1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright 2016 6WIND S.A.
3 * Copyright 2016 Mellanox Technologies, Ltd
11 * RTE generic flow API
13 * This interface provides the ability to program packet matching and
14 * associated actions in hardware through flow rules.
21 #include <rte_ether.h>
22 #include <rte_eth_ctrl.h>
28 #include <rte_byteorder.h>
36 * Flow rule attributes.
38 * Priorities are set on two levels: per group and per rule within groups.
40 * Lower values denote higher priority, the highest priority for both levels
41 * is 0, so that a rule with priority 0 in group 8 is always matched after a
42 * rule with priority 8 in group 0.
44 * Although optional, applications are encouraged to group similar rules as
45 * much as possible to fully take advantage of hardware capabilities
46 * (e.g. optimized matching) and work around limitations (e.g. a single
47 * pattern type possibly allowed in a given group).
49 * Group and priority levels are arbitrary and up to the application, they
50 * do not need to be contiguous nor start from 0, however the maximum number
51 * varies between devices and may be affected by existing flow rules.
53 * If a packet is matched by several rules of a given group for a given
54 * priority level, the outcome is undefined. It can take any path, may be
55 * duplicated or even cause unrecoverable errors.
57 * Note that support for more than a single group and priority level is not
60 * Flow rules can apply to inbound and/or outbound traffic (ingress/egress).
62 * Several pattern items and actions are valid and can be used in both
63 * directions. Those valid for only one direction are described as such.
65 * At least one direction must be specified.
67 * Specifying both directions at once for a given rule is not recommended
68 * but may be valid in a few cases (e.g. shared counter).
70 struct rte_flow_attr {
71 uint32_t group; /**< Priority group. */
72 uint32_t priority; /**< Priority level within group. */
73 uint32_t ingress:1; /**< Rule applies to ingress traffic. */
74 uint32_t egress:1; /**< Rule applies to egress traffic. */
75 uint32_t reserved:30; /**< Reserved, must be zero. */
79 * Matching pattern item types.
81 * Pattern items fall in two categories:
83 * - Matching protocol headers and packet data, usually associated with a
84 * specification structure. These must be stacked in the same order as the
85 * protocol layers to match inside packets, starting from the lowest.
87 * - Matching meta-data or affecting pattern processing, often without a
88 * specification structure. Since they do not match packet contents, their
89 * position in the list is usually not relevant.
91 * See the description of individual types for more information. Those
92 * marked with [META] fall into the second category.
94 enum rte_flow_item_type {
98 * End marker for item lists. Prevents further processing of items,
99 * thereby ending the pattern.
101 * No associated specification structure.
103 RTE_FLOW_ITEM_TYPE_END,
108 * Used as a placeholder for convenience. It is ignored and simply
111 * No associated specification structure.
113 RTE_FLOW_ITEM_TYPE_VOID,
118 * Inverted matching, i.e. process packets that do not match the
121 * No associated specification structure.
123 RTE_FLOW_ITEM_TYPE_INVERT,
126 * Matches any protocol in place of the current layer, a single ANY
127 * may also stand for several protocol layers.
129 * See struct rte_flow_item_any.
131 RTE_FLOW_ITEM_TYPE_ANY,
136 * Matches packets addressed to the physical function of the device.
138 * If the underlying device function differs from the one that would
139 * normally receive the matched traffic, specifying this item
140 * prevents it from reaching that device unless the flow rule
141 * contains a PF action. Packets are not duplicated between device
142 * instances by default.
144 * No associated specification structure.
146 RTE_FLOW_ITEM_TYPE_PF,
151 * Matches packets addressed to a virtual function ID of the device.
153 * If the underlying device function differs from the one that would
154 * normally receive the matched traffic, specifying this item
155 * prevents it from reaching that device unless the flow rule
156 * contains a VF action. Packets are not duplicated between device
157 * instances by default.
159 * See struct rte_flow_item_vf.
161 RTE_FLOW_ITEM_TYPE_VF,
166 * Matches packets coming from the specified physical port of the
169 * The first PORT item overrides the physical port normally
170 * associated with the specified DPDK input port (port_id). This
171 * item can be provided several times to match additional physical
174 * See struct rte_flow_item_port.
176 RTE_FLOW_ITEM_TYPE_PORT,
179 * Matches a byte string of a given length at a given offset.
181 * See struct rte_flow_item_raw.
183 RTE_FLOW_ITEM_TYPE_RAW,
186 * Matches an Ethernet header.
188 * See struct rte_flow_item_eth.
190 RTE_FLOW_ITEM_TYPE_ETH,
193 * Matches an 802.1Q/ad VLAN tag.
195 * See struct rte_flow_item_vlan.
197 RTE_FLOW_ITEM_TYPE_VLAN,
200 * Matches an IPv4 header.
202 * See struct rte_flow_item_ipv4.
204 RTE_FLOW_ITEM_TYPE_IPV4,
207 * Matches an IPv6 header.
209 * See struct rte_flow_item_ipv6.
211 RTE_FLOW_ITEM_TYPE_IPV6,
214 * Matches an ICMP header.
216 * See struct rte_flow_item_icmp.
218 RTE_FLOW_ITEM_TYPE_ICMP,
221 * Matches a UDP header.
223 * See struct rte_flow_item_udp.
225 RTE_FLOW_ITEM_TYPE_UDP,
228 * Matches a TCP header.
230 * See struct rte_flow_item_tcp.
232 RTE_FLOW_ITEM_TYPE_TCP,
235 * Matches a SCTP header.
237 * See struct rte_flow_item_sctp.
