qemu

FORK: QEMU emulator
git clone https://git.neptards.moe/neptards/qemu.git
Log | Files | Refs | Submodules | LICENSE

rocker.txt (36775B)


      1 Rocker Network Switch Register Programming Guide
      2 Copyright (c) Scott Feldman <sfeldma@gmail.com>
      3 Copyright (c) Neil Horman <nhorman@tuxdriver.com>
      4 Version 0.11, 12/29/2014
      5 
      6 LICENSE
      7 =======
      8 
      9 This program is free software; you can redistribute it and/or modify
     10 it under the terms of the GNU General Public License as published by
     11 the Free Software Foundation; either version 2 of the License, or
     12 (at your option) any later version.
     13 
     14 This program is distributed in the hope that it will be useful,
     15 but WITHOUT ANY WARRANTY; without even the implied warranty of
     16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
     17 GNU General Public License for more details.
     18 
     19 SECTION 1: Introduction
     20 =======================
     21 
     22 Overview
     23 --------
     24 
     25 This document describes the hardware/software interface for the Rocker switch
     26 device.  The intended audience is authors of OS drivers and device emulation
     27 software.
     28 
     29 Notations and Conventions
     30 -------------------------
     31 
     32 o In register descriptions, [n:m] indicates a range from bit n to bit m,
     33 inclusive.
     34 o Use of leading 0x indicates a hexadecimal number.
     35 o Use of leading 0b indicates a binary number.
     36 o The use of RSVD or Reserved indicates that a bit or field is reserved for
     37 future use.
     38 o Field width is in bytes, unless otherwise noted.
     39 o Register are (R) read-only, (R/W) read/write, (W) write-only, or (COR) clear
     40 on read
     41 o TLV values in network-byte-order are designated with (N).
     42 
     43 
     44 SECTION 2: PCI Configuration Registers
     45 ======================================
     46 
     47 PCI Configuration Space
     48 -----------------------
     49 
     50 Each switch instance registers as a PCI device with PCI configuration space:
     51 
     52 	offset	width	description		value
     53 	---------------------------------------------
     54 	0x0	2	Vendor ID		0x1b36
     55 	0x2	2	Device ID		0x0006
     56 	0x4	4	Command/Status
     57 	0x8	1	Revision ID		0x01
     58 	0x9	3	Class code		0x2800
     59 	0xC	1	Cache line size
     60 	0xD	1	Latency timer
     61 	0xE	1	Header type
     62 	0xF	1	Built-in self test
     63 	0x10	4	Base address low
     64 	0x14	4	Base address high
     65 	0x18-28		Reserved
     66 	0x2C	2	Subsystem vendor ID	*
     67 	0x2E	2	Subsystem ID		*
     68 	0x30-38		Reserved
     69 	0x3C	1	Interrupt line
     70 	0x3D	1	Interrupt pin		0x00
     71 	0x3E	1	Min grant		0x00
     72 	0x3D	1	Max latency		0x00
     73 	0x40	1	TRDY timeout
     74 	0x41	1	Retry count
     75 	0x42	2	Reserved
     76 
     77 
     78 * Assigned by sub-system implementation
     79 
     80 SECTION 3: Memory-Mapped Register Space
     81 =======================================
     82 
     83 There are two memory-mapped BARs.  BAR0 maps device register space and is
     84 0x2000 in size.  BAR1 maps MSI-X vector and PBA tables and is also 0x2000 in
     85 size, allowing for 256 MSI-X vectors.
     86 
     87 All registers are 4 or 8 bytes long.  It is assumed host software will access 4
     88 byte registers with one 4-byte access, and 8 byte registers with either two
     89 4-byte accesses or a single 8-byte access.  In the case of two 4-byte accesses,
     90 access must be lower and then upper 4-bytes, in that order.
     91 
     92 BAR0 device register space is organized as follows:
     93 
     94 	offset		description
     95 	------------------------------------------------------
     96 	0x0000-0x000f	Bogus registers to catch misbehaving
     97 			drivers.  Writes do nothing.  Reads
     98 			back as 0xDEADBABE.
     99 	0x0010-0x00ff	Test registers
    100 	0x0300-0x03ff	General purpose registers
    101 	0x1000-0x1fff	Descriptor control
    102 
    103 Holes in register space are reserved.  Writes to reserved registers do nothing.
    104 Reads to reserved registers read back as 0.
    105 
    106 No fancy stuff like write-combining is enabled on any of the registers.
    107 
    108 BAR1 MSI-X register space is organized as follows:
    109 
    110 	offset		description
    111 	------------------------------------------------------
    112 	0x0000-0x0fff	MSI-X vector table (256 vectors total)
    113 	0x1000-0x1fff	MSI-X PBA table
    114 
    115 
    116 SECTION 4: Interrupts, DMA, and Endianness
    117 ==========================================
    118 
    119 PCI Interrupts
    120 --------------
    121 
    122 The device supports only MSI-X interrupts.  BAR1 memory-mapped region contains
    123 the MSI-X vector and PBA tables, with support for up to 256 MSI-X vectors.
    124 
    125 The vector assignment is:
    126 
    127 	vector		description
    128 	-----------------------------------------------------
    129 	0		Command descriptor ring completion
    130 	1		Event descriptor ring completion
    131 	2		Test operation completion
    132 	3		RSVD
    133 	4-255		Tx and Rx descriptor ring completion
    134 			  Tx vector is even
    135 			  Rx vector is odd
    136 
    137 A MSI-X vector table entry is 16 bytes:
    138 
    139 	field		offset	width	description
    140 	-------------------------------------------------------------
    141 	lower_addr	0x0	4	[31:2] message address[31:2]
    142 					[1:0] Rsvd (4 byte alignment
    143 						    required)
    144 	upper_addr	0x4	4	[31:19] Rsvd
    145 					[14:0] message address[46:32]
    146 	data		0x8	4	message data[31:0]
    147 	control		0xc	4	[31:1] Rsvd
    148 					[0] mask (0 = enable,
    149 						  1 = masked)
    150 
    151 Software should install the Interrupt Service Routine (ISR) before any ports
    152 are enabled or any commands are issued on the command ring.
