xref: /linux/Documentation/networking/device_drivers/ethernet/amazon/ena.rst (revision c532de5a67a70f8533d495f8f2aaa9a0491c3ad0)
1.. SPDX-License-Identifier: GPL-2.0
2
3============================================================
4Linux kernel driver for Elastic Network Adapter (ENA) family
5============================================================
6
7Overview
8========
9
10ENA is a networking interface designed to make good use of modern CPU
11features and system architectures.
12
13The ENA device exposes a lightweight management interface with a
14minimal set of memory mapped registers and extendible command set
15through an Admin Queue.
16
17The driver supports a range of ENA devices, is link-speed independent
18(i.e., the same driver is used for 10GbE, 25GbE, 40GbE, etc), and has
19a negotiated and extendible feature set.
20
21Some ENA devices support SR-IOV. This driver is used for both the
22SR-IOV Physical Function (PF) and Virtual Function (VF) devices.
23
24ENA devices enable high speed and low overhead network traffic
25processing by providing multiple Tx/Rx queue pairs (the maximum number
26is advertised by the device via the Admin Queue), a dedicated MSI-X
27interrupt vector per Tx/Rx queue pair, adaptive interrupt moderation,
28and CPU cacheline optimized data placement.
29
30The ENA driver supports industry standard TCP/IP offload features such as
31checksum offload. Receive-side scaling (RSS) is supported for multi-core
32scaling.
33
34The ENA driver and its corresponding devices implement health
35monitoring mechanisms such as watchdog, enabling the device and driver
36to recover in a manner transparent to the application, as well as
37debug logs.
38
39Some of the ENA devices support a working mode called Low-latency
40Queue (LLQ), which saves several more microseconds.
41
42ENA Source Code Directory Structure
43===================================
44
45=================   ======================================================
46ena_com.[ch]        Management communication layer. This layer is
47                    responsible for the handling all the management
48                    (admin) communication between the device and the
49                    driver.
50ena_eth_com.[ch]    Tx/Rx data path.
51ena_admin_defs.h    Definition of ENA management interface.
52ena_eth_io_defs.h   Definition of ENA data path interface.
53ena_common_defs.h   Common definitions for ena_com layer.
54ena_regs_defs.h     Definition of ENA PCI memory-mapped (MMIO) registers.
55ena_netdev.[ch]     Main Linux kernel driver.
56ena_ethtool.c       ethtool callbacks.
57ena_xdp.[ch]        XDP files
58ena_pci_id_tbl.h    Supported device IDs.
59=================   ======================================================
60
61Management Interface:
62=====================
63
64ENA management interface is exposed by means of:
65
66- PCIe Configuration Space
67- Device Registers
68- Admin Queue (AQ) and Admin Completion Queue (ACQ)
69- Asynchronous Event Notification Queue (AENQ)
70
71ENA device MMIO Registers are accessed only during driver
72initialization and are not used during further normal device
73operation.
74
75AQ is used for submitting management commands, and the
76results/responses are reported asynchronously through ACQ.
77
78ENA introduces a small set of management commands with room for
79vendor-specific extensions. Most of the management operations are
80framed in a generic Get/Set feature command.
81
82The following admin queue commands are supported:
83
84- Create I/O submission queue
85- Create I/O completion queue
86- Destroy I/O submission queue
87- Destroy I/O completion queue
88- Get feature
89- Set feature
90- Configure AENQ
91- Get statistics
92
93Refer to ena_admin_defs.h for the list of supported Get/Set Feature
94properties.
95
96The Asynchronous Event Notification Queue (AENQ) is a uni-directional
97queue used by the ENA device to send to the driver events that cannot
98be reported using ACQ. AENQ events are subdivided into groups. Each
99group may have multiple syndromes, as shown below
100
101The events are:
102
103====================    ===============
104Group                   Syndrome
105====================    ===============
106Link state change       **X**
107Fatal error             **X**
108Notification            Suspend traffic
109Notification            Resume traffic
110Keep-Alive              **X**
111====================    ===============
112
113ACQ and AENQ share the same MSI-X vector.
114
115Keep-Alive is a special mechanism that allows monitoring the device's health.
116A Keep-Alive event is delivered by the device every second.
117The driver maintains a watchdog (WD) handler which logs the current state and
118statistics. If the keep-alive events aren't delivered as expected the WD resets
119the device and the driver.
120
121Data Path Interface
122===================
123
124I/O operations are based on Tx and Rx Submission Queues (Tx SQ and Rx
125SQ correspondingly). Each SQ has a completion queue (CQ) associated
126with it.
