freebsd-nq/share/man/man4/ptnet.4
Vincenzo Maffione 76ed4255b5 netmap: add man page for ptnet
Add a man page for ptnet(4), describing the guest driver for netmap passthrough.

Reviewed by:	bcr
MFC after:	3 days
Differential Revision:	https://reviews.freebsd.org/D18518
2018-12-12 17:05:47 +00:00

141 lines
4.7 KiB
Groff

.\" Copyright (c) 2018 Vincenzo Maffione
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd December 11, 2018
.Dt PTNET 4
.Os
.Sh NAME
.Nm ptnet
.Nd Ethernet driver for passed-through netmap ports
.Sh SYNOPSIS
This network driver is included in
.Xr netmap 4 ,
and it can be compiled into the kernel by adding the following
line in your kernel configuration file:
.Bd -ragged -offset indent
.Cd "device netmap"
.Ed
.Sh DESCRIPTION
The
.Nm
device driver provides direct access to host netmap ports,
from within a Virtual Machine (VM).
Applications running inside
the VM can access the TX/RX rings and buffers of a netmap port
that the hypervisor has passed-through to the VM.
Hypervisor support for
.Nm
is currently available for QEMU/KVM.
Any
.Xr netmap 4
port can be passed-through, including physical NICs,
.Xr vale 4
ports, netmap pipes, etc.
.Pp
The main use-case for netmap passthrough is Network Function
Virtualization (NFV), where middlebox applications running within
VMs may want to process very high packet rates (e.g., 1-10 millions
packets per second or more).
Note, however, that those applications
must use the device in netmap mode in order to achieve such rates.
In addition to the general advantages of netmap, the improved
performance of
.Nm
when compared to hypervisor device emulation or paravirtualization (e.g.,
.Xr vtnet 4 ,
.Xr vmx 4 )
comes from the hypervisor being completely bypassed in the data-path.
For example, when using
.Xr vtnet 4
the VM has to convert each
.Xr mbuf 9
to a VirtIO-specific packet representation
and publish that to a VirtIO queue; on the hypervisor side, the
packet is extracted from the VirtIO queue and converted to a
hypervisor-specific packet representation.
The overhead of format conversions (and packet copies, in same cases) is not
incured by
.Nm
in netmap mode, because mbufs are not used at all, and the packet format
is the one defined by netmap (e.g.,
.Ar struct netmap_slot )
along the whole data-path.
No format conversions or copies happen.
.Pp
It is also possible to use a
.Nm
device like a regular network interface, which interacts with the
.Fx
network stack (i.e., not in netmap mode).
However, in that case it is necessary to pay the cost of data copies
between mbufs and netmap buffers, which generally results in lower
TCP/UDP performance than
.Xr vtnet 4
or other paravirtualized network devices.
If the passed-through netmap port supports the VirtIO network header,
.Nm
is able to use it, and support TCP/UDP checksum offload (for both transmit
and receive), TCP segmentation offload (TSO) and TCP large receive offload
(LRO).
Currently,
.Xr vale 4
ports support the header.
Note that the VirtIO network header is generally not used in NFV
use-cases, because middleboxes are not endpoints of TCP/UDP connections.
.Sh TUNABLES
Tunables can be set at the
.Xr loader 8
prompt before booting the kernel or stored in
.Xr loader.conf 5 .
.Bl -tag -width "xxxxxx"
.It Va dev.netmap.ptnet_vnet_hdr
This tunable enables (1) or disables (0) the VirtIO network header.
If enabled,
.Nm
uses the same header used by
.Xr vtnet 4
to exchange offload metadata with the hypervisor.
If disabled, no header is prepended to transmitted and received
packets.
The metadata is necessary to support TCP/UDP checksum offloads,
TSO, and LRO.
The default value is 1.
.El
.Sh SEE ALSO
.Xr netintro 4 ,
.Xr netmap 4 ,
.Xr vale 4 ,
.Xr virtio 4 ,
.Xr vmx 4 ,
.Xr ifconfig 8
.Sh HISTORY
The
.Nm
driver was written by
.An Vincenzo Maffione Aq Mt vmaffione@FreeBSD.org .
It first appeared in
.Fx 12.0 .