| 1 |
.. SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) |
| 2 |
|
| 3 |
==================================== |
| 4 |
Marvell OcteonTx2 RVU Kernel Drivers |
| 5 |
==================================== |
| 6 |
|
| 7 |
Copyright (c) 2020 Marvell International Ltd. |
| 8 |
|
| 9 |
Contents |
| 10 |
======== |
| 11 |
|
| 12 |
- `Overview`_ |
| 13 |
- `Drivers`_ |
| 14 |
- `Basic packet flow`_ |
| 15 |
- `Devlink health reporters`_ |
| 16 |
- `Quality of service`_ |
| 17 |
- `RVU representors`_ |
| 18 |
|
| 19 |
Overview |
| 20 |
======== |
| 21 |
|
| 22 |
Resource virtualization unit (RVU) on Marvell's OcteonTX2 SOC maps HW |
| 23 |
resources from the network, crypto and other functional blocks into |
| 24 |
PCI-compatible physical and virtual functions. Each functional block |
| 25 |
again has multiple local functions (LFs) for provisioning to PCI devices. |
| 26 |
RVU supports multiple PCIe SRIOV physical functions (PFs) and virtual |
| 27 |
functions (VFs). PF0 is called the administrative / admin function (AF) |
| 28 |
and has privileges to provision RVU functional block's LFs to each of the |
| 29 |
PF/VF. |
| 30 |
|
| 31 |
RVU managed networking functional blocks |
| 32 |
- Network pool or buffer allocator (NPA) |
| 33 |
- Network interface controller (NIX) |
| 34 |
- Network parser CAM (NPC) |
| 35 |
- Schedule/Synchronize/Order unit (SSO) |
| 36 |
- Loopback interface (LBK) |
| 37 |
|
| 38 |
RVU managed non-networking functional blocks |
| 39 |
- Crypto accelerator (CPT) |
| 40 |
- Scheduled timers unit (TIM) |
| 41 |
- Schedule/Synchronize/Order unit (SSO) |
| 42 |
Used for both networking and non networking usecases |
| 43 |
|
| 44 |
Resource provisioning examples |
| 45 |
- A PF/VF with NIX-LF & NPA-LF resources works as a pure network device |
| 46 |
- A PF/VF with CPT-LF resource works as a pure crypto offload device. |
| 47 |
|
| 48 |
RVU functional blocks are highly configurable as per software requirements. |
| 49 |
|
| 50 |
Firmware setups following stuff before kernel boots |
| 51 |
- Enables required number of RVU PFs based on number of physical links. |
| 52 |
- Number of VFs per PF are either static or configurable at compile time. |
| 53 |
Based on config, firmware assigns VFs to each of the PFs. |
| 54 |
- Also assigns MSIX vectors to each of PF and VFs. |
| 55 |
- These are not changed after kernel boot. |
| 56 |
|
| 57 |
Drivers |
| 58 |
======= |
| 59 |
|
| 60 |
Linux kernel will have multiple drivers registering to different PF and VFs |
| 61 |
of RVU. Wrt networking there will be 3 flavours of drivers. |
| 62 |
|
| 63 |
Admin Function driver |
| 64 |
--------------------- |
| 65 |
|
| 66 |
As mentioned above RVU PF0 is called the admin function (AF), this driver |
| 67 |
supports resource provisioning and configuration of functional blocks. |
| 68 |
Doesn't handle any I/O. It sets up few basic stuff but most of the |
| 69 |
functionality is achieved via configuration requests from PFs and VFs. |
| 70 |
|
| 71 |
PF/VFs communicates with AF via a shared memory region (mailbox). Upon |
| 72 |
receiving requests AF does resource provisioning and other HW configuration. |
| 73 |
AF is always attached to host kernel, but PFs and their VFs may be used by host |
| 74 |
kernel itself, or attached to VMs or to userspace applications like |
| 75 |
DPDK etc. So AF has to handle provisioning/configuration requests sent |
| 76 |