    153 
    154 DMA Operations
    155 --------------
    156 
    157 DMA operations are used for packet DMA to/from the CPU, command and event
    158 processing.  Command processing includes statistical counters and table dumps,
    159 table insertion/deletion, and more.  Event processing provides an async
    160 notification method for device-originating events.  Each DMA operation has a
    161 set of control registers to manage a descriptor ring.  The descriptor rings are
    162 allocated from contiguous host DMA-able memory and registers specify the rings
    163 base address, size and current head and tail indices.  Software always writes
    164 the head, and hardware always writes the tail.
    165 
    166 The higher-order bit of DMA_DESC_COMP_ERR is used to mark hardware completion
    167 of a descriptor.  Software will clear this bit when posting a descriptor to the
    168 ring, and hardware will set this bit when the descriptor is complete.
    169 
    170 Descriptor ring sizes must be a power of 2 and range from 2 to 64K entries.
    171 Descriptor rings' base address must be 8-byte aligned.  Descriptors must be
    172 packed within ring.  Each descriptor in each ring must also be aligned on an 8
    173 byte boundary.  Each descriptor ring will have these registers:
    174 
    175 	DMA_DESC_xxx_BASE_ADDR, offset 0x1000 + (x * 32), 64-bit, (R/W)
    176 	DMA_DESC_xxx_SIZE, offset 0x1008 + (x * 32), 32-bit, (R/W)
    177 	DMA_DESC_xxx_HEAD, offset 0x100c + (x * 32), 32-bit, (R/W)
    178 	DMA_DESC_xxx_TAIL, offset 0x1010 + (x * 32), 32-bit, (R)
    179 	DMA_DESC_xxx_CTRL, offset 0x1014 + (x * 32), 32-bit, (W)
    180 	DMA_DESC_xxx_CREDITS, offset 0x1018 + (x * 32), 32-bit, (R/W)
    181 	DMA_DESC_xxx_RSVD1, offset 0x101c + (x * 32), 32-bit, (R/W)
    182 
    183 Where x is descriptor ring index:
    184 
    185 	index		ring
    186 	--------------------
    187 	0		CMD
    188 	1		EVENT
    189 	2		TX (port 0)
    190 	3		RX (port 0)
    191 	4		TX (port 1)
    192 	5		RX (port 1)
    193 	.
    194 	.
    195 	.
    196 	124		TX (port 61)
    197 	125		RX (port 61)
    198 	126		Resv
    199 	127		Resv
    200 
    201 Writing BASE_ADDR or SIZE will reset HEAD and TAIL to zero.  HEAD cannot be
    202 written past TAIL.  To do so would wrap the ring.  An empty ring is when HEAD
    203 == TAIL.  A full ring is when HEAD is one position behind TAIL.  Both HEAD and
    204 TAIL increment and modulo wrap at the ring size.
    205 
    206 CTRL register bits:
    207 
    208 	bit	name		description
    209 	------------------------------------------------------------------------
    210 	[0]	CTRL_RESET	Reset the descriptor ring
    211 	[1:31]	Reserved
    212 
    213 All descriptor types share some common fields:
    214 
    215 	field			width	description
    216 	-------------------------------------------------------------------
    217 	DMA_DESC_BUF_ADDR	8	Phys addr of desc payload, 8-byte
    218 					aligned
    219 	DMA_DESC_COOKIE		8	Desc cookie for completion matching,
    220 					upper-most bit is reserved
    221 	DMA_DESC_BUF_SIZE	2	Desc payload size in bytes
    222 	DMA_DESC_TLV_SIZE	2	Desc payload total size in bytes
    223 					used for TLVs.  Must be <=
    224 					DMA_DESC_BUF_SIZE.
    225 	DMA_DESC_COMP_ERR	2	Completion status of associated
    226 					desc payload.  High order bit is
    227 					clear on new descs, toggled by
    228 					hw for completed items.
    229 
    230 To support forward- and backward-compatibility, descriptor and completion
    231 payloads are specified in TLV format.  Fields are packed with Type=field name,
    232 Length=field length, and Value=field value.  Software will ignore unknown fields
    233 filled in by the switch.  Likewise, the switch will ignore unknown fields
    234 filled in by software.
    235 
    236 Descriptor payload buffer is 8-byte aligned and TLVs are 8-byte aligned.  The
    237 value within a TLV is also 8-byte aligned.  The (packed, 8 byte) TLV header is:
    238 
    239 	field	width	description
    240 	-----------------------------
    241 	type	4	TLV type
    242 	len	2	TLV value length
    243 	pad	2	Reserved
    244 
    245 The alignment requirements for descriptors and TLVs are to avoid unaligned
    246 access exceptions in software.  Note that the payload for each TLV is also
    247 8 byte aligned.
