]>
Commit | Line | Data |
---|---|---|
a489b168 | 1 | /* |
4ea96698 | 2 | * Copyright (c) 2015-2019 Nicira, Inc. |
a489b168 DDP |
3 | * |
4 | * Licensed under the Apache License, Version 2.0 (the "License"); | |
5 | * you may not use this file except in compliance with the License. | |
6 | * You may obtain a copy of the License at: | |
7 | * | |
8 | * http://www.apache.org/licenses/LICENSE-2.0 | |
9 | * | |
10 | * Unless required by applicable law or agreed to in writing, software | |
11 | * distributed under the License is distributed on an "AS IS" BASIS, | |
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
13 | * See the License for the specific language governing permissions and | |
14 | * limitations under the License. | |
15 | */ | |
16 | ||
17 | #include <config.h> | |
bd5e81a0 | 18 | #include <ctype.h> |
a489b168 | 19 | #include <errno.h> |
ff6aa424 | 20 | #include <sys/types.h> |
a489b168 DDP |
21 | #include <netinet/in.h> |
22 | #include <netinet/icmp6.h> | |
bd5e81a0 | 23 | #include <string.h> |
a489b168 DDP |
24 | |
25 | #include "bitmap.h" | |
bd5e81a0 | 26 | #include "conntrack.h" |
a489b168 DDP |
27 | #include "conntrack-private.h" |
28 | #include "coverage.h" | |
29 | #include "csum.h" | |
4d4e68ed | 30 | #include "ct-dpif.h" |
a489b168 DDP |
31 | #include "dp-packet.h" |
32 | #include "flow.h" | |
33 | #include "netdev.h" | |
34 | #include "odp-netlink.h" | |
35 | #include "openvswitch/hmap.h" | |
36 | #include "openvswitch/vlog.h" | |
37 | #include "ovs-rcu.h" | |
e6ef6cc6 | 38 | #include "ovs-thread.h" |
fd016ae3 | 39 | #include "openvswitch/poll-loop.h" |
a489b168 DDP |
40 | #include "random.h" |
41 | #include "timeval.h" | |
42 | ||
43 | VLOG_DEFINE_THIS_MODULE(conntrack); | |
44 | ||
45 | COVERAGE_DEFINE(conntrack_full); | |
e6ef6cc6 | 46 | COVERAGE_DEFINE(conntrack_long_cleanup); |
a489b168 DDP |
47 | |
48 | struct conn_lookup_ctx { | |
49 | struct conn_key key; | |
50 | struct conn *conn; | |
51 | uint32_t hash; | |
52 | bool reply; | |
dbb597d3 | 53 | bool icmp_related; |
a489b168 DDP |
54 | }; |
55 | ||
bd5e81a0 DB |
56 | enum ftp_ctl_pkt { |
57 | /* Control packets with address and/or port specifiers. */ | |
58 | CT_FTP_CTL_INTEREST, | |
59 | /* Control packets without address and/or port specifiers. */ | |
60 | CT_FTP_CTL_OTHER, | |
61 | CT_FTP_CTL_INVALID, | |
62 | }; | |
63 | ||
64 | enum ct_alg_mode { | |
65 | CT_FTP_MODE_ACTIVE, | |
66 | CT_FTP_MODE_PASSIVE, | |
7be77cb0 | 67 | CT_TFTP_MODE, |
bd5e81a0 DB |
68 | }; |
69 | ||
94e71143 DB |
70 | enum ct_alg_ctl_type { |
71 | CT_ALG_CTL_NONE, | |
72 | CT_ALG_CTL_FTP, | |
73 | CT_ALG_CTL_TFTP, | |
be38342d DB |
74 | /* SIP is not enabled through Openflow and presently only used as |
75 | * an example of an alg that allows a wildcard src ip. */ | |
76 | CT_ALG_CTL_SIP, | |
94e71143 DB |
77 | }; |
78 | ||
a489b168 | 79 | static bool conn_key_extract(struct conntrack *, struct dp_packet *, |
66e4ad8a DDP |
80 | ovs_be16 dl_type, struct conn_lookup_ctx *, |
81 | uint16_t zone); | |
a489b168 DDP |
82 | static uint32_t conn_key_hash(const struct conn_key *, uint32_t basis); |
83 | static void conn_key_reverse(struct conn_key *); | |
a489b168 | 84 | static bool valid_new(struct dp_packet *pkt, struct conn_key *); |
967bb5c5 | 85 | static struct conn *new_conn(struct conntrack *ct, struct dp_packet *pkt, |
e6ef6cc6 | 86 | struct conn_key *, long long now); |
967bb5c5 | 87 | static void delete_conn_cmn(struct conn *); |
a489b168 | 88 | static void delete_conn(struct conn *); |
967bb5c5 DB |
89 | static void delete_conn_one(struct conn *conn); |
90 | static enum ct_update_res conn_update(struct conntrack *ct, struct conn *conn, | |
91 | struct dp_packet *pkt, | |
92 | struct conn_lookup_ctx *ctx, | |
e6ef6cc6 | 93 | long long now); |
a489b168 DDP |
94 | static bool conn_expired(struct conn *, long long now); |
95 | static void set_mark(struct dp_packet *, struct conn *, | |
96 | uint32_t val, uint32_t mask); | |
97 | static void set_label(struct dp_packet *, struct conn *, | |
98 | const struct ovs_key_ct_labels *val, | |
99 | const struct ovs_key_ct_labels *mask); | |
e6ef6cc6 | 100 | static void *clean_thread_main(void *f_); |
a489b168 | 101 | |
286de272 DB |
102 | static bool |
103 | nat_select_range_tuple(struct conntrack *ct, const struct conn *conn, | |
104 | struct conn *nat_conn); | |
105 | ||
106 | static uint8_t | |
107 | reverse_icmp_type(uint8_t type); | |
108 | static uint8_t | |
109 | reverse_icmp6_type(uint8_t type); | |
110 | static inline bool | |
111 | extract_l3_ipv4(struct conn_key *key, const void *data, size_t size, | |
112 | const char **new_data, bool validate_checksum); | |
113 | static inline bool | |
114 | extract_l3_ipv6(struct conn_key *key, const void *data, size_t size, | |
115 | const char **new_data); | |
bd5e81a0 | 116 | static struct alg_exp_node * |
be38342d DB |
117 | expectation_lookup(struct hmap *alg_expectations, const struct conn_key *key, |
118 | uint32_t basis, bool src_ip_wc); | |
bd5e81a0 DB |
119 | |
120 | static int | |
121 | repl_ftp_v4_addr(struct dp_packet *pkt, ovs_be32 v4_addr_rep, | |
122 | char *ftp_data_v4_start, | |
cd7c99a6 | 123 | size_t addr_offset_from_ftp_data_start, size_t addr_size); |
bd5e81a0 DB |
124 | |
125 | static enum ftp_ctl_pkt | |
126 | process_ftp_ctl_v4(struct conntrack *ct, | |
127 | struct dp_packet *pkt, | |
128 | const struct conn *conn_for_expectation, | |
4417ca3d | 129 | ovs_be32 *v4_addr_rep, |
bd5e81a0 | 130 | char **ftp_data_v4_start, |
cd7c99a6 DB |
131 | size_t *addr_offset_from_ftp_data_start, |
132 | size_t *addr_size); | |
bd5e81a0 DB |
133 | |
134 | static enum ftp_ctl_pkt | |
135 | detect_ftp_ctl_type(const struct conn_lookup_ctx *ctx, | |
136 | struct dp_packet *pkt); | |
137 | ||
4417ca3d | 138 | static void |
967bb5c5 | 139 | expectation_clean(struct conntrack *ct, const struct conn_key *master_key); |
4417ca3d | 140 | |
94e71143 DB |
141 | static struct ct_l4_proto *l4_protos[] = { |
142 | [IPPROTO_TCP] = &ct_proto_tcp, | |
143 | [IPPROTO_UDP] = &ct_proto_other, | |
144 | [IPPROTO_ICMP] = &ct_proto_icmp4, | |
145 | [IPPROTO_ICMPV6] = &ct_proto_icmp6, | |
146 | }; | |
147 | ||
bd5e81a0 DB |
148 | static void |
149 | handle_ftp_ctl(struct conntrack *ct, const struct conn_lookup_ctx *ctx, | |
967bb5c5 DB |
150 | struct dp_packet *pkt, struct conn *ec, long long now, |
151 | enum ftp_ctl_pkt ftp_ctl, bool nat); | |
bd5e81a0 | 152 | |
7be77cb0 DB |
153 | static void |
154 | handle_tftp_ctl(struct conntrack *ct, | |
94e71143 | 155 | const struct conn_lookup_ctx *ctx OVS_UNUSED, |
967bb5c5 DB |
156 | struct dp_packet *pkt, struct conn *conn_for_expectation, |
157 | long long now OVS_UNUSED, enum ftp_ctl_pkt ftp_ctl OVS_UNUSED, | |
158 | bool nat OVS_UNUSED); | |
94e71143 DB |
159 | |
160 | typedef void (*alg_helper)(struct conntrack *ct, | |
161 | const struct conn_lookup_ctx *ctx, | |
162 | struct dp_packet *pkt, | |
967bb5c5 | 163 | struct conn *conn_for_expectation, |
94e71143 DB |
164 | long long now, enum ftp_ctl_pkt ftp_ctl, |
165 | bool nat); | |
166 | ||
167 | static alg_helper alg_helpers[] = { | |
168 | [CT_ALG_CTL_NONE] = NULL, | |
169 | [CT_ALG_CTL_FTP] = handle_ftp_ctl, | |
170 | [CT_ALG_CTL_TFTP] = handle_tftp_ctl, | |
a489b168 DDP |
171 | }; |
172 | ||
173 | long long ct_timeout_val[] = { | |
174 | #define CT_TIMEOUT(NAME, VAL) [CT_TM_##NAME] = VAL, | |
175 | CT_TIMEOUTS | |
176 | #undef CT_TIMEOUT | |
177 | }; | |
178 | ||
bd5e81a0 DB |
179 | /* The maximum TCP or UDP port number. */ |
180 | #define CT_MAX_L4_PORT 65535 | |
bd5e81a0 DB |
181 | /* String buffer used for parsing FTP string messages. |
182 | * This is sized about twice what is needed to leave some | |
183 | * margin of error. */ | |
184 | #define LARGEST_FTP_MSG_OF_INTEREST 128 | |
185 | /* FTP port string used in active mode. */ | |
186 | #define FTP_PORT_CMD "PORT" | |
187 | /* FTP pasv string used in passive mode. */ | |
188 | #define FTP_PASV_REPLY_CODE "227" | |
189 | /* Maximum decimal digits for port in FTP command. | |
190 | * The port is represented as two 3 digit numbers with the | |
191 | * high part a multiple of 256. */ | |
192 | #define MAX_FTP_PORT_DGTS 3 | |
193 | ||
194 | /* FTP extension EPRT string used for active mode. */ | |
195 | #define FTP_EPRT_CMD "EPRT" | |
196 | /* FTP extension EPSV string used for passive mode. */ | |
197 | #define FTP_EPSV_REPLY "EXTENDED PASSIVE" | |
198 | /* Maximum decimal digits for port in FTP extended command. */ | |
199 | #define MAX_EXT_FTP_PORT_DGTS 5 | |
200 | /* FTP extended command code for IPv6. */ | |
201 | #define FTP_AF_V6 '2' | |
202 | /* Used to indicate a wildcard L4 source port number for ALGs. | |
203 | * This is used for port numbers that we cannot predict in | |
204 | * expectations. */ | |
205 | #define ALG_WC_SRC_PORT 0 | |
206 | ||
a489b168 | 207 | /* If the total number of connections goes above this value, no new connections |
286de272 | 208 | * are accepted; this is for CT_CONN_TYPE_DEFAULT connections. */ |
a489b168 DDP |
209 | #define DEFAULT_N_CONN_LIMIT 3000000 |
210 | ||
5ed7a0b4 DB |
211 | /* Does a member by member comparison of two conn_keys; this |
212 | * function must be kept in sync with struct conn_key; returns 0 | |
213 | * if the keys are equal or 1 if the keys are not equal. */ | |
214 | static int | |
215 | conn_key_cmp(const struct conn_key *key1, const struct conn_key *key2) | |
216 | { | |
217 | if (!memcmp(&key1->src.addr, &key2->src.addr, sizeof key1->src.addr) && | |
218 | !memcmp(&key1->dst.addr, &key2->dst.addr, sizeof key1->dst.addr) && | |
219 | (key1->src.icmp_id == key2->src.icmp_id) && | |
220 | (key1->src.icmp_type == key2->src.icmp_type) && | |
221 | (key1->src.icmp_code == key2->src.icmp_code) && | |
222 | (key1->dst.icmp_id == key2->dst.icmp_id) && | |
223 | (key1->dst.icmp_type == key2->dst.icmp_type) && | |
224 | (key1->dst.icmp_code == key2->dst.icmp_code) && | |
225 | (key1->dl_type == key2->dl_type) && | |
226 | (key1->zone == key2->zone) && | |
227 | (key1->nw_proto == key2->nw_proto)) { | |
228 | ||
229 | return 0; | |
230 | } | |
231 | return 1; | |
232 | } | |
233 | ||
d8682ee5 | 234 | static void |
dec0dbbc DB |
235 | ct_print_conn_info(const struct conn *c, const char *log_msg, |
236 | enum vlog_level vll, bool force, bool rl_on) | |
66f400f5 DB |
237 | { |
238 | #define CT_VLOG(RL_ON, LEVEL, ...) \ | |
239 | do { \ | |
240 | if (RL_ON) { \ | |
241 | static struct vlog_rate_limit rl_ = VLOG_RATE_LIMIT_INIT(5, 5); \ | |
242 | vlog_rate_limit(&this_module, LEVEL, &rl_, __VA_ARGS__); \ | |
243 | } else { \ | |
244 | vlog(&this_module, LEVEL, __VA_ARGS__); \ | |
245 | } \ | |
246 | } while (0) | |
247 | ||
248 | if (OVS_UNLIKELY(force || vlog_is_enabled(&this_module, vll))) { | |
249 | if (c->key.dl_type == htons(ETH_TYPE_IP)) { | |
250 | CT_VLOG(rl_on, vll, "%s: src ip "IP_FMT" dst ip "IP_FMT" rev src " | |
251 | "ip "IP_FMT" rev dst ip "IP_FMT" src/dst ports " | |
252 | "%"PRIu16"/%"PRIu16" rev src/dst ports " | |
253 | "%"PRIu16"/%"PRIu16" zone/rev zone " | |
254 | "%"PRIu16"/%"PRIu16" nw_proto/rev nw_proto " | |
255 | "%"PRIu8"/%"PRIu8, log_msg, | |
cda1b109 DB |
256 | IP_ARGS(c->key.src.addr.ipv4), |
257 | IP_ARGS(c->key.dst.addr.ipv4), | |
258 | IP_ARGS(c->rev_key.src.addr.ipv4), | |
259 | IP_ARGS(c->rev_key.dst.addr.ipv4), | |
66f400f5 DB |
260 | ntohs(c->key.src.port), ntohs(c->key.dst.port), |
261 | ntohs(c->rev_key.src.port), ntohs(c->rev_key.dst.port), | |
262 | c->key.zone, c->rev_key.zone, c->key.nw_proto, | |
263 | c->rev_key.nw_proto); | |
264 | } else { | |
265 | char ip6_s[INET6_ADDRSTRLEN]; | |
266 | inet_ntop(AF_INET6, &c->key.src.addr.ipv6, ip6_s, sizeof ip6_s); | |
267 | char ip6_d[INET6_ADDRSTRLEN]; | |
268 | inet_ntop(AF_INET6, &c->key.dst.addr.ipv6, ip6_d, sizeof ip6_d); | |
269 | char ip6_rs[INET6_ADDRSTRLEN]; | |
270 | inet_ntop(AF_INET6, &c->rev_key.src.addr.ipv6, ip6_rs, | |
271 | sizeof ip6_rs); | |
272 | char ip6_rd[INET6_ADDRSTRLEN]; | |
273 | inet_ntop(AF_INET6, &c->rev_key.dst.addr.ipv6, ip6_rd, | |
274 | sizeof ip6_rd); | |
275 | ||
276 | CT_VLOG(rl_on, vll, "%s: src ip %s dst ip %s rev src ip %s" | |
277 | " rev dst ip %s src/dst ports %"PRIu16"/%"PRIu16 | |
278 | " rev src/dst ports %"PRIu16"/%"PRIu16" zone/rev zone " | |
279 | "%"PRIu16"/%"PRIu16" nw_proto/rev nw_proto " | |
280 | "%"PRIu8"/%"PRIu8, log_msg, ip6_s, ip6_d, ip6_rs, | |
281 | ip6_rd, ntohs(c->key.src.port), ntohs(c->key.dst.port), | |
282 | ntohs(c->rev_key.src.port), ntohs(c->rev_key.dst.port), | |
283 | c->key.zone, c->rev_key.zone, c->key.nw_proto, | |
284 | c->rev_key.nw_proto); | |
285 | } | |
286 | } | |
287 | } | |
288 | ||
a489b168 DDP |
289 | /* Initializes the connection tracker 'ct'. The caller is responsible for |
290 | * calling 'conntrack_destroy()', when the instance is not needed anymore */ | |
57593fd2 DB |
291 | struct conntrack * |
292 | conntrack_init(void) | |
a489b168 | 293 | { |
57593fd2 DB |
294 | struct conntrack *ct = xzalloc(sizeof *ct); |
295 | ||
967bb5c5 DB |
296 | ovs_rwlock_init(&ct->resources_lock); |
297 | ovs_rwlock_wrlock(&ct->resources_lock); | |
bd5e81a0 | 298 | hmap_init(&ct->alg_expectations); |
4417ca3d | 299 | hindex_init(&ct->alg_expectation_refs); |
967bb5c5 | 300 | ovs_rwlock_unlock(&ct->resources_lock); |
a489b168 | 301 | |
967bb5c5 DB |
302 | ovs_mutex_init_adaptive(&ct->ct_lock); |
303 | ovs_mutex_lock(&ct->ct_lock); | |
304 | cmap_init(&ct->conns); | |
305 | for (unsigned i = 0; i < ARRAY_SIZE(ct->exp_lists); i++) { | |
306 | ovs_list_init(&ct->exp_lists[i]); | |
a489b168 | 307 | } |
967bb5c5 DB |
308 | ovs_mutex_unlock(&ct->ct_lock); |
309 | ||
a489b168 DDP |
310 | ct->hash_basis = random_uint32(); |
311 | atomic_count_init(&ct->n_conn, 0); | |
312 | atomic_init(&ct->n_conn_limit, DEFAULT_N_CONN_LIMIT); | |
e6ef6cc6 DDP |
313 | latch_init(&ct->clean_thread_exit); |
314 | ct->clean_thread = ovs_thread_create("ct_clean", clean_thread_main, ct); | |
4ea96698 | 315 | ct->ipf = ipf_init(); |
57593fd2 DB |
316 | |
317 | return ct; | |
a489b168 DDP |
318 | } |
319 | ||
967bb5c5 DB |
320 | static void |
321 | conn_clean_cmn(struct conntrack *ct, struct conn *conn) | |
322 | OVS_REQUIRES(ct->ct_lock) | |
323 | { | |
324 | if (conn->alg) { | |
325 | expectation_clean(ct, &conn->key); | |
326 | } | |
327 | ||
328 | uint32_t hash = conn_key_hash(&conn->key, ct->hash_basis); | |
329 | cmap_remove(&ct->conns, &conn->cm_node, hash); | |
330 | } | |
331 | ||
332 | /* Must be called with 'conn' of 'conn_type' CT_CONN_TYPE_DEFAULT. Also | |
333 | * removes the associated nat 'conn' from the lookup datastructures. */ | |
334 | static void | |
335 | conn_clean(struct conntrack *ct, struct conn *conn) | |
336 | OVS_REQUIRES(ct->ct_lock) | |
337 | { | |
338 | ovs_assert(conn->conn_type == CT_CONN_TYPE_DEFAULT); | |
339 | ||
340 | conn_clean_cmn(ct, conn); | |
341 | if (conn->nat_conn) { | |
342 | uint32_t hash = conn_key_hash(&conn->nat_conn->key, ct->hash_basis); | |
343 | cmap_remove(&ct->conns, &conn->nat_conn->cm_node, hash); | |
344 | } | |
345 | ovs_list_remove(&conn->exp_node); | |
5f918a8a | 346 | conn->cleaned = true; |
967bb5c5 DB |
347 | ovsrcu_postpone(delete_conn, conn); |
348 | atomic_count_dec(&ct->n_conn); | |
349 | } | |
350 | ||
351 | static void | |
352 | conn_clean_one(struct conntrack *ct, struct conn *conn) | |
353 | OVS_REQUIRES(ct->ct_lock) | |
354 | { | |
355 | conn_clean_cmn(ct, conn); | |
356 | if (conn->conn_type == CT_CONN_TYPE_DEFAULT) { | |
357 | ovs_list_remove(&conn->exp_node); | |
5f918a8a | 358 | conn->cleaned = true; |
967bb5c5 DB |
359 | atomic_count_dec(&ct->n_conn); |
360 | } | |
361 | ovsrcu_postpone(delete_conn_one, conn); | |
362 | } | |
363 | ||
364 | /* Destroys the connection tracker 'ct' and frees all the allocated memory. | |
365 | * The caller of this function must already have shut down packet input | |
