1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2018 6WIND S.A.
4 .. _switch_representation:
6 Switch Representation within DPDK Applications
7 ==============================================
14 Network adapters with multiple physical ports and/or SR-IOV capabilities
15 usually support the offload of traffic steering rules between their virtual
16 functions (VFs), sub functions (SFs), physical functions (PFs) and ports.
18 Like for standard Ethernet switches, this involves a combination of
19 automatic MAC learning and manual configuration. For most purposes it is
20 managed by the host system and fully transparent to users and applications.
22 On the other hand, applications typically found on hypervisors that process
23 layer 2 (L2) traffic (such as OVS) need to steer traffic themselves
24 according on their own criteria.
26 Without a standard software interface to manage traffic steering rules
27 between VFs, SFs, PFs and the various physical ports of a given device,
28 applications cannot take advantage of these offloads; software processing is
29 mandatory even for traffic which ends up re-injected into the device it
32 This document describes how such steering rules can be configured through
33 the DPDK flow API (**rte_flow**), with emphasis on the SR-IOV use case
34 (PF/VF steering) using a single physical port for clarity, however the same
35 logic applies to any number of ports without necessarily involving SR-IOV.
39 Besides SR-IOV, Sub function is a portion of the PCI device, a SF netdev
40 has its own dedicated queues(txq, rxq). A SF netdev supports E-Switch
41 representation offload similar to existing PF and VF representors.
42 A SF shares PCI level resources with other SFs and/or with its parent PCI
45 Sub function is created on-demand, coexists with VFs. Number of SFs is
46 limited by hardware resources.
51 In many cases, traffic steering rules cannot be determined in advance;
52 applications usually have to process a bit of traffic in software before
53 thinking about offloading specific flows to hardware.
55 Applications therefore need the ability to receive and inject traffic to
56 various device endpoints (other VFs, SFs, PFs or physical ports) before
57 connecting them together. Device drivers must provide means to hook the
58 "other end" of these endpoints and to refer them when configuring flow
61 This role is left to so-called "port representors" (also known as "VF
62 representors" in the specific context of VFs, "SF representors" in the
63 specific context of SFs), which are to DPDK what the Ethernet switch
64 device driver model (**switchdev**) [1]_ is to Linux, and which can be
65 thought as a software "patch panel" front-end for applications.
67 - DPDK port representors are implemented as additional virtual Ethernet
68 device (**ethdev**) instances, spawned on an as needed basis through
69 configuration parameters passed to the driver of the underlying
74 -a pci:dbdf,representor=vf0
75 -a pci:dbdf,representor=vf[0-3]
76 -a pci:dbdf,representor=vf[0,5-11]
77 -a pci:dbdf,representor=sf1
78 -a pci:dbdf,representor=sf[0-1023]
79 -a pci:dbdf,representor=sf[0,2-1023]
81 - As virtual devices, they may be more limited than their physical
82 counterparts, for instance by exposing only a subset of device
83 configuration callbacks and/or by not necessarily having Rx/Tx capability.
85 - Among other things, they can be used to assign MAC addresses to the
86 resource they represent.
88 - Applications can tell port representors apart from other physical of virtual
89 port by checking the dev_flags field within their device information
90 structure for the RTE_ETH_DEV_REPRESENTOR bit-field.
94 struct rte_eth_dev_info {
96 uint32_t dev_flags; /**< Device flags */
100 - The device or group relationship of ports can be discovered using the
101 switch ``domain_id`` field within the devices switch information structure. By
102 default the switch ``domain_id`` of a port will be
103 ``RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID`` to indicate that the port doesn't
104 support the concept of a switch domain, but ports which do support the concept
105 will be allocated a unique switch ``domain_id``, ports within the same switch
106 domain will share the same ``domain_id``. The switch ``port_id`` is used to
107 specify the port_id in terms of the switch, so in the case of SR-IOV devices
108 the switch ``port_id`` would represent the virtual function identifier of the
114 * Ethernet device associated switch information
116 struct rte_eth_switch_info {
117 const char *name; /**< switch name */
118 uint16_t domain_id; /**< switch domain id */
119 uint16_t port_id; /**< switch port id */
123 .. [1] `Ethernet switch device driver model (switchdev)
124 <https://www.kernel.org/doc/Documentation/networking/switchdev.txt>`_
129 "Basic" in the sense that it is not managed by applications, which
130 nonetheless expect traffic to flow between the various endpoints and the
131 outside as if everything was linked by an Ethernet hub.