by any device from any domain. |
| 77 |
|
| 78 |
AF driver also interacts with underlying firmware to |
| 79 |
- Manage physical ethernet links ie CGX LMACs. |
| 80 |
- Retrieve information like speed, duplex, autoneg etc |
| 81 |
- Retrieve PHY EEPROM and stats. |
| 82 |
- Configure FEC, PAM modes |
| 83 |
- etc |
| 84 |
|
| 85 |
From pure networking side AF driver supports following functionality. |
| 86 |
- Map a physical link to a RVU PF to which a netdev is registered. |
| 87 |
- Attach NIX and NPA block LFs to RVU PF/VF which provide buffer pools, RQs, SQs |
| 88 |
for regular networking functionality. |
| 89 |
- Flow control (pause frames) enable/disable/config. |
| 90 |
- HW PTP timestamping related config. |
| 91 |
- NPC parser profile config, basically how to parse pkt and what info to extract. |
| 92 |
- NPC extract profile config, what to extract from the pkt to match data in MCAM entries. |
| 93 |
- Manage NPC MCAM entries, upon request can frame and install requested packet forwarding rules. |
| 94 |
- Defines receive side scaling (RSS) algorithms. |
| 95 |
- Defines segmentation offload algorithms (eg TSO) |
| 96 |
- VLAN stripping, capture and insertion config. |
| 97 |
- SSO and TIM blocks config which provide packet scheduling support. |
| 98 |
- Debugfs support, to check current resource provising, current status of |
| 99 |
NPA pools, NIX RQ, SQ and CQs, various stats etc which helps in debugging issues. |
| 100 |
- And many more. |
| 101 |
|
| 102 |
Physical Function driver |
| 103 |
------------------------ |
| 104 |
|
| 105 |
This RVU PF handles IO, is mapped to a physical ethernet link and this |
| 106 |
driver registers a netdev. This supports SR-IOV. As said above this driver |
| 107 |
communicates with AF with a mailbox. To retrieve information from physical |
| 108 |
links this driver talks to AF and AF gets that info from firmware and responds |
| 109 |
back ie cannot talk to firmware directly. |
| 110 |
|
| 111 |
Supports ethtool for configuring links, RSS, queue count, queue size, |
| 112 |
flow control, ntuple filters, dump PHY EEPROM, config FEC etc. |
| 113 |
|
| 114 |
Virtual Function driver |
| 115 |
----------------------- |
| 116 |
|
| 117 |
There are two types VFs, VFs that share the physical link with their parent |
| 118 |
SR-IOV PF and the VFs which work in pairs using internal HW loopback channels (LBK). |
| 119 |
|
| 120 |
Type1: |
| 121 |
- These VFs and their parent PF share a physical link and used for outside communication. |
| 122 |
- VFs cannot communicate with AF directly, they send mbox message to PF and PF |
| 123 |
forwards that to AF. AF after processing, responds back to PF and PF forwards |
| 124 |
the reply to VF. |
| 125 |
- From functionality point of view there is no difference between PF and VF as same type |
| 126 |
HW resources are attached to both. But user would be able to configure few stuff only |
| 127 |
from PF as PF is treated as owner/admin of the link. |
| 128 |
|
| 129 |
Type2: |
| 130 |
- RVU PF0 ie admin function creates these VFs and maps them to loopback block's channels. |
| 131 |
- A set of two VFs (VF0 & VF1, VF2 & VF3 .. so on) works as a pair ie pkts sent out of |
| 132 |
VF0 will be received by VF1 and vice versa. |
| 133 |
- These VFs can be used by applications or virtual machines to communicate between them |
| 134 |
without sending traffic outside. There is no switch present in HW, hence the support |