366 | * and PMD threads (which would have been quiesced). */ | |
a489b168 DDP |
367 | void |
368 | conntrack_destroy(struct conntrack *ct) | |
369 | { | |
967bb5c5 | 370 | struct conn *conn; |
e6ef6cc6 DDP |
371 | latch_set(&ct->clean_thread_exit); |
372 | pthread_join(ct->clean_thread, NULL); | |
373 | latch_destroy(&ct->clean_thread_exit); | |
a489b168 | 374 | |
967bb5c5 DB |
375 | ovs_mutex_lock(&ct->ct_lock); |
376 | CMAP_FOR_EACH (conn, cm_node, &ct->conns) { | |
377 | conn_clean_one(ct, conn); | |
a489b168 | 378 | } |
967bb5c5 DB |
379 | cmap_destroy(&ct->conns); |
380 | ovs_mutex_unlock(&ct->ct_lock); | |
381 | ovs_mutex_destroy(&ct->ct_lock); | |
bd5e81a0 | 382 | |
967bb5c5 | 383 | ovs_rwlock_wrlock(&ct->resources_lock); |
bd5e81a0 DB |
384 | struct alg_exp_node *alg_exp_node; |
385 | HMAP_FOR_EACH_POP (alg_exp_node, node, &ct->alg_expectations) { | |
386 | free(alg_exp_node); | |
387 | } | |
bd5e81a0 | 388 | hmap_destroy(&ct->alg_expectations); |
4417ca3d | 389 | hindex_destroy(&ct->alg_expectation_refs); |
967bb5c5 DB |
390 | ovs_rwlock_unlock(&ct->resources_lock); |
391 | ovs_rwlock_destroy(&ct->resources_lock); | |
392 | ||
4ea96698 | 393 | ipf_destroy(ct->ipf); |
21ffe409 | 394 | free(ct); |
a489b168 DDP |
395 | } |
396 | \f | |
967bb5c5 DB |
397 | |
398 | static bool | |
399 | conn_key_lookup(struct conntrack *ct, const struct conn_key *key, | |
400 | uint32_t hash, long long now, struct conn **conn_out, | |
401 | bool *reply) | |
a489b168 | 402 | { |
967bb5c5 DB |
403 | struct conn *conn; |
404 | bool found = false; | |
405 | ||
406 | CMAP_FOR_EACH_WITH_HASH (conn, cm_node, hash, &ct->conns) { | |
407 | if (!conn_key_cmp(&conn->key, key) && !conn_expired(conn, now)) { | |
408 | found = true; | |
409 | if (reply) { | |
410 | *reply = false; | |
411 | } | |
412 | break; | |
413 | } | |
414 | if (!conn_key_cmp(&conn->rev_key, key) && !conn_expired(conn, now)) { | |
415 | found = true; | |
416 | if (reply) { | |
417 | *reply = true; | |
418 | } | |
419 | break; | |
420 | } | |
421 | } | |
a489b168 | 422 | |
967bb5c5 DB |
423 | if (found && conn_out) { |
424 | *conn_out = conn; | |
425 | } else if (conn_out) { | |
426 | *conn_out = NULL; | |
427 | } | |
428 | return found; | |
a489b168 DDP |
429 | } |
430 | ||
4048c508 DB |
431 | static bool |
432 | conn_lookup(struct conntrack *ct, const struct conn_key *key, | |
433 | long long now, struct conn **conn_out, bool *reply) | |
434 | { | |
435 | uint32_t hash = conn_key_hash(key, ct->hash_basis); | |
436 | return conn_key_lookup(ct, key, hash, now, conn_out, reply); | |
437 | } | |
438 | ||
a489b168 | 439 | static void |
286de272 | 440 | write_ct_md(struct dp_packet *pkt, uint16_t zone, const struct conn *conn, |
bd5e81a0 | 441 | const struct conn_key *key, const struct alg_exp_node *alg_exp) |
a489b168 | 442 | { |
286de272 | 443 | pkt->md.ct_state |= CS_TRACKED; |
a489b168 | 444 | pkt->md.ct_zone = zone; |
967bb5c5 DB |
445 | |
446 | if (conn) { | |
447 | ovs_mutex_lock(&conn->lock); | |
448 | pkt->md.ct_mark = conn->mark; | |
449 | pkt->md.ct_label = conn->label; | |
450 | ovs_mutex_unlock(&conn->lock); | |
451 | } else { | |
452 | pkt->md.ct_mark = 0; | |
453 | pkt->md.ct_label = OVS_U128_ZERO; | |
454 | } | |
daf4d3c1 JR |
455 | |
456 | /* Use the original direction tuple if we have it. */ | |
457 | if (conn) { | |
bd5e81a0 DB |
458 | if (conn->alg_related) { |
459 | key = &conn->master_key; | |
460 | } else { | |
461 | key = &conn->key; | |
462 | } | |
463 | } else if (alg_exp) { | |
464 | pkt->md.ct_mark = alg_exp->master_mark; | |
465 | pkt->md.ct_label = alg_exp->master_label; | |
466 | key = &alg_exp->master_key; | |
daf4d3c1 | 467 | } |
dec0dbbc | 468 | |
daf4d3c1 | 469 | pkt->md.ct_orig_tuple_ipv6 = false; |
dec0dbbc | 470 | |
daf4d3c1 JR |
471 | if (key) { |
472 | if (key->dl_type == htons(ETH_TYPE_IP)) { | |
473 | pkt->md.ct_orig_tuple.ipv4 = (struct ovs_key_ct_tuple_ipv4) { | |
cda1b109 DB |
474 | key->src.addr.ipv4, |
475 | key->dst.addr.ipv4, | |
daf4d3c1 JR |
476 | key->nw_proto != IPPROTO_ICMP |
477 | ? key->src.port : htons(key->src.icmp_type), | |
478 | key->nw_proto != IPPROTO_ICMP | |
479 | ? key->dst.port : htons(key->src.icmp_code), | |
480 | key->nw_proto, | |
481 | }; | |
286de272 | 482 | } else { |
daf4d3c1 JR |
483 | pkt->md.ct_orig_tuple_ipv6 = true; |
484 | pkt->md.ct_orig_tuple.ipv6 = (struct ovs_key_ct_tuple_ipv6) { | |
cda1b109 DB |
485 | key->src.addr.ipv6, |
486 | key->dst.addr.ipv6, | |
daf4d3c1 JR |
487 | key->nw_proto != IPPROTO_ICMPV6 |
488 | ? key->src.port : htons(key->src.icmp_type), | |
489 | key->nw_proto != IPPROTO_ICMPV6 | |
490 | ? key->dst.port : htons(key->src.icmp_code), | |
491 | key->nw_proto, | |
492 | }; | |
493 | } | |
494 | } else { | |
495 | memset(&pkt->md.ct_orig_tuple, 0, sizeof pkt->md.ct_orig_tuple); | |
496 | } | |
bd5e81a0 DB |
497 | } |
498 | ||
499 | static uint8_t | |
500 | get_ip_proto(const struct dp_packet *pkt) | |
501 | { | |
502 | uint8_t ip_proto; | |
503 | struct eth_header *l2 = dp_packet_eth(pkt); | |
504 | if (l2->eth_type == htons(ETH_TYPE_IPV6)) { | |
505 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
506 | ip_proto = nh6->ip6_ctlun.ip6_un1.ip6_un1_nxt; | |
507 | } else { | |
508 | struct ip_header *l3_hdr = dp_packet_l3(pkt); | |
509 | ip_proto = l3_hdr->ip_proto; | |
510 | } | |
286de272 | 511 | |
bd5e81a0 DB |
512 | return ip_proto; |
513 | } | |
514 | ||
515 | static bool | |
94e71143 | 516 | is_ftp_ctl(const enum ct_alg_ctl_type ct_alg_ctl) |
bd5e81a0 | 517 | { |
94e71143 | 518 | return ct_alg_ctl == CT_ALG_CTL_FTP; |
bd5e81a0 DB |
519 | } |
520 | ||
94e71143 | 521 | static enum ct_alg_ctl_type |
bd7d93f8 DB |
522 | get_alg_ctl_type(const struct dp_packet *pkt, ovs_be16 tp_src, ovs_be16 tp_dst, |
523 | const char *helper) | |
7be77cb0 | 524 | { |
94e71143 DB |
525 | /* CT_IPPORT_FTP/TFTP is used because IPPORT_FTP/TFTP in not defined |
526 | * in OSX, at least in in.h. Since these values will never change, remove | |
7be77cb0 | 527 | * the external dependency. */ |
94e71143 DB |
528 | enum { CT_IPPORT_FTP = 21 }; |
529 | enum { CT_IPPORT_TFTP = 69 }; | |
bd7d93f8 DB |
530 | uint8_t ip_proto = get_ip_proto(pkt); |
531 | struct udp_header *uh = dp_packet_l4(pkt); | |
532 | struct tcp_header *th = dp_packet_l4(pkt); | |
533 | ovs_be16 ftp_src_port = htons(CT_IPPORT_FTP); | |
534 | ovs_be16 ftp_dst_port = htons(CT_IPPORT_FTP); | |
535 | ovs_be16 tftp_dst_port = htons(CT_IPPORT_TFTP); | |
536 | ||
537 | if (OVS_UNLIKELY(tp_dst)) { | |
538 | if (helper && !strncmp(helper, "ftp", strlen("ftp"))) { | |
539 | ftp_dst_port = tp_dst; | |
540 | } else if (helper && !strncmp(helper, "tftp", strlen("tftp"))) { | |
541 | tftp_dst_port = tp_dst; | |
542 | } | |
543 | } else if (OVS_UNLIKELY(tp_src)) { | |
544 | if (helper && !strncmp(helper, "ftp", strlen("ftp"))) { | |
545 | ftp_src_port = tp_src; | |
546 | } | |
547 | } | |
7be77cb0 | 548 | |
bd7d93f8 | 549 | if (ip_proto == IPPROTO_UDP && uh->udp_dst == tftp_dst_port) { |
94e71143 DB |
550 | return CT_ALG_CTL_TFTP; |
551 | } else if (ip_proto == IPPROTO_TCP && | |
bd7d93f8 | 552 | (th->tcp_src == ftp_src_port || th->tcp_dst == ftp_dst_port)) { |
94e71143 DB |
553 | return CT_ALG_CTL_FTP; |
554 | } | |
555 | return CT_ALG_CTL_NONE; | |
556 | } | |
557 | ||
be38342d DB |
558 | static bool |
559 | alg_src_ip_wc(enum ct_alg_ctl_type alg_ctl_type) | |
560 | { | |
561 | if (alg_ctl_type == CT_ALG_CTL_SIP) { | |
562 | return true; | |
563 | } | |
564 | return false; | |
565 | } | |
566 | ||
94e71143 DB |
567 | static void |
568 | handle_alg_ctl(struct conntrack *ct, const struct conn_lookup_ctx *ctx, | |
569 | struct dp_packet *pkt, enum ct_alg_ctl_type ct_alg_ctl, | |
967bb5c5 | 570 | struct conn *conn, long long now, bool nat) |
94e71143 DB |
571 | { |
572 | /* ALG control packet handling with expectation creation. */ | |
3a2a425b | 573 | if (OVS_UNLIKELY(alg_helpers[ct_alg_ctl] && conn && conn->alg)) { |
967bb5c5 DB |
574 | ovs_mutex_lock(&conn->lock); |
575 | alg_helpers[ct_alg_ctl](ct, ctx, pkt, conn, now, CT_FTP_CTL_INTEREST, | |
576 | nat); | |
577 | ovs_mutex_unlock(&conn->lock); | |
94e71143 | 578 | } |
7be77cb0 DB |
579 | } |
580 | ||
286de272 DB |
581 | static void |
582 | pat_packet(struct dp_packet *pkt, const struct conn *conn) | |
583 | { | |
584 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
585 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
586 | struct tcp_header *th = dp_packet_l4(pkt); | |
587 | packet_set_tcp_port(pkt, conn->rev_key.dst.port, th->tcp_dst); | |
588 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
589 | struct udp_header *uh = dp_packet_l4(pkt); | |
590 | packet_set_udp_port(pkt, conn->rev_key.dst.port, uh->udp_dst); | |
591 | } | |
592 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { | |
593 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
594 | struct tcp_header *th = dp_packet_l4(pkt); | |
595 | packet_set_tcp_port(pkt, th->tcp_src, conn->rev_key.src.port); | |
596 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
597 | struct udp_header *uh = dp_packet_l4(pkt); | |
598 | packet_set_udp_port(pkt, uh->udp_src, conn->rev_key.src.port); | |
599 | } | |
600 | } | |
601 | } | |
602 | ||
603 | static void | |
604 | nat_packet(struct dp_packet *pkt, const struct conn *conn, bool related) | |
605 | { | |
606 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
607 | pkt->md.ct_state |= CS_SRC_NAT; | |
608 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
609 | struct ip_header *nh = dp_packet_l3(pkt); | |
610 | packet_set_ipv4_addr(pkt, &nh->ip_src, | |
cda1b109 | 611 | conn->rev_key.dst.addr.ipv4); |
286de272 DB |
612 | } else { |
613 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
614 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
615 | nh6->ip6_src.be32, | |
cda1b109 | 616 | &conn->rev_key.dst.addr.ipv6, true); |
286de272 DB |
617 | } |
618 | if (!related) { | |
619 | pat_packet(pkt, conn); | |
620 | } | |
621 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { | |
622 | pkt->md.ct_state |= CS_DST_NAT; | |
623 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
624 | struct ip_header *nh = dp_packet_l3(pkt); | |
625 | packet_set_ipv4_addr(pkt, &nh->ip_dst, | |
cda1b109 | 626 | conn->rev_key.src.addr.ipv4); |
286de272 DB |
627 | } else { |
628 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
629 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
630 | nh6->ip6_dst.be32, | |
cda1b109 | 631 | &conn->rev_key.src.addr.ipv6, true); |
286de272 DB |
632 | } |
633 | if (!related) { | |
634 | pat_packet(pkt, conn); | |
635 | } | |
636 | } | |
637 | } | |
638 | ||
639 | static void | |
640 | un_pat_packet(struct dp_packet *pkt, const struct conn *conn) | |
641 | { | |
642 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
643 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
644 | struct tcp_header *th = dp_packet_l4(pkt); | |
645 | packet_set_tcp_port(pkt, th->tcp_src, conn->key.src.port); | |
646 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
647 | struct udp_header *uh = dp_packet_l4(pkt); | |
648 | packet_set_udp_port(pkt, uh->udp_src, conn->key.src.port); | |
649 | } | |
650 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { | |
651 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
652 | struct tcp_header *th = dp_packet_l4(pkt); | |
653 | packet_set_tcp_port(pkt, conn->key.dst.port, th->tcp_dst); | |
654 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
655 | struct udp_header *uh = dp_packet_l4(pkt); | |
656 | packet_set_udp_port(pkt, conn->key.dst.port, uh->udp_dst); | |
657 | } | |
658 | } | |
659 | } | |
660 | ||
edd1bef4 DB |
661 | static void |
662 | reverse_pat_packet(struct dp_packet *pkt, const struct conn *conn) | |
663 | { | |
664 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
665 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
666 | struct tcp_header *th_in = dp_packet_l4(pkt); | |
667 | packet_set_tcp_port(pkt, conn->key.src.port, | |
668 | th_in->tcp_dst); | |
669 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
670 | struct udp_header *uh_in = dp_packet_l4(pkt); | |
671 | packet_set_udp_port(pkt, conn->key.src.port, | |
672 | uh_in->udp_dst); | |
673 | } | |
674 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { | |
675 | if (conn->key.nw_proto == IPPROTO_TCP) { | |
676 | struct tcp_header *th_in = dp_packet_l4(pkt); | |
677 | packet_set_tcp_port(pkt, th_in->tcp_src, | |
678 | conn->key.dst.port); | |
679 | } else if (conn->key.nw_proto == IPPROTO_UDP) { | |
680 | struct udp_header *uh_in = dp_packet_l4(pkt); | |
681 | packet_set_udp_port(pkt, uh_in->udp_src, | |
682 | conn->key.dst.port); | |
683 | } | |
684 | } | |
685 | } | |
686 | ||
687 | static void | |
688 | reverse_nat_packet(struct dp_packet *pkt, const struct conn *conn) | |
689 | { | |
690 | char *tail = dp_packet_tail(pkt); | |
ba5ca284 | 691 | uint8_t pad = dp_packet_l2_pad_size(pkt); |
edd1bef4 DB |
692 | struct conn_key inner_key; |
693 | const char *inner_l4 = NULL; | |
694 | uint16_t orig_l3_ofs = pkt->l3_ofs; | |
695 | uint16_t orig_l4_ofs = pkt->l4_ofs; | |
696 | ||
697 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
698 | struct ip_header *nh = dp_packet_l3(pkt); | |
699 | struct icmp_header *icmp = dp_packet_l4(pkt); | |
700 | struct ip_header *inner_l3 = (struct ip_header *) (icmp + 1); | |
ba5ca284 DB |
701 | /* This call is already verified to succeed during the code path from |
702 | * 'conn_key_extract()' which calls 'extract_l4_icmp()'. */ | |
bd5e81a0 DB |
703 | extract_l3_ipv4(&inner_key, inner_l3, tail - ((char *)inner_l3) - pad, |
704 | &inner_l4, false); | |
edd1bef4 DB |
705 | pkt->l3_ofs += (char *) inner_l3 - (char *) nh; |
706 | pkt->l4_ofs += inner_l4 - (char *) icmp; | |
707 | ||
708 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
709 | packet_set_ipv4_addr(pkt, &inner_l3->ip_src, | |
cda1b109 | 710 | conn->key.src.addr.ipv4); |
edd1bef4 DB |
711 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { |
712 | packet_set_ipv4_addr(pkt, &inner_l3->ip_dst, | |
cda1b109 | 713 | conn->key.dst.addr.ipv4); |
edd1bef4 | 714 | } |
dec0dbbc | 715 | |
edd1bef4 DB |
716 | reverse_pat_packet(pkt, conn); |
717 | icmp->icmp_csum = 0; | |
718 | icmp->icmp_csum = csum(icmp, tail - (char *) icmp - pad); | |
719 | } else { | |
720 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
721 | struct icmp6_error_header *icmp6 = dp_packet_l4(pkt); | |
722 | struct ovs_16aligned_ip6_hdr *inner_l3_6 = | |
723 | (struct ovs_16aligned_ip6_hdr *) (icmp6 + 1); | |
ba5ca284 DB |
724 | /* This call is already verified to succeed during the code path from |
725 | * 'conn_key_extract()' which calls 'extract_l4_icmp6()'. */ | |
edd1bef4 DB |
726 | extract_l3_ipv6(&inner_key, inner_l3_6, |
727 | tail - ((char *)inner_l3_6) - pad, | |
728 | &inner_l4); | |
729 | pkt->l3_ofs += (char *) inner_l3_6 - (char *) nh6; | |
730 | pkt->l4_ofs += inner_l4 - (char *) icmp6; | |
731 | ||
732 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
733 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
734 | inner_l3_6->ip6_src.be32, | |
cda1b109 | 735 | &conn->key.src.addr.ipv6, true); |
edd1bef4 DB |
736 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { |
737 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
738 | inner_l3_6->ip6_dst.be32, | |
cda1b109 | 739 | &conn->key.dst.addr.ipv6, true); |
edd1bef4 DB |
740 | } |
741 | reverse_pat_packet(pkt, conn); | |
edd1bef4 | 742 | icmp6->icmp6_base.icmp6_cksum = 0; |
76d85771 DB |
743 | icmp6->icmp6_base.icmp6_cksum = packet_csum_upperlayer6(nh6, icmp6, |
744 | IPPROTO_ICMPV6, tail - (char *) icmp6 - pad); | |
edd1bef4 DB |
745 | } |
746 | pkt->l3_ofs = orig_l3_ofs; | |
747 | pkt->l4_ofs = orig_l4_ofs; | |
748 | } | |
749 | ||
286de272 DB |
750 | static void |
751 | un_nat_packet(struct dp_packet *pkt, const struct conn *conn, | |
