1 Design Decisions In Open vSwitch
2 ================================
4 This document describes design decisions that went into implementing
5 Open vSwitch. While we believe these to be reasonable decisions, it is
6 impossible to predict how Open vSwitch will be used in all environments.
7 Understanding assumptions made by Open vSwitch is critical to a
8 successful deployment. The end of this document contains contact
9 information that can be used to let us know how we can make Open vSwitch
10 more generally useful.
15 Over time, Open vSwitch has added many knobs that control whether a
16 given controller receives OpenFlow asynchronous messages. This
17 section describes how all of these features interact.
19 First, a service controller never receives any asynchronous messages
20 unless it changes its miss_send_len from the service controller
21 default of zero in one of the following ways:
23 - Sending an OFPT_SET_CONFIG message with nonzero miss_send_len.
25 - Sending any NXT_SET_ASYNC_CONFIG message: as a side effect, this
26 message changes the miss_send_len to
27 OFP_DEFAULT_MISS_SEND_LEN (128) for service controllers.
29 Second, OFPT_FLOW_REMOVED and NXT_FLOW_REMOVED messages are generated
30 only if the flow that was removed had the OFPFF_SEND_FLOW_REM flag
33 Third, OFPT_PACKET_IN and NXT_PACKET_IN messages are sent only to
34 OpenFlow controller connections that have the correct connection ID
35 (see "struct nx_controller_id" and "struct nx_action_controller"):
37 - For packet-in messages generated by a NXAST_CONTROLLER action,
38 the controller ID specified in the action.
40 - For other packet-in messages, controller ID zero. (This is the
41 default ID when an OpenFlow controller does not configure one.)
43 Finally, Open vSwitch consults a per-connection table indexed by the
44 message type, reason code, and current role. The following table
45 shows how this table is initialized by default when an OpenFlow
46 connection is made. An entry labeled "yes" means that the message is
47 sent, an entry labeled "---" means that the message is suppressed.
51 message and reason code other slave
52 ---------------------------------------- ------- -----
53 OFPT_PACKET_IN / NXT_PACKET_IN
56 OFPR_INVALID_TTL --- ---
57 OFPR_ACTION_SET (OF1.4+) yes ---
58 OFPR_GROUP (OF1.4+) yes ---
60 OFPT_FLOW_REMOVED / NXT_FLOW_REMOVED
61 OFPRR_IDLE_TIMEOUT yes ---
62 OFPRR_HARD_TIMEOUT yes ---
71 The NXT_SET_ASYNC_CONFIG message directly sets all of the values in
72 this table for the current connection. The
73 OFPC_INVALID_TTL_TO_CONTROLLER bit in the OFPT_SET_CONFIG message
74 controls the setting for OFPR_INVALID_TTL for the "master" role.
80 The OpenFlow 1.0 specification requires the output port of the OFPAT_ENQUEUE
81 action to "refer to a valid physical port (i.e. < OFPP_MAX) or OFPP_IN_PORT".
82 Although OFPP_LOCAL is not less than OFPP_MAX, it is an 'internal' port which
83 can have QoS applied to it in Linux. Since we allow the OFPAT_ENQUEUE to apply
84 to 'internal' ports whose port numbers are less than OFPP_MAX, we interpret
85 OFPP_LOCAL as a physical port and support OFPAT_ENQUEUE on it as well.
91 The OpenFlow specification for the behavior of OFPT_FLOW_MOD is
92 confusing. The following tables summarize the Open vSwitch
93 implementation of its behavior in the following categories:
95 - "match on priority": Whether the flow_mod acts only on flows
96 whose priority matches that included in the flow_mod message.
98 - "match on out_port": Whether the flow_mod acts only on flows
99 that output to the out_port included in the flow_mod message (if
100 out_port is not OFPP_NONE). OpenFlow 1.1 and later have a
101 similar feature (not listed separately here) for out_group.
103 - "match on flow_cookie": Whether the flow_mod acts only on flows
104 whose flow_cookie matches an optional controller-specified value
107 - "updates flow_cookie": Whether the flow_mod changes the
108 flow_cookie of the flow or flows that it matches to the
109 flow_cookie included in the flow_mod message.
111 - "updates OFPFF_ flags": Whether the flow_mod changes the
112 OFPFF_SEND_FLOW_REM flag of the flow or flows that it matches to
113 the setting included in the flags of the flow_mod message.
115 - "honors OFPFF_CHECK_OVERLAP": Whether the OFPFF_CHECK_OVERLAP
116 flag in the flow_mod is significant.
118 - "updates idle_timeout" and "updates hard_timeout": Whether the
119 idle_timeout and hard_timeout in the flow_mod, respectively,
120 have an effect on the flow or flows matched by the flow_mod.
122 - "updates idle timer": Whether the flow_mod resets the per-flow
123 timer that measures how long a flow has been idle.
125 - "updates hard timer": Whether the flow_mod resets the per-flow
126 timer that measures how long it has been since a flow was
129 - "zeros counters": Whether the flow_mod resets per-flow packet
130 and byte counters to zero.
