1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2015 Intel Corporation.
4 Packet Classification and Access Control
5 ========================================
7 The DPDK provides an Access Control library that gives the ability
8 to classify an input packet based on a set of classification rules.
10 The ACL library is used to perform an N-tuple search over a set of rules with multiple categories
11 and find the best match (highest priority) for each category.
12 The library API provides the following basic operations:
14 * Create a new Access Control (AC) context.
16 * Add rules into the context.
18 * For all rules in the context, build the runtime structures necessary to perform packet classification.
20 * Perform input packet classifications.
22 * Destroy an AC context and its runtime structures and free the associated memory.
30 The current implementation allows the user for each AC context to specify its own rule (set of fields)
31 over which packet classification will be performed.
32 Though there are few restrictions on the rule fields layout:
34 * First field in the rule definition has to be one byte long.
35 * All subsequent fields has to be grouped into sets of 4 consecutive bytes.
37 This is done mainly for performance reasons - search function processes the first input byte as part of the flow setup and then the inner loop of the search function is unrolled to process four input bytes at a time.
39 To define each field inside an AC rule, the following structure is used:
43 struct rte_acl_field_def {
44 uint8_t type; /*< type - ACL_FIELD_TYPE. */
45 uint8_t size; /*< size of field 1,2,4, or 8. */
46 uint8_t field_index; /*< index of field inside the rule. */
47 uint8_t input_index; /*< 0-N input index. */
48 uint32_t offset; /*< offset to start of field. */
52 The field type is one of three choices:
54 * _MASK - for fields such as IP addresses that have a value and a mask defining the number of relevant bits.
56 * _RANGE - for fields such as ports that have a lower and upper value for the field.
58 * _BITMASK - for fields such as protocol identifiers that have a value and a bit mask.
61 The size parameter defines the length of the field in bytes. Allowable values are 1, 2, 4, or 8 bytes.
62 Note that due to the grouping of input bytes, 1 or 2 byte fields must be defined as consecutive fields
63 that make up 4 consecutive input bytes.
64 Also, it is best to define fields of 8 or more bytes as 4 byte fields so that
65 the build processes can eliminate fields that are all wild.
68 A zero-based value that represents the position of the field inside the rule; 0 to N-1 for N fields.
71 As mentioned above, all input fields, except the very first one, must be in groups of 4 consecutive bytes.
72 The input index specifies to which input group that field belongs to.
75 The offset field defines the offset for the field.
76 This is the offset from the beginning of the buffer parameter for the search.
78 For example, to define classification for the following IPv4 5-tuple structure:
90 The following array of field definitions can be used:
94 struct rte_acl_field_def ipv4_defs[5] = {
95 /* first input field - always one byte long. */
97 .type = RTE_ACL_FIELD_TYPE_BITMASK,
98 .size = sizeof (uint8_t),
101 .offset = offsetof (struct ipv4_5tuple, proto),
104 /* next input field (IPv4 source address) - 4 consecutive bytes. */
106 .type = RTE_ACL_FIELD_TYPE_MASK,
107 .size = sizeof (uint32_t),
110 .offset = offsetof (struct ipv4_5tuple, ip_src),
113 /* next input field (IPv4 destination address) - 4 consecutive bytes. */
115 .type = RTE_ACL_FIELD_TYPE_MASK,
116 .size = sizeof (uint32_t),
119 .offset = offsetof (struct ipv4_5tuple, ip_dst),
123 * Next 2 fields (src & dst ports) form 4 consecutive bytes.
124 * They share the same input index.
127 .type = RTE_ACL_FIELD_TYPE_RANGE,
128 .size = sizeof (uint16_t),
131 .offset = offsetof (struct ipv4_5tuple, port_src),
135 .type = RTE_ACL_FIELD_TYPE_RANGE,
136 .size = sizeof (uint16_t),
139 .offset = offsetof (struct ipv4_5tuple, port_dst),
143 A typical example of such an IPv4 5-tuple rule is a follows:
147 source addr/mask destination addr/mask source ports dest ports protocol/mask
148 192.168.1.0/24 192.168.2.31/32 0:65535 1234:1234 17/0xff
150 Any IPv4 packets with protocol ID 17 (UDP), source address 192.168.1.[0-255], destination address 192.168.2.31,
151 source port [0-65535] and destination port 1234 matches the above rule.