752 | bool related) | |
753 | { | |
754 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
755 | pkt->md.ct_state |= CS_DST_NAT; | |
756 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
757 | struct ip_header *nh = dp_packet_l3(pkt); | |
758 | packet_set_ipv4_addr(pkt, &nh->ip_dst, | |
cda1b109 | 759 | conn->key.src.addr.ipv4); |
286de272 DB |
760 | } else { |
761 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
762 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
763 | nh6->ip6_dst.be32, | |
cda1b109 | 764 | &conn->key.src.addr.ipv6, true); |
286de272 | 765 | } |
edd1bef4 DB |
766 | |
767 | if (OVS_UNLIKELY(related)) { | |
768 | reverse_nat_packet(pkt, conn); | |
769 | } else { | |
286de272 DB |
770 | un_pat_packet(pkt, conn); |
771 | } | |
772 | } else if (conn->nat_info->nat_action & NAT_ACTION_DST) { | |
773 | pkt->md.ct_state |= CS_SRC_NAT; | |
774 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
775 | struct ip_header *nh = dp_packet_l3(pkt); | |
776 | packet_set_ipv4_addr(pkt, &nh->ip_src, | |
cda1b109 | 777 | conn->key.dst.addr.ipv4); |
286de272 DB |
778 | } else { |
779 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); | |
780 | packet_set_ipv6_addr(pkt, conn->key.nw_proto, | |
781 | nh6->ip6_src.be32, | |
cda1b109 | 782 | &conn->key.dst.addr.ipv6, true); |
286de272 | 783 | } |
edd1bef4 DB |
784 | |
785 | if (OVS_UNLIKELY(related)) { | |
786 | reverse_nat_packet(pkt, conn); | |
787 | } else { | |
286de272 DB |
788 | un_pat_packet(pkt, conn); |
789 | } | |
790 | } | |
791 | } | |
792 | ||
bd5e81a0 | 793 | static void |
967bb5c5 | 794 | conn_seq_skew_set(struct conntrack *ct, const struct conn *conn_in, |
bd5e81a0 | 795 | long long now, int seq_skew, bool seq_skew_dir) |
967bb5c5 | 796 | OVS_NO_THREAD_SAFETY_ANALYSIS |
bd5e81a0 | 797 | { |
967bb5c5 | 798 | struct conn *conn; |
967bb5c5 | 799 | ovs_mutex_unlock(&conn_in->lock); |
4048c508 | 800 | conn_lookup(ct, &conn_in->key, now, &conn, NULL); |
967bb5c5 DB |
801 | ovs_mutex_lock(&conn_in->lock); |
802 | ||
bd5e81a0 DB |
803 | if (conn && seq_skew) { |
804 | conn->seq_skew = seq_skew; | |
805 | conn->seq_skew_dir = seq_skew_dir; | |
806 | } | |
a720a7fa DB |
807 | } |
808 | ||
3a2a425b DB |
809 | static bool |
810 | ct_verify_helper(const char *helper, enum ct_alg_ctl_type ct_alg_ctl) | |
811 | { | |
812 | if (ct_alg_ctl == CT_ALG_CTL_NONE) { | |
813 | return true; | |
814 | } else if (helper) { | |
815 | if ((ct_alg_ctl == CT_ALG_CTL_FTP) && | |
816 | !strncmp(helper, "ftp", strlen("ftp"))) { | |
817 | return true; | |
818 | } else if ((ct_alg_ctl == CT_ALG_CTL_TFTP) && | |
819 | !strncmp(helper, "tftp", strlen("tftp"))) { | |
820 | return true; | |
821 | } else { | |
822 | return false; | |
823 | } | |
824 | } else { | |
825 | return false; | |
826 | } | |
827 | } | |
828 | ||
a489b168 DDP |
829 | static struct conn * |
830 | conn_not_found(struct conntrack *ct, struct dp_packet *pkt, | |
286de272 DB |
831 | struct conn_lookup_ctx *ctx, bool commit, long long now, |
832 | const struct nat_action_info_t *nat_action_info, | |
967bb5c5 | 833 | const char *helper, const struct alg_exp_node *alg_exp, |
3a2a425b | 834 | enum ct_alg_ctl_type ct_alg_ctl) |
967bb5c5 | 835 | OVS_REQUIRES(ct->ct_lock) |
a489b168 | 836 | { |
a489b168 | 837 | struct conn *nc = NULL; |
967bb5c5 | 838 | struct conn *nat_conn = NULL; |
a489b168 DDP |
839 | |
840 | if (!valid_new(pkt, &ctx->key)) { | |
286de272 | 841 | pkt->md.ct_state = CS_INVALID; |
a489b168 DDP |
842 | return nc; |
843 | } | |
dec0dbbc | 844 | |
286de272 | 845 | pkt->md.ct_state = CS_NEW; |
dec0dbbc | 846 | |
bd5e81a0 DB |
847 | if (alg_exp) { |
848 | pkt->md.ct_state |= CS_RELATED; | |
849 | } | |
a489b168 DDP |
850 | |
851 | if (commit) { | |
852 | unsigned int n_conn_limit; | |
a489b168 | 853 | atomic_read_relaxed(&ct->n_conn_limit, &n_conn_limit); |
a489b168 DDP |
854 | if (atomic_count_get(&ct->n_conn) >= n_conn_limit) { |
855 | COVERAGE_INC(conntrack_full); | |
856 | return nc; | |
857 | } | |
858 | ||
967bb5c5 | 859 | nc = new_conn(ct, pkt, &ctx->key, now); |
a720a7fa | 860 | memcpy(&nc->key, &ctx->key, sizeof nc->key); |
82b9ac94 | 861 | memcpy(&nc->rev_key, &nc->key, sizeof nc->rev_key); |
286de272 | 862 | conn_key_reverse(&nc->rev_key); |
a489b168 | 863 | |
3a2a425b DB |
864 | if (ct_verify_helper(helper, ct_alg_ctl)) { |
865 | nc->alg = nullable_xstrdup(helper); | |
bd5e81a0 DB |
866 | } |
867 | ||
868 | if (alg_exp) { | |
869 | nc->alg_related = true; | |
870 | nc->mark = alg_exp->master_mark; | |
871 | nc->label = alg_exp->master_label; | |
872 | nc->master_key = alg_exp->master_key; | |
873 | } | |
874 | ||
286de272 DB |
875 | if (nat_action_info) { |
876 | nc->nat_info = xmemdup(nat_action_info, sizeof *nc->nat_info); | |
967bb5c5 | 877 | nat_conn = xzalloc(sizeof *nat_conn); |
a489b168 | 878 | |
bd5e81a0 | 879 | if (alg_exp) { |
be38342d | 880 | if (alg_exp->nat_rpl_dst) { |
bd5e81a0 DB |
881 | nc->rev_key.dst.addr = alg_exp->alg_nat_repl_addr; |
882 | nc->nat_info->nat_action = NAT_ACTION_SRC; | |
883 | } else { | |
884 | nc->rev_key.src.addr = alg_exp->alg_nat_repl_addr; | |
885 | nc->nat_info->nat_action = NAT_ACTION_DST; | |
886 | } | |
bd5e81a0 | 887 | } else { |
967bb5c5 DB |
888 | memcpy(nat_conn, nc, sizeof *nat_conn); |
889 | bool nat_res = nat_select_range_tuple(ct, nc, nat_conn); | |
286de272 | 890 | |
bd5e81a0 DB |
891 | if (!nat_res) { |
892 | goto nat_res_exhaustion; | |
893 | } | |
286de272 | 894 | |
967bb5c5 DB |
895 | /* Update nc with nat adjustments made to nat_conn by |
896 | * nat_select_range_tuple(). */ | |
897 | memcpy(nc, nat_conn, sizeof *nc); | |
286de272 | 898 | } |
967bb5c5 | 899 | |
dbb597d3 | 900 | nat_packet(pkt, nc, ctx->icmp_related); |
967bb5c5 DB |
901 | memcpy(&nat_conn->key, &nc->rev_key, sizeof nat_conn->key); |
902 | memcpy(&nat_conn->rev_key, &nc->key, sizeof nat_conn->rev_key); | |
903 | nat_conn->conn_type = CT_CONN_TYPE_UN_NAT; | |
904 | nat_conn->nat_info = NULL; | |
905 | nat_conn->alg = NULL; | |
906 | nat_conn->nat_conn = NULL; | |
907 | uint32_t nat_hash = conn_key_hash(&nat_conn->key, ct->hash_basis); | |
908 | cmap_insert(&ct->conns, &nat_conn->cm_node, nat_hash); | |
909 | } | |
910 | ||
911 | nc->nat_conn = nat_conn; | |
912 | ovs_mutex_init_adaptive(&nc->lock); | |
913 | nc->conn_type = CT_CONN_TYPE_DEFAULT; | |
914 | cmap_insert(&ct->conns, &nc->cm_node, ctx->hash); | |
a489b168 | 915 | atomic_count_inc(&ct->n_conn); |
967bb5c5 | 916 | ctx->conn = nc; /* For completeness. */ |
a489b168 | 917 | } |
bd5e81a0 | 918 | |
a489b168 | 919 | return nc; |
bd5e81a0 | 920 | |
967bb5c5 DB |
921 | /* This would be a user error or a DOS attack. A user error is prevented |
922 | * by allocating enough combinations of NAT addresses when combined with | |
923 | * ephemeral ports. A DOS attack should be protected against with | |
924 | * firewall rules or a separate firewall. Also using zone partitioning | |
925 | * can limit DoS impact. */ | |
bd5e81a0 | 926 | nat_res_exhaustion: |
967bb5c5 DB |
927 | free(nat_conn); |
928 | ovs_list_remove(&nc->exp_node); | |
929 | delete_conn_cmn(nc); | |
bd5e81a0 DB |
930 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5); |
931 | VLOG_WARN_RL(&rl, "Unable to NAT due to tuple space exhaustion - " | |
932 | "if DoS attack, use firewalling and/or zone partitioning."); | |
933 | return NULL; | |
a489b168 DDP |
934 | } |
935 | ||
286de272 DB |
936 | static bool |
937 | conn_update_state(struct conntrack *ct, struct dp_packet *pkt, | |
967bb5c5 DB |
938 | struct conn_lookup_ctx *ctx, struct conn *conn, |
939 | long long now) | |
286de272 | 940 | { |
967bb5c5 | 941 | ovs_assert(conn->conn_type == CT_CONN_TYPE_DEFAULT); |
286de272 DB |
942 | bool create_new_conn = false; |
943 | ||
dbb597d3 | 944 | if (ctx->icmp_related) { |
286de272 DB |
945 | pkt->md.ct_state |= CS_RELATED; |
946 | if (ctx->reply) { | |
947 | pkt->md.ct_state |= CS_REPLY_DIR; | |
948 | } | |
949 | } else { | |
967bb5c5 | 950 | if (conn->alg_related) { |
bd5e81a0 DB |
951 | pkt->md.ct_state |= CS_RELATED; |
952 | } | |
dec0dbbc | 953 | |
967bb5c5 | 954 | enum ct_update_res res = conn_update(ct, conn, pkt, ctx, now); |
286de272 DB |
955 | |
956 | switch (res) { | |
957 | case CT_UPDATE_VALID: | |
958 | pkt->md.ct_state |= CS_ESTABLISHED; | |
959 | pkt->md.ct_state &= ~CS_NEW; | |
960 | if (ctx->reply) { | |
961 | pkt->md.ct_state |= CS_REPLY_DIR; | |
962 | } | |
963 | break; | |
964 | case CT_UPDATE_INVALID: | |
965 | pkt->md.ct_state = CS_INVALID; | |
966 | break; | |
967 | case CT_UPDATE_NEW: | |
967bb5c5 | 968 | ovs_mutex_lock(&ct->ct_lock); |
4048c508 | 969 | if (conn_lookup(ct, &conn->key, now, NULL, NULL)) { |
28274f77 DB |
970 | conn_clean(ct, conn); |
971 | } | |
967bb5c5 | 972 | ovs_mutex_unlock(&ct->ct_lock); |
286de272 DB |
973 | create_new_conn = true; |
974 | break; | |
975 | default: | |
976 | OVS_NOT_REACHED(); | |
977 | } | |
978 | } | |
979 | return create_new_conn; | |
980 | } | |
981 | ||
286de272 DB |
982 | static void |
983 | handle_nat(struct dp_packet *pkt, struct conn *conn, | |
984 | uint16_t zone, bool reply, bool related) | |
985 | { | |
986 | if (conn->nat_info && | |
987 | (!(pkt->md.ct_state & (CS_SRC_NAT | CS_DST_NAT)) || | |
988 | (pkt->md.ct_state & (CS_SRC_NAT | CS_DST_NAT) && | |
989 | zone != pkt->md.ct_zone))) { | |
bd5e81a0 | 990 | |
286de272 DB |
991 | if (pkt->md.ct_state & (CS_SRC_NAT | CS_DST_NAT)) { |
992 | pkt->md.ct_state &= ~(CS_SRC_NAT | CS_DST_NAT); | |
993 | } | |
994 | if (reply) { | |
995 | un_nat_packet(pkt, conn, related); | |
996 | } else { | |
997 | nat_packet(pkt, conn, related); | |
998 | } | |
999 | } | |
1000 | } | |
1001 | ||
f8016041 DB |
1002 | static bool |
1003 | check_orig_tuple(struct conntrack *ct, struct dp_packet *pkt, | |
1004 | struct conn_lookup_ctx *ctx_in, long long now, | |
967bb5c5 | 1005 | struct conn **conn, |
f8016041 | 1006 | const struct nat_action_info_t *nat_action_info) |
f8016041 | 1007 | { |
a0b89c51 DB |
1008 | if (!(pkt->md.ct_state & (CS_SRC_NAT | CS_DST_NAT)) || |
1009 | (ctx_in->key.dl_type == htons(ETH_TYPE_IP) && | |
f8016041 DB |
1010 | !pkt->md.ct_orig_tuple.ipv4.ipv4_proto) || |
1011 | (ctx_in->key.dl_type == htons(ETH_TYPE_IPV6) && | |
1012 | !pkt->md.ct_orig_tuple.ipv6.ipv6_proto) || | |
f8016041 DB |
1013 | nat_action_info) { |
1014 | return false; | |
1015 | } | |
1016 | ||
967bb5c5 DB |
1017 | struct conn_key key; |
1018 | memset(&key, 0 , sizeof key); | |
f8016041 DB |
1019 | |
1020 | if (ctx_in->key.dl_type == htons(ETH_TYPE_IP)) { | |
967bb5c5 DB |
1021 | key.src.addr.ipv4 = pkt->md.ct_orig_tuple.ipv4.ipv4_src; |
1022 | key.dst.addr.ipv4 = pkt->md.ct_orig_tuple.ipv4.ipv4_dst; | |
f8016041 DB |
1023 | |
1024 | if (ctx_in->key.nw_proto == IPPROTO_ICMP) { | |
967bb5c5 DB |
1025 | key.src.icmp_id = ctx_in->key.src.icmp_id; |
1026 | key.dst.icmp_id = ctx_in->key.dst.icmp_id; | |
f8016041 | 1027 | uint16_t src_port = ntohs(pkt->md.ct_orig_tuple.ipv4.src_port); |
967bb5c5 DB |
1028 | key.src.icmp_type = (uint8_t) src_port; |
1029 | key.dst.icmp_type = reverse_icmp_type(key.src.icmp_type); | |
f8016041 | 1030 | } else { |
967bb5c5 DB |
1031 | key.src.port = pkt->md.ct_orig_tuple.ipv4.src_port; |
1032 | key.dst.port = pkt->md.ct_orig_tuple.ipv4.dst_port; | |
f8016041 | 1033 | } |
967bb5c5 | 1034 | key.nw_proto = pkt->md.ct_orig_tuple.ipv4.ipv4_proto; |
f8016041 | 1035 | } else { |
967bb5c5 DB |
1036 | key.src.addr.ipv6 = pkt->md.ct_orig_tuple.ipv6.ipv6_src; |
1037 | key.dst.addr.ipv6 = pkt->md.ct_orig_tuple.ipv6.ipv6_dst; | |
f8016041 DB |
1038 | |
1039 | if (ctx_in->key.nw_proto == IPPROTO_ICMPV6) { | |
967bb5c5 DB |
1040 | key.src.icmp_id = ctx_in->key.src.icmp_id; |
1041 | key.dst.icmp_id = ctx_in->key.dst.icmp_id; | |
f8016041 | 1042 | uint16_t src_port = ntohs(pkt->md.ct_orig_tuple.ipv6.src_port); |
967bb5c5 DB |
1043 | key.src.icmp_type = (uint8_t) src_port; |
1044 | key.dst.icmp_type = reverse_icmp6_type(key.src.icmp_type); | |
f8016041 | 1045 | } else { |
967bb5c5 DB |
1046 | key.src.port = pkt->md.ct_orig_tuple.ipv6.src_port; |
1047 | key.dst.port = pkt->md.ct_orig_tuple.ipv6.dst_port; | |
f8016041 | 1048 | } |
967bb5c5 | 1049 | key.nw_proto = pkt->md.ct_orig_tuple.ipv6.ipv6_proto; |
f8016041 DB |
1050 | } |
1051 | ||
967bb5c5 DB |
1052 | key.dl_type = ctx_in->key.dl_type; |
1053 | key.zone = pkt->md.ct_zone; | |
4048c508 | 1054 | conn_lookup(ct, &key, now, conn, NULL); |
f8016041 DB |
1055 | return *conn ? true : false; |
1056 | } | |
1057 | ||
94e71143 DB |
1058 | static bool |
1059 | conn_update_state_alg(struct conntrack *ct, struct dp_packet *pkt, | |
1060 | struct conn_lookup_ctx *ctx, struct conn *conn, | |
1061 | const struct nat_action_info_t *nat_action_info, | |
1062 | enum ct_alg_ctl_type ct_alg_ctl, long long now, | |
967bb5c5 | 1063 | bool *create_new_conn) |
94e71143 DB |
1064 | { |
1065 | if (is_ftp_ctl(ct_alg_ctl)) { | |
1066 | /* Keep sequence tracking in sync with the source of the | |
1067 | * sequence skew. */ | |
967bb5c5 | 1068 | ovs_mutex_lock(&conn->lock); |
94e71143 DB |
1069 | if (ctx->reply != conn->seq_skew_dir) { |
1070 | handle_ftp_ctl(ct, ctx, pkt, conn, now, CT_FTP_CTL_OTHER, | |
1071 | !!nat_action_info); | |
967bb5c5 DB |
1072 | /* conn_update_state locks for unrelated fields, so unlock. */ |
1073 | ovs_mutex_unlock(&conn->lock); | |
1074 | *create_new_conn = conn_update_state(ct, pkt, ctx, conn, now); | |
94e71143 | 1075 | } else { |
967bb5c5 DB |
1076 | /* conn_update_state locks for unrelated fields, so unlock. */ |
1077 | ovs_mutex_unlock(&conn->lock); | |
1078 | *create_new_conn = conn_update_state(ct, pkt, ctx, conn, now); | |
1079 | ovs_mutex_lock(&conn->lock); | |
030958a0 DB |
1080 | if (*create_new_conn == false) { |
1081 | handle_ftp_ctl(ct, ctx, pkt, conn, now, CT_FTP_CTL_OTHER, | |
1082 | !!nat_action_info); | |
1083 | } | |
967bb5c5 | 1084 | ovs_mutex_unlock(&conn->lock); |
94e71143 DB |
1085 | } |
1086 | return true; | |
1087 | } | |
1088 | return false; | |
1089 | } | |
1090 | ||
286de272 | 1091 | static void |
a489b168 DDP |
1092 | process_one(struct conntrack *ct, struct dp_packet *pkt, |
1093 | struct conn_lookup_ctx *ctx, uint16_t zone, | |
286de272 DB |
1094 | bool force, bool commit, long long now, const uint32_t *setmark, |
1095 | const struct ovs_key_ct_labels *setlabel, | |
bd5e81a0 | 1096 | const struct nat_action_info_t *nat_action_info, |
bd7d93f8 | 1097 | ovs_be16 tp_src, ovs_be16 tp_dst, const char *helper) |
a489b168 | 1098 | { |
967bb5c5 DB |
1099 | bool create_new_conn = false; |
1100 | conn_key_lookup(ct, &ctx->key, ctx->hash, now, &ctx->conn, &ctx->reply); | |
1101 | struct conn *conn = ctx->conn; | |
a489b168 | 1102 | |
a76a37ef | 1103 | /* Delete found entry if in wrong direction. 'force' implies commit. */ |
a720a7fa | 1104 | if (OVS_UNLIKELY(force && ctx->reply && conn)) { |
967bb5c5 | 1105 | ovs_mutex_lock(&ct->ct_lock); |
4048c508 | 1106 | if (conn_lookup(ct, &conn->key, now, NULL, NULL)) { |
28274f77 DB |
1107 | conn_clean(ct, conn); |
1108 | } | |
967bb5c5 | 1109 | ovs_mutex_unlock(&ct->ct_lock); |
a76a37ef JR |
1110 | conn = NULL; |
1111 | } | |
1112 | ||
286de272 DB |
1113 | if (OVS_LIKELY(conn)) { |
1114 | if (conn->conn_type == CT_CONN_TYPE_UN_NAT) { | |
a489b168 | 1115 | |
286de272 | 1116 | ctx->reply = true; |
967bb5c5 | 1117 | struct conn *rev_conn = conn; /* Save for debugging. */ |
4048c508 | 1118 | uint32_t hash = conn_key_hash(&conn->rev_key, ct->hash_basis); |
967bb5c5 | 1119 | conn_key_lookup(ct, &ctx->key, hash, now, &conn, &ctx->reply); |
a489b168 | 1120 | |
967bb5c5 | 1121 | if (!conn) { |
286de272 | 1122 | pkt->md.ct_state |= CS_TRACKED | CS_INVALID; |
967bb5c5 DB |
1123 | char *log_msg = xasprintf("Missing master conn %p", rev_conn); |
1124 | ct_print_conn_info(conn, log_msg, VLL_INFO, true, true); | |
1125 | free(log_msg); | |