133 The following diagram pictures a setup involving a device with one PF, two
134 VFs and one shared physical port
138 .-------------. .-------------. .-------------.
139 | hypervisor | | VM 1 | | VM 2 |
140 | application | | application | | application |
141 `--+----------' `----------+--' `--+----------'
147 .-+--. .---+--. .--+---.
148 | PF | | VF 1 | | VF 2 |
149 `-+--' `---+--' `--+---'
151 `---------. .-----------------------' |
152 | | .-------------------------'
163 - A DPDK application running on the hypervisor owns the PF device, which is
164 arbitrarily assigned port index 3.
166 - Both VFs are assigned to VMs and used by unknown applications; they may be
167 DPDK-based or anything else.
169 - Interconnection is not necessarily done through a true Ethernet switch and
170 may not even exist as a separate entity. The role of this block is to show
171 that something brings PF, VFs and physical ports together and enables
172 communication between them, with a number of built-in restrictions.
174 Subsequent sections in this document describe means for DPDK applications
175 running on the hypervisor to freely assign specific flows between PF, VFs
176 and physical ports based on traffic properties, by managing this
185 When a DPDK application gets assigned a PF device and is deliberately not
186 started in `basic SR-IOV`_ mode, any traffic coming from physical ports is
187 received by PF according to default rules, while VFs remain isolated.
191 .-------------. .-------------. .-------------.
192 | hypervisor | | VM 1 | | VM 2 |
193 | application | | application | | application |
194 `--+----------' `----------+--' `--+----------'
200 .-+--. .---+--. .--+---.
201 | PF | | VF 1 | | VF 2 |
202 `-+--' `------' `------'
206 .--+----------------------.
207 | managed interconnection |
208 `------------+------------'
215 In this mode, interconnection must be configured by the application to
216 enable VF communication, for instance by explicitly directing traffic with a
217 given destination MAC address to VF 1 and allowing that with the same source
218 MAC address to come out of it.
220 For this to work, hypervisor applications need a way to refer to either VF 1
221 or VF 2 in addition to the PF. This is addressed by `VF representors`_.
226 VF representors are virtual but standard DPDK network devices (albeit with
227 limited capabilities) created by PMDs when managing a PF device.
229 Since they represent VF instances used by other applications, configuring
230 them (e.g. assigning a MAC address or setting up promiscuous mode) affects
231 interconnection accordingly. If supported, they may also be used as two-way
232 communication ports with VFs (assuming **switchdev** topology)
237 .-------------. .-------------. .-------------.
238 | hypervisor | | VM 1 | | VM 2 |
239 | application | | application | | application |
240 `--+---+---+--' `----------+--' `--+----------'
242 | | `-------------------. | |
245 .-----+-----. .-----+-----. .-----+-----. | |
246 | port_id 3 | | port_id 4 | | port_id 5 | | |
247 `-----+-----' `-----+-----' `-----+-----' | |
249 .-+--. .-----+-----. .-----+-----. .---+--. .--+---.
250 | PF | | VF 1 rep. | | VF 2 rep. | | VF 1 | | VF 2 |
251 `-+--' `-----+-----' `-----+-----' `---+--' `--+---'
254 `-----. | | .-----------------' |
255 | | | | .---------------------'
257 .--+-------+---+---+---+--.
258 | managed interconnection |
259 `------------+------------'
266 - VF representors are assigned arbitrary port indices 4 and 5 in the
267 hypervisor application and are respectively associated with VF 1 and VF 2.
269 - They can't be dissociated; even if VF 1 and VF 2 were not connected,
270 representors could still be used for configuration.
