]>
Commit | Line | Data |
---|---|---|
d13ee228 | 1 | /* |
d40533fc | 2 | * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2019 Nicira, Inc. |
d13ee228 BP |
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> | |
18 | ||
19 | #include "ofproto-dpif-trace.h" | |
20 | ||
0f2f05bb | 21 | #include "conntrack.h" |
d13ee228 BP |
22 | #include "dpif.h" |
23 | #include "ofproto-dpif-xlate.h" | |
d13ee228 BP |
24 | #include "unixctl.h" |
25 | ||
2d9b49dd BP |
26 | static void oftrace_node_destroy(struct oftrace_node *); |
27 | ||
28 | /* Creates a new oftrace_node, populates it with the given 'type' and a copy of | |
29 | * 'text', and appends it to list 'super'. The caller retains ownership of | |
30 | * 'text'. */ | |
31 | struct oftrace_node * | |
32 | oftrace_report(struct ovs_list *super, enum oftrace_node_type type, | |
33 | const char *text) | |
d13ee228 | 34 | { |
2d9b49dd BP |
35 | struct oftrace_node *node = xmalloc(sizeof *node); |
36 | ovs_list_push_back(super, &node->node); | |
37 | node->type = type; | |
38 | node->text = xstrdup(text); | |
39 | ovs_list_init(&node->subs); | |
d13ee228 | 40 | |
2d9b49dd | 41 | return node; |
d13ee228 BP |
42 | } |
43 | ||
2d9b49dd BP |
44 | static bool |
45 | oftrace_node_type_is_terminal(enum oftrace_node_type type) | |
d13ee228 | 46 | { |
2d9b49dd BP |
47 | switch (type) { |
48 | case OFT_ACTION: | |
49 | case OFT_DETAIL: | |
50 | case OFT_WARN: | |
51 | case OFT_ERROR: | |
fac4786a | 52 | case OFT_BUCKET: |
2d9b49dd BP |
53 | return true; |
54 | ||
55 | case OFT_BRIDGE: | |
56 | case OFT_TABLE: | |
57 | case OFT_THAW: | |
58 | return false; | |
d13ee228 | 59 | } |
2d9b49dd BP |
60 | |
61 | OVS_NOT_REACHED(); | |
d13ee228 BP |
62 | } |
63 | ||
64 | static void | |
2d9b49dd | 65 | oftrace_node_list_destroy(struct ovs_list *nodes) |
d13ee228 | 66 | { |
2d9b49dd BP |
67 | if (nodes) { |
68 | struct oftrace_node *node, *next; | |
69 | LIST_FOR_EACH_SAFE (node, next, node, nodes) { | |
70 | ovs_list_remove(&node->node); | |
71 | oftrace_node_destroy(node); | |
72 | } | |
d13ee228 | 73 | } |
d13ee228 BP |
74 | } |
75 | ||
76 | static void | |
2d9b49dd | 77 | oftrace_node_destroy(struct oftrace_node *node) |
d13ee228 | 78 | { |
2d9b49dd BP |
79 | if (node) { |
80 | oftrace_node_list_destroy(&node->subs); | |
81 | free(node->text); | |
82 | free(node); | |
83 | } | |
d13ee228 BP |
84 | } |
85 | ||
e6bc8e74 YHW |
86 | bool |
87 | oftrace_add_recirc_node(struct ovs_list *recirc_queue, | |
88 | enum oftrace_recirc_type type, const struct flow *flow, | |
d072d2de | 89 | const struct ofpact_nat *ofn, |
5fdd80cc YHW |
90 | const struct dp_packet *packet, uint32_t recirc_id, |
91 | const uint16_t zone) | |
e6bc8e74 YHW |
92 | { |
93 | if (!recirc_id_node_find_and_ref(recirc_id)) { | |
94 | return false; | |
95 | } | |
96 | ||
97 | struct oftrace_recirc_node *node = xmalloc(sizeof *node); | |
98 | ovs_list_push_back(recirc_queue, &node->node); | |
99 | ||
100 | node->type = type; | |
101 | node->recirc_id = recirc_id; | |
102 | node->flow = *flow; | |
103 | node->flow.recirc_id = recirc_id; | |
5fdd80cc | 104 | node->flow.ct_zone = zone; |
d072d2de | 105 | node->nat_act = ofn; |
e6bc8e74 YHW |
106 | node->packet = packet ? dp_packet_clone(packet) : NULL; |
107 | ||
108 | return true; | |
109 | } | |
110 | ||
111 | static void | |
112 | oftrace_recirc_node_destroy(struct oftrace_recirc_node *node) | |
113 | { | |
114 | if (node) { | |
115 | recirc_free_id(node->recirc_id); | |
116 | dp_packet_delete(node->packet); | |
117 | free(node); | |
118 | } | |
119 | } | |
120 | ||
0f2f05bb YHW |
121 | static void |
122 | oftrace_push_ct_state(struct ovs_list *next_ct_states, uint32_t ct_state) | |
123 | { | |
124 | struct oftrace_next_ct_state *next_ct_state = | |
125 | xmalloc(sizeof *next_ct_state); | |
126 | next_ct_state->state = ct_state; | |
127 | ovs_list_push_back(next_ct_states, &next_ct_state->node); | |
128 | } | |
129 | ||
130 | static uint32_t | |
131 | oftrace_pop_ct_state(struct ovs_list *next_ct_states) | |
132 | { | |
133 | struct oftrace_next_ct_state *s; | |
