } |
993410fb BP |
653 | } |
654 | ||
d8d9c698 EJ |
655 | void |
656 | flow_unwildcard_tp_ports(const struct flow *flow, struct flow_wildcards *wc) | |
657 | { | |
658 | if (flow->nw_proto != IPPROTO_ICMP) { | |
659 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
660 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
661 | } else { | |
662 | wc->masks.tp_src = htons(0xff); | |
663 | wc->masks.tp_dst = htons(0xff); | |
664 | } | |
665 | } | |
666 | ||
5d6c3af0 EJ |
667 | /* Initializes 'fmd' with the metadata found in 'flow'. */ |
668 | void | |
669 | flow_get_metadata(const struct flow *flow, struct flow_metadata *fmd) | |
670 | { | |
0e612675 | 671 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 27); |
e9358af6 | 672 | |
a79f29f2 AZ |
673 | fmd->dp_hash = flow->dp_hash; |
674 | fmd->recirc_id = flow->recirc_id; | |
296e07ac | 675 | fmd->tun_id = flow->tunnel.tun_id; |
0ad90c84 JR |
676 | fmd->tun_src = flow->tunnel.ip_src; |
677 | fmd->tun_dst = flow->tunnel.ip_dst; | |
969fc56c | 678 | fmd->metadata = flow->metadata; |
5d6c3af0 | 679 | memcpy(fmd->regs, flow->regs, sizeof fmd->regs); |
ac923e91 | 680 | fmd->pkt_mark = flow->pkt_mark; |
4e022ec0 | 681 | fmd->in_port = flow->in_port.ofp_port; |
5d6c3af0 EJ |
682 | } |
683 | ||
064af421 | 684 | char * |
ae412e7d | 685 | flow_to_string(const struct flow *flow) |
064af421 BP |
686 | { |
687 | struct ds ds = DS_EMPTY_INITIALIZER; | |
688 | flow_format(&ds, flow); | |
689 | return ds_cstr(&ds); | |
690 | } | |
691 | ||
4fe3445a PS |
692 | const char * |
693 | flow_tun_flag_to_string(uint32_t flags) | |
694 | { | |
695 | switch (flags) { | |
696 | case FLOW_TNL_F_DONT_FRAGMENT: | |
697 | return "df"; | |
698 | case FLOW_TNL_F_CSUM: | |
699 | return "csum"; | |
700 | case FLOW_TNL_F_KEY: | |
701 | return "key"; | |
94872594 JG |
702 | case FLOW_TNL_F_OAM: |
703 | return "oam"; | |
4fe3445a PS |
704 | default: |
705 | return NULL; | |
706 | } | |
707 | } | |
708 | ||
709 | void | |
710 | format_flags(struct ds *ds, const char *(*bit_to_string)(uint32_t), | |
711 | uint32_t flags, char del) | |
712 | { | |
713 | uint32_t bad = 0; | |
714 | ||
715 | if (!flags) { | |
716 | return; | |
717 | } | |
718 | while (flags) { | |
719 | uint32_t bit = rightmost_1bit(flags); | |
720 | const char *s; | |
721 | ||
722 | s = bit_to_string(bit); | |
723 | if (s) { | |
724 | ds_put_format(ds, "%s%c", s, del); | |
725 | } else { | |
726 | bad |= bit; | |
727 | } | |
728 | ||
729 | flags &= ~bit; | |
730 | } | |
731 | ||
732 | if (bad) { | |
733 | ds_put_format(ds, "0x%"PRIx32"%c", bad, del); | |
734 | } | |
735 | ds_chomp(ds, del); | |
736 | } | |
737 | ||
61bf6666 JR |
738 | void |
739 | format_flags_masked(struct ds *ds, const char *name, | |
740 | const char *(*bit_to_string)(uint32_t), uint32_t flags, | |
741 | uint32_t mask) | |
742 | { | |
743 | if (name) { | |
744 | ds_put_format(ds, "%s=", name); | |
745 | } | |
746 | while (mask) { | |
747 | uint32_t bit = rightmost_1bit(mask); | |
748 | const char *s = bit_to_string(bit); | |
749 | ||
750 | ds_put_format(ds, "%s%s", (flags & bit) ? "+" : "-", | |
751 | s ? s : "[Unknown]"); | |
752 | mask &= ~bit; | |
753 | } | |
754 | } | |
755 | ||
064af421 | 756 | void |
ae412e7d | 757 | flow_format(struct ds *ds, const struct flow *flow) |
064af421 | 758 | { |
aa6c9932 | 759 | struct match match; |
78c9486d | 760 | struct flow_wildcards *wc = &match.wc; |
296e07ac | 761 | |
aa6c9932 | 762 | match_wc_init(&match, flow); |
78c9486d JR |
763 | |
764 | /* As this function is most often used for formatting a packet in a | |
765 | * packet-in message, skip formatting the packet context fields that are | |
766 | * all-zeroes (Openflow spec encourages leaving out all-zeroes context | |
767 | * fields from the packet-in messages). We make an exception with the | |
768 | * 'in_port' field, which we always format, as packets usually have an | |
769 | * in_port, and 0 is a port just like any other port. */ | |
770 | if (!flow->skb_priority) { | |
771 | WC_UNMASK_FIELD(wc, skb_priority); | |
772 | } | |
773 | if (!flow->pkt_mark) { | |
774 | WC_UNMASK_FIELD(wc, pkt_mark); | |
775 | } | |
776 | if (!flow->recirc_id) { | |
777 | WC_UNMASK_FIELD(wc, recirc_id); | |
778 | } | |
779 | for (int i = 0; i < FLOW_N_REGS; i++) { | |
780 | if (!flow->regs[i]) { | |
781 | WC_UNMASK_FIELD(wc, regs[i]); | |
782 | } | |
783 | } | |
784 | if (!flow->metadata) { | |
785 | WC_UNMASK_FIELD(wc, metadata); | |
786 | } | |
787 | ||
3f78c3cc | 788 | match_format(&match, ds, OFP_DEFAULT_PRIORITY); |
064af421 BP |
789 | } |
790 | ||
791 | void | |
ae412e7d | 792 | flow_print(FILE *stream, const struct flow *flow) |
064af421 BP |
793 | { |
794 | char *s = flow_to_string(flow); | |
795 | fputs(s, stream); | |
796 | free(s); | |
797 | } | |
54363004 BP |
798 | \f |
799 | /* flow_wildcards functions. */ | |
800 | ||
d8ae4d67 | 801 | /* Initializes 'wc' as a set of wildcards that matches every packet. */ |
54363004 | 802 | void |
d8ae4d67 | 803 | flow_wildcards_init_catchall(struct flow_wildcards *wc) |
54363004 | 804 | { |
659c2346 | 805 | memset(&wc->masks, 0, sizeof wc->masks); |
54363004 BP |
806 | } |
807 | ||
78c9486d JR |
808 | /* Converts a flow into flow wildcards. It sets the wildcard masks based on |
809 | * the packet headers extracted to 'flow'. It will not set the mask for fields | |
810 | * that do not make sense for the packet type. OpenFlow-only metadata is | |
811 | * wildcarded, but other metadata is unconditionally exact-matched. */ | |
812 | void flow_wildcards_init_for_packet(struct flow_wildcards *wc, | |
813 | const struct flow *flow) | |
814 | { | |
815 | memset(&wc->masks, 0x0, sizeof wc->masks); | |
816 | ||
817 | if (flow->tunnel.ip_dst) { | |
818 | if (flow->tunnel.flags & FLOW_TNL_F_KEY) { | |
819 | WC_MASK_FIELD(wc, tunnel.tun_id); | |
820 | } | |
821 | WC_MASK_FIELD(wc, tunnel.ip_src); | |
822 | WC_MASK_FIELD(wc, tunnel.ip_dst); | |
823 | WC_MASK_FIELD(wc, tunnel.flags); | |
824 | WC_MASK_FIELD(wc, tunnel.ip_tos); | |
825 | WC_MASK_FIELD(wc, tunnel.ip_ttl); | |
826 | WC_MASK_FIELD(wc, tunnel.tp_src); | |
827 | WC_MASK_FIELD(wc, tunnel.tp_dst); | |
828 | } else if (flow->tunnel.tun_id) { | |
829 | WC_MASK_FIELD(wc, tunnel.tun_id); | |
830 | } | |
831 | ||
832 | /* metadata and regs wildcarded. */ | |
833 | ||
834 | WC_MASK_FIELD(wc, skb_priority); | |
835 | WC_MASK_FIELD(wc, pkt_mark); | |
836 | WC_MASK_FIELD(wc, recirc_id); | |
837 | WC_MASK_FIELD(wc, dp_hash); | |
838 | WC_MASK_FIELD(wc, in_port); | |
839 | ||
840 | WC_MASK_FIELD(wc, dl_dst); | |
841 | WC_MASK_FIELD(wc, dl_src); | |
842 | WC_MASK_FIELD(wc, dl_type); | |
843 | WC_MASK_FIELD(wc, vlan_tci); | |
844 | ||
845 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
846 | WC_MASK_FIELD(wc, nw_src); | |
847 | WC_MASK_FIELD(wc, nw_dst); | |
848 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
849 | WC_MASK_FIELD(wc, ipv6_src); | |
850 | WC_MASK_FIELD(wc, ipv6_dst); | |
851 | WC_MASK_FIELD(wc, ipv6_label); | |