132 - "may add a new flow": Whether the flow_mod may add a new flow to
133 the flow table. (Obviously this is always true for "add"
134 commands but in some OpenFlow versions "modify" and
135 "modify-strict" can also add new flows.)
137 - "sends flow_removed message": Whether the flow_mod generates a
138 flow_removed message for the flow or flows that it affects.
140 An entry labeled "yes" means that the flow mod type does have the
141 indicated behavior, "---" means that it does not, an empty cell means
142 that the property is not applicable, and other values are explained
150 ADD MODIFY STRICT DELETE STRICT
151 === ====== ====== ====== ======
152 match on priority yes --- yes --- yes
153 match on out_port --- --- --- yes yes
154 match on flow_cookie --- --- --- --- ---
155 match on table_id --- --- --- --- ---
156 controller chooses table_id --- --- ---
157 updates flow_cookie yes yes yes
158 updates OFPFF_SEND_FLOW_REM yes + +
159 honors OFPFF_CHECK_OVERLAP yes + +
160 updates idle_timeout yes + +
161 updates hard_timeout yes + +
162 resets idle timer yes + +
163 resets hard timer yes yes yes
164 zeros counters yes + +
165 may add a new flow yes yes yes
166 sends flow_removed message --- --- --- % %
168 (+) "modify" and "modify-strict" only take these actions when they
169 create a new flow, not when they update an existing flow.
171 (%) "delete" and "delete_strict" generates a flow_removed message if
172 the deleted flow or flows have the OFPFF_SEND_FLOW_REM flag set.
173 (Each controller can separately control whether it wants to
174 receive the generated messages.)
180 OpenFlow 1.1 makes these changes:
182 - The controller now must specify the table_id of the flow match
183 searched and into which a flow may be inserted. Behavior for a
184 table_id of 255 is undefined.
186 - A flow_mod, except an "add", can now match on the flow_cookie.
188 - When a flow_mod matches on the flow_cookie, "modify" and
189 "modify-strict" never insert a new flow.
193 ADD MODIFY STRICT DELETE STRICT
194 === ====== ====== ====== ======
195 match on priority yes --- yes --- yes
196 match on out_port --- --- --- yes yes
197 match on flow_cookie --- yes yes yes yes
198 match on table_id yes yes yes yes yes
199 controller chooses table_id yes yes yes
200 updates flow_cookie yes --- ---
201 updates OFPFF_SEND_FLOW_REM yes + +
202 honors OFPFF_CHECK_OVERLAP yes + +
203 updates idle_timeout yes + +
204 updates hard_timeout yes + +
205 resets idle timer yes + +
206 resets hard timer yes yes yes
207 zeros counters yes + +
208 may add a new flow yes # #
209 sends flow_removed message --- --- --- % %
211 (+) "modify" and "modify-strict" only take these actions when they
212 create a new flow, not when they update an existing flow.
214 (%) "delete" and "delete_strict" generates a flow_removed message if
215 the deleted flow or flows have the OFPFF_SEND_FLOW_REM flag set.
216 (Each controller can separately control whether it wants to
217 receive the generated messages.)
219 (#) "modify" and "modify-strict" only add a new flow if the flow_mod
220 does not match on any bits of the flow cookie
226 OpenFlow 1.2 makes these changes:
228 - Only "add" commands ever add flows, "modify" and "modify-strict"
231 - A new flag OFPFF_RESET_COUNTS now controls whether "modify" and
232 "modify-strict" reset counters, whereas previously they never
233 reset counters (except when they inserted a new flow).
237 ADD MODIFY STRICT DELETE STRICT
238 === ====== ====== ====== ======
239 match on priority yes --- yes --- yes
240 match on out_port --- --- --- yes yes
241 match on flow_cookie --- yes yes yes yes
242 match on table_id yes yes yes yes yes
243 controller chooses table_id yes yes yes
244 updates flow_cookie yes --- ---
245 updates OFPFF_SEND_FLOW_REM yes --- ---
246 honors OFPFF_CHECK_OVERLAP yes --- ---
247 updates idle_timeout yes --- ---
248 updates hard_timeout yes --- ---
249 resets idle timer yes --- ---
250 resets hard timer yes yes yes
251 zeros counters yes & &
252 may add a new flow yes --- ---
253 sends flow_removed message --- --- --- % %
255 (%) "delete" and "delete_strict" generates a flow_removed message if
256 the deleted flow or flows have the OFPFF_SEND_FLOW_REM flag set.
257 (Each controller can separately control whether it wants to
258 receive the generated messages.)
260 (&) "modify" and "modify-strict" reset counters if the
261 OFPFF_RESET_COUNTS flag is specified.
267 OpenFlow 1.3 makes these changes:
269 - Behavior for a table_id of 255 is now defined, for "delete" and
270 "delete-strict" commands, as meaning to delete from all tables.