134 | LIST_FOR_EACH_POP (s, node, next_ct_states) { | |
36e67140 YHW |
135 | uint32_t state = s->state; |
136 | free(s); | |
137 | return state; | |
0f2f05bb YHW |
138 | } |
139 | OVS_NOT_REACHED(); | |
140 | } | |
141 | ||
d13ee228 | 142 | static void |
2d9b49dd BP |
143 | oftrace_node_print_details(struct ds *output, |
144 | const struct ovs_list *nodes, int level) | |
d13ee228 | 145 | { |
2d9b49dd BP |
146 | const struct oftrace_node *sub; |
147 | LIST_FOR_EACH (sub, node, nodes) { | |
148 | if (sub->type == OFT_BRIDGE) { | |
149 | ds_put_char(output, '\n'); | |
150 | } | |
d13ee228 | 151 | |
2d9b49dd BP |
152 | bool more = (sub->node.next != nodes |
153 | || oftrace_node_type_is_terminal(sub->type)); | |
154 | ||
155 | ds_put_char_multiple(output, ' ', (level + more) * 4); | |
156 | switch (sub->type) { | |
157 | case OFT_DETAIL: | |
158 | ds_put_format(output, " -> %s\n", sub->text); | |
159 | break; | |
160 | case OFT_WARN: | |
161 | ds_put_format(output, " >> %s\n", sub->text); | |
162 | break; | |
163 | case OFT_ERROR: | |
164 | ds_put_format(output, " >>>> %s <<<<\n", sub->text); | |
165 | break; | |
166 | case OFT_BRIDGE: | |
167 | ds_put_format(output, "%s\n", sub->text); | |
168 | ds_put_char_multiple(output, ' ', (level + more) * 4); | |
169 | ds_put_char_multiple(output, '-', strlen(sub->text)); | |
170 | ds_put_char(output, '\n'); | |
171 | break; | |
172 | case OFT_TABLE: | |
fac4786a | 173 | case OFT_BUCKET: |
2d9b49dd BP |
174 | case OFT_THAW: |
175 | case OFT_ACTION: | |
176 | ds_put_format(output, "%s\n", sub->text); | |
177 | break; | |
d13ee228 | 178 | } |
d13ee228 | 179 | |
2d9b49dd | 180 | oftrace_node_print_details(output, &sub->subs, level + more + more); |
d13ee228 | 181 | } |
d13ee228 BP |
182 | } |
183 | ||
d072d2de DC |
184 | static void |
185 | oftrace_print_ip_flow(const struct flow *flow, int af, struct ds *output) | |
186 | { | |
187 | if (af == AF_INET) { | |
188 | ds_put_format(output, "nw_src="IP_FMT",tp_src=%"PRIu16"," | |
189 | "nw_dst="IP_FMT",tp_dst=%"PRIu16, | |
190 | IP_ARGS(flow->nw_src), ntohs(flow->tp_src), | |
191 | IP_ARGS(flow->nw_dst), ntohs(flow->tp_dst)); | |
192 | } else if (af == AF_INET6) { | |
193 | ds_put_cstr(output, "ipv6_src="); | |
194 | ipv6_format_addr_bracket(&flow->ipv6_src, output, true); | |
195 | ds_put_format(output, ",tp_src=%"PRIu16, ntohs(flow->tp_src)); | |
196 | ds_put_cstr(output, ",ipv6_dst="); | |
197 | ipv6_format_addr_bracket(&flow->ipv6_dst, output, true); | |
198 | ds_put_format(output, ",tp_dst=%"PRIu16, ntohs(flow->tp_dst)); | |
199 | } | |
200 | ds_put_char(output, '\n'); | |
201 | } | |
202 | ||
d13ee228 BP |
203 | /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following |
204 | * forms are supported: | |
205 | * | |
b490b189 BP |
206 | * - [options] [dpname] odp_flow [packet] |
207 | * - [options] bridge br_flow [packet] | |
d13ee228 BP |
208 | * |
209 | * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure | |
210 | * returns a nonnull malloced error message. */ | |
211 | static char * OVS_WARN_UNUSED_RESULT | |
212 | parse_flow_and_packet(int argc, const char *argv[], | |
213 | struct ofproto_dpif **ofprotop, struct flow *flow, | |
0f2f05bb | 214 | struct dp_packet **packetp, |
b490b189 BP |
215 | struct ovs_list *next_ct_states, |
216 | bool *consistent) | |
d13ee228 BP |
217 | { |
218 | const struct dpif_backer *backer = NULL; | |
d40533fc | 219 | char *error = NULL; |
d13ee228 | 220 | struct simap port_names = SIMAP_INITIALIZER(&port_names); |
b490b189 | 221 | struct dp_packet *packet = NULL; |
6f068379 BP |
222 | uint8_t *l7 = NULL; |
223 | size_t l7_len = 64; | |
d13ee228 BP |
224 | struct ofpbuf odp_key; |
225 | struct ofpbuf odp_mask; | |
226 | ||
227 | ofpbuf_init(&odp_key, 0); | |
228 | ofpbuf_init(&odp_mask, 0); | |
229 | ||
b490b189 BP |
230 | const char *args[3]; |
231 | int n_args = 0; | |
232 | bool generate_packet = false; | |
233 | if (consistent) { | |
234 | *consistent = false; | |
d13ee228 | 235 | } |
b490b189 BP |
236 | for (int i = 1; i < argc; i++) { |
237 | const char *arg = argv[i]; | |