153 To define classification for the IPv6 2-tuple: <protocol, IPv6 source address> over the following IPv6 header structure:
157 struct rte_ipv6_hdr {
158 uint32_t vtc_flow; /* IP version, traffic class & flow label. */
159 uint16_t payload_len; /* IP packet length - includes sizeof(ip_header). */
160 uint8_t proto; /* Protocol, next header. */
161 uint8_t hop_limits; /* Hop limits. */
162 uint8_t src_addr[16]; /* IP address of source host. */
163 uint8_t dst_addr[16]; /* IP address of destination host(s). */
166 The following array of field definitions can be used:
170 struct rte_acl_field_def ipv6_2tuple_defs[5] = {
172 .type = RTE_ACL_FIELD_TYPE_BITMASK,
173 .size = sizeof (uint8_t),
176 .offset = offsetof (struct rte_ipv6_hdr, proto),
180 .type = RTE_ACL_FIELD_TYPE_MASK,
181 .size = sizeof (uint32_t),
184 .offset = offsetof (struct rte_ipv6_hdr, src_addr[0]),
188 .type = RTE_ACL_FIELD_TYPE_MASK,
189 .size = sizeof (uint32_t),
192 .offset = offsetof (struct rte_ipv6_hdr, src_addr[4]),
196 .type = RTE_ACL_FIELD_TYPE_MASK,
197 .size = sizeof (uint32_t),
200 .offset = offsetof (struct rte_ipv6_hdr, src_addr[8]),
204 .type = RTE_ACL_FIELD_TYPE_MASK,
205 .size = sizeof (uint32_t),
208 .offset = offsetof (struct rte_ipv6_hdr, src_addr[12]),
212 A typical example of such an IPv6 2-tuple rule is a follows:
216 source addr/mask protocol/mask
217 2001:db8:1234:0000:0000:0000:0000:0000/48 6/0xff
219 Any IPv6 packets with protocol ID 6 (TCP), and source address inside the range
220 [2001:db8:1234:0000:0000:0000:0000:0000 - 2001:db8:1234:ffff:ffff:ffff:ffff:ffff] matches the above rule.
222 In the following example the last element of the search key is 8-bit long.
223 So it is a case where the 4 consecutive bytes of an input field are not fully occupied.
224 The structure for the classification is:
232 uint8_t tos; /*< This is partially using a 32-bit input element */
235 The following array of field definitions can be used:
239 struct rte_acl_field_def ipv4_defs[4] = {
240 /* first input field - always one byte long. */
242 .type = RTE_ACL_FIELD_TYPE_BITMASK,
243 .size = sizeof (uint8_t),
246 .offset = offsetof (struct acl_key, ip_proto),
249 /* next input field (IPv4 source address) - 4 consecutive bytes. */
251 .type = RTE_ACL_FIELD_TYPE_MASK,
252 .size = sizeof (uint32_t),
255 .offset = offsetof (struct acl_key, ip_src),
258 /* next input field (IPv4 destination address) - 4 consecutive bytes. */
260 .type = RTE_ACL_FIELD_TYPE_MASK,
261 .size = sizeof (uint32_t),
264 .offset = offsetof (struct acl_key, ip_dst),
268 * Next element of search key (Type of Service) is indeed 1 byte long.
269 * Anyway we need to allocate all the 4 consecutive bytes for it.
272 .type = RTE_ACL_FIELD_TYPE_BITMASK,
273 .size = sizeof (uint32_t), /* All the 4 consecutive bytes are allocated */
276 .offset = offsetof (struct acl_key, tos),
280 A typical example of such an IPv4 4-tuple rule is as follows:
284 source addr/mask destination addr/mask tos/mask protocol/mask
285 192.168.1.0/24 192.168.2.31/32 1/0xff 6/0xff
287 Any IPv4 packets with protocol ID 6 (TCP), source address 192.168.1.[0-255], destination address 192.168.2.31,
288 ToS 1 matches the above rule.
290 When creating a set of rules, for each rule, additional information must be supplied also:
292 * **priority**: A weight to measure the priority of the rules (higher is better).
293 If the input tuple matches more than one rule, then the rule with the higher priority is returned.
294 Note that if the input tuple matches more than one rule and these rules have equal priority,
295 it is undefined which rule is returned as a match.
296 It is recommended to assign a unique priority for each rule.
298 * **category_mask**: Each rule uses a bit mask value to select the relevant category(s) for the rule.