271 A table_id of 255 is now explicitly invalid for other commands.
273 - New flags OFPFF_NO_PKT_COUNTS and OFPFF_NO_BYT_COUNTS for "add"
276 The table for 1.3 is the same as the one shown above for 1.2.
282 OpenFlow 1.4 adds the "importance" field to flow_mods, but it does not
283 explicitly specify which kinds of flow_mods set the importance. For
284 consistency, Open vSwitch uses the same rule for importance as for
285 idle_timeout and hard_timeout, that is, only an "ADD" flow_mod sets
286 the importance. (This issue has been filed with the ONF as EXT-496.)
292 Open vSwitch makes all flow table modifications atomically, i.e., any
293 datapath packet only sees flow table configurations either before or
294 after any change made by any flow_mod. For example, if a controller
295 removes all flows with a single OpenFlow "flow_mod", no packet sees an
296 intermediate version of the OpenFlow pipeline where only some of the
297 flows have been deleted.
299 It should be noted that Open vSwitch caches datapath flows, and that
300 the cached flows are NOT flushed immediately when a flow table
301 changes. Instead, the datapath flows are revalidated against the new
302 flow table as soon as possible, and usually within one second of the
303 modification. This design amortizes the cost of datapath cache
304 flushing across multiple flow table changes, and has a significant
305 performance effect during simultaneous heavy flow table churn and high
306 traffic load. This means that different cached datapath flows may
307 have been computed based on a different flow table configurations, but
308 each of the datapath flows is guaranteed to have been computed over a
309 coherent view of the flow tables, as described above.
311 With OpenFlow 1.4 bundles this atomicity can be extended across an
312 arbitrary set of flow_mods. Bundles are supported for flow_mod and
313 port_mod messages only. For flow_mods, both 'atomic' and 'ordered'
314 bundle flags are trivially supported, as all bundled messages are
315 executed in the order they were added and all flow table modifications
316 are now atomic to the datapath. Port mods may not appear in atomic
317 bundles, as port status modifications are not atomic.
319 To support bundles, ovs-ofctl has a '--bundle' option that makes the
320 flow mod commands ('add-flow', 'add-flows', 'mod-flows', 'del-flows',
321 and 'replace-flows') use an OpenFlow 1.4 bundle to operate the
322 modifications as a single atomic transaction. If any of the flow mods
323 in a transaction fail, none of them are executed. All flow mods in a
324 bundle appear to datapath lookups simultaneously.
326 Furthermore, ovs-ofctl 'add-flow' and 'add-flows' commands now accept
327 arbitrary flow mods as an input by allowing the flow specification to
328 start with an explicit 'add', 'modify', 'modify_strict', 'delete', or
329 'delete_strict' keyword. A missing keyword is treated as 'add', so
330 this is fully backwards compatible. With the new '--bundle' option
331 all the flow mods are executed as a single atomic transaction using an
332 OpenFlow 1.4 bundle. Without the '--bundle' option the flow mods are
333 executed in order up to the first failing flow_mod, and in case of an
334 error the earlier successful flow_mods are not rolled back.
340 The OpenFlow 1.1 specification for OFPT_PACKET_IN is confusing. The
341 definition in OF1.1 openflow.h is[*]:
344 /* Packet received on port (datapath -> controller). */
345 struct ofp_packet_in {
346 struct ofp_header header;
347 uint32_t buffer_id; /* ID assigned by datapath. */
348 uint32_t in_port; /* Port on which frame was received. */
349 uint32_t in_phy_port; /* Physical Port on which frame was received. */
350 uint16_t total_len; /* Full length of frame. */
351 uint8_t reason; /* Reason packet is being sent (one of OFPR_*) */
352 uint8_t table_id; /* ID of the table that was looked up */
353 uint8_t data[0]; /* Ethernet frame, halfway through 32-bit word,
354 so the IP header is 32-bit aligned. The
355 amount of data is inferred from the length
356 field in the header. Because of padding,
357 offsetof(struct ofp_packet_in, data) ==
358 sizeof(struct ofp_packet_in) - 2. */
360 OFP_ASSERT(sizeof(struct ofp_packet_in) == 24);
363 The confusing part is the comment on the data[] member. This comment
364 is a leftover from OF1.0 openflow.h, in which the comment was correct:
365 sizeof(struct ofp_packet_in) is 20 in OF1.0 and offsetof(struct
366 ofp_packet_in, data) is 18. When OF1.1 was written, the structure
367 members were changed but the comment was carelessly not updated, and
368 the comment became wrong: sizeof(struct ofp_packet_in) and
369 offsetof(struct ofp_packet_in, data) are both 24 in OF1.1.
371 That leaves the question of how to implement ofp_packet_in in OF1.1.
372 The OpenFlow reference implementation for OF1.1 does not include any
373 padding, that is, the first byte of the encapsulated frame immediately
374 follows the 'table_id' member without a gap. Open vSwitch therefore
375 implements it the same way for compatibility.