238 | if (!strcmp(arg, "-generate") || !strcmp(arg, "--generate")) { | |
239 | generate_packet = true; | |
6f068379 BP |
240 | } else if (!strcmp(arg, "--l7")) { |
241 | if (i + 1 >= argc) { | |
d40533fc | 242 | error = xasprintf("Missing argument for option %s", arg); |
6f068379 BP |
243 | goto exit; |
244 | } | |
245 | ||
246 | struct dp_packet payload; | |
247 | memset(&payload, 0, sizeof payload); | |
248 | dp_packet_init(&payload, 0); | |
249 | if (dp_packet_put_hex(&payload, argv[++i], NULL)[0] != '\0') { | |
250 | dp_packet_uninit(&payload); | |
d40533fc | 251 | error = xstrdup("Trailing garbage in packet data"); |
6f068379 BP |
252 | goto exit; |
253 | } | |
254 | free(l7); | |
255 | l7_len = dp_packet_size(&payload); | |
256 | l7 = dp_packet_steal_data(&payload); | |
257 | } else if (!strcmp(arg, "--l7-len")) { | |
258 | if (i + 1 >= argc) { | |
d40533fc | 259 | error = xasprintf("Missing argument for option %s", arg); |
6f068379 BP |
260 | goto exit; |
261 | } | |
262 | free(l7); | |
263 | l7 = NULL; | |
264 | l7_len = atoi(argv[++i]); | |
265 | if (l7_len > 64000) { | |
d40533fc | 266 | error = xasprintf("%s: too much L7 data", argv[i]); |
6f068379 BP |
267 | goto exit; |
268 | } | |
b490b189 BP |
269 | } else if (consistent |
270 | && (!strcmp(arg, "-consistent") || | |
271 | !strcmp(arg, "--consistent"))) { | |
272 | *consistent = true; | |
273 | } else if (!strcmp(arg, "--ct-next")) { | |
274 | if (i + 1 >= argc) { | |
d40533fc | 275 | error = xasprintf("Missing argument for option %s", arg); |
b490b189 BP |
276 | goto exit; |
277 | } | |
d13ee228 | 278 | |
b490b189 BP |
279 | uint32_t ct_state; |
280 | struct ds ds = DS_EMPTY_INITIALIZER; | |
281 | if (!parse_ct_state(argv[++i], 0, &ct_state, &ds) | |
282 | || !validate_ct_state(ct_state, &ds)) { | |
d40533fc | 283 | error = ds_steal_cstr(&ds); |
b490b189 BP |
284 | goto exit; |
285 | } | |
286 | oftrace_push_ct_state(next_ct_states, ct_state); | |
287 | } else if (arg[0] == '-') { | |
d40533fc | 288 | error = xasprintf("%s: unknown option", arg); |
b490b189 BP |
289 | goto exit; |
290 | } else if (n_args >= ARRAY_SIZE(args)) { | |
d40533fc | 291 | error = xstrdup("too many arguments"); |
0f2f05bb | 292 | goto exit; |
b490b189 BP |
293 | } else { |
294 | args[n_args++] = arg; | |
0f2f05bb | 295 | } |
b490b189 | 296 | } |
0f2f05bb | 297 | |
b490b189 BP |
298 | /* 'args' must now have one of the following forms: |
299 | * | |
300 | * odp_flow | |
301 | * dpname odp_flow | |
302 | * bridge br_flow | |
303 | * odp_flow packet | |
304 | * dpname odp_flow packet | |
305 | * bridge br_flow packet | |
306 | * | |
307 | * Parse the packet if it's there. Note that: | |
308 | * | |
309 | * - If there is one argument, there cannot be a packet. | |
310 | * | |
311 | * - If there are three arguments, there must be a packet. | |
312 | * | |
313 | * If there is a packet, we strip it off. | |
314 | */ | |
315 | if (!generate_packet && n_args > 1) { | |
d40533fc BP |
316 | const char *const_error = eth_from_hex(args[n_args - 1], &packet); |
317 | if (!const_error) { | |
b490b189 BP |
318 | n_args--; |
319 | } else if (n_args > 2) { | |
320 | /* The 3-argument form must end in a hex string. */ | |
d40533fc | 321 | error = xstrdup(const_error); |
0f2f05bb YHW |
322 | goto exit; |
323 | } | |
0f2f05bb YHW |
324 | } |
325 | ||
b490b189 BP |
326 | /* We stripped off the packet if there was one, so 'args' now has one of |
327 | * the following forms: | |
328 | * | |
329 | * odp_flow | |
330 | * dpname odp_flow | |
331 | * bridge br_flow | |
332 | * | |
333 | * Before we parse the flow, try to identify the backer, then use that | |
334 | * backer to assemble a collection of port names. The port names are | |
335 | * useful so that the user can specify ports by name instead of number in | |
336 | * the flow. */ | |
337 | if (n_args == 2) { | |
338 | /* args[0] might be dpname. */ | |
d13ee228 | 339 | const char *dp_type; |
b490b189 BP |
340 | if (!strncmp(args[0], "ovs-", 4)) { |
341 | dp_type = args[0] + 4; | |
d13ee228 | 342 | } else { |
b490b189 | 343 | dp_type = args[0]; |
d13ee228 BP |