852 | } else if (flow->dl_type == htons(ETH_TYPE_ARP) || | |
853 | flow->dl_type == htons(ETH_TYPE_RARP)) { | |
854 | WC_MASK_FIELD(wc, nw_src); | |
855 | WC_MASK_FIELD(wc, nw_dst); | |
856 | WC_MASK_FIELD(wc, nw_proto); | |
857 | WC_MASK_FIELD(wc, arp_sha); | |
858 | WC_MASK_FIELD(wc, arp_tha); | |
859 | return; | |
860 | } else if (eth_type_mpls(flow->dl_type)) { | |
861 | for (int i = 0; i < FLOW_MAX_MPLS_LABELS; i++) { | |
862 | WC_MASK_FIELD(wc, mpls_lse[i]); | |
863 | if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { | |
864 | break; | |
865 | } | |
866 | } | |
867 | return; | |
868 | } else { | |
869 | return; /* Unknown ethertype. */ | |
870 | } | |
871 | ||
872 | /* IPv4 or IPv6. */ | |
873 | WC_MASK_FIELD(wc, nw_frag); | |
874 | WC_MASK_FIELD(wc, nw_tos); | |
875 | WC_MASK_FIELD(wc, nw_ttl); | |
876 | WC_MASK_FIELD(wc, nw_proto); | |
877 | ||
878 | /* No transport layer header in later fragments. */ | |
879 | if (!(flow->nw_frag & FLOW_NW_FRAG_LATER) && | |
880 | (flow->nw_proto == IPPROTO_ICMP || | |
881 | flow->nw_proto == IPPROTO_ICMPV6 || | |
882 | flow->nw_proto == IPPROTO_TCP || | |
883 | flow->nw_proto == IPPROTO_UDP || | |
884 | flow->nw_proto == IPPROTO_SCTP || | |
885 | flow->nw_proto == IPPROTO_IGMP)) { | |
886 | WC_MASK_FIELD(wc, tp_src); | |
887 | WC_MASK_FIELD(wc, tp_dst); | |
888 | ||
889 | if (flow->nw_proto == IPPROTO_TCP) { | |
890 | WC_MASK_FIELD(wc, tcp_flags); | |
891 | } else if (flow->nw_proto == IPPROTO_ICMPV6) { | |
892 | WC_MASK_FIELD(wc, arp_sha); | |
893 | WC_MASK_FIELD(wc, arp_tha); | |
894 | WC_MASK_FIELD(wc, nd_target); | |
895 | } else if (flow->nw_proto == IPPROTO_IGMP) { | |
896 | WC_MASK_FIELD(wc, igmp_group_ip4); | |
897 | } | |
898 | } | |
899 | } | |
900 | ||
c11c6faa AZ |
901 | /* Clear the metadata and register wildcard masks. They are not packet |
902 | * header fields. */ | |
903 | void | |
904 | flow_wildcards_clear_non_packet_fields(struct flow_wildcards *wc) | |
905 | { | |
906 | memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata); | |
907 | memset(&wc->masks.regs, 0, sizeof wc->masks.regs); | |
908 | } | |
909 | ||
ecf1e7ac BP |
910 | /* Returns true if 'wc' matches every packet, false if 'wc' fixes any bits or |
911 | * fields. */ | |
912 | bool | |
913 | flow_wildcards_is_catchall(const struct flow_wildcards *wc) | |
914 | { | |
659c2346 BP |
915 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; |
916 | size_t i; | |
ecf1e7ac | 917 | |
659c2346 BP |
918 | for (i = 0; i < FLOW_U32S; i++) { |
919 | if (wc_u32[i]) { | |
ecf1e7ac BP |
920 | return false; |
921 | } | |
922 | } | |
ecf1e7ac BP |
923 | return true; |
924 | } | |
925 | ||
368eefac EJ |
926 | /* Sets 'dst' as the bitwise AND of wildcards in 'src1' and 'src2'. |
927 | * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded | |
928 | * in 'src1' or 'src2' or both. */ | |
b5d97350 | 929 | void |
368eefac EJ |
930 | flow_wildcards_and(struct flow_wildcards *dst, |
931 | const struct flow_wildcards *src1, | |
932 | const struct flow_wildcards *src2) | |
b5d97350 | 933 | { |
659c2346 BP |
934 | uint32_t *dst_u32 = (uint32_t *) &dst->masks; |
935 | const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; | |
936 | const uint32_t *src2_u32 = (const uint32_t *) &src2->masks; | |
937 | size_t i; | |
a79c50f3 | 938 | |
659c2346 BP |
939 | for (i = 0; i < FLOW_U32S; i++) { |
940 | dst_u32[i] = src1_u32[i] & src2_u32[i]; | |
26720e24 | 941 | } |
b5d97350 BP |
942 | } |
943 | ||
368eefac EJ |
944 | /* Sets 'dst' as the bitwise OR of wildcards in 'src1' and 'src2'. That |
945 | * is, a bit or a field is wildcarded in 'dst' if it is neither | |
946 | * wildcarded in 'src1' nor 'src2'. */ | |
947 | void | |
948 | flow_wildcards_or(struct flow_wildcards *dst, | |
949 | const struct flow_wildcards *src1, | |
950 | const struct flow_wildcards *src2) | |
951 | { | |
952 | uint32_t *dst_u32 = (uint32_t *) &dst->masks; | |
953 | const uint32_t *src1_u32 = (const uint32_t *) &src1->masks; | |
954 | const uint32_t *src2_u32 = (const uint32_t *) &src2->masks; | |
955 | size_t i; | |
956 | ||
957 | for (i = 0; i < FLOW_U32S; i++) { | |
958 | dst_u32[i] = src1_u32[i] | src2_u32[i]; | |
959 | } | |
960 | } | |
961 | ||
b5d97350 BP |
962 | /* Returns a hash of the wildcards in 'wc'. */ |
963 | uint32_t | |
1006cda6 | 964 | flow_wildcards_hash(const struct flow_wildcards *wc, uint32_t basis) |
b5d97350 | 965 | { |
ac31c5af | 966 | return flow_hash(&wc->masks, basis); |
b5d97350 BP |
967 | } |
968 | ||
969 | /* Returns true if 'a' and 'b' represent the same wildcards, false if they are | |
970 | * different. */ | |
971 | bool | |
972 | flow_wildcards_equal(const struct flow_wildcards *a, | |
973 | const struct flow_wildcards *b) | |
974 | { | |
659c2346 | 975 | return flow_equal(&a->masks, &b->masks); |
b5d97350 BP |
976 | } |
977 | ||
978 | /* Returns true if at least one bit or field is wildcarded in 'a' but not in | |
979 | * 'b', false otherwise. */ | |
980 | bool | |
981 | flow_wildcards_has_extra(const struct flow_wildcards *a, | |
982 | const struct flow_wildcards *b) | |
983 | { | |
659c2346 BP |
984 | const uint32_t *a_u32 = (const uint32_t *) &a->masks; |
985 | const uint32_t *b_u32 = (const uint32_t *) &b->masks; | |
986 | size_t i; | |
a79c50f3 | 987 | |
659c2346 BP |
988 | for (i = 0; i < FLOW_U32S; i++) { |
989 | if ((a_u32[i] & b_u32[i]) != b_u32[i]) { | |
b6c9e612 BP |
990 | return true; |
991 | } | |
992 | } | |
659c2346 BP |
993 | return false; |
994 | } | |
b6c9e612 | 995 | |
659c2346 BP |
996 | /* Returns true if 'a' and 'b' are equal, except that 0-bits (wildcarded bits) |
997 | * in 'wc' do not need to be equal in 'a' and 'b'. */ | |
998 | bool | |
999 | flow_equal_except(const struct flow *a, const struct flow *b, | |
1000 | const struct flow_wildcards *wc) | |
1001 | { | |
1002 | const uint32_t *a_u32 = (const uint32_t *) a; | |
1003 | const uint32_t *b_u32 = (const uint32_t *) b; | |
1004 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; | |
1005 | size_t i; | |
d31f1109 | 1006 | |
659c2346 BP |
1007 | for (i = 0; i < FLOW_U32S; i++) { |
1008 | if ((a_u32[i] ^ b_u32[i]) & wc_u32[i]) { | |
1009 | return false; | |
1010 | } | |
47284b1f | 1011 | } |
659c2346 | 1012 | return true; |
b5d97350 BP |
1013 | } |
1014 | ||
b6c9e612 BP |
1015 | /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. |
1016 | * (A 0-bit indicates a wildcard bit.) */ | |
1017 | void | |
1018 | flow_wildcards_set_reg_mask(struct flow_wildcards *wc, int idx, uint32_t mask) | |
1019 | { | |
26720e24 | 1020 | wc->masks.regs[idx] = mask; |
b6c9e612 | 1021 | } |
ff55ea1f | 1022 | |
79fe0f46 BP |
1023 | /* Sets the wildcard mask for register 'idx' in 'wc' to 'mask'. |
1024 | * (A 0-bit indicates a wildcard bit.) */ | |
1025 | void | |
1026 | flow_wildcards_set_xreg_mask(struct flow_wildcards *wc, int idx, uint64_t mask) | |
1027 | { | |
1028 | flow_set_xreg(&wc->masks, idx, mask); | |
1029 | } | |
1030 | ||
28a560d9 JR |
1031 | /* Calculates the 5-tuple hash from the given miniflow. |
