AF_XDP is a network socket family that allows communication directly
with the network device driver in the kernel, bypassing most or all
of the kernel networking stack. In the essence, the technology is
pretty similar to netmap. But, unlike netmap, AF_XDP is Linux-native
and works with any network interfaces without driver modifications.
Unlike vhost-based backends (kernel, user, vdpa), AF_XDP doesn't
require access to character devices or unix sockets. Only access to
the network interface itself is necessary.
This patch implements a network backend that communicates with the
kernel by creating an AF_XDP socket. A chunk of userspace memory
is shared between QEMU and the host kernel. 4 ring buffers (Tx, Rx,
Fill and Completion) are placed in that memory along with a pool of
memory buffers for the packet data. Data transmission is done by
allocating one of the buffers, copying packet data into it and
placing the pointer into Tx ring. After transmission, device will
return the buffer via Completion ring. On Rx, device will take
a buffer form a pre-populated Fill ring, write the packet data into
it and place the buffer into Rx ring.
AF_XDP network backend takes on the communication with the host
kernel and the network interface and forwards packets to/from the
peer device in QEMU.
Usage example:
-device virtio-net-pci,netdev=guest1,mac=00:16:35:AF:AA:5C
-netdev af-xdp,ifname=ens6f1np1,id=guest1,mode=native,queues=1
XDP program bridges the socket with a network interface. It can be
attached to the interface in 2 different modes:
1. skb - this mode should work for any interface and doesn't require
driver support. With a caveat of lower performance.
2. native - this does require support from the driver and allows to
bypass skb allocation in the kernel and potentially use
zero-copy while getting packets in/out userspace.
By default, QEMU will try to use native mode and fall back to skb.
Mode can be forced via 'mode' option. To force 'copy' even in native
mode, use 'force-copy=on' option. This might be useful if there is
some issue with the driver.
Option 'queues=N' allows to specify how many device queues should
be open. Note that all the queues that are not open are still
functional and can receive traffic, but it will not be delivered to
QEMU. So, the number of device queues should generally match the
QEMU configuration, unless the device is shared with something
else and the traffic re-direction to appropriate queues is correctly
configured on a device level (e.g. with ethtool -N).
'start-queue=M' option can be used to specify from which queue id
QEMU should start configuring 'N' queues. It might also be necessary
to use this option with certain NICs, e.g. MLX5 NICs. See the docs
for examples.
In a general case QEMU will need CAP_NET_ADMIN and CAP_SYS_ADMIN
or CAP_BPF capabilities in order to load default XSK/XDP programs to
the network interface and configure BPF maps. It is possible, however,
to run with no capabilities. For that to work, an external process
with enough capabilities will need to pre-load default XSK program,
create AF_XDP sockets and pass their file descriptors to QEMU process
on startup via 'sock-fds' option. Network backend will need to be
configured with 'inhibit=on' to avoid loading of the program.
QEMU will need 32 MB of locked memory (RLIMIT_MEMLOCK) per queue
or CAP_IPC_LOCK.
There are few performance challenges with the current network backends.
First is that they do not support IO threads. This means that data
path is handled by the main thread in QEMU and may slow down other
work or may be slowed down by some other work. This also means that
taking advantage of multi-queue is generally not possible today.
Another thing is that data path is going through the device emulation
code, which is not really optimized for performance. The fastest
"frontend" device is virtio-net. But it's not optimized for heavy
traffic either, because it expects such use-cases to be handled via
some implementation of vhost (user, kernel, vdpa). In practice, we
have virtio notifications and rcu lock/unlock on a per-packet basis
and not very efficient accesses to the guest memory. Communication
channels between backend and frontend devices do not allow passing
more than one packet at a time as well.
Some of these challenges can be avoided in the future by adding better
batching into device emulation or by implementing vhost-af-xdp variant.
There are also a few kernel limitations. AF_XDP sockets do not
support any kinds of checksum or segmentation offloading. Buffers
are limited to a page size (4K), i.e. MTU is limited. Multi-buffer
support implementation for AF_XDP is in progress, but not ready yet.
Also, transmission in all non-zero-copy modes is synchronous, i.e.
done in a syscall. That doesn't allow high packet rates on virtual
interfaces.
However, keeping in mind all of these challenges, current implementation
of the AF_XDP backend shows a decent performance while running on top
of a physical NIC with zero-copy support.
Test setup:
2 VMs running on 2 physical hosts connected via ConnectX6-Dx card.
Network backend is configured to open the NIC directly in native mode.
The driver supports zero-copy. NIC is configured to use 1 queue.
Inside a VM - iperf3 for basic TCP performance testing and dpdk-testpmd
for PPS testing.
iperf3 result:
TCP stream : 19.1 Gbps
dpdk-testpmd (single queue, single CPU core, 64 B packets) results:
Tx only : 3.4 Mpps
Rx only : 2.0 Mpps
L2 FWD Loopback : 1.5 Mpps
In skb mode the same setup shows much lower performance, similar to
the setup where pair of physical NICs is replaced with veth pair:
iperf3 result:
TCP stream : 9 Gbps
dpdk-testpmd (single queue, single CPU core, 64 B packets) results:
Tx only : 1.2 Mpps
Rx only : 1.0 Mpps
L2 FWD Loopback : 0.7 Mpps
Results in skb mode or over the veth are close to results of a tap
backend with vhost=on and disabled segmentation offloading bridged
with a NIC.