344 | } |
345 | backer = shash_find_data(&all_dpif_backers, dp_type); | |
b490b189 BP |
346 | } else if (n_args == 1) { |
347 | /* Pick default backer. */ | |
d13ee228 BP |
348 | struct shash_node *node; |
349 | if (shash_count(&all_dpif_backers) == 1) { | |
350 | node = shash_first(&all_dpif_backers); | |
351 | backer = node->data; | |
352 | } | |
353 | } else { | |
d40533fc | 354 | error = xstrdup("Syntax error"); |
d13ee228 BP |
355 | goto exit; |
356 | } | |
357 | if (backer && backer->dpif) { | |
358 | struct dpif_port dpif_port; | |
359 | struct dpif_port_dump port_dump; | |
360 | DPIF_PORT_FOR_EACH (&dpif_port, &port_dump, backer->dpif) { | |
361 | simap_put(&port_names, dpif_port.name, | |
362 | odp_to_u32(dpif_port.port_no)); | |
363 | } | |
364 | } | |
365 | ||
366 | /* Parse the flow and determine whether a datapath or | |
367 | * bridge is specified. If function odp_flow_key_from_string() | |
368 | * returns 0, the flow is a odp_flow. If function | |
369 | * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */ | |
b490b189 | 370 | if (!odp_flow_from_string(args[n_args - 1], &port_names, |
d40533fc | 371 | &odp_key, &odp_mask, &error)) { |
d13ee228 | 372 | if (!backer) { |
d40533fc | 373 | error = xstrdup("Cannot find the datapath"); |
d13ee228 BP |
374 | goto exit; |
375 | } | |
376 | ||
d40533fc BP |
377 | if (odp_flow_key_to_flow(odp_key.data, odp_key.size, flow, &error) |
378 | == ODP_FIT_ERROR) { | |
d13ee228 BP |
379 | goto exit; |
380 | } | |
381 | ||
382 | *ofprotop = xlate_lookup_ofproto(backer, flow, | |
d40533fc | 383 | &flow->in_port.ofp_port, &error); |
d13ee228 | 384 | if (*ofprotop == NULL) { |
d13ee228 BP |
385 | goto exit; |
386 | } | |
387 | ||
388 | flow->tunnel.metadata.tab = ofproto_get_tun_tab(&(*ofprotop)->up); | |
389 | ||
390 | /* Convert Geneve options to OpenFlow format now. This isn't actually | |
391 | * required in order to get the right results since the ofproto xlate | |
392 | * actions will handle this for us. However, converting now ensures | |
393 | * that our formatting code will always be able to consistently print | |
394 | * in OpenFlow format, which is what we use here. */ | |
395 | if (flow->tunnel.flags & FLOW_TNL_F_UDPIF) { | |
396 | struct flow_tnl tnl; | |
d13ee228 | 397 | memcpy(&tnl, &flow->tunnel, sizeof tnl); |
d40533fc BP |
398 | int err = tun_metadata_from_geneve_udpif( |
399 | flow->tunnel.metadata.tab, &tnl, &tnl, &flow->tunnel); | |
d13ee228 | 400 | if (err) { |
d40533fc | 401 | error = xstrdup("Failed to parse Geneve options"); |
d13ee228 BP |
402 | goto exit; |
403 | } | |
404 | } | |
d40533fc BP |
405 | } else if (n_args != 2) { |
406 | char *s = error; | |
407 | error = xasprintf("%s (or the bridge name was omitted)", s); | |
408 | free(s); | |
409 | goto exit; | |
d13ee228 | 410 | } else { |
d40533fc BP |
411 | free(error); |
412 | error = NULL; | |
d13ee228 | 413 | |
b490b189 | 414 | *ofprotop = ofproto_dpif_lookup_by_name(args[0]); |
d13ee228 | 415 | if (!*ofprotop) { |
d40533fc | 416 | error = xasprintf("%s: unknown bridge", args[0]); |
d13ee228 BP |
417 | goto exit; |
418 | } | |
419 | ||
50f96b10 BP |
420 | struct ofputil_port_map map = OFPUTIL_PORT_MAP_INITIALIZER(&map); |
421 | const struct ofport *ofport; | |
422 | HMAP_FOR_EACH (ofport, hmap_node, &(*ofprotop)->up.ports) { | |
423 | ofputil_port_map_put(&map, ofport->ofp_port, | |
424 | netdev_get_name(ofport->netdev)); | |
425 | } | |
d40533fc BP |
426 | char *err = parse_ofp_exact_flow(flow, NULL, |
427 | ofproto_get_tun_tab(&(*ofprotop)->up), | |
428 | args[n_args - 1], &map); | |
50f96b10 | 429 | ofputil_port_map_destroy(&map); |
d13ee228 | 430 | if (err) { |
d40533fc | 431 | error = xasprintf("Bad openflow flow syntax: %s", err); |
d13ee228 BP |
432 | free(err); |
433 | goto exit; | |
434 | } | |
435 | } | |
436 | ||
b490b189 BP |
437 | if (generate_packet) { |
438 | /* Generate a packet, as requested. */ | |
439 | packet = dp_packet_new(0); | |
6f068379 | 440 | flow_compose(packet, flow, l7, l7_len); |
b490b189 BP |
441 | } else if (packet) { |