1032 | * This returns the same value as flow_hash_5tuple for the corresponding | |
1033 | * flow. */ | |
4f150744 JR |
1034 | uint32_t |
1035 | miniflow_hash_5tuple(const struct miniflow *flow, uint32_t basis) | |
1036 | { | |
28a560d9 | 1037 | uint32_t hash = basis; |
4f150744 | 1038 | |
28a560d9 JR |
1039 | if (flow) { |
1040 | ovs_be16 dl_type = MINIFLOW_GET_BE16(flow, dl_type); | |
1041 | ||
33c6a1b9 | 1042 | hash = hash_add(hash, MINIFLOW_GET_U8(flow, nw_proto)); |
28a560d9 JR |
1043 | |
1044 | /* Separate loops for better optimization. */ | |
1045 | if (dl_type == htons(ETH_TYPE_IPV6)) { | |
1046 | uint64_t map = MINIFLOW_MAP(ipv6_src) | MINIFLOW_MAP(ipv6_dst) | |
1047 | | MINIFLOW_MAP(tp_src); /* Covers both ports */ | |
1048 | uint32_t value; | |
4f150744 | 1049 | |
28a560d9 | 1050 | MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { |
33c6a1b9 | 1051 | hash = hash_add(hash, value); |
28a560d9 JR |
1052 | } |
1053 | } else { | |
1054 | uint64_t map = MINIFLOW_MAP(nw_src) | MINIFLOW_MAP(nw_dst) | |
1055 | | MINIFLOW_MAP(tp_src); /* Covers both ports */ | |
1056 | uint32_t value; | |
4f150744 | 1057 | |
28a560d9 | 1058 | MINIFLOW_FOR_EACH_IN_MAP(value, flow, map) { |
33c6a1b9 | 1059 | hash = hash_add(hash, value); |
28a560d9 JR |
1060 | } |
1061 | } | |
33c6a1b9 | 1062 | hash = hash_finish(hash, 42); /* Arbitrary number. */ |
28a560d9 JR |
1063 | } |
1064 | return hash; | |
4f150744 JR |
1065 | } |
1066 | ||
1067 | BUILD_ASSERT_DECL(offsetof(struct flow, tp_src) + 2 | |
1068 | == offsetof(struct flow, tp_dst) && | |
1069 | offsetof(struct flow, tp_src) / 4 | |
1070 | == offsetof(struct flow, tp_dst) / 4); | |
28a560d9 JR |
1071 | BUILD_ASSERT_DECL(offsetof(struct flow, ipv6_src) + 16 |
1072 | == offsetof(struct flow, ipv6_dst)); | |
4f150744 | 1073 | |
63be20be AW |
1074 | /* Calculates the 5-tuple hash from the given flow. */ |
1075 | uint32_t | |
1076 | flow_hash_5tuple(const struct flow *flow, uint32_t basis) | |
1077 | { | |
28a560d9 | 1078 | uint32_t hash = basis; |
63be20be | 1079 | |
28a560d9 JR |
1080 | if (flow) { |
1081 | const uint32_t *flow_u32 = (const uint32_t *)flow; | |
1082 | ||
33c6a1b9 | 1083 | hash = hash_add(hash, flow->nw_proto); |
28a560d9 JR |
1084 | |
1085 | if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
1086 | int ofs = offsetof(struct flow, ipv6_src) / 4; | |
1087 | int end = ofs + 2 * sizeof flow->ipv6_src / 4; | |
63be20be | 1088 | |
28a560d9 | 1089 | while (ofs < end) { |
33c6a1b9 | 1090 | hash = hash_add(hash, flow_u32[ofs++]); |
28a560d9 JR |
1091 | } |
1092 | } else { | |
33c6a1b9 JR |
1093 | hash = hash_add(hash, (OVS_FORCE uint32_t) flow->nw_src); |
1094 | hash = hash_add(hash, (OVS_FORCE uint32_t) flow->nw_dst); | |
28a560d9 | 1095 | } |
33c6a1b9 | 1096 | hash = hash_add(hash, flow_u32[offsetof(struct flow, tp_src) / 4]); |
63be20be | 1097 | |
33c6a1b9 | 1098 | hash = hash_finish(hash, 42); /* Arbitrary number. */ |
28a560d9 JR |
1099 | } |
1100 | return hash; | |
63be20be AW |
1101 | } |
1102 | ||
ff55ea1f EJ |
1103 | /* Hashes 'flow' based on its L2 through L4 protocol information. */ |
1104 | uint32_t | |
1105 | flow_hash_symmetric_l4(const struct flow *flow, uint32_t basis) | |
1106 | { | |
1107 | struct { | |
d31f1109 JP |
1108 | union { |
1109 | ovs_be32 ipv4_addr; | |
1110 | struct in6_addr ipv6_addr; | |
1111 | }; | |
ff55ea1f EJ |
1112 | ovs_be16 eth_type; |
1113 | ovs_be16 vlan_tci; | |
5b909cbb | 1114 | ovs_be16 tp_port; |
ff55ea1f EJ |
1115 | uint8_t eth_addr[ETH_ADDR_LEN]; |
1116 | uint8_t ip_proto; | |
1117 | } fields; | |
1118 | ||
1119 | int i; | |
1120 | ||
1121 | memset(&fields, 0, sizeof fields); | |
1122 | for (i = 0; i < ETH_ADDR_LEN; i++) { | |
1123 | fields.eth_addr[i] = flow->dl_src[i] ^ flow->dl_dst[i]; | |
1124 | } | |
1125 | fields.vlan_tci = flow->vlan_tci & htons(VLAN_VID_MASK); | |
1126 | fields.eth_type = flow->dl_type; | |
3e3eda95 EJ |
1127 | |
1128 | /* UDP source and destination port are not taken into account because they | |
1129 | * will not necessarily be symmetric in a bidirectional flow. */ | |
ff55ea1f | 1130 | if (fields.eth_type == htons(ETH_TYPE_IP)) { |
d31f1109 JP |
1131 | fields.ipv4_addr = flow->nw_src ^ flow->nw_dst; |
1132 | fields.ip_proto = flow->nw_proto; | |
c6bcb685 | 1133 | if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { |
5b909cbb | 1134 | fields.tp_port = flow->tp_src ^ flow->tp_dst; |
d31f1109 JP |
1135 | } |
1136 | } else if (fields.eth_type == htons(ETH_TYPE_IPV6)) { | |
1137 | const uint8_t *a = &flow->ipv6_src.s6_addr[0]; | |
1138 | const uint8_t *b = &flow->ipv6_dst.s6_addr[0]; | |
1139 | uint8_t *ipv6_addr = &fields.ipv6_addr.s6_addr[0]; | |
1140 | ||
1141 | for (i=0; i<16; i++) { | |
1142 | ipv6_addr[i] = a[i] ^ b[i]; | |
1143 | } | |
ff55ea1f | 1144 | fields.ip_proto = flow->nw_proto; |
c6bcb685 | 1145 | if (fields.ip_proto == IPPROTO_TCP || fields.ip_proto == IPPROTO_SCTP) { |
5b909cbb | 1146 | fields.tp_port = flow->tp_src ^ flow->tp_dst; |
ff55ea1f | 1147 | } |
ff55ea1f | 1148 | } |
c49d1dd1 | 1149 | return jhash_bytes(&fields, sizeof fields, basis); |
ff55ea1f | 1150 | } |
520e9a2a | 1151 | |
94639963 JR |
1152 | /* Initialize a flow with random fields that matter for nx_hash_fields. */ |
1153 | void | |
1154 | flow_random_hash_fields(struct flow *flow) | |
1155 | { | |
1156 | uint16_t rnd = random_uint16(); | |
1157 | ||
1158 | /* Initialize to all zeros. */ | |
1159 | memset(flow, 0, sizeof *flow); | |
1160 | ||
1161 | eth_addr_random(flow->dl_src); | |
1162 | eth_addr_random(flow->dl_dst); | |
1163 | ||
1164 | flow->vlan_tci = (OVS_FORCE ovs_be16) (random_uint16() & VLAN_VID_MASK); | |
1165 | ||
1166 | /* Make most of the random flows IPv4, some IPv6, and rest random. */ | |
1167 | flow->dl_type = rnd < 0x8000 ? htons(ETH_TYPE_IP) : | |
1168 | rnd < 0xc000 ? htons(ETH_TYPE_IPV6) : (OVS_FORCE ovs_be16)rnd; | |
1169 | ||
1170 | if (dl_type_is_ip_any(flow->dl_type)) { | |
1171 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
1172 | flow->nw_src = (OVS_FORCE ovs_be32)random_uint32(); | |
1173 | flow->nw_dst = (OVS_FORCE ovs_be32)random_uint32(); | |
1174 | } else { | |
1175 | random_bytes(&flow->ipv6_src, sizeof flow->ipv6_src); | |
1176 | random_bytes(&flow->ipv6_dst, sizeof flow->ipv6_dst); | |
1177 | } | |
1178 | /* Make most of IP flows TCP, some UDP or SCTP, and rest random. */ | |
1179 | rnd = random_uint16(); | |
1180 | flow->nw_proto = rnd < 0x8000 ? IPPROTO_TCP : | |
1181 | rnd < 0xc000 ? IPPROTO_UDP : | |
1182 | rnd < 0xd000 ? IPPROTO_SCTP : (uint8_t)rnd; | |
1183 | if (flow->nw_proto == IPPROTO_TCP || | |
1184 | flow->nw_proto == IPPROTO_UDP || | |
1185 | flow->nw_proto == IPPROTO_SCTP) { | |
1186 | flow->tp_src = (OVS_FORCE ovs_be16)random_uint16(); | |
1187 | flow->tp_dst = (OVS_FORCE ovs_be16)random_uint16(); | |
1188 | } | |
1189 | } | |
1190 | } | |
1191 | ||
bcd2633a JP |
1192 | /* Masks the fields in 'wc' that are used by the flow hash 'fields'. */ |
1193 | void | |
6cdd5145 JP |
