1.. SPDX-License-Identifier: GPL-2.0
2
3====================
4mlx5 devlink support
5====================
6
7This document describes the devlink features implemented by the ``mlx5``
8device driver.
9
10Parameters
11==========
12
13.. list-table:: Generic parameters implemented
14
15   * - Name
16     - Mode
17     - Validation
18   * - ``enable_roce``
19     - driverinit
20     - Type: Boolean
21
22       If the device supports RoCE disablement, RoCE enablement state controls
23       device support for RoCE capability. Otherwise, the control occurs in the
24       driver stack. When RoCE is disabled at the driver level, only raw
25       ethernet QPs are supported.
26   * - ``io_eq_size``
27     - driverinit
28     - The range is between 64 and 4096.
29   * - ``event_eq_size``
30     - driverinit
31     - The range is between 64 and 4096.
32   * - ``max_macs``
33     - driverinit
34     - The range is between 1 and 2^31. Only power of 2 values are supported.
35
36The ``mlx5`` driver also implements the following driver-specific
37parameters.
38
39.. list-table:: Driver-specific parameters implemented
40   :widths: 5 5 5 85
41
42   * - Name
43     - Type
44     - Mode
45     - Description
46   * - ``flow_steering_mode``
47     - string
48     - runtime
49     - Controls the flow steering mode of the driver
50
51       * ``dmfs`` Device managed flow steering. In DMFS mode, the HW
52         steering entities are created and managed through firmware.
53       * ``smfs`` Software managed flow steering. In SMFS mode, the HW
54         steering entities are created and manage through the driver without
55         firmware intervention.
56
57       SMFS mode is faster and provides better rule insertion rate compared to
58       default DMFS mode.
59   * - ``fdb_large_groups``
60     - u32
61     - driverinit
62     - Control the number of large groups (size > 1) in the FDB table.
63
64       * The default value is 15, and the range is between 1 and 1024.
65   * - ``esw_multiport``
66     - Boolean
67     - runtime
68     - Control MultiPort E-Switch shared fdb mode.
69
70       An experimental mode where a single E-Switch is used and all the vports
71       and physical ports on the NIC are connected to it.
72
73       An example is to send traffic from a VF that is created on PF0 to an
74       uplink that is natively associated with the uplink of PF1
75
76       Note: Future devices, ConnectX-8 and onward, will eventually have this
77       as the default to allow forwarding between all NIC ports in a single
78       E-switch environment and the dual E-switch mode will likely get
79       deprecated.
80
81       Default: disabled
82   * - ``esw_port_metadata``
83     - Boolean
84     - runtime
85     - When applicable, disabling eswitch metadata can increase packet rate up
86       to 20% depending on the use case and packet sizes.
87
88       Eswitch port metadata state controls whether to internally tag packets
89       with metadata. Metadata tagging must be enabled for multi-port RoCE,
90       failover between representors and stacked devices. By default metadata is
91       enabled on the supported devices in E-switch. Metadata is applicable only
92       for E-switch in switchdev mode and users may disable it when NONE of the
93       below use cases will be in use:
94       1. HCA is in Dual/multi-port RoCE mode.
95       2. VF/SF representor bonding (Usually used for Live migration)
96       3. Stacked devices
97
98       When metadata is disabled, the above use cases will fail to initialize if
99       users try to enable them.
100   * - ``hairpin_num_queues``
101     - u32
102     - driverinit
103     - We refer to a TC NIC rule that involves forwarding as "hairpin".
104       Hairpin queues are mlx5 hardware specific implementation for hardware
105       forwarding of such packets.
106
107       Control the number of hairpin queues.
108   * - ``hairpin_queue_size``
109     - u32
110     - driverinit
111     - Control the size (in packets) of the hairpin queues.
112
113The ``mlx5`` driver supports reloading via ``DEVLINK_CMD_RELOAD``
114
115Info versions
116=============
117
118The ``mlx5`` driver reports the following versions
119
120.. list-table:: devlink info versions implemented
121   :widths: 5 5 90
122
123   * - Name
124     - Type
125     - Description
126   * - ``fw.psid``
127     - fixed
128     - Used to represent the board id of the device.
129   * - ``fw.version``
130     - stored, running
131     - Three digit major.minor.subminor firmware version number.
132
133Health reporters
134================
135
136tx reporter
137-----------
138The tx reporter is responsible for reporting and recovering of the following three error scenarios:
139
140- tx timeout
141    Report on kernel tx timeout detection.
142    Recover by searching lost interrupts.
143- tx error completion
144    Report on error tx completion.
145    Recover by flushing the tx queue and reset it.
146- tx PTP port timestamping CQ unhealthy
147    Report too many CQEs never delivered on port ts CQ.
148    Recover by flushing and re-creating all PTP channels.
149
150tx reporter also support on demand diagnose callback, on which it provides
151real time information of its send queues status.