442 | /* Use the metadata from the flow and the packet argument to | |
443 | * reconstruct the flow. */ | |
444 | pkt_metadata_from_flow(&packet->md, flow); | |
445 | flow_extract(packet, flow); | |
d13ee228 BP |
446 | } |
447 | ||
448 | exit: | |
d40533fc | 449 | if (error) { |
d13ee228 BP |
450 | dp_packet_delete(packet); |
451 | packet = NULL; | |
452 | } | |
453 | *packetp = packet; | |
454 | ofpbuf_uninit(&odp_key); | |
455 | ofpbuf_uninit(&odp_mask); | |
456 | simap_destroy(&port_names); | |
6f068379 | 457 | free(l7); |
d40533fc | 458 | return error; |
d13ee228 BP |
459 | } |
460 | ||
0f2f05bb YHW |
461 | static void |
462 | free_ct_states(struct ovs_list *ct_states) | |
463 | { | |
464 | while (!ovs_list_is_empty(ct_states)) { | |
465 | oftrace_pop_ct_state(ct_states); | |
466 | } | |
467 | } | |
468 | ||
d13ee228 BP |
469 | static void |
470 | ofproto_unixctl_trace(struct unixctl_conn *conn, int argc, const char *argv[], | |
471 | void *aux OVS_UNUSED) | |
472 | { | |
473 | struct ofproto_dpif *ofproto; | |
474 | struct dp_packet *packet; | |
475 | char *error; | |
476 | struct flow flow; | |
0f2f05bb | 477 | struct ovs_list next_ct_states = OVS_LIST_INITIALIZER(&next_ct_states); |
d13ee228 | 478 | |
0f2f05bb | 479 | error = parse_flow_and_packet(argc, argv, &ofproto, &flow, &packet, |
b490b189 | 480 | &next_ct_states, NULL); |
d13ee228 BP |
481 | if (!error) { |
482 | struct ds result; | |
483 | ||
484 | ds_init(&result); | |
0f2f05bb YHW |
485 | ofproto_trace(ofproto, &flow, packet, NULL, 0, &next_ct_states, |
486 | &result); | |
d13ee228 BP |
487 | unixctl_command_reply(conn, ds_cstr(&result)); |
488 | ds_destroy(&result); | |
489 | dp_packet_delete(packet); | |
490 | } else { | |
491 | unixctl_command_reply_error(conn, error); | |
492 | free(error); | |
493 | } | |
0f2f05bb | 494 | free_ct_states(&next_ct_states); |
d13ee228 BP |
495 | } |
496 | ||
497 | static void | |
498 | ofproto_unixctl_trace_actions(struct unixctl_conn *conn, int argc, | |
499 | const char *argv[], void *aux OVS_UNUSED) | |
500 | { | |
501 | enum ofputil_protocol usable_protocols; | |
502 | struct ofproto_dpif *ofproto; | |
503 | bool enforce_consistency; | |
504 | struct ofpbuf ofpacts; | |
505 | struct dp_packet *packet; | |
506 | struct ds result; | |
67210a55 | 507 | struct match match; |
d13ee228 | 508 | uint16_t in_port; |
0f2f05bb | 509 | struct ovs_list next_ct_states = OVS_LIST_INITIALIZER(&next_ct_states); |
d13ee228 BP |
510 | |
511 | /* Three kinds of error return values! */ | |
512 | enum ofperr retval; | |
513 | char *error; | |
514 | ||
515 | packet = NULL; | |
516 | ds_init(&result); | |
517 | ofpbuf_init(&ofpacts, 0); | |
518 | ||
519 | /* Parse actions. */ | |
efefbcae BP |
520 | struct ofpact_parse_params pp = { |
521 | .port_map = NULL, | |
522 | .ofpacts = &ofpacts, | |
523 | .usable_protocols = &usable_protocols, | |
524 | }; | |
525 | error = ofpacts_parse_actions(argv[--argc], &pp); | |
d13ee228 BP |
526 | if (error) { |
527 | unixctl_command_reply_error(conn, error); | |
528 | free(error); | |
529 | goto exit; | |
530 | } | |
531 | ||
0f2f05bb | 532 | error = parse_flow_and_packet(argc, argv, &ofproto, &match.flow, &packet, |
b490b189 | 533 | &next_ct_states, &enforce_consistency); |
d13ee228 BP |
534 | if (error) { |
535 | unixctl_command_reply_error(conn, error); | |
536 | free(error); | |
537 | goto exit; | |
538 | } | |
67210a55 | 539 | match_wc_init(&match, &match.flow); |
d13ee228 BP |
540 | |
541 | /* Do the same checks as handle_packet_out() in ofproto.c. | |
542 | * | |
543 | * We pass a 'table_id' of 0 to ofpacts_check(), which isn't | |
544 | * strictly correct because these actions aren't in any table, but it's OK | |
545 | * because it 'table_id' is used only to check goto_table instructions, but | |
546 | * packet-outs take a list of actions and therefore it can't include | |
547 | * instructions. | |
548 | * | |
549 | * We skip the "meter" check here because meter is an instruction, not an | |
550 | * action, and thus cannot appear in ofpacts. */ | |
67210a55 | 551 | in_port = ofp_to_u16(match.flow.in_port.ofp_port); |