1194 | flow_mask_hash_fields(const struct flow *flow, struct flow_wildcards *wc, |
1195 | enum nx_hash_fields fields) | |
bcd2633a JP |
1196 | { |
1197 | switch (fields) { | |
1198 | case NX_HASH_FIELDS_ETH_SRC: | |
1199 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); | |
1200 | break; | |
1201 | ||
1202 | case NX_HASH_FIELDS_SYMMETRIC_L4: | |
1203 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); | |
1204 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
6cdd5145 JP |
1205 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
1206 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1207 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
7f8a65ca | 1208 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { |
6cdd5145 JP |
1209 | memset(&wc->masks.ipv6_src, 0xff, sizeof wc->masks.ipv6_src); |
1210 | memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); | |
1211 | } | |
1212 | if (is_ip_any(flow)) { | |
1213 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
d8d9c698 | 1214 | flow_unwildcard_tp_ports(flow, wc); |
6cdd5145 | 1215 | } |
1dd35f8a | 1216 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
bcd2633a JP |
1217 | break; |
1218 | ||
1219 | default: | |
428b2edd | 1220 | OVS_NOT_REACHED(); |
bcd2633a JP |
1221 | } |
1222 | } | |
1223 | ||
520e9a2a EJ |
1224 | /* Hashes the portions of 'flow' designated by 'fields'. */ |
1225 | uint32_t | |
1226 | flow_hash_fields(const struct flow *flow, enum nx_hash_fields fields, | |
1227 | uint16_t basis) | |
1228 | { | |
1229 | switch (fields) { | |
1230 | ||
1231 | case NX_HASH_FIELDS_ETH_SRC: | |
c49d1dd1 | 1232 | return jhash_bytes(flow->dl_src, sizeof flow->dl_src, basis); |
520e9a2a EJ |
1233 | |
1234 | case NX_HASH_FIELDS_SYMMETRIC_L4: | |
1235 | return flow_hash_symmetric_l4(flow, basis); | |
1236 | } | |
1237 | ||
428b2edd | 1238 | OVS_NOT_REACHED(); |
520e9a2a EJ |
1239 | } |
1240 | ||
1241 | /* Returns a string representation of 'fields'. */ | |
1242 | const char * | |
1243 | flow_hash_fields_to_str(enum nx_hash_fields fields) | |
1244 | { | |
1245 | switch (fields) { | |
1246 | case NX_HASH_FIELDS_ETH_SRC: return "eth_src"; | |
1247 | case NX_HASH_FIELDS_SYMMETRIC_L4: return "symmetric_l4"; | |
1248 | default: return "<unknown>"; | |
1249 | } | |
1250 | } | |
1251 | ||
1252 | /* Returns true if the value of 'fields' is supported. Otherwise false. */ | |
1253 | bool | |
1254 | flow_hash_fields_valid(enum nx_hash_fields fields) | |
1255 | { | |
1256 | return fields == NX_HASH_FIELDS_ETH_SRC | |
1257 | || fields == NX_HASH_FIELDS_SYMMETRIC_L4; | |
1258 | } | |
8b3b8dd1 | 1259 | |
368eefac EJ |
1260 | /* Returns a hash value for the bits of 'flow' that are active based on |
1261 | * 'wc', given 'basis'. */ | |
1262 | uint32_t | |
1263 | flow_hash_in_wildcards(const struct flow *flow, | |
1264 | const struct flow_wildcards *wc, uint32_t basis) | |
1265 | { | |
1266 | const uint32_t *wc_u32 = (const uint32_t *) &wc->masks; | |
1267 | const uint32_t *flow_u32 = (const uint32_t *) flow; | |
1268 | uint32_t hash; | |
1269 | size_t i; | |
1270 | ||
1271 | hash = basis; | |
1272 | for (i = 0; i < FLOW_U32S; i++) { | |
33c6a1b9 | 1273 | hash = hash_add(hash, flow_u32[i] & wc_u32[i]); |
368eefac | 1274 | } |
33c6a1b9 | 1275 | return hash_finish(hash, 4 * FLOW_U32S); |
368eefac EJ |
1276 | } |
1277 | ||
3719455c BP |
1278 | /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an |
1279 | * OpenFlow 1.0 "dl_vlan" value: | |
1280 | * | |
1281 | * - If it is in the range 0...4095, 'flow->vlan_tci' is set to match | |
1282 | * that VLAN. Any existing PCP match is unchanged (it becomes 0 if | |
1283 | * 'flow' previously matched packets without a VLAN header). | |
1284 | * | |
1285 | * - If it is OFP_VLAN_NONE, 'flow->vlan_tci' is set to match a packet | |
1286 | * without a VLAN tag. | |
1287 | * | |
1288 | * - Other values of 'vid' should not be used. */ | |
1289 | void | |
fb0451d9 | 1290 | flow_set_dl_vlan(struct flow *flow, ovs_be16 vid) |
3719455c | 1291 | { |
0c436519 | 1292 | if (vid == htons(OFP10_VLAN_NONE)) { |
3719455c BP |
1293 | flow->vlan_tci = htons(0); |
1294 | } else { | |
1295 | vid &= htons(VLAN_VID_MASK); | |
1296 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
1297 | flow->vlan_tci |= htons(VLAN_CFI) | vid; | |
1298 | } | |
1299 | } | |
1300 | ||
cc34bc8c BP |
1301 | /* Sets the VLAN VID that 'flow' matches to 'vid', which is interpreted as an |
1302 | * OpenFlow 1.2 "vlan_vid" value, that is, the low 13 bits of 'vlan_tci' (VID | |
1303 | * plus CFI). */ | |
1304 | void | |
1305 | flow_set_vlan_vid(struct flow *flow, ovs_be16 vid) | |
1306 | { | |
1307 | ovs_be16 mask = htons(VLAN_VID_MASK | VLAN_CFI); | |
1308 | flow->vlan_tci &= ~mask; | |
1309 | flow->vlan_tci |= vid & mask; | |
1310 | } | |
1311 | ||
3719455c BP |
1312 | /* Sets the VLAN PCP that 'flow' matches to 'pcp', which should be in the |
1313 | * range 0...7. | |
1314 | * | |
1315 | * This function has no effect on the VLAN ID that 'flow' matches. | |
1316 | * | |
1317 | * After calling this function, 'flow' will not match packets without a VLAN | |
1318 | * header. */ | |
1319 | void | |
1320 | flow_set_vlan_pcp(struct flow *flow, uint8_t pcp) | |
1321 | { | |
1322 | pcp &= 0x07; | |
1323 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
1324 | flow->vlan_tci |= htons((pcp << VLAN_PCP_SHIFT) | VLAN_CFI); | |
1325 | } | |
1326 | ||
8bfd0fda BP |
1327 | /* Returns the number of MPLS LSEs present in 'flow' |
1328 | * | |
1329 | * Returns 0 if the 'dl_type' of 'flow' is not an MPLS ethernet type. | |
1330 | * Otherwise traverses 'flow''s MPLS label stack stopping at the | |
1331 | * first entry that has the BoS bit set. If no such entry exists then | |
1332 | * the maximum number of LSEs that can be stored in 'flow' is returned. | |
1333 | */ | |
1334 | int | |
1335 | flow_count_mpls_labels(const struct flow *flow, struct flow_wildcards *wc) | |
1336 | { | |
1337 | if (wc) { | |
1338 | wc->masks.dl_type = OVS_BE16_MAX; | |
1339 | } | |
1340 | if (eth_type_mpls(flow->dl_type)) { | |
1341 | int i; | |
1342 | int len = FLOW_MAX_MPLS_LABELS; | |
1343 | ||
1344 | for (i = 0; i < len; i++) { | |
1345 | if (wc) { | |
1346 | wc->masks.mpls_lse[i] |= htonl(MPLS_BOS_MASK); | |
1347 | } | |
1348 | if (flow->mpls_lse[i] & htonl(MPLS_BOS_MASK)) { | |
1349 | return i + 1; | |
1350 | } | |
1351 | } | |
1352 | ||
1353 | return len; | |
1354 | } else { | |
1355 | return 0; | |
1356 | } | |
1357 | } | |
1358 | ||
1359 | /* Returns the number consecutive of MPLS LSEs, starting at the | |
1360 | * innermost LSE, that are common in 'a' and 'b'. | |
1361 | * | |
1362 | * 'an' must be flow_count_mpls_labels(a). | |
1363 | * 'bn' must be flow_count_mpls_labels(b). | |
1364 | */ | |
1365 | int | |
1366 | flow_count_common_mpls_labels(const struct flow *a, int an, | |
1367 | const struct flow *b, int bn, | |
1368 | struct flow_wildcards *wc) | |
1369 | { | |
1370 | int min_n = MIN(an, bn); | |
1371 | if (min_n == 0) { | |
1372 | return 0; | |
1373 | } else { | |