377 For an earlier discussion, please see the thread archived at:
378 https://mailman.stanford.edu/pipermail/openflow-discuss/2011-August/002604.html
380 [*] The quoted definition is directly from OF1.1. Definitions used
381 inside OVS omit the 8-byte ofp_header members, so the sizes in
382 this discussion are 8 bytes larger than those declared in OVS
389 The 802.1Q VLAN header causes more trouble than any other 4 bytes in
390 networking. More specifically, three versions of OpenFlow and Open
391 vSwitch have among them four different ways to match the contents and
392 presence of the VLAN header. The following table describes how each
395 Match NXM OF1.0 OF1.1 OF1.2
396 ----- --------- ----------- ----------- ------------
397 [1] 0000/0000 ????/1,??/? ????/1,??/? 0000/0000,--
398 [2] 0000/ffff ffff/0,??/? ffff/0,??/? 0000/ffff,--
399 [3] 1xxx/1fff 0xxx/0,??/1 0xxx/0,??/1 1xxx/ffff,--
400 [4] z000/f000 ????/1,0y/0 fffe/0,0y/0 1000/1000,0y
401 [5] zxxx/ffff 0xxx/0,0y/0 0xxx/0,0y/0 1xxx/ffff,0y
402 [6] 0000/0fff <none> <none> <none>
403 [7] 0000/f000 <none> <none> <none>
404 [8] 0000/efff <none> <none> <none>
405 [9] 1001/1001 <none> <none> 1001/1001,--
406 [10] 3000/3000 <none> <none> <none>
408 Each column is interpreted as follows.
410 - Match: See the list below.
412 - NXM: xxxx/yyyy means NXM_OF_VLAN_TCI_W with value xxxx and mask
413 yyyy. A mask of 0000 is equivalent to omitting
414 NXM_OF_VLAN_TCI(_W), a mask of ffff is equivalent to
417 - OF1.0 and OF1.1: wwww/x,yy/z means dl_vlan wwww, OFPFW_DL_VLAN
418 x, dl_vlan_pcp yy, and OFPFW_DL_VLAN_PCP z. ? means that the
419 given nibble is ignored (and conventionally 0 for wwww or yy,
420 conventionally 1 for x or z). <none> means that the given match
423 - OF1.2: xxxx/yyyy,zz means OXM_OF_VLAN_VID_W with value xxxx and
424 mask yyyy, and OXM_OF_VLAN_PCP (which is not maskable) with
425 value zz. A mask of 0000 is equivalent to omitting
426 OXM_OF_VLAN_VID(_W), a mask of ffff is equivalent to
427 OXM_OF_VLAN_VID. -- means that OXM_OF_VLAN_PCP is omitted.
428 <none> means that the given match is not supported.
432 [1] Matches any packet, that is, one without an 802.1Q header or with
433 an 802.1Q header with any TCI value.
435 [2] Matches only packets without an 802.1Q header.
437 NXM: Any match with (vlan_tci == 0) and (vlan_tci_mask & 0x1000)
438 != 0 is equivalent to the one listed in the table.
440 OF1.0: The spec doesn't define behavior if dl_vlan is set to
441 0xffff and OFPFW_DL_VLAN_PCP is not set.
443 OF1.1: The spec says explicitly to ignore dl_vlan_pcp when
444 dl_vlan is set to 0xffff.
446 OF1.2: The spec doesn't say what should happen if (vlan_vid == 0)
447 and (vlan_vid_mask & 0x1000) != 0 but (vlan_vid_mask != 0x1000),
448 but it would be straightforward to also interpret as [2].
450 [3] Matches only packets that have an 802.1Q header with VID xxx (and
453 [4] Matches only packets that have an 802.1Q header with PCP y (and
456 NXM: z is ((y << 1) | 1).
458 OF1.0: The spec isn't very clear, but OVS implements it this way.
460 OF1.2: Presumably other masks such that (vlan_vid_mask & 0x1fff)
461 == 0x1000 would also work, but the spec doesn't define their
464 [5] Matches only packets that have an 802.1Q header with VID xxx and
467 NXM: z is ((y << 1) | 1).
469 OF1.2: Presumably other masks such that (vlan_vid_mask & 0x1fff)
470 == 0x1fff would also work.
472 [6] Matches packets with no 802.1Q header or with an 802.1Q header
473 with a VID of 0. Only possible with NXM.
475 [7] Matches packets with no 802.1Q header or with an 802.1Q header
476 with a PCP of 0. Only possible with NXM.
478 [8] Matches packets with no 802.1Q header or with an 802.1Q header
479 with both VID and PCP of 0. Only possible with NXM.
