1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright 2016 6WIND S.A.
3 Copyright 2016 Mellanox Technologies, Ltd
7 Generic flow API (rte_flow)
8 ===========================
13 This API provides a generic means to configure hardware to match specific
14 ingress or egress traffic, alter its fate and query related counters
15 according to any number of user-defined rules.
17 It is named *rte_flow* after the prefix used for all its symbols, and is
18 defined in ``rte_flow.h``.
20 - Matching can be performed on packet data (protocol headers, payload) and
21 properties (e.g. associated physical port, virtual device function ID).
23 - Possible operations include dropping traffic, diverting it to specific
24 queues, to virtual/physical device functions or ports, performing tunnel
25 offloads, adding marks and so on.
27 It is slightly higher-level than the legacy filtering framework which it
28 encompasses and supersedes (including all functions and filter types) in
29 order to expose a single interface with an unambiguous behavior that is
30 common to all poll-mode drivers (PMDs).
38 A flow rule is the combination of attributes with a matching pattern and a
39 list of actions. Flow rules form the basis of this API.
41 Flow rules can have several distinct actions (such as counting,
42 encapsulating, decapsulating before redirecting packets to a particular
43 queue, etc.), instead of relying on several rules to achieve this and having
44 applications deal with hardware implementation details regarding their
47 Support for different priority levels on a rule basis is provided, for
48 example in order to force a more specific rule to come before a more generic
49 one for packets matched by both. However hardware support for more than a
50 single priority level cannot be guaranteed. When supported, the number of
51 available priority levels is usually low, which is why they can also be
52 implemented in software by PMDs (e.g. missing priority levels may be
53 emulated by reordering rules).
55 In order to remain as hardware-agnostic as possible, by default all rules
56 are considered to have the same priority, which means that the order between
57 overlapping rules (when a packet is matched by several filters) is
60 PMDs may refuse to create overlapping rules at a given priority level when
61 they can be detected (e.g. if a pattern matches an existing filter).
63 Thus predictable results for a given priority level can only be achieved
64 with non-overlapping rules, using perfect matching on all protocol layers.
66 Flow rules can also be grouped, the flow rule priority is specific to the
67 group they belong to. All flow rules in a given group are thus processed within
68 the context of that group. Groups are not linked by default, so the logical
69 hierarchy of groups must be explicitly defined by flow rules themselves in each
70 group using the JUMP action to define the next group to redirect too. Only flow
71 rules defined in the default group 0 are guarantee to be matched against, this
72 makes group 0 the origin of any group hierarchy defined by an application.
74 Support for multiple actions per rule may be implemented internally on top
75 of non-default hardware priorities, as a result both features may not be
76 simultaneously available to applications.
78 Considering that allowed pattern/actions combinations cannot be known in
79 advance and would result in an impractically large number of capabilities to
80 expose, a method is provided to validate a given rule from the current
81 device configuration state.
83 This enables applications to check if the rule types they need is supported
84 at initialization time, before starting their data path. This method can be
85 used anytime, its only requirement being that the resources needed by a rule
86 should exist (e.g. a target RX queue should be configured first).
88 Each defined rule is associated with an opaque handle managed by the PMD,
89 applications are responsible for keeping it. These can be used for queries
90 and rules management, such as retrieving counters or other data and
93 To avoid resource leaks on the PMD side, handles must be explicitly
94 destroyed by the application before releasing associated resources such as
97 The following sections cover:
99 - **Attributes** (represented by ``struct rte_flow_attr``): properties of a
100 flow rule such as its direction (ingress or egress) and priority.
102 - **Pattern item** (represented by ``struct rte_flow_item``): part of a
103 matching pattern that either matches specific packet data or traffic
104 properties. It can also describe properties of the pattern itself, such as
107 - **Matching pattern**: traffic properties to look for, a combination of any
110 - **Actions** (represented by ``struct rte_flow_action``): operations to
111 perform whenever a packet is matched by a pattern.
119 Flow rules can be grouped by assigning them a common group number. Groups
120 allow a logical hierarchy of flow rule groups (tables) to be defined. These
121 groups can be supported virtually in the PMD or in the physical device.
122 Group 0 is the default group and this is the only group which flows are
123 guarantee to matched against, all subsequent groups can only be reached by
124 way of the JUMP action from a matched flow rule.
126 Although optional, applications are encouraged to group similar rules as
127 much as possible to fully take advantage of hardware capabilities
128 (e.g. optimized matching) and work around limitations (e.g. a single pattern
129 type possibly allowed in a given group), while being aware that the groups
130 hierarchies must be programmed explicitly.
132 Note that support for more than a single group is not guaranteed.
137 A priority level can be assigned to a flow rule, lower values
138 denote higher priority, with 0 as the maximum.
140 Priority levels are arbitrary and up to the application, they do
141 not need to be contiguous nor start from 0, however the maximum number
142 varies between devices and may be affected by existing flow rules.
144 A flow which matches multiple rules in the same group will always matched by
145 the rule with the highest priority in that group.
147 If a packet is matched by several rules of a given group for a given
148 priority level, the outcome is undefined. It can take any path, may be
149 duplicated or even cause unrecoverable errors.
151 Note that support for more than a single priority level is not guaranteed.
153 Attribute: Traffic direction
154 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
156 Flow rule patterns apply to inbound and/or outbound traffic.
158 In the context of this API, **ingress** and **egress** respectively stand
159 for **inbound** and **outbound** based on the standpoint of the application
160 creating a flow rule.
162 There are no exceptions to this definition.
164 Several pattern items and actions are valid and can be used in both
165 directions. At least one direction must be specified.
167 Specifying both directions at once for a given rule is not recommended but
168 may be valid in a few cases (e.g. shared counters).
173 Instead of simply matching the properties of traffic as it would appear on a
174 given DPDK port ID, enabling this attribute transfers a flow rule to the
175 lowest possible level of any device endpoints found in the pattern.
177 When supported, this effectively enables an application to reroute traffic
178 not necessarily intended for it (e.g. coming from or addressed to different
179 physical ports, VFs or applications) at the device level.
181 It complements the behavior of some pattern items such as `Item: PHY_PORT`_
182 and is meaningless without them.
184 When transferring flow rules, **ingress** and **egress** attributes
185 (`Attribute: Traffic direction`_) keep their original meaning, as if
186 processing traffic emitted or received by the application.
191 Pattern items fall in two categories:
193 - Matching protocol headers and packet data, usually associated with a
194 specification structure. These must be stacked in the same order as the
195 protocol layers to match inside packets, starting from the lowest.
197 - Matching meta-data or affecting pattern processing, often without a
198 specification structure. Since they do not match packet contents, their
199 position in the list is usually not relevant.
201 Item specification structures are used to match specific values among
202 protocol fields (or item properties). Documentation describes for each item
203 whether they are associated with one and their type name if so.
205 Up to three structures of the same type can be set for a given item:
207 - ``spec``: values to match (e.g. a given IPv4 address).
209 - ``last``: upper bound for an inclusive range with corresponding fields in
212 - ``mask``: bit-mask applied to both ``spec`` and ``last`` whose purpose is
213 to distinguish the values to take into account and/or partially mask them
214 out (e.g. in order to match an IPv4 address prefix).
216 Usage restrictions and expected behavior:
218 - Setting either ``mask`` or ``last`` without ``spec`` is an error.
220 - Field values in ``last`` which are either 0 or equal to the corresponding
221 values in ``spec`` are ignored; they do not generate a range. Nonzero
222 values lower than those in ``spec`` are not supported.
224 - Setting ``spec`` and optionally ``last`` without ``mask`` causes the PMD
225 to use the default mask defined for that item (defined as
226 ``rte_flow_item_{name}_mask`` constants).
228 - Not setting any of them (assuming item type allows it) is equivalent to
229 providing an empty (zeroed) ``mask`` for broad (nonspecific) matching.
231 - ``mask`` is a simple bit-mask applied before interpreting the contents of
232 ``spec`` and ``last``, which may yield unexpected results if not used
233 carefully. For example, if for an IPv4 address field, ``spec`` provides
234 *10.1.2.3*, ``last`` provides *10.3.4.5* and ``mask`` provides
235 *255.255.0.0*, the effective range becomes *10.1.0.0* to *10.3.255.255*.
237 Example of an item specification matching an Ethernet header:
239 .. _table_rte_flow_pattern_item_example:
241 .. table:: Ethernet item
243 +----------+----------+--------------------+
244 | Field | Subfield | Value |
245 +==========+==========+====================+
246 | ``spec`` | ``src`` | ``00:01:02:03:04`` |
247 | +----------+--------------------+
248 | | ``dst`` | ``00:2a:66:00:01`` |
249 | +----------+--------------------+
250 | | ``type`` | ``0x22aa`` |
251 +----------+----------+--------------------+
252 | ``last`` | unspecified |
253 +----------+----------+--------------------+
254 | ``mask`` | ``src`` | ``00:ff:ff:ff:00`` |
255 | +----------+--------------------+
256 | | ``dst`` | ``00:00:00:00:ff`` |
257 | +----------+--------------------+
258 | | ``type`` | ``0x0000`` |
259 +----------+----------+--------------------+
261 Non-masked bits stand for any value (shown as ``?`` below), Ethernet headers
262 with the following properties are thus matched:
264 - ``src``: ``??:01:02:03:??``
265 - ``dst``: ``??:??:??:??:01``
266 - ``type``: ``0x????``
271 A pattern is formed by stacking items starting from the lowest protocol
272 layer to match. This stacking restriction does not apply to meta items which
273 can be placed anywhere in the stack without affecting the meaning of the
276 Patterns are terminated by END items.
280 .. _table_rte_flow_tcpv4_as_l4:
282 .. table:: TCPv4 as L4
298 .. _table_rte_flow_tcpv6_in_vxlan:
300 .. table:: TCPv6 in VXLAN
302 +-------+------------+
304 +=======+============+
306 +-------+------------+
308 +-------+------------+
310 +-------+------------+
312 +-------+------------+
314 +-------+------------+
316 +-------+------------+
318 +-------+------------+
320 +-------+------------+
324 .. _table_rte_flow_tcpv4_as_l4_meta:
326 .. table:: TCPv4 as L4 with meta items
348 The above example shows how meta items do not affect packet data matching
349 items, as long as those remain stacked properly. The resulting matching
350 pattern is identical to "TCPv4 as L4".