152
153User commands examples:
154
155- Diagnose send queues status::
156
157    $ devlink health diagnose pci/0000:82:00.0 reporter tx
158
159.. note::
160   This command has valid output only when interface is up, otherwise the command has empty output.
161
162- Show number of tx errors indicated, number of recover flows ended successfully,
163  is autorecover enabled and graceful period from last recover::
164
165    $ devlink health show pci/0000:82:00.0 reporter tx
166
167rx reporter
168-----------
169The rx reporter is responsible for reporting and recovering of the following two error scenarios:
170
171- rx queues' initialization (population) timeout
172    Population of rx queues' descriptors on ring initialization is done
173    in napi context via triggering an irq. In case of a failure to get
174    the minimum amount of descriptors, a timeout would occur, and
175    descriptors could be recovered by polling the EQ (Event Queue).
176- rx completions with errors (reported by HW on interrupt context)
177    Report on rx completion error.
178    Recover (if needed) by flushing the related queue and reset it.
179
180rx reporter also supports on demand diagnose callback, on which it
181provides real time information of its receive queues' status.
182
183- Diagnose rx queues' status and corresponding completion queue::
184
185    $ devlink health diagnose pci/0000:82:00.0 reporter rx
186
187.. note::
188   This command has valid output only when interface is up. Otherwise, the command has empty output.
189
190- Show number of rx errors indicated, number of recover flows ended successfully,
191  is autorecover enabled, and graceful period from last recover::
192
193    $ devlink health show pci/0000:82:00.0 reporter rx
194
195fw reporter
196-----------
197The fw reporter implements `diagnose` and `dump` callbacks.
198It follows symptoms of fw error such as fw syndrome by triggering
199fw core dump and storing it into the dump buffer.
200The fw reporter diagnose command can be triggered any time by the user to check
201current fw status.
202
203User commands examples:
204
205- Check fw heath status::
206
207    $ devlink health diagnose pci/0000:82:00.0 reporter fw
208
209- Read FW core dump if already stored or trigger new one::
210
211    $ devlink health dump show pci/0000:82:00.0 reporter fw
212
213.. note::
214   This command can run only on the PF which has fw tracer ownership,
215   running it on other PF or any VF will return "Operation not permitted".
216
217fw fatal reporter
218-----------------
219The fw fatal reporter implements `dump` and `recover` callbacks.
220It follows fatal errors indications by CR-space dump and recover flow.
221The CR-space dump uses vsc interface which is valid even if the FW command
222interface is not functional, which is the case in most FW fatal errors.
223The recover function runs recover flow which reloads the driver and triggers fw
224reset if needed.
225On firmware error, the health buffer is dumped into the dmesg. The log
226level is derived from the error's severity (given in health buffer).
227
228User commands examples:
229
230- Run fw recover flow manually::
231
232    $ devlink health recover pci/0000:82:00.0 reporter fw_fatal
233
234- Read FW CR-space dump if already stored or trigger new one::
235
236    $ devlink health dump show pci/0000:82:00.1 reporter fw_fatal
237
238.. note::
239   This command can run only on PF.
240
241vnic reporter
242-------------
243The vnic reporter implements only the `diagnose` callback.
244It is responsible for querying the vnic diagnostic counters from fw and displaying
245them in realtime.
246
247Description of the vnic counters:
248
249- total_q_under_processor_handle
250        number of queues in an error state due to
251        an async error or errored command.
252- send_queue_priority_update_flow
253        number of QP/SQ priority/SL update events.
254- cq_overrun
255        number of times CQ entered an error state due to an overflow.
256- async_eq_overrun
257        number of times an EQ mapped to async events was overrun.
258        comp_eq_overrun number of times an EQ mapped to completion events was
259        overrun.
260- quota_exceeded_command
261        number of commands issued and failed due to quota exceeded.
262- invalid_command
263        number of commands issued and failed dues to any reason other than quota
264        exceeded.
265- nic_receive_steering_discard
266        number of packets that completed RX flow
267        steering but were discarded due to a mismatch in flow table.
268- generated_pkt_steering_fail
269	number of packets generated by the VNIC experiencing unexpected steering
270	failure (at any point in steering flow).
271- handled_pkt_steering_fail
272	number of packets handled by the VNIC experiencing unexpected steering
273	failure (at any point in steering flow owned by the VNIC, including the FDB
274	for the eswitch owner).
275
276User commands examples:
277
278- Diagnose PF/VF vnic counters::
279
280        $ devlink health diagnose pci/0000:82:00.1 reporter vnic
281
282- Diagnose representor vnic counters (performed by supplying devlink port of the
283  representor, which can be obtained via devlink port command)::
284
285        $ devlink health diagnose pci/0000:82:00.1/65537 reporter vnic
286
287.. note::
288   This command can run over all interfaces such as PF/VF and representor ports.
289