239 RTE_FLOW_ITEM_TYPE_SCTP,
242 * Matches a VXLAN header.
244 * See struct rte_flow_item_vxlan.
246 RTE_FLOW_ITEM_TYPE_VXLAN,
249 * Matches a E_TAG header.
251 * See struct rte_flow_item_e_tag.
253 RTE_FLOW_ITEM_TYPE_E_TAG,
256 * Matches a NVGRE header.
258 * See struct rte_flow_item_nvgre.
260 RTE_FLOW_ITEM_TYPE_NVGRE,
263 * Matches a MPLS header.
265 * See struct rte_flow_item_mpls.
267 RTE_FLOW_ITEM_TYPE_MPLS,
270 * Matches a GRE header.
272 * See struct rte_flow_item_gre.
274 RTE_FLOW_ITEM_TYPE_GRE,
279 * Fuzzy pattern match, expect faster than default.
281 * This is for device that support fuzzy matching option.
282 * Usually a fuzzy matching is fast but the cost is accuracy.
284 * See struct rte_flow_item_fuzzy.
286 RTE_FLOW_ITEM_TYPE_FUZZY,
289 * Matches a GTP header.
291 * Configure flow for GTP packets.
293 * See struct rte_flow_item_gtp.
295 RTE_FLOW_ITEM_TYPE_GTP,
298 * Matches a GTP header.
300 * Configure flow for GTP-C packets.
302 * See struct rte_flow_item_gtp.
304 RTE_FLOW_ITEM_TYPE_GTPC,
307 * Matches a GTP header.
309 * Configure flow for GTP-U packets.
311 * See struct rte_flow_item_gtp.
313 RTE_FLOW_ITEM_TYPE_GTPU,
316 * Matches a ESP header.
318 * See struct rte_flow_item_esp.
320 RTE_FLOW_ITEM_TYPE_ESP,
323 * Matches a GENEVE header.
325 * See struct rte_flow_item_geneve.
327 RTE_FLOW_ITEM_TYPE_GENEVE,
331 * RTE_FLOW_ITEM_TYPE_ANY
333 * Matches any protocol in place of the current layer, a single ANY may also
334 * stand for several protocol layers.
336 * This is usually specified as the first pattern item when looking for a
337 * protocol anywhere in a packet.
339 * A zeroed mask stands for any number of layers.
341 struct rte_flow_item_any {
342 uint32_t num; /**< Number of layers covered. */
345 /** Default mask for RTE_FLOW_ITEM_TYPE_ANY. */
347 static const struct rte_flow_item_any rte_flow_item_any_mask = {
353 * RTE_FLOW_ITEM_TYPE_VF
355 * Matches packets addressed to a virtual function ID of the device.
357 * If the underlying device function differs from the one that would
358 * normally receive the matched traffic, specifying this item prevents it
359 * from reaching that device unless the flow rule contains a VF
360 * action. Packets are not duplicated between device instances by default.
362 * - Likely to return an error or never match any traffic if this causes a
363 * VF device to match traffic addressed to a different VF.
364 * - Can be specified multiple times to match traffic addressed to several
366 * - Can be combined with a PF item to match both PF and VF traffic.
368 * A zeroed mask can be used to match any VF ID.
370 struct rte_flow_item_vf {
371 uint32_t id; /**< Destination VF ID. */
374 /** Default mask for RTE_FLOW_ITEM_TYPE_VF. */
376 static const struct rte_flow_item_vf rte_flow_item_vf_mask = {
382 * RTE_FLOW_ITEM_TYPE_PORT
384 * Matches packets coming from the specified physical port of the underlying
387 * The first PORT item overrides the physical port normally associated with
388 * the specified DPDK input port (port_id). This item can be provided
389 * several times to match additional physical ports.
391 * Note that physical ports are not necessarily tied to DPDK input ports
392 * (port_id) when those are not under DPDK control. Possible values are
393 * specific to each device, they are not necessarily indexed from zero and
394 * may not be contiguous.
396 * As a device property, the list of allowed values as well as the value
397 * associated with a port_id should be retrieved by other means.
399 * A zeroed mask can be used to match any port index.
401 struct rte_flow_item_port {
402 uint32_t index; /**< Physical port index. */
405 /** Default mask for RTE_FLOW_ITEM_TYPE_PORT. */
407 static const struct rte_flow_item_port rte_flow_item_port_mask = {
413 * RTE_FLOW_ITEM_TYPE_RAW
415 * Matches a byte string of a given length at a given offset.
417 * Offset is either absolute (using the start of the packet) or relative to
418 * the end of the previous matched item in the stack, in which case negative
419 * values are allowed.
421 * If search is enabled, offset is used as the starting point. The search
422 * area can be delimited by setting limit to a nonzero value, which is the
423 * maximum number of bytes after offset where the pattern may start.
425 * Matching a zero-length pattern is allowed, doing so resets the relative
426 * offset for subsequent items.
428 * This type does not support ranges (struct rte_flow_item.last).
430 struct rte_flow_item_raw {
431 uint32_t relative:1; /**< Look for pattern after the previous item. */
432 uint32_t search:1; /**< Search pattern from offset (see also limit). */
433 uint32_t reserved:30; /**< Reserved, must be set to zero. */
434 int32_t offset; /**< Absolute or relative offset for pattern. */
435 uint16_t limit; /**< Search area limit for start of pattern. */
436 uint16_t length; /**< Pattern length. */
437 const uint8_t *pattern; /**< Byte string to look for. */
440 /** Default mask for RTE_FLOW_ITEM_TYPE_RAW. */
442 static const struct rte_flow_item_raw rte_flow_item_raw_mask = {
445 .reserved = 0x3fffffff,
446 .offset = 0xffffffff,
454 * RTE_FLOW_ITEM_TYPE_ETH
456 * Matches an Ethernet header.