272 - In this context, port index 3 can be thought as a representor for physical
275 As previously described, the "interconnection" block represents a logical
276 concept. Interconnection occurs when hardware configuration enables traffic
277 flows from one place to another (e.g. physical port 0 to VF 1) according to
280 This is discussed in more detail in `traffic steering`_.
285 In the following diagram, each meaningful traffic origin or endpoint as seen
286 by the hypervisor application is tagged with a unique letter from A to F.
290 .-------------. .-------------. .-------------.
291 | hypervisor | | VM 1 | | VM 2 |
292 | application | | application | | application |
293 `--+---+---+--' `----------+--' `--+----------'
295 | | `-------------------. | |
298 .----(A)----. .----(B)----. .----(C)----. | |
299 | port_id 3 | | port_id 4 | | port_id 5 | | |
300 `-----+-----' `-----+-----' `-----+-----' | |
302 .-+--. .-----+-----. .-----+-----. .---+--. .--+---.
303 | PF | | VF 1 rep. | | VF 2 rep. | | VF 1 | | VF 2 |
304 `-+--' `-----+-----' `-----+-----' `--(D)-' `-(E)--'
307 `-----. | | .-----------------' |
308 | | | | .---------------------'
310 .--+-------+---+---+---+--.
311 | managed interconnection |
312 `------------+------------'
320 - **B**: port representor for VF 1.
321 - **C**: port representor for VF 2.
322 - **D**: VF 1 proper.
323 - **E**: VF 2 proper.
324 - **F**: physical port.
326 Although uncommon, some devices do not enforce a one to one mapping between
327 PF and physical ports. For instance, by default all ports of **mlx4**
328 adapters are available to all their PF/VF instances, in which case
329 additional ports appear next to **F** in the above diagram.
331 Assuming no interconnection is provided by default in this mode, setting up
332 a `basic SR-IOV`_ configuration involving physical port 0 could be broken
337 - **A to F**: let everything through.
338 - **F to A**: PF MAC as destination.
342 - **A to D**, **E to D** and **F to D**: VF 1 MAC as destination.
343 - **D to A**: VF 1 MAC as source and PF MAC as destination.
344 - **D to E**: VF 1 MAC as source and VF 2 MAC as destination.
345 - **D to F**: VF 1 MAC as source.
349 - **A to E**, **D to E** and **F to E**: VF 2 MAC as destination.
350 - **E to A**: VF 2 MAC as source and PF MAC as destination.
351 - **E to D**: VF 2 MAC as source and VF 1 MAC as destination.
352 - **E to F**: VF 2 MAC as source.
354 Devices may additionally support advanced matching criteria such as
355 IPv4/IPv6 addresses or TCP/UDP ports.
357 The combination of matching criteria with target endpoints fits well with
358 **rte_flow** [6]_, which expresses flow rules as combinations of patterns
361 Enhancing **rte_flow** with the ability to make flow rules match and target
362 these endpoints provides a standard interface to manage their
363 interconnection without introducing new concepts and whole new API to
364 implement them. This is described in `flow API (rte_flow)`_.
366 .. [6] :doc:`Generic flow API (rte_flow) <rte_flow>`
374 Compared to creating a brand new dedicated interface, **rte_flow** was
375 deemed flexible enough to manage representor traffic only with minor
378 - Using physical ports, PF, SF, VF or port representors as targets.
380 - Affecting traffic that is not necessarily addressed to the DPDK port ID a
381 flow rule is associated with (e.g. forcing VF traffic redirection to PF).
385 - Rule-based packet counters.
387 - The ability to combine several identical actions for traffic duplication
388 (e.g. VF representor in addition to a physical port).
390 - Dedicated actions for traffic encapsulation / decapsulation before
391 reaching an endpoint.
396 From an application standpoint, "ingress" and "egress" flow rule attributes
397 apply to the DPDK port ID they are associated with. They select a traffic
398 direction for matching patterns, but have no impact on actions.
400 When matching traffic coming from or going to a different place than the
401 immediate port ID a flow rule is associated with, these attributes keep
402 their meaning while applying to the chosen origin, as highlighted by the
407 .-------------. .-------------. .-------------.
