man/man4/netmap.4

.\" Copyright (c) 2011 Matteo Landi, Luigi Rizzo, Universita` di Pisa
.\" 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.
.\"
.\" This document is derived in part from the enet man page (enet.4)
.\" distributed with 4.3BSD Unix.
.\"
.\" $FreeBSD$
.\" $Id: netmap.4 9662 2011-11-16 13:18:06Z luigi $: stable/8/share/man/man4/bpf.4 181694 2008-08-13 17:45:06Z ed $
.\"
.Dd February 27, 2012
.Dt NETMAP 4
.Os
.Sh NAME
.Nm netmap
.Nd a framework for fast packet I/O
.Sh SYNOPSIS
.Cd device netmap
.Sh DESCRIPTION
.Nm
is a framework for fast and safe access to network devices
(reaching 14.88 Mpps at less than 1 GHz).
.Nm
uses memory mapped buffers and metadata
(buffer indexes and lengths) to communicate with the kernel,
which is in charge of validating information through
.Pa ioctl()
and
.Pa select()/poll().
.Nm
can exploit the parallelism in multiqueue devices and
multicore systems.
.Pp
.Pp
.Nm
requires explicit support in device drivers.
For a list of supported devices, see the end of this manual page.
.Sh OPERATION
.Nm
clients must first open the
.Pa open("/dev/netmap") ,
and then issue an
.Pa ioctl(...,NIOCREGIF,...)
to bind the file descriptor to a network device.
.Pp
When a device is put in
.Nm
mode, its data path is disconnected from the host stack.
The processes owning the file descriptor
can exchange packets with the device, or with the host stack,
through an mmapped memory region that contains pre-allocated
buffers and metadata.
.Pp
Non blocking I/O is done with special
.Pa ioctl()'s ,
whereas the file descriptor can be passed to
.Pa select()/poll()
to be notified about incoming packet or available transmit buffers.
.Ss Data structures
All data structures for all devices in
.Nm
mode are in a memory
region shared by the kernel and all processes
who open
.Pa /dev/netmap
(NOTE: visibility may be restricted in future implementations).
All references between the shared data structure
are relative (offsets or indexes). Some macros help converting
them into actual pointers.
.Pp
The data structures in shared memory are the following:
.Pp
.Bl -tag -width XXX
.It Dv struct netmap_if (one per interface)
indicates the number of rings supported by an interface, their
sizes, and the offsets of the
.Pa netmap_rings
associated to the interface.
The offset of a
.Pa struct netmap_if
in the shared memory region is indicated by the
.Pa nr_offset
field in the structure returned by the
.Pa NIOCREGIF
(see below).
.Bd -literal
struct netmap_if {
    char ni_name[IFNAMSIZ]; /* name of the interface. */
    const u_int ni_num_queues; /* number of hw ring pairs */
    const ssize_t   ring_ofs[]; /* offset of tx and rx rings */
};
.Ed
.It Dv struct netmap_ring (one per ring)
contains the index of the current read or write slot (cur),
the number of slots available for reception or transmission (avail),
and an array of
.Pa slots
describing the buffers.
There is one ring pair for each of the N hardware ring pairs
supported by the card (numbered 0..N-1), plus
one ring pair (numbered N) for packets from/to the host stack.
.Bd -literal
struct netmap_ring {
    const ssize_t buf_ofs;
    const uint32_t num_slots; /* number of slots in the ring. */
    uint32_t avail;           /* number of usable slots */
    uint32_t cur;             /* 'current' index for the user side */
    uint32_t reserved;        /* not refilled before current */

    const uint16_t nr_buf_size;
    uint16_t flags;
    struct netmap_slot slot[0]; /* array of slots. */
}
.Ed
.It Dv struct netmap_slot (one per packet)
contains the metadata for a packet: a buffer index (buf_idx),
a buffer length (len), and some flags.
.Bd -literal
struct netmap_slot {
    uint32_t buf_idx; /* buffer index */
    uint16_t len;   /* packet length */
    uint16_t flags; /* buf changed, etc. */
#define NS_BUF_CHANGED  0x0001  /* must resync, buffer changed */
#define NS_REPORT       0x0002  /* tell hw to report results
                                 * e.g. by generating an interrupt
                                 */
};
.Ed
.It Dv packet buffers
are fixed size (approximately 2k) buffers allocated by the kernel
that contain packet data. Buffers addresses are computed through
macros.
.El
.Pp
Some macros support the access to objects in the shared memory
region. In particular:
.Bd -literal
struct netmap_if *nifp;
struct netmap_ring *txring = NETMAP_TXRING(nifp, i);
struct netmap_ring *rxring = NETMAP_RXRING(nifp, i);
int i = txring->slot[txring->cur].buf_idx;
char *buf = NETMAP_BUF(txring, i);
.Ed
.Ss IOCTLS
.Pp
.Nm
supports some ioctl() to synchronize the state of the rings
between the kernel and the user processes, plus some
to query and configure the interface.
The former do not require any argument, whereas the latter
use a
.Pa struct netmap_req
defined as follows:
.Bd -literal
struct nmreq {
        char      nr_name[IFNAMSIZ];
        uint32_t  nr_version;     /* API version */
#define NETMAP_API      2         /* current version */
        uint32_t  nr_offset;      /* nifp offset in the shared region */
        uint32_t  nr_memsize;     /* size of the shared region */
        uint32_t  nr_tx_slots;    /* slots in tx rings */
        uint32_t  nr_rx_slots;    /* slots in rx rings */
        uint16_t  nr_tx_rings;    /* number of tx rings */
        uint16_t  nr_rx_rings;    /* number of tx rings */
        uint16_t  nr_ringid;      /* ring(s) we care about */
#define NETMAP_HW_RING  0x4000    /* low bits indicate one hw ring */
#define NETMAP_SW_RING  0x2000    /* we process the sw ring */
#define NETMAP_NO_TX_POLL 0x1000  /* no gratuitous txsync on poll */
#define NETMAP_RING_MASK 0xfff    /* the actual ring number */
};

.Ed
A device descriptor obtained through
.Pa /dev/netmap
also supports the ioctl supported by network devices.
.Pp
The netmap-specific
.Xr ioctl 2
command codes below are defined in
.In net/netmap.h
and are:
.Bl -tag -width XXXX
.It Dv NIOCGINFO
returns information about the interface named in nr_name.
On return, nr_memsize indicates the size of the shared netmap