1374 | int common_n = 0; | |
1375 | int a_last = an - 1; | |
1376 | int b_last = bn - 1; | |
1377 | int i; | |
1378 | ||
1379 | for (i = 0; i < min_n; i++) { | |
1380 | if (wc) { | |
1381 | wc->masks.mpls_lse[a_last - i] = OVS_BE32_MAX; | |
1382 | wc->masks.mpls_lse[b_last - i] = OVS_BE32_MAX; | |
1383 | } | |
1384 | if (a->mpls_lse[a_last - i] != b->mpls_lse[b_last - i]) { | |
1385 | break; | |
1386 | } else { | |
1387 | common_n++; | |
1388 | } | |
1389 | } | |
1390 | ||
1391 | return common_n; | |
1392 | } | |
1393 | } | |
1394 | ||
1395 | /* Adds a new outermost MPLS label to 'flow' and changes 'flow''s Ethernet type | |
1396 | * to 'mpls_eth_type', which must be an MPLS Ethertype. | |
1397 | * | |
1398 | * If the new label is the first MPLS label in 'flow', it is generated as; | |
1399 | * | |
1400 | * - label: 2, if 'flow' is IPv6, otherwise 0. | |
1401 | * | |
1402 | * - TTL: IPv4 or IPv6 TTL, if present and nonzero, otherwise 64. | |
1403 | * | |
1404 | * - TC: IPv4 or IPv6 TOS, if present, otherwise 0. | |
1405 | * | |
1406 | * - BoS: 1. | |
1407 | * | |
1408 | * If the new label is the second or label MPLS label in 'flow', it is | |
1409 | * generated as; | |
1410 | * | |
368fb7e6 | 1411 | * - label: Copied from outer label. |
8bfd0fda BP |
1412 | * |
1413 | * - TTL: Copied from outer label. | |
1414 | * | |
1415 | * - TC: Copied from outer label. | |
1416 | * | |
1417 | * - BoS: 0. | |
1418 | * | |
1419 | * 'n' must be flow_count_mpls_labels(flow). 'n' must be less than | |
1420 | * FLOW_MAX_MPLS_LABELS (because otherwise flow->mpls_lse[] would overflow). | |
1421 | */ | |
1422 | void | |
1423 | flow_push_mpls(struct flow *flow, int n, ovs_be16 mpls_eth_type, | |
1424 | struct flow_wildcards *wc) | |
1425 | { | |
1426 | ovs_assert(eth_type_mpls(mpls_eth_type)); | |
1427 | ovs_assert(n < FLOW_MAX_MPLS_LABELS); | |
1428 | ||
1429 | memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); | |
1430 | if (n) { | |
1431 | int i; | |
1432 | ||
1433 | for (i = n; i >= 1; i--) { | |
1434 | flow->mpls_lse[i] = flow->mpls_lse[i - 1]; | |
1435 | } | |
1436 | flow->mpls_lse[0] = (flow->mpls_lse[1] | |
368fb7e6 | 1437 | & htonl(~MPLS_BOS_MASK)); |
8bfd0fda BP |
1438 | } else { |
1439 | int label = 0; /* IPv4 Explicit Null. */ | |
1440 | int tc = 0; | |
1441 | int ttl = 64; | |
1442 | ||
1443 | if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
1444 | label = 2; | |
1445 | } | |
1446 | ||
1447 | if (is_ip_any(flow)) { | |
1448 | tc = (flow->nw_tos & IP_DSCP_MASK) >> 2; | |
1449 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
1450 | ||
1451 | if (flow->nw_ttl) { | |
1452 | ttl = flow->nw_ttl; | |
1453 | } | |
1454 | wc->masks.nw_ttl = 0xff; | |
1455 | } | |
1456 | ||
1457 | flow->mpls_lse[0] = set_mpls_lse_values(ttl, tc, 1, htonl(label)); | |
1458 | ||
1459 | /* Clear all L3 and L4 fields. */ | |
0e612675 | 1460 | BUILD_ASSERT(FLOW_WC_SEQ == 27); |
8bfd0fda BP |
1461 | memset((char *) flow + FLOW_SEGMENT_2_ENDS_AT, 0, |
1462 | sizeof(struct flow) - FLOW_SEGMENT_2_ENDS_AT); | |
1463 | } | |
1464 | flow->dl_type = mpls_eth_type; | |
1465 | } | |
1466 | ||
1467 | /* Tries to remove the outermost MPLS label from 'flow'. Returns true if | |
1468 | * successful, false otherwise. On success, sets 'flow''s Ethernet type to | |
1469 | * 'eth_type'. | |
1470 | * | |
1471 | * 'n' must be flow_count_mpls_labels(flow). */ | |
1472 | bool | |
1473 | flow_pop_mpls(struct flow *flow, int n, ovs_be16 eth_type, | |
1474 | struct flow_wildcards *wc) | |
1475 | { | |
1476 | int i; | |
1477 | ||
1478 | if (n == 0) { | |
1479 | /* Nothing to pop. */ | |
1480 | return false; | |
1481 | } else if (n == FLOW_MAX_MPLS_LABELS | |
1482 | && !(flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK))) { | |
1483 | /* Can't pop because we don't know what to fill in mpls_lse[n - 1]. */ | |
1484 | return false; | |
1485 | } | |
1486 | ||
1487 | memset(wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse); | |
1488 | for (i = 1; i < n; i++) { | |
1489 | flow->mpls_lse[i - 1] = flow->mpls_lse[i]; | |
1490 | } | |
1491 | flow->mpls_lse[n - 1] = 0; | |
1492 | flow->dl_type = eth_type; | |
1493 | return true; | |
1494 | } | |
1495 | ||
b02475c5 SH |
1496 | /* Sets the MPLS Label that 'flow' matches to 'label', which is interpreted |
1497 | * as an OpenFlow 1.1 "mpls_label" value. */ | |
1498 | void | |
8bfd0fda | 1499 | flow_set_mpls_label(struct flow *flow, int idx, ovs_be32 label) |
b02475c5 | 1500 | { |
8bfd0fda | 1501 | set_mpls_lse_label(&flow->mpls_lse[idx], label); |
b02475c5 SH |
1502 | } |
1503 | ||
b676167a SH |
1504 | /* Sets the MPLS TTL that 'flow' matches to 'ttl', which should be in the |
1505 | * range 0...255. */ | |
1506 | void | |
8bfd0fda | 1507 | flow_set_mpls_ttl(struct flow *flow, int idx, uint8_t ttl) |
b676167a | 1508 | { |
8bfd0fda | 1509 | set_mpls_lse_ttl(&flow->mpls_lse[idx], ttl); |
b676167a SH |
1510 | } |
1511 | ||
b02475c5 SH |
1512 | /* Sets the MPLS TC that 'flow' matches to 'tc', which should be in the |
1513 | * range 0...7. */ | |
1514 | void | |
8bfd0fda | 1515 | flow_set_mpls_tc(struct flow *flow, int idx, uint8_t tc) |
b02475c5 | 1516 | { |
8bfd0fda | 1517 | set_mpls_lse_tc(&flow->mpls_lse[idx], tc); |
b02475c5 SH |
1518 | } |
1519 | ||
1520 | /* Sets the MPLS BOS bit that 'flow' matches to which should be 0 or 1. */ | |
1521 | void | |
8bfd0fda | 1522 | flow_set_mpls_bos(struct flow *flow, int idx, uint8_t bos) |
b02475c5 | 1523 | { |
8bfd0fda | 1524 | set_mpls_lse_bos(&flow->mpls_lse[idx], bos); |
b02475c5 SH |
1525 | } |
1526 | ||
8bfd0fda BP |
1527 | /* Sets the entire MPLS LSE. */ |
1528 | void | |
1529 | flow_set_mpls_lse(struct flow *flow, int idx, ovs_be32 lse) | |
1530 | { | |
1531 | flow->mpls_lse[idx] = lse; | |
1532 | } | |
52105b67 | 1533 | |
437d0d22 | 1534 | static size_t |
52105b67 JR |
1535 | flow_compose_l4(struct ofpbuf *b, const struct flow *flow) |
1536 | { | |
437d0d22 JR |
1537 | size_t l4_len = 0; |
1538 | ||
52105b67 JR |
1539 | if (!(flow->nw_frag & FLOW_NW_FRAG_ANY) |
1540 | || !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { | |
1541 | if (flow->nw_proto == IPPROTO_TCP) { | |
1542 | struct tcp_header *tcp; | |
1543 | ||
437d0d22 JR |
1544 | l4_len = sizeof *tcp; |
1545 | tcp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1546 | tcp->tcp_src = flow->tp_src; |
1547 | tcp->tcp_dst = flow->tp_dst; | |
1548 | tcp->tcp_ctl = TCP_CTL(ntohs(flow->tcp_flags), 5); | |
52105b67 JR |
1549 | } else if (flow->nw_proto == IPPROTO_UDP) { |
1550 | struct udp_header *udp; | |
1551 | ||
437d0d22 JR |
1552 | l4_len = sizeof *udp; |
1553 | udp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1554 | udp->udp_src = flow->tp_src; |
1555 | udp->udp_dst = flow->tp_dst; | |
52105b67 JR |
1556 | } else if (flow->nw_proto == IPPROTO_SCTP) { |
1557 | struct sctp_header *sctp; | |
1558 | ||
437d0d22 JR |
1559 | l4_len = sizeof *sctp; |
1560 | sctp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1561 | sctp->sctp_src = flow->tp_src; |