286de272 | 1126 | return; |
a489b168 DDP |
1127 | } |
1128 | } | |
286de272 DB |
1129 | } |
1130 | ||
bd7d93f8 DB |
1131 | enum ct_alg_ctl_type ct_alg_ctl = get_alg_ctl_type(pkt, tp_src, tp_dst, |
1132 | helper); | |
bd5e81a0 | 1133 | |
286de272 | 1134 | if (OVS_LIKELY(conn)) { |
94e71143 DB |
1135 | if (OVS_LIKELY(!conn_update_state_alg(ct, pkt, ctx, conn, |
1136 | nat_action_info, | |
967bb5c5 | 1137 | ct_alg_ctl, now, |
94e71143 | 1138 | &create_new_conn))) { |
967bb5c5 | 1139 | create_new_conn = conn_update_state(ct, pkt, ctx, conn, now); |
bd5e81a0 | 1140 | } |
286de272 | 1141 | if (nat_action_info && !create_new_conn) { |
dbb597d3 | 1142 | handle_nat(pkt, conn, zone, ctx->reply, ctx->icmp_related); |
286de272 | 1143 | } |
bd5e81a0 | 1144 | |
a0b89c51 | 1145 | } else if (check_orig_tuple(ct, pkt, ctx, now, &conn, nat_action_info)) { |
967bb5c5 | 1146 | create_new_conn = conn_update_state(ct, pkt, ctx, conn, now); |
a489b168 | 1147 | } else { |
dbb597d3 | 1148 | if (ctx->icmp_related) { |
bd5e81a0 DB |
1149 | /* An icmp related conn should always be found; no new |
1150 | connection is created based on an icmp related packet. */ | |
286de272 | 1151 | pkt->md.ct_state = CS_INVALID; |
5c2e106b | 1152 | } else { |
286de272 | 1153 | create_new_conn = true; |
5c2e106b | 1154 | } |
a489b168 DDP |
1155 | } |
1156 | ||
bd5e81a0 | 1157 | const struct alg_exp_node *alg_exp = NULL; |
96bbcbf7 | 1158 | struct alg_exp_node alg_exp_entry; |
dec0dbbc | 1159 | |
286de272 | 1160 | if (OVS_UNLIKELY(create_new_conn)) { |
bd5e81a0 | 1161 | |
967bb5c5 | 1162 | ovs_rwlock_rdlock(&ct->resources_lock); |
bd5e81a0 | 1163 | alg_exp = expectation_lookup(&ct->alg_expectations, &ctx->key, |
be38342d DB |
1164 | ct->hash_basis, |
1165 | alg_src_ip_wc(ct_alg_ctl)); | |
bd5e81a0 | 1166 | if (alg_exp) { |
c3f6bae2 | 1167 | memcpy(&alg_exp_entry, alg_exp, sizeof alg_exp_entry); |
bd5e81a0 DB |
1168 | alg_exp = &alg_exp_entry; |
1169 | } | |
967bb5c5 | 1170 | ovs_rwlock_unlock(&ct->resources_lock); |
bd5e81a0 | 1171 | |
967bb5c5 | 1172 | ovs_mutex_lock(&ct->ct_lock); |
4048c508 | 1173 | if (!conn_lookup(ct, &ctx->key, now, NULL, NULL)) { |
28274f77 DB |
1174 | conn = conn_not_found(ct, pkt, ctx, commit, now, nat_action_info, |
1175 | helper, alg_exp, ct_alg_ctl); | |
1176 | } | |
967bb5c5 | 1177 | ovs_mutex_unlock(&ct->ct_lock); |
286de272 DB |
1178 | } |
1179 | ||
bd5e81a0 DB |
1180 | write_ct_md(pkt, zone, conn, &ctx->key, alg_exp); |
1181 | ||
286de272 DB |
1182 | if (conn && setmark) { |
1183 | set_mark(pkt, conn, setmark[0], setmark[1]); | |
1184 | } | |
a489b168 | 1185 | |
286de272 DB |
1186 | if (conn && setlabel) { |
1187 | set_label(pkt, conn, &setlabel[0], &setlabel[1]); | |
1188 | } | |
1189 | ||
967bb5c5 | 1190 | handle_alg_ctl(ct, ctx, pkt, ct_alg_ctl, conn, now, !!nat_action_info); |
a489b168 DDP |
1191 | } |
1192 | ||
1193 | /* Sends the packets in '*pkt_batch' through the connection tracker 'ct'. All | |
51b9a533 | 1194 | * the packets must have the same 'dl_type' (IPv4 or IPv6) and should have |
4ea96698 DB |
1195 | * the l3 and and l4 offset properly set. Performs fragment reassembly with |
1196 | * the help of ipf_preprocess_conntrack(). | |
a489b168 DDP |
1197 | * |
1198 | * If 'commit' is true, the packets are allowed to create new entries in the | |
1199 | * connection tables. 'setmark', if not NULL, should point to a two | |
1200 | * elements array containing a value and a mask to set the connection mark. | |
1201 | * 'setlabel' behaves similarly for the connection label.*/ | |
1202 | int | |
1203 | conntrack_execute(struct conntrack *ct, struct dp_packet_batch *pkt_batch, | |
a76a37ef | 1204 | ovs_be16 dl_type, bool force, bool commit, uint16_t zone, |
66e4ad8a | 1205 | const uint32_t *setmark, |
a489b168 | 1206 | const struct ovs_key_ct_labels *setlabel, |
bd7d93f8 | 1207 | ovs_be16 tp_src, ovs_be16 tp_dst, const char *helper, |
94053e66 FA |
1208 | const struct nat_action_info_t *nat_action_info, |
1209 | long long now) | |
a489b168 | 1210 | { |
4ea96698 DB |
1211 | ipf_preprocess_conntrack(ct->ipf, pkt_batch, now, dl_type, zone, |
1212 | ct->hash_basis); | |
1213 | ||
43495c45 | 1214 | struct dp_packet *packet; |
61ce32b9 | 1215 | struct conn_lookup_ctx ctx; |
a489b168 | 1216 | |
e883448e | 1217 | DP_PACKET_BATCH_FOR_EACH (i, packet, pkt_batch) { |
4ea96698 DB |
1218 | if (packet->md.ct_state == CS_INVALID |
1219 | || !conn_key_extract(ct, packet, dl_type, &ctx, zone)) { | |
43495c45 BB |
1220 | packet->md.ct_state = CS_INVALID; |
1221 | write_ct_md(packet, zone, NULL, NULL, NULL); | |
a489b168 DDP |
1222 | continue; |
1223 | } | |
94e71143 | 1224 | process_one(ct, packet, &ctx, zone, force, commit, now, setmark, |
bd7d93f8 | 1225 | setlabel, nat_action_info, tp_src, tp_dst, helper); |
a489b168 DDP |
1226 | } |
1227 | ||
4ea96698 DB |
1228 | ipf_postprocess_conntrack(ct->ipf, pkt_batch, now, dl_type); |
1229 | ||
a489b168 DDP |
1230 | return 0; |
1231 | } | |
1232 | ||
1fe178d2 EG |
1233 | void |
1234 | conntrack_clear(struct dp_packet *packet) | |
1235 | { | |
1236 | /* According to pkt_metadata_init(), ct_state == 0 is enough to make all of | |
1237 | * the conntrack fields invalid. */ | |
1238 | packet->md.ct_state = 0; | |
1239 | } | |
1240 | ||
a489b168 DDP |
1241 | static void |
1242 | set_mark(struct dp_packet *pkt, struct conn *conn, uint32_t val, uint32_t mask) | |
1243 | { | |
967bb5c5 | 1244 | ovs_mutex_lock(&conn->lock); |
bd5e81a0 DB |
1245 | if (conn->alg_related) { |
1246 | pkt->md.ct_mark = conn->mark; | |
1247 | } else { | |
1248 | pkt->md.ct_mark = val | (pkt->md.ct_mark & ~(mask)); | |
1249 | conn->mark = pkt->md.ct_mark; | |
1250 | } | |
967bb5c5 | 1251 | ovs_mutex_unlock(&conn->lock); |
a489b168 DDP |
1252 | } |
1253 | ||
1254 | static void | |
1255 | set_label(struct dp_packet *pkt, struct conn *conn, | |
1256 | const struct ovs_key_ct_labels *val, | |
1257 | const struct ovs_key_ct_labels *mask) | |
1258 | { | |
967bb5c5 | 1259 | ovs_mutex_lock(&conn->lock); |
bd5e81a0 DB |
1260 | if (conn->alg_related) { |
1261 | pkt->md.ct_label = conn->label; | |
1262 | } else { | |
1263 | ovs_u128 v, m; | |
a489b168 | 1264 | |
bd5e81a0 DB |
1265 | memcpy(&v, val, sizeof v); |
1266 | memcpy(&m, mask, sizeof m); | |
a489b168 | 1267 | |
bd5e81a0 | 1268 | pkt->md.ct_label.u64.lo = v.u64.lo |
a489b168 | 1269 | | (pkt->md.ct_label.u64.lo & ~(m.u64.lo)); |
bd5e81a0 | 1270 | pkt->md.ct_label.u64.hi = v.u64.hi |
a489b168 | 1271 | | (pkt->md.ct_label.u64.hi & ~(m.u64.hi)); |
bd5e81a0 DB |
1272 | conn->label = pkt->md.ct_label; |
1273 | } | |
967bb5c5 | 1274 | ovs_mutex_unlock(&conn->lock); |
a489b168 | 1275 | } |
286de272 | 1276 | |
a489b168 | 1277 | \f |
e6ef6cc6 DDP |
1278 | /* Delete the expired connections from 'ctb', up to 'limit'. Returns the |
1279 | * earliest expiration time among the remaining connections in 'ctb'. Returns | |
1280 | * LLONG_MAX if 'ctb' is empty. The return value might be smaller than 'now', | |
1281 | * if 'limit' is reached */ | |
1282 | static long long | |
967bb5c5 | 1283 | ct_sweep(struct conntrack *ct, long long now, size_t limit) |
e6ef6cc6 DDP |
1284 | { |
1285 | struct conn *conn, *next; | |
1286 | long long min_expiration = LLONG_MAX; | |
e6ef6cc6 DDP |
1287 | size_t count = 0; |
1288 | ||
967bb5c5 DB |
1289 | ovs_mutex_lock(&ct->ct_lock); |
1290 | ||
dec0dbbc | 1291 | for (unsigned i = 0; i < N_CT_TM; i++) { |
967bb5c5 DB |
1292 | LIST_FOR_EACH_SAFE (conn, next, exp_node, &ct->exp_lists[i]) { |
1293 | ovs_mutex_lock(&conn->lock); | |
1294 | if (now < conn->expiration || count >= limit) { | |
a720a7fa | 1295 | min_expiration = MIN(min_expiration, conn->expiration); |
967bb5c5 | 1296 | ovs_mutex_unlock(&conn->lock); |
a720a7fa DB |
1297 | if (count >= limit) { |
1298 | /* Do not check other lists. */ | |
1299 | COVERAGE_INC(conntrack_long_cleanup); | |
967bb5c5 | 1300 | goto out; |
e6ef6cc6 | 1301 | } |
a720a7fa | 1302 | break; |
967bb5c5 DB |
1303 | } else { |
1304 | ovs_mutex_unlock(&conn->lock); | |
1305 | conn_clean(ct, conn); | |
e6ef6cc6 | 1306 | } |
a720a7fa | 1307 | count++; |
e6ef6cc6 DDP |
1308 | } |
1309 | } | |
967bb5c5 DB |
1310 | |
1311 | out: | |
1312 | VLOG_DBG("conntrack cleanup %"PRIuSIZE" entries in %lld msec", count, | |
1313 | time_msec() - now); | |
1314 | ovs_mutex_unlock(&ct->ct_lock); | |
e6ef6cc6 DDP |
1315 | return min_expiration; |
1316 | } | |
1317 | ||
1318 | /* Cleans up old connection entries from 'ct'. Returns the time when the | |
1319 | * next expiration might happen. The return value might be smaller than | |
1320 | * 'now', meaning that an internal limit has been reached, and some expired | |
1321 | * connections have not been deleted. */ | |
1322 | static long long | |
1323 | conntrack_clean(struct conntrack *ct, long long now) | |
1324 | { | |
e6ef6cc6 | 1325 | unsigned int n_conn_limit; |
e6ef6cc6 | 1326 | atomic_read_relaxed(&ct->n_conn_limit, &n_conn_limit); |
967bb5c5 DB |
1327 | size_t clean_max = n_conn_limit > 10 ? n_conn_limit / 10 : 1; |
1328 | long long min_exp = ct_sweep(ct, now, clean_max); | |
1329 | long long next_wakeup = MIN(min_exp, now + CT_TM_MIN); | |
e6ef6cc6 DDP |
1330 | |
1331 | return next_wakeup; | |
1332 | } | |
1333 | ||
1334 | /* Cleanup: | |
e6ef6cc6 DDP |
1335 | * |
1336 | * We must call conntrack_clean() periodically. conntrack_clean() return | |
1337 | * value gives an hint on when the next cleanup must be done (either because | |
1338 | * there is an actual connection that expires, or because a new connection | |
1339 | * might be created with the minimum timeout). | |
1340 | * | |
1341 | * The logic below has two goals: | |
1342 | * | |
6c54734e DDP |
1343 | * - We want to reduce the number of wakeups and batch connection cleanup |
1344 | * when the load is not very high. CT_CLEAN_INTERVAL ensures that if we | |
1345 | * are coping with the current cleanup tasks, then we wait at least | |
1346 | * 5 seconds to do further cleanup. | |
e6ef6cc6 | 1347 | * |
967bb5c5 | 1348 | * - We don't want to keep the map locked too long, as we might prevent |
6c54734e | 1349 | * traffic from flowing. CT_CLEAN_MIN_INTERVAL ensures that if cleanup is |
967bb5c5 | 1350 | * behind, there is at least some 200ms blocks of time when the map will be |
6c54734e | 1351 | * left alone, so the datapath can operate unhindered. |
e6ef6cc6 DDP |
1352 | */ |
1353 | #define CT_CLEAN_INTERVAL 5000 /* 5 seconds */ | |
1354 | #define CT_CLEAN_MIN_INTERVAL 200 /* 0.2 seconds */ | |
1355 | ||
1356 | static void * | |
1357 | clean_thread_main(void *f_) | |
1358 | { | |
1359 | struct conntrack *ct = f_; | |
1360 | ||
1361 | while (!latch_is_set(&ct->clean_thread_exit)) { | |
1362 | long long next_wake; | |
1363 | long long now = time_msec(); | |
e6ef6cc6 DDP |
1364 | next_wake = conntrack_clean(ct, now); |
1365 | ||
1366 | if (next_wake < now) { | |
1367 | poll_timer_wait_until(now + CT_CLEAN_MIN_INTERVAL); | |
1368 | } else { | |
1369 | poll_timer_wait_until(MAX(next_wake, now + CT_CLEAN_INTERVAL)); | |
1370 | } | |
1371 | latch_wait(&ct->clean_thread_exit); | |
1372 | poll_block(); | |
1373 | } | |
1374 | ||
1375 | return NULL; | |
1376 | } | |
1377 | \f | |
e917d3ee DB |
1378 | /* 'Data' is a pointer to the beginning of the L3 header and 'new_data' is |
1379 | * used to store a pointer to the first byte after the L3 header. 'Size' is | |
1380 | * the size of the packet beyond the data pointer. */ | |
a489b168 DDP |
1381 | static inline bool |
1382 | extract_l3_ipv4(struct conn_key *key, const void *data, size_t size, | |
1383 | const char **new_data, bool validate_checksum) | |
1384 | { | |
e917d3ee DB |
1385 | if (OVS_UNLIKELY(size < IP_HEADER_LEN)) { |
1386 | return false; | |
a489b168 DDP |
1387 | } |
1388 | ||
dec0dbbc DB |
1389 | const struct ip_header *ip = data; |
1390 | size_t ip_len = IP_IHL(ip->ip_ihl_ver) * 4; | |
a489b168 | 1391 | |
e917d3ee DB |
1392 | if (OVS_UNLIKELY(ip_len < IP_HEADER_LEN)) { |
1393 | return false; | |
1394 | } | |
a489b168 | 1395 | |
e917d3ee DB |
1396 | if (OVS_UNLIKELY(size < ip_len)) { |
1397 | return false; | |
1398 | } | |
a489b168 | 1399 | |
e917d3ee DB |
1400 | if (IP_IS_FRAGMENT(ip->ip_frag_off)) { |
1401 | return false; | |
a489b168 DDP |
1402 | } |
1403 | ||
1404 | if (validate_checksum && csum(data, ip_len) != 0) { | |
1405 | return false; | |
1406 | } | |
1407 | ||
e917d3ee DB |
1408 | if (new_data) { |
1409 | *new_data = (char *) data + ip_len; | |
1410 | } | |
1411 | ||
cda1b109 DB |
1412 | key->src.addr.ipv4 = get_16aligned_be32(&ip->ip_src); |
1413 | key->dst.addr.ipv4 = get_16aligned_be32(&ip->ip_dst); | |
a489b168 DDP |
1414 | key->nw_proto = ip->ip_proto; |
1415 | ||
1416 | return true; | |
1417 | } | |
1418 | ||
e917d3ee DB |
1419 | /* 'Data' is a pointer to the beginning of the L3 header and 'new_data' is |
1420 | * used to store a pointer to the first byte after the L3 header. 'Size' is | |
1421 | * the size of the packet beyond the data pointer. */ | |
a489b168 DDP |
1422 | static inline bool |
1423 | extract_l3_ipv6(struct conn_key *key, const void *data, size_t size, | |
1424 | const char **new_data) | |
1425 | { | |
1426 | const struct ovs_16aligned_ip6_hdr *ip6 = data; | |
286de272 | 1427 | |
e917d3ee DB |
1428 | if (OVS_UNLIKELY(size < sizeof *ip6)) { |
1429 | return false; | |
a489b168 DDP |
1430 | } |
1431 | ||
1432 | data = ip6 + 1; | |
1433 | size -= sizeof *ip6; | |
dec0dbbc DB |
1434 | uint8_t nw_proto = ip6->ip6_nxt; |
1435 | uint8_t nw_frag = 0; | |
a489b168 | 1436 | |
523464ab DB |
1437 | const struct ovs_16aligned_ip6_frag *frag_hdr; |
1438 | if (!parse_ipv6_ext_hdrs(&data, &size, &nw_proto, &nw_frag, &frag_hdr)) { | |
a489b168 DDP |
1439 | return false; |
1440 | } | |
1441 | ||
a489b168 DDP |
1442 | if (nw_frag) { |
1443 | return false; | |
1444 | } | |
1445 | ||
c8b1ad49 DB |
1446 | if (new_data) { |
1447 | *new_data = data; | |
1448 | } | |
1449 | ||
cda1b109 DB |
1450 | memcpy(&key->src.addr.ipv6, &ip6->ip6_src, sizeof key->src.addr); |
1451 | memcpy(&key->dst.addr.ipv6, &ip6->ip6_dst, sizeof key->dst.addr); | |
a489b168 DDP |
1452 | key->nw_proto = nw_proto; |
1453 | ||
1454 | return true; | |
1455 | } | |
1456 | ||
1457 | static inline bool | |
1458 | checksum_valid(const struct conn_key *key, const void *data, size_t size, | |
1459 | const void *l3) | |
1460 | { | |
a489b168 | 1461 | if (key->dl_type == htons(ETH_TYPE_IP)) { |
76d85771 DB |
1462 | uint32_t csum = packet_csum_pseudoheader(l3); |
1463 | return csum_finish(csum_continue(csum, data, size)) == 0; | |
a489b168 | 1464 | } else if (key->dl_type == htons(ETH_TYPE_IPV6)) { |
76d85771 | 1465 | return packet_csum_upperlayer6(l3, data, key->nw_proto, size) == 0; |
a489b168 DDP |
1466 | } else { |
1467 | return false; | |
1468 | } | |
a489b168 DDP |
1469 | } |
1470 | ||
1471 | static inline bool | |
1472 | check_l4_tcp(const struct conn_key *key, const void *data, size_t size, | |
324459a3 | 1473 | const void *l3, bool validate_checksum) |
a489b168 DDP |
1474 | { |
1475 | const struct tcp_header *tcp = data; | |
40225b0c BP |
1476 | if (size < sizeof *tcp) { |
1477 | return false; | |
1478 | } | |
a489b168 | 1479 | |
40225b0c | 1480 | size_t tcp_len = TCP_OFFSET(tcp->tcp_ctl) * 4; |
a489b168 DDP |
1481 | if (OVS_UNLIKELY(tcp_len < TCP_HEADER_LEN || tcp_len > size)) { |
1482 | return false; | |
1483 | } | |
1484 | ||
324459a3 | 1485 | return validate_checksum ? checksum_valid(key, data, size, l3) : true; |
a489b168 DDP |
1486 | } |
1487 | ||
1488 | static inline bool | |
1489 | check_l4_udp(const struct conn_key *key, const void *data, size_t size, | |
324459a3 | 1490 | const void *l3, bool validate_checksum) |
a489b168 DDP |
1491 | { |
1492 | const struct udp_header *udp = data; | |
40225b0c BP |
1493 | if (size < sizeof *udp) { |
1494 | return false; | |
1495 | } | |
a489b168 | 1496 | |
40225b0c | 1497 | size_t udp_len = ntohs(udp->udp_len); |