352 .. _table_rte_flow_udpv6_anywhere:
354 .. table:: UDPv6 anywhere
366 If supported by the PMD, omitting one or several protocol layers at the
367 bottom of the stack as in the above example (missing an Ethernet
368 specification) enables looking up anywhere in packets.
370 It is unspecified whether the payload of supported encapsulations
371 (e.g. VXLAN payload) is matched by such a pattern, which may apply to inner,
372 outer or both packets.
374 .. _table_rte_flow_invalid_l3:
376 .. table:: Invalid, missing L3
388 The above pattern is invalid due to a missing L3 specification between L2
389 (Ethernet) and L4 (UDP). Doing so is only allowed at the bottom and at the
395 They match meta-data or affect pattern processing instead of matching packet
396 data directly, most of them do not need a specification structure. This
397 particularity allows them to be specified anywhere in the stack without
398 causing any side effect.
403 End marker for item lists. Prevents further processing of items, thereby
406 - Its numeric value is 0 for convenience.
407 - PMD support is mandatory.
408 - ``spec``, ``last`` and ``mask`` are ignored.
410 .. _table_rte_flow_item_end:
414 +----------+---------+
416 +==========+=========+
417 | ``spec`` | ignored |
418 +----------+---------+
419 | ``last`` | ignored |
420 +----------+---------+
421 | ``mask`` | ignored |
422 +----------+---------+
427 Used as a placeholder for convenience. It is ignored and simply discarded by
430 - PMD support is mandatory.
431 - ``spec``, ``last`` and ``mask`` are ignored.
433 .. _table_rte_flow_item_void:
437 +----------+---------+
439 +==========+=========+
440 | ``spec`` | ignored |
441 +----------+---------+
442 | ``last`` | ignored |
443 +----------+---------+
444 | ``mask`` | ignored |
445 +----------+---------+
447 One usage example for this type is generating rules that share a common
448 prefix quickly without reallocating memory, only by updating item types:
450 .. _table_rte_flow_item_void_example:
452 .. table:: TCP, UDP or ICMP as L4
454 +-------+--------------------+
456 +=======+====================+
458 +-------+--------------------+
460 +-------+------+------+------+
461 | 2 | UDP | VOID | VOID |
462 +-------+------+------+------+
463 | 3 | VOID | TCP | VOID |
464 +-------+------+------+------+
465 | 4 | VOID | VOID | ICMP |
466 +-------+------+------+------+
468 +-------+--------------------+
473 Inverted matching, i.e. process packets that do not match the pattern.
475 - ``spec``, ``last`` and ``mask`` are ignored.
477 .. _table_rte_flow_item_invert:
481 +----------+---------+
483 +==========+=========+
484 | ``spec`` | ignored |
485 +----------+---------+
486 | ``last`` | ignored |
487 +----------+---------+
488 | ``mask`` | ignored |
489 +----------+---------+
491 Usage example, matching non-TCPv4 packets only:
493 .. _table_rte_flow_item_invert_example:
495 .. table:: Anything but TCPv4
514 Matches traffic originating from (ingress) or going to (egress) the physical
515 function of the current device.
517 If supported, should work even if the physical function is not managed by
518 the application and thus not associated with a DPDK port ID.
520 - Can be combined with any number of `Item: VF`_ to match both PF and VF
522 - ``spec``, ``last`` and ``mask`` must not be set.
524 .. _table_rte_flow_item_pf:
541 Matches traffic originating from (ingress) or going to (egress) a given
542 virtual function of the current device.
544 If supported, should work even if the virtual function is not managed by the
545 application and thus not associated with a DPDK port ID.
547 Note this pattern item does not match VF representors traffic which, as
548 separate entities, should be addressed through their own DPDK port IDs.
550 - Can be specified multiple times to match traffic addressed to several VF
552 - Can be combined with a PF item to match both PF and VF traffic.
553 - Default ``mask`` matches any VF ID.
555 .. _table_rte_flow_item_vf:
559 +----------+----------+---------------------------+
560 | Field | Subfield | Value |
561 +==========+==========+===========================+
562 | ``spec`` | ``id`` | destination VF ID |
563 +----------+----------+---------------------------+
564 | ``last`` | ``id`` | upper range value |
565 +----------+----------+---------------------------+
566 | ``mask`` | ``id`` | zeroed to match any VF ID |
567 +----------+----------+---------------------------+
572 Matches traffic originating from (ingress) or going to (egress) a physical
573 port of the underlying device.
575 The first PHY_PORT item overrides the physical port normally associated with
576 the specified DPDK input port (port_id). This item can be provided several
577 times to match additional physical ports.
579 Note that physical ports are not necessarily tied to DPDK input ports
580 (port_id) when those are not under DPDK control. Possible values are
581 specific to each device, they are not necessarily indexed from zero and may
584 As a device property, the list of allowed values as well as the value
585 associated with a port_id should be retrieved by other means.
587 - Default ``mask`` matches any port index.
589 .. _table_rte_flow_item_phy_port:
593 +----------+-----------+--------------------------------+
594 | Field | Subfield | Value |
595 +==========+===========+================================+
596 | ``spec`` | ``index`` | physical port index |
597 +----------+-----------+--------------------------------+
598 | ``last`` | ``index`` | upper range value |
599 +----------+-----------+--------------------------------+
600 | ``mask`` | ``index`` | zeroed to match any port index |
601 +----------+-----------+--------------------------------+
606 Matches traffic originating from (ingress) or going to (egress) a given DPDK
609 Normally only supported if the port ID in question is known by the
610 underlying PMD and related to the device the flow rule is created against.
612 This must not be confused with `Item: PHY_PORT`_ which refers to the
613 physical port of a device, whereas `Item: PORT_ID`_ refers to a ``struct
614 rte_eth_dev`` object on the application side (also known as "port
615 representor" depending on the kind of underlying device).
617 - Default ``mask`` matches the specified DPDK port ID.
619 .. _table_rte_flow_item_port_id:
623 +----------+----------+-----------------------------+
624 | Field | Subfield | Value |
625 +==========+==========+=============================+
626 | ``spec`` | ``id`` | DPDK port ID |
627 +----------+----------+-----------------------------+
628 | ``last`` | ``id`` | upper range value |
629 +----------+----------+-----------------------------+
630 | ``mask`` | ``id`` | zeroed to match any port ID |
631 +----------+----------+-----------------------------+
636 Matches an arbitrary integer value which was set using the ``MARK`` action in
637 a previously matched rule.
639 This item can only specified once as a match criteria as the ``MARK`` action can
640 only be specified once in a flow action.
642 Note the value of MARK field is arbitrary and application defined.
644 Depending on the underlying implementation the MARK item may be supported on
645 the physical device, with virtual groups in the PMD or not at all.
647 - Default ``mask`` matches any integer value.
649 .. _table_rte_flow_item_mark:
653 +----------+----------+---------------------------+
654 | Field | Subfield | Value |
655 +==========+==========+===========================+
656 | ``spec`` | ``id`` | integer value |
657 +----------+--------------------------------------+
658 | ``last`` | ``id`` | upper range value |
659 +----------+----------+---------------------------+
660 | ``mask`` | ``id`` | zeroed to match any value |
661 +----------+----------+---------------------------+
663 Data matching item types
664 ~~~~~~~~~~~~~~~~~~~~~~~~
666 Most of these are basically protocol header definitions with associated
667 bit-masks. They must be specified (stacked) from lowest to highest protocol
668 layer to form a matching pattern.
670 The following list is not exhaustive, new protocols will be added in the
676 Matches any protocol in place of the current layer, a single ANY may also
677 stand for several protocol layers.
679 This is usually specified as the first pattern item when looking for a
680 protocol anywhere in a packet.
682 - Default ``mask`` stands for any number of layers.
684 .. _table_rte_flow_item_any:
688 +----------+----------+--------------------------------------+
689 | Field | Subfield | Value |
690 +==========+==========+======================================+
691 | ``spec`` | ``num`` | number of layers covered |
692 +----------+----------+--------------------------------------+
693 | ``last`` | ``num`` | upper range value |
694 +----------+----------+--------------------------------------+
695 | ``mask`` | ``num`` | zeroed to cover any number of layers |
696 +----------+----------+--------------------------------------+
698 Example for VXLAN TCP payload matching regardless of outer L3 (IPv4 or IPv6)
699 and L4 (UDP) both matched by the first ANY specification, and inner L3 (IPv4
700 or IPv6) matched by the second ANY specification:
702 .. _table_rte_flow_item_any_example:
704 .. table:: TCP in VXLAN with wildcards
706 +-------+------+----------+----------+-------+
707 | Index | Item | Field | Subfield | Value |
708 +=======+======+==========+==========+=======+
710 +-------+------+----------+----------+-------+
711 | 1 | ANY | ``spec`` | ``num`` | 2 |
712 +-------+------+----------+----------+-------+
714 +-------+------------------------------------+
716 +-------+------+----------+----------+-------+
717 | 4 | ANY | ``spec`` | ``num`` | 1 |
718 +-------+------+----------+----------+-------+
720 +-------+------------------------------------+
722 +-------+------------------------------------+
727 Matches a byte string of a given length at a given offset.
729 Offset is either absolute (using the start of the packet) or relative to the
730 end of the previous matched item in the stack, in which case negative values
733 If search is enabled, offset is used as the starting point. The search area
734 can be delimited by setting limit to a nonzero value, which is the maximum
735 number of bytes after offset where the pattern may start.
737 Matching a zero-length pattern is allowed, doing so resets the relative
738 offset for subsequent items.
740 - This type does not support ranges (``last`` field).
741 - Default ``mask`` matches all fields exactly.