1562 | sctp->sctp_dst = flow->tp_dst; | |
52105b67 JR |
1563 | } else if (flow->nw_proto == IPPROTO_ICMP) { |
1564 | struct icmp_header *icmp; | |
1565 | ||
437d0d22 JR |
1566 | l4_len = sizeof *icmp; |
1567 | icmp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1568 | icmp->icmp_type = ntohs(flow->tp_src); |
1569 | icmp->icmp_code = ntohs(flow->tp_dst); | |
1570 | icmp->icmp_csum = csum(icmp, ICMP_HEADER_LEN); | |
0e612675 FL |
1571 | } else if (flow->nw_proto == IPPROTO_IGMP) { |
1572 | struct igmp_header *igmp; | |
1573 | ||
1574 | l4_len = sizeof *igmp; | |
1575 | igmp = ofpbuf_put_zeros(b, l4_len); | |
1576 | igmp->igmp_type = ntohs(flow->tp_src); | |
1577 | igmp->igmp_code = ntohs(flow->tp_dst); | |
1578 | put_16aligned_be32(&igmp->group, flow->igmp_group_ip4); | |
1579 | igmp->igmp_csum = csum(igmp, IGMP_HEADER_LEN); | |
52105b67 JR |
1580 | } else if (flow->nw_proto == IPPROTO_ICMPV6) { |
1581 | struct icmp6_hdr *icmp; | |
1582 | ||
437d0d22 JR |
1583 | l4_len = sizeof *icmp; |
1584 | icmp = ofpbuf_put_zeros(b, l4_len); | |
52105b67 JR |
1585 | icmp->icmp6_type = ntohs(flow->tp_src); |
1586 | icmp->icmp6_code = ntohs(flow->tp_dst); | |
1587 | ||
1588 | if (icmp->icmp6_code == 0 && | |
1589 | (icmp->icmp6_type == ND_NEIGHBOR_SOLICIT || | |
1590 | icmp->icmp6_type == ND_NEIGHBOR_ADVERT)) { | |
1591 | struct in6_addr *nd_target; | |
1592 | struct nd_opt_hdr *nd_opt; | |
1593 | ||
437d0d22 | 1594 | l4_len += sizeof *nd_target; |
52105b67 JR |
1595 | nd_target = ofpbuf_put_zeros(b, sizeof *nd_target); |
1596 | *nd_target = flow->nd_target; | |
1597 | ||
1598 | if (!eth_addr_is_zero(flow->arp_sha)) { | |
437d0d22 | 1599 | l4_len += 8; |
52105b67 JR |
1600 | nd_opt = ofpbuf_put_zeros(b, 8); |
1601 | nd_opt->nd_opt_len = 1; | |
1602 | nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; | |
1603 | memcpy(nd_opt + 1, flow->arp_sha, ETH_ADDR_LEN); | |
1604 | } | |
1605 | if (!eth_addr_is_zero(flow->arp_tha)) { | |
437d0d22 | 1606 | l4_len += 8; |
52105b67 JR |
1607 | nd_opt = ofpbuf_put_zeros(b, 8); |
1608 | nd_opt->nd_opt_len = 1; | |
1609 | nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; | |
1610 | memcpy(nd_opt + 1, flow->arp_tha, ETH_ADDR_LEN); | |
1611 | } | |
1612 | } | |
1613 | icmp->icmp6_cksum = (OVS_FORCE uint16_t) | |
1614 | csum(icmp, (char *)ofpbuf_tail(b) - (char *)icmp); | |
52105b67 JR |
1615 | } |
1616 | } | |
437d0d22 | 1617 | return l4_len; |
52105b67 JR |
1618 | } |
1619 | ||
8b3b8dd1 BP |
1620 | /* Puts into 'b' a packet that flow_extract() would parse as having the given |
1621 | * 'flow'. | |
1622 | * | |
1623 | * (This is useful only for testing, obviously, and the packet isn't really | |
dc5a7ce7 | 1624 | * valid. It hasn't got some checksums filled in, for one, and lots of fields |
8b3b8dd1 BP |
1625 | * are just zeroed.) */ |
1626 | void | |
1627 | flow_compose(struct ofpbuf *b, const struct flow *flow) | |
1628 | { | |
437d0d22 JR |
1629 | size_t l4_len; |
1630 | ||
52105b67 | 1631 | /* eth_compose() sets l3 pointer and makes sure it is 32-bit aligned. */ |
8b3b8dd1 BP |
1632 | eth_compose(b, flow->dl_dst, flow->dl_src, ntohs(flow->dl_type), 0); |
1633 | if (flow->dl_type == htons(FLOW_DL_TYPE_NONE)) { | |
cf3b7538 | 1634 | struct eth_header *eth = ofpbuf_l2(b); |
1f317cb5 | 1635 | eth->eth_type = htons(ofpbuf_size(b)); |
8b3b8dd1 BP |
1636 | return; |
1637 | } | |
1638 | ||
1639 | if (flow->vlan_tci & htons(VLAN_CFI)) { | |
1bf02876 | 1640 | eth_push_vlan(b, htons(ETH_TYPE_VLAN), flow->vlan_tci); |
8b3b8dd1 BP |
1641 | } |
1642 | ||
cff78c88 | 1643 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
8b3b8dd1 BP |
1644 | struct ip_header *ip; |
1645 | ||
52105b67 | 1646 | ip = ofpbuf_put_zeros(b, sizeof *ip); |
8b3b8dd1 | 1647 | ip->ip_ihl_ver = IP_IHL_VER(5, 4); |
eadef313 | 1648 | ip->ip_tos = flow->nw_tos; |
aabf5352 | 1649 | ip->ip_ttl = flow->nw_ttl; |
8b3b8dd1 | 1650 | ip->ip_proto = flow->nw_proto; |
7c457c33 BP |
1651 | put_16aligned_be32(&ip->ip_src, flow->nw_src); |
1652 | put_16aligned_be32(&ip->ip_dst, flow->nw_dst); | |
8b3b8dd1 | 1653 | |
eadef313 | 1654 | if (flow->nw_frag & FLOW_NW_FRAG_ANY) { |
7257b535 | 1655 | ip->ip_frag_off |= htons(IP_MORE_FRAGMENTS); |
eadef313 | 1656 | if (flow->nw_frag & FLOW_NW_FRAG_LATER) { |
7257b535 BP |
1657 | ip->ip_frag_off |= htons(100); |
1658 | } | |
1659 | } | |
df9b6612 | 1660 | |
437d0d22 | 1661 | ofpbuf_set_l4(b, ofpbuf_tail(b)); |
52105b67 | 1662 | |
437d0d22 | 1663 | l4_len = flow_compose_l4(b, flow); |
52105b67 | 1664 | |
1c98d0ad | 1665 | ip = ofpbuf_l3(b); |
437d0d22 | 1666 | ip->ip_tot_len = htons(b->l4_ofs - b->l3_ofs + l4_len); |
dc5a7ce7 | 1667 | ip->ip_csum = csum(ip, sizeof *ip); |
cff78c88 | 1668 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { |
52105b67 JR |
1669 | struct ovs_16aligned_ip6_hdr *nh; |
1670 | ||
1671 | nh = ofpbuf_put_zeros(b, sizeof *nh); | |
1672 | put_16aligned_be32(&nh->ip6_flow, htonl(6 << 28) | | |
1673 | htonl(flow->nw_tos << 20) | flow->ipv6_label); | |
1674 | nh->ip6_hlim = flow->nw_ttl; | |
1675 | nh->ip6_nxt = flow->nw_proto; | |
1676 | ||
1677 | memcpy(&nh->ip6_src, &flow->ipv6_src, sizeof(nh->ip6_src)); | |
1678 | memcpy(&nh->ip6_dst, &flow->ipv6_dst, sizeof(nh->ip6_dst)); | |
1679 | ||
437d0d22 | 1680 | ofpbuf_set_l4(b, ofpbuf_tail(b)); |
52105b67 | 1681 | |
437d0d22 | 1682 | l4_len = flow_compose_l4(b, flow); |
52105b67 | 1683 | |
1c98d0ad | 1684 | nh = ofpbuf_l3(b); |
437d0d22 | 1685 | nh->ip6_plen = htons(l4_len); |
cff78c88 SH |
1686 | } else if (flow->dl_type == htons(ETH_TYPE_ARP) || |
1687 | flow->dl_type == htons(ETH_TYPE_RARP)) { | |
8b3b8dd1 BP |
1688 | struct arp_eth_header *arp; |
1689 | ||
437d0d22 JR |
1690 | arp = ofpbuf_put_zeros(b, sizeof *arp); |
1691 | ofpbuf_set_l3(b, arp); | |
8b3b8dd1 BP |
1692 | arp->ar_hrd = htons(1); |
1693 | arp->ar_pro = htons(ETH_TYPE_IP); | |
1694 | arp->ar_hln = ETH_ADDR_LEN; | |
1695 | arp->ar_pln = 4; | |
1696 | arp->ar_op = htons(flow->nw_proto); | |
1697 | ||
1698 | if (flow->nw_proto == ARP_OP_REQUEST || | |
1699 | flow->nw_proto == ARP_OP_REPLY) { | |
7c457c33 BP |
1700 | put_16aligned_be32(&arp->ar_spa, flow->nw_src); |
1701 | put_16aligned_be32(&arp->ar_tpa, flow->nw_dst); | |
8b3b8dd1 BP |
1702 | memcpy(arp->ar_sha, flow->arp_sha, ETH_ADDR_LEN); |
1703 | memcpy(arp->ar_tha, flow->arp_tha, ETH_ADDR_LEN); | |
1704 | } | |
1705 | } | |
b02475c5 SH |
1706 | |
1707 | if (eth_type_mpls(flow->dl_type)) { | |
8bfd0fda BP |
1708 | int n; |
1709 | ||
437d0d22 | 1710 | b->l2_5_ofs = b->l3_ofs; |
8bfd0fda BP |
1711 | for (n = 1; n < FLOW_MAX_MPLS_LABELS; n++) { |
1712 | if (flow->mpls_lse[n - 1] & htonl(MPLS_BOS_MASK)) { | |
1713 | break; | |
1714 | } | |
1715 | } | |
1716 | while (n > 0) { | |
1717 | push_mpls(b, flow->dl_type, flow->mpls_lse[--n]); | |
1718 | } | |
b02475c5 | 1719 | } |
8b3b8dd1 | 1720 | } |
5cb7a798 BP |
1721 | \f |
1722 | /* Compressed flow. */ | |
1723 | ||
1724 | static int | |
1725 | miniflow_n_values(const struct miniflow *flow) | |
1726 | { | |