481 [9] Matches only packets that have an 802.1Q header with an
482 odd-numbered VID (and any PCP). Only possible with NXM and
483 OF1.2. (This is just an example; one can match on any desired
486 [10] Matches only packets that have an 802.1Q header with an
487 odd-numbered PCP (and any VID). Only possible with NXM. (This
488 is just an example; one can match on any desired VID bit
493 - OF1.2: The top three bits of OXM_OF_VLAN_VID are fixed to zero,
494 so bits 13, 14, and 15 in the masks listed in the table may be
495 set to arbitrary values, as long as the corresponding value bits
496 are also zero. The suggested ffff mask for [2], [3], and [5]
497 allows a shorter OXM representation (the mask is omitted) than
498 the minimal 1fff mask.
504 OpenFlow 1.0 and later versions have the concept of a "flow cookie",
505 which is a 64-bit integer value attached to each flow. The treatment
506 of the flow cookie has varied greatly across OpenFlow versions,
511 - OFPFC_ADD set the cookie in the flow that it added.
513 - OFPFC_MODIFY and OFPFC_MODIFY_STRICT updated the cookie for
514 the flow or flows that it modified.
516 - OFPST_FLOW messages included the flow cookie.
518 - OFPT_FLOW_REMOVED messages reported the cookie of the flow
521 OpenFlow 1.1 made the following changes:
523 - Flow mod operations OFPFC_MODIFY, OFPFC_MODIFY_STRICT,
524 OFPFC_DELETE, and OFPFC_DELETE_STRICT, plus flow stats
525 requests and aggregate stats requests, gained the ability to
526 match on flow cookies with an arbitrary mask.
528 - OFPFC_MODIFY and OFPFC_MODIFY_STRICT were changed to add a
529 new flow, in the case of no match, only if the flow table
530 modification operation did not match on the cookie field.
531 (In OpenFlow 1.0, modify operations always added a new flow
532 when there was no match.)
534 - OFPFC_MODIFY and OFPFC_MODIFY_STRICT no longer updated flow
537 OpenFlow 1.2 made the following changes:
539 - OFPC_MODIFY and OFPFC_MODIFY_STRICT were changed to never
540 add a new flow, regardless of whether the flow cookie was
543 Open vSwitch support for OpenFlow 1.0 implements the OpenFlow 1.0
544 behavior with the following extensions:
546 - An NXM extension field NXM_NX_COOKIE(_W) allows the NXM
547 versions of OFPFC_MODIFY, OFPFC_MODIFY_STRICT, OFPFC_DELETE,
548 and OFPFC_DELETE_STRICT flow_mods, plus flow stats requests
549 and aggregate stats requests, to match on flow cookies with
550 arbitrary masks. This is much like the equivalent OpenFlow
553 - Like OpenFlow 1.1, OFPC_MODIFY and OFPFC_MODIFY_STRICT add a
554 new flow if there is no match and the mask is zero (or not
557 - The "cookie" field in OFPT_FLOW_MOD and NXT_FLOW_MOD messages
558 is used as the cookie value for OFPFC_ADD commands, as
559 described in OpenFlow 1.0. For OFPFC_MODIFY and
560 OFPFC_MODIFY_STRICT commands, the "cookie" field is used as a
561 new cookie for flows that match unless it is UINT64_MAX, in
562 which case the flow's cookie is not updated.
564 - NXT_PACKET_IN (the Nicira extended version of
565 OFPT_PACKET_IN) reports the cookie of the rule that
566 generated the packet, or all-1-bits if no rule generated the
567 packet. (Older versions of OVS used all-0-bits instead of
570 The following table shows the handling of different protocols when
571 receiving OFPFC_MODIFY and OFPFC_MODIFY_STRICT messages. A mask of 0
572 indicates either an explicit mask of zero or an implicit one by not
573 specifying the NXM_NX_COOKIE(_W) field.
576 Match Update Add on miss Add on miss
577 cookie cookie mask!=0 mask==0
578 ====== ====== =========== ===========
579 OpenFlow 1.0 no yes <always add on miss>
580 OpenFlow 1.1 yes no no yes
581 OpenFlow 1.2 yes no no no
584 * Updates the flow's cookie unless the "cookie" field is UINT64_MAX.
587 Multiple Table Support
588 ======================
590 OpenFlow 1.0 has only rudimentary support for multiple flow tables.
591 Notably, OpenFlow 1.0 does not allow the controller to specify the
592 flow table to which a flow is to be added. Open vSwitch adds an
593 extension for this purpose, which is enabled on a per-OpenFlow
594 connection basis using the NXT_FLOW_MOD_TABLE_ID message. When the
595 extension is enabled, the upper 8 bits of the 'command' member in an
596 OFPT_FLOW_MOD or NXT_FLOW_MOD message designates the table to which a
599 The Open vSwitch software switch implementation offers 255 flow
600 tables. On packet ingress, only the first flow table (table 0) is
601 searched, and the contents of the remaining tables are not considered
602 in any way. Tables other than table 0 only come into play when an
603 NXAST_RESUBMIT_TABLE action specifies another table to search.
605 Tables 128 and above are reserved for use by the switch itself.
606 Controllers should use only tables 0 through 127.
612 Open vSwitch supports stateless handling of IPv6 packets. Flows can be
613 written to support matching TCP, UDP, and ICMPv6 headers within an IPv6
614 packet. Deeper matching of some Neighbor Discovery messages is also
617 IPv6 was not designed to interact well with middle-boxes. This,
618 combined with Open vSwitch's stateless nature, have affected the
619 processing of IPv6 traffic, which is detailed below.