458 * The @p type field either stands for "EtherType" or "TPID" when followed
459 * by so-called layer 2.5 pattern items such as RTE_FLOW_ITEM_TYPE_VLAN. In
460 * the latter case, @p type refers to that of the outer header, with the
461 * inner EtherType/TPID provided by the subsequent pattern item. This is the
462 * same order as on the wire.
464 struct rte_flow_item_eth {
465 struct ether_addr dst; /**< Destination MAC. */
466 struct ether_addr src; /**< Source MAC. */
467 rte_be16_t type; /**< EtherType or TPID. */
470 /** Default mask for RTE_FLOW_ITEM_TYPE_ETH. */
472 static const struct rte_flow_item_eth rte_flow_item_eth_mask = {
473 .dst.addr_bytes = "\xff\xff\xff\xff\xff\xff",
474 .src.addr_bytes = "\xff\xff\xff\xff\xff\xff",
475 .type = RTE_BE16(0x0000),
480 * RTE_FLOW_ITEM_TYPE_VLAN
482 * Matches an 802.1Q/ad VLAN tag.
484 * The corresponding standard outer EtherType (TPID) values are
485 * ETHER_TYPE_VLAN or ETHER_TYPE_QINQ. It can be overridden by the preceding
488 struct rte_flow_item_vlan {
489 rte_be16_t tci; /**< Tag control information. */
490 rte_be16_t inner_type; /**< Inner EtherType or TPID. */
493 /** Default mask for RTE_FLOW_ITEM_TYPE_VLAN. */
495 static const struct rte_flow_item_vlan rte_flow_item_vlan_mask = {
496 .tci = RTE_BE16(0xffff),
497 .inner_type = RTE_BE16(0x0000),
502 * RTE_FLOW_ITEM_TYPE_IPV4
504 * Matches an IPv4 header.
506 * Note: IPv4 options are handled by dedicated pattern items.
508 struct rte_flow_item_ipv4 {
509 struct ipv4_hdr hdr; /**< IPv4 header definition. */
512 /** Default mask for RTE_FLOW_ITEM_TYPE_IPV4. */
514 static const struct rte_flow_item_ipv4 rte_flow_item_ipv4_mask = {
516 .src_addr = RTE_BE32(0xffffffff),
517 .dst_addr = RTE_BE32(0xffffffff),
523 * RTE_FLOW_ITEM_TYPE_IPV6.
525 * Matches an IPv6 header.
527 * Note: IPv6 options are handled by dedicated pattern items.
529 struct rte_flow_item_ipv6 {
530 struct ipv6_hdr hdr; /**< IPv6 header definition. */
533 /** Default mask for RTE_FLOW_ITEM_TYPE_IPV6. */
535 static const struct rte_flow_item_ipv6 rte_flow_item_ipv6_mask = {
538 "\xff\xff\xff\xff\xff\xff\xff\xff"
539 "\xff\xff\xff\xff\xff\xff\xff\xff",
541 "\xff\xff\xff\xff\xff\xff\xff\xff"
542 "\xff\xff\xff\xff\xff\xff\xff\xff",
548 * RTE_FLOW_ITEM_TYPE_ICMP.
550 * Matches an ICMP header.
552 struct rte_flow_item_icmp {
553 struct icmp_hdr hdr; /**< ICMP header definition. */
556 /** Default mask for RTE_FLOW_ITEM_TYPE_ICMP. */
558 static const struct rte_flow_item_icmp rte_flow_item_icmp_mask = {
567 * RTE_FLOW_ITEM_TYPE_UDP.
569 * Matches a UDP header.
571 struct rte_flow_item_udp {
572 struct udp_hdr hdr; /**< UDP header definition. */
575 /** Default mask for RTE_FLOW_ITEM_TYPE_UDP. */
577 static const struct rte_flow_item_udp rte_flow_item_udp_mask = {
579 .src_port = RTE_BE16(0xffff),
580 .dst_port = RTE_BE16(0xffff),
586 * RTE_FLOW_ITEM_TYPE_TCP.
588 * Matches a TCP header.
590 struct rte_flow_item_tcp {
591 struct tcp_hdr hdr; /**< TCP header definition. */
594 /** Default mask for RTE_FLOW_ITEM_TYPE_TCP. */
596 static const struct rte_flow_item_tcp rte_flow_item_tcp_mask = {
598 .src_port = RTE_BE16(0xffff),
599 .dst_port = RTE_BE16(0xffff),
605 * RTE_FLOW_ITEM_TYPE_SCTP.
607 * Matches a SCTP header.
609 struct rte_flow_item_sctp {
610 struct sctp_hdr hdr; /**< SCTP header definition. */
613 /** Default mask for RTE_FLOW_ITEM_TYPE_SCTP. */
615 static const struct rte_flow_item_sctp rte_flow_item_sctp_mask = {
617 .src_port = RTE_BE16(0xffff),
618 .dst_port = RTE_BE16(0xffff),
624 * RTE_FLOW_ITEM_TYPE_VXLAN.
626 * Matches a VXLAN header (RFC 7348).