a489b168 DDP |
1498 | if (OVS_UNLIKELY(udp_len < UDP_HEADER_LEN || udp_len > size)) { |
1499 | return false; | |
1500 | } | |
1501 | ||
1502 | /* Validation must be skipped if checksum is 0 on IPv4 packets */ | |
1503 | return (udp->udp_csum == 0 && key->dl_type == htons(ETH_TYPE_IP)) | |
324459a3 | 1504 | || (validate_checksum ? checksum_valid(key, data, size, l3) : true); |
a489b168 DDP |
1505 | } |
1506 | ||
1507 | static inline bool | |
324459a3 | 1508 | check_l4_icmp(const void *data, size_t size, bool validate_checksum) |
a489b168 | 1509 | { |
324459a3 | 1510 | return validate_checksum ? csum(data, size) == 0 : true; |
a489b168 DDP |
1511 | } |
1512 | ||
1513 | static inline bool | |
1514 | check_l4_icmp6(const struct conn_key *key, const void *data, size_t size, | |
324459a3 | 1515 | const void *l3, bool validate_checksum) |
a489b168 | 1516 | { |
324459a3 | 1517 | return validate_checksum ? checksum_valid(key, data, size, l3) : true; |
a489b168 DDP |
1518 | } |
1519 | ||
1520 | static inline bool | |
6c2a9306 DB |
1521 | extract_l4_tcp(struct conn_key *key, const void *data, size_t size, |
1522 | size_t *chk_len) | |
a489b168 | 1523 | { |
6c2a9306 | 1524 | if (OVS_UNLIKELY(size < (chk_len ? *chk_len : TCP_HEADER_LEN))) { |
a489b168 DDP |
1525 | return false; |
1526 | } | |
1527 | ||
dec0dbbc | 1528 | const struct tcp_header *tcp = data; |
a489b168 DDP |
1529 | key->src.port = tcp->tcp_src; |
1530 | key->dst.port = tcp->tcp_dst; | |
1531 | ||
1532 | /* Port 0 is invalid */ | |
1533 | return key->src.port && key->dst.port; | |
1534 | } | |
1535 | ||
1536 | static inline bool | |
6c2a9306 DB |
1537 | extract_l4_udp(struct conn_key *key, const void *data, size_t size, |
1538 | size_t *chk_len) | |
a489b168 | 1539 | { |
6c2a9306 | 1540 | if (OVS_UNLIKELY(size < (chk_len ? *chk_len : UDP_HEADER_LEN))) { |
a489b168 DDP |
1541 | return false; |
1542 | } | |
1543 | ||
dec0dbbc | 1544 | const struct udp_header *udp = data; |
a489b168 DDP |
1545 | key->src.port = udp->udp_src; |
1546 | key->dst.port = udp->udp_dst; | |
1547 | ||
1548 | /* Port 0 is invalid */ | |
1549 | return key->src.port && key->dst.port; | |
1550 | } | |
1551 | ||
1552 | static inline bool extract_l4(struct conn_key *key, const void *data, | |
324459a3 | 1553 | size_t size, bool *related, const void *l3, |
6c2a9306 | 1554 | bool validate_checksum, size_t *chk_len); |
a489b168 | 1555 | |
b269a122 DDP |
1556 | static uint8_t |
1557 | reverse_icmp_type(uint8_t type) | |
1558 | { | |
1559 | switch (type) { | |
1560 | case ICMP4_ECHO_REQUEST: | |
1561 | return ICMP4_ECHO_REPLY; | |
1562 | case ICMP4_ECHO_REPLY: | |
1563 | return ICMP4_ECHO_REQUEST; | |
1564 | ||
1565 | case ICMP4_TIMESTAMP: | |
1566 | return ICMP4_TIMESTAMPREPLY; | |
1567 | case ICMP4_TIMESTAMPREPLY: | |
1568 | return ICMP4_TIMESTAMP; | |
1569 | ||
1570 | case ICMP4_INFOREQUEST: | |
1571 | return ICMP4_INFOREPLY; | |
1572 | case ICMP4_INFOREPLY: | |
1573 | return ICMP4_INFOREQUEST; | |
1574 | default: | |
1575 | OVS_NOT_REACHED(); | |
1576 | } | |
1577 | } | |
1578 | ||
a489b168 DDP |
1579 | /* If 'related' is not NULL and the function is processing an ICMP |
1580 | * error packet, extract the l3 and l4 fields from the nested header | |
1581 | * instead and set *related to true. If 'related' is NULL we're | |
1582 | * already processing a nested header and no such recursion is | |
1583 | * possible */ | |
1584 | static inline int | |
1585 | extract_l4_icmp(struct conn_key *key, const void *data, size_t size, | |
6c2a9306 | 1586 | bool *related, size_t *chk_len) |
a489b168 | 1587 | { |
6c2a9306 | 1588 | if (OVS_UNLIKELY(size < (chk_len ? *chk_len : ICMP_HEADER_LEN))) { |
a489b168 DDP |
1589 | return false; |
1590 | } | |
1591 | ||
dec0dbbc DB |
1592 | const struct icmp_header *icmp = data; |
1593 | ||
a489b168 DDP |
1594 | switch (icmp->icmp_type) { |
1595 | case ICMP4_ECHO_REQUEST: | |
1596 | case ICMP4_ECHO_REPLY: | |
1597 | case ICMP4_TIMESTAMP: | |
1598 | case ICMP4_TIMESTAMPREPLY: | |
1599 | case ICMP4_INFOREQUEST: | |
1600 | case ICMP4_INFOREPLY: | |
b269a122 DDP |
1601 | if (icmp->icmp_code != 0) { |
1602 | return false; | |
1603 | } | |
a489b168 | 1604 | /* Separate ICMP connection: identified using id */ |
b269a122 DDP |
1605 | key->src.icmp_id = key->dst.icmp_id = icmp->icmp_fields.echo.id; |
1606 | key->src.icmp_type = icmp->icmp_type; | |
1607 | key->dst.icmp_type = reverse_icmp_type(icmp->icmp_type); | |
a489b168 DDP |
1608 | break; |
1609 | case ICMP4_DST_UNREACH: | |
1610 | case ICMP4_TIME_EXCEEDED: | |
1611 | case ICMP4_PARAM_PROB: | |
1612 | case ICMP4_SOURCEQUENCH: | |
1613 | case ICMP4_REDIRECT: { | |
1614 | /* ICMP packet part of another connection. We should | |
1615 | * extract the key from embedded packet header */ | |
1616 | struct conn_key inner_key; | |
1617 | const char *l3 = (const char *) (icmp + 1); | |
1618 | const char *tail = (const char *) data + size; | |
1619 | const char *l4; | |
a489b168 DDP |
1620 | |
1621 | if (!related) { | |
1622 | return false; | |
1623 | } | |
1624 | ||
1625 | memset(&inner_key, 0, sizeof inner_key); | |
1626 | inner_key.dl_type = htons(ETH_TYPE_IP); | |
dec0dbbc | 1627 | bool ok = extract_l3_ipv4(&inner_key, l3, tail - l3, &l4, false); |
a489b168 DDP |
1628 | if (!ok) { |
1629 | return false; | |
1630 | } | |
1631 | ||
cda1b109 | 1632 | if (inner_key.src.addr.ipv4 != key->dst.addr.ipv4) { |
a489b168 DDP |
1633 | return false; |
1634 | } | |
1635 | ||
1636 | key->src = inner_key.src; | |
1637 | key->dst = inner_key.dst; | |
1638 | key->nw_proto = inner_key.nw_proto; | |
6c2a9306 | 1639 | size_t check_len = ICMP_ERROR_DATA_L4_LEN; |
a489b168 | 1640 | |
6c2a9306 | 1641 | ok = extract_l4(key, l4, tail - l4, NULL, l3, false, &check_len); |
a489b168 DDP |
1642 | if (ok) { |
1643 | conn_key_reverse(key); | |
1644 | *related = true; | |
1645 | } | |
1646 | return ok; | |
1647 | } | |
1648 | default: | |
1649 | return false; | |
1650 | } | |
1651 | ||
1652 | return true; | |
1653 | } | |
1654 | ||
b269a122 DDP |
1655 | static uint8_t |
1656 | reverse_icmp6_type(uint8_t type) | |
1657 | { | |
1658 | switch (type) { | |
1659 | case ICMP6_ECHO_REQUEST: | |
1660 | return ICMP6_ECHO_REPLY; | |
1661 | case ICMP6_ECHO_REPLY: | |
1662 | return ICMP6_ECHO_REQUEST; | |
1663 | default: | |
1664 | OVS_NOT_REACHED(); | |
1665 | } | |
1666 | } | |
1667 | ||
a489b168 DDP |
1668 | /* If 'related' is not NULL and the function is processing an ICMP |
1669 | * error packet, extract the l3 and l4 fields from the nested header | |
1670 | * instead and set *related to true. If 'related' is NULL we're | |
1671 | * already processing a nested header and no such recursion is | |
1672 | * possible */ | |
1673 | static inline bool | |
1674 | extract_l4_icmp6(struct conn_key *key, const void *data, size_t size, | |
1675 | bool *related) | |
1676 | { | |
1677 | const struct icmp6_header *icmp6 = data; | |
1678 | ||
1679 | /* All the messages that we support need at least 4 bytes after | |
1680 | * the header */ | |
1681 | if (size < sizeof *icmp6 + 4) { | |
1682 | return false; | |
1683 | } | |
1684 | ||
1685 | switch (icmp6->icmp6_type) { | |
1686 | case ICMP6_ECHO_REQUEST: | |
1687 | case ICMP6_ECHO_REPLY: | |
b269a122 DDP |
1688 | if (icmp6->icmp6_code != 0) { |
1689 | return false; | |
1690 | } | |
a489b168 | 1691 | /* Separate ICMP connection: identified using id */ |
b269a122 DDP |
1692 | key->src.icmp_id = key->dst.icmp_id = *(ovs_be16 *) (icmp6 + 1); |
1693 | key->src.icmp_type = icmp6->icmp6_type; | |
1694 | key->dst.icmp_type = reverse_icmp6_type(icmp6->icmp6_type); | |
a489b168 DDP |
1695 | break; |
1696 | case ICMP6_DST_UNREACH: | |
1697 | case ICMP6_PACKET_TOO_BIG: | |
1698 | case ICMP6_TIME_EXCEEDED: | |
1699 | case ICMP6_PARAM_PROB: { | |
1700 | /* ICMP packet part of another connection. We should | |
1701 | * extract the key from embedded packet header */ | |
1702 | struct conn_key inner_key; | |
1703 | const char *l3 = (const char *) icmp6 + 8; | |
1704 | const char *tail = (const char *) data + size; | |
1705 | const char *l4 = NULL; | |
a489b168 DDP |
1706 | |
1707 | if (!related) { | |
1708 | return false; | |
1709 | } | |
1710 | ||
1711 | memset(&inner_key, 0, sizeof inner_key); | |
1712 | inner_key.dl_type = htons(ETH_TYPE_IPV6); | |
dec0dbbc | 1713 | bool ok = extract_l3_ipv6(&inner_key, l3, tail - l3, &l4); |
a489b168 DDP |
1714 | if (!ok) { |
1715 | return false; | |
1716 | } | |
1717 | ||
1718 | /* pf doesn't do this, but it seems a good idea */ | |
cda1b109 DB |
1719 | if (!ipv6_addr_equals(&inner_key.src.addr.ipv6, |
1720 | &key->dst.addr.ipv6)) { | |
a489b168 DDP |
1721 | return false; |
1722 | } | |
1723 | ||
1724 | key->src = inner_key.src; | |
1725 | key->dst = inner_key.dst; | |
1726 | key->nw_proto = inner_key.nw_proto; | |
1727 | ||
6c2a9306 | 1728 | ok = extract_l4(key, l4, tail - l4, NULL, l3, false, NULL); |
a489b168 DDP |
1729 | if (ok) { |
1730 | conn_key_reverse(key); | |
1731 | *related = true; | |
1732 | } | |
1733 | return ok; | |
1734 | } | |
1735 | default: | |
1736 | return false; | |
1737 | } | |
1738 | ||
1739 | return true; | |
1740 | } | |
1741 | ||
1742 | /* Extract l4 fields into 'key', which must already contain valid l3 | |
1743 | * members. | |
1744 | * | |
1745 | * If 'related' is not NULL and an ICMP error packet is being | |
1746 | * processed, the function will extract the key from the packet nested | |
1401f6de | 1747 | * in the ICMP payload and set '*related' to true. |
a489b168 | 1748 | * |
9171c635 DB |
1749 | * 'size' here is the layer 4 size, which can be a nested size if parsing |
1750 | * an ICMP or ICMP6 header. | |
1751 | * | |
a489b168 | 1752 | * If 'related' is NULL, it means that we're already parsing a header nested |
6c2a9306 DB |
1753 | * in an ICMP error. In this case, we skip the checksum and some length |
1754 | * validations. */ | |
a489b168 DDP |
1755 | static inline bool |
1756 | extract_l4(struct conn_key *key, const void *data, size_t size, bool *related, | |
6c2a9306 | 1757 | const void *l3, bool validate_checksum, size_t *chk_len) |
a489b168 DDP |
1758 | { |
1759 | if (key->nw_proto == IPPROTO_TCP) { | |
324459a3 | 1760 | return (!related || check_l4_tcp(key, data, size, l3, |
6c2a9306 DB |
1761 | validate_checksum)) |
1762 | && extract_l4_tcp(key, data, size, chk_len); | |
a489b168 | 1763 | } else if (key->nw_proto == IPPROTO_UDP) { |
324459a3 | 1764 | return (!related || check_l4_udp(key, data, size, l3, |
6c2a9306 DB |
1765 | validate_checksum)) |
1766 | && extract_l4_udp(key, data, size, chk_len); | |
a489b168 DDP |
1767 | } else if (key->dl_type == htons(ETH_TYPE_IP) |
1768 | && key->nw_proto == IPPROTO_ICMP) { | |
324459a3 | 1769 | return (!related || check_l4_icmp(data, size, validate_checksum)) |
6c2a9306 | 1770 | && extract_l4_icmp(key, data, size, related, chk_len); |
a489b168 DDP |
1771 | } else if (key->dl_type == htons(ETH_TYPE_IPV6) |
1772 | && key->nw_proto == IPPROTO_ICMPV6) { | |
324459a3 | 1773 | return (!related || check_l4_icmp6(key, data, size, l3, |
6c2a9306 DB |
1774 | validate_checksum)) |
1775 | && extract_l4_icmp6(key, data, size, related); | |
a489b168 DDP |
1776 | } else { |
1777 | return false; | |
1778 | } | |
1779 | } | |
1780 | ||
1781 | static bool | |
66e4ad8a | 1782 | conn_key_extract(struct conntrack *ct, struct dp_packet *pkt, ovs_be16 dl_type, |
a489b168 DDP |
1783 | struct conn_lookup_ctx *ctx, uint16_t zone) |
1784 | { | |
2482b0b0 | 1785 | const struct eth_header *l2 = dp_packet_eth(pkt); |
a489b168 DDP |
1786 | const struct ip_header *l3 = dp_packet_l3(pkt); |
1787 | const char *l4 = dp_packet_l4(pkt); | |
a489b168 DDP |
1788 | |
1789 | memset(ctx, 0, sizeof *ctx); | |
1790 | ||
1791 | if (!l2 || !l3 || !l4) { | |
1792 | return false; | |
1793 | } | |
1794 | ||
1795 | ctx->key.zone = zone; | |
1796 | ||
1797 | /* XXX In this function we parse the packet (again, it has already | |
1798 | * gone through miniflow_extract()) for two reasons: | |
1799 | * | |
1800 | * 1) To extract the l3 addresses and l4 ports. | |
1801 | * We already have the l3 and l4 headers' pointers. Extracting | |
1802 | * the l3 addresses and the l4 ports is really cheap, since they | |
1803 | * can be found at fixed locations. | |
66e4ad8a DDP |
1804 | * 2) To extract the l4 type. |
1805 | * Extracting the l4 types, for IPv6 can be quite expensive, because | |
1806 | * it's not at a fixed location. | |
a489b168 DDP |
1807 | * |
1808 | * Here's a way to avoid (2) with the help of the datapath. | |
66e4ad8a | 1809 | * The datapath doesn't keep the packet's extracted flow[1], so |
a489b168 | 1810 | * using that is not an option. We could use the packet's matching |
66e4ad8a DDP |
1811 | * megaflow, but we have to make sure that the l4 type (nw_proto) |
1812 | * is unwildcarded. This means either: | |
a489b168 | 1813 | * |
66e4ad8a DDP |
1814 | * a) dpif-netdev unwildcards the l4 type when a new flow is installed |
1815 | * if the actions contains ct(). | |
a489b168 | 1816 | * |
66e4ad8a DDP |
1817 | * b) ofproto-dpif-xlate unwildcards the l4 type when translating a ct() |
1818 | * action. This is already done in different actions, but it's | |
1819 | * unnecessary for the kernel. | |
a489b168 DDP |
1820 | * |
1821 | * --- | |
66e4ad8a | 1822 | * [1] The reasons for this are that keeping the flow increases |
a489b168 DDP |
1823 | * (slightly) the cache footprint and increases computation |
1824 | * time as we move the packet around. Most importantly, the flow | |
1825 | * should be updated by the actions and this can be slow, as | |
1826 | * we use a sparse representation (miniflow). | |
1827 | * | |
1828 | */ | |
dec0dbbc | 1829 | bool ok; |
66e4ad8a | 1830 | ctx->key.dl_type = dl_type; |
dec0dbbc | 1831 | |
a489b168 | 1832 | if (ctx->key.dl_type == htons(ETH_TYPE_IP)) { |
dec0dbbc | 1833 | bool hwol_bad_l3_csum = dp_packet_ip_checksum_bad(pkt); |
324459a3 SC |
1834 | if (hwol_bad_l3_csum) { |
1835 | ok = false; | |
1836 | } else { | |
dec0dbbc | 1837 | bool hwol_good_l3_csum = dp_packet_ip_checksum_valid(pkt); |
324459a3 | 1838 | /* Validate the checksum only when hwol is not supported. */ |
9171c635 | 1839 | ok = extract_l3_ipv4(&ctx->key, l3, dp_packet_l3_size(pkt), NULL, |
324459a3 SC |
1840 | !hwol_good_l3_csum); |
1841 | } | |
a489b168 | 1842 | } else if (ctx->key.dl_type == htons(ETH_TYPE_IPV6)) { |
9171c635 | 1843 | ok = extract_l3_ipv6(&ctx->key, l3, dp_packet_l3_size(pkt), NULL); |
a489b168 DDP |
1844 | } else { |
1845 | ok = false; | |
1846 | } | |
1847 | ||
1848 | if (ok) { | |
324459a3 SC |
1849 | bool hwol_bad_l4_csum = dp_packet_l4_checksum_bad(pkt); |
1850 | if (!hwol_bad_l4_csum) { | |
1851 | bool hwol_good_l4_csum = dp_packet_l4_checksum_valid(pkt); | |
1852 | /* Validate the checksum only when hwol is not supported. */ | |
9171c635 | 1853 | if (extract_l4(&ctx->key, l4, dp_packet_l4_size(pkt), |
6c2a9306 DB |
1854 | &ctx->icmp_related, l3, !hwol_good_l4_csum, |
1855 | NULL)) { | |
324459a3 SC |
1856 | ctx->hash = conn_key_hash(&ctx->key, ct->hash_basis); |
1857 | return true; | |
1858 | } | |
a489b168 DDP |
1859 | } |
1860 | } | |
1861 | ||
1862 | return false; | |
1863 | } | |
92edd073 DB |
1864 | |
1865 | static uint32_t | |
cda1b109 | 1866 | ct_addr_hash_add(uint32_t hash, const union ct_addr *addr) |
92edd073 DB |
1867 | { |
1868 | BUILD_ASSERT_DECL(sizeof *addr % 4 == 0); | |
1869 | return hash_add_bytes32(hash, (const uint32_t *) addr, sizeof *addr); | |
1870 | } | |
1871 | ||
1872 | static uint32_t | |
1873 | ct_endpoint_hash_add(uint32_t hash, const struct ct_endpoint *ep) | |
1874 | { | |
1875 | BUILD_ASSERT_DECL(sizeof *ep % 4 == 0); | |
1876 | return hash_add_bytes32(hash, (const uint32_t *) ep, sizeof *ep); | |
1877 | } | |
a489b168 DDP |
1878 | \f |
1879 | /* Symmetric */ | |
1880 | static uint32_t | |
1881 | conn_key_hash(const struct conn_key *key, uint32_t basis) | |