624 The base IPv6 header is incredibly simple with the intention of only
625 containing information relevant for routing packets between two
626 endpoints. IPv6 relies heavily on the use of extension headers to
627 provide any other functionality. Unfortunately, the extension headers
628 were designed in such a way that it is impossible to move to the next
629 header (including the layer-4 payload) unless the current header is
632 Open vSwitch will process the following extension headers and continue
635 * Fragment (see the next section)
636 * AH (Authentication Header)
639 * Destination Options
641 When a header is encountered that is not in that list, it is considered
642 "terminal". A terminal header's IPv6 protocol value is stored in
643 "nw_proto" for matching purposes. If a terminal header is TCP, UDP, or
644 ICMPv6, the packet will be further processed in an attempt to extract
650 IPv6 requires that every link in the internet have an MTU of 1280 octets
651 or greater (RFC 2460). As such, a terminal header (as described above in
652 "Extension Headers") in the first fragment should generally be
653 reachable. In this case, the terminal header's IPv6 protocol type is
654 stored in the "nw_proto" field for matching purposes. If a terminal
655 header cannot be found in the first fragment (one with a fragment offset
656 of zero), the "nw_proto" field is set to 0. Subsequent fragments (those
657 with a non-zero fragment offset) have the "nw_proto" field set to the
658 IPv6 protocol type for fragments (44).
663 An IPv6 jumbogram (RFC 2675) is a packet containing a payload longer
664 than 65,535 octets. A jumbogram is only relevant in subnets with a link
665 MTU greater than 65,575 octets, and are not required to be supported on
666 nodes that do not connect to link with such large MTUs. Currently, Open
667 vSwitch doesn't process jumbograms.
676 An OpenFlow switch must establish and maintain a TCP network
677 connection to its controller. There are two basic ways to categorize
678 the network that this connection traverses: either it is completely
679 separate from the one that the switch is otherwise controlling, or its
680 path may overlap the network that the switch controls. We call the
681 former case "out-of-band control", the latter case "in-band control".
683 Out-of-band control has the following benefits:
685 - Simplicity: Out-of-band control slightly simplifies the switch
688 - Reliability: Excessive switch traffic volume cannot interfere
689 with control traffic.
691 - Integrity: Machines not on the control network cannot
692 impersonate a switch or a controller.
694 - Confidentiality: Machines not on the control network cannot
695 snoop on control traffic.
697 In-band control, on the other hand, has the following advantages:
699 - No dedicated port: There is no need to dedicate a physical
700 switch port to control, which is important on switches that have
701 few ports (e.g. wireless routers, low-end embedded platforms).
703 - No dedicated network: There is no need to build and maintain a
704 separate control network. This is important in many
705 environments because it reduces proliferation of switches and
708 Open vSwitch supports both out-of-band and in-band control. This
709 section describes the principles behind in-band control. See the
710 description of the Controller table in ovs-vswitchd.conf.db(5) to
711 configure OVS for in-band control.
716 The fundamental principle of in-band control is that an OpenFlow
717 switch must recognize and switch control traffic without involving the
718 OpenFlow controller. All the details of implementing in-band control
719 are special cases of this principle.
721 The rationale for this principle is simple. If the switch does not
722 handle in-band control traffic itself, then it will be caught in a
723 contradiction: it must contact the controller, but it cannot, because
724 only the controller can set up the flows that are needed to contact
727 The following points describe important special cases of this
730 - In-band control must be implemented regardless of whether the
733 It is tempting to implement the in-band control rules only when
734 the switch is not connected to the controller, using the
735 reasoning that the controller should have complete control once
736 it has established a connection with the switch.
738 This does not work in practice. Consider the case where the
739 switch is connected to the controller. Occasionally it can
740 happen that the controller forgets or otherwise needs to obtain
741 the MAC address of the switch. To do so, the controller sends a
742 broadcast ARP request. A switch that implements the in-band
743 control rules only when it is disconnected will then send an
744 OFPT_PACKET_IN message up to the controller. The controller will
745 be unable to respond, because it does not know the MAC address of
746 the switch. This is a deadlock situation that can only be
747 resolved by the switch noticing that its connection to the
748 controller has hung and reconnecting.
750 - In-band control must override flows set up by the controller.
752 It is reasonable to assume that flows set up by the OpenFlow
753 controller should take precedence over in-band control, on the
754 basis that the controller should be in charge of the switch.
756 Again, this does not work in practice. Reasonable controller
757 implementations may set up a "last resort" fallback rule that
758 wildcards every field and, e.g., sends it up to the controller or
759 discards it. If a controller does that, then it will isolate
760 itself from the switch.
762 - The switch must recognize all control traffic.