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> (docker/lcitool)
Signed-off-by: Jason Wang <jasowang@redhat.com>
Copied from socket netdev file and modified to use SocketAddress
to be able to introduce new features like unix socket.
"udp" and "mcast" are squashed into dgram netdev, multicast is detected
according to the IP address type.
"listen" and "connect" modes are managed by stream netdev. An optional
parameter "server" defines the mode (off by default)
The two new types need to be parsed the modern way with -netdev, because
with the traditional way, the "type" field of netdev structure collides with
the "type" field of SocketAddress and prevents the correct evaluation of the
command line option. Moreover the traditional way doesn't allow to use
the same type (SocketAddress) several times with the -netdev option
(needed to specify "local" and "remote" addresses).
The previous commit paved the way for parsing the modern way, but
omitted one detail: how to pick modern vs. traditional, in
netdev_is_modern().
We want to pick based on the value of parameter "type". But how to
extract it from the option argument?
Parsing the option argument, either the modern or the traditional way,
extracts it for us, but only if parsing succeeds.
If parsing fails, there is no good option. No matter which parser we
pick, it'll be the wrong one for some arguments, and the error
reporting will be confusing.
Fortunately, the traditional parser accepts *anything* when called in
a certain way. This maximizes our chance to extract the value of
"type", and in turn minimizes the risk of confusing error reporting.
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
This patch set introduces a new net client type: vhost-vdpa.
vhost-vdpa net client will set up a vDPA device which is specified
by a "vhostdev" parameter.
Signed-off-by: Lingshan Zhu <lingshan.zhu@intel.com>
Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20200701145538.22333-15-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20180201111846.21846-4-armbru@redhat.com>
This way we no longer need NetClientOptions and can convert Netdev
into a flat union.
Signed-off-by: Kővágó, Zoltán <DirtY.iCE.hu@gmail.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <93ffdfed7054529635e6acb935150d95dc173a12.1441627176.git.DirtY.iCE.hu@gmail.com>
[rework net_client_init1() to pass Netdev by copying from NetdevLegacy,
rather than merging the two types - which means that we still need
NetClientOptions after all. Rebase to qapi changes. The bulk of the
patch is mechanical, replacing 'opts' by 'netdev->opts', while
net_client_init1() takes care of converting between legacy and modern
types.]
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1468468228-27827-2-git-send-email-eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Error reporting for netdev_add is broken: the net_client_init_fun[]
report the actual errors with (at best) error_report(), and their
caller net_client_init1() makes up a generic error on top.
For command line and HMP, this produces an mildly ugly error cascade.
In QMP, the actual errors go to stderr, and the generic error becomes
the command's error reply.
To fix this, we need to convert the net_client_init_fun[] to Error.
To permit fixing them one by one, add an Error ** parameter to the
net_client_init_fun[]. If the call fails without returning an Error,
make up the same generic Error as before. But if it returns one, use
that instead. Since none of them does so far, no functional change.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1431691143-1015-3-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This transport allows to connect a QEMU nic to a static Ethernet
over L2TPv3 tunnel. The transport supports all options present
in the Linux kernel implementation. It allows QEMU to connect
to any Linux host running kernel 3.3+, most routers and network
devices as well as other QEMU instances.
[Fixed up net_client_init1() switch statement to support -netdev
--Stefan]
Signed-off-by: Anton Ivanov <antivano@cisco.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Add a new QEMU netdev backend that is intended to invoke vhost_net with the
vhost-user backend. It uses an Unix socket chardev to establish a
communication with the 'slave' (client and server mode supported).
At runtime the netdev will handle OPEN/CLOSE events from the chardev. Upon
disconnection it will set link_down accordingly and notify virtio-net; the
virtio-net interface will go down.
Signed-off-by: Antonios Motakis <a.motakis@virtualopensystems.com>
Signed-off-by: Nikolay Nikolaev <n.nikolaev@virtualopensystems.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch adds support for a network backend based on netmap.
netmap is a framework for high speed packet I/O. You can use it
to build extremely fast traffic generators, monitors, software
switches or network middleboxes. Its companion software switch
VALE lets you interconnect virtual machines.
netmap and VALE are implemented as a non-intrusive kernel module,
support NICs from multiple vendors, are part of standard FreeBSD
distributions and available in source format for Linux too.
To compile QEMU with netmap support, use the following configure
options:
./configure [...] --enable-netmap --extra-cflags=-I/path/to/netmap/sys
where "/path/to/netmap" contains the netmap source code, available at
http://info.iet.unipi.it/~luigi/netmap/
The same webpage contains more information about the netmap project
(together with papers and presentations).
Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Move public headers to include/net, and leave private headers in net/.
Put the virtio headers in include/net/tap.h, removing the multiple copies
that existed. Leave include/net/tap.h as the interface for NICs, and
net/tap_int.h as the interface for OS-specific parts of the tap backend.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch doesn't seem much useful alone, I must admit. However,
it makes sense as part of the upcoming directory reorganization,
where I want to have include/net/tap.h as the net<->hw interface
for tap. Then having both net/tap.h and include/net/tap.h does
not work. "Fixed" by moving all the init functions to a single
header file net/clients.h.
The patch also adopts a uniform style for including net/*.h files
from net/*.c, without the net/ path.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