1882 | { | |
1883 | uint32_t hsrc, hdst, hash; | |
a489b168 | 1884 | hsrc = hdst = basis; |
6b1d4625 DB |
1885 | hsrc = ct_endpoint_hash_add(hsrc, &key->src); |
1886 | hdst = ct_endpoint_hash_add(hdst, &key->dst); | |
a489b168 DDP |
1887 | |
1888 | /* Even if source and destination are swapped the hash will be the same. */ | |
1889 | hash = hsrc ^ hdst; | |
1890 | ||
1891 | /* Hash the rest of the key(L3 and L4 types and zone). */ | |
763b40b0 | 1892 | return hash_words((uint32_t *) (&key->dst + 1), |
a489b168 DDP |
1893 | (uint32_t *) (key + 1) - (uint32_t *) (&key->dst + 1), |
1894 | hash); | |
a489b168 DDP |
1895 | } |
1896 | ||
1897 | static void | |
1898 | conn_key_reverse(struct conn_key *key) | |
1899 | { | |
dec0dbbc | 1900 | struct ct_endpoint tmp = key->src; |
a489b168 DDP |
1901 | key->src = key->dst; |
1902 | key->dst = tmp; | |
1903 | } | |
1904 | ||
286de272 | 1905 | static uint32_t |
cda1b109 | 1906 | nat_ipv6_addrs_delta(struct in6_addr *ipv6_min, struct in6_addr *ipv6_max) |
286de272 | 1907 | { |
cda1b109 DB |
1908 | uint8_t *ipv6_min_hi = &ipv6_min->s6_addr[0]; |
1909 | uint8_t *ipv6_min_lo = &ipv6_min->s6_addr[0] + sizeof(uint64_t); | |
1910 | uint8_t *ipv6_max_hi = &ipv6_max->s6_addr[0]; | |
1911 | uint8_t *ipv6_max_lo = &ipv6_max->s6_addr[0] + sizeof(uint64_t); | |
286de272 DB |
1912 | |
1913 | ovs_be64 addr6_64_min_hi; | |
1914 | ovs_be64 addr6_64_min_lo; | |
1915 | memcpy(&addr6_64_min_hi, ipv6_min_hi, sizeof addr6_64_min_hi); | |
1916 | memcpy(&addr6_64_min_lo, ipv6_min_lo, sizeof addr6_64_min_lo); | |
1917 | ||
1918 | ovs_be64 addr6_64_max_hi; | |
1919 | ovs_be64 addr6_64_max_lo; | |
1920 | memcpy(&addr6_64_max_hi, ipv6_max_hi, sizeof addr6_64_max_hi); | |
1921 | memcpy(&addr6_64_max_lo, ipv6_max_lo, sizeof addr6_64_max_lo); | |
1922 | ||
1923 | uint64_t diff; | |
dec0dbbc | 1924 | |
286de272 DB |
1925 | if (addr6_64_min_hi == addr6_64_max_hi && |
1926 | ntohll(addr6_64_min_lo) <= ntohll(addr6_64_max_lo)) { | |
1927 | diff = ntohll(addr6_64_max_lo) - ntohll(addr6_64_min_lo); | |
1928 | } else if (ntohll(addr6_64_min_hi) + 1 == ntohll(addr6_64_max_hi) && | |
1929 | ntohll(addr6_64_min_lo) > ntohll(addr6_64_max_lo)) { | |
1930 | diff = UINT64_MAX - (ntohll(addr6_64_min_lo) - | |
1931 | ntohll(addr6_64_max_lo) - 1); | |
1932 | } else { | |
1933 | /* Limit address delta supported to 32 bits or 4 billion approximately. | |
1934 | * Possibly, this should be visible to the user through a datapath | |
1935 | * support check, however the practical impact is probably nil. */ | |
1936 | diff = 0xfffffffe; | |
1937 | } | |
dec0dbbc | 1938 | |
286de272 DB |
1939 | if (diff > 0xfffffffe) { |
1940 | diff = 0xfffffffe; | |
1941 | } | |
1942 | return diff; | |
1943 | } | |
1944 | ||
1945 | /* This function must be used in tandem with nat_ipv6_addrs_delta(), which | |
1946 | * restricts the input parameters. */ | |
a489b168 | 1947 | static void |
cda1b109 | 1948 | nat_ipv6_addr_increment(struct in6_addr *ipv6, uint32_t increment) |
286de272 | 1949 | { |
cda1b109 DB |
1950 | uint8_t *ipv6_hi = &ipv6->s6_addr[0]; |
1951 | uint8_t *ipv6_lo = &ipv6->s6_addr[0] + sizeof(ovs_be64); | |
286de272 DB |
1952 | ovs_be64 addr6_64_hi; |
1953 | ovs_be64 addr6_64_lo; | |
1954 | memcpy(&addr6_64_hi, ipv6_hi, sizeof addr6_64_hi); | |
1955 | memcpy(&addr6_64_lo, ipv6_lo, sizeof addr6_64_lo); | |
1956 | ||
1957 | if (UINT64_MAX - increment >= ntohll(addr6_64_lo)) { | |
1958 | addr6_64_lo = htonll(increment + ntohll(addr6_64_lo)); | |
1959 | } else if (addr6_64_hi != OVS_BE64_MAX) { | |
1960 | addr6_64_hi = htonll(1 + ntohll(addr6_64_hi)); | |
1961 | addr6_64_lo = htonll(increment - (UINT64_MAX - | |
1962 | ntohll(addr6_64_lo) + 1)); | |
1963 | } else { | |
1964 | OVS_NOT_REACHED(); | |
1965 | } | |
1966 | ||
1967 | memcpy(ipv6_hi, &addr6_64_hi, sizeof addr6_64_hi); | |
1968 | memcpy(ipv6_lo, &addr6_64_lo, sizeof addr6_64_lo); | |
286de272 DB |
1969 | } |
1970 | ||
1971 | static uint32_t | |
1972 | nat_range_hash(const struct conn *conn, uint32_t basis) | |
1973 | { | |
1974 | uint32_t hash = basis; | |
286de272 | 1975 | |
92edd073 DB |
1976 | hash = ct_addr_hash_add(hash, &conn->nat_info->min_addr); |
1977 | hash = ct_addr_hash_add(hash, &conn->nat_info->max_addr); | |
1978 | hash = hash_add(hash, | |
1979 | (conn->nat_info->max_port << 16) | |
1980 | | conn->nat_info->min_port); | |
92edd073 DB |
1981 | hash = ct_endpoint_hash_add(hash, &conn->key.src); |
1982 | hash = ct_endpoint_hash_add(hash, &conn->key.dst); | |
286de272 DB |
1983 | hash = hash_add(hash, (OVS_FORCE uint32_t) conn->key.dl_type); |
1984 | hash = hash_add(hash, conn->key.nw_proto); | |
1985 | hash = hash_add(hash, conn->key.zone); | |
92edd073 DB |
1986 | |
1987 | /* The purpose of the second parameter is to distinguish hashes of data of | |
1988 | * different length; our data always has the same length so there is no | |
1989 | * value in counting. */ | |
1990 | return hash_finish(hash, 0); | |
286de272 DB |
1991 | } |
1992 | ||
1993 | static bool | |
1994 | nat_select_range_tuple(struct conntrack *ct, const struct conn *conn, | |
1995 | struct conn *nat_conn) | |
1996 | { | |
bd5e81a0 DB |
1997 | enum { MIN_NAT_EPHEMERAL_PORT = 1024, |
1998 | MAX_NAT_EPHEMERAL_PORT = 65535 }; | |
286de272 DB |
1999 | |
2000 | uint16_t min_port; | |
2001 | uint16_t max_port; | |
2002 | uint16_t first_port; | |
286de272 DB |
2003 | uint32_t hash = nat_range_hash(conn, ct->hash_basis); |
2004 | ||
2005 | if ((conn->nat_info->nat_action & NAT_ACTION_SRC) && | |
2006 | (!(conn->nat_info->nat_action & NAT_ACTION_SRC_PORT))) { | |
2007 | min_port = ntohs(conn->key.src.port); | |
2008 | max_port = ntohs(conn->key.src.port); | |
2009 | first_port = min_port; | |
2010 | } else if ((conn->nat_info->nat_action & NAT_ACTION_DST) && | |
2011 | (!(conn->nat_info->nat_action & NAT_ACTION_DST_PORT))) { | |
2012 | min_port = ntohs(conn->key.dst.port); | |
2013 | max_port = ntohs(conn->key.dst.port); | |
2014 | first_port = min_port; | |
2015 | } else { | |
2016 | uint16_t deltap = conn->nat_info->max_port - conn->nat_info->min_port; | |
2017 | uint32_t port_index = hash % (deltap + 1); | |
2018 | first_port = conn->nat_info->min_port + port_index; | |
2019 | min_port = conn->nat_info->min_port; | |
2020 | max_port = conn->nat_info->max_port; | |
2021 | } | |
2022 | ||
2023 | uint32_t deltaa = 0; | |
2024 | uint32_t address_index; | |
cda1b109 | 2025 | union ct_addr ct_addr; |
286de272 | 2026 | memset(&ct_addr, 0, sizeof ct_addr); |
cda1b109 | 2027 | union ct_addr max_ct_addr; |
286de272 DB |
2028 | memset(&max_ct_addr, 0, sizeof max_ct_addr); |
2029 | max_ct_addr = conn->nat_info->max_addr; | |
2030 | ||
2031 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
cda1b109 DB |
2032 | deltaa = ntohl(conn->nat_info->max_addr.ipv4) - |
2033 | ntohl(conn->nat_info->min_addr.ipv4); | |
286de272 | 2034 | address_index = hash % (deltaa + 1); |
cda1b109 DB |
2035 | ct_addr.ipv4 = htonl( |
2036 | ntohl(conn->nat_info->min_addr.ipv4) + address_index); | |
286de272 | 2037 | } else { |
cda1b109 DB |
2038 | deltaa = nat_ipv6_addrs_delta(&conn->nat_info->min_addr.ipv6, |
2039 | &conn->nat_info->max_addr.ipv6); | |
286de272 DB |
2040 | /* deltaa must be within 32 bits for full hash coverage. A 64 or |
2041 | * 128 bit hash is unnecessary and hence not used here. Most code | |
2042 | * is kept common with V4; nat_ipv6_addrs_delta() will do the | |
2043 | * enforcement via max_ct_addr. */ | |
2044 | max_ct_addr = conn->nat_info->min_addr; | |
cda1b109 | 2045 | nat_ipv6_addr_increment(&max_ct_addr.ipv6, deltaa); |
286de272 | 2046 | address_index = hash % (deltaa + 1); |
cda1b109 DB |
2047 | ct_addr.ipv6 = conn->nat_info->min_addr.ipv6; |
2048 | nat_ipv6_addr_increment(&ct_addr.ipv6, address_index); | |
286de272 DB |
2049 | } |
2050 | ||
2051 | uint16_t port = first_port; | |
2052 | bool all_ports_tried = false; | |
32b2c81f DB |
2053 | /* For DNAT or for specified port ranges, we don't use ephemeral ports. */ |
2054 | bool ephemeral_ports_tried | |
2055 | = conn->nat_info->nat_action & NAT_ACTION_DST || | |
2056 | conn->nat_info->nat_action & NAT_ACTION_SRC_PORT | |
2057 | ? true : false; | |
cda1b109 | 2058 | union ct_addr first_addr = ct_addr; |
4cd0481c DB |
2059 | bool pat_enabled = conn->key.nw_proto != IPPROTO_ICMP && |
2060 | conn->key.nw_proto != IPPROTO_ICMPV6; | |
286de272 DB |
2061 | |
2062 | while (true) { | |
2063 | if (conn->nat_info->nat_action & NAT_ACTION_SRC) { | |
2064 | nat_conn->rev_key.dst.addr = ct_addr; | |
e32cd4c6 | 2065 | if (pat_enabled) { |
2066 | nat_conn->rev_key.dst.port = htons(port); | |
2067 | } | |
286de272 | 2068 | } else { |
1c8689d7 | 2069 | nat_conn->rev_key.src.addr = ct_addr; |
e32cd4c6 | 2070 | if (pat_enabled) { |
2071 | nat_conn->rev_key.src.port = htons(port); | |
2072 | } | |
286de272 DB |
2073 | } |
2074 | ||
e32cd4c6 | 2075 | bool found = conn_lookup(ct, &nat_conn->rev_key, time_msec(), NULL, |
2076 | NULL); | |
967bb5c5 | 2077 | if (!found) { |
286de272 | 2078 | return true; |
4cd0481c | 2079 | } else if (pat_enabled && !all_ports_tried) { |
286de272 DB |
2080 | if (min_port == max_port) { |
2081 | all_ports_tried = true; | |
2082 | } else if (port == max_port) { | |
2083 | port = min_port; | |
2084 | } else { | |
2085 | port++; | |
2086 | } | |
2087 | if (port == first_port) { | |
2088 | all_ports_tried = true; | |
2089 | } | |
2090 | } else { | |
2091 | if (memcmp(&ct_addr, &max_ct_addr, sizeof ct_addr)) { | |
2092 | if (conn->key.dl_type == htons(ETH_TYPE_IP)) { | |
cda1b109 | 2093 | ct_addr.ipv4 = htonl(ntohl(ct_addr.ipv4) + 1); |
286de272 | 2094 | } else { |
cda1b109 | 2095 | nat_ipv6_addr_increment(&ct_addr.ipv6, 1); |
286de272 DB |
2096 | } |
2097 | } else { | |
2098 | ct_addr = conn->nat_info->min_addr; | |
2099 | } | |
2100 | if (!memcmp(&ct_addr, &first_addr, sizeof ct_addr)) { | |
4cd0481c | 2101 | if (pat_enabled && !ephemeral_ports_tried) { |
ac04639a | 2102 | ephemeral_ports_tried = true; |
286de272 | 2103 | ct_addr = conn->nat_info->min_addr; |
8417e688 | 2104 | first_addr = ct_addr; |
286de272 DB |
2105 | min_port = MIN_NAT_EPHEMERAL_PORT; |
2106 | max_port = MAX_NAT_EPHEMERAL_PORT; | |
2107 | } else { | |
2108 | break; | |
2109 | } | |
2110 | } | |
2111 | first_port = min_port; | |
2112 | port = first_port; | |
2113 | all_ports_tried = false; | |
2114 | } | |
2115 | } | |
2116 | return false; | |
2117 | } | |
2118 | ||
a489b168 | 2119 | static enum ct_update_res |
967bb5c5 DB |
2120 | conn_update(struct conntrack *ct, struct conn *conn, struct dp_packet *pkt, |
2121 | struct conn_lookup_ctx *ctx, long long now) | |
a489b168 | 2122 | { |
967bb5c5 DB |
2123 | ovs_mutex_lock(&conn->lock); |
2124 | enum ct_update_res update_res = | |
2125 | l4_protos[conn->key.nw_proto]->conn_update(ct, conn, pkt, ctx->reply, | |
2126 | now); | |
2127 | ovs_mutex_unlock(&conn->lock); | |
2128 | return update_res; | |
a489b168 DDP |
2129 | } |
2130 | ||
2131 | static bool | |
2132 | conn_expired(struct conn *conn, long long now) | |
2133 | { | |
286de272 | 2134 | if (conn->conn_type == CT_CONN_TYPE_DEFAULT) { |
967bb5c5 DB |
2135 | ovs_mutex_lock(&conn->lock); |
2136 | bool expired = now >= conn->expiration ? true : false; | |
2137 | ovs_mutex_unlock(&conn->lock); | |
2138 | return expired; | |
286de272 DB |
2139 | } |
2140 | return false; | |
a489b168 DDP |
2141 | } |
2142 | ||
2143 | static bool | |
2144 | valid_new(struct dp_packet *pkt, struct conn_key *key) | |
2145 | { | |
2146 | return l4_protos[key->nw_proto]->valid_new(pkt); | |
2147 | } | |
2148 | ||
2149 | static struct conn * | |
967bb5c5 DB |
2150 | new_conn(struct conntrack *ct, struct dp_packet *pkt, struct conn_key *key, |
2151 | long long now) | |
a489b168 | 2152 | { |
967bb5c5 | 2153 | return l4_protos[key->nw_proto]->new_conn(ct, pkt, now); |
a489b168 DDP |
2154 | } |
2155 | ||
2156 | static void | |
967bb5c5 | 2157 | delete_conn_cmn(struct conn *conn) |
a489b168 | 2158 | { |
286de272 | 2159 | free(conn->nat_info); |
bd5e81a0 | 2160 | free(conn->alg); |
a489b168 DDP |
2161 | free(conn); |
2162 | } | |
967bb5c5 DB |
2163 | |
2164 | static void | |
2165 | delete_conn(struct conn *conn) | |
2166 | { | |
2167 | ovs_assert(conn->conn_type == CT_CONN_TYPE_DEFAULT); | |
2168 | ovs_mutex_destroy(&conn->lock); | |
2169 | free(conn->nat_conn); | |
2170 | delete_conn_cmn(conn); | |
2171 | } | |
2172 | ||
2173 | /* Only used by conn_clean_one(). */ | |
2174 | static void | |
2175 | delete_conn_one(struct conn *conn) | |
2176 | { | |
2177 | if (conn->conn_type == CT_CONN_TYPE_DEFAULT) { | |
2178 | ovs_mutex_destroy(&conn->lock); | |
2179 | } | |
2180 | delete_conn_cmn(conn); | |
2181 | } | |
4d4e68ed | 2182 | \f |
271e48a0 YHW |
2183 | /* Convert a conntrack address 'a' into an IP address 'b' based on 'dl_type'. |
2184 | * | |
2185 | * Note that 'dl_type' should be either "ETH_TYPE_IP" or "ETH_TYPE_IPv6" | |
2186 | * in network-byte order. */ | |
4d4e68ed | 2187 | static void |
cda1b109 | 2188 | ct_endpoint_to_ct_dpif_inet_addr(const union ct_addr *a, |
4d4e68ed DDP |
2189 | union ct_dpif_inet_addr *b, |
2190 | ovs_be16 dl_type) | |
2191 | { | |
2192 | if (dl_type == htons(ETH_TYPE_IP)) { | |
cda1b109 | 2193 | b->ip = a->ipv4; |
4d4e68ed | 2194 | } else if (dl_type == htons(ETH_TYPE_IPV6)){ |
cda1b109 | 2195 | b->in6 = a->ipv6; |
4d4e68ed DDP |
2196 | } |
2197 | } | |
2198 | ||
271e48a0 YHW |
2199 | /* Convert an IP address 'a' into a conntrack address 'b' based on 'dl_type'. |
2200 | * | |
2201 | * Note that 'dl_type' should be either "ETH_TYPE_IP" or "ETH_TYPE_IPv6" | |
2202 | * in network-byte order. */ | |
2203 | static void | |
2204 | ct_dpif_inet_addr_to_ct_endpoint(const union ct_dpif_inet_addr *a, | |
cda1b109 | 2205 | union ct_addr *b, ovs_be16 dl_type) |
271e48a0 YHW |
2206 | { |
2207 | if (dl_type == htons(ETH_TYPE_IP)) { | |
cda1b109 | 2208 | b->ipv4 = a->ip; |
271e48a0 | 2209 | } else if (dl_type == htons(ETH_TYPE_IPV6)){ |
cda1b109 | 2210 | b->ipv6 = a->in6; |
271e48a0 YHW |
2211 | } |
2212 | } | |
2213 | ||
4d4e68ed DDP |
2214 | static void |
2215 | conn_key_to_tuple(const struct conn_key *key, struct ct_dpif_tuple *tuple) | |
2216 | { | |
2217 | if (key->dl_type == htons(ETH_TYPE_IP)) { | |
2218 | tuple->l3_type = AF_INET; | |
2219 | } else if (key->dl_type == htons(ETH_TYPE_IPV6)) { | |
2220 | tuple->l3_type = AF_INET6; | |
2221 | } | |
2222 | tuple->ip_proto = key->nw_proto; | |
2223 | ct_endpoint_to_ct_dpif_inet_addr(&key->src.addr, &tuple->src, | |
2224 | key->dl_type); | |
2225 | ct_endpoint_to_ct_dpif_inet_addr(&key->dst.addr, &tuple->dst, | |
2226 | key->dl_type); | |
2227 | ||
2228 | if (key->nw_proto == IPPROTO_ICMP || key->nw_proto == IPPROTO_ICMPV6) { | |
b269a122 DDP |
2229 | tuple->icmp_id = key->src.icmp_id; |
2230 | tuple->icmp_type = key->src.icmp_type; | |
2231 | tuple->icmp_code = key->src.icmp_code; | |
4d4e68ed DDP |
2232 | } else { |
2233 | tuple->src_port = key->src.port; | |
2234 | tuple->dst_port = key->dst.port; | |
2235 | } | |
2236 | } | |
2237 | ||
271e48a0 YHW |
2238 | static void |
2239 | tuple_to_conn_key(const struct ct_dpif_tuple *tuple, uint16_t zone, | |
2240 | struct conn_key *key) | |
2241 | { | |
2242 | if (tuple->l3_type == AF_INET) { | |
2243 | key->dl_type = htons(ETH_TYPE_IP); | |
2244 | } else if (tuple->l3_type == AF_INET6) { | |
2245 | key->dl_type = htons(ETH_TYPE_IPV6); | |
2246 | } | |
2247 | key->nw_proto = tuple->ip_proto; | |
2248 | ct_dpif_inet_addr_to_ct_endpoint(&tuple->src, &key->src.addr, | |
2249 | key->dl_type); | |
2250 | ct_dpif_inet_addr_to_ct_endpoint(&tuple->dst, &key->dst.addr, | |
2251 | key->dl_type); | |
2252 | ||
2253 | if (tuple->ip_proto == IPPROTO_ICMP || tuple->ip_proto == IPPROTO_ICMPV6) { | |
2254 | key->src.icmp_id = tuple->icmp_id; | |
2255 | key->src.icmp_type = tuple->icmp_type; | |
2256 | key->src.icmp_code = tuple->icmp_code; | |