| 135 |
for loopback VFs. |
| 136 |
- These communicate directly with AF (PF0) via mbox. |
| 137 |
|
| 138 |
Except for the IO channels or links used for packet reception and transmission there is |
| 139 |
no other difference between these VF types. AF driver takes care of IO channel mapping, |
| 140 |
hence same VF driver works for both types of devices. |
| 141 |
|
| 142 |
Basic packet flow |
| 143 |
================= |
| 144 |
|
| 145 |
Ingress |
| 146 |
------- |
| 147 |
|
| 148 |
1. CGX LMAC receives packet. |
| 149 |
2. Forwards the packet to the NIX block. |
| 150 |
3. Then submitted to NPC block for parsing and then MCAM lookup to get the destination RVU device. |
| 151 |
4. NIX LF attached to the destination RVU device allocates a buffer from RQ mapped buffer pool of NPA block LF. |
| 152 |
5. RQ may be selected by RSS or by configuring MCAM rule with a RQ number. |
| 153 |
6. Packet is DMA'ed and driver is notified. |
| 154 |
|
| 155 |
Egress |
| 156 |
------ |
| 157 |
|
| 158 |
1. Driver prepares a send descriptor and submits to SQ for transmission. |
| 159 |
2. The SQ is already configured (by AF) to transmit on a specific link/channel. |
| 160 |
3. The SQ descriptor ring is maintained in buffers allocated from SQ mapped pool of NPA block LF. |
| 161 |
4. NIX block transmits the pkt on the designated channel. |
| 162 |
5. NPC MCAM entries can be installed to divert pkt onto a different channel. |
| 163 |
|
| 164 |
Devlink health reporters |
| 165 |
======================== |
| 166 |
|
| 167 |
NPA Reporters |
| 168 |
------------- |
| 169 |
The NPA reporters are responsible for reporting and recovering the following group of errors: |
| 170 |
|
| 171 |
1. GENERAL events |
| 172 |
|
| 173 |
- Error due to operation of unmapped PF. |
| 174 |
- Error due to disabled alloc/free for other HW blocks (NIX, SSO, TIM, DPI and AURA). |
| 175 |
|
| 176 |
2. ERROR events |
| 177 |
|
| 178 |
- Fault due to NPA_AQ_INST_S read or NPA_AQ_RES_S write. |
| 179 |
- AQ Doorbell Error. |
| 180 |
|
| 181 |
3. RAS events |
| 182 |
|
| 183 |
- RAS Error Reporting for NPA_AQ_INST_S/NPA_AQ_RES_S. |
| 184 |
|
| 185 |
4. RVU events |
| 186 |
|
| 187 |
- Error due to unmapped slot. |
| 188 |
|
| 189 |
Sample Output:: |
| 190 |
|
| 191 |
~# devlink health |
| 192 |
pci/0002:01:00.0: |
| 193 |
reporter hw_npa_intr |
| 194 |
state healthy error 2872 recover 2872 last_dump_date 2020-12-10 last_dump_time 09:39:09 grace_period 0 auto_recover true auto_dump true |
| 195 |
reporter hw_npa_gen |
| 196 |
state healthy error 2872 recover 2872 last_dump_date 2020-12-11 last_dump_time 04:43:04 grace_period 0 auto_recover true auto_dump true |
| 197 |
reporter hw_npa_err |
| 198 |
state healthy error 2871 recover 2871 last_dump_date 2020-12-10 last_dump_time 09:39:17 grace_period 0 auto_recover true auto_dump true |
| 199 |
reporter hw_npa_ras |
| 200 |
state healthy error 0 recover 0 last_dump_date 2020-12-10 last_dump_time 09:32:40 grace_period 0 auto_recover true auto_dump true |
| 201 |
|
| 202 |
Each reporter dumps the |
| 203 |
|
| 204 |
- Error Type |
| 205 |
- Error Register value |
| 206 |
- Reason in words |
| 207 |
|
| 208 |
For example:: |
| 209 |
|
| 210 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_npa_gen |
| 211 |
NPA_AF_GENERAL: |
| 212 |
NPA General Interrupt Reg : 1 |
| 213 |
NIX0: free disabled RX |