d13ee228 BP |
552 | if (in_port >= ofproto->up.max_ports && in_port < ofp_to_u16(OFPP_MAX)) { |
553 | unixctl_command_reply_error(conn, "invalid in_port"); | |
554 | goto exit; | |
555 | } | |
ae6f7530 BP |
556 | |
557 | struct ofpact_check_params cp = { | |
558 | .match = &match, | |
559 | .max_ports = u16_to_ofp(ofproto->up.max_ports), | |
560 | .table_id = 0, | |
561 | .n_tables = ofproto->up.n_tables, | |
562 | }; | |
563 | retval = ofpacts_check_consistency( | |
564 | ofpacts.data, ofpacts.size, | |
565 | enforce_consistency ? usable_protocols : 0, &cp); | |
d13ee228 | 566 | if (!retval) { |
d61973d6 | 567 | ovs_mutex_lock(&ofproto_mutex); |
d13ee228 BP |
568 | retval = ofproto_check_ofpacts(&ofproto->up, ofpacts.data, |
569 | ofpacts.size); | |
d61973d6 | 570 | ovs_mutex_unlock(&ofproto_mutex); |
d13ee228 BP |
571 | } |
572 | ||
573 | if (retval) { | |
574 | ds_clear(&result); | |
575 | ds_put_format(&result, "Bad actions: %s", ofperr_to_string(retval)); | |
576 | unixctl_command_reply_error(conn, ds_cstr(&result)); | |
577 | goto exit; | |
578 | } | |
579 | ||
67210a55 | 580 | ofproto_trace(ofproto, &match.flow, packet, |
0f2f05bb | 581 | ofpacts.data, ofpacts.size, &next_ct_states, &result); |
d13ee228 BP |
582 | unixctl_command_reply(conn, ds_cstr(&result)); |
583 | ||
584 | exit: | |
585 | ds_destroy(&result); | |
586 | dp_packet_delete(packet); | |
587 | ofpbuf_uninit(&ofpacts); | |
0f2f05bb | 588 | free_ct_states(&next_ct_states); |
d13ee228 BP |
589 | } |
590 | ||
418a7a84 BP |
591 | static void |
592 | explain_slow_path(enum slow_path_reason slow, struct ds *output) | |
593 | { | |
594 | ds_put_cstr(output, "\nThis flow is handled by the userspace " | |
595 | "slow path because it:"); | |
596 | for (; slow; slow = zero_rightmost_1bit(slow)) { | |
597 | enum slow_path_reason bit = rightmost_1bit(slow); | |
598 | ds_put_format(output, "\n - %s.", | |
599 | slow_path_reason_to_explanation(bit)); | |
600 | } | |
601 | } | |
602 | ||
603 | /* Copies ODP actions from 'in' to 'out', dropping OVS_ACTION_ATTR_OUTPUT and | |
604 | * OVS_ACTION_ATTR_RECIRC along the way. */ | |
605 | static void | |
606 | prune_output_actions(const struct ofpbuf *in, struct ofpbuf *out) | |
607 | { | |
608 | const struct nlattr *a; | |
609 | unsigned int left; | |
610 | NL_ATTR_FOR_EACH (a, left, in->data, in->size) { | |
611 | if (a->nla_type == OVS_ACTION_ATTR_CLONE) { | |
612 | struct ofpbuf in_nested; | |
613 | nl_attr_get_nested(a, &in_nested); | |
614 | ||
615 | size_t ofs = nl_msg_start_nested(out, OVS_ACTION_ATTR_CLONE); | |
616 | prune_output_actions(&in_nested, out); | |
617 | nl_msg_end_nested(out, ofs); | |
618 | } else if (a->nla_type != OVS_ACTION_ATTR_OUTPUT && | |
619 | a->nla_type != OVS_ACTION_ATTR_RECIRC) { | |
620 | ofpbuf_put(out, a, NLA_ALIGN(a->nla_len)); | |
621 | } | |
622 | } | |
623 | } | |
624 | ||
625 | /* Executes all of the datapath actions, except for any OVS_ACTION_ATTR_OUTPUT | |
626 | * and OVS_ACTION_ATTR_RECIRC actions, in 'actions' on 'packet', which has the | |
627 | * given 'flow', on 'dpif'. The actions have slow path reason 'slow' (if any). | |
628 | * Appends any error message to 'output'. | |
629 | * | |
630 | * With output and recirculation actions dropped, the only remaining side | |
631 | * effects are from OVS_ACTION_ATTR_USERSPACE actions for executing actions to | |
632 | * send a packet to an OpenFlow controller, IPFIX, NetFlow, and sFlow, etc. */ | |
633 | static void | |
634 | execute_actions_except_outputs(struct dpif *dpif, | |
635 | const struct dp_packet *packet, | |
636 | const struct flow *flow, | |
637 | const struct ofpbuf *actions, | |
638 | enum slow_path_reason slow, | |
639 | struct ds *output) | |
640 | { | |
641 | struct ofpbuf pruned_actions; | |
642 | ofpbuf_init(&pruned_actions, 0); | |
643 | prune_output_actions(actions, &pruned_actions); | |
644 | ||
645 | struct dpif_execute execute = { | |
646 | .actions = pruned_actions.data, | |
647 | .actions_len = pruned_actions.size, | |
648 | .needs_help = (slow & SLOW_ACTION) != 0, | |
649 | .flow = flow, | |