2257 | key->dst.icmp_id = tuple->icmp_id; | |
2258 | key->dst.icmp_type = reverse_icmp_type(tuple->icmp_type); | |
2259 | key->dst.icmp_code = tuple->icmp_code; | |
2260 | } else { | |
2261 | key->src.port = tuple->src_port; | |
2262 | key->dst.port = tuple->dst_port; | |
2263 | } | |
2264 | key->zone = zone; | |
2265 | } | |
2266 | ||
4d4e68ed DDP |
2267 | static void |
2268 | conn_to_ct_dpif_entry(const struct conn *conn, struct ct_dpif_entry *entry, | |
f1a0469e | 2269 | long long now) |
4d4e68ed | 2270 | { |
4d4e68ed DDP |
2271 | memset(entry, 0, sizeof *entry); |
2272 | conn_key_to_tuple(&conn->key, &entry->tuple_orig); | |
2273 | conn_key_to_tuple(&conn->rev_key, &entry->tuple_reply); | |
2274 | ||
2275 | entry->zone = conn->key.zone; | |
4d4e68ed | 2276 | |
967bb5c5 DB |
2277 | ovs_mutex_lock(&conn->lock); |
2278 | entry->mark = conn->mark; | |
286de272 | 2279 | memcpy(&entry->labels, &conn->label, sizeof entry->labels); |
4d4e68ed | 2280 | |
dec0dbbc | 2281 | long long expiration = conn->expiration - now; |
4d4e68ed | 2282 | |
dec0dbbc | 2283 | struct ct_l4_proto *class = l4_protos[conn->key.nw_proto]; |
4d4e68ed DDP |
2284 | if (class->conn_get_protoinfo) { |
2285 | class->conn_get_protoinfo(conn, &entry->protoinfo); | |
2286 | } | |
f1a0469e | 2287 | ovs_mutex_unlock(&conn->lock); |
bd5e81a0 | 2288 | |
f1a0469e | 2289 | entry->timeout = (expiration > 0) ? expiration / 1000 : 0; |
bd5e81a0 DB |
2290 | |
2291 | if (conn->alg) { | |
2292 | /* Caller is responsible for freeing. */ | |
2293 | entry->helper.name = xstrdup(conn->alg); | |
2294 | } | |
4d4e68ed DDP |
2295 | } |
2296 | ||
4ea96698 DB |
2297 | struct ipf * |
2298 | conntrack_ipf_ctx(struct conntrack *ct) | |
2299 | { | |
2300 | return ct->ipf; | |
2301 | } | |
2302 | ||
4d4e68ed DDP |
2303 | int |
2304 | conntrack_dump_start(struct conntrack *ct, struct conntrack_dump *dump, | |
ded30c74 | 2305 | const uint16_t *pzone, int *ptot_bkts) |
4d4e68ed DDP |
2306 | { |
2307 | memset(dump, 0, sizeof(*dump)); | |
dec0dbbc | 2308 | |
4d4e68ed DDP |
2309 | if (pzone) { |
2310 | dump->zone = *pzone; | |
2311 | dump->filter_zone = true; | |
2312 | } | |
4d4e68ed | 2313 | |
dec0dbbc | 2314 | dump->ct = ct; |
967bb5c5 | 2315 | *ptot_bkts = 1; /* Need to clean up the callers. */ |
4d4e68ed DDP |
2316 | return 0; |
2317 | } | |
2318 | ||
2319 | int | |
2320 | conntrack_dump_next(struct conntrack_dump *dump, struct ct_dpif_entry *entry) | |
2321 | { | |
2322 | struct conntrack *ct = dump->ct; | |
2323 | long long now = time_msec(); | |
2324 | ||
967bb5c5 DB |
2325 | for (;;) { |
2326 | struct cmap_node *cm_node = cmap_next_position(&ct->conns, | |
2327 | &dump->cm_pos); | |
2328 | if (!cm_node) { | |
2329 | break; | |
4d4e68ed | 2330 | } |
967bb5c5 DB |
2331 | struct conn *conn; |
2332 | INIT_CONTAINER(conn, cm_node, cm_node); | |
2333 | if ((!dump->filter_zone || conn->key.zone == dump->zone) && | |
2334 | (conn->conn_type != CT_CONN_TYPE_UN_NAT)) { | |
f1a0469e | 2335 | conn_to_ct_dpif_entry(conn, entry, now); |
4d4e68ed DDP |
2336 | return 0; |
2337 | } | |
2338 | } | |
967bb5c5 | 2339 | |
4d4e68ed DDP |
2340 | return EOF; |
2341 | } | |
2342 | ||
2343 | int | |
2344 | conntrack_dump_done(struct conntrack_dump *dump OVS_UNUSED) | |
2345 | { | |
2346 | return 0; | |
2347 | } | |
5d9cbb4c DDP |
2348 | |
2349 | int | |
2350 | conntrack_flush(struct conntrack *ct, const uint16_t *zone) | |
2351 | { | |
967bb5c5 DB |
2352 | struct conn *conn; |
2353 | ||
2354 | ovs_mutex_lock(&ct->ct_lock); | |
2355 | CMAP_FOR_EACH (conn, cm_node, &ct->conns) { | |
2356 | if (!zone || *zone == conn->key.zone) { | |
2357 | conn_clean_one(ct, conn); | |
5d9cbb4c | 2358 | } |
5d9cbb4c | 2359 | } |
967bb5c5 | 2360 | ovs_mutex_unlock(&ct->ct_lock); |
bd5e81a0 | 2361 | |
5d9cbb4c DDP |
2362 | return 0; |
2363 | } | |
bd5e81a0 | 2364 | |
271e48a0 YHW |
2365 | int |
2366 | conntrack_flush_tuple(struct conntrack *ct, const struct ct_dpif_tuple *tuple, | |
2367 | uint16_t zone) | |
2368 | { | |
271e48a0 | 2369 | int error = 0; |
4048c508 DB |
2370 | struct conn_key key; |
2371 | struct conn *conn; | |
271e48a0 | 2372 | |
4048c508 DB |
2373 | memset(&key, 0, sizeof(key)); |
2374 | tuple_to_conn_key(tuple, zone, &key); | |
967bb5c5 | 2375 | ovs_mutex_lock(&ct->ct_lock); |
4048c508 | 2376 | conn_lookup(ct, &key, time_msec(), &conn, NULL); |
271e48a0 | 2377 | |
4048c508 DB |
2378 | if (conn && conn->conn_type == CT_CONN_TYPE_DEFAULT) { |
2379 | conn_clean(ct, conn); | |
271e48a0 | 2380 | } else { |
a1d5eeff | 2381 | VLOG_WARN("Must flush tuple using the original pre-NATed tuple"); |
271e48a0 YHW |
2382 | error = ENOENT; |
2383 | } | |
967bb5c5 DB |
2384 | |
2385 | ovs_mutex_unlock(&ct->ct_lock); | |
271e48a0 YHW |
2386 | return error; |
2387 | } | |
2388 | ||
c92339ad DB |
2389 | int |
2390 | conntrack_set_maxconns(struct conntrack *ct, uint32_t maxconns) | |
2391 | { | |
2392 | atomic_store_relaxed(&ct->n_conn_limit, maxconns); | |
2393 | return 0; | |
2394 | } | |
2395 | ||
2396 | int | |
2397 | conntrack_get_maxconns(struct conntrack *ct, uint32_t *maxconns) | |
2398 | { | |
2399 | atomic_read_relaxed(&ct->n_conn_limit, maxconns); | |
2400 | return 0; | |
2401 | } | |
2402 | ||
875075b3 DB |
2403 | int |
2404 | conntrack_get_nconns(struct conntrack *ct, uint32_t *nconns) | |
2405 | { | |
2406 | *nconns = atomic_count_get(&ct->n_conn); | |
2407 | return 0; | |
2408 | } | |
2409 | ||
bd5e81a0 DB |
2410 | /* This function must be called with the ct->resources read lock taken. */ |
2411 | static struct alg_exp_node * | |
be38342d DB |
2412 | expectation_lookup(struct hmap *alg_expectations, const struct conn_key *key, |
2413 | uint32_t basis, bool src_ip_wc) | |
bd5e81a0 | 2414 | { |
c3f6bae2 DB |
2415 | struct conn_key check_key; |
2416 | memcpy(&check_key, key, sizeof check_key); | |
bd5e81a0 | 2417 | check_key.src.port = ALG_WC_SRC_PORT; |
dec0dbbc | 2418 | |
be38342d DB |
2419 | if (src_ip_wc) { |
2420 | memset(&check_key.src.addr, 0, sizeof check_key.src.addr); | |
2421 | } | |
dec0dbbc | 2422 | |
bd5e81a0 DB |
2423 | struct alg_exp_node *alg_exp_node; |
2424 | ||
bd5e81a0 | 2425 | HMAP_FOR_EACH_WITH_HASH (alg_exp_node, node, |
dec0dbbc | 2426 | conn_key_hash(&check_key, basis), |
bd5e81a0 DB |
2427 | alg_expectations) { |
2428 | if (!conn_key_cmp(&alg_exp_node->key, &check_key)) { | |
2429 | return alg_exp_node; | |
2430 | } | |
2431 | } | |
2432 | return NULL; | |
2433 | } | |
2434 | ||
4417ca3d DB |
2435 | /* This function must be called with the ct->resources write lock taken. */ |
2436 | static void | |
2437 | expectation_remove(struct hmap *alg_expectations, | |
2438 | const struct conn_key *key, uint32_t basis) | |
2439 | { | |
2440 | struct alg_exp_node *alg_exp_node; | |
2441 | ||
2442 | HMAP_FOR_EACH_WITH_HASH (alg_exp_node, node, conn_key_hash(key, basis), | |
2443 | alg_expectations) { | |
2444 | if (!conn_key_cmp(&alg_exp_node->key, key)) { | |
2445 | hmap_remove(alg_expectations, &alg_exp_node->node); | |
2446 | break; | |
2447 | } | |
2448 | } | |
2449 | } | |
2450 | ||
2451 | /* This function must be called with the ct->resources read lock taken. */ | |
2452 | static struct alg_exp_node * | |
2453 | expectation_ref_lookup_unique(const struct hindex *alg_expectation_refs, | |
2454 | const struct conn_key *master_key, | |
2455 | const struct conn_key *alg_exp_key, | |
2456 | uint32_t basis) | |
2457 | { | |
2458 | struct alg_exp_node *alg_exp_node; | |
2459 | ||
2460 | HINDEX_FOR_EACH_WITH_HASH (alg_exp_node, node_ref, | |
2461 | conn_key_hash(master_key, basis), | |
2462 | alg_expectation_refs) { | |
2463 | if (!conn_key_cmp(&alg_exp_node->master_key, master_key) && | |
2464 | !conn_key_cmp(&alg_exp_node->key, alg_exp_key)) { | |
2465 | return alg_exp_node; | |
2466 | } | |
2467 | } | |
2468 | return NULL; | |
2469 | } | |
2470 | ||
2471 | /* This function must be called with the ct->resources write lock taken. */ | |
2472 | static void | |
2473 | expectation_ref_create(struct hindex *alg_expectation_refs, | |
2474 | struct alg_exp_node *alg_exp_node, | |
2475 | uint32_t basis) | |
2476 | { | |
2477 | if (!expectation_ref_lookup_unique(alg_expectation_refs, | |
2478 | &alg_exp_node->master_key, | |
2479 | &alg_exp_node->key, basis)) { | |
2480 | hindex_insert(alg_expectation_refs, &alg_exp_node->node_ref, | |
2481 | conn_key_hash(&alg_exp_node->master_key, basis)); | |
2482 | } | |
2483 | } | |
2484 | ||
2485 | static void | |
967bb5c5 | 2486 | expectation_clean(struct conntrack *ct, const struct conn_key *master_key) |
4417ca3d | 2487 | { |
967bb5c5 | 2488 | ovs_rwlock_wrlock(&ct->resources_lock); |
4417ca3d DB |
2489 | |
2490 | struct alg_exp_node *node, *next; | |
2491 | HINDEX_FOR_EACH_WITH_HASH_SAFE (node, next, node_ref, | |
967bb5c5 | 2492 | conn_key_hash(master_key, ct->hash_basis), |
4417ca3d DB |
2493 | &ct->alg_expectation_refs) { |
2494 | if (!conn_key_cmp(&node->master_key, master_key)) { | |
967bb5c5 DB |
2495 | expectation_remove(&ct->alg_expectations, &node->key, |
2496 | ct->hash_basis); | |
4417ca3d DB |
2497 | hindex_remove(&ct->alg_expectation_refs, &node->node_ref); |
2498 | free(node); | |
2499 | } | |
2500 | } | |
2501 | ||
967bb5c5 | 2502 | ovs_rwlock_unlock(&ct->resources_lock); |
4417ca3d DB |
2503 | } |
2504 | ||
bd5e81a0 | 2505 | static void |
be38342d DB |
2506 | expectation_create(struct conntrack *ct, ovs_be16 dst_port, |
2507 | const struct conn *master_conn, bool reply, bool src_ip_wc, | |
2508 | bool skip_nat) | |
bd5e81a0 | 2509 | { |
cda1b109 DB |
2510 | union ct_addr src_addr; |
2511 | union ct_addr dst_addr; | |
2512 | union ct_addr alg_nat_repl_addr; | |
be38342d | 2513 | struct alg_exp_node *alg_exp_node = xzalloc(sizeof *alg_exp_node); |
bd5e81a0 | 2514 | |
be38342d | 2515 | if (reply) { |
bd5e81a0 DB |
2516 | src_addr = master_conn->key.src.addr; |
2517 | dst_addr = master_conn->key.dst.addr; | |
efa29a89 | 2518 | alg_exp_node->nat_rpl_dst = true; |
be38342d DB |
2519 | if (skip_nat) { |
2520 | alg_nat_repl_addr = dst_addr; | |
efa29a89 DM |
2521 | } else if (master_conn->nat_info && |
2522 | master_conn->nat_info->nat_action & NAT_ACTION_DST) { | |
2523 | alg_nat_repl_addr = master_conn->rev_key.src.addr; | |
2524 | alg_exp_node->nat_rpl_dst = false; | |
be38342d DB |
2525 | } else { |
2526 | alg_nat_repl_addr = master_conn->rev_key.dst.addr; | |
2527 | } | |
be38342d DB |
2528 | } else { |
2529 | src_addr = master_conn->rev_key.src.addr; | |
2530 | dst_addr = master_conn->rev_key.dst.addr; | |
efa29a89 | 2531 | alg_exp_node->nat_rpl_dst = false; |
be38342d DB |
2532 | if (skip_nat) { |
2533 | alg_nat_repl_addr = src_addr; | |
efa29a89 DM |
2534 | } else if (master_conn->nat_info && |
2535 | master_conn->nat_info->nat_action & NAT_ACTION_DST) { | |
2536 | alg_nat_repl_addr = master_conn->key.dst.addr; | |
2537 | alg_exp_node->nat_rpl_dst = true; | |
be38342d DB |
2538 | } else { |
2539 | alg_nat_repl_addr = master_conn->key.src.addr; | |
2540 | } | |
be38342d DB |
2541 | } |
2542 | if (src_ip_wc) { | |
2543 | memset(&src_addr, 0, sizeof src_addr); | |
bd5e81a0 DB |
2544 | } |
2545 | ||
bd5e81a0 DB |
2546 | alg_exp_node->key.dl_type = master_conn->key.dl_type; |
2547 | alg_exp_node->key.nw_proto = master_conn->key.nw_proto; | |
2548 | alg_exp_node->key.zone = master_conn->key.zone; | |
2549 | alg_exp_node->key.src.addr = src_addr; | |
2550 | alg_exp_node->key.dst.addr = dst_addr; | |
2551 | alg_exp_node->key.src.port = ALG_WC_SRC_PORT; | |
2552 | alg_exp_node->key.dst.port = dst_port; | |
2553 | alg_exp_node->master_mark = master_conn->mark; | |
2554 | alg_exp_node->master_label = master_conn->label; | |
82b9ac94 DB |
2555 | memcpy(&alg_exp_node->master_key, &master_conn->key, |
2556 | sizeof alg_exp_node->master_key); | |
bd5e81a0 DB |
2557 | /* Take the write lock here because it is almost 100% |
2558 | * likely that the lookup will fail and | |
2559 | * expectation_create() will be called below. */ | |
967bb5c5 | 2560 | ovs_rwlock_wrlock(&ct->resources_lock); |
bd5e81a0 | 2561 | struct alg_exp_node *alg_exp = expectation_lookup( |
be38342d | 2562 | &ct->alg_expectations, &alg_exp_node->key, ct->hash_basis, src_ip_wc); |
bd5e81a0 DB |
2563 | if (alg_exp) { |
2564 | free(alg_exp_node); | |
967bb5c5 | 2565 | ovs_rwlock_unlock(&ct->resources_lock); |
bd5e81a0 DB |
2566 | return; |
2567 | } | |
2568 | ||
2569 | alg_exp_node->alg_nat_repl_addr = alg_nat_repl_addr; | |
4417ca3d | 2570 | hmap_insert(&ct->alg_expectations, &alg_exp_node->node, |
dec0dbbc | 2571 | conn_key_hash(&alg_exp_node->key, ct->hash_basis)); |
4417ca3d DB |
2572 | expectation_ref_create(&ct->alg_expectation_refs, alg_exp_node, |
2573 | ct->hash_basis); | |
967bb5c5 | 2574 | ovs_rwlock_unlock(&ct->resources_lock); |
bd5e81a0 DB |
2575 | } |
2576 | ||
bd5e81a0 DB |
2577 | static void |
2578 | replace_substring(char *substr, uint8_t substr_size, | |
2579 | uint8_t total_size, char *rep_str, | |
2580 | uint8_t rep_str_size) | |
2581 | { | |
2582 | memmove(substr + rep_str_size, substr + substr_size, | |
2583 | total_size - substr_size); | |
2584 | memcpy(substr, rep_str, rep_str_size); | |
2585 | } | |
2586 | ||
cd7c99a6 DB |
2587 | static void |
2588 | repl_bytes(char *str, char c1, char c2) | |
2589 | { | |
2590 | while (*str) { | |
2591 | if (*str == c1) { | |
2592 | *str = c2; | |
2593 | } | |
2594 | str++; | |
2595 | } | |
2596 | } | |
2597 | ||
2598 | static void | |
2599 | modify_packet(struct dp_packet *pkt, char *pkt_str, size_t size, | |
2600 | char *repl_str, size_t repl_size, | |
2601 | uint32_t orig_used_size) | |
2602 | { | |
2603 | replace_substring(pkt_str, size, | |
2604 | (const char *) dp_packet_tail(pkt) - pkt_str, | |
2605 | repl_str, repl_size); | |
2606 | dp_packet_set_size(pkt, orig_used_size + (int) repl_size - (int) size); | |
2607 | } | |
2608 | ||
bd5e81a0 DB |
2609 | /* Replace IPV4 address in FTP message with NATed address. */ |
2610 | static int | |
2611 | repl_ftp_v4_addr(struct dp_packet *pkt, ovs_be32 v4_addr_rep, | |
2612 | char *ftp_data_start, | |
cd7c99a6 DB |
2613 | size_t addr_offset_from_ftp_data_start, |
2614 | size_t addr_size OVS_UNUSED) | |
bd5e81a0 DB |
2615 | { |
2616 | enum { MAX_FTP_V4_NAT_DELTA = 8 }; | |
2617 | ||
2618 | /* Do conservative check for pathological MTU usage. */ | |
2619 | uint32_t orig_used_size = dp_packet_size(pkt); | |
cd7c99a6 DB |
2620 | if (orig_used_size + MAX_FTP_V4_NAT_DELTA > |
2621 | dp_packet_get_allocated(pkt)) { | |
2622 | ||
bd5e81a0 | 2623 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5); |
cd7c99a6 DB |
2624 | VLOG_WARN_RL(&rl, "Unsupported effective MTU %u used with FTP V4", |
2625 | dp_packet_get_allocated(pkt)); | |
bd5e81a0 DB |
2626 | return 0; |
2627 | } | |
2628 | ||
cd7c99a6 DB |
2629 | char v4_addr_str[INET_ADDRSTRLEN] = {0}; |
2630 | ovs_assert(inet_ntop(AF_INET, &v4_addr_rep, v4_addr_str, | |
2631 | sizeof v4_addr_str)); | |
2632 | repl_bytes(v4_addr_str, '.', ','); | |
2633 | modify_packet(pkt, ftp_data_start + addr_offset_from_ftp_data_start, | |
2634 | addr_size, v4_addr_str, strlen(v4_addr_str), | |
2635 | orig_used_size); | |
2636 | return (int) strlen(v4_addr_str) - (int) addr_size; | |
bd5e81a0 DB |
2637 | } |
2638 | ||
2639 | static char * | |
2640 | skip_non_digits(char *str) | |
2641 | { | |
2642 | while (!isdigit(*str) && *str != 0) { | |
2643 | str++; | |
2644 | } | |
2645 | return str; | |
2646 | } | |
2647 | ||
2648 | static char * | |
2649 | terminate_number_str(char *str, uint8_t max_digits) | |
2650 | { | |
2651 | uint8_t digits_found = 0; | |
2652 | while (isdigit(*str) && digits_found <= max_digits) { | |
2653 | str++; | |
2654 | digits_found++; | |
2655 | } | |
2656 | ||
2657 | *str = 0; | |
2658 | return str; | |
2659 | } | |
2660 | ||
2661 | ||
2662 | static void | |
2663 | get_ftp_ctl_msg(struct dp_packet *pkt, char *ftp_msg) | |
2664 | { | |
2665 | struct tcp_header *th = dp_packet_l4(pkt); | |