| 214 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_npa_intr |
| 215 |
NPA_AF_RVU: |
| 216 |
NPA RVU Interrupt Reg : 1 |
| 217 |
Unmap Slot Error |
| 218 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_npa_err |
| 219 |
NPA_AF_ERR: |
| 220 |
NPA Error Interrupt Reg : 4096 |
| 221 |
AQ Doorbell Error |
| 222 |
|
| 223 |
|
| 224 |
NIX Reporters |
| 225 |
------------- |
| 226 |
The NIX reporters are responsible for reporting and recovering the following group of errors: |
| 227 |
|
| 228 |
1. GENERAL events |
| 229 |
|
| 230 |
- Receive mirror/multicast packet drop due to insufficient buffer. |
| 231 |
- SMQ Flush operation. |
| 232 |
|
| 233 |
2. ERROR events |
| 234 |
|
| 235 |
- Memory Fault due to WQE read/write from multicast/mirror buffer. |
| 236 |
- Receive multicast/mirror replication list error. |
| 237 |
- Receive packet on an unmapped PF. |
| 238 |
- Fault due to NIX_AQ_INST_S read or NIX_AQ_RES_S write. |
| 239 |
- AQ Doorbell Error. |
| 240 |
|
| 241 |
3. RAS events |
| 242 |
|
| 243 |
- RAS Error Reporting for NIX Receive Multicast/Mirror Entry Structure. |
| 244 |
- RAS Error Reporting for WQE/Packet Data read from Multicast/Mirror Buffer.. |
| 245 |
- RAS Error Reporting for NIX_AQ_INST_S/NIX_AQ_RES_S. |
| 246 |
|
| 247 |
4. RVU events |
| 248 |
|
| 249 |
- Error due to unmapped slot. |
| 250 |
|
| 251 |
Sample Output:: |
| 252 |
|
| 253 |
~# ./devlink health |
| 254 |
pci/0002:01:00.0: |
| 255 |
reporter hw_npa_intr |
| 256 |
state healthy error 0 recover 0 grace_period 0 auto_recover true auto_dump true |
| 257 |
reporter hw_npa_gen |
| 258 |
state healthy error 0 recover 0 grace_period 0 auto_recover true auto_dump true |
| 259 |
reporter hw_npa_err |
| 260 |
state healthy error 0 recover 0 grace_period 0 auto_recover true auto_dump true |
| 261 |
reporter hw_npa_ras |
| 262 |
state healthy error 0 recover 0 grace_period 0 auto_recover true auto_dump true |
| 263 |
reporter hw_nix_intr |
| 264 |
state healthy error 1121 recover 1121 last_dump_date 2021-01-19 last_dump_time 05:42:26 grace_period 0 auto_recover true auto_dump true |
| 265 |
reporter hw_nix_gen |
| 266 |
state healthy error 949 recover 949 last_dump_date 2021-01-19 last_dump_time 05:42:43 grace_period 0 auto_recover true auto_dump true |
| 267 |
reporter hw_nix_err |
| 268 |
state healthy error 1147 recover 1147 last_dump_date 2021-01-19 last_dump_time 05:42:59 grace_period 0 auto_recover true auto_dump true |
| 269 |
reporter hw_nix_ras |
| 270 |
state healthy error 409 recover 409 last_dump_date 2021-01-19 last_dump_time 05:43:16 grace_period 0 auto_recover true auto_dump true |
| 271 |
|
| 272 |
Each reporter dumps the |
| 273 |
|
| 274 |
- Error Type |
| 275 |
- Error Register value |
| 276 |
- Reason in words |
| 277 |
|
| 278 |
For example:: |
| 279 |
|
| 280 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_nix_intr |
| 281 |
NIX_AF_RVU: |
| 282 |
NIX RVU Interrupt Reg : 1 |
| 283 |
Unmap Slot Error |
| 284 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_nix_gen |
| 285 |
NIX_AF_GENERAL: |
| 286 |
NIX General Interrupt Reg : 1 |
| 287 |
Rx multicast pkt drop |
| 288 |
~# devlink health dump show pci/0002:01:00.0 reporter hw_nix_err |
| 289 |
NIX_AF_ERR: |
| 290 |
NIX Error Interrupt Reg : 64 |
| 291 |
Rx on unmapped PF_FUNC |
| 292 |
|
| 293 |
|
| 294 |
Quality of service |
| 295 |
================== |
| 296 |
|
| 297 |
|
| 298 |