    248 
    249 Figure 1 shows an example descriptor buffer with two TLVs.
    250 
    251                   <------- 8 bytes ------->
    252 
    253   8-byte  +––––+  +–––––––––––+–––––+–––––+                     +–+
    254   align           |   type    | len | pad |    TLV#1 hdr          |
    255                   +–––––––––––+–––––+–––––+    (len=22)           |
    256                   |                       |                       |
    257                   |  value                |    TVL#1 value        |
    258                   |                       |    (padded to 8-byte  |
    259                   |                 +–––––+     alignment)        |
    260                   |                 |/////|                       |
    261    8-byte +––––+  +–––––––––––+–––––––––––+                       |
    262    align          |   type    | len | pad |    TLV#2 hdr    DESC_BUF_SIZE
    263                   +–––––+–––––+–––––+–––––+    (len=2)            |
    264                   |value|/////////////////|    TLV#2 value        |
    265                   +–––––+/////////////////|                       |
    266                   |///////////////////////|                       |
    267                   |///////////////////////|                       |
    268                   |///////////////////////|                       |
    269                   |////////unused/////////|                       |
    270                   |////////space//////////|                       |
    271                   |///////////////////////|                       |
    272                   |///////////////////////|                       |
    273                   |///////////////////////|                       |
    274                   +–––––––––––––––––––––––+                     +–+
    275 
    276 				fig. 1
    277 
    278 TLVs can be nested within the NEST TLV type.
    279 
    280 Interrupt credits
    281 ^^^^^^^^^^^^^^^^^
    282 
    283 MSI-X vectors used for descriptor ring completions use a credit mechanism for
    284 efficient device, PCIe bus, OS and driver operations.  Each descriptor ring has
    285 a credit count which represents the number of outstanding descriptors to be
    286 processed by the driver.  As the device marks descriptors complete, the credit
    287 count is incremented.  As the driver processes those outstanding descriptors,
    288 it returns credits back to the device.  This way, the device knows the driver's
    289 progress and can make decisions about when to fire the next interrupt or not.
    290 When the credit count is zero, and the first descriptors are posted for the
    291 driver, a single interrupt is fired.  Once the interrupt is fired, the
    292 interrupt is disabled (auto-masked*).  In response to the interrupt, the driver
    293 will process descriptors and PIO write a returned credit value for that
    294 descriptor ring.  If the driver returns all credits (the driver caught up with
    295 the device and there is no outstanding work), then the interrupt is unmasked,
    296 but not fired.  If only partial credits are returned, the interrupt remains
    297 masked but the device generates an interrupt, signaling the driver that more
    298 outstanding work is available.
    299 
    300 (* this masking is unrelated to the MSI-X interrupt mask register)
    301 
    302 Endianness
    303 ----------
    304 
    305 Device registers are hard-coded to little-endian (LE).  The driver should
    306 convert to/from host endianness to LE for device register accesses.
    307 
    308 Descriptors are LE.  Descriptor buffer TLVs will have LE type and length
    309 fields, but the value field can either be LE or network-byte-order, depending
    310 on context.  TLV values containing network packet data will be in network-byte
    311 order.  A TLV value containing a field or mask used to compare against network
    312 packet data is network-byte order.  For example, flow match fields (and masks)
    313 are network-byte-order since they're matched directly, byte-by-byte, against
    314 network packet data.  All non-network-packet TLV multi-byte values will be LE.
    315 
    316 TLV values in network-byte-order are designated with (N).
    317 
    318 
    319 SECTION 5: Test Registers
    320 =========================
    321 
    322 Rocker has several test registers to support troubleshooting register access,
    323 interrupt generation, and DMA operations:
    324 
    325 	TEST_REG, offset 0x0010, 32-bit (R/W)
    326 	TEST_REG64, offset 0x0018, 64-bit (R/W)
    327 	TEST_IRQ, offset 0x0020, 32-bit (R/W)
    328 	TEST_DMA_ADDR, offset 0x0028, 64-bit (R/W)
    329 	TEST_DMA_SIZE, offset 0x0030, 32-bit (R/W)
    330 	TEST_DMA_CTRL, offset 0x0034, 32-bit (R/W)
    331 
    332 Reads to TEST_REG and TEST_REG64 will read a value equal to twice the last
    333 value written to the register.  The 32-bit and 64-bit versions are for testing
    334 32-bit and 64-bit host accesses.
    335 
    336 A vector can be written to TEST_IRQ and the device will generate an interrupt
    337 for that vector.
    338 
    339 To test basic DMA operations, allocate a DMA-able host buffer and put the
    340 buffer address into TEST_DMA_ADDR and size into TEST_DMA_SIZE.  Then, write to
    341 TEST_DMA_CTRL to manipulate the buffer contents.  TEST_DMA_CTRL operations are:
    342 
    343 	operation		value	description
    344 	-----------------------------------------------------------
    345 	TEST_DMA_CTRL_CLEAR	1	clear buffer
    346 	TEST_DMA_CTRL_FILL	2	fill buffer bytes with 0x96
    347 	TEST_DMA_CTRL_INVERT	4	invert bytes in buffer
    348 
    349 Various buffer address and sizes should be tested to verify no address boundary
    350 issue exists.  In particular, buffers that start on odd-8-byte boundary and/or
    351 span multiple PAGE sizes should be tested.
    352 
    353 
    354 SECTION 6: Ports
    355 ================
    356 
    357 Physical and Logical Ports
    358 ------------------------------------
    359 
    360 The switch supports up to 62 physical (front-panel) ports.  Register
    361 PORT_PHYS_COUNT returns the actual number of physical ports available:
    362 
    363 	PORT_PHYS_COUNT, offset 0x0304, 32-bit, (R)
    364 
    365 In addition to front-panel ports, the switch supports logical ports for
    366 tunnels.
    367 
    368 Front-panel ports and logical tunnel ports are mapped into a single 32-bit port
    369 space.  A special CPU port is assigned port 0.  The front-panel ports are
    370 mapped to ports 1-62.  A special loopback port is assigned port 63.  Logical
    371 tunnel ports are assigned ports 0x0001000-0x0001ffff.
    372 To summarize the port assignments:
    373 
    374 	port			mapping
    375 	-------------------------------------------------------
    376 	0			CPU port (for packets to/from host CPU)
    377 	1-62			front-panel physical ports
    378 	63			loopback port
    379 	64-0x0000ffff		RSVD
    380 	0x00010000-0x0001ffff	logical tunnel ports
    381 	0x00020000-0xffffffff	RSVD
    382 
    383 Physical Port Mode
    384 ------------------
    385 
    386 Switch front-panel ports operate in a mode.  Currently, the only mode is
    387 OF-DPA.  OF-DPA[1] mode is based on OpenFlow Data Plane Abstraction (OF-DPA)
    388 Abstract Switch Specification, Version 1.0, from Broadcom Corporation.  To
    389 set/get the mode for front-panel ports, see port settings, below.