127
128The SQs and CQs are implemented as descriptor rings in contiguous
129physical memory.
130
131The ENA driver supports two Queue Operation modes for Tx SQs:
132
133- **Regular mode:**
134  In this mode the Tx SQs reside in the host's memory. The ENA
135  device fetches the ENA Tx descriptors and packet data from host
136  memory.
137
138- **Low Latency Queue (LLQ) mode or "push-mode":**
139  In this mode the driver pushes the transmit descriptors and the
140  first 96 bytes of the packet directly to the ENA device memory
141  space. The rest of the packet payload is fetched by the
142  device. For this operation mode, the driver uses a dedicated PCI
143  device memory BAR, which is mapped with write-combine capability.
144
145  **Note that** not all ENA devices support LLQ, and this feature is negotiated
146  with the device upon initialization. If the ENA device does not
147  support LLQ mode, the driver falls back to the regular mode.
148
149The Rx SQs support only the regular mode.
150
151The driver supports multi-queue for both Tx and Rx. This has various
152benefits:
153
154- Reduced CPU/thread/process contention on a given Ethernet interface.
155- Cache miss rate on completion is reduced, particularly for data
156  cache lines that hold the sk_buff structures.
157- Increased process-level parallelism when handling received packets.
158- Increased data cache hit rate, by steering kernel processing of
159  packets to the CPU, where the application thread consuming the
160  packet is running.
161- In hardware interrupt re-direction.
162
163Interrupt Modes
164===============
165
166The driver assigns a single MSI-X vector per queue pair (for both Tx
167and Rx directions). The driver assigns an additional dedicated MSI-X vector
168for management (for ACQ and AENQ).
169
170Management interrupt registration is performed when the Linux kernel
171probes the adapter, and it is de-registered when the adapter is
172removed. I/O queue interrupt registration is performed when the Linux
173interface of the adapter is opened, and it is de-registered when the
174interface is closed.
175
176The management interrupt is named::
177
178   ena-mgmnt@pci:<PCI domain:bus:slot.function>
179
180and for each queue pair, an interrupt is named::
181
182   <interface name>-Tx-Rx-<queue index>
183
184The ENA device operates in auto-mask and auto-clear interrupt
185modes. That is, once MSI-X is delivered to the host, its Cause bit is
186automatically cleared and the interrupt is masked. The interrupt is
187unmasked by the driver after NAPI processing is complete.
188
189Interrupt Moderation
190====================
191
192ENA driver and device can operate in conventional or adaptive interrupt
193moderation mode.
194
195**In conventional mode** the driver instructs device to postpone interrupt
196posting according to static interrupt delay value. The interrupt delay
197value can be configured through `ethtool(8)`. The following `ethtool`
198parameters are supported by the driver: ``tx-usecs``, ``rx-usecs``
199
200**In adaptive interrupt** moderation mode the interrupt delay value is
201updated by the driver dynamically and adjusted every NAPI cycle
202according to the traffic nature.
203
204Adaptive coalescing can be switched on/off through `ethtool(8)`'s
205:code:`adaptive_rx on|off` parameter.
206
207More information about Adaptive Interrupt Moderation (DIM) can be found in
208Documentation/networking/net_dim.rst
209
210.. _`RX copybreak`:
211
212RX copybreak
213============
214
215The rx_copybreak is initialized by default to ENA_DEFAULT_RX_COPYBREAK
216and can be configured by the ETHTOOL_STUNABLE command of the
217SIOCETHTOOL ioctl.
218
219This option controls the maximum packet length for which the RX
220descriptor it was received on would be recycled. When a packet smaller
221than RX copybreak bytes is received, it is copied into a new memory
222buffer and the RX descriptor is returned to HW.
223
224Statistics
225==========
226
227The user can obtain ENA device and driver statistics using `ethtool`.
228The driver can collect regular or extended statistics (including
229per-queue stats) from the device.
230
231In addition the driver logs the stats to syslog upon device reset.
232
233On supported instance types, the statistics will also include the
234ENA Express data (fields prefixed with `ena_srd`). For a complete
235documentation of ENA Express data refer to
236https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ena-express.html#ena-express-monitor
237
238MTU
239===
240
241The driver supports an arbitrarily large MTU with a maximum that is
242negotiated with the device. The driver configures MTU using the
243SetFeature command (ENA_ADMIN_MTU property). The user can change MTU
244via `ip(8)` and similar legacy tools.