Hardware algorithms used in scheduling |
| 299 |
-------------------------------------- |
| 300 |
|
| 301 |
octeontx2 silicon and CN10K transmit interface consists of five transmit levels |
| 302 |
starting from SMQ/MDQ, TL4 to TL1. Each packet will traverse MDQ, TL4 to TL1 |
| 303 |
levels. Each level contains an array of queues to support scheduling and shaping. |
| 304 |
The hardware uses the below algorithms depending on the priority of scheduler queues. |
| 305 |
once the usercreates tc classes with different priorities, the driver configures |
| 306 |
schedulers allocated to the class with specified priority along with rate-limiting |
| 307 |
configuration. |
| 308 |
|
| 309 |
1. Strict Priority |
| 310 |
|
| 311 |
- Once packets are submitted to MDQ, hardware picks all active MDQs having different priority |
| 312 |
using strict priority. |
| 313 |
|
| 314 |
2. Round Robin |
| 315 |
|
| 316 |
- Active MDQs having the same priority level are chosen using round robin. |
| 317 |
|
| 318 |
|
| 319 |
Setup HTB offload |
| 320 |
----------------- |
| 321 |
|
| 322 |
1. Enable HW TC offload on the interface:: |
| 323 |
|
| 324 |
# ethtool -K <interface> hw-tc-offload on |
| 325 |
|
| 326 |
2. Crate htb root:: |
| 327 |
|
| 328 |
# tc qdisc add dev <interface> clsact |
| 329 |
# tc qdisc replace dev <interface> root handle 1: htb offload |
| 330 |
|
| 331 |
3. Create tc classes with different priorities:: |
| 332 |
|
| 333 |
# tc class add dev <interface> parent 1: classid 1:1 htb rate 10Gbit prio 1 |
| 334 |
|
| 335 |
# tc class add dev <interface> parent 1: classid 1:2 htb rate 10Gbit prio 7 |
| 336 |
|
| 337 |
4. Create tc classes with same priorities and different quantum:: |
| 338 |
|
| 339 |
# tc class add dev <interface> parent 1: classid 1:1 htb rate 10Gbit prio 2 quantum 409600 |
| 340 |
|
| 341 |
# tc class add dev <interface> parent 1: classid 1:2 htb rate 10Gbit prio 2 quantum 188416 |
| 342 |
|
| 343 |
# tc class add dev <interface> parent 1: classid 1:3 htb rate 10Gbit prio 2 quantum 32768 |
| 344 |
|
| 345 |
|
| 346 |
RVU Representors |
| 347 |
================ |
| 348 |
|
| 349 |
RVU representor driver adds support for creation of representor devices for |
| 350 |
RVU PFs' VFs in the system. Representor devices are created when user enables |
| 351 |
the switchdev mode. |
| 352 |
Switchdev mode can be enabled either before or after setting up SRIOV numVFs. |
| 353 |
All representor devices share a single NIXLF but each has a dedicated Rx/Tx |
| 354 |
queues. RVU PF representor driver registers a separate netdev for each |
| 355 |
Rx/Tx queue pair. |
| 356 |
|
| 357 |
Current HW does not support built-in switch which can do L2 learning and |
| 358 |
forwarding packets between representee and representor. Hence, packet path |
| 359 |
between representee and it's representor is achieved by setting up appropriate |
| 360 |
NPC MCAM filters. |
| 361 |
Transmit packets matching these filters will be loopbacked through hardware |
| 362 |
loopback channel/interface (i.e, instead of sending them out of MAC interface). |
| 363 |
Which will again match the installed filters and will be forwarded. |
| 364 |
This way representee => representor and representor => representee packet |
| 365 |
path is achieved. These rules get installed when representors are created |
| 366 |
and gets active/deactivate based on the representor/representee interface state. |
| 367 |
