xref: /freebsd/share/man/man4/gve.4 (revision fa38579f317d5c2ff2926fab9b12ee6d429bd155)
1.\" SPDX-License-Identifier: BSD-3-Clause
2.\"
3.\" Copyright (c) 2023-2024 Google LLC
4.\"
5.\" Redistribution and use in source and binary forms, with or without modification,
6.\" are permitted provided that the following conditions are met:
7.\"
8.\" 1. Redistributions of source code must retain the above copyright notice, this
9.\"    list of conditions and the following disclaimer.
10.\"
11.\" 2. Redistributions in binary form must reproduce the above copyright notice,
12.\"    this list of conditions and the following disclaimer in the documentation
13.\"    and/or other materials provided with the distribution.
14.\"
15.\" 3. Neither the name of the copyright holder nor the names of its contributors
16.\"    may be used to endorse or promote products derived from this software without
17.\"    specific prior written permission.
18.\"
19.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
20.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
21.\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22.\" DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
23.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
24.\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25.\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
26.\" ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
28.\" SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29.Dd October 14, 2024
30.Dt GVE 4
31.Os
32.Sh NAME
33.Nm gve
34.Nd "Ethernet driver for Google Virtual NIC (gVNIC)"
35.Sh SYNOPSIS
36To compile this driver into the kernel,
37place the following lines in your
38kernel configuration file:
39.Bd -ragged -offset indent
40.Cd "device gve"
41.Ed
42.Pp
43Alternatively, to load the driver as a
44module at boot time, place the following line in
45.Xr loader.conf 5 :
46.Bd -literal -offset indent
47if_gve_load="YES"
48.Ed
49.Sh DESCRIPTION
50gVNIC is a virtual network interface designed specifically for Google Compute Engine (GCE).
51It is required to support per-VM Tier-1 networking performance, and for using certain VM shapes on GCE.
52.Pp
53.Nm
54is the driver for gVNIC.
55It supports the following features:
56.Pp
57.Bl -bullet -compact
58.It
59RX checksum offload
60.It
61TX chesksum offload
62.It
63TCP Segmentation Offload (TSO)
64.It
65Large Receive Offload (LRO) in software
66.It
67Jumbo frames
68.It
69Receive Side Scaling (RSS)
70.El
71.Pp
72For more information on configuring this device, see
73.Xr ifconfig 8 .
74.Sh HARDWARE
75.Nm
76binds to a single PCI device ID presented by gVNIC:
77.Pp
78.Bl -bullet -compact
79.It
800x1AE0:0x0042
81.El
82.Sh EXAMPLES
83.Pp
84Change the TX queue count to 4 for the gve0 interface:
85.D1 sysctl dev.gve.0.num_tx_queues=4
86.Pp
87Change the RX queue count to 4 for the gve0 interface:
88.D1 sysctl dev.gve.0.num_rx_queues=4
89.Pp
90Change the TX ring size to 512 for the gve0 interface:
91.D1 sysctl dev.gve.0.tx_ring_size=512
92.Pp
93Change the RX ring size to 512 for the gve0 interface:
94.D1 sysctl dev.gve.0.rx_ring_size=512
95.Sh DIAGNOSTICS
96The following messages are recorded during driver initialization:
97.Bl -diag
98.It "Enabled MSIX with %d vectors"
99.It "Configured device resources"
100.It "Successfully attached %s"
101.It "Deconfigured device resources"
102.El
103.Pp
104These messages are seen if driver initialization fails.
105Global (across-queues) allocation failures:
106.Bl -diag
107.It "Failed to configure device resources: err=%d"
108.It "No compatible queue formats"
109.It "Failed to allocate ifnet struct"
110.It "Failed to allocate admin queue mem"
111.It "Failed to alloc DMA mem for DescribeDevice"
112.It "Failed to allocate QPL page"
113.El
114.Pp
115irq and BAR allocation failures:
116.Bl -diag
117.It "Failed to acquire any msix vectors"
118.It "Tried to acquire %d msix vectors, got only %d"
119.It "Failed to setup irq %d for Tx queue %d "
120.It "Failed to setup irq %d for Rx queue %d "
121.It "Failed to allocate irq %d for mgmnt queue"
122.It "Failed to setup irq %d for mgmnt queue, err: %d"
123.It "Failed to allocate BAR0"
124.It "Failed to allocate BAR2"
125.It "Failed to allocate msix table"
126.El
127.Pp
128Rx queue-specific allocation failures:
129.Bl -diag
130.It "No QPL left for rx ring %d"
131.It "Failed to alloc queue resources for rx ring %d"
132.It "Failed to alloc desc ring for rx ring %d"
133.It "Failed to alloc data ring for rx ring %d"
134.El
135.Pp
136Tx queue-specific allocation failures:
137.Bl -diag
138.It "No QPL left for tx ring %d"
139.It "Failed to alloc queue resources for tx ring %d"
140.It "Failed to alloc desc ring for tx ring %d"
141.It "Failed to vmap fifo, qpl_id = %d"
142.El
143.Pp
144The following messages are recorded when the interface detach fails:
145.Bl -diag
146.It "Failed to deconfigure device resources: err=%d"
147.El
148.Pp
149If bootverbose is on, the following messages are recorded when the interface is being brought up:
150.Bl -diag
151.It "Created %d rx queues"
152.It "Created %d tx queues"
153.It "MTU set to %d"
154.El
155.Pp
156The following messages are recorded when the interface is being brought down:
157.Bl -diag
158.It "Destroyed %d rx queues"
159.It "Destroyed %d tx queues"
160.El
161.Pp
162These messages are seen if errors are encountered when bringing the interface up or down:
163.Bl -diag
164.It "Failed to destroy rxq %d, err: %d"
165.It "Failed to destroy txq %d, err: %d"
166.It "Failed to create rxq %d, err: %d"
167.It "Failed to create txq %d, err: %d"
168.It "Failed to set MTU to %d"
169.It "Invalid new MTU setting. new mtu: %d max mtu: %d min mtu: %d"
170.It "Cannot bring the iface up when detached"
171.It "Reached max number of registered pages %lu > %lu"
172.It "Failed to init lro for rx ring %d"
173.El
174.Pp
175These messages are seen if any admin queue command fails:
176.Bl -diag
177.It "AQ command(%u): failed with status %d"
178.It "AQ command(%u): unknown status code %d"
179.It "AQ commands timed out, need to reset AQ"
180.It "Unknown AQ command opcode %d"
181.El
182.Pp
183These messages are recorded when the device is being reset due to an error:
184.Bl -diag
185.It "Scheduling reset task!"