    390 
    391 Port Settings
    392 -------------
    393 
    394 Link status for all front-panel ports is available via PORT_PHYS_LINK_STATUS:
    395 
    396 	PORT_PHYS_LINK_STATUS, offset 0x0310, 64-bit, (R)
    397 
    398 	Value is port bitmap.  Bits 0 and 63 always read 0.  Bits 1-62
    399 	read 1 for link UP and 0 for link DOWN for respective front-panel ports.
    400 
    401 Other properties for front-panel ports are available via DMA CMD descriptors:
    402 
    403 	Get PORT_SETTINGS descriptor:
    404 
    405 		field		width	description
    406 		----------------------------------------------
    407 		PORT_SETTINGS	2	CMD_GET
    408 		PPORT		4	Physical port #
    409 
    410 	Get PORT_SETTINGS completion:
    411 
    412 		field		width	description
    413 		----------------------------------------------
    414 		PPORT		4	Physical port #
    415 		SPEED		4	Current port interface speed, in Mbps
    416 		DUPLEX		1	1 = Full, 0 = Half
    417 		AUTONEG		1	1 = enabled, 0 = disabled
    418 		MACADDR		6	Port MAC address
    419 		MODE		1	0 = OF-DPA
    420 		LEARNING	1	MAC address learning on port
    421 						1 = enabled
    422 						0 = disabled
    423 		PHYS_NAME	<var>	Physical port name (string)
    424 
    425 	Set PORT_SETTINGS descriptor:
    426 
    427 		field		width	description
    428 		----------------------------------------------
    429 		PORT_SETTINGS	2	CMD_SET
    430 		PPORT		4	Physical port #
    431 		SPEED		4	Port interface speed, in Mbps
    432 		DUPLEX		1	1 = Full, 0 = Half
    433 		AUTONEG		1	1 = enabled, 0 = disabled
    434 		MACADDR		6	Port MAC address
    435 		MODE		1	0 = OF-DPA
    436 
    437 Port Enable
    438 -----------
    439 
    440 Front-panel ports are initially disabled, which means port ingress and egress
    441 packets will be dropped.  To enable or disable a port, use PORT_PHYS_ENABLE:
    442 
    443 	PORT_PHYS_ENABLE: offset 0x0318, 64-bit, (R/W)
    444 
    445 	Value is bitmap of first 64 ports.  Bits 0 and 63 are ignored
    446 	and always read as 0.  Write 1 to enable port; write 0 to disable it.
    447 	Default is 0.
    448 
    449 
    450 SECTION 7: Switch Control
    451 =========================
    452 
    453 This section covers switch-wide register settings.
    454 
    455 Control
    456 -------
    457 
    458 This register is used for low level control of the switch.
    459 
    460 	CONTROL: offset 0x0300, 32-bit, (W)
    461 
    462 	bit	name		description
    463 	------------------------------------------------------------------------
    464 	[0]	CONTROL_RESET	If set, device will perform reset
    465 	[1:31]	Reserved
    466 
    467 Switch ID
    468 ---------
    469 
    470 The switch has a SWITCH_ID to be used by software to uniquely identify the
    471 switch:
    472 
    473 	SWITCH_ID: offset 0x0320, 64-bit, (R)
    474 
    475 	Value is opaque to switch software and no special encoding is implied.
    476 
    477 
    478 SECTION 8: Events
    479 =================
    480 
    481 Non-I/O asynchronous events from the device are notified to the host using the
    482 event ring.  The TLV structure for events is:
    483 
    484 	field		width	description
    485 	---------------------------------------------------
    486 	TYPE		4	Event type, one of:
    487 					1: LINK_CHANGED
    488 					2: MAC_VLAN_SEEN
    489 	INFO		<nest>	Event info (details below)
    490 
    491 Link Changed Event
    492 ------------------
    493 
    494 When link status changes on a physical port, this event is generated.
    495 
    496 	field		width	description
    497 	---------------------------------------------------
    498 	INFO		<nest>
    499 	  PPORT		4	Physical port
    500 	  LINKUP	1	Link status:
    501 					0: down
    502 					1: up
    503 
    504 MAC VLAN Seen Event
    505 -------------------
    506 
    507 When a packet ingresses on a port and the source MAC/VLAN isn't known to the
    508 device, the device will generate this event.  In response to the event, the
    509 driver should install to the device the MAC/VLAN on the port into the bridge
    510 table.  Once installed, the MAC/VLAN is known on the port and this event will
    511 no longer be generated.
    512 
    513 	field		width	description
    514 	---------------------------------------------------
    515 	INFO		<nest>
    516 	  PPORT		4	Physical port
    517 	  MAC		6	MAC address
    518 	  VLAN		2	VLAN ID
    519 
    520 
    521 SECTION 9: CPU Packet Processing
    522 ================================
    523 
    524 Ingress packets directed to the host CPU for further processing are delivered
    525 in the DMA RX ring.  Likewise, host CPU originating packets destined to egress
    526 on switch ports are scheduled by software using the DMA TX ring.
    527 
    528 Tx Packet Processing
    529 --------------------
    530 
    531 Software schedules packets for egress on switch ports using the DMA TX ring.  A
    532 TX descriptor buffer describes the packet location and size in host DMA-able
    533 memory, the destination port, and any hardware-offload functions (such as L3
    534 payload checksum offload).  Software then bumps the descriptor head to signal
    535 hardware of new Tx work.  In response, hardware will DMA read Tx descriptors up
    536 to head, DMA read descriptor buffer and packet data, perform offloading
    537 functions, and finally frame packet on wire (network).  Once packet processing
    538 is complete, hardware will writeback status to descriptor(s) to signal to
    539 software that Tx is complete and software resources (e.g. skb) backing packet
    540 can be released.