650 | .packet = dp_packet_clone_with_headroom(packet, 2), | |
651 | }; | |
652 | int error = dpif_execute(dpif, &execute); | |
653 | if (error) { | |
654 | ds_put_format(output, "\nAction execution failed (%s)\n.", | |
655 | ovs_strerror(error)); | |
656 | } | |
657 | dp_packet_delete(execute.packet); | |
658 | ofpbuf_uninit(&pruned_actions); | |
659 | } | |
660 | ||
d072d2de DC |
661 | static void |
662 | ofproto_trace_recirc_node(struct oftrace_recirc_node *node, | |
663 | struct ovs_list *next_ct_states, | |
664 | struct ds *output) | |
665 | { | |
666 | ds_put_cstr(output, "\n\n"); | |
667 | ds_put_char_multiple(output, '=', 79); | |
668 | ds_put_format(output, "\nrecirc(%#"PRIx32")", node->recirc_id); | |
669 | ||
670 | if (next_ct_states && node->type == OFT_RECIRC_CONNTRACK) { | |
671 | uint32_t ct_state; | |
672 | if (ovs_list_is_empty(next_ct_states)) { | |
673 | ct_state = CS_TRACKED | CS_NEW; | |
674 | ds_put_cstr(output, " - resume conntrack with default " | |
675 | "ct_state=trk|new (use --ct-next to customize)"); | |
676 | } else { | |
677 | ct_state = oftrace_pop_ct_state(next_ct_states); | |
678 | struct ds s = DS_EMPTY_INITIALIZER; | |
679 | format_flags(&s, ct_state_to_string, ct_state, '|'); | |
680 | ds_put_format(output, " - resume conntrack with ct_state=%s", | |
681 | ds_cstr(&s)); | |
682 | ds_destroy(&s); | |
683 | } | |
684 | node->flow.ct_state = ct_state; | |
685 | } | |
686 | ds_put_char(output, '\n'); | |
687 | ||
688 | /* If there's any snat/dnat information assume we always translate to | |
689 | * the first IP/port to make sure we don't match on incorrect flows later | |
690 | * on. | |
691 | */ | |
692 | if (node->nat_act) { | |
693 | const struct ofpact_nat *ofn = node->nat_act; | |
694 | ||
695 | ds_put_cstr(output, "Replacing src/dst IP/ports to simulate NAT:\n"); | |
696 | ds_put_cstr(output, " Initial flow: "); | |
697 | oftrace_print_ip_flow(&node->flow, ofn->range_af, output); | |
698 | ||
699 | if (ofn->flags & NX_NAT_F_SRC) { | |
700 | if (ofn->range_af == AF_INET) { | |
701 | node->flow.nw_src = ofn->range.addr.ipv4.min; | |
702 | } else if (ofn->range_af == AF_INET6) { | |
703 | node->flow.ipv6_src = ofn->range.addr.ipv6.min; | |
704 | } | |
705 | ||
706 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
707 | node->flow.tp_src = htons(ofn->range.proto.min); | |
708 | } | |
709 | } | |
710 | if (ofn->flags & NX_NAT_F_DST) { | |
711 | if (ofn->range_af == AF_INET) { | |
712 | node->flow.nw_dst = ofn->range.addr.ipv4.min; | |
713 | } else if (ofn->range_af == AF_INET6) { | |
714 | node->flow.ipv6_dst = ofn->range.addr.ipv6.min; | |
715 | } | |
716 | ||
717 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
718 | node->flow.tp_dst = htons(ofn->range.proto.min); | |
719 | } | |
720 | } | |
721 | ds_put_cstr(output, " Modified flow: "); | |
722 | oftrace_print_ip_flow(&node->flow, ofn->range_af, output); | |
723 | } | |
724 | ds_put_char_multiple(output, '=', 79); | |
725 | ds_put_cstr(output, "\n\n"); | |
726 | } | |
727 | ||
d13ee228 | 728 | static void |
e6bc8e74 YHW |
729 | ofproto_trace__(struct ofproto_dpif *ofproto, const struct flow *flow, |
730 | const struct dp_packet *packet, struct ovs_list *recirc_queue, | |
731 | const struct ofpact ofpacts[], size_t ofpacts_len, | |
732 | struct ds *output) | |
d13ee228 | 733 | { |
2d9b49dd BP |
734 | struct ofpbuf odp_actions; |
735 | ofpbuf_init(&odp_actions, 0); | |
d13ee228 | 736 | |
2d9b49dd BP |
737 | struct xlate_in xin; |
738 | struct flow_wildcards wc; | |
739 | struct ovs_list trace = OVS_LIST_INITIALIZER(&trace); | |
740 | xlate_in_init(&xin, ofproto, | |
d13ee228 BP |
741 | ofproto_dpif_get_tables_version(ofproto), flow, |
742 | flow->in_port.ofp_port, NULL, ntohs(flow->tcp_flags), | |
2d9b49dd BP |
743 | packet, &wc, &odp_actions); |
744 | xin.ofpacts = ofpacts; | |
745 | xin.ofpacts_len = ofpacts_len; | |
746 | xin.trace = &trace; | |
e6bc8e74 | 747 | xin.recirc_queue = recirc_queue; |
2d9b49dd BP |
748 | |
749 | /* Copy initial flow out of xin.flow. It differs from '*flow' because | |