743 .. _table_rte_flow_item_raw:
747 +----------+--------------+-------------------------------------------------+
748 | Field | Subfield | Value |
749 +==========+==============+=================================================+
750 | ``spec`` | ``relative`` | look for pattern after the previous item |
751 | +--------------+-------------------------------------------------+
752 | | ``search`` | search pattern from offset (see also ``limit``) |
753 | +--------------+-------------------------------------------------+
754 | | ``reserved`` | reserved, must be set to zero |
755 | +--------------+-------------------------------------------------+
756 | | ``offset`` | absolute or relative offset for ``pattern`` |
757 | +--------------+-------------------------------------------------+
758 | | ``limit`` | search area limit for start of ``pattern`` |
759 | +--------------+-------------------------------------------------+
760 | | ``length`` | ``pattern`` length |
761 | +--------------+-------------------------------------------------+
762 | | ``pattern`` | byte string to look for |
763 +----------+--------------+-------------------------------------------------+
764 | ``last`` | if specified, either all 0 or with the same values as ``spec`` |
765 +----------+----------------------------------------------------------------+
766 | ``mask`` | bit-mask applied to ``spec`` values with usual behavior |
767 +----------+----------------------------------------------------------------+
769 Example pattern looking for several strings at various offsets of a UDP
770 payload, using combined RAW items:
772 .. _table_rte_flow_item_raw_example:
774 .. table:: UDP payload matching
776 +-------+------+----------+--------------+-------+
777 | Index | Item | Field | Subfield | Value |
778 +=======+======+==========+==============+=======+
780 +-------+----------------------------------------+
782 +-------+----------------------------------------+
784 +-------+------+----------+--------------+-------+
785 | 3 | RAW | ``spec`` | ``relative`` | 1 |
786 | | | +--------------+-------+
787 | | | | ``search`` | 1 |
788 | | | +--------------+-------+
789 | | | | ``offset`` | 10 |
790 | | | +--------------+-------+
791 | | | | ``limit`` | 0 |
792 | | | +--------------+-------+
793 | | | | ``length`` | 3 |
794 | | | +--------------+-------+
795 | | | | ``pattern`` | "foo" |
796 +-------+------+----------+--------------+-------+
797 | 4 | RAW | ``spec`` | ``relative`` | 1 |
798 | | | +--------------+-------+
799 | | | | ``search`` | 0 |
800 | | | +--------------+-------+
801 | | | | ``offset`` | 20 |
802 | | | +--------------+-------+
803 | | | | ``limit`` | 0 |
804 | | | +--------------+-------+
805 | | | | ``length`` | 3 |
806 | | | +--------------+-------+
807 | | | | ``pattern`` | "bar" |
808 +-------+------+----------+--------------+-------+
809 | 5 | RAW | ``spec`` | ``relative`` | 1 |
810 | | | +--------------+-------+
811 | | | | ``search`` | 0 |
812 | | | +--------------+-------+
813 | | | | ``offset`` | -29 |
814 | | | +--------------+-------+
815 | | | | ``limit`` | 0 |
816 | | | +--------------+-------+
817 | | | | ``length`` | 3 |
818 | | | +--------------+-------+
819 | | | | ``pattern`` | "baz" |
820 +-------+------+----------+--------------+-------+
822 +-------+----------------------------------------+
826 - Locate "foo" at least 10 bytes deep inside UDP payload.
827 - Locate "bar" after "foo" plus 20 bytes.
828 - Locate "baz" after "bar" minus 29 bytes.
830 Such a packet may be represented as follows (not to scale)::
833 | |<--------->| |<--------->|
835 |-----|------|-----|-----|-----|-----|-----------|-----|------|
836 | ETH | IPv4 | UDP | ... | baz | foo | ......... | bar | .... |
837 |-----|------|-----|-----|-----|-----|-----------|-----|------|
839 |<--------------------------->|
842 Note that matching subsequent pattern items would resume after "baz", not
843 "bar" since matching is always performed after the previous item of the
849 Matches an Ethernet header.
851 The ``type`` field either stands for "EtherType" or "TPID" when followed by
852 so-called layer 2.5 pattern items such as ``RTE_FLOW_ITEM_TYPE_VLAN``. In
853 the latter case, ``type`` refers to that of the outer header, with the inner
854 EtherType/TPID provided by the subsequent pattern item. This is the same
855 order as on the wire.
857 - ``dst``: destination MAC.
858 - ``src``: source MAC.
859 - ``type``: EtherType or TPID.
860 - Default ``mask`` matches destination and source addresses only.
865 Matches an 802.1Q/ad VLAN tag.
867 The corresponding standard outer EtherType (TPID) values are
868 ``ETHER_TYPE_VLAN`` or ``ETHER_TYPE_QINQ``. It can be overridden by the
869 preceding pattern item.
871 - ``tci``: tag control information.
872 - ``inner_type``: inner EtherType or TPID.
873 - Default ``mask`` matches the VID part of TCI only (lower 12 bits).
878 Matches an IPv4 header.
880 Note: IPv4 options are handled by dedicated pattern items.
882 - ``hdr``: IPv4 header definition (``rte_ip.h``).
883 - Default ``mask`` matches source and destination addresses only.
888 Matches an IPv6 header.
890 Note: IPv6 options are handled by dedicated pattern items, see `Item:
893 - ``hdr``: IPv6 header definition (``rte_ip.h``).
894 - Default ``mask`` matches source and destination addresses only.
899 Matches an ICMP header.
901 - ``hdr``: ICMP header definition (``rte_icmp.h``).
902 - Default ``mask`` matches ICMP type and code only.
907 Matches a UDP header.
909 - ``hdr``: UDP header definition (``rte_udp.h``).
910 - Default ``mask`` matches source and destination ports only.
915 Matches a TCP header.
917 - ``hdr``: TCP header definition (``rte_tcp.h``).
918 - Default ``mask`` matches source and destination ports only.
923 Matches a SCTP header.
925 - ``hdr``: SCTP header definition (``rte_sctp.h``).
926 - Default ``mask`` matches source and destination ports only.
931 Matches a VXLAN header (RFC 7348).
933 - ``flags``: normally 0x08 (I flag).
934 - ``rsvd0``: reserved, normally 0x000000.
935 - ``vni``: VXLAN network identifier.
936 - ``rsvd1``: reserved, normally 0x00.
937 - Default ``mask`` matches VNI only.
942 Matches an IEEE 802.1BR E-Tag header.
944 The corresponding standard outer EtherType (TPID) value is
945 ``ETHER_TYPE_ETAG``. It can be overridden by the preceding pattern item.
947 - ``epcp_edei_in_ecid_b``: E-Tag control information (E-TCI), E-PCP (3b),
948 E-DEI (1b), ingress E-CID base (12b).
949 - ``rsvd_grp_ecid_b``: reserved (2b), GRP (2b), E-CID base (12b).
950 - ``in_ecid_e``: ingress E-CID ext.
951 - ``ecid_e``: E-CID ext.
952 - ``inner_type``: inner EtherType or TPID.
953 - Default ``mask`` simultaneously matches GRP and E-CID base.
958 Matches a NVGRE header (RFC 7637).
960 - ``c_k_s_rsvd0_ver``: checksum (1b), undefined (1b), key bit (1b),
961 sequence number (1b), reserved 0 (9b), version (3b). This field must have
962 value 0x2000 according to RFC 7637.
963 - ``protocol``: protocol type (0x6558).
964 - ``tni``: virtual subnet ID.
965 - ``flow_id``: flow ID.
966 - Default ``mask`` matches TNI only.
971 Matches a MPLS header.
973 - ``label_tc_s_ttl``: label, TC, Bottom of Stack and TTL.
974 - Default ``mask`` matches label only.
979 Matches a GRE header.
981 - ``c_rsvd0_ver``: checksum, reserved 0 and version.
982 - ``protocol``: protocol type.
983 - Default ``mask`` matches protocol only.
988 Fuzzy pattern match, expect faster than default.
990 This is for device that support fuzzy match option. Usually a fuzzy match is
991 fast but the cost is accuracy. i.e. Signature Match only match pattern's hash
992 value, but it is possible two different patterns have the same hash value.
994 Matching accuracy level can be configured by threshold. Driver can divide the
995 range of threshold and map to different accuracy levels that device support.
997 Threshold 0 means perfect match (no fuzziness), while threshold 0xffffffff
998 means fuzziest match.
1000 .. _table_rte_flow_item_fuzzy:
1004 +----------+---------------+--------------------------------------------------+
1005 | Field | Subfield | Value |
1006 +==========+===============+==================================================+
1007 | ``spec`` | ``threshold`` | 0 as perfect match, 0xffffffff as fuzziest match |
1008 +----------+---------------+--------------------------------------------------+
1009 | ``last`` | ``threshold`` | upper range value |
1010 +----------+---------------+--------------------------------------------------+
1011 | ``mask`` | ``threshold`` | bit-mask apply to "spec" and "last" |
1012 +----------+---------------+--------------------------------------------------+
1014 Usage example, fuzzy match a TCPv4 packets:
1016 .. _table_rte_flow_item_fuzzy_example:
1018 .. table:: Fuzzy matching
1020 +-------+----------+
1022 +=======+==========+
1024 +-------+----------+
1026 +-------+----------+
1028 +-------+----------+
1030 +-------+----------+
1032 +-------+----------+
1034 Item: ``GTP``, ``GTPC``, ``GTPU``
1035 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1037 Matches a GTPv1 header.
1039 Note: GTP, GTPC and GTPU use the same structure. GTPC and GTPU item
1040 are defined for a user-friendly API when creating GTP-C and GTP-U
1043 - ``v_pt_rsv_flags``: version (3b), protocol type (1b), reserved (1b),
1044 extension header flag (1b), sequence number flag (1b), N-PDU number
1046 - ``msg_type``: message type.
1047 - ``msg_len``: message length.
1048 - ``teid``: tunnel endpoint identifier.
1049 - Default ``mask`` matches teid only.
1054 Matches an ESP header.
1056 - ``hdr``: ESP header definition (``rte_esp.h``).
1057 - Default ``mask`` matches SPI only.
1062 Matches a GENEVE header.
1064 - ``ver_opt_len_o_c_rsvd0``: version (2b), length of the options fields (6b),
1065 OAM packet (1b), critical options present (1b), reserved 0 (6b).
1066 - ``protocol``: protocol type.
1067 - ``vni``: virtual network identifier.
1068 - ``rsvd1``: reserved, normally 0x00.
1069 - Default ``mask`` matches VNI only.
1074 Matches a VXLAN-GPE header (draft-ietf-nvo3-vxlan-gpe-05).
1076 - ``flags``: normally 0x0C (I and P flags).