628 struct rte_flow_item_vxlan {
629 uint8_t flags; /**< Normally 0x08 (I flag). */
630 uint8_t rsvd0[3]; /**< Reserved, normally 0x000000. */
631 uint8_t vni[3]; /**< VXLAN identifier. */
632 uint8_t rsvd1; /**< Reserved, normally 0x00. */
635 /** Default mask for RTE_FLOW_ITEM_TYPE_VXLAN. */
637 static const struct rte_flow_item_vxlan rte_flow_item_vxlan_mask = {
638 .vni = "\xff\xff\xff",
643 * RTE_FLOW_ITEM_TYPE_E_TAG.
645 * Matches a E-tag header.
647 * The corresponding standard outer EtherType (TPID) value is
648 * ETHER_TYPE_ETAG. It can be overridden by the preceding pattern item.
650 struct rte_flow_item_e_tag {
652 * E-Tag control information (E-TCI).
653 * E-PCP (3b), E-DEI (1b), ingress E-CID base (12b).
655 rte_be16_t epcp_edei_in_ecid_b;
656 /** Reserved (2b), GRP (2b), E-CID base (12b). */
657 rte_be16_t rsvd_grp_ecid_b;
658 uint8_t in_ecid_e; /**< Ingress E-CID ext. */
659 uint8_t ecid_e; /**< E-CID ext. */
660 rte_be16_t inner_type; /**< Inner EtherType or TPID. */
663 /** Default mask for RTE_FLOW_ITEM_TYPE_E_TAG. */
665 static const struct rte_flow_item_e_tag rte_flow_item_e_tag_mask = {
666 .rsvd_grp_ecid_b = RTE_BE16(0x3fff),
671 * RTE_FLOW_ITEM_TYPE_NVGRE.
673 * Matches a NVGRE header.
675 struct rte_flow_item_nvgre {
677 * Checksum (1b), undefined (1b), key bit (1b), sequence number (1b),
678 * reserved 0 (9b), version (3b).
680 * c_k_s_rsvd0_ver must have value 0x2000 according to RFC 7637.
682 rte_be16_t c_k_s_rsvd0_ver;
683 rte_be16_t protocol; /**< Protocol type (0x6558). */
684 uint8_t tni[3]; /**< Virtual subnet ID. */
685 uint8_t flow_id; /**< Flow ID. */
688 /** Default mask for RTE_FLOW_ITEM_TYPE_NVGRE. */
690 static const struct rte_flow_item_nvgre rte_flow_item_nvgre_mask = {
691 .tni = "\xff\xff\xff",
696 * RTE_FLOW_ITEM_TYPE_MPLS.
698 * Matches a MPLS header.
700 struct rte_flow_item_mpls {
702 * Label (20b), TC (3b), Bottom of Stack (1b).
704 uint8_t label_tc_s[3];
705 uint8_t ttl; /** Time-to-Live. */
708 /** Default mask for RTE_FLOW_ITEM_TYPE_MPLS. */
710 static const struct rte_flow_item_mpls rte_flow_item_mpls_mask = {
711 .label_tc_s = "\xff\xff\xf0",
716 * RTE_FLOW_ITEM_TYPE_GRE.
718 * Matches a GRE header.
720 struct rte_flow_item_gre {
722 * Checksum (1b), reserved 0 (12b), version (3b).
725 rte_be16_t c_rsvd0_ver;
726 rte_be16_t protocol; /**< Protocol type. */
729 /** Default mask for RTE_FLOW_ITEM_TYPE_GRE. */
731 static const struct rte_flow_item_gre rte_flow_item_gre_mask = {
732 .protocol = RTE_BE16(0xffff),
737 * RTE_FLOW_ITEM_TYPE_FUZZY
739 * Fuzzy pattern match, expect faster than default.
741 * This is for device that support fuzzy match option.
742 * Usually a fuzzy match is fast but the cost is accuracy.
743 * i.e. Signature Match only match pattern's hash value, but it is
744 * possible two different patterns have the same hash value.
746 * Matching accuracy level can be configure by threshold.
747 * Driver can divide the range of threshold and map to different
748 * accuracy levels that device support.
750 * Threshold 0 means perfect match (no fuzziness), while threshold
751 * 0xffffffff means fuzziest match.
753 struct rte_flow_item_fuzzy {
754 uint32_t thresh; /**< Accuracy threshold. */
757 /** Default mask for RTE_FLOW_ITEM_TYPE_FUZZY. */
759 static const struct rte_flow_item_fuzzy rte_flow_item_fuzzy_mask = {
760 .thresh = 0xffffffff,
765 * RTE_FLOW_ITEM_TYPE_GTP.
767 * Matches a GTPv1 header.
769 struct rte_flow_item_gtp {
771 * Version (3b), protocol type (1b), reserved (1b),
772 * Extension header flag (1b),
773 * Sequence number flag (1b),
774 * N-PDU number flag (1b).
776 uint8_t v_pt_rsv_flags;
777 uint8_t msg_type; /**< Message type. */
778 rte_be16_t msg_len; /**< Message length. */
779 rte_be32_t teid; /**< Tunnel endpoint identifier. */
782 /** Default mask for RTE_FLOW_ITEM_TYPE_GTP. */
784 static const struct rte_flow_item_gtp rte_flow_item_gtp_mask = {
785 .teid = RTE_BE32(0xffffffff),
790 * RTE_FLOW_ITEM_TYPE_ESP
792 * Matches an ESP header.
794 struct rte_flow_item_esp {
795 struct esp_hdr hdr; /**< ESP header definition. */
798 /** Default mask for RTE_FLOW_ITEM_TYPE_ESP. */
800 static const struct rte_flow_item_esp rte_flow_item_esp_mask = {
808 * RTE_FLOW_ITEM_TYPE_GENEVE.
810 * Matches a GENEVE header.