186.It "Waiting until admin queue is released."
187.It "Admin queue released"
188.El
189.Pp
190If it was the NIC that requested the reset, this message is recorded:
191.Bl -diag
192.It "Device requested reset"
193.El
194.Pp
195If the reset fails during the reinitialization phase, this message is recorded:
196.Bl -diag
197.It "Restore failed!"
198.El
199.Pp
200These two messages correspond to the NIC alerting the driver to link state changes:
201.Bl -diag
202.It "Device link is up."
203.It "Device link is down."
204.El
205.Pp
206Apart from these messages, the driver exposes per-queue packet and error counters as sysctl nodes.
207Global (across queues) counters can be read using
208.Xr netstat 1 .
209.Sh SYSCTL VARIABLES
210.Nm
211exposes the following
212.Xr sysctl 8
213variables:
214.Bl -tag -width indent
215.It Va hw.gve.driver_version
216The driver version.
217This is read-only.
218.It Va hw.gve.queue_format
219The queue format in use.
220This is read-only.
221.It Va hw.gve.disable_hw_lro
222Setting this boot-time tunable to 1 disables Large Receive Offload (LRO) in the NIC.
223The default value is 0, which means hardware LRO is enabled by default.
224The software LRO stack in the kernel is always used.
225This sysctl variable needs to be set before loading the driver, using
226.Xr loader.conf 5 .
227.It Va dev.gve.X.num_rx_queues and dev.gve.X.num_tx_queues
228Run-time tunables that represent the number of currently used RX/TX queues.
229The default value is the max number of RX/TX queues the device can support.
230.Pp
231This call turns down the interface while setting up the new queues,
232which may potentially cause any new packets to be dropped.
233This call can fail if the system is not able to provide the driver with enough resources.
234In that situation, the driver will revert to the previous number of RX/TX queues.
235If this also fails, a device reset will be triggered.
236.Pp
237Note: sysctl nodes for queue stats remain available even if a queue is removed.
238.Pp
239.It Va dev.gve.X.rx_ring_size and dev.gve.X.tx_ring_size
240Run-time tunables that represent the current ring size for RX/TX queues.
241The default value is set to device defaults for ring size.
242.Pp
243This call turns down the interface while setting up the queues with the new ring size,
244which may potentially cause any new packets to be dropped.
245This call can fail if the system is not able to provide the driver with enough resources.
246In that situation, the driver will try to revert to the previous ring size for RX/TX queues.
247If this also fails, the device will be in an unhealthy state and will need to be reloaded.
248This value must be a power of 2 and within the defined range.
249.Pp
250.El
251.Sh LIMITATIONS
252.Nm
253does not support the transmission of VLAN-tagged packets.
254All VLAN-tagged traffic is dropped.
255.Sh QUEUE FORMATS
256.Nm
257features different datapath modes called queue formats:
258.Pp
259.Bl -bullet -compact
260.It
261GQI_QPL: "QPL" stands for "Queue Page List" and refers to the fact that
262hardware expects a fixed bounce buffer and cannot access arbitrary memory.
263GQI is the older descriptor format.
264The G in "GQI" refers to an older generation of hardware, and the "QI"
265stands for "Queue In-order" referring to the fact that the NIC sends
266Tx and Rx completions in the same order as the one in which the corresponding
267descriptors were posted by the driver.
268.It
269DQO_RDA: DQO is the descriptor format required to take full advantage of
270next generation VM shapes.
271"RDA" stands for "Raw DMA Addressing" and refers to the fact that hardware
272can work with DMA-ed packets and does not expect them to be copied into or
273out of a fixed bounce buffer.
274The D in "DQO" refers to a newer generation of hardware, and the "QO"
275stands for "Queue Out-of-order" referring to the fact that the NIC might
276send Tx and Rx completions in an order different from the one in which
277the corresponding descriptors were posted by the driver.
278.It
279DQO_QPL: The next generation descriptor format in the "QPL" mode.
280.El
281.Sh SUPPORT
282Please email gvnic-drivers@google.com with the specifics of the issue encountered.
283.Sh SEE ALSO
284.Xr netstat 1 ,
285.Xr loader.conf 5 ,
286.Xr ifconfig 8 ,
287.Xr sysctl 8
288.Sh HISTORY
289The
290.Nm
291device driver first appeared in
292.Fx 13.3 .
293.Sh AUTHORS
294The
295.Nm
296driver was written by Google.
297