1077 - ``rsvd0``: reserved, normally 0x0000.
1078 - ``protocol``: protocol type.
1079 - ``vni``: VXLAN network identifier.
1080 - ``rsvd1``: reserved, normally 0x00.
1081 - Default ``mask`` matches VNI only.
1083 Item: ``ARP_ETH_IPV4``
1084 ^^^^^^^^^^^^^^^^^^^^^^
1086 Matches an ARP header for Ethernet/IPv4.
1088 - ``hdr``: hardware type, normally 1.
1089 - ``pro``: protocol type, normally 0x0800.
1090 - ``hln``: hardware address length, normally 6.
1091 - ``pln``: protocol address length, normally 4.
1092 - ``op``: opcode (1 for request, 2 for reply).
1093 - ``sha``: sender hardware address.
1094 - ``spa``: sender IPv4 address.
1095 - ``tha``: target hardware address.
1096 - ``tpa``: target IPv4 address.
1097 - Default ``mask`` matches SHA, SPA, THA and TPA.
1102 Matches the presence of any IPv6 extension header.
1104 - ``next_hdr``: next header.
1105 - Default ``mask`` matches ``next_hdr``.
1107 Normally preceded by any of:
1115 Matches any ICMPv6 header.
1117 - ``type``: ICMPv6 type.
1118 - ``code``: ICMPv6 code.
1119 - ``checksum``: ICMPv6 checksum.
1120 - Default ``mask`` matches ``type`` and ``code``.
1122 Item: ``ICMP6_ND_NS``
1123 ^^^^^^^^^^^^^^^^^^^^^
1125 Matches an ICMPv6 neighbor discovery solicitation.
1127 - ``type``: ICMPv6 type, normally 135.
1128 - ``code``: ICMPv6 code, normally 0.
1129 - ``checksum``: ICMPv6 checksum.
1130 - ``reserved``: reserved, normally 0.
1131 - ``target_addr``: target address.
1132 - Default ``mask`` matches target address only.
1134 Item: ``ICMP6_ND_NA``
1135 ^^^^^^^^^^^^^^^^^^^^^
1137 Matches an ICMPv6 neighbor discovery advertisement.
1139 - ``type``: ICMPv6 type, normally 136.
1140 - ``code``: ICMPv6 code, normally 0.
1141 - ``checksum``: ICMPv6 checksum.
1142 - ``rso_reserved``: route flag (1b), solicited flag (1b), override flag
1143 (1b), reserved (29b).
1144 - ``target_addr``: target address.
1145 - Default ``mask`` matches target address only.
1147 Item: ``ICMP6_ND_OPT``
1148 ^^^^^^^^^^^^^^^^^^^^^^
1150 Matches the presence of any ICMPv6 neighbor discovery option.
1152 - ``type``: ND option type.
1153 - ``length``: ND option length.
1154 - Default ``mask`` matches type only.
1156 Normally preceded by any of:
1158 - `Item: ICMP6_ND_NA`_
1159 - `Item: ICMP6_ND_NS`_
1160 - `Item: ICMP6_ND_OPT`_
1162 Item: ``ICMP6_ND_OPT_SLA_ETH``
1163 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1165 Matches an ICMPv6 neighbor discovery source Ethernet link-layer address
1168 - ``type``: ND option type, normally 1.
1169 - ``length``: ND option length, normally 1.
1170 - ``sla``: source Ethernet LLA.
1171 - Default ``mask`` matches source link-layer address only.
1173 Normally preceded by any of:
1175 - `Item: ICMP6_ND_NA`_
1176 - `Item: ICMP6_ND_OPT`_
1178 Item: ``ICMP6_ND_OPT_TLA_ETH``
1179 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1181 Matches an ICMPv6 neighbor discovery target Ethernet link-layer address
1184 - ``type``: ND option type, normally 2.
1185 - ``length``: ND option length, normally 1.
1186 - ``tla``: target Ethernet LLA.
1187 - Default ``mask`` matches target link-layer address only.
1189 Normally preceded by any of:
1191 - `Item: ICMP6_ND_NS`_
1192 - `Item: ICMP6_ND_OPT`_
1197 Each possible action is represented by a type. Some have associated
1198 configuration structures. Several actions combined in a list can be assigned
1199 to a flow rule and are performed in order.
1201 They fall in three categories:
1203 - Actions that modify the fate of matching traffic, for instance by dropping
1204 or assigning it a specific destination.
1206 - Actions that modify matching traffic contents or its properties. This
1207 includes adding/removing encapsulation, encryption, compression and marks.
1209 - Actions related to the flow rule itself, such as updating counters or
1210 making it non-terminating.
1212 Flow rules being terminating by default, not specifying any action of the
1213 fate kind results in undefined behavior. This applies to both ingress and
1216 PASSTHRU, when supported, makes a flow rule non-terminating.
1218 Like matching patterns, action lists are terminated by END items.
1220 Example of action that redirects packets to queue index 10:
1222 .. _table_rte_flow_action_example:
1224 .. table:: Queue action
1226 +-----------+-------+
1228 +===========+=======+
1230 +-----------+-------+
1232 Actions are performed in list order:
1234 .. _table_rte_flow_count_then_drop:
1236 .. table:: Count then drop
1250 .. _table_rte_flow_mark_count_redirect:
1252 .. table:: Mark, count then redirect
1254 +-------+--------+------------+-------+
1255 | Index | Action | Field | Value |
1256 +=======+========+============+=======+
1257 | 0 | MARK | ``mark`` | 0x2a |
1258 +-------+--------+------------+-------+
1259 | 1 | COUNT | ``shared`` | 0 |
1260 | | +------------+-------+
1262 +-------+--------+------------+-------+
1263 | 2 | QUEUE | ``queue`` | 10 |
1264 +-------+--------+------------+-------+
1266 +-------+-----------------------------+
1270 .. _table_rte_flow_redirect_queue_5:
1272 .. table:: Redirect to queue 5
1274 +-------+--------+-----------+-------+
1275 | Index | Action | Field | Value |
1276 +=======+========+===========+=======+
1278 +-------+--------+-----------+-------+
1279 | 1 | QUEUE | ``queue`` | 5 |
1280 +-------+--------+-----------+-------+
1282 +-------+----------------------------+
1284 In the above example, while DROP and QUEUE must be performed in order, both
1285 have to happen before reaching END. Only QUEUE has a visible effect.
1287 Note that such a list may be thought as ambiguous and rejected on that
1290 .. _table_rte_flow_redirect_queue_5_3:
1292 .. table:: Redirect to queues 5 and 3
1294 +-------+--------+-----------+-------+
1295 | Index | Action | Field | Value |
1296 +=======+========+===========+=======+
1297 | 0 | QUEUE | ``queue`` | 5 |
1298 +-------+--------+-----------+-------+
1300 +-------+--------+-----------+-------+
1301 | 2 | QUEUE | ``queue`` | 3 |
1302 +-------+--------+-----------+-------+
1304 +-------+----------------------------+
1306 As previously described, all actions must be taken into account. This
1307 effectively duplicates traffic to both queues. The above example also shows
1308 that VOID is ignored.
1313 Common action types are described in this section. Like pattern item types,
1314 this list is not exhaustive as new actions will be added in the future.
1319 End marker for action lists. Prevents further processing of actions, thereby
1322 - Its numeric value is 0 for convenience.
1323 - PMD support is mandatory.
1324 - No configurable properties.
1326 .. _table_rte_flow_action_end:
1339 Used as a placeholder for convenience. It is ignored and simply discarded by
1342 - PMD support is mandatory.
1343 - No configurable properties.
1345 .. _table_rte_flow_action_void:
1355 Action: ``PASSTHRU``
1356 ^^^^^^^^^^^^^^^^^^^^
1358 Leaves traffic up for additional processing by subsequent flow rules; makes
1359 a flow rule non-terminating.
1361 - No configurable properties.
1363 .. _table_rte_flow_action_passthru:
1373 Example to copy a packet to a queue and continue processing by subsequent
1376 .. _table_rte_flow_action_passthru_example:
1378 .. table:: Copy to queue 8
1380 +-------+--------+-----------+-------+
1381 | Index | Action | Field | Value |
1382 +=======+========+===========+=======+
1384 +-------+--------+-----------+-------+
1385 | 1 | QUEUE | ``queue`` | 8 |
1386 +-------+--------+-----------+-------+
1388 +-------+----------------------------+
1393 Redirects packets to a group on the current device.
1395 In a hierarchy of groups, which can be used to represent physical or logical
1396 flow group/tables on the device, this action redirects the matched flow to
1397 the specified group on that device.
1399 If a matched flow is redirected to a table which doesn't contain a matching
1400 rule for that flow then the behavior is undefined and the resulting behavior
1401 is up to the specific device. Best practice when using groups would be define
1402 a default flow rule for each group which a defines the default actions in that
1403 group so a consistent behavior is defined.
1405 Defining an action for matched flow in a group to jump to a group which is
1406 higher in the group hierarchy may not be supported by physical devices,
1407 depending on how groups are mapped to the physical devices. In the
1408 definitions of jump actions, applications should be aware that it may be
1409 possible to define flow rules which trigger an undefined behavior causing
1410 flows to loop between groups.
1412 .. _table_rte_flow_action_jump:
1416 +-----------+------------------------------+
1418 +===========+==============================+
1419 | ``group`` | Group to redirect packets to |
1420 +-----------+------------------------------+
1425 Attaches an integer value to packets and sets ``PKT_RX_FDIR`` and
1426 ``PKT_RX_FDIR_ID`` mbuf flags.
1428 This value is arbitrary and application-defined. Maximum allowed value
1429 depends on the underlying implementation. It is returned in the
1430 ``hash.fdir.hi`` mbuf field.
1432 .. _table_rte_flow_action_mark:
1436 +--------+--------------------------------------+
1438 +========+======================================+
1439 | ``id`` | integer value to return with packets |
1440 +--------+--------------------------------------+
1445 Flags packets. Similar to `Action: MARK`_ without a specific value; only
1446 sets the ``PKT_RX_FDIR`` mbuf flag.
1448 - No configurable properties.