812 struct rte_flow_item_geneve {
814 * Version (2b), length of the options fields (6b), OAM packet (1b),
815 * critical options present (1b), reserved 0 (6b).
817 rte_be16_t ver_opt_len_o_c_rsvd0;
818 rte_be16_t protocol; /**< Protocol type. */
819 uint8_t vni[3]; /**< Virtual Network Identifier. */
820 uint8_t rsvd1; /**< Reserved, normally 0x00. */
823 /** Default mask for RTE_FLOW_ITEM_TYPE_GENEVE. */
825 static const struct rte_flow_item_geneve rte_flow_item_geneve_mask = {
826 .vni = "\xff\xff\xff",
831 * Matching pattern item definition.
833 * A pattern is formed by stacking items starting from the lowest protocol
834 * layer to match. This stacking restriction does not apply to meta items
835 * which can be placed anywhere in the stack without affecting the meaning
836 * of the resulting pattern.
838 * Patterns are terminated by END items.
840 * The spec field should be a valid pointer to a structure of the related
841 * item type. It may remain unspecified (NULL) in many cases to request
842 * broad (nonspecific) matching. In such cases, last and mask must also be
845 * Optionally, last can point to a structure of the same type to define an
846 * inclusive range. This is mostly supported by integer and address fields,
847 * may cause errors otherwise. Fields that do not support ranges must be set
848 * to 0 or to the same value as the corresponding fields in spec.
850 * Only the fields defined to nonzero values in the default masks (see
851 * rte_flow_item_{name}_mask constants) are considered relevant by
852 * default. This can be overridden by providing a mask structure of the
853 * same type with applicable bits set to one. It can also be used to
854 * partially filter out specific fields (e.g. as an alternate mean to match
855 * ranges of IP addresses).
857 * Mask is a simple bit-mask applied before interpreting the contents of
858 * spec and last, which may yield unexpected results if not used
859 * carefully. For example, if for an IPv4 address field, spec provides
860 * 10.1.2.3, last provides 10.3.4.5 and mask provides 255.255.0.0, the
861 * effective range becomes 10.1.0.0 to 10.3.255.255.
863 struct rte_flow_item {
864 enum rte_flow_item_type type; /**< Item type. */
865 const void *spec; /**< Pointer to item specification structure. */
866 const void *last; /**< Defines an inclusive range (spec to last). */
867 const void *mask; /**< Bit-mask applied to spec and last. */
873 * Each possible action is represented by a type. Some have associated
874 * configuration structures. Several actions combined in a list can be
875 * assigned to a flow rule and are performed in order.
877 * They fall in three categories:
879 * - Actions that modify the fate of matching traffic, for instance by
880 * dropping or assigning it a specific destination.
882 * - Actions that modify matching traffic contents or its properties. This
883 * includes adding/removing encapsulation, encryption, compression and
886 * - Actions related to the flow rule itself, such as updating counters or
887 * making it non-terminating.
889 * Flow rules being terminating by default, not specifying any action of the
890 * fate kind results in undefined behavior. This applies to both ingress and
893 * PASSTHRU, when supported, makes a flow rule non-terminating.
895 enum rte_flow_action_type {
897 * End marker for action lists. Prevents further processing of
898 * actions, thereby ending the list.
900 * No associated configuration structure.
902 RTE_FLOW_ACTION_TYPE_END,
905 * Used as a placeholder for convenience. It is ignored and simply
908 * No associated configuration structure.
910 RTE_FLOW_ACTION_TYPE_VOID,
913 * Leaves traffic up for additional processing by subsequent flow
914 * rules; makes a flow rule non-terminating.
916 * No associated configuration structure.
918 RTE_FLOW_ACTION_TYPE_PASSTHRU,
921 * Attaches an integer value to packets and sets PKT_RX_FDIR and
922 * PKT_RX_FDIR_ID mbuf flags.
924 * See struct rte_flow_action_mark.
926 RTE_FLOW_ACTION_TYPE_MARK,
929 * Flags packets. Similar to MARK without a specific value; only
930 * sets the PKT_RX_FDIR mbuf flag.
932 * No associated configuration structure.
934 RTE_FLOW_ACTION_TYPE_FLAG,
937 * Assigns packets to a given queue index.
939 * See struct rte_flow_action_queue.
941 RTE_FLOW_ACTION_TYPE_QUEUE,
946 * PASSTHRU overrides this action if both are specified.
948 * No associated configuration structure.
950 RTE_FLOW_ACTION_TYPE_DROP,
953 * Enables counters for this flow rule.
955 * These counters can be retrieved and reset through rte_flow_query(),
956 * see struct rte_flow_query_count.
958 * No associated configuration structure.
960 RTE_FLOW_ACTION_TYPE_COUNT,
963 * Similar to QUEUE, except RSS is additionally performed on packets
964 * to spread them among several queues according to the provided
967 * See struct rte_flow_action_rss.
969 RTE_FLOW_ACTION_TYPE_RSS,
972 * Redirects packets to the physical function (PF) of the current
975 * No associated configuration structure.
977 RTE_FLOW_ACTION_TYPE_PF,
980 * Redirects packets to the virtual function (VF) of the current
981 * device with the specified ID.
983 * See struct rte_flow_action_vf.
985 RTE_FLOW_ACTION_TYPE_VF,
988 * Traffic metering and policing (MTR).
990 * See struct rte_flow_action_meter.
991 * See file rte_mtr.h for MTR object configuration.
993 RTE_FLOW_ACTION_TYPE_METER,
996 * Redirects packets to security engine of current device for security
997 * processing as specified by security session.
