1 Distributed Switch Architecture
2 ===============================
7 This document describes the Distributed Switch Architecture (DSA) subsystem
8 design principles, limitations, interactions with other subsystems, and how to
9 develop drivers for this subsystem as well as a TODO for developers interested
10 in joining the effort.
15 The Distributed Switch Architecture is a subsystem which was primarily designed
16 to support Marvell Ethernet switches (MV88E6xxx, a.k.a Linkstreet product line)
17 using Linux, but has since evolved to support other vendors as well.
19 The original philosophy behind this design was to be able to use unmodified
20 Linux tools such as bridge, iproute2, ifconfig to work transparently whether
21 they configured/queried a switch port network device or a regular network
24 An Ethernet switch is typically comprised of multiple front-panel ports, and one
25 or more CPU or management port. The DSA subsystem currently relies on the
26 presence of a management port connected to an Ethernet controller capable of
27 receiving Ethernet frames from the switch. This is a very common setup for all
28 kinds of Ethernet switches found in Small Home and Office products: routers,
29 gateways, or even top-of-the rack switches. This host Ethernet controller will
30 be later referred to as "master" and "cpu" in DSA terminology and code.
32 The D in DSA stands for Distributed, because the subsystem has been designed
33 with the ability to configure and manage cascaded switches on top of each other
34 using upstream and downstream Ethernet links between switches. These specific
35 ports are referred to as "dsa" ports in DSA terminology and code. A collection
36 of multiple switches connected to each other is called a "switch tree".
38 For each front-panel port, DSA will create specialized network devices which are
39 used as controlling and data-flowing endpoints for use by the Linux networking
40 stack. These specialized network interfaces are referred to as "slave" network
41 interfaces in DSA terminology and code.
43 The ideal case for using DSA is when an Ethernet switch supports a "switch tag"
44 which is a hardware feature making the switch insert a specific tag for each
45 Ethernet frames it received to/from specific ports to help the management
48 - what port is this frame coming from
49 - what was the reason why this frame got forwarded
50 - how to send CPU originated traffic to specific ports
52 The subsystem does support switches not capable of inserting/stripping tags, but
53 the features might be slightly limited in that case (traffic separation relies
54 on Port-based VLAN IDs).
56 Note that DSA does not currently create network interfaces for the "cpu" and
59 - the "cpu" port is the Ethernet switch facing side of the management
60 controller, and as such, would create a duplication of feature, since you
61 would get two interfaces for the same conduit: master netdev, and "cpu" netdev
63 - the "dsa" port(s) are just conduits between two or more switches, and as such
64 cannot really be used as proper network interfaces either, only the
65 downstream, or the top-most upstream interface makes sense with that model
67 Switch tagging protocols
68 ------------------------
70 DSA currently supports 5 different tagging protocols, and a tag-less mode as
71 well. The different protocols are implemented in:
73 net/dsa/tag_trailer.c: Marvell's 4 trailer tag mode (legacy)
74 net/dsa/tag_dsa.c: Marvell's original DSA tag
75 net/dsa/tag_edsa.c: Marvell's enhanced DSA tag
76 net/dsa/tag_brcm.c: Broadcom's 4 bytes tag
77 net/dsa/tag_qca.c: Qualcomm's 2 bytes tag
79 The exact format of the tag protocol is vendor specific, but in general, they
80 all contain something which:
82 - identifies which port the Ethernet frame came from/should be sent to
83 - provides a reason why this frame was forwarded to the management interface
85 Master network devices
86 ----------------------
88 Master network devices are regular, unmodified Linux network device drivers for
89 the CPU/management Ethernet interface. Such a driver might occasionally need to
90 know whether DSA is enabled (e.g.: to enable/disable specific offload features),
91 but the DSA subsystem has been proven to work with industry standard drivers:
92 e1000e, mv643xx_eth etc. without having to introduce modifications to these
93 drivers. Such network devices are also often referred to as conduit network
94 devices since they act as a pipe between the host processor and the hardware
97 Networking stack hooks
98 ----------------------
100 When a master netdev is used with DSA, a small hook is placed in in the
101 networking stack is in order to have the DSA subsystem process the Ethernet
102 switch specific tagging protocol. DSA accomplishes this by registering a
103 specific (and fake) Ethernet type (later becoming skb->protocol) with the
104 networking stack, this is also known as a ptype or packet_type. A typical
105 Ethernet Frame receive sequence looks like this:
107 Master network device (e.g.: e1000e):
109 Receive interrupt fires:
110 - receive function is invoked
111 - basic packet processing is done: getting length, status etc.