408 | hypervisor | | VM 1 | | VM 2 |
409 | application | | application | | application |
410 `--+---+---+--' `----------+--' `--+----------'
412 | | `-------------------. | |
415 | | ingress | | ingress | | ingress | |
416 | | egress | | egress | | egress | |
418 .----(A)----. .----(B)----. .----(C)----. | |
419 | port_id 3 | | port_id 4 | | port_id 5 | | |
420 `-----+-----' `-----+-----' `-----+-----' | |
422 .-+--. .-----+-----. .-----+-----. .---+--. .--+---.
423 | PF | | VF 1 rep. | | VF 2 rep. | | VF 1 | | VF 2 |
424 `-+--' `-----+-----' `-----+-----' `--(D)-' `-(E)--'
426 | | | egress | | | | egress
427 | | | ingress | | | | ingress
428 | | .---------' v | | v
429 `-----. | | .-----------------' |
430 | | | | .---------------------'
432 .--+-------+---+---+---+--.
433 | managed interconnection |
434 `------------+------------'
444 Ingress and egress are defined as relative to the application creating the
447 For instance, matching traffic sent by VM 2 would be done through an ingress
448 flow rule on VF 2 (**E**). Likewise for incoming traffic on physical port
449 (**F**). This also applies to **C** and **A** respectively.
454 Without Port Representors
455 ^^^^^^^^^^^^^^^^^^^^^^^^^
457 `Traffic direction`_ describes how an application could match traffic coming
458 from or going to a specific place reachable from a DPDK port ID. This makes
459 sense when the traffic in question is normally seen (i.e. sent or received)
460 by the application creating the flow rule (e.g. as in "redirect all traffic
461 coming from VF 1 to local queue 6").
463 However this does not force such traffic to take a specific route. Creating
464 a flow rule on **A** matching traffic coming from **D** is only meaningful
465 if it can be received by **A** in the first place, otherwise doing so simply
468 A new flow rule attribute named "transfer" is necessary for that. Combining
469 it with "ingress" or "egress" and a specific origin requests a flow rule to
470 be applied at the lowest level
474 ingress only : ingress + transfer
476 .-------------. .-------------. : .-------------. .-------------.
477 | hypervisor | | VM 1 | : | hypervisor | | VM 1 |
478 | application | | application | : | application | | application |
479 `------+------' `--+----------' : `------+------' `--+----------'
480 | | | traffic : | | | traffic
481 .----(A)----. | v : .----(A)----. | v
482 | port_id 3 | | : | port_id 3 | |
483 `-----+-----' | : `-----+-----' |
486 .-+--. .---+--. : .-+--. .---+--.
487 | PF | | VF 1 | : | PF | | VF 1 |
488 `-+--' `--(D)-' : `-+--' `--(D)-'
489 | | | traffic : | ^ | | traffic
490 | | v : | | traffic | v
491 .--+-----------+--. : .--+-----------+--.
492 | interconnection | : | interconnection |
493 `--------+--------' : `--------+--------'
496 .---(F)----. : .---(F)----.
497 | physical | : | physical |
498 | port 0 | : | port 0 |
499 `----------' : `----------'
501 With "ingress" only, traffic is matched on **A** thus still goes to physical
502 port **F** by default
507 testpmd> flow create 3 ingress pattern vf id is 1 / end
508 actions queue index 6 / end
510 With "ingress + transfer", traffic is matched on **D** and is therefore
511 successfully assigned to queue 6 on **A**
516 testpmd> flow create 3 ingress transfer pattern vf id is 1 / end
517 actions queue index 6 / end
520 With Port Representors
521 ^^^^^^^^^^^^^^^^^^^^^^
523 When port representors exist, implicit flow rules with the "transfer"
524 attribute (described in `without port representors`_) are be assumed to
525 exist between them and their represented resources. These may be immutable.
527 In this case, traffic is received by default through the representor and
528 neither the "transfer" attribute nor traffic origin in flow rule patterns
529 are necessary. They simply have to be created on the representor port
530 directly and may target a different representor as described in `PORT_ID
533 Implicit traffic flow with port representor
537 .-------------. .-------------.
538 | hypervisor | | VM 1 |
539 | application | | application |
540 `--+-------+--' `----------+--'
545 .----(A)----. .----(B)----. |
546 | port_id 3 | | port_id 4 | |
547 `-----+-----' `-----+-----' |
549 .-+--. .-----+-----. .---+--.