999 * See struct rte_flow_action_security.
1001 RTE_FLOW_ACTION_TYPE_SECURITY
1005 * RTE_FLOW_ACTION_TYPE_MARK
1007 * Attaches an integer value to packets and sets PKT_RX_FDIR and
1008 * PKT_RX_FDIR_ID mbuf flags.
1010 * This value is arbitrary and application-defined. Maximum allowed value
1011 * depends on the underlying implementation. It is returned in the
1012 * hash.fdir.hi mbuf field.
1014 struct rte_flow_action_mark {
1015 uint32_t id; /**< Integer value to return with packets. */
1019 * RTE_FLOW_ACTION_TYPE_QUEUE
1021 * Assign packets to a given queue index.
1023 struct rte_flow_action_queue {
1024 uint16_t index; /**< Queue index to use. */
1028 * RTE_FLOW_ACTION_TYPE_COUNT (query)
1030 * Query structure to retrieve and reset flow rule counters.
1032 struct rte_flow_query_count {
1033 uint32_t reset:1; /**< Reset counters after query [in]. */
1034 uint32_t hits_set:1; /**< hits field is set [out]. */
1035 uint32_t bytes_set:1; /**< bytes field is set [out]. */
1036 uint32_t reserved:29; /**< Reserved, must be zero [in, out]. */
1037 uint64_t hits; /**< Number of hits for this rule [out]. */
1038 uint64_t bytes; /**< Number of bytes through this rule [out]. */
1042 * RTE_FLOW_ACTION_TYPE_RSS
1044 * Similar to QUEUE, except RSS is additionally performed on packets to
1045 * spread them among several queues according to the provided parameters.
1047 * Unlike global RSS settings used by other DPDK APIs, unsetting the
1048 * @p types field does not disable RSS in a flow rule. Doing so instead
1049 * requests safe unspecified "best-effort" settings from the underlying PMD,
1050 * which depending on the flow rule, may result in anything ranging from
1051 * empty (single queue) to all-inclusive RSS.
1053 * Note: RSS hash result is stored in the hash.rss mbuf field which overlaps
1054 * hash.fdir.lo. Since the MARK action sets the hash.fdir.hi field only,
1055 * both can be requested simultaneously.
1057 struct rte_flow_action_rss {
1058 enum rte_eth_hash_function func; /**< RSS hash function to apply. */
1060 * Packet encapsulation level RSS hash @p types apply to.
1062 * - @p 0 requests the default behavior. Depending on the packet
1063 * type, it can mean outermost, innermost, anything in between or
1066 * It basically stands for the innermost encapsulation level RSS
1067 * can be performed on according to PMD and device capabilities.
1069 * - @p 1 requests RSS to be performed on the outermost packet
1070 * encapsulation level.
1072 * - @p 2 and subsequent values request RSS to be performed on the
1073 * specified inner packet encapsulation level, from outermost to
1074 * innermost (lower to higher values).
1076 * Values other than @p 0 are not necessarily supported.
1078 * Requesting a specific RSS level on unrecognized traffic results
1079 * in undefined behavior. For predictable results, it is recommended
1080 * to make the flow rule pattern match packet headers up to the
1081 * requested encapsulation level so that only matching traffic goes
1085 uint64_t types; /**< Specific RSS hash types (see ETH_RSS_*). */
1086 uint32_t key_len; /**< Hash key length in bytes. */
1087 uint32_t queue_num; /**< Number of entries in @p queue. */
1088 const uint8_t *key; /**< Hash key. */
1089 const uint16_t *queue; /**< Queue indices to use. */
1093 * RTE_FLOW_ACTION_TYPE_VF
1095 * Redirects packets to a virtual function (VF) of the current device.
1097 * Packets matched by a VF pattern item can be redirected to their original
1098 * VF ID instead of the specified one. This parameter may not be available
1099 * and is not guaranteed to work properly if the VF part is matched by a
1100 * prior flow rule or if packets are not addressed to a VF in the first
1103 struct rte_flow_action_vf {
1104 uint32_t original:1; /**< Use original VF ID if possible. */
1105 uint32_t reserved:31; /**< Reserved, must be zero. */
1106 uint32_t id; /**< VF ID to redirect packets to. */
1110 * RTE_FLOW_ACTION_TYPE_METER
1112 * Traffic metering and policing (MTR).
1114 * Packets matched by items of this type can be either dropped or passed to the
1115 * next item with their color set by the MTR object.
1117 struct rte_flow_action_meter {
1118 uint32_t mtr_id; /**< MTR object ID created with rte_mtr_create(). */
1122 * RTE_FLOW_ACTION_TYPE_SECURITY
1124 * Perform the security action on flows matched by the pattern items
1125 * according to the configuration of the security session.
1127 * This action modifies the payload of matched flows. For INLINE_CRYPTO, the
1128 * security protocol headers and IV are fully provided by the application as
1129 * specified in the flow pattern. The payload of matching packets is
1130 * encrypted on egress, and decrypted and authenticated on ingress.
1131 * For INLINE_PROTOCOL, the security protocol is fully offloaded to HW,
1132 * providing full encapsulation and decapsulation of packets in security
1133 * protocols. The flow pattern specifies both the outer security header fields
1134 * and the inner packet fields. The security session specified in the action
1135 * must match the pattern parameters.
1137 * The security session specified in the action must be created on the same
1138 * port as the flow action that is being specified.
1140 * The ingress/egress flow attribute should match that specified in the
1141 * security session if the security session supports the definition of the
1144 * Multiple flows can be configured to use the same security session.