|
| 368 |
Usage example: |
| 369 |
|
| 370 |
- Change device to switchdev mode:: |
| 371 |
|
| 372 |
# devlink dev eswitch set pci/0002:1c:00.0 mode switchdev |
| 373 |
|
| 374 |
- List of representor devices on the system:: |
| 375 |
|
| 376 |
# ip link show |
| 377 |
Rpf1vf0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000 link/ether f6:43:83:ee:26:21 brd ff:ff:ff:ff:ff:ff |
| 378 |
Rpf1vf1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000 link/ether 12:b2:54:0e:24:54 brd ff:ff:ff:ff:ff:ff |
| 379 |
Rpf1vf2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000 link/ether 4a:12:c4:4c:32:62 brd ff:ff:ff:ff:ff:ff |
| 380 |
Rpf1vf3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000 link/ether ca:cb:68:0e:e2:6e brd ff:ff:ff:ff:ff:ff |
| 381 |
Rpf2vf0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000 link/ether 06:cc:ad:b4:f0:93 brd ff:ff:ff:ff:ff:ff |
| 382 |
|
| 383 |
|
| 384 |
To delete the representors devices from the system. Change the device to legacy mode. |
| 385 |
|
| 386 |
- Change device to legacy mode:: |
| 387 |
|
| 388 |
# devlink dev eswitch set pci/0002:1c:00.0 mode legacy |
| 389 |
|
| 390 |
RVU representors can be managed using devlink ports |
| 391 |
(see :ref:`Documentation/networking/devlink/devlink-port.rst <devlink_port>`) interface. |
| 392 |
|
| 393 |
- Show devlink ports of representors:: |
| 394 |
|
| 395 |
# devlink port |
| 396 |
pci/0002:1c:00.0/0: type eth netdev Rpf1vf0 flavour physical port 0 splittable false |
| 397 |
pci/0002:1c:00.0/1: type eth netdev Rpf1vf1 flavour pcivf controller 0 pfnum 1 vfnum 1 external false splittable false |
| 398 |
pci/0002:1c:00.0/2: type eth netdev Rpf1vf2 flavour pcivf controller 0 pfnum 1 vfnum 2 external false splittable false |
| 399 |
pci/0002:1c:00.0/3: type eth netdev Rpf1vf3 flavour pcivf controller 0 pfnum 1 vfnum 3 external false splittable false |
| 400 |
|
| 401 |
Function attributes |
| 402 |
=================== |
| 403 |
|
| 404 |
The RVU representor support function attributes for representors. |
| 405 |
Port function configuration of the representors are supported through devlink eswitch port. |
| 406 |
|
| 407 |
MAC address setup |
| 408 |
----------------- |
| 409 |
|
| 410 |
RVU representor driver support devlink port function attr mechanism to setup MAC |
| 411 |
address. (refer to Documentation/networking/devlink/devlink-port.rst) |
| 412 |
|
| 413 |
- To setup MAC address for port 2:: |
| 414 |
|
| 415 |
# devlink port function set pci/0002:1c:00.0/2 hw_addr 5c:a1:1b:5e:43:11 |
| 416 |
# devlink port show pci/0002:1c:00.0/2 |
| 417 |
pci/0002:1c:00.0/2: type eth netdev Rpf1vf2 flavour pcivf controller 0 pfnum 1 vfnum 2 external false splittable false |
| 418 |
function: |
| 419 |
hw_addr 5c:a1:1b:5e:43:11 |
| 420 |
|
| 421 |
|
| 422 |
TC offload |
| 423 |
========== |
| 424 |
|
| 425 |
The rvu representor driver implements support for offloading tc rules using port representors. |
| 426 |
|
| 427 |
- Drop packets with vlan id 3:: |
| 428 |
|
| 429 |
# tc filter add dev Rpf1vf0 protocol 802.1Q parent ffff: flower vlan_id 3 vlan_ethtype ipv4 skip_sw action drop |
| 430 |
|
| 431 |
- Redirect packets with vlan id 5 and IPv4 packets to eth1, after stripping vlan header.:: |
| 432 |
|
| 433 |
# tc filter add dev Rpf1vf0 ingress protocol 802.1Q flower vlan_id 5 vlan_ethtype ipv4 skip_sw action vlan pop action mirred ingress redirect dev eth1 |