1450 .. _table_rte_flow_action_flag:
1463 Assigns packets to a given queue index.
1465 .. _table_rte_flow_action_queue:
1469 +-----------+--------------------+
1471 +===========+====================+
1472 | ``index`` | queue index to use |
1473 +-----------+--------------------+
1480 - No configurable properties.
1482 .. _table_rte_flow_action_drop:
1495 Adds a counter action to a matched flow.
1497 If more than one count action is specified in a single flow rule, then each
1498 action must specify a unique id.
1500 Counters can be retrieved and reset through ``rte_flow_query()``, see
1501 ``struct rte_flow_query_count``.
1503 The shared flag indicates whether the counter is unique to the flow rule the
1504 action is specified with, or whether it is a shared counter.
1506 For a count action with the shared flag set, then then a global device
1507 namespace is assumed for the counter id, so that any matched flow rules using
1508 a count action with the same counter id on the same port will contribute to
1511 For ports within the same switch domain then the counter id namespace extends
1512 to all ports within that switch domain.
1514 .. _table_rte_flow_action_count:
1518 +------------+---------------------+
1520 +============+=====================+
1521 | ``shared`` | shared counter flag |
1522 +------------+---------------------+
1523 | ``id`` | counter id |
1524 +------------+---------------------+
1526 Query structure to retrieve and reset flow rule counters:
1528 .. _table_rte_flow_query_count:
1530 .. table:: COUNT query
1532 +---------------+-----+-----------------------------------+
1533 | Field | I/O | Value |
1534 +===============+=====+===================================+
1535 | ``reset`` | in | reset counter after query |
1536 +---------------+-----+-----------------------------------+
1537 | ``hits_set`` | out | ``hits`` field is set |
1538 +---------------+-----+-----------------------------------+
1539 | ``bytes_set`` | out | ``bytes`` field is set |
1540 +---------------+-----+-----------------------------------+
1541 | ``hits`` | out | number of hits for this rule |
1542 +---------------+-----+-----------------------------------+
1543 | ``bytes`` | out | number of bytes through this rule |
1544 +---------------+-----+-----------------------------------+
1549 Similar to QUEUE, except RSS is additionally performed on packets to spread
1550 them among several queues according to the provided parameters.
1552 Unlike global RSS settings used by other DPDK APIs, unsetting the ``types``
1553 field does not disable RSS in a flow rule. Doing so instead requests safe
1554 unspecified "best-effort" settings from the underlying PMD, which depending
1555 on the flow rule, may result in anything ranging from empty (single queue)
1556 to all-inclusive RSS.
1558 Note: RSS hash result is stored in the ``hash.rss`` mbuf field which
1559 overlaps ``hash.fdir.lo``. Since `Action: MARK`_ sets the ``hash.fdir.hi``
1560 field only, both can be requested simultaneously.
1562 Also, regarding packet encapsulation ``level``:
1564 - ``0`` requests the default behavior. Depending on the packet type, it can
1565 mean outermost, innermost, anything in between or even no RSS.
1567 It basically stands for the innermost encapsulation level RSS can be
1568 performed on according to PMD and device capabilities.
1570 - ``1`` requests RSS to be performed on the outermost packet encapsulation
1573 - ``2`` and subsequent values request RSS to be performed on the specified
1574 inner packet encapsulation level, from outermost to innermost (lower to
1577 Values other than ``0`` are not necessarily supported.
1579 Requesting a specific RSS level on unrecognized traffic results in undefined
1580 behavior. For predictable results, it is recommended to make the flow rule
1581 pattern match packet headers up to the requested encapsulation level so that
1582 only matching traffic goes through.
1584 .. _table_rte_flow_action_rss:
1588 +---------------+---------------------------------------------+
1590 +===============+=============================================+
1591 | ``func`` | RSS hash function to apply |
1592 +---------------+---------------------------------------------+
1593 | ``level`` | encapsulation level for ``types`` |
1594 +---------------+---------------------------------------------+
1595 | ``types`` | specific RSS hash types (see ``ETH_RSS_*``) |
1596 +---------------+---------------------------------------------+
1597 | ``key_len`` | hash key length in bytes |
1598 +---------------+---------------------------------------------+
1599 | ``queue_num`` | number of entries in ``queue`` |
1600 +---------------+---------------------------------------------+
1601 | ``key`` | hash key |
1602 +---------------+---------------------------------------------+
1603 | ``queue`` | queue indices to use |
1604 +---------------+---------------------------------------------+
1609 Directs matching traffic to the physical function (PF) of the current
1614 - No configurable properties.
1616 .. _table_rte_flow_action_pf:
1629 Directs matching traffic to a given virtual function of the current device.
1631 Packets matched by a VF pattern item can be redirected to their original VF
1632 ID instead of the specified one. This parameter may not be available and is
1633 not guaranteed to work properly if the VF part is matched by a prior flow
1634 rule or if packets are not addressed to a VF in the first place.
1638 .. _table_rte_flow_action_vf:
1642 +--------------+--------------------------------+
1644 +==============+================================+
1645 | ``original`` | use original VF ID if possible |
1646 +--------------+--------------------------------+
1648 +--------------+--------------------------------+
1650 Action: ``PHY_PORT``
1651 ^^^^^^^^^^^^^^^^^^^^
1653 Directs matching traffic to a given physical port index of the underlying
1656 See `Item: PHY_PORT`_.
1658 .. _table_rte_flow_action_phy_port:
1662 +--------------+-------------------------------------+
1664 +==============+=====================================+
1665 | ``original`` | use original port index if possible |
1666 +--------------+-------------------------------------+
1667 | ``index`` | physical port index |
1668 +--------------+-------------------------------------+
1672 Directs matching traffic to a given DPDK port ID.
1674 See `Item: PORT_ID`_.
1676 .. _table_rte_flow_action_port_id:
1680 +--------------+---------------------------------------+
1682 +==============+=======================================+
1683 | ``original`` | use original DPDK port ID if possible |
1684 +--------------+---------------------------------------+
1685 | ``id`` | DPDK port ID |
1686 +--------------+---------------------------------------+
1691 Applies a stage of metering and policing.
1693 The metering and policing (MTR) object has to be first created using the
1694 rte_mtr_create() API function. The ID of the MTR object is specified as
1695 action parameter. More than one flow can use the same MTR object through
1696 the meter action. The MTR object can be further updated or queried using
1699 .. _table_rte_flow_action_meter:
1703 +--------------+---------------+
1705 +==============+===============+
1706 | ``mtr_id`` | MTR object ID |
1707 +--------------+---------------+
1709 Action: ``SECURITY``
1710 ^^^^^^^^^^^^^^^^^^^^
1712 Perform the security action on flows matched by the pattern items
1713 according to the configuration of the security session.
1715 This action modifies the payload of matched flows. For INLINE_CRYPTO, the
1716 security protocol headers and IV are fully provided by the application as
1717 specified in the flow pattern. The payload of matching packets is
1718 encrypted on egress, and decrypted and authenticated on ingress.
1719 For INLINE_PROTOCOL, the security protocol is fully offloaded to HW,
1720 providing full encapsulation and decapsulation of packets in security
1721 protocols. The flow pattern specifies both the outer security header fields
1722 and the inner packet fields. The security session specified in the action
1723 must match the pattern parameters.
1725 The security session specified in the action must be created on the same
1726 port as the flow action that is being specified.
1728 The ingress/egress flow attribute should match that specified in the
1729 security session if the security session supports the definition of the
1732 Multiple flows can be configured to use the same security session.
1734 .. _table_rte_flow_action_security:
1738 +----------------------+--------------------------------------+
1740 +======================+======================================+
1741 | ``security_session`` | security session to apply |
1742 +----------------------+--------------------------------------+
1744 The following is an example of configuring IPsec inline using the
1745 INLINE_CRYPTO security session:
1747 The encryption algorithm, keys and salt are part of the opaque
1748 ``rte_security_session``. The SA is identified according to the IP and ESP
1749 fields in the pattern items.
1751 .. _table_rte_flow_item_esp_inline_example:
1753 .. table:: IPsec inline crypto flow pattern items.
1755 +-------+----------+
1757 +=======+==========+
1759 +-------+----------+
1761 +-------+----------+
1763 +-------+----------+
1765 +-------+----------+
1767 .. _table_rte_flow_action_esp_inline_example:
1769 .. table:: IPsec inline flow actions.
1771 +-------+----------+
1773 +=======+==========+
1775 +-------+----------+
1777 +-------+----------+
1779 Action: ``OF_SET_MPLS_TTL``
1780 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1782 Implements ``OFPAT_SET_MPLS_TTL`` ("MPLS TTL") as defined by the `OpenFlow
1783 Switch Specification`_.
1785 .. _table_rte_flow_action_of_set_mpls_ttl:
1787 .. table:: OF_SET_MPLS_TTL
1789 +--------------+----------+
1791 +==============+==========+
1792 | ``mpls_ttl`` | MPLS TTL |
1793 +--------------+----------+
1795 Action: ``OF_DEC_MPLS_TTL``
1796 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1798 Implements ``OFPAT_DEC_MPLS_TTL`` ("decrement MPLS TTL") as defined by the
1799 `OpenFlow Switch Specification`_.
1801 .. _table_rte_flow_action_of_dec_mpls_ttl:
1803 .. table:: OF_DEC_MPLS_TTL
1811 Action: ``OF_SET_NW_TTL``
1812 ^^^^^^^^^^^^^^^^^^^^^^^^^
1814 Implements ``OFPAT_SET_NW_TTL`` ("IP TTL") as defined by the `OpenFlow
1815 Switch Specification`_.
1817 .. _table_rte_flow_action_of_set_nw_ttl:
1819 .. table:: OF_SET_NW_TTL
1821 +------------+--------+
1823 +============+========+
1824 | ``nw_ttl`` | IP TTL |
1825 +------------+--------+
1827 Action: ``OF_DEC_NW_TTL``
1828 ^^^^^^^^^^^^^^^^^^^^^^^^^
1830 Implements ``OFPAT_DEC_NW_TTL`` ("decrement IP TTL") as defined by the
1831 `OpenFlow Switch Specification`_.