1146 struct rte_flow_action_security {
1147 void *security_session; /**< Pointer to security session structure. */
1151 * Definition of a single action.
1153 * A list of actions is terminated by a END action.
1155 * For simple actions without a configuration structure, conf remains NULL.
1157 struct rte_flow_action {
1158 enum rte_flow_action_type type; /**< Action type. */
1159 const void *conf; /**< Pointer to action configuration structure. */
1163 * Opaque type returned after successfully creating a flow.
1165 * This handle can be used to manage and query the related flow (e.g. to
1166 * destroy it or retrieve counters).
1171 * Verbose error types.
1173 * Most of them provide the type of the object referenced by struct
1174 * rte_flow_error.cause.
1176 enum rte_flow_error_type {
1177 RTE_FLOW_ERROR_TYPE_NONE, /**< No error. */
1178 RTE_FLOW_ERROR_TYPE_UNSPECIFIED, /**< Cause unspecified. */
1179 RTE_FLOW_ERROR_TYPE_HANDLE, /**< Flow rule (handle). */
1180 RTE_FLOW_ERROR_TYPE_ATTR_GROUP, /**< Group field. */
1181 RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, /**< Priority field. */
1182 RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, /**< Ingress field. */
1183 RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, /**< Egress field. */
1184 RTE_FLOW_ERROR_TYPE_ATTR, /**< Attributes structure. */
1185 RTE_FLOW_ERROR_TYPE_ITEM_NUM, /**< Pattern length. */
1186 RTE_FLOW_ERROR_TYPE_ITEM_SPEC, /**< Item specification. */
1187 RTE_FLOW_ERROR_TYPE_ITEM_LAST, /**< Item specification range. */
1188 RTE_FLOW_ERROR_TYPE_ITEM_MASK, /**< Item specification mask. */
1189 RTE_FLOW_ERROR_TYPE_ITEM, /**< Specific pattern item. */
1190 RTE_FLOW_ERROR_TYPE_ACTION_NUM, /**< Number of actions. */
1191 RTE_FLOW_ERROR_TYPE_ACTION_CONF, /**< Action configuration. */
1192 RTE_FLOW_ERROR_TYPE_ACTION, /**< Specific action. */
1196 * Verbose error structure definition.
1198 * This object is normally allocated by applications and set by PMDs, the
1199 * message points to a constant string which does not need to be freed by
1200 * the application, however its pointer can be considered valid only as long
1201 * as its associated DPDK port remains configured. Closing the underlying
1202 * device or unloading the PMD invalidates it.
1204 * Both cause and message may be NULL regardless of the error type.
1206 struct rte_flow_error {
1207 enum rte_flow_error_type type; /**< Cause field and error types. */
1208 const void *cause; /**< Object responsible for the error. */
1209 const char *message; /**< Human-readable error message. */
1213 * Check whether a flow rule can be created on a given port.
1215 * The flow rule is validated for correctness and whether it could be accepted
1216 * by the device given sufficient resources. The rule is checked against the
1217 * current device mode and queue configuration. The flow rule may also
1218 * optionally be validated against existing flow rules and device resources.
1219 * This function has no effect on the target device.
1221 * The returned value is guaranteed to remain valid only as long as no
1222 * successful calls to rte_flow_create() or rte_flow_destroy() are made in
1223 * the meantime and no device parameter affecting flow rules in any way are
1224 * modified, due to possible collisions or resource limitations (although in
1225 * such cases EINVAL should not be returned).
1228 * Port identifier of Ethernet device.
1230 * Flow rule attributes.
1231 * @param[in] pattern
1232 * Pattern specification (list terminated by the END pattern item).
1233 * @param[in] actions
1234 * Associated actions (list terminated by the END action).
1236 * Perform verbose error reporting if not NULL. PMDs initialize this
1237 * structure in case of error only.
1240 * 0 if flow rule is valid and can be created. A negative errno value
1241 * otherwise (rte_errno is also set), the following errors are defined:
1243 * -ENOSYS: underlying device does not support this functionality.
1245 * -EIO: underlying device is removed.
1247 * -EINVAL: unknown or invalid rule specification.
1249 * -ENOTSUP: valid but unsupported rule specification (e.g. partial
1250 * bit-masks are unsupported).
1252 * -EEXIST: collision with an existing rule. Only returned if device
1253 * supports flow rule collision checking and there was a flow rule
1254 * collision. Not receiving this return code is no guarantee that creating
1255 * the rule will not fail due to a collision.
1257 * -ENOMEM: not enough memory to execute the function, or if the device
1258 * supports resource validation, resource limitation on the device.
1260 * -EBUSY: action cannot be performed due to busy device resources, may
1261 * succeed if the affected queues or even the entire port are in a stopped
1262 * state (see rte_eth_dev_rx_queue_stop() and rte_eth_dev_stop()).
1265 rte_flow_validate(uint16_t port_id,
1266 const struct rte_flow_attr *attr,
1267 const struct rte_flow_item pattern[],
1268 const struct rte_flow_action actions[],
1269 struct rte_flow_error *error);
1272 * Create a flow rule on a given port.
1275 * Port identifier of Ethernet device.
1277 * Flow rule attributes.
1278 * @param[in] pattern
1279 * Pattern specification (list terminated by the END pattern item).
1280 * @param[in] actions
1281 * Associated actions (list terminated by the END action).
1283 * Perform verbose error reporting if not NULL. PMDs initialize this
1284 * structure in case of error only.
1287 * A valid handle in case of success, NULL otherwise and rte_errno is set
1288 * to the positive version of one of the error codes defined for
1289 * rte_flow_validate().