112 - packet is prepared to be processed by the Ethernet layer by calling
117 eth_type_trans(skb, dev)
118 if (dev->dsa_ptr != NULL)
119 -> skb->protocol = ETH_P_XDSA
121 drivers/net/ethernet/*:
123 netif_receive_skb(skb)
124 -> iterate over registered packet_type
125 -> invoke handler for ETH_P_XDSA, calls dsa_switch_rcv()
129 -> invoke switch tag specific protocol handler in
133 -> inspect and strip switch tag protocol to determine originating port
134 -> locate per-port network device
135 -> invoke eth_type_trans() with the DSA slave network device
136 -> invoked netif_receive_skb()
138 Past this point, the DSA slave network devices get delivered regular Ethernet
139 frames that can be processed by the networking stack.
141 Slave network devices
142 ---------------------
144 Slave network devices created by DSA are stacked on top of their master network
145 device, each of these network interfaces will be responsible for being a
146 controlling and data-flowing end-point for each front-panel port of the switch.
147 These interfaces are specialized in order to:
149 - insert/remove the switch tag protocol (if it exists) when sending traffic
150 to/from specific switch ports
151 - query the switch for ethtool operations: statistics, link state,
152 Wake-on-LAN, register dumps...
153 - external/internal PHY management: link, auto-negotiation etc.
155 These slave network devices have custom net_device_ops and ethtool_ops function
156 pointers which allow DSA to introduce a level of layering between the networking
157 stack/ethtool, and the switch driver implementation.
159 Upon frame transmission from these slave network devices, DSA will look up which
160 switch tagging protocol is currently registered with these network devices, and
161 invoke a specific transmit routine which takes care of adding the relevant
162 switch tag in the Ethernet frames.
164 These frames are then queued for transmission using the master network device
165 ndo_start_xmit() function, since they contain the appropriate switch tag, the
166 Ethernet switch will be able to process these incoming frames from the
167 management interface and delivers these frames to the physical switch port.
169 Graphical representation
170 ------------------------
172 Summarized, this is basically how DSA looks like from a network device
176 |---------------------------
177 | CPU network device (eth0)|
178 ----------------------------
179 | <tag added by switch |
182 | tag added by CPU> |
183 |--------------------------------------------|
185 |--------------------------------------------|
187 |-------| |-------| |-------|
188 | sw0p0 | | sw0p1 | | sw0p2 |
189 |-------| |-------| |-------|
194 In order to be able to read to/from a switch PHY built into it, DSA creates a
195 slave MDIO bus which allows a specific switch driver to divert and intercept
196 MDIO reads/writes towards specific PHY addresses. In most MDIO-connected
197 switches, these functions would utilize direct or indirect PHY addressing mode
198 to return standard MII registers from the switch builtin PHYs, allowing the PHY
199 library and/or to return link status, link partner pages, auto-negotiation
202 For Ethernet switches which have both external and internal MDIO busses, the
203 slave MII bus can be utilized to mux/demux MDIO reads and writes towards either
204 internal or external MDIO devices this switch might be connected to: internal
205 PHYs, external PHYs, or even external switches.
210 DSA data structures are defined in include/net/dsa.h as well as
213 dsa_chip_data: platform data configuration for a given switch device, this
214 structure describes a switch device's parent device, its address, as well as
215 various properties of its ports: names/labels, and finally a routing table
216 indication (when cascading switches)
218 dsa_platform_data: platform device configuration data which can reference a
219 collection of dsa_chip_data structure if multiples switches are cascaded, the
220 master network device this switch tree is attached to needs to be referenced
222 dsa_switch_tree: structure assigned to the master network device under
223 "dsa_ptr", this structure references a dsa_platform_data structure as well as
224 the tagging protocol supported by the switch tree, and which receive/transmit
225 function hooks should be invoked, information about the directly attached switch
226 is also provided: CPU port. Finally, a collection of dsa_switch are referenced
227 to address individual switches in the tree.
229 dsa_switch: structure describing a switch device in the tree, referencing a
230 dsa_switch_tree as a backpointer, slave network devices, master network device,
231 and a reference to the backing dsa_switch_ops
233 dsa_switch_ops: structure referencing function pointers, see below for a full
239 Limits on the number of devices and ports
240 -----------------------------------------
242 DSA currently limits the number of maximum switches within a tree to 4
243 (DSA_MAX_SWITCHES), and the number of ports per switch to 12 (DSA_MAX_PORTS).