2666 | char *tcp_hdr = (char *) th; | |
2667 | uint32_t tcp_payload_len = tcp_payload_length(pkt); | |
2668 | size_t tcp_payload_of_interest = MIN(tcp_payload_len, | |
2669 | LARGEST_FTP_MSG_OF_INTEREST); | |
2670 | size_t tcp_hdr_len = TCP_OFFSET(th->tcp_ctl) * 4; | |
2671 | ||
2672 | ovs_strlcpy(ftp_msg, tcp_hdr + tcp_hdr_len, | |
2673 | tcp_payload_of_interest); | |
2674 | } | |
2675 | ||
2676 | static enum ftp_ctl_pkt | |
2677 | detect_ftp_ctl_type(const struct conn_lookup_ctx *ctx, | |
2678 | struct dp_packet *pkt) | |
2679 | { | |
bd5e81a0 DB |
2680 | char ftp_msg[LARGEST_FTP_MSG_OF_INTEREST + 1] = {0}; |
2681 | get_ftp_ctl_msg(pkt, ftp_msg); | |
dec0dbbc | 2682 | |
bd5e81a0 DB |
2683 | if (ctx->key.dl_type == htons(ETH_TYPE_IPV6)) { |
2684 | if (strncasecmp(ftp_msg, FTP_EPRT_CMD, strlen(FTP_EPRT_CMD)) && | |
2685 | !strcasestr(ftp_msg, FTP_EPSV_REPLY)) { | |
2686 | return CT_FTP_CTL_OTHER; | |
2687 | } | |
2688 | } else { | |
2689 | if (strncasecmp(ftp_msg, FTP_PORT_CMD, strlen(FTP_PORT_CMD)) && | |
2690 | strncasecmp(ftp_msg, FTP_PASV_REPLY_CODE, | |
2691 | strlen(FTP_PASV_REPLY_CODE))) { | |
2692 | return CT_FTP_CTL_OTHER; | |
2693 | } | |
2694 | } | |
2695 | ||
2696 | return CT_FTP_CTL_INTEREST; | |
2697 | } | |
2698 | ||
2699 | static enum ftp_ctl_pkt | |
2700 | process_ftp_ctl_v4(struct conntrack *ct, | |
2701 | struct dp_packet *pkt, | |
2702 | const struct conn *conn_for_expectation, | |
4417ca3d | 2703 | ovs_be32 *v4_addr_rep, |
bd5e81a0 | 2704 | char **ftp_data_v4_start, |
cd7c99a6 DB |
2705 | size_t *addr_offset_from_ftp_data_start, |
2706 | size_t *addr_size) | |
bd5e81a0 DB |
2707 | { |
2708 | struct tcp_header *th = dp_packet_l4(pkt); | |
2709 | size_t tcp_hdr_len = TCP_OFFSET(th->tcp_ctl) * 4; | |
2710 | char *tcp_hdr = (char *) th; | |
2711 | *ftp_data_v4_start = tcp_hdr + tcp_hdr_len; | |
2712 | char ftp_msg[LARGEST_FTP_MSG_OF_INTEREST + 1] = {0}; | |
2713 | get_ftp_ctl_msg(pkt, ftp_msg); | |
bd5e81a0 DB |
2714 | char *ftp = ftp_msg; |
2715 | enum ct_alg_mode mode; | |
dec0dbbc | 2716 | |
23bea975 | 2717 | if (!strncasecmp(ftp, FTP_PORT_CMD, strlen(FTP_PORT_CMD))) { |
bd5e81a0 DB |
2718 | ftp = ftp_msg + strlen(FTP_PORT_CMD); |
2719 | mode = CT_FTP_MODE_ACTIVE; | |
2720 | } else { | |
2721 | ftp = ftp_msg + strlen(FTP_PASV_REPLY_CODE); | |
2722 | mode = CT_FTP_MODE_PASSIVE; | |
2723 | } | |
2724 | ||
2725 | /* Find first space. */ | |
2726 | ftp = strchr(ftp, ' '); | |
2727 | if (!ftp) { | |
2728 | return CT_FTP_CTL_INVALID; | |
2729 | } | |
2730 | ||
2731 | /* Find the first digit, after space. */ | |
2732 | ftp = skip_non_digits(ftp); | |
2733 | if (*ftp == 0) { | |
2734 | return CT_FTP_CTL_INVALID; | |
2735 | } | |
2736 | ||
2737 | char *ip_addr_start = ftp; | |
2738 | *addr_offset_from_ftp_data_start = ip_addr_start - ftp_msg; | |
bd5e81a0 | 2739 | |
dec0dbbc | 2740 | uint8_t comma_count = 0; |
bd5e81a0 DB |
2741 | while (comma_count < 4 && *ftp) { |
2742 | if (*ftp == ',') { | |
2743 | comma_count++; | |
2744 | if (comma_count == 4) { | |
2745 | *ftp = 0; | |
2746 | } else { | |
2747 | *ftp = '.'; | |
2748 | } | |
2749 | } | |
2750 | ftp++; | |
2751 | } | |
2752 | if (comma_count != 4) { | |
2753 | return CT_FTP_CTL_INVALID; | |
2754 | } | |
2755 | ||
2756 | struct in_addr ip_addr; | |
2757 | int rc2 = inet_pton(AF_INET, ip_addr_start, &ip_addr); | |
2758 | if (rc2 != 1) { | |
2759 | return CT_FTP_CTL_INVALID; | |
2760 | } | |
2761 | ||
cd7c99a6 | 2762 | *addr_size = ftp - ip_addr_start - 1; |
bd5e81a0 DB |
2763 | char *save_ftp = ftp; |
2764 | ftp = terminate_number_str(ftp, MAX_FTP_PORT_DGTS); | |
2765 | if (!ftp) { | |
2766 | return CT_FTP_CTL_INVALID; | |
2767 | } | |
2768 | int value; | |
2769 | if (!str_to_int(save_ftp, 10, &value)) { | |
2770 | return CT_FTP_CTL_INVALID; | |
2771 | } | |
2772 | ||
2773 | /* This is derived from the L4 port maximum is 65535. */ | |
2774 | if (value > 255) { | |
2775 | return CT_FTP_CTL_INVALID; | |
2776 | } | |
2777 | ||
2778 | uint16_t port_hs = value; | |
2779 | port_hs <<= 8; | |
2780 | ||
2781 | /* Skip over comma. */ | |
2782 | ftp++; | |
2783 | save_ftp = ftp; | |
2784 | bool digit_found = false; | |
2785 | while (isdigit(*ftp)) { | |
2786 | ftp++; | |
2787 | digit_found = true; | |
2788 | } | |
2789 | if (!digit_found) { | |
2790 | return CT_FTP_CTL_INVALID; | |
2791 | } | |
2792 | *ftp = 0; | |
2793 | if (!str_to_int(save_ftp, 10, &value)) { | |
2794 | return CT_FTP_CTL_INVALID; | |
2795 | } | |
2796 | ||
2797 | if (value > 255) { | |
2798 | return CT_FTP_CTL_INVALID; | |
2799 | } | |
2800 | ||
78a0b272 | 2801 | port_hs |= value; |
bd5e81a0 DB |
2802 | ovs_be16 port = htons(port_hs); |
2803 | ovs_be32 conn_ipv4_addr; | |
2804 | ||
2805 | switch (mode) { | |
2806 | case CT_FTP_MODE_ACTIVE: | |
cda1b109 DB |
2807 | *v4_addr_rep = conn_for_expectation->rev_key.dst.addr.ipv4; |
2808 | conn_ipv4_addr = conn_for_expectation->key.src.addr.ipv4; | |
bd5e81a0 DB |
2809 | break; |
2810 | case CT_FTP_MODE_PASSIVE: | |
cda1b109 DB |
2811 | *v4_addr_rep = conn_for_expectation->key.dst.addr.ipv4; |
2812 | conn_ipv4_addr = conn_for_expectation->rev_key.src.addr.ipv4; | |
bd5e81a0 | 2813 | break; |
7be77cb0 | 2814 | case CT_TFTP_MODE: |
bd5e81a0 DB |
2815 | default: |
2816 | OVS_NOT_REACHED(); | |
2817 | } | |
2818 | ||
2819 | ovs_be32 ftp_ipv4_addr; | |
2820 | ftp_ipv4_addr = ip_addr.s_addr; | |
2821 | /* Although most servers will block this exploit, there may be some | |
2822 | * less well managed. */ | |
2823 | if (ftp_ipv4_addr != conn_ipv4_addr && ftp_ipv4_addr != *v4_addr_rep) { | |
2824 | return CT_FTP_CTL_INVALID; | |
2825 | } | |
2826 | ||
be38342d DB |
2827 | expectation_create(ct, port, conn_for_expectation, |
2828 | !!(pkt->md.ct_state & CS_REPLY_DIR), false, false); | |
bd5e81a0 DB |
2829 | return CT_FTP_CTL_INTEREST; |
2830 | } | |
2831 | ||
2832 | static char * | |
2833 | skip_ipv6_digits(char *str) | |
2834 | { | |
2835 | while (isxdigit(*str) || *str == ':' || *str == '.') { | |
2836 | str++; | |
2837 | } | |
2838 | return str; | |
2839 | } | |
2840 | ||
2841 | static enum ftp_ctl_pkt | |
2842 | process_ftp_ctl_v6(struct conntrack *ct, | |
2843 | struct dp_packet *pkt, | |
2844 | const struct conn *conn_for_expectation, | |
cda1b109 | 2845 | union ct_addr *v6_addr_rep, char **ftp_data_start, |
bd5e81a0 DB |
2846 | size_t *addr_offset_from_ftp_data_start, |
2847 | size_t *addr_size, enum ct_alg_mode *mode) | |
2848 | { | |
2849 | struct tcp_header *th = dp_packet_l4(pkt); | |
2850 | size_t tcp_hdr_len = TCP_OFFSET(th->tcp_ctl) * 4; | |
2851 | char *tcp_hdr = (char *) th; | |
2852 | char ftp_msg[LARGEST_FTP_MSG_OF_INTEREST + 1] = {0}; | |
bd5e81a0 DB |
2853 | get_ftp_ctl_msg(pkt, ftp_msg); |
2854 | *ftp_data_start = tcp_hdr + tcp_hdr_len; | |
bd5e81a0 DB |
2855 | char *ftp = ftp_msg; |
2856 | struct in6_addr ip6_addr; | |
dec0dbbc | 2857 | |
23bea975 | 2858 | if (!strncasecmp(ftp, FTP_EPRT_CMD, strlen(FTP_EPRT_CMD))) { |
bd5e81a0 DB |
2859 | ftp = ftp_msg + strlen(FTP_EPRT_CMD); |
2860 | ftp = skip_non_digits(ftp); | |
2861 | if (*ftp != FTP_AF_V6 || isdigit(ftp[1])) { | |
2862 | return CT_FTP_CTL_INVALID; | |
2863 | } | |
2864 | /* Jump over delimiter. */ | |
2865 | ftp += 2; | |
2866 | ||
bd5e81a0 | 2867 | memset(&ip6_addr, 0, sizeof ip6_addr); |
dec0dbbc | 2868 | char *ip_addr_start = ftp; |
bd5e81a0 DB |
2869 | *addr_offset_from_ftp_data_start = ip_addr_start - ftp_msg; |
2870 | ftp = skip_ipv6_digits(ftp); | |
2871 | *ftp = 0; | |
2872 | *addr_size = ftp - ip_addr_start; | |
2873 | int rc2 = inet_pton(AF_INET6, ip_addr_start, &ip6_addr); | |
2874 | if (rc2 != 1) { | |
2875 | return CT_FTP_CTL_INVALID; | |
2876 | } | |
2877 | ftp++; | |
2878 | *mode = CT_FTP_MODE_ACTIVE; | |
2879 | } else { | |
2880 | ftp = ftp_msg + strcspn(ftp_msg, "("); | |
2881 | ftp = skip_non_digits(ftp); | |
2882 | if (!isdigit(*ftp)) { | |
2883 | return CT_FTP_CTL_INVALID; | |
2884 | } | |
2885 | ||
2886 | /* Not used for passive mode. */ | |
2887 | *addr_offset_from_ftp_data_start = 0; | |
2888 | *addr_size = 0; | |
2889 | ||
2890 | *mode = CT_FTP_MODE_PASSIVE; | |
2891 | } | |
2892 | ||
2893 | char *save_ftp = ftp; | |
2894 | ftp = terminate_number_str(ftp, MAX_EXT_FTP_PORT_DGTS); | |
2895 | if (!ftp) { | |
2896 | return CT_FTP_CTL_INVALID; | |
2897 | } | |
dec0dbbc | 2898 | |
bd5e81a0 DB |
2899 | int value; |
2900 | if (!str_to_int(save_ftp, 10, &value)) { | |
2901 | return CT_FTP_CTL_INVALID; | |
2902 | } | |
2903 | if (value > CT_MAX_L4_PORT) { | |
2904 | return CT_FTP_CTL_INVALID; | |
2905 | } | |
2906 | ||
2907 | uint16_t port_hs = value; | |
2908 | ovs_be16 port = htons(port_hs); | |
2909 | ||
2910 | switch (*mode) { | |
2911 | case CT_FTP_MODE_ACTIVE: | |
2912 | *v6_addr_rep = conn_for_expectation->rev_key.dst.addr; | |
2913 | /* Although most servers will block this exploit, there may be some | |
2914 | * less well managed. */ | |
cda1b109 DB |
2915 | if (memcmp(&ip6_addr, &v6_addr_rep->ipv6, sizeof ip6_addr) && |
2916 | memcmp(&ip6_addr, &conn_for_expectation->key.src.addr.ipv6, | |
bd5e81a0 DB |
2917 | sizeof ip6_addr)) { |
2918 | return CT_FTP_CTL_INVALID; | |
2919 | } | |
2920 | break; | |
2921 | case CT_FTP_MODE_PASSIVE: | |
2922 | *v6_addr_rep = conn_for_expectation->key.dst.addr; | |
2923 | break; | |
7be77cb0 | 2924 | case CT_TFTP_MODE: |
bd5e81a0 DB |
2925 | default: |
2926 | OVS_NOT_REACHED(); | |
2927 | } | |
2928 | ||
be38342d DB |
2929 | expectation_create(ct, port, conn_for_expectation, |
2930 | !!(pkt->md.ct_state & CS_REPLY_DIR), false, false); | |
bd5e81a0 DB |
2931 | return CT_FTP_CTL_INTEREST; |
2932 | } | |
2933 | ||
2934 | static int | |
cda1b109 | 2935 | repl_ftp_v6_addr(struct dp_packet *pkt, union ct_addr v6_addr_rep, |
bd5e81a0 DB |
2936 | char *ftp_data_start, |
2937 | size_t addr_offset_from_ftp_data_start, | |
2938 | size_t addr_size, enum ct_alg_mode mode) | |
2939 | { | |
2940 | /* This is slightly bigger than really possible. */ | |
2941 | enum { MAX_FTP_V6_NAT_DELTA = 45 }; | |
2942 | ||
2943 | if (mode == CT_FTP_MODE_PASSIVE) { | |
2944 | return 0; | |
2945 | } | |
2946 | ||
2947 | /* Do conservative check for pathological MTU usage. */ | |
2948 | uint32_t orig_used_size = dp_packet_size(pkt); | |
cd7c99a6 DB |
2949 | if (orig_used_size + MAX_FTP_V6_NAT_DELTA > |
2950 | dp_packet_get_allocated(pkt)) { | |
2951 | ||
bd5e81a0 | 2952 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5); |
cd7c99a6 DB |
2953 | VLOG_WARN_RL(&rl, "Unsupported effective MTU %u used with FTP V6", |
2954 | dp_packet_get_allocated(pkt)); | |
bd5e81a0 DB |
2955 | return 0; |
2956 | } | |
2957 | ||
298530b8 | 2958 | char v6_addr_str[INET6_ADDRSTRLEN] = {0}; |
cda1b109 | 2959 | ovs_assert(inet_ntop(AF_INET6, &v6_addr_rep.ipv6, v6_addr_str, |
298530b8 | 2960 | sizeof v6_addr_str)); |
cd7c99a6 DB |
2961 | modify_packet(pkt, ftp_data_start + addr_offset_from_ftp_data_start, |
2962 | addr_size, v6_addr_str, strlen(v6_addr_str), | |
2963 | orig_used_size); | |
2964 | return (int) strlen(v6_addr_str) - (int) addr_size; | |
bd5e81a0 DB |
2965 | } |
2966 | ||
d13d7115 DB |
2967 | /* Increment/decrement a TCP sequence number. */ |
2968 | static void | |
2969 | adj_seqnum(ovs_16aligned_be32 *val, int32_t inc) | |
2970 | { | |
2971 | put_16aligned_be32(val, htonl(ntohl(get_16aligned_be32(val)) + inc)); | |
2972 | } | |
2973 | ||
bd5e81a0 DB |
2974 | static void |
2975 | handle_ftp_ctl(struct conntrack *ct, const struct conn_lookup_ctx *ctx, | |
967bb5c5 | 2976 | struct dp_packet *pkt, struct conn *ec, long long now, |
253e4dc0 | 2977 | enum ftp_ctl_pkt ftp_ctl, bool nat) |
bd5e81a0 DB |
2978 | { |
2979 | struct ip_header *l3_hdr = dp_packet_l3(pkt); | |
2980 | ovs_be32 v4_addr_rep = 0; | |
cda1b109 | 2981 | union ct_addr v6_addr_rep; |
faa0826d | 2982 | size_t addr_offset_from_ftp_data_start = 0; |
bd5e81a0 DB |
2983 | size_t addr_size = 0; |
2984 | char *ftp_data_start; | |
bd5e81a0 DB |
2985 | enum ct_alg_mode mode = CT_FTP_MODE_ACTIVE; |
2986 | ||
2987 | if (detect_ftp_ctl_type(ctx, pkt) != ftp_ctl) { | |
2988 | return; | |
2989 | } | |
2990 | ||
bd5e81a0 DB |
2991 | struct ovs_16aligned_ip6_hdr *nh6 = dp_packet_l3(pkt); |
2992 | int64_t seq_skew = 0; | |
dec0dbbc | 2993 | |
253e4dc0 | 2994 | if (ftp_ctl == CT_FTP_CTL_INTEREST) { |
bd5e81a0 DB |
2995 | enum ftp_ctl_pkt rc; |
2996 | if (ctx->key.dl_type == htons(ETH_TYPE_IPV6)) { | |
253e4dc0 | 2997 | rc = process_ftp_ctl_v6(ct, pkt, ec, |
4417ca3d | 2998 | &v6_addr_rep, &ftp_data_start, |
bd5e81a0 DB |
2999 | &addr_offset_from_ftp_data_start, |
3000 | &addr_size, &mode); | |
3001 | } else { | |
253e4dc0 | 3002 | rc = process_ftp_ctl_v4(ct, pkt, ec, |
4417ca3d | 3003 | &v4_addr_rep, &ftp_data_start, |
cd7c99a6 DB |
3004 | &addr_offset_from_ftp_data_start, |
3005 | &addr_size); | |
bd5e81a0 DB |
3006 | } |
3007 | if (rc == CT_FTP_CTL_INVALID) { | |
3008 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5); | |
3009 | VLOG_WARN_RL(&rl, "Invalid FTP control packet format"); | |
3010 | pkt->md.ct_state |= CS_TRACKED | CS_INVALID; | |
3011 | return; | |
3012 | } else if (rc == CT_FTP_CTL_INTEREST) { | |
3013 | uint16_t ip_len; | |
dec0dbbc | 3014 | |
bd5e81a0 | 3015 | if (ctx->key.dl_type == htons(ETH_TYPE_IPV6)) { |
253e4dc0 DM |
3016 | if (nat) { |
3017 | seq_skew = repl_ftp_v6_addr(pkt, v6_addr_rep, | |
3018 | ftp_data_start, | |
3019 | addr_offset_from_ftp_data_start, | |
3020 | addr_size, mode); | |
3021 | } | |
3022 | ||
bd5e81a0 | 3023 | if (seq_skew) { |
253e4dc0 DM |
3024 | ip_len = ntohs(nh6->ip6_ctlun.ip6_un1.ip6_un1_plen) + |
3025 | seq_skew; | |
bd5e81a0 | 3026 | nh6->ip6_ctlun.ip6_un1.ip6_un1_plen = htons(ip_len); |
bd5e81a0 DB |
3027 | } |
3028 | } else { | |
253e4dc0 DM |
3029 | if (nat) { |
3030 | seq_skew = repl_ftp_v4_addr(pkt, v4_addr_rep, | |
3031 | ftp_data_start, | |
cd7c99a6 DB |
3032 | addr_offset_from_ftp_data_start, |
3033 | addr_size); | |
253e4dc0 | 3034 | } |
bd5e81a0 | 3035 | if (seq_skew) { |
253e4dc0 | 3036 | ip_len = ntohs(l3_hdr->ip_tot_len) + seq_skew; |
bd5e81a0 DB |
3037 | l3_hdr->ip_csum = recalc_csum16(l3_hdr->ip_csum, |
3038 | l3_hdr->ip_tot_len, htons(ip_len)); | |
3039 | l3_hdr->ip_tot_len = htons(ip_len); | |
bd5e81a0 DB |
3040 | } |
3041 | } | |
3042 | } else { | |
3043 | OVS_NOT_REACHED(); | |
3044 | } | |
bd5e81a0 DB |
3045 | } |
3046 | ||
3047 | struct tcp_header *th = dp_packet_l4(pkt); | |
dec0dbbc | 3048 | |
253e4dc0 | 3049 | if (nat && ec->seq_skew != 0) { |
d13d7115 DB |
3050 | ctx->reply != ec->seq_skew_dir ? |
3051 | adj_seqnum(&th->tcp_ack, -ec->seq_skew) : | |
3052 | adj_seqnum(&th->tcp_seq, ec->seq_skew); | |
bd5e81a0 DB |
3053 | } |
3054 | ||
bd5e81a0 | 3055 | th->tcp_csum = 0; |
bd5e81a0 | 3056 | if (ctx->key.dl_type == htons(ETH_TYPE_IPV6)) { |
76d85771 DB |
3057 | th->tcp_csum = packet_csum_upperlayer6(nh6, th, ctx->key.nw_proto, |
3058 | dp_packet_l4_size(pkt)); | |
bd5e81a0 | 3059 | } else { |
76d85771 DB |
3060 | uint32_t tcp_csum = packet_csum_pseudoheader(l3_hdr); |
3061 | th->tcp_csum = csum_finish( | |
3062 | csum_continue(tcp_csum, th, dp_packet_l4_size(pkt))); | |
bd5e81a0 | 3063 | } |
253e4dc0 DM |
3064 | |
3065 | if (seq_skew) { | |
967bb5c5 | 3066 | conn_seq_skew_set(ct, ec, now, seq_skew + ec->seq_skew, |
253e4dc0 DM |
3067 | ctx->reply); |
3068 | } | |
bd5e81a0 | 3069 | } |
7be77cb0 DB |
3070 | |
3071 | static void | |
3072 | handle_tftp_ctl(struct conntrack *ct, | |
94e71143 | 3073 | const struct conn_lookup_ctx *ctx OVS_UNUSED, |
967bb5c5 DB |
3074 | struct dp_packet *pkt, struct conn *conn_for_expectation, |
3075 | long long now OVS_UNUSED, enum ftp_ctl_pkt ftp_ctl OVS_UNUSED, | |
3076 | bool nat OVS_UNUSED) | |
7be77cb0 | 3077 | { |
be38342d DB |
3078 | expectation_create(ct, conn_for_expectation->key.src.port, |
3079 | conn_for_expectation, | |
3080 | !!(pkt->md.ct_state & CS_REPLY_DIR), false, false); | |
7be77cb0 | 3081 | } |