550 | PF | | VF 1 rep. | | VF 1 |
551 `-+--' `-----+-----' `--(D)-'
553 .--|-------------|-----------|--.
557 `--|----------------------------'
564 Pattern Items And Actions
565 ~~~~~~~~~~~~~~~~~~~~~~~~~
570 Matches traffic originating from (ingress) or going to (egress) a physical
571 port of the underlying device.
573 Using this pattern item without specifying a port index matches the physical
574 port associated with the current DPDK port ID by default. As described in
575 `traffic steering`_, specifying it should be rarely needed.
577 - Matches **F** in `traffic steering`_.
582 Directs matching traffic to a given physical port index.
584 - Targets **F** in `traffic steering`_.
589 Matches traffic originating from (ingress) or going to (egress) a given DPDK
592 Normally only supported if the port ID in question is known by the
593 underlying PMD and related to the device the flow rule is created against.
595 This must not be confused with the `PORT pattern item`_ which refers to the
596 physical port of a device. ``PORT_ID`` refers to a ``struct rte_eth_dev``
597 object on the application side (also known as "port representor" depending
598 on the kind of underlying device).
600 - Matches **A**, **B** or **C** in `traffic steering`_.
605 Directs matching traffic to a given DPDK port ID.
607 Same restrictions as `PORT_ID pattern item`_.
609 - Targets **A**, **B** or **C** in `traffic steering`_.
614 Matches traffic originating from (ingress) or going to (egress) the physical
615 function of the current device.
617 If supported, should work even if the physical function is not managed by
618 the application and thus not associated with a DPDK port ID. Its behavior is
619 otherwise similar to `PORT_ID pattern item`_ using PF port ID.
621 - Matches **A** in `traffic steering`_.
626 Directs matching traffic to the physical function of the current device.
628 Same restrictions as `PF pattern item`_.
630 - Targets **A** in `traffic steering`_.
635 Matches traffic originating from (ingress) or going to (egress) a given
636 virtual function of the current device.
638 If supported, should work even if the virtual function is not managed by
639 the application and thus not associated with a DPDK port ID. Its behavior is
640 otherwise similar to `PORT_ID pattern item`_ using VF port ID.
642 Note this pattern item does not match VF representors traffic which, as
643 separate entities, should be addressed through their own port IDs.
645 - Matches **D** or **E** in `traffic steering`_.
650 Directs matching traffic to a given virtual function of the current device.
652 Same restrictions as `VF pattern item`_.
654 - Targets **D** or **E** in `traffic steering`_.
659 These actions are named according to the protocol they encapsulate traffic
660 with (e.g. ``VXLAN_ENCAP``) and using specific parameters (e.g. VNI for
663 While they modify traffic and can be used multiple times (order matters),
664 unlike `PORT_ID action`_ and friends, they have no impact on steering.
666 As described in `actions order and repetition`_ this means they are useless
667 if used alone in an action list, the resulting traffic gets dropped unless
668 combined with either ``PASSTHRU`` or other endpoint-targeting actions.
673 They perform the reverse of `\*_ENCAP actions`_ by popping protocol headers
674 from traffic instead of pushing them. They can be used multiple times as
677 Note that using these actions on non-matching traffic results in undefined
678 behavior. It is recommended to match the protocol headers to decapsulate on
679 the pattern side of a flow rule in order to use these actions or otherwise
680 make sure only matching traffic goes through.
682 Actions Order and Repetition
683 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
685 Flow rules are currently restricted to at most a single action of each
686 supported type, performed in an unpredictable order (or all at once). To
687 repeat actions in a predictable fashion, applications have to make rules
688 pass-through and use priority levels.
690 It's now clear that PMD support for chaining multiple non-terminating flow
691 rules of varying priority levels is prohibitively difficult to implement
692 compared to simply allowing multiple identical actions performed in a
693 defined order by a single flow rule.
695 - This change is required to support protocol encapsulation offloads and the
696 ability to perform them multiple times (e.g. VLAN then VXLAN).