244 These limits could be extended to support larger configurations would this need
247 Lack of CPU/DSA network devices
248 -------------------------------
250 DSA does not currently create slave network devices for the CPU or DSA ports, as
251 described before. This might be an issue in the following cases:
253 - inability to fetch switch CPU port statistics counters using ethtool, which
254 can make it harder to debug MDIO switch connected using xMII interfaces
256 - inability to configure the CPU port link parameters based on the Ethernet
257 controller capabilities attached to it: http://patchwork.ozlabs.org/patch/509806/
259 - inability to configure specific VLAN IDs / trunking VLANs between switches
260 when using a cascaded setup
262 Common pitfalls using DSA setups
263 --------------------------------
265 Once a master network device is configured to use DSA (dev->dsa_ptr becomes
266 non-NULL), and the switch behind it expects a tagging protocol, this network
267 interface can only exclusively be used as a conduit interface. Sending packets
268 directly through this interface (e.g.: opening a socket using this interface)
269 will not make us go through the switch tagging protocol transmit function, so
270 the Ethernet switch on the other end, expecting a tag will typically drop this
273 Slave network devices check that the master network device is UP before allowing
274 you to administratively bring UP these slave network devices. A common
275 configuration mistake is forgetting to bring UP the master network device first.
277 Interactions with other subsystems
278 ==================================
280 DSA currently leverages the following subsystems:
282 - MDIO/PHY library: drivers/net/phy/phy.c, mdio_bus.c
283 - Switchdev: net/switchdev/*
284 - Device Tree for various of_* functions
289 Slave network devices exposed by DSA may or may not be interfacing with PHY
290 devices (struct phy_device as defined in include/linux/phy.h), but the DSA
291 subsystem deals with all possible combinations:
293 - internal PHY devices, built into the Ethernet switch hardware
294 - external PHY devices, connected via an internal or external MDIO bus
295 - internal PHY devices, connected via an internal MDIO bus
296 - special, non-autonegotiated or non MDIO-managed PHY devices: SFPs, MoCA; a.k.a
299 The PHY configuration is done by the dsa_slave_phy_setup() function and the
300 logic basically looks like this:
302 - if Device Tree is used, the PHY device is looked up using the standard
303 "phy-handle" property, if found, this PHY device is created and registered
304 using of_phy_connect()
306 - if Device Tree is used, and the PHY device is "fixed", that is, conforms to
307 the definition of a non-MDIO managed PHY as defined in
308 Documentation/devicetree/bindings/net/fixed-link.txt, the PHY is registered
309 and connected transparently using the special fixed MDIO bus driver
311 - finally, if the PHY is built into the switch, as is very common with
312 standalone switch packages, the PHY is probed using the slave MII bus created
319 DSA directly utilizes SWITCHDEV when interfacing with the bridge layer, and
320 more specifically with its VLAN filtering portion when configuring VLANs on top
321 of per-port slave network devices. Since DSA primarily deals with
322 MDIO-connected switches, although not exclusively, SWITCHDEV's
323 prepare/abort/commit phases are often simplified into a prepare phase which
324 checks whether the operation is supported by the DSA switch driver, and a commit
325 phase which applies the changes.
327 As of today, the only SWITCHDEV objects supported by DSA are the FDB and VLAN
333 DSA features a standardized binding which is documented in
334 Documentation/devicetree/bindings/net/dsa/dsa.txt. PHY/MDIO library helper
335 functions such as of_get_phy_mode(), of_phy_connect() are also used to query
336 per-port PHY specific details: interface connection, MDIO bus location etc..
341 DSA switch drivers need to implement a dsa_switch_ops structure which will
342 contain the various members described below.
344 register_switch_driver() registers this dsa_switch_ops in its internal list
345 of drivers to probe for. unregister_switch_driver() does the exact opposite.
347 Unless requested differently by setting the priv_size member accordingly, DSA
348 does not allocate any driver private context space.
353 - tag_protocol: this is to indicate what kind of tagging protocol is supported,
354 should be a valid value from the dsa_tag_protocol enum
356 - probe: probe routine which will be invoked by the DSA platform device upon
357 registration to test for the presence/absence of a switch device. For MDIO
358 devices, it is recommended to issue a read towards internal registers using
359 the switch pseudo-PHY and return whether this is a supported device. For other
360 buses, return a non-NULL string
362 - setup: setup function for the switch, this function is responsible for setting
363 up the dsa_switch_ops private structure with all it needs: register maps,
364 interrupts, mutexes, locks etc.. This function is also expected to properly
365 configure the switch to separate all network interfaces from each other, that
366 is, they should be isolated by the switch hardware itself, typically by creating
367 a Port-based VLAN ID for each port and allowing only the CPU port and the
368 specific port to be in the forwarding vector. Ports that are unused by the
369 platform should be disabled. Past this function, the switch is expected to be
370 fully configured and ready to serve any kind of request. It is recommended
371 to issue a software reset of the switch during this setup function in order to
372 avoid relying on what a previous software agent such as a bootloader/firmware
373 may have previously configured.