    541 
    542 Figure 2 shows an example 3-fragment packet queued with one Tx descriptor.  A
    543 TLV is used for each packet fragment.
    544 
    545 	                                           pkt frag 1
    546 	                                           +–––––––+  +–+
    547 	                                       +–––+       |    |
    548 	                         desc buf      |   |       |    |
    549 	                        +––––––––+     |   |       |    |
    550 	        Tx ring     +–––+        +–––––+   |       |    |
    551 	      +–––––––––+   |   |  TLVs  |         +–––––––+    |
    552 	      |         +–––+   +––––––––+         pkt frag 2   |
    553 	      | desc 0  |       |        +–––––+   +–––––––+    |
    554 	      +–––––––––+       |  TLVs  |     +–––+       |    |
    555 	head+–+         |       +––––––––+         |       |    |
    556 	      | desc 1  |       |        +–––––+   +–––––––+    |pkt
    557 	      +–––––––––+       |  TLVs  |     |                |
    558 	      |         |       +––––––––+     |   pkt frag 3   |
    559 	      |         |                      |   +–––––––+    |
    560 	      +–––––––––+                      +–––+       |    |
    561 	      |         |                          |       |    |
    562 	      |         |                          |       |    |
    563 	      +–––––––––+                          |       |    |
    564 	      |         |                          |       |    |
    565 	      |         |                          |       |    |
    566 	      +–––––––––+                          |       |    |
    567 	      |         |                          +–––––––+  +–+
    568 	      |         |
    569 	      +–––––––––+
    570 
    571 				fig 2.
    572 
    573 The TLVs for Tx descriptor buffer are:
    574 
    575 	field			width	description
    576 	---------------------------------------------------------------------
    577 	PPORT			4	Destination physical port #
    578 	TX_OFFLOAD		1	Hardware offload modes:
    579 					  0: no offload
    580 					  1: insert IP csum (ipv4 only)
    581 					  2: insert TCP/UDP csum
    582 					  3: L3 csum calc and insert
    583                         	             into csum offset (TX_L3_CSUM_OFF)
    584                  	                    16-bit 1's complement csum value.
    585                                 	     IPv4 pseudo-header and IP
    586                         	             already calculated by OS
    587                   	                   and inserted.
    588 					  4: TSO (TCP Segmentation Offload)
    589 	TX_L3_CSUM_OFF		2	For L3 csum offload mode, the offset,
    590 					from the beginning of the packet,
    591 					of the csum field in the L3 header
    592 	TX_TSO_MSS		2	For TSO offload mode, the
    593 					Maximum Segment Size in bytes
    594         TX_TSO_HDR_LEN		2	For TSO offload mode, the
    595 					length of ethernet, IP, and
    596 					TCP/UDP headers, including IP
    597 					and TCP options.
    598 	TX_FRAGS		<array>	Packet fragments
    599 	  TX_FRAG		<nest>	Packet fragment
    600 	    TX_FRAG_ADDR	8	DMA address of packet fragment
    601 	    TX_FRAG_LEN		2	Packet fragment length
    602 
    603 Possible status return codes in descriptor on completion are:
    604 
    605 	DESC_COMP_ERR	reason
    606 	--------------------------------------------------------------------
    607 	0		OK
    608 	-ROCKER_ENXIO	address or data read err on desc buf or packet
    609 			fragment
    610 	-ROCKER_EINVAL	bad pport or TSO or csum offloading error
    611 	-ROCKER_ENOMEM	no memory for internal staging tx fragment
    612 
    613 Rx Packet Processing
    614 --------------------
    615 
    616 For packets ingressing on switch ports that are not forwarded by the switch but
    617 rather directed to the host CPU for further processing are delivered in the DMA
    618 RX ring.  Rx descriptor buffers are allocated by software and placed on the
    619 ring.  Hardware will fill Rx descriptor buffers with packet data, write the
    620 completion, and signal to software that a new packet is ready.  Since Rx packet
    621 size is not known a-priori, the Rx descriptor buffer must be allocated for
    622 worst-case packet size.  A single Rx descriptor will contain the entire Rx
    623 packet data in one RX_FRAG.  Other Rx TLVs describe and hardware offloads
    624 performed on the packet, such as checksum validation.
    625 
    626 The TLVs for Rx descriptor buffer are:
    627 
    628 	field		width	description
    629 	---------------------------------------------------
    630 	PPORT		4	Source physical port #
    631 	RX_FLAGS	2	Packet parsing flags:
    632 				  (1 << 0): IPv4 packet
    633 				  (1 << 1): IPv6 packet
    634 				  (1 << 2): csum calculated
    635 				  (1 << 3): IPv4 csum good
    636 				  (1 << 4): IP fragment
    637 				  (1 << 5): TCP packet
    638 				  (1 << 6): UDP packet
    639 				  (1 << 7): TCP/UDP csum good
    640 				  (1 << 8): Offload forward
    641 	RX_CSUM		2	IP calculated checksum:
    642 				  IPv4: IP payload csum
    643 				  IPv6: header and payload csum
    644 				(Only valid is RX_FLAGS:csum calc is set)
    645 	RX_FRAG_ADDR	8	DMA address of packet fragment
    646 	RX_FRAG_MAX_LEN	2	Packet maximum fragment length
    647 	RX_FRAG_LEN	2	Actual packet fragment length after receive
    648 
    649 Offload forward RX_FLAG indicates the device has already forwarded the packet
    650 so the host CPU should not also forward the packet.
    651 
    652 Possible status return codes in descriptor on completion are:
    653 
    654 	DESC_COMP_ERR	reason
    655 	--------------------------------------------------------------------
    656 	0		OK
    657 	-ROCKER_ENXIO	address or data read err on desc buf
    658 	-ROCKER_ENOMEM	no memory for internal staging desc buf
    659 	-ROCKER_EMSGSIZE Rx descriptor buffer wasn't big enough to contain
    660 			packet data TLV and other TLVs.