1292 rte_flow_create(uint16_t port_id,
1293 const struct rte_flow_attr *attr,
1294 const struct rte_flow_item pattern[],
1295 const struct rte_flow_action actions[],
1296 struct rte_flow_error *error);
1299 * Destroy a flow rule on a given port.
1301 * Failure to destroy a flow rule handle may occur when other flow rules
1302 * depend on it, and destroying it would result in an inconsistent state.
1304 * This function is only guaranteed to succeed if handles are destroyed in
1305 * reverse order of their creation.
1308 * Port identifier of Ethernet device.
1310 * Flow rule handle to destroy.
1312 * Perform verbose error reporting if not NULL. PMDs initialize this
1313 * structure in case of error only.
1316 * 0 on success, a negative errno value otherwise and rte_errno is set.
1319 rte_flow_destroy(uint16_t port_id,
1320 struct rte_flow *flow,
1321 struct rte_flow_error *error);
1324 * Destroy all flow rules associated with a port.
1326 * In the unlikely event of failure, handles are still considered destroyed
1327 * and no longer valid but the port must be assumed to be in an inconsistent
1331 * Port identifier of Ethernet device.
1333 * Perform verbose error reporting if not NULL. PMDs initialize this
1334 * structure in case of error only.
1337 * 0 on success, a negative errno value otherwise and rte_errno is set.
1340 rte_flow_flush(uint16_t port_id,
1341 struct rte_flow_error *error);
1344 * Query an existing flow rule.
1346 * This function allows retrieving flow-specific data such as counters.
1347 * Data is gathered by special actions which must be present in the flow
1350 * \see RTE_FLOW_ACTION_TYPE_COUNT
1353 * Port identifier of Ethernet device.
1355 * Flow rule handle to query.
1357 * Action type to query.
1358 * @param[in, out] data
1359 * Pointer to storage for the associated query data type.
1361 * Perform verbose error reporting if not NULL. PMDs initialize this
1362 * structure in case of error only.
1365 * 0 on success, a negative errno value otherwise and rte_errno is set.
1368 rte_flow_query(uint16_t port_id,
1369 struct rte_flow *flow,
1370 enum rte_flow_action_type action,
1372 struct rte_flow_error *error);
1375 * Restrict ingress traffic to the defined flow rules.
1377 * Isolated mode guarantees that all ingress traffic comes from defined flow
1378 * rules only (current and future).
1380 * Besides making ingress more deterministic, it allows PMDs to safely reuse
1381 * resources otherwise assigned to handle the remaining traffic, such as
1382 * global RSS configuration settings, VLAN filters, MAC address entries,
1383 * legacy filter API rules and so on in order to expand the set of possible
1386 * Calling this function as soon as possible after device initialization,
1387 * ideally before the first call to rte_eth_dev_configure(), is recommended
1388 * to avoid possible failures due to conflicting settings.
1390 * Once effective, leaving isolated mode may not be possible depending on
1391 * PMD implementation.
1393 * Additionally, the following functionality has no effect on the underlying
1394 * port and may return errors such as ENOTSUP ("not supported"):
1396 * - Toggling promiscuous mode.
1397 * - Toggling allmulticast mode.
1398 * - Configuring MAC addresses.
1399 * - Configuring multicast addresses.
1400 * - Configuring VLAN filters.
1401 * - Configuring Rx filters through the legacy API (e.g. FDIR).
1402 * - Configuring global RSS settings.
1405 * Port identifier of Ethernet device.
1407 * Nonzero to enter isolated mode, attempt to leave it otherwise.
1409 * Perform verbose error reporting if not NULL. PMDs initialize this
1410 * structure in case of error only.
1413 * 0 on success, a negative errno value otherwise and rte_errno is set.
1416 rte_flow_isolate(uint16_t port_id, int set, struct rte_flow_error *error);
1419 * Initialize flow error structure.
1422 * Pointer to flow error structure (may be NULL).
1424 * Related error code (rte_errno).
1426 * Cause field and error types.
1428 * Object responsible for the error.
1430 * Human-readable error message.
1433 * Negative error code (errno value) and rte_errno is set.
1436 rte_flow_error_set(struct rte_flow_error *error,
1438 enum rte_flow_error_type type,
1440 const char *message);
1443 * Generic flow representation.
1445 * This form is sufficient to describe an rte_flow independently from any
1446 * PMD implementation and allows for replayability and identification.
1448 struct rte_flow_desc {
1449 size_t size; /**< Allocated space including data[]. */
1450 struct rte_flow_attr attr; /**< Attributes. */
1451 struct rte_flow_item *items; /**< Items. */
1452 struct rte_flow_action *actions; /**< Actions. */
1453 uint8_t data[]; /**< Storage for items/actions. */
1457 * Copy an rte_flow rule description.
1460 * Flow rule description.
1462 * Total size of allocated data for the flow description.
1464 * Flow rule attributes.
1466 * Pattern specification (list terminated by the END pattern item).
1467 * @param[in] actions
1468 * Associated actions (list terminated by the END action).
1471 * If len is greater or equal to the size of the flow, the total size of the
1472 * flow description and its data.
1473 * If len is lower than the size of the flow, the number of bytes that would
1474 * have been written to desc had it been sufficient. Nothing is written.
1477 rte_flow_copy(struct rte_flow_desc *fd, size_t len,
1478 const struct rte_flow_attr *attr,
1479 const struct rte_flow_item *items,
1480 const struct rte_flow_action *actions);
1486 #endif /* RTE_FLOW_H_ */