299 When a lookup is performed, the result for each category is returned.
300 This effectively provides a "parallel lookup" by enabling a single search to return multiple results if,
301 for example, there were four different sets of ACL rules, one for access control, one for routing, and so on.
302 Each set could be assigned its own category and by combining them into a single database,
303 one lookup returns a result for each of the four sets.
305 * **userdata**: A user-defined value.
306 For each category, a successful match returns the userdata field of the highest priority matched rule.
307 When no rules match, returned value is zero.
311 When adding new rules into an ACL context, all fields must be in host byte order (LSB).
312 When the search is performed for an input tuple, all fields in that tuple must be in network byte order (MSB).
317 Build phase (rte_acl_build()) creates for a given set of rules internal structure for further run-time traversal.
318 With current implementation it is a set of multi-bit tries (with stride == 8).
319 Depending on the rules set, that could consume significant amount of memory.
320 In attempt to conserve some space ACL build process tries to split the given
321 rule-set into several non-intersecting subsets and construct a separate trie
323 Depending on the rule-set, it might reduce RT memory requirements but might
324 increase classification time.
325 There is a possibility at build-time to specify maximum memory limit for internal RT structures for given AC context.
326 It could be done via **max_size** field of the **rte_acl_config** structure.
327 Setting it to the value greater than zero, instructs rte_acl_build() to:
329 * attempt to minimize number of tries in the RT table, but
330 * make sure that size of RT table wouldn't exceed given value.
332 Setting it to zero makes rte_acl_build() to use the default behavior:
333 try to minimize size of the RT structures, but doesn't expose any hard limit on it.
335 That gives the user the ability to decisions about performance/space trade-off.
340 struct rte_acl_ctx * acx;
341 struct rte_acl_config cfg;
345 * assuming that acx points to already created and
346 * populated with rules AC context and cfg filled properly.
349 /* try to build AC context, with RT structures less then 8MB. */
350 cfg.max_size = 0x800000;
351 ret = rte_acl_build(acx, &cfg);
354 * RT structures can't fit into 8MB for given context.
355 * Try to build without exposing any hard limit.
357 if (ret == -ERANGE) {
359 ret = rte_acl_build(acx, &cfg);
364 Classification methods
365 ~~~~~~~~~~~~~~~~~~~~~~
367 After rte_acl_build() over given AC context has finished successfully, it can be used to perform classification - search for a rule with highest priority over the input data.
368 There are several implementations of classify algorithm:
370 * **RTE_ACL_CLASSIFY_SCALAR**: generic implementation, doesn't require any specific HW support.
372 * **RTE_ACL_CLASSIFY_SSE**: vector implementation, can process up to 8 flows in parallel. Requires SSE 4.1 support.
374 * **RTE_ACL_CLASSIFY_AVX2**: vector implementation, can process up to 16 flows in parallel. Requires AVX2 support.
376 * **RTE_ACL_CLASSIFY_NEON**: vector implementation, can process up to 8 flows
377 in parallel. Requires NEON support.
379 * **RTE_ACL_CLASSIFY_ALTIVEC**: vector implementation, can process up to 8
380 flows in parallel. Requires ALTIVEC support.
382 * **RTE_ACL_CLASSIFY_AVX512X16**: vector implementation, can process up to 16
383 flows in parallel. Uses 256-bit width SIMD registers.
384 Requires AVX512 support.
386 * **RTE_ACL_CLASSIFY_AVX512X32**: vector implementation, can process up to 32
387 flows in parallel. Uses 512-bit width SIMD registers.
388 Requires AVX512 support.
390 It is purely a runtime decision which method to choose, there is no build-time difference.
391 All implementations operates over the same internal RT structures and use similar principles. The main difference is that vector implementations can manually exploit IA SIMD instructions and process several input data flows in parallel.
392 At startup ACL library determines the highest available classify method for the given platform and sets it as default one. Though the user has an ability to override the default classifier function for a given ACL context or perform particular search using non-default classify method. In that case it is user responsibility to make sure that given platform supports selected classify implementation.
396 Right now ``RTE_ACL_CLASSIFY_AVX512X32`` is not selected by default
397 (due to possible frequency level change), but it can be selected at
398 runtime by apps through the use of ACL API: ``rte_acl_set_ctx_classify``.
400 Application Programming Interface (API) Usage
401 ---------------------------------------------
405 For more details about the Access Control API, please refer to the *DPDK API Reference*.
407 The following example demonstrates IPv4, 5-tuple classification for rules defined above
408 with multiple categories in more detail.