fb9aefa3 | 1727 | return count_1bits(flow->map); |
5cb7a798 BP |
1728 | } |
1729 | ||
1730 | static uint32_t * | |
1731 | miniflow_alloc_values(struct miniflow *flow, int n) | |
1732 | { | |
3016f3e4 JR |
1733 | int size = MINIFLOW_VALUES_SIZE(n); |
1734 | ||
1735 | if (size <= sizeof flow->inline_values) { | |
27bbe15d | 1736 | flow->values_inline = true; |
5cb7a798 BP |
1737 | return flow->inline_values; |
1738 | } else { | |
1739 | COVERAGE_INC(miniflow_malloc); | |
27bbe15d | 1740 | flow->values_inline = false; |
3016f3e4 | 1741 | flow->offline_values = xmalloc(size); |
27bbe15d | 1742 | return flow->offline_values; |
5cb7a798 BP |
1743 | } |
1744 | } | |
1745 | ||
df40c152 BP |
1746 | /* Completes an initialization of 'dst' as a miniflow copy of 'src' begun by |
1747 | * the caller. The caller must have already initialized 'dst->map' properly | |
13751fd8 JR |
1748 | * to indicate the significant uint32_t elements of 'src'. 'n' must be the |
1749 | * number of 1-bits in 'dst->map'. | |
1750 | * | |
1751 | * Normally the significant elements are the ones that are non-zero. However, | |
1752 | * when a miniflow is initialized from a (mini)mask, the values can be zeroes, | |
1753 | * so that the flow and mask always have the same maps. | |
df40c152 | 1754 | * |
27bbe15d | 1755 | * This function initializes values (either inline if possible or with |
13751fd8 JR |
1756 | * malloc() otherwise) and copies the uint32_t elements of 'src' indicated by |
1757 | * 'dst->map' into it. */ | |
df40c152 BP |
1758 | static void |
1759 | miniflow_init__(struct miniflow *dst, const struct flow *src, int n) | |
1760 | { | |
1761 | const uint32_t *src_u32 = (const uint32_t *) src; | |
27bbe15d | 1762 | uint32_t *dst_u32 = miniflow_alloc_values(dst, n); |
080e28d0 | 1763 | uint64_t map; |
df40c152 | 1764 | |
080e28d0 | 1765 | for (map = dst->map; map; map = zero_rightmost_1bit(map)) { |
27bbe15d | 1766 | *dst_u32++ = src_u32[raw_ctz(map)]; |
df40c152 BP |
1767 | } |
1768 | } | |
1769 | ||
5cb7a798 | 1770 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' |
27bbe15d JR |
1771 | * with miniflow_destroy(). |
1772 | * Always allocates offline storage. */ | |
5cb7a798 BP |
1773 | void |
1774 | miniflow_init(struct miniflow *dst, const struct flow *src) | |
1775 | { | |
1776 | const uint32_t *src_u32 = (const uint32_t *) src; | |
5cb7a798 BP |
1777 | unsigned int i; |
1778 | int n; | |
1779 | ||
1780 | /* Initialize dst->map, counting the number of nonzero elements. */ | |
1781 | n = 0; | |
080e28d0 JR |
1782 | dst->map = 0; |
1783 | ||
5cb7a798 BP |
1784 | for (i = 0; i < FLOW_U32S; i++) { |
1785 | if (src_u32[i]) { | |
080e28d0 | 1786 | dst->map |= UINT64_C(1) << i; |
5cb7a798 BP |
1787 | n++; |
1788 | } | |
1789 | } | |
1790 | ||
df40c152 BP |
1791 | miniflow_init__(dst, src, n); |
1792 | } | |
5cb7a798 | 1793 | |
df40c152 BP |
1794 | /* Initializes 'dst' as a copy of 'src', using 'mask->map' as 'dst''s map. The |
1795 | * caller must eventually free 'dst' with miniflow_destroy(). */ | |
1796 | void | |
1797 | miniflow_init_with_minimask(struct miniflow *dst, const struct flow *src, | |
1798 | const struct minimask *mask) | |
1799 | { | |
080e28d0 | 1800 | dst->map = mask->masks.map; |
df40c152 | 1801 | miniflow_init__(dst, src, miniflow_n_values(dst)); |
5cb7a798 BP |
1802 | } |
1803 | ||
1804 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
1805 | * with miniflow_destroy(). */ | |
1806 | void | |
1807 | miniflow_clone(struct miniflow *dst, const struct miniflow *src) | |
1808 | { | |
3016f3e4 JR |
1809 | int size = MINIFLOW_VALUES_SIZE(miniflow_n_values(src)); |
1810 | uint32_t *values; | |
1811 | ||
080e28d0 | 1812 | dst->map = src->map; |
3016f3e4 JR |
1813 | if (size <= sizeof dst->inline_values) { |
1814 | dst->values_inline = true; | |
1815 | values = dst->inline_values; | |
1816 | } else { | |
1817 | dst->values_inline = false; | |
1818 | COVERAGE_INC(miniflow_malloc); | |
1819 | dst->offline_values = xmalloc(size); | |
1820 | values = dst->offline_values; | |
1821 | } | |
1822 | memcpy(values, miniflow_get_values(src), size); | |
1823 | } | |
1824 | ||
1825 | /* Initializes 'dst' as a copy of 'src'. The caller must have allocated | |
1826 | * 'dst' to have inline space all data in 'src'. */ | |
1827 | void | |
1828 | miniflow_clone_inline(struct miniflow *dst, const struct miniflow *src, | |
1829 | size_t n_values) | |
1830 | { | |
1831 | dst->values_inline = true; | |
1832 | dst->map = src->map; | |
1833 | memcpy(dst->inline_values, miniflow_get_values(src), | |
1834 | MINIFLOW_VALUES_SIZE(n_values)); | |
5cb7a798 BP |
1835 | } |
1836 | ||
b2c1f00b | 1837 | /* Initializes 'dst' with the data in 'src', destroying 'src'. |
3016f3e4 JR |
1838 | * The caller must eventually free 'dst' with miniflow_destroy(). |
1839 | * 'dst' must be regularly sized miniflow, but 'src' can have | |
3c30d111 JR |
1840 | * storage for more than the default MINI_N_INLINE inline |
1841 | * values. */ | |
b2c1f00b BP |
1842 | void |
1843 | miniflow_move(struct miniflow *dst, struct miniflow *src) | |
1844 | { | |
3016f3e4 JR |
1845 | int size = MINIFLOW_VALUES_SIZE(miniflow_n_values(src)); |
1846 | ||
1847 | dst->map = src->map; | |
1848 | if (size <= sizeof dst->inline_values) { | |
1849 | dst->values_inline = true; | |
1850 | memcpy(dst->inline_values, miniflow_get_values(src), size); | |
1851 | miniflow_destroy(src); | |
097673a6 JR |
1852 | } else if (src->values_inline) { |
1853 | dst->values_inline = false; | |
1854 | COVERAGE_INC(miniflow_malloc); | |
1855 | dst->offline_values = xmalloc(size); | |
1856 | memcpy(dst->offline_values, src->inline_values, size); | |
3016f3e4 JR |
1857 | } else { |
1858 | dst->values_inline = false; | |
1859 | dst->offline_values = src->offline_values; | |
1860 | } | |
b2c1f00b BP |
1861 | } |
1862 | ||
5cb7a798 BP |
1863 | /* Frees any memory owned by 'flow'. Does not free the storage in which 'flow' |
1864 | * itself resides; the caller is responsible for that. */ | |
1865 | void | |
1866 | miniflow_destroy(struct miniflow *flow) | |
1867 | { | |
27bbe15d JR |
1868 | if (!flow->values_inline) { |
1869 | free(flow->offline_values); | |
5cb7a798 BP |
1870 | } |
1871 | } | |
1872 | ||
1873 | /* Initializes 'dst' as a copy of 'src'. */ | |
1874 | void | |
1875 | miniflow_expand(const struct miniflow *src, struct flow *dst) | |
1876 | { | |
ad77e3c5 EJ |
1877 | memset(dst, 0, sizeof *dst); |
1878 | flow_union_with_miniflow(dst, src); | |
5cb7a798 BP |
1879 | } |
1880 | ||
5cb7a798 BP |
1881 | /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'flow' |
1882 | * were expanded into a "struct flow". */ | |
b7807e4f | 1883 | static uint32_t |
5cb7a798 BP |
1884 | miniflow_get(const struct miniflow *flow, unsigned int u32_ofs) |
1885 | { | |
b7807e4f | 1886 | return (flow->map & UINT64_C(1) << u32_ofs) |
27bbe15d | 1887 | ? *(miniflow_get_u32_values(flow) + |
b7807e4f JR |
1888 | count_1bits(flow->map & ((UINT64_C(1) << u32_ofs) - 1))) |
1889 | : 0; | |
5cb7a798 BP |