764 The fundamental principle of in-band control states, in part,
765 that a switch must recognize control traffic without involving
766 the OpenFlow controller. More specifically, the switch must
767 recognize *all* control traffic. "False negatives", that is,
768 packets that constitute control traffic but that the switch does
769 not recognize as control traffic, lead to control traffic storms.
771 Consider an OpenFlow switch that only recognizes control packets
772 sent to or from that switch. Now suppose that two switches of
773 this type, named A and B, are connected to ports on an Ethernet
774 hub (not a switch) and that an OpenFlow controller is connected
775 to a third hub port. In this setup, control traffic sent by
776 switch A will be seen by switch B, which will send it to the
777 controller as part of an OFPT_PACKET_IN message. Switch A will
778 then see the OFPT_PACKET_IN message's packet, re-encapsulate it
779 in another OFPT_PACKET_IN, and send it to the controller. Switch
780 B will then see that OFPT_PACKET_IN, and so on in an infinite
783 Incidentally, the consequences of "false positives", where
784 packets that are not control traffic are nevertheless recognized
785 as control traffic, are much less severe. The controller will
786 not be able to control their behavior, but the network will
787 remain in working order. False positives do constitute a
790 - The switch should use echo-requests to detect disconnection.
792 TCP will notice that a connection has hung, but this can take a
793 considerable amount of time. For example, with default settings
794 the Linux kernel TCP implementation will retransmit for between
795 13 and 30 minutes, depending on the connection's retransmission
796 timeout, according to kernel documentation. This is far too long
797 for a switch to be disconnected, so an OpenFlow switch should
798 implement its own connection timeout. OpenFlow OFPT_ECHO_REQUEST
799 messages are the best way to do this, since they test the
800 OpenFlow connection itself.
805 This section describes how Open vSwitch implements in-band control.
806 Correctly implementing in-band control has proven difficult due to its
807 many subtleties, and has thus gone through many iterations. Please
808 read through and understand the reasoning behind the chosen rules
809 before making modifications.
811 Open vSwitch implements in-band control as "hidden" flows, that is,
812 flows that are not visible through OpenFlow, and at a higher priority
813 than wildcarded flows can be set up through OpenFlow. This is done so
814 that the OpenFlow controller cannot interfere with them and possibly
815 break connectivity with its switches. It is possible to see all
816 flows, including in-band ones, with the ovs-appctl "bridge/dump-flows"
819 The Open vSwitch implementation of in-band control can hide traffic to
820 arbitrary "remotes", where each remote is one TCP port on one IP address.
821 Currently the remotes are automatically configured as the in-band OpenFlow
822 controllers plus the OVSDB managers, if any. (The latter is a requirement
823 because OVSDB managers are responsible for configuring OpenFlow controllers,
824 so if the manager cannot be reached then OpenFlow cannot be reconfigured.)
826 The following rules (with the OFPP_NORMAL action) are set up on any bridge
827 that has any remotes:
829 (a) DHCP requests sent from the local port.
830 (b) ARP replies to the local port's MAC address.
831 (c) ARP requests from the local port's MAC address.
833 In-band also sets up the following rules for each unique next-hop MAC
834 address for the remotes' IPs (the "next hop" is either the remote
835 itself, if it is on a local subnet, or the gateway to reach the remote):
837 (d) ARP replies to the next hop's MAC address.
838 (e) ARP requests from the next hop's MAC address.
840 In-band also sets up the following rules for each unique remote IP address:
842 (f) ARP replies containing the remote's IP address as a target.
843 (g) ARP requests containing the remote's IP address as a source.
845 In-band also sets up the following rules for each unique remote (IP,port)
848 (h) TCP traffic to the remote's IP and port.
849 (i) TCP traffic from the remote's IP and port.
851 The goal of these rules is to be as narrow as possible to allow a
852 switch to join a network and be able to communicate with the
853 remotes. As mentioned earlier, these rules have higher priority
854 than the controller's rules, so if they are too broad, they may
855 prevent the controller from implementing its policy. As such,
856 in-band actively monitors some aspects of flow and packet processing
857 so that the rules can be made more precise.
859 In-band control monitors attempts to add flows into the datapath that
860 could interfere with its duties. The datapath only allows exact
861 match entries, so in-band control is able to be very precise about
862 the flows it prevents. Flows that miss in the datapath are sent to
863 userspace to be processed, so preventing these flows from being
864 cached in the "fast path" does not affect correctness. The only type
865 of flow that is currently prevented is one that would prevent DHCP
866 replies from being seen by the local port. For example, a rule that
867 forwarded all DHCP traffic to the controller would not be allowed,
868 but one that forwarded to all ports (including the local port) would.
870 As mentioned earlier, packets that miss in the datapath are sent to
871 the userspace for processing. The userspace has its own flow table,
872 the "classifier", so in-band checks whether any special processing
873 is needed before the classifier is consulted. If a packet is a DHCP
874 response to a request from the local port, the packet is forwarded to
875 the local port, regardless of the flow table. Note that this requires
876 L7 processing of DHCP replies to determine whether the 'chaddr' field
877 matches the MAC address of the local port.