410 Classify with Multiple Categories
411 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
415 struct rte_acl_ctx * acx;
416 struct rte_acl_config cfg;
419 /* define a structure for the rule with up to 5 fields. */
421 RTE_ACL_RULE_DEF(acl_ipv4_rule, RTE_DIM(ipv4_defs));
423 /* AC context creation parameters. */
425 struct rte_acl_param prm = {
426 .name = "ACL_example",
427 .socket_id = SOCKET_ID_ANY,
428 .rule_size = RTE_ACL_RULE_SZ(RTE_DIM(ipv4_defs)),
430 /* number of fields per rule. */
432 .max_rule_num = 8, /* maximum number of rules in the AC context. */
435 struct acl_ipv4_rule acl_rules[] = {
437 /* matches all packets traveling to 192.168.0.0/16, applies for categories: 0,1 */
439 .data = {.userdata = 1, .category_mask = 3, .priority = 1},
441 /* destination IPv4 */
442 .field[2] = {.value.u32 = RTE_IPV4(192,168,0,0),. mask_range.u32 = 16,},
445 .field[3] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
447 /* destination port */
448 .field[4] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
451 /* matches all packets traveling to 192.168.1.0/24, applies for categories: 0 */
453 .data = {.userdata = 2, .category_mask = 1, .priority = 2},
455 /* destination IPv4 */
456 .field[2] = {.value.u32 = RTE_IPV4(192,168,1,0),. mask_range.u32 = 24,},
459 .field[3] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
461 /* destination port */
462 .field[4] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
465 /* matches all packets traveling from 10.1.1.1, applies for categories: 1 */
467 .data = {.userdata = 3, .category_mask = 2, .priority = 3},
470 .field[1] = {.value.u32 = RTE_IPV4(10,1,1,1),. mask_range.u32 = 32,},
473 .field[3] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
475 /* destination port */
476 .field[4] = {.value.u16 = 0, .mask_range.u16 = 0xffff,},
482 /* create an empty AC context */
484 if ((acx = rte_acl_create(&prm)) == NULL) {
486 /* handle context create failure. */
490 /* add rules to the context */
492 ret = rte_acl_add_rules(acx, acl_rules, RTE_DIM(acl_rules));
494 /* handle error at adding ACL rules. */
497 /* prepare AC build config. */
499 cfg.num_categories = 2;
500 cfg.num_fields = RTE_DIM(ipv4_defs);
502 memcpy(cfg.defs, ipv4_defs, sizeof (ipv4_defs));
504 /* build the runtime structures for added rules, with 2 categories. */
506 ret = rte_acl_build(acx, &cfg);
508 /* handle error at build runtime structures for ACL context. */
511 For a tuple with source IP address: 10.1.1.1 and destination IP address: 192.168.1.15,
512 once the following lines are executed:
516 uint32_t results[4]; /* make classify for 4 categories. */
518 rte_acl_classify(acx, data, results, 1, 4);
520 then the results[] array contains:
524 results[4] = {2, 3, 0, 0};
526 * For category 0, both rules 1 and 2 match, but rule 2 has higher priority,
527 therefore results[0] contains the userdata for rule 2.
529 * For category 1, both rules 1 and 3 match, but rule 3 has higher priority,
530 therefore results[1] contains the userdata for rule 3.
532 * For categories 2 and 3, there are no matches, so results[2] and results[3] contain zero,
533 which indicates that no matches were found for those categories.
535 For a tuple with source IP address: 192.168.1.1 and destination IP address: 192.168.2.11,
536 once the following lines are executed:
540 uint32_t results[4]; /* make classify by 4 categories. */
542 rte_acl_classify(acx, data, results, 1, 4);
544 the results[] array contains:
548 results[4] = {1, 1, 0, 0};
550 * For categories 0 and 1, only rule 1 matches.
552 * For categories 2 and 3, there are no matches.
554 For a tuple with source IP address: 10.1.1.1 and destination IP address: 201.212.111.12,
555 once the following lines are executed:
559 uint32_t results[4]; /* make classify by 4 categories. */
560 rte_acl_classify(acx, data, results, 1, 4);
562 the results[] array contains:
566 results[4] = {0, 3, 0, 0};
568 * For category 1, only rule 3 matches.
570 * For categories 0, 2 and 3, there are no matches.