1890 | } |
1891 | ||
8e86f090 | 1892 | /* Returns true if 'a' and 'b' are the equal miniflow, false otherwise. */ |
5cb7a798 BP |
1893 | bool |
1894 | miniflow_equal(const struct miniflow *a, const struct miniflow *b) | |
1895 | { | |
27bbe15d JR |
1896 | const uint32_t *ap = miniflow_get_u32_values(a); |
1897 | const uint32_t *bp = miniflow_get_u32_values(b); | |
080e28d0 JR |
1898 | const uint64_t a_map = a->map; |
1899 | const uint64_t b_map = b->map; | |
5cb7a798 | 1900 | |
27bbe15d JR |
1901 | if (OVS_LIKELY(a_map == b_map)) { |
1902 | int count = miniflow_n_values(a); | |
1903 | ||
8cc86801 | 1904 | return !memcmp(ap, bp, count * sizeof *ap); |
080e28d0 | 1905 | } else { |
27bbe15d JR |
1906 | uint64_t map; |
1907 | ||
080e28d0 JR |
1908 | for (map = a_map | b_map; map; map = zero_rightmost_1bit(map)) { |
1909 | uint64_t bit = rightmost_1bit(map); | |
1910 | uint64_t a_value = a_map & bit ? *ap++ : 0; | |
1911 | uint64_t b_value = b_map & bit ? *bp++ : 0; | |
df40c152 | 1912 | |
080e28d0 JR |
1913 | if (a_value != b_value) { |
1914 | return false; | |
df40c152 | 1915 | } |
5cb7a798 BP |
1916 | } |
1917 | } | |
1918 | ||
df40c152 | 1919 | return true; |
5cb7a798 BP |
1920 | } |
1921 | ||
1922 | /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits | |
1923 | * in 'mask', false if they differ. */ | |
1924 | bool | |
1925 | miniflow_equal_in_minimask(const struct miniflow *a, const struct miniflow *b, | |
1926 | const struct minimask *mask) | |
1927 | { | |
27bbe15d | 1928 | const uint32_t *p = miniflow_get_u32_values(&mask->masks); |
080e28d0 | 1929 | uint64_t map; |
5cb7a798 | 1930 | |
080e28d0 | 1931 | for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { |
d43d314e | 1932 | int ofs = raw_ctz(map); |
5cb7a798 | 1933 | |
27bbe15d | 1934 | if ((miniflow_get(a, ofs) ^ miniflow_get(b, ofs)) & *p++) { |
080e28d0 | 1935 | return false; |
5cb7a798 BP |
1936 | } |
1937 | } | |
1938 | ||
1939 | return true; | |
1940 | } | |
1941 | ||
1942 | /* Returns true if 'a' and 'b' are equal at the places where there are 1-bits | |
1943 | * in 'mask', false if they differ. */ | |
1944 | bool | |
1945 | miniflow_equal_flow_in_minimask(const struct miniflow *a, const struct flow *b, | |
1946 | const struct minimask *mask) | |
1947 | { | |
1948 | const uint32_t *b_u32 = (const uint32_t *) b; | |
27bbe15d | 1949 | const uint32_t *p = miniflow_get_u32_values(&mask->masks); |
080e28d0 | 1950 | uint64_t map; |
5cb7a798 | 1951 | |
080e28d0 | 1952 | for (map = mask->masks.map; map; map = zero_rightmost_1bit(map)) { |
d43d314e | 1953 | int ofs = raw_ctz(map); |
5cb7a798 | 1954 | |
27bbe15d | 1955 | if ((miniflow_get(a, ofs) ^ b_u32[ofs]) & *p++) { |
080e28d0 | 1956 | return false; |
5cb7a798 BP |
1957 | } |
1958 | } | |
1959 | ||
1960 | return true; | |
1961 | } | |
1962 | ||
5cb7a798 BP |
1963 | \f |
1964 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
1965 | * with minimask_destroy(). */ | |
1966 | void | |
1967 | minimask_init(struct minimask *mask, const struct flow_wildcards *wc) | |
1968 | { | |
1969 | miniflow_init(&mask->masks, &wc->masks); | |
1970 | } | |
1971 | ||
1972 | /* Initializes 'dst' as a copy of 'src'. The caller must eventually free 'dst' | |
1973 | * with minimask_destroy(). */ | |
1974 | void | |
1975 | minimask_clone(struct minimask *dst, const struct minimask *src) | |
1976 | { | |
1977 | miniflow_clone(&dst->masks, &src->masks); | |
1978 | } | |
1979 | ||
b2c1f00b BP |
1980 | /* Initializes 'dst' with the data in 'src', destroying 'src'. |
1981 | * The caller must eventually free 'dst' with minimask_destroy(). */ | |
1982 | void | |
1983 | minimask_move(struct minimask *dst, struct minimask *src) | |
1984 | { | |
a24de7ee | 1985 | miniflow_move(&dst->masks, &src->masks); |
b2c1f00b BP |
1986 | } |
1987 | ||
5cb7a798 BP |
1988 | /* Initializes 'dst_' as the bit-wise "and" of 'a_' and 'b_'. |
1989 | * | |
1990 | * The caller must provide room for FLOW_U32S "uint32_t"s in 'storage', for use | |
1991 | * by 'dst_'. The caller must *not* free 'dst_' with minimask_destroy(). */ | |
1992 | void | |
1993 | minimask_combine(struct minimask *dst_, | |
1994 | const struct minimask *a_, const struct minimask *b_, | |
1995 | uint32_t storage[FLOW_U32S]) | |
1996 | { | |
1997 | struct miniflow *dst = &dst_->masks; | |
27bbe15d | 1998 | uint32_t *dst_values = storage; |
5cb7a798 BP |
1999 | const struct miniflow *a = &a_->masks; |
2000 | const struct miniflow *b = &b_->masks; | |
080e28d0 JR |
2001 | uint64_t map; |
2002 | int n = 0; | |
5cb7a798 | 2003 | |
27bbe15d JR |
2004 | dst->values_inline = false; |
2005 | dst->offline_values = storage; | |
080e28d0 JR |
2006 | |
2007 | dst->map = 0; | |
2008 | for (map = a->map & b->map; map; map = zero_rightmost_1bit(map)) { | |
d43d314e | 2009 | int ofs = raw_ctz(map); |
080e28d0 JR |
2010 | uint32_t mask = miniflow_get(a, ofs) & miniflow_get(b, ofs); |
2011 | ||
2012 | if (mask) { | |
2013 | dst->map |= rightmost_1bit(map); | |
27bbe15d | 2014 | dst_values[n++] = mask; |
5cb7a798 BP |
2015 | } |
2016 | } | |
2017 | } | |
2018 | ||
2019 | /* Frees any memory owned by 'mask'. Does not free the storage in which 'mask' | |
2020 | * itself resides; the caller is responsible for that. */ | |
2021 | void | |
2022 | minimask_destroy(struct minimask *mask) | |
2023 | { | |
2024 | miniflow_destroy(&mask->masks); | |
2025 | } | |
2026 | ||
2027 | /* Initializes 'dst' as a copy of 'src'. */ | |
2028 | void | |
2029 | minimask_expand(const struct minimask *mask, struct flow_wildcards *wc) | |
2030 | { | |
2031 | miniflow_expand(&mask->masks, &wc->masks); | |
2032 | } | |
2033 | ||
2034 | /* Returns the uint32_t that would be at byte offset '4 * u32_ofs' if 'mask' | |
2035 | * were expanded into a "struct flow_wildcards". */ | |
2036 | uint32_t | |
2037 | minimask_get(const struct minimask *mask, unsigned int u32_ofs) | |
2038 | { | |
2039 | return miniflow_get(&mask->masks, u32_ofs); | |
2040 | } | |
2041 | ||
5cb7a798 BP |
2042 | /* Returns true if 'a' and 'b' are the same flow mask, false otherwise. */ |
2043 | bool | |
2044 | minimask_equal(const struct minimask *a, const struct minimask *b) | |
2045 | { | |
2046 | return miniflow_equal(&a->masks, &b->masks); | |
2047 | } | |
2048 | ||
d4570fd8 | 2049 | /* Returns true if at least one bit matched by 'b' is wildcarded by 'a', |
5cb7a798 BP |
2050 | * false otherwise. */ |
2051 | bool | |
d4570fd8 | 2052 | minimask_has_extra(const struct minimask *a, const struct minimask *b) |
5cb7a798 | 2053 | { |
27bbe15d | 2054 | const uint32_t *p = miniflow_get_u32_values(&b->masks); |
080e28d0 | 2055 | uint64_t map; |
5cb7a798 | 2056 | |
d4570fd8 JR |
2057 | for (map = b->masks.map; map; map = zero_rightmost_1bit(map)) { |
2058 | uint32_t a_u32 = minimask_get(a, raw_ctz(map)); | |
2059 | uint32_t b_u32 = *p++; | |
5cb7a798 | 2060 | |
080e28d0 JR |
2061 | if ((a_u32 & b_u32) != b_u32) { |
2062 | return true; | |
5cb7a798 BP |
2063 | } |
2064 | } | |
2065 | ||
2066 | return false; | |
2067 | } |