1833 .. _table_rte_flow_action_of_dec_nw_ttl:
1835 .. table:: OF_DEC_NW_TTL
1843 Action: ``OF_COPY_TTL_OUT``
1844 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1846 Implements ``OFPAT_COPY_TTL_OUT`` ("copy TTL "outwards" -- from
1847 next-to-outermost to outermost") as defined by the `OpenFlow Switch
1850 .. _table_rte_flow_action_of_copy_ttl_out:
1852 .. table:: OF_COPY_TTL_OUT
1860 Action: ``OF_COPY_TTL_IN``
1861 ^^^^^^^^^^^^^^^^^^^^^^^^^^
1863 Implements ``OFPAT_COPY_TTL_IN`` ("copy TTL "inwards" -- from outermost to
1864 next-to-outermost") as defined by the `OpenFlow Switch Specification`_.
1866 .. _table_rte_flow_action_of_copy_ttl_in:
1868 .. table:: OF_COPY_TTL_IN
1876 Action: ``OF_POP_VLAN``
1877 ^^^^^^^^^^^^^^^^^^^^^^^
1879 Implements ``OFPAT_POP_VLAN`` ("pop the outer VLAN tag") as defined
1880 by the `OpenFlow Switch Specification`_.
1882 .. _table_rte_flow_action_of_pop_vlan:
1884 .. table:: OF_POP_VLAN
1892 Action: ``OF_PUSH_VLAN``
1893 ^^^^^^^^^^^^^^^^^^^^^^^^
1895 Implements ``OFPAT_PUSH_VLAN`` ("push a new VLAN tag") as defined by the
1896 `OpenFlow Switch Specification`_.
1898 .. _table_rte_flow_action_of_push_vlan:
1900 .. table:: OF_PUSH_VLAN
1902 +---------------+-----------+
1904 +===============+===========+
1905 | ``ethertype`` | EtherType |
1906 +---------------+-----------+
1908 Action: ``OF_SET_VLAN_VID``
1909 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1911 Implements ``OFPAT_SET_VLAN_VID`` ("set the 802.1q VLAN id") as defined by
1912 the `OpenFlow Switch Specification`_.
1914 .. _table_rte_flow_action_of_set_vlan_vid:
1916 .. table:: OF_SET_VLAN_VID
1918 +--------------+---------+
1920 +==============+=========+
1921 | ``vlan_vid`` | VLAN id |
1922 +--------------+---------+
1924 Action: ``OF_SET_VLAN_PCP``
1925 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
1927 Implements ``OFPAT_SET_LAN_PCP`` ("set the 802.1q priority") as defined by
1928 the `OpenFlow Switch Specification`_.
1930 .. _table_rte_flow_action_of_set_vlan_pcp:
1932 .. table:: OF_SET_VLAN_PCP
1934 +--------------+---------------+
1936 +==============+===============+
1937 | ``vlan_pcp`` | VLAN priority |
1938 +--------------+---------------+
1940 Action: ``OF_POP_MPLS``
1941 ^^^^^^^^^^^^^^^^^^^^^^^
1943 Implements ``OFPAT_POP_MPLS`` ("pop the outer MPLS tag") as defined by the
1944 `OpenFlow Switch Specification`_.
1946 .. _table_rte_flow_action_of_pop_mpls:
1948 .. table:: OF_POP_MPLS
1950 +---------------+-----------+
1952 +===============+===========+
1953 | ``ethertype`` | EtherType |
1954 +---------------+-----------+
1956 Action: ``OF_PUSH_MPLS``
1957 ^^^^^^^^^^^^^^^^^^^^^^^^
1959 Implements ``OFPAT_PUSH_MPLS`` ("push a new MPLS tag") as defined by the
1960 `OpenFlow Switch Specification`_.
1962 .. _table_rte_flow_action_of_push_mpls:
1964 .. table:: OF_PUSH_MPLS
1966 +---------------+-----------+
1968 +===============+===========+
1969 | ``ethertype`` | EtherType |
1970 +---------------+-----------+
1972 Action: ``VXLAN_ENCAP``
1973 ^^^^^^^^^^^^^^^^^^^^^^^
1975 Performs a VXLAN encapsulation action by encapsulating the matched flow in the
1976 VXLAN tunnel as defined in the``rte_flow_action_vxlan_encap`` flow items
1979 This action modifies the payload of matched flows. The flow definition specified
1980 in the ``rte_flow_action_tunnel_encap`` action structure must define a valid
1981 VLXAN network overlay which conforms with RFC 7348 (Virtual eXtensible Local
1982 Area Network (VXLAN): A Framework for Overlaying Virtualized Layer 2 Networks
1983 over Layer 3 Networks). The pattern must be terminated with the
1984 RTE_FLOW_ITEM_TYPE_END item type.
1986 .. _table_rte_flow_action_vxlan_encap:
1988 .. table:: VXLAN_ENCAP
1990 +----------------+-------------------------------------+
1992 +================+=====================================+
1993 | ``definition`` | Tunnel end-point overlay definition |
1994 +----------------+-------------------------------------+
1996 .. _table_rte_flow_action_vxlan_encap_example:
1998 .. table:: IPv4 VxLAN flow pattern example.
2000 +-------+----------+
2002 +=======+==========+
2004 +-------+----------+
2006 +-------+----------+
2008 +-------+----------+
2010 +-------+----------+
2012 +-------+----------+
2014 Action: ``VXLAN_DECAP``
2015 ^^^^^^^^^^^^^^^^^^^^^^^
2017 Performs a decapsulation action by stripping all headers of the VXLAN tunnel
2018 network overlay from the matched flow.
2020 The flow items pattern defined for the flow rule with which a ``VXLAN_DECAP``
2021 action is specified, must define a valid VXLAN tunnel as per RFC7348. If the
2022 flow pattern does not specify a valid VXLAN tunnel then a
2023 RTE_FLOW_ERROR_TYPE_ACTION error should be returned.
2025 This action modifies the payload of matched flows.
2027 Action: ``NVGRE_ENCAP``
2028 ^^^^^^^^^^^^^^^^^^^^^^^
2030 Performs a NVGRE encapsulation action by encapsulating the matched flow in the
2031 NVGRE tunnel as defined in the``rte_flow_action_tunnel_encap`` flow item
2034 This action modifies the payload of matched flows. The flow definition specified
2035 in the ``rte_flow_action_tunnel_encap`` action structure must defined a valid
2036 NVGRE network overlay which conforms with RFC 7637 (NVGRE: Network
2037 Virtualization Using Generic Routing Encapsulation). The pattern must be
2038 terminated with the RTE_FLOW_ITEM_TYPE_END item type.
2040 .. _table_rte_flow_action_nvgre_encap:
2042 .. table:: NVGRE_ENCAP
2044 +----------------+-------------------------------------+
2046 +================+=====================================+
2047 | ``definition`` | NVGRE end-point overlay definition |
2048 +----------------+-------------------------------------+
2050 .. _table_rte_flow_action_nvgre_encap_example:
2052 .. table:: IPv4 NVGRE flow pattern example.
2054 +-------+----------+
2056 +=======+==========+
2058 +-------+----------+
2060 +-------+----------+
2062 +-------+----------+
2064 +-------+----------+
2066 Action: ``NVGRE_DECAP``
2067 ^^^^^^^^^^^^^^^^^^^^^^^
2069 Performs a decapsulation action by stripping all headers of the NVGRE tunnel
2070 network overlay from the matched flow.
2072 The flow items pattern defined for the flow rule with which a ``NVGRE_DECAP``
2073 action is specified, must define a valid NVGRE tunnel as per RFC7637. If the
2074 flow pattern does not specify a valid NVGRE tunnel then a
2075 RTE_FLOW_ERROR_TYPE_ACTION error should be returned.
2077 This action modifies the payload of matched flows.
2079 Action: ``SET_IPV4_SRC``
2080 ^^^^^^^^^^^^^^^^^^^^^^^^
2082 Set a new IPv4 source address in the outermost IPv4 header.
2084 It must be used with a valid RTE_FLOW_ITEM_TYPE_IPV4 flow pattern item.
2085 Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2087 .. _table_rte_flow_action_set_ipv4_src:
2089 .. table:: SET_IPV4_SRC
2091 +-----------------------------------------+
2093 +===============+=========================+
2094 | ``ipv4_addr`` | new IPv4 source address |
2095 +---------------+-------------------------+
2097 Action: ``SET_IPV4_DST``
2098 ^^^^^^^^^^^^^^^^^^^^^^^^
2100 Set a new IPv4 destination address in the outermost IPv4 header.
2102 It must be used with a valid RTE_FLOW_ITEM_TYPE_IPV4 flow pattern item.
2103 Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2105 .. _table_rte_flow_action_set_ipv4_dst:
2107 .. table:: SET_IPV4_DST
2109 +---------------+------------------------------+
2111 +===============+==============================+
2112 | ``ipv4_addr`` | new IPv4 destination address |
2113 +---------------+------------------------------+
2115 Action: ``SET_IPV6_SRC``
2116 ^^^^^^^^^^^^^^^^^^^^^^^^
2118 Set a new IPv6 source address in the outermost IPv6 header.
2120 It must be used with a valid RTE_FLOW_ITEM_TYPE_IPV6 flow pattern item.
2121 Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2123 .. _table_rte_flow_action_set_ipv6_src:
2125 .. table:: SET_IPV6_SRC
2127 +---------------+-------------------------+
2129 +===============+=========================+
2130 | ``ipv6_addr`` | new IPv6 source address |
2131 +---------------+-------------------------+
2133 Action: ``SET_IPV6_DST``
2134 ^^^^^^^^^^^^^^^^^^^^^^^^
2136 Set a new IPv6 destination address in the outermost IPv6 header.
2138 It must be used with a valid RTE_FLOW_ITEM_TYPE_IPV6 flow pattern item.
2139 Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2141 .. _table_rte_flow_action_set_ipv6_dst:
2143 .. table:: SET_IPV6_DST
2145 +---------------+------------------------------+
2147 +===============+==============================+
2148 | ``ipv6_addr`` | new IPv6 destination address |
2149 +---------------+------------------------------+
2151 Action: ``SET_TP_SRC``
2152 ^^^^^^^^^^^^^^^^^^^^^^^^^
2154 Set a new source port number in the outermost TCP/UDP header.
2156 It must be used with a valid RTE_FLOW_ITEM_TYPE_TCP or RTE_FLOW_ITEM_TYPE_UDP
2157 flow pattern item. Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2159 .. _table_rte_flow_action_set_tp_src:
2161 .. table:: SET_TP_SRC
2163 +----------+-------------------------+
2165 +==========+=========================+
2166 | ``port`` | new TCP/UDP source port |
2167 +---------------+--------------------+
2169 Action: ``SET_TP_DST``
2170 ^^^^^^^^^^^^^^^^^^^^^^^^^
2172 Set a new destination port number in the outermost TCP/UDP header.
2174 It must be used with a valid RTE_FLOW_ITEM_TYPE_TCP or RTE_FLOW_ITEM_TYPE_UDP
2175 flow pattern item. Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2177 .. _table_rte_flow_action_set_tp_dst:
2179 .. table:: SET_TP_DST
2181 +----------+------------------------------+
2183 +==========+==============================+
2184 | ``port`` | new TCP/UDP destination port |
2185 +---------------+-------------------------+
2187 Action: ``MAC_SWAP``
2188 ^^^^^^^^^^^^^^^^^^^^^^^^^
2190 Swap the source and destination MAC addresses in the outermost Ethernet
2193 It must be used with a valid RTE_FLOW_ITEM_TYPE_ETH flow pattern item.
2194 Otherwise, RTE_FLOW_ERROR_TYPE_ACTION error will be returned.
2196 .. _table_rte_flow_action_mac_swap:
2209 All specified pattern items (``enum rte_flow_item_type``) and actions
2210 (``enum rte_flow_action_type``) use positive identifiers.
2212 The negative space is reserved for dynamic types generated by PMDs during
2213 run-time. PMDs may encounter them as a result but must not accept negative
2214 identifiers they are not aware of.
2216 A method to generate them remains to be defined.
2221 Pattern item types will be added as new protocols are implemented.
2223 Variable headers support through dedicated pattern items, for example in
2224 order to match specific IPv4 options and IPv6 extension headers would be
2225 stacked after IPv4/IPv6 items.
2227 Other action types are planned but are not defined yet. These include the
2228 ability to alter packet data in several ways, such as performing
2229 encapsulation/decapsulation of tunnel headers.
2234 A rather simple API with few functions is provided to fully manage flow
2237 Each created flow rule is associated with an opaque, PMD-specific handle
2238 pointer. The application is responsible for keeping it until the rule is
2241 Flows rules are represented by ``struct rte_flow`` objects.
2246 Given that expressing a definite set of device capabilities is not
2247 practical, a dedicated function is provided to check if a flow rule is
2248 supported and can be created.
2253 rte_flow_validate(uint16_t port_id,
2254 const struct rte_flow_attr *attr,
2255 const struct rte_flow_item pattern[],
2256 const struct rte_flow_action actions[],
2257 struct rte_flow_error *error);
2259 The flow rule is validated for correctness and whether it could be accepted
2260 by the device given sufficient resources. The rule is checked against the
2261 current device mode and queue configuration. The flow rule may also
2262 optionally be validated against existing flow rules and device resources.
2263 This function has no effect on the target device.
2265 The returned value is guaranteed to remain valid only as long as no
2266 successful calls to ``rte_flow_create()`` or ``rte_flow_destroy()`` are made
2267 in the meantime and no device parameter affecting flow rules in any way are
2268 modified, due to possible collisions or resource limitations (although in
2269 such cases ``EINVAL`` should not be returned).
2273 - ``port_id``: port identifier of Ethernet device.
2274 - ``attr``: flow rule attributes.
2275 - ``pattern``: pattern specification (list terminated by the END pattern
2277 - ``actions``: associated actions (list terminated by the END action).
2278 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2279 this structure in case of error only.
2283 - 0 if flow rule is valid and can be created. A negative errno value
2284 otherwise (``rte_errno`` is also set), the following errors are defined.
2285 - ``-ENOSYS``: underlying device does not support this functionality.
2286 - ``-EINVAL``: unknown or invalid rule specification.
2287 - ``-ENOTSUP``: valid but unsupported rule specification (e.g. partial
2288 bit-masks are unsupported).
2289 - ``EEXIST``: collision with an existing rule. Only returned if device
2290 supports flow rule collision checking and there was a flow rule
2291 collision. Not receiving this return code is no guarantee that creating
2292 the rule will not fail due to a collision.
2293 - ``ENOMEM``: not enough memory to execute the function, or if the device
2294 supports resource validation, resource limitation on the device.
2295 - ``-EBUSY``: action cannot be performed due to busy device resources, may
2296 succeed if the affected queues or even the entire port are in a stopped
2297 state (see ``rte_eth_dev_rx_queue_stop()`` and ``rte_eth_dev_stop()``).
2302 Creating a flow rule is similar to validating one, except the rule is
2303 actually created and a handle returned.
2308 rte_flow_create(uint16_t port_id,
2309 const struct rte_flow_attr *attr,
2310 const struct rte_flow_item pattern[],
2311 const struct rte_flow_action *actions[],
2312 struct rte_flow_error *error);
2316 - ``port_id``: port identifier of Ethernet device.
2317 - ``attr``: flow rule attributes.
2318 - ``pattern``: pattern specification (list terminated by the END pattern
2320 - ``actions``: associated actions (list terminated by the END action).
2321 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2322 this structure in case of error only.
2326 A valid handle in case of success, NULL otherwise and ``rte_errno`` is set
2327 to the positive version of one of the error codes defined for
2328 ``rte_flow_validate()``.
2333 Flow rules destruction is not automatic, and a queue or a port should not be
2334 released if any are still attached to them. Applications must take care of
2335 performing this step before releasing resources.
2340 rte_flow_destroy(uint16_t port_id,
2341 struct rte_flow *flow,
2342 struct rte_flow_error *error);
2345 Failure to destroy a flow rule handle may occur when other flow rules depend
2346 on it, and destroying it would result in an inconsistent state.
2348 This function is only guaranteed to succeed if handles are destroyed in
2349 reverse order of their creation.
2353 - ``port_id``: port identifier of Ethernet device.
2354 - ``flow``: flow rule handle to destroy.
2355 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2356 this structure in case of error only.
2360 - 0 on success, a negative errno value otherwise and ``rte_errno`` is set.
2365 Convenience function to destroy all flow rule handles associated with a
2366 port. They are released as with successive calls to ``rte_flow_destroy()``.
2371 rte_flow_flush(uint16_t port_id,
2372 struct rte_flow_error *error);
2374 In the unlikely event of failure, handles are still considered destroyed and
2375 no longer valid but the port must be assumed to be in an inconsistent state.
2379 - ``port_id``: port identifier of Ethernet device.
2380 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2381 this structure in case of error only.
2385 - 0 on success, a negative errno value otherwise and ``rte_errno`` is set.
2390 Query an existing flow rule.
2392 This function allows retrieving flow-specific data such as counters. Data
2393 is gathered by special actions which must be present in the flow rule
2399 rte_flow_query(uint16_t port_id,
2400 struct rte_flow *flow,
2401 const struct rte_flow_action *action,
2403 struct rte_flow_error *error);
2407 - ``port_id``: port identifier of Ethernet device.
2408 - ``flow``: flow rule handle to query.
2409 - ``action``: action to query, this must match prototype from flow rule.
2410 - ``data``: pointer to storage for the associated query data type.
2411 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2412 this structure in case of error only.
2416 - 0 on success, a negative errno value otherwise and ``rte_errno`` is set.
2421 The general expectation for ingress traffic is that flow rules process it
2422 first; the remaining unmatched or pass-through traffic usually ends up in a
2423 queue (with or without RSS, locally or in some sub-device instance)
2424 depending on the global configuration settings of a port.
2426 While fine from a compatibility standpoint, this approach makes drivers more
2427 complex as they have to check for possible side effects outside of this API
2428 when creating or destroying flow rules. It results in a more limited set of
2429 available rule types due to the way device resources are assigned (e.g. no
2430 support for the RSS action even on capable hardware).
2432 Given that nonspecific traffic can be handled by flow rules as well,
2433 isolated mode is a means for applications to tell a driver that ingress on
2434 the underlying port must be injected from the defined flow rules only; that
2435 no default traffic is expected outside those rules.
2437 This has the following benefits:
2439 - Applications get finer-grained control over the kind of traffic they want
2440 to receive (no traffic by default).
2442 - More importantly they control at what point nonspecific traffic is handled
2443 relative to other flow rules, by adjusting priority levels.
2445 - Drivers can assign more hardware resources to flow rules and expand the
2446 set of supported rule types.
2448 Because toggling isolated mode may cause profound changes to the ingress
2449 processing path of a driver, it may not be possible to leave it once
2450 entered. Likewise, existing flow rules or global configuration settings may
2451 prevent a driver from entering isolated mode.
2453 Applications relying on this mode are therefore encouraged to toggle it as
2454 soon as possible after device initialization, ideally before the first call
2455 to ``rte_eth_dev_configure()`` to avoid possible failures due to conflicting
2458 Once effective, the following functionality has no effect on the underlying
2459 port and may return errors such as ``ENOTSUP`` ("not supported"):
2461 - Toggling promiscuous mode.
2462 - Toggling allmulticast mode.
2463 - Configuring MAC addresses.
2464 - Configuring multicast addresses.
2465 - Configuring VLAN filters.
2466 - Configuring Rx filters through the legacy API (e.g. FDIR).
2467 - Configuring global RSS settings.
2472 rte_flow_isolate(uint16_t port_id, int set, struct rte_flow_error *error);
2476 - ``port_id``: port identifier of Ethernet device.
2477 - ``set``: nonzero to enter isolated mode, attempt to leave it otherwise.
2478 - ``error``: perform verbose error reporting if not NULL. PMDs initialize
2479 this structure in case of error only.
2483 - 0 on success, a negative errno value otherwise and ``rte_errno`` is set.