750 | * xlate_in_init() initializes actset_output to OFPP_UNSET. */ | |
751 | struct flow initial_flow = xin.flow; | |
752 | ds_put_cstr(output, "Flow: "); | |
50f96b10 | 753 | flow_format(output, &initial_flow, NULL); |
2d9b49dd BP |
754 | ds_put_char(output, '\n'); |
755 | ||
756 | struct xlate_out xout; | |
757 | enum xlate_error error = xlate_actions(&xin, &xout); | |
758 | ||
759 | oftrace_node_print_details(output, &trace, 0); | |
760 | ||
761 | ds_put_cstr(output, "\nFinal flow: "); | |
762 | if (flow_equal(&initial_flow, &xin.flow)) { | |
763 | ds_put_cstr(output, "unchanged"); | |
764 | } else { | |
50f96b10 | 765 | flow_format(output, &xin.flow, NULL); |
2d9b49dd BP |
766 | } |
767 | ds_put_char(output, '\n'); | |
d13ee228 | 768 | |
2d9b49dd BP |
769 | ds_put_cstr(output, "Megaflow: "); |
770 | struct match match; | |
771 | match_init(&match, flow, &wc); | |
50f96b10 | 772 | match_format(&match, NULL, output, OFP_DEFAULT_PRIORITY); |
2d9b49dd | 773 | ds_put_char(output, '\n'); |
d13ee228 | 774 | |
2d9b49dd | 775 | ds_put_cstr(output, "Datapath actions: "); |
0722f341 | 776 | format_odp_actions(output, odp_actions.data, odp_actions.size, NULL); |
d13ee228 BP |
777 | |
778 | if (error != XLATE_OK) { | |
2d9b49dd BP |
779 | ds_put_format(output, |
780 | "\nTranslation failed (%s), packet is dropped.\n", | |
d13ee228 | 781 | xlate_strerror(error)); |
418a7a84 BP |
782 | } else { |
783 | if (xout.slow) { | |
784 | explain_slow_path(xout.slow, output); | |
785 | } | |
786 | if (packet) { | |
787 | execute_actions_except_outputs(ofproto->backer->dpif, packet, | |
788 | &initial_flow, &odp_actions, | |
789 | xout.slow, output); | |
d13ee228 BP |
790 | } |
791 | } | |
792 | ||
418a7a84 | 793 | |
2d9b49dd BP |
794 | xlate_out_uninit(&xout); |
795 | ofpbuf_uninit(&odp_actions); | |
796 | oftrace_node_list_destroy(&trace); | |
d13ee228 BP |
797 | } |
798 | ||
e6bc8e74 YHW |
799 | /* Implements a "trace" through 'ofproto''s flow table, appending a textual |
800 | * description of the results to 'output'. | |
801 | * | |
802 | * The trace follows a packet with the specified 'flow' through the flow | |
803 | * table. 'packet' may be nonnull to trace an actual packet, with consequent | |
804 | * side effects (if it is nonnull then its flow must be 'flow'). | |
805 | * | |
806 | * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to | |
807 | * trace, otherwise the actions are determined by a flow table lookup. */ | |
d1ea2cc3 | 808 | void |
e6bc8e74 YHW |
809 | ofproto_trace(struct ofproto_dpif *ofproto, const struct flow *flow, |
810 | const struct dp_packet *packet, | |
811 | const struct ofpact ofpacts[], size_t ofpacts_len, | |
0f2f05bb | 812 | struct ovs_list *next_ct_states, struct ds *output) |
e6bc8e74 YHW |
813 | { |
814 | struct ovs_list recirc_queue = OVS_LIST_INITIALIZER(&recirc_queue); | |
815 | ofproto_trace__(ofproto, flow, packet, &recirc_queue, | |
816 | ofpacts, ofpacts_len, output); | |
817 | ||
818 | struct oftrace_recirc_node *recirc_node; | |
819 | LIST_FOR_EACH_POP (recirc_node, node, &recirc_queue) { | |
d072d2de | 820 | ofproto_trace_recirc_node(recirc_node, next_ct_states, output); |
e6bc8e74 YHW |
821 | ofproto_trace__(ofproto, &recirc_node->flow, recirc_node->packet, |
822 | &recirc_queue, ofpacts, ofpacts_len, output); | |
823 | oftrace_recirc_node_destroy(recirc_node); | |
824 | } | |
825 | } | |
826 | ||
d13ee228 BP |
827 | void |
828 | ofproto_dpif_trace_init(void) | |
829 | { | |
830 | static bool registered; | |
831 | if (registered) { | |
832 | return; | |
833 | } | |
834 | registered = true; | |
835 | ||
836 | unixctl_command_register( | |
837 | "ofproto/trace", | |
0f2f05bb YHW |
838 | "{[dp_name] odp_flow | bridge br_flow} [OPTIONS...] " |
839 | "[-generate|packet]", 1, INT_MAX, ofproto_unixctl_trace, NULL); | |
d13ee228 BP |
840 | unixctl_command_register( |
841 | "ofproto/trace-packet-out", | |
0f2f05bb YHW |
842 | "[-consistent] {[dp_name] odp_flow | bridge br_flow} [OPTIONS...] " |
843 | "[-generate|packet] actions", | |
844 | 2, INT_MAX, ofproto_unixctl_trace_actions, NULL); | |
d13ee228 | 845 | } |