    661 
    662 
    663 SECTION 10: OF-DPA Mode
    664 ======================
    665 
    666 OF-DPA mode allows the switch to offload flow packet processing functions to
    667 hardware.  An OpenFlow controller would communicate with an OpenFlow agent
    668 installed on the switch.  The OpenFlow agent would (directly or indirectly)
    669 communicate with the Rocker switch driver, which in turn would program switch
    670 hardware with flow functionality, as defined in OF-DPA.  The block diagram is:
    671 
    672 		+–––––––––––––––----–––+
    673 		|        OF            |
    674 		|  Remote Controller   |
    675 		+––––––––+––----–––––––+
    676 		         |
    677 		         |
    678 		+––––––––+–––––––––+
    679 		|       OF         |
    680 		|   Local Agent    |
    681 		+––––––––––––––––––+
    682 		|                  |
    683 		|   Rocker Driver  |
    684 		+––––––––––––––––––+
    685 		    <this spec>
    686 		+––––––––––––––––––+
    687 		|                  |
    688 		|   Rocker Switch  |
    689 		+––––––––––––––––––+
    690 
    691 To participate in flow functions, ports must be configure for OF-DPA mode
    692 during switch initialization.
    693 
    694 OF-DPA Flow Table Interface
    695 ---------------------------
    696 
    697 There are commands to add, modify, delete, and get stats of flow table entries.
    698 The commands are issued using the DMA CMD descriptor ring.  The following
    699 commands are defined:
    700 
    701 	CMD_ADD:		add an entry to flow table
    702 	CMD_MOD:		modify an entry in flow table
    703 	CMD_DEL:		delete an entry from flow table
    704 	CMD_GET_STATS:		get stats for flow entry
    705 
    706 TLVs for add and modify commands are:
    707 
    708 	field			width	description
    709 	----------------------------------------------------
    710 	OF_DPA_CMD		2	CMD_[ADD|MOD]
    711 	OF_DPA_TBL		2	Flow table ID
    712 					  0: ingress port
    713 					  10: vlan
    714 					  20: termination mac
    715 					  30: unicast routing
    716 					  40: multicast routing
    717 					  50: bridging
    718 					  60: ACL policy
    719 	OF_DPA_PRIORITY		4	Flow priority
    720 	OF_DPA_HARDTIME		4	Hard timeout for flow
    721 	OF_DPA_IDLETIME		4	Idle timeout for flow
    722 	OF_DPA_COOKIE		8	Cookie
    723 
    724 Additional TLVs based on flow table ID:
    725 
    726 Table ID 0: ingress port
    727 
    728 	field			width	description
    729 	----------------------------------------------------
    730 	OF_DPA_IN_PPORT		4	ingress physical port number
    731 	OF_DPA_GOTO_TBL		2	goto table ID; zero to drop
    732 
    733 Table ID 10: vlan
    734 
    735 	field			width	description
    736 	----------------------------------------------------
    737 	OF_DPA_IN_PPORT		4	ingress physical port number
    738 	OF_DPA_VLAN_ID		2 (N)	vlan ID
    739 	OF_DPA_VLAN_ID_MASK	2 (N)	vlan ID mask
    740 	OF_DPA_GOTO_TBL		2	goto table ID; zero to drop
    741 	OF_DPA_NEW_VLAN_ID	2 (N)	new vlan ID
    742 
    743 Table ID 20: termination mac
    744 
    745 	field			width	description
    746 	----------------------------------------------------
    747 	OF_DPA_IN_PPORT		4	ingress physical port number
    748 	OF_DPA_IN_PPORT_MASK	4	ingress physical port number mask
    749 	OF_DPA_ETHERTYPE	2 (N)	must be either 0x0800 or 0x86dd
    750 	OF_DPA_DST_MAC		6 (N)	destination MAC
    751 	OF_DPA_DST_MAC_MASK	6 (N)	destination MAC mask
    752 	OF_DPA_VLAN_ID		2 (N)	vlan ID
    753 	OF_DPA_VLAN_ID_MASK	2 (N)	vlan ID mask
    754 	OF_DPA_GOTO_TBL		2	only acceptable values are
    755 					unicast or multicast routing
    756 					table IDs
    757 	OF_DPA_OUT_PPORT	2	if specified, must be
    758 					controller, set zero otherwise
    759 
    760 Table ID 30: unicast routing
    761 
    762 	field			width	description
    763 	----------------------------------------------------
    764 	OF_DPA_ETHERTYPE	2 (N)	must be either 0x0800 or 0x86dd
    765 	OF_DPA_DST_IP		4 (N)	destination IPv4 address.
    766 					Must be unicast address
    767 	OF_DPA_DST_IP_MASK	4 (N)	IP mask.  Must be prefix mask
    768 	OF_DPA_DST_IPV6		16 (N)	destination IPv6 address.
    769 					Must be unicast address
    770 	OF_DPA_DST_IPV6_MASK	16 (N)	IPv6 mask. Must be prefix mask
    771 	OF_DPA_GOTO_TBL		2	goto table ID; zero to drop
    772 	OF_DPA_GROUP_ID		4	data for GROUP action must
    773 					be an L3 Unicast group entry
    774 
    775 Table ID 40: multicast routing
    776 
    777 	field			width	description
    778 	----------------------------------------------------
    779 	OF_DPA_ETHERTYPE	2 (N)	must be either 0x0800 or 0x86dd
    780 	OF_DPA_VLAN_ID		2 (N)	vlan ID
    781 	OF_DPA_SRC_IP		4 (N)	source IPv4. Optional,
    782 					can contain IPv4 address,
    783 					must be completely masked
    784 					if not used
    785 	OF_DPA_SRC_IP_MASK	4 (N)	IP Mask
    786 	OF_DPA_DST_IP		4 (N)	destination IPv4 address.
    787 					Must be multicast address
    788 	OF_DPA_SRC_IPV6		16 (N)	source IPv6 Address. Optional.
    789 					Can contain IPv6 address,
    790 					must be completely masked
    791 					if not used
    792 	OF_DPA_SRC_IPV6_MASK	16 (N)	IPv6 mask.
    793 	OF_DPA_DST_IPV6		16 (N)	destination IPv6 Address. Must
    794 					be multicast address
    795 					Must be multicast address
    796 	OF_DPA_GOTO_TBL		2	goto table ID; zero to drop
    797 	OF_DPA_GROUP_ID		4	data for GROUP action must
    798 					be an L3 multicast group entry
    799 
    800 Table ID 50: bridging
    801 
    802 	field			width	description
    803 	----------------------------------------------------
    804 	OF_DPA_VLAN_ID		2 (N)	vlan ID
    805 	OF_DPA_TUNNEL_ID	4	tunnel ID
    806 	OF_DPA_DST_MAC		6 (N)	destination MAC
    807 	OF_DPA_DST_MAC_MASK	6 (N)	destination MAC mask
    808 	OF_DPA_GOTO_TBL		2	goto table ID; zero to drop
    809 	OF_DPA_GROUP_ID		4	data for GROUP action must
    810 					be a L2 Interface, L2
    811 					Multicast, L2 Flood,
    812 					or L2 Overlay group entry
    813 					as appropriate
    814 	OF_DPA_TUNNEL_LPORT	4	unicast Tenant Bridging
    815 					flows specify a tunnel
    816 					logical port ID
    817 	OF_DPA_OUT_PPORT	2	data for OUTPUT action,
    818 					restricted to CONTROLLER,
    819 					set to 0 otherwise
    820 
    821 Table ID 60: acl policy
    822 
    823 	field			width	description
    824 	----------------------------------------------------
    825 	OF_DPA_IN_PPORT		4	ingress physical port number
    826 	OF_DPA_IN_PPORT_MASK	4	ingress physical port number mask
    827 	OF_DPA_ETHERTYPE	2 (N)	ethertype
    828 	OF_DPA_VLAN_ID		2 (N)	vlan ID
    829 	OF_DPA_VLAN_ID_MASK	2 (N)	vlan ID mask
    830 	OF_DPA_VLAN_PCP		2 (N)	vlan Priority Code Point
    831 	OF_DPA_VLAN_PCP_MASK	2 (N)	vlan Priority Code Point mask
    832 	OF_DPA_SRC_MAC		6 (N)	source MAC
    833 	OF_DPA_SRC_MAC_MASK	6 (N)	source MAC mask
    834 	OF_DPA_DST_MAC		6 (N)	destination MAC
    835 	OF_DPA_DST_MAC_MASK	6 (N)	destination MAC mask
    836 	OF_DPA_TUNNEL_ID	4	tunnel ID
    837 	OF_DPA_SRC_IP		4 (N)	source IPv4. Optional,
    838 					can contain IPv4 address,
    839 					must be completely masked
    840 					if not used
    841 	OF_DPA_SRC_IP_MASK	4 (N)	IP Mask
    842 	OF_DPA_DST_IP		4 (N)	destination IPv4 address.
    843 					Must be multicast address
    844 	OF_DPA_DST_IP_MASK	4 (N)	IP Mask
    845 	OF_DPA_SRC_IPV6		16 (N)	source IPv6 Address. Optional.
    846 					Can contain IPv6 address,
    847 					must be completely masked
    848 					if not used
    849 	OF_DPA_SRC_IPV6_MASK	16 (N)	IPv6 mask
    850 	OF_DPA_DST_IPV6		16 (N)	destination IPv6 Address. Must
    851 					be multicast address.
    852 	OF_DPA_DST_IPV6_MASK	16 (N)	IPv6 mask
    853 	OF_DPA_SRC_ARP_IP	4 (N)	source IPv4 address in the ARP
    854 					payload.  Only used if ethertype
    855 					== 0x0806.
    856 	OF_DPA_SRC_ARP_IP_MASK	4 (N)	IP Mask
    857 	OF_DPA_IP_PROTO		1	IP protocol
    858 	OF_DPA_IP_PROTO_MASK	1	IP protocol mask
    859 	OF_DPA_IP_DSCP		1	DSCP
    860 	OF_DPA_IP_DSCP_MASK	1	DSCP mask
    861 	OF_DPA_IP_ECN		1	ECN
    862 	OF_DPA_IP_ECN_MASK		1	ECN mask
    863 	OF_DPA_L4_SRC_PORT	2 (N)	L4 source port, only for
    864 					TCP, UDP, or SCTP
    865 	OF_DPA_L4_SRC_PORT_MASK	2 (N)	L4 source port mask
    866 	OF_DPA_L4_DST_PORT	2 (N)	L4 source port, only for
    867 					TCP, UDP, or SCTP
    868 	OF_DPA_L4_DST_PORT_MASK	2 (N)	L4 source port mask
    869 	OF_DPA_ICMP_TYPE	1	ICMP type, only if IP
    870 					protocol is 1
    871 	OF_DPA_ICMP_TYPE_MASK	1	ICMP type mask
    872 	OF_DPA_ICMP_CODE	1	ICMP code
    873 	OF_DPA_ICMP_CODE_MASK	1	ICMP code mask
    874 	OF_DPA_IPV6_LABEL	4 (N)	IPv6 flow label
    875 	OF_DPA_IPV6_LABEL_MASK	4 (N)	IPv6 flow label mask
    876 	OF_DPA_GROUP_ID		4	data for GROUP action
    877 	OF_DPA_QUEUE_ID_ACTION	1	write the queue ID
    878 	OF_DPA_NEW_QUEUE_ID	1	queue ID
    879 	OF_DPA_VLAN_PCP_ACTION	1	write the VLAN priority
    880 	OF_DPA_NEW_VLAN_PCP	1	VLAN priority
    881 	OF_DPA_IP_DSCP_ACTION	1	write the DSCP
    882 	OF_DPA_NEW_IP_DSCP	1	new DSCP
    883 	OF_DPA_TUNNEL_LPORT	4	restrct to valid tunnel
    884 					logical port, set to 0
    885 					otherwise.