245
246Stateless Offloads
247==================
248
249The ENA driver supports:
250
251- IPv4 header checksum offload
252- TCP/UDP over IPv4/IPv6 checksum offloads
253
254RSS
255===
256
257- The ENA device supports RSS that allows flexible Rx traffic
258  steering.
259- Toeplitz and CRC32 hash functions are supported.
260- Different combinations of L2/L3/L4 fields can be configured as
261  inputs for hash functions.
262- The driver configures RSS settings using the AQ SetFeature command
263  (ENA_ADMIN_RSS_HASH_FUNCTION, ENA_ADMIN_RSS_HASH_INPUT and
264  ENA_ADMIN_RSS_INDIRECTION_TABLE_CONFIG properties).
265- If the NETIF_F_RXHASH flag is set, the 32-bit result of the hash
266  function delivered in the Rx CQ descriptor is set in the received
267  SKB.
268- The user can provide a hash key, hash function, and configure the
269  indirection table through `ethtool(8)`.
270
271DATA PATH
272=========
273
274Tx
275--
276
277:code:`ena_start_xmit()` is called by the stack. This function does the following:
278
279- Maps data buffers (``skb->data`` and frags).
280- Populates ``ena_buf`` for the push buffer (if the driver and device are
281  in push mode).
282- Prepares ENA bufs for the remaining frags.
283- Allocates a new request ID from the empty ``req_id`` ring. The request
284  ID is the index of the packet in the Tx info. This is used for
285  out-of-order Tx completions.
286- Adds the packet to the proper place in the Tx ring.
287- Calls :code:`ena_com_prepare_tx()`, an ENA communication layer that converts
288  the ``ena_bufs`` to ENA descriptors (and adds meta ENA descriptors as
289  needed).
290
291  * This function also copies the ENA descriptors and the push buffer
292    to the Device memory space (if in push mode).
293
294- Writes a doorbell to the ENA device.
295- When the ENA device finishes sending the packet, a completion
296  interrupt is raised.
297- The interrupt handler schedules NAPI.
298- The :code:`ena_clean_tx_irq()` function is called. This function handles the
299  completion descriptors generated by the ENA, with a single
300  completion descriptor per completed packet.
301
302  * ``req_id`` is retrieved from the completion descriptor. The ``tx_info`` of
303    the packet is retrieved via the ``req_id``. The data buffers are
304    unmapped and ``req_id`` is returned to the empty ``req_id`` ring.
305  * The function stops when the completion descriptors are completed or
306    the budget is reached.
307
308Rx
309--
310
311- When a packet is received from the ENA device.
312- The interrupt handler schedules NAPI.
313- The :code:`ena_clean_rx_irq()` function is called. This function calls
314  :code:`ena_com_rx_pkt()`, an ENA communication layer function, which returns the
315  number of descriptors used for a new packet, and zero if
316  no new packet is found.
317- :code:`ena_rx_skb()` checks packet length:
318
319  * If the packet is small (len < rx_copybreak), the driver allocates
320    a SKB for the new packet, and copies the packet payload into the
321    SKB data buffer.
322
323    - In this way the original data buffer is not passed to the stack
324      and is reused for future Rx packets.
325
326  * Otherwise the function unmaps the Rx buffer, sets the first
327    descriptor as `skb`'s linear part and the other descriptors as the
328    `skb`'s frags.
329
330- The new SKB is updated with the necessary information (protocol,
331  checksum hw verify result, etc), and then passed to the network
332  stack, using the NAPI interface function :code:`napi_gro_receive()`.
333
334Dynamic RX Buffers (DRB)
335------------------------
336
337Each RX descriptor in the RX ring is a single memory page (which is either 4KB
338or 16KB long depending on system's configurations).
339To reduce the memory allocations required when dealing with a high rate of small
340packets, the driver tries to reuse the remaining RX descriptor's space if more
341than 2KB of this page remain unused.
342
343A simple example of this mechanism is the following sequence of events:
344
345::
346
347        1. Driver allocates page-sized RX buffer and passes it to hardware
348                +----------------------+
349                |4KB RX Buffer         |
350                +----------------------+
351
352        2. A 300Bytes packet is received on this buffer
353
354        3. The driver increases the ref count on this page and returns it back to
355           HW as an RX buffer of size 4KB - 300Bytes = 3796 Bytes
356               +----+--------------------+
357               |****|3796 Bytes RX Buffer|
358               +----+--------------------+
359
360This mechanism isn't used when an XDP program is loaded, or when the
361RX packet is less than rx_copybreak bytes (in which case the packet is
362copied out of the RX buffer into the linear part of a new skb allocated
363for it and the RX buffer remains the same size, see `RX copybreak`_).
364