2485 Verbose error reporting
2486 -----------------------
2488 The defined *errno* values may not be accurate enough for users or
2489 application developers who want to investigate issues related to flow rules
2490 management. A dedicated error object is defined for this purpose:
2494 enum rte_flow_error_type {
2495 RTE_FLOW_ERROR_TYPE_NONE, /**< No error. */
2496 RTE_FLOW_ERROR_TYPE_UNSPECIFIED, /**< Cause unspecified. */
2497 RTE_FLOW_ERROR_TYPE_HANDLE, /**< Flow rule (handle). */
2498 RTE_FLOW_ERROR_TYPE_ATTR_GROUP, /**< Group field. */
2499 RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, /**< Priority field. */
2500 RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, /**< Ingress field. */
2501 RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, /**< Egress field. */
2502 RTE_FLOW_ERROR_TYPE_ATTR, /**< Attributes structure. */
2503 RTE_FLOW_ERROR_TYPE_ITEM_NUM, /**< Pattern length. */
2504 RTE_FLOW_ERROR_TYPE_ITEM, /**< Specific pattern item. */
2505 RTE_FLOW_ERROR_TYPE_ACTION_NUM, /**< Number of actions. */
2506 RTE_FLOW_ERROR_TYPE_ACTION, /**< Specific action. */
2509 struct rte_flow_error {
2510 enum rte_flow_error_type type; /**< Cause field and error types. */
2511 const void *cause; /**< Object responsible for the error. */
2512 const char *message; /**< Human-readable error message. */
2515 Error type ``RTE_FLOW_ERROR_TYPE_NONE`` stands for no error, in which case
2516 remaining fields can be ignored. Other error types describe the type of the
2517 object pointed by ``cause``.
2519 If non-NULL, ``cause`` points to the object responsible for the error. For a
2520 flow rule, this may be a pattern item or an individual action.
2522 If non-NULL, ``message`` provides a human-readable error message.
2524 This object is normally allocated by applications and set by PMDs in case of
2525 error, the message points to a constant string which does not need to be
2526 freed by the application, however its pointer can be considered valid only
2527 as long as its associated DPDK port remains configured. Closing the
2528 underlying device or unloading the PMD invalidates it.
2539 rte_flow_error_set(struct rte_flow_error *error,
2541 enum rte_flow_error_type type,
2543 const char *message);
2545 This function initializes ``error`` (if non-NULL) with the provided
2546 parameters and sets ``rte_errno`` to ``code``. A negative error ``code`` is
2555 rte_flow_conv(enum rte_flow_conv_op op,
2559 struct rte_flow_error *error);
2561 Convert ``src`` to ``dst`` according to operation ``op``. Possible
2564 - Attributes, pattern item or action duplication.
2565 - Duplication of an entire pattern or list of actions.
2566 - Duplication of a complete flow rule description.
2567 - Pattern item or action name retrieval.
2572 - DPDK does not keep track of flow rules definitions or flow rule objects
2573 automatically. Applications may keep track of the former and must keep
2574 track of the latter. PMDs may also do it for internal needs, however this
2575 must not be relied on by applications.
2577 - Flow rules are not maintained between successive port initializations. An
2578 application exiting without releasing them and restarting must re-create
2581 - API operations are synchronous and blocking (``EAGAIN`` cannot be
2584 - There is no provision for reentrancy/multi-thread safety, although nothing
2585 should prevent different devices from being configured at the same
2586 time. PMDs may protect their control path functions accordingly.
2588 - Stopping the data path (TX/RX) should not be necessary when managing flow
2589 rules. If this cannot be achieved naturally or with workarounds (such as
2590 temporarily replacing the burst function pointers), an appropriate error
2591 code must be returned (``EBUSY``).
2593 - PMDs, not applications, are responsible for maintaining flow rules
2594 configuration when stopping and restarting a port or performing other
2595 actions which may affect them. They can only be destroyed explicitly by
2598 For devices exposing multiple ports sharing global settings affected by flow
2601 - All ports under DPDK control must behave consistently, PMDs are
2602 responsible for making sure that existing flow rules on a port are not
2603 affected by other ports.
2605 - Ports not under DPDK control (unaffected or handled by other applications)
2606 are user's responsibility. They may affect existing flow rules and cause
2607 undefined behavior. PMDs aware of this may prevent flow rules creation
2608 altogether in such cases.
2613 The PMD interface is defined in ``rte_flow_driver.h``. It is not subject to
2614 API/ABI versioning constraints as it is not exposed to applications and may
2615 evolve independently.
2617 It is currently implemented on top of the legacy filtering framework through
2618 filter type *RTE_ETH_FILTER_GENERIC* that accepts the single operation
2619 *RTE_ETH_FILTER_GET* to return PMD-specific *rte_flow* callbacks wrapped
2620 inside ``struct rte_flow_ops``.
2622 This overhead is temporarily necessary in order to keep compatibility with
2623 the legacy filtering framework, which should eventually disappear.
2625 - PMD callbacks implement exactly the interface described in `Rules
2626 management`_, except for the port ID argument which has already been
2627 converted to a pointer to the underlying ``struct rte_eth_dev``.
2629 - Public API functions do not process flow rules definitions at all before
2630 calling PMD functions (no basic error checking, no validation
2631 whatsoever). They only make sure these callbacks are non-NULL or return
2632 the ``ENOSYS`` (function not supported) error.
2634 This interface additionally defines the following helper function:
2636 - ``rte_flow_ops_get()``: get generic flow operations structure from a
2639 More will be added over time.
2641 Device compatibility
2642 --------------------
2644 No known implementation supports all the described features.
2646 Unsupported features or combinations are not expected to be fully emulated
2647 in software by PMDs for performance reasons. Partially supported features
2648 may be completed in software as long as hardware performs most of the work
2649 (such as queue redirection and packet recognition).
2651 However PMDs are expected to do their best to satisfy application requests
2652 by working around hardware limitations as long as doing so does not affect
2653 the behavior of existing flow rules.
2655 The following sections provide a few examples of such cases and describe how
2656 PMDs should handle them, they are based on limitations built into the
2662 Each flow rule comes with its own, per-layer bit-masks, while hardware may
2663 support only a single, device-wide bit-mask for a given layer type, so that
2664 two IPv4 rules cannot use different bit-masks.
2666 The expected behavior in this case is that PMDs automatically configure
2667 global bit-masks according to the needs of the first flow rule created.
2669 Subsequent rules are allowed only if their bit-masks match those, the
2670 ``EEXIST`` error code should be returned otherwise.
2672 Unsupported layer types
2673 ~~~~~~~~~~~~~~~~~~~~~~~
2675 Many protocols can be simulated by crafting patterns with the `Item: RAW`_
2678 PMDs can rely on this capability to simulate support for protocols with
2679 headers not directly recognized by hardware.
2681 ``ANY`` pattern item
2682 ~~~~~~~~~~~~~~~~~~~~
2684 This pattern item stands for anything, which can be difficult to translate
2685 to something hardware would understand, particularly if followed by more
2688 Consider the following pattern:
2690 .. _table_rte_flow_unsupported_any:
2692 .. table:: Pattern with ANY as L3
2694 +-------+-----------------------+
2696 +=======+=======================+
2698 +-------+-----+---------+-------+
2699 | 1 | ANY | ``num`` | ``1`` |
2700 +-------+-----+---------+-------+
2702 +-------+-----------------------+
2704 +-------+-----------------------+
2706 Knowing that TCP does not make sense with something other than IPv4 and IPv6
2707 as L3, such a pattern may be translated to two flow rules instead:
2709 .. _table_rte_flow_unsupported_any_ipv4:
2711 .. table:: ANY replaced with IPV4
2713 +-------+--------------------+
2715 +=======+====================+
2717 +-------+--------------------+
2718 | 1 | IPV4 (zeroed mask) |
2719 +-------+--------------------+
2721 +-------+--------------------+
2723 +-------+--------------------+
2727 .. _table_rte_flow_unsupported_any_ipv6:
2729 .. table:: ANY replaced with IPV6
2731 +-------+--------------------+
2733 +=======+====================+
2735 +-------+--------------------+
2736 | 1 | IPV6 (zeroed mask) |
2737 +-------+--------------------+
2739 +-------+--------------------+
2741 +-------+--------------------+
2743 Note that as soon as a ANY rule covers several layers, this approach may
2744 yield a large number of hidden flow rules. It is thus suggested to only
2745 support the most common scenarios (anything as L2 and/or L3).
2750 - When combined with `Action: QUEUE`_, packet counting (`Action: COUNT`_)
2751 and tagging (`Action: MARK`_ or `Action: FLAG`_) may be implemented in
2752 software as long as the target queue is used by a single rule.
2754 - When a single target queue is provided, `Action: RSS`_ can also be
2755 implemented through `Action: QUEUE`_.
2760 While it would naturally make sense, flow rules cannot be assumed to be
2761 processed by hardware in the same order as their creation for several
2764 - They may be managed internally as a tree or a hash table instead of a
2766 - Removing a flow rule before adding another one can either put the new rule
2767 at the end of the list or reuse a freed entry.
2768 - Duplication may occur when packets are matched by several rules.
2770 For overlapping rules (particularly in order to use `Action: PASSTHRU`_)
2771 predictable behavior is only guaranteed by using different priority levels.
2773 Priority levels are not necessarily implemented in hardware, or may be
2774 severely limited (e.g. a single priority bit).
2776 For these reasons, priority levels may be implemented purely in software by
2779 - For devices expecting flow rules to be added in the correct order, PMDs
2780 may destroy and re-create existing rules after adding a new one with
2783 - A configurable number of dummy or empty rules can be created at
2784 initialization time to save high priority slots for later.
2786 - In order to save priority levels, PMDs may evaluate whether rules are
2787 likely to collide and adjust their priority accordingly.
2792 - A device profile selection function which could be used to force a
2793 permanent profile instead of relying on its automatic configuration based
2794 on existing flow rules.
2796 - A method to optimize *rte_flow* rules with specific pattern items and
2797 action types generated on the fly by PMDs. DPDK should assign negative
2798 numbers to these in order to not collide with the existing types. See
2801 - Adding specific egress pattern items and actions as described in
2802 `Attribute: Traffic direction`_.
2804 - Optional software fallback when PMDs are unable to handle requested flow
2805 rules so applications do not have to implement their own.
2807 .. _OpenFlow Switch Specification: https://www.opennetworking.org/software-defined-standards/specifications/
git.droids-corp.org / dpdk.git / blob