698 - It makes the ``DUP`` action redundant since multiple ``QUEUE`` actions can
699 be combined for duplication.
701 - The (non-)terminating property of actions must be discarded. Instead, flow
702 rules themselves must be considered terminating by default (i.e. dropping
703 traffic if there is no specific target) unless a ``PASSTHRU`` action is
709 This section provides practical examples based on the established testpmd
710 flow command syntax [2]_, in the context described in `traffic steering`_
714 .-------------. .-------------. .-------------.
715 | hypervisor | | VM 1 | | VM 2 |
716 | application | | application | | application |
717 `--+---+---+--' `----------+--' `--+----------'
719 | | `-------------------. | |
722 .----(A)----. .----(B)----. .----(C)----. | |
723 | port_id 3 | | port_id 4 | | port_id 5 | | |
724 `-----+-----' `-----+-----' `-----+-----' | |
726 .-+--. .-----+-----. .-----+-----. .---+--. .--+---.
727 | PF | | VF 1 rep. | | VF 2 rep. | | VF 1 | | VF 2 |
728 `-+--' `-----+-----' `-----+-----' `--(D)-' `-(E)--'
731 `-----. | | .-----------------' |
732 | | | | .---------------------'
734 .--|-------|---|---|---|--.
738 `------------|------------'
745 By default, PF (**A**) can communicate with the physical port it is
746 associated with (**F**), while VF 1 (**D**) and VF 2 (**E**) are isolated
747 and restricted to communicate with the hypervisor application through their
748 respective representors (**B** and **C**) if supported.
750 Examples in subsequent sections apply to hypervisor applications only and
751 are based on port representors **A**, **B** and **C**.
753 .. [2] :ref:`Flow syntax <testpmd_rte_flow>`
755 Associating VF 1 with Physical Port 0
756 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
758 Assign all port traffic (**F**) to VF 1 (**D**) indiscriminately through
763 flow create 3 ingress pattern / end actions port_id id 4 / end
764 flow create 4 ingress pattern / end actions port_id id 3 / end
766 More practical example with MAC address restrictions
770 flow create 3 ingress
771 pattern eth dst is {VF 1 MAC} / end
772 actions port_id id 4 / end
776 flow create 4 ingress
777 pattern eth src is {VF 1 MAC} / end
778 actions port_id id 3 / end
784 From outside to PF and VFs
788 flow create 3 ingress
789 pattern eth dst is ff:ff:ff:ff:ff:ff / end
790 actions port_id id 3 / port_id id 4 / port_id id 5 / end
792 Note ``port_id id 3`` is necessary otherwise only VFs would receive matching
795 From PF to outside and VFs
800 pattern eth dst is ff:ff:ff:ff:ff:ff / end
801 actions port / port_id id 4 / port_id id 5 / end
803 From VFs to outside and PF
807 flow create 4 ingress
808 pattern eth dst is ff:ff:ff:ff:ff:ff src is {VF 1 MAC} / end
809 actions port_id id 3 / port_id id 5 / end
811 flow create 5 ingress
812 pattern eth dst is ff:ff:ff:ff:ff:ff src is {VF 2 MAC} / end
813 actions port_id id 4 / port_id id 4 / end
815 Similar ``33:33:*`` rules based on known MAC addresses should be added for
818 Encapsulating VF 2 Traffic in VXLAN
819 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
821 Assuming pass-through flow rules are supported
825 flow create 5 ingress
827 actions vxlan_encap vni 42 / passthru / end
832 pattern vxlan vni is 42 / end
833 actions vxlan_decap / passthru / end
835 Here ``passthru`` is needed since as described in `actions order and
836 repetition`_, flow rules are otherwise terminating; if supported, a rule
837 without a target endpoint will drop traffic.
839 Without pass-through support, ingress encapsulation on the destination
840 endpoint might not be supported and action list must provide one
844 flow create 5 ingress
845 pattern eth src is {VF 2 MAC} / end
846 actions vxlan_encap vni 42 / port_id id 3 / end
848 flow create 3 ingress
849 pattern vxlan vni is 42 / end
850 actions vxlan_decap / port_id id 5 / end