    886 	OF_DPA_OUT_PPORT	2	data for OUTPUT action,
    887 					restricted to CONTROLLER,
    888 					set to 0 otherwise
    889 	OF_DPA_CLEAR_ACTIONS	4	if 1 packets matching flow are
    890 					dropped (all other instructions
    891 					ignored)
    892 
    893 TLVs for flow delete and get stats command are:
    894 
    895 	field			width	description
    896 	---------------------------------------------------
    897 	OF_DPA_CMD		2	CMD_[DEL|GET_STATS]
    898 	OF_DPA_COOKIE		8	Cookie
    899 
    900 On completion of get stats command, the descriptor buffer is written back with
    901 the following TLVs:
    902 
    903 	field			width	description
    904 	---------------------------------------------------
    905 	OF_DPA_STAT_DURATION	4	Flow duration
    906 	OF_DPA_STAT_RX_PKTS	8	Received packets
    907 	OF_DPA_STAT_TX_PKTS	8	Transmit packets
    908 
    909 Possible status return codes in descriptor on completion are:
    910 
    911 	DESC_COMP_ERR	command			reason
    912 	--------------------------------------------------------------------
    913 	0		all			OK
    914 	-ROCKER_EFAULT	all			head or tail index outside
    915 						of ring
    916 	-ROCKER_ENXIO	all			address or data read err on
    917 						desc buf
    918 	-ROCKER_EMSGSIZE GET_STATS		cmd descriptor buffer wasn't
    919 						big enough to contain write-back
    920 						TLVs
    921 	-ROCKER_EINVAL	all			invalid parameters passed in
    922 	-ROCKER_EEXIST	ADD			entry already exists
    923 	-ROCKER_ENOSPC	ADD			no space left in flow table
    924 	-ROCKER_ENOENT	MOD|DEL|GET_STATS	cookie invalid
    925 
    926 Group Table Interface
    927 ---------------------
    928 
    929 There are commands to add, modify, delete, and get stats of group table
    930 entries.  The commands are issued using the DMA CMD descriptor ring.  The
    931 following commands are defined:
    932 
    933 	CMD_ADD:		add an entry to group table
    934 	CMD_MOD:		modify an entry in group table
    935 	CMD_DEL:		delete an entry from group table
    936 	CMD_GET_STATS:		get stats for group entry
    937 
    938 TLVs for add and modify commands are:
    939 
    940 	field			width	description
    941 	-----------------------------------------------------------
    942 	FLOW_GROUP_CMD		2	CMD_[ADD|MOD]
    943 	FLOW_GROUP_ID		2	Flow group ID
    944 	FLOW_GROUP_TYPE		1	Group type:
    945 					  0: L2 interface
    946 					  1: L2 rewrite
    947 					  2: L3 unicast
    948 					  3: L2 multicast
    949 					  4: L2 flood
    950 					  5: L3 interface
    951 					  6: L3 multicast
    952 					  7: L3 ECMP
    953 					  8: L2 overlay
    954 	FLOW_VLAN_ID		2	Vlan ID (types 0, 3, 4, 6)
    955 	FLOW_L2_PORT		2	Port (types 0)
    956 	FLOW_INDEX		4	Index (all types but 0)
    957 	FLOW_OVERLAY_TYPE	1	Overlay sub-type (type 8):
    958 					  0: Flood unicast tunnel
    959 					  1: Flood multicast tunnel
    960 					  2: Multicast unicast tunnel
    961 					  3: Multicast multicast tunnel
    962 	FLOW_GROUP_ACTION		nest
    963 	  FLOW_GROUP_ID		2	next group ID in chain (all
    964 					types except 0)
    965 	  FLOW_OUT_PORT		4	egress port (types 0, 8)
    966 	  FLOW_POP_VLAN_TAG	1	strip outer VLAN tag (type 1
    967 					only)
    968 	  FLOW_VLAN_ID		2	(types 1, 5)
    969 	  FLOW_SRC_MAC		6	(types 1, 2, 5)
    970 	  FLOW_DST_MAC		6	(types 1, 2)
    971 
    972 TLVs for flow delete and get stats command are:
    973 
    974 	field			width	description
    975 	-----------------------------------------------------------
    976 	FLOW_GROUP_CMD		2	CMD_[DEL|GET_STATS]
    977 	FLOW_GROUP_ID		2	Flow group ID
    978 
    979 On completion of get stats command, the descriptor buffer is written back with
    980 the following TLVs:
    981 
    982 	field			width	description
    983 	---------------------------------------------------
    984 	FLOW_GROUP_ID		2	Flow group ID
    985 	FLOW_STAT_DURATION	4	Flow duration
    986 	FLOW_STAT_REF_COUNT	4	Flow reference count
    987 	FLOW_STAT_BUCKET_COUNT	4	Flow bucket count
    988 
    989 Possible status return codes in descriptor on completion are:
    990 
    991 	DESC_COMP_ERR	command			reason
    992 	--------------------------------------------------------------------
    993 	0		all			OK
    994 	-ROCKER_EFAULT	all			head or tail index outside
    995 						of ring
    996 	-ROCKER_ENXIO	all			address or data read err on
    997 						desc buf
    998 	-ROCKER_ENOSPC	GET_STATS		cmd descriptor buffer wasn't
    999 						big enough to contain write-back
   1000 						TLVs
   1001 	-ROCKER_EINVAL	ADD|MOD			invalid parameters passed in
   1002 	-ROCKER_EEXIST	ADD			entry already exists
   1003 	-ROCKER_ENOSPC	ADD			no space left in flow table
   1004 	-ROCKER_ENOENT	MOD|DEL|GET_STATS	group ID invalid
   1005 	-ROCKER_EBUSY	DEL			group reference count non-zero
   1006 	-ROCKER_ENODEV	ADD			next group ID doesn't exist
   1007 
   1008 
   1009 
   1010 References
   1011 ==========
   1012 
   1013 [1] OpenFlow Data Plane Abstraction (OF-DPA) Abstract Switch Specification,
   1014 Version 1.0, from Broadcom Corporation, February 21, 2014.