879 It is interesting to note that for an L3-based in-band control
880 mechanism, the majority of rules are devoted to ARP traffic. At first
881 glance, some of these rules appear redundant. However, each serves an
882 important role. First, in order to determine the MAC address of the
883 remote side (controller or gateway) for other ARP rules, we must allow
884 ARP traffic for our local port with rules (b) and (c). If we are
885 between a switch and its connection to the remote, we have to
886 allow the other switch's ARP traffic to through. This is done with
887 rules (d) and (e), since we do not know the addresses of the other
888 switches a priori, but do know the remote's or gateway's. Finally,
889 if the remote is running in a local guest VM that is not reached
890 through the local port, the switch that is connected to the VM must
891 allow ARP traffic based on the remote's IP address, since it will
892 not know the MAC address of the local port that is sending the traffic
893 or the MAC address of the remote in the guest VM.
895 With a few notable exceptions below, in-band should work in most
896 network setups. The following are considered "supported" in the
897 current implementation:
899 - Locally Connected. The switch and remote are on the same
900 subnet. This uses rules (a), (b), (c), (h), and (i).
902 - Reached through Gateway. The switch and remote are on
903 different subnets and must go through a gateway. This uses
904 rules (a), (b), (c), (h), and (i).
906 - Between Switch and Remote. This switch is between another
907 switch and the remote, and we want to allow the other
908 switch's traffic through. This uses rules (d), (e), (h), and
909 (i). It uses (b) and (c) indirectly in order to know the MAC
910 address for rules (d) and (e). Note that DHCP for the other
911 switch will not work unless an OpenFlow controller explicitly lets this
912 switch pass the traffic.
914 - Between Switch and Gateway. This switch is between another
915 switch and the gateway, and we want to allow the other switch's
916 traffic through. This uses the same rules and logic as the
917 "Between Switch and Remote" configuration described earlier.
919 - Remote on Local VM. The remote is a guest VM on the
920 system running in-band control. This uses rules (a), (b), (c),
923 - Remote on Local VM with Different Networks. The remote
924 is a guest VM on the system running in-band control, but the
925 local port is not used to connect to the remote. For
926 example, an IP address is configured on eth0 of the switch. The
927 remote's VM is connected through eth1 of the switch, but an
928 IP address has not been configured for that port on the switch.
929 As such, the switch will use eth0 to connect to the remote,
930 and eth1's rules about the local port will not work. In the
931 example, the switch attached to eth0 would use rules (a), (b),
932 (c), (h), and (i) on eth0. The switch attached to eth1 would use
933 rules (f), (g), (h), and (i).
935 The following are explicitly *not* supported by in-band control:
937 - Specify Remote by Name. Currently, the remote must be
938 identified by IP address. A naive approach would be to permit
939 all DNS traffic. Unfortunately, this would prevent the
940 controller from defining any policy over DNS. Since switches
941 that are located behind us need to connect to the remote,
942 in-band cannot simply add a rule that allows DNS traffic from
943 the local port. The "correct" way to support this is to parse
944 DNS requests to allow all traffic related to a request for the
945 remote's name through. Due to the potential security
946 problems and amount of processing, we decided to hold off for
949 - Differing Remotes for Switches. All switches must know
950 the L3 addresses for all the remotes that other switches
951 may use, since rules need to be set up to allow traffic related
952 to those remotes through. See rules (f), (g), (h), and (i).
954 - Differing Routes for Switches. In order for the switch to
955 allow other switches to connect to a remote through a
956 gateway, it allows the gateway's traffic through with rules (d)
957 and (e). If the routes to the remote differ for the two
958 switches, we will not know the MAC address of the alternate
965 It seems likely that many controllers, at least at startup, use the
966 OpenFlow "flow statistics" request to obtain existing flows, then
967 compare the flows' actions against the actions that they expect to
968 find. Before version 1.8.0, Open vSwitch always returned exact,
969 byte-for-byte copies of the actions that had been added to the flow
970 table. The current version of Open vSwitch does not always do this in
971 some exceptional cases. This section lists the exceptions that
972 controller authors must keep in mind if they compare actual actions
973 against desired actions in a bytewise fashion:
975 - Open vSwitch zeros padding bytes in action structures,
976 regardless of their values when the flows were added.
978 - Open vSwitch "normalizes" the instructions in OpenFlow 1.1
979 (and later) in the following way:
981 * OVS sorts the instructions into the following order:
982 Apply-Actions, Clear-Actions, Write-Actions,
983 Write-Metadata, Goto-Table.
985 * OVS drops Apply-Actions instructions that have empty
988 * OVS drops Write-Actions instructions that have empty
991 Please report other discrepancies, if you notice any, so that we can
992 fix or document them.
998 Suggestions to improve Open vSwitch are welcome at discuss@openvswitch.org.