375 PHY devices and link management
376 -------------------------------
378 - get_phy_flags: Some switches are interfaced to various kinds of Ethernet PHYs,
379 if the PHY library PHY driver needs to know about information it cannot obtain
380 on its own (e.g.: coming from switch memory mapped registers), this function
381 should return a 32-bits bitmask of "flags", that is private between the switch
382 driver and the Ethernet PHY driver in drivers/net/phy/*.
384 - phy_read: Function invoked by the DSA slave MDIO bus when attempting to read
385 the switch port MDIO registers. If unavailable, return 0xffff for each read.
386 For builtin switch Ethernet PHYs, this function should allow reading the link
387 status, auto-negotiation results, link partner pages etc..
389 - phy_write: Function invoked by the DSA slave MDIO bus when attempting to write
390 to the switch port MDIO registers. If unavailable return a negative error
393 - adjust_link: Function invoked by the PHY library when a slave network device
394 is attached to a PHY device. This function is responsible for appropriately
395 configuring the switch port link parameters: speed, duplex, pause based on
396 what the phy_device is providing.
398 - fixed_link_update: Function invoked by the PHY library, and specifically by
399 the fixed PHY driver asking the switch driver for link parameters that could
400 not be auto-negotiated, or obtained by reading the PHY registers through MDIO.
401 This is particularly useful for specific kinds of hardware such as QSGMII,
402 MoCA or other kinds of non-MDIO managed PHYs where out of band link
403 information is obtained
408 - get_strings: ethtool function used to query the driver's strings, will
409 typically return statistics strings, private flags strings etc.
411 - get_ethtool_stats: ethtool function used to query per-port statistics and
412 return their values. DSA overlays slave network devices general statistics:
413 RX/TX counters from the network device, with switch driver specific statistics
416 - get_sset_count: ethtool function used to query the number of statistics items
418 - get_wol: ethtool function used to obtain Wake-on-LAN settings per-port, this
419 function may, for certain implementations also query the master network device
420 Wake-on-LAN settings if this interface needs to participate in Wake-on-LAN
422 - set_wol: ethtool function used to configure Wake-on-LAN settings per-port,
423 direct counterpart to set_wol with similar restrictions
425 - set_eee: ethtool function which is used to configure a switch port EEE (Green
426 Ethernet) settings, can optionally invoke the PHY library to enable EEE at the
427 PHY level if relevant. This function should enable EEE at the switch port MAC
428 controller and data-processing logic
430 - get_eee: ethtool function which is used to query a switch port EEE settings,
431 this function should return the EEE state of the switch port MAC controller
432 and data-processing logic as well as query the PHY for its currently configured
435 - get_eeprom_len: ethtool function returning for a given switch the EEPROM
438 - get_eeprom: ethtool function returning for a given switch the EEPROM contents
440 - set_eeprom: ethtool function writing specified data to a given switch EEPROM
442 - get_regs_len: ethtool function returning the register length for a given
445 - get_regs: ethtool function returning the Ethernet switch internal register
446 contents. This function might require user-land code in ethtool to
447 pretty-print register values and registers
452 - suspend: function invoked by the DSA platform device when the system goes to
453 suspend, should quiesce all Ethernet switch activities, but keep ports
454 participating in Wake-on-LAN active as well as additional wake-up logic if
457 - resume: function invoked by the DSA platform device when the system resumes,
458 should resume all Ethernet switch activities and re-configure the switch to be
459 in a fully active state
461 - port_enable: function invoked by the DSA slave network device ndo_open
462 function when a port is administratively brought up, this function should be
463 fully enabling a given switch port. DSA takes care of marking the port with
464 BR_STATE_BLOCKING if the port is a bridge member, or BR_STATE_FORWARDING if it
465 was not, and propagating these changes down to the hardware
467 - port_disable: function invoked by the DSA slave network device ndo_close
468 function when a port is administratively brought down, this function should be
469 fully disabling a given switch port. DSA takes care of marking the port with
470 BR_STATE_DISABLED and propagating changes to the hardware if this port is
471 disabled while being a bridge member
476 - port_bridge_join: bridge layer function invoked when a given switch port is
477 added to a bridge, this function should be doing the necessary at the switch
478 level to permit the joining port from being added to the relevant logical
479 domain for it to ingress/egress traffic with other members of the bridge.
481 - port_bridge_leave: bridge layer function invoked when a given switch port is
482 removed from a bridge, this function should be doing the necessary at the
483 switch level to deny the leaving port from ingress/egress traffic from the
484 remaining bridge members. When the port leaves the bridge, it should be aged
485 out at the switch hardware for the switch to (re) learn MAC addresses behind
488 - port_stp_state_set: bridge layer function invoked when a given switch port STP
489 state is computed by the bridge layer and should be propagated to switch
490 hardware to forward/block/learn traffic. The switch driver is responsible for
491 computing a STP state change based on current and asked parameters and perform
492 the relevant ageing based on the intersection results
494 Bridge VLAN filtering
495 ---------------------
497 - port_vlan_filtering: bridge layer function invoked when the bridge gets
498 configured for turning on or off VLAN filtering. If nothing specific needs to
499 be done at the hardware level, this callback does not need to be implemented.
500 When VLAN filtering is turned on, the hardware must be programmed with
501 rejecting 802.1Q frames which have VLAN IDs outside of the programmed allowed
502 VLAN ID map/rules. If there is no PVID programmed into the switch port,
503 untagged frames must be rejected as well. When turned off the switch must
504 accept any 802.1Q frames irrespective of their VLAN ID, and untagged frames are
507 - port_vlan_prepare: bridge layer function invoked when the bridge prepares the
508 configuration of a VLAN on the given port. If the operation is not supported
509 by the hardware, this function should return -EOPNOTSUPP to inform the bridge
510 code to fallback to a software implementation. No hardware setup must be done
511 in this function. See port_vlan_add for this and details.
513 - port_vlan_add: bridge layer function invoked when a VLAN is configured
514 (tagged or untagged) for the given switch port
516 - port_vlan_del: bridge layer function invoked when a VLAN is removed from the
519 - port_vlan_dump: bridge layer function invoked with a switchdev callback
520 function that the driver has to call for each VLAN the given port is a member
521 of. A switchdev object is used to carry the VID and bridge flags.
523 - port_fdb_add: bridge layer function invoked when the bridge wants to install a
524 Forwarding Database entry, the switch hardware should be programmed with the
525 specified address in the specified VLAN Id in the forwarding database
526 associated with this VLAN ID. If the operation is not supported, this
527 function should return -EOPNOTSUPP to inform the bridge code to fallback to
528 a software implementation.
530 Note: VLAN ID 0 corresponds to the port private database, which, in the context
531 of DSA, would be the its port-based VLAN, used by the associated bridge device.
533 - port_fdb_del: bridge layer function invoked when the bridge wants to remove a
534 Forwarding Database entry, the switch hardware should be programmed to delete
535 the specified MAC address from the specified VLAN ID if it was mapped into
536 this port forwarding database
538 - port_fdb_dump: bridge layer function invoked with a switchdev callback
539 function that the driver has to call for each MAC address known to be behind
540 the given port. A switchdev object is used to carry the VID and FDB info.
542 - port_mdb_prepare: bridge layer function invoked when the bridge prepares the
543 installation of a multicast database entry. If the operation is not supported,
544 this function should return -EOPNOTSUPP to inform the bridge code to fallback
545 to a software implementation. No hardware setup must be done in this function.
546 See port_fdb_add for this and details.
548 - port_mdb_add: bridge layer function invoked when the bridge wants to install
549 a multicast database entry, the switch hardware should be programmed with the
550 specified address in the specified VLAN ID in the forwarding database
551 associated with this VLAN ID.
553 Note: VLAN ID 0 corresponds to the port private database, which, in the context
554 of DSA, would be the its port-based VLAN, used by the associated bridge device.
556 - port_mdb_del: bridge layer function invoked when the bridge wants to remove a
557 multicast database entry, the switch hardware should be programmed to delete
558 the specified MAC address from the specified VLAN ID if it was mapped into
559 this port forwarding database.
561 - port_mdb_dump: bridge layer function invoked with a switchdev callback
562 function that the driver has to call for each MAC address known to be behind
563 the given port. A switchdev object is used to carry the VID and MDB info.
568 Making SWITCHDEV and DSA converge towards an unified codebase
569 -------------------------------------------------------------
571 SWITCHDEV properly takes care of abstracting the networking stack with offload
572 capable hardware, but does not enforce a strict switch device driver model. On
573 the other DSA enforces a fairly strict device driver model, and deals with most
574 of the switch specific. At some point we should envision a merger between these
575 two subsystems and get the best of both worlds.
580 - making the number of ports fully dynamic and not dependent on DSA_MAX_PORTS
581 - allowing more than one CPU/management interface:
582 http://comments.gmane.org/gmane.linux.network/365657
583 - porting more drivers from other vendors:
584 http://comments.gmane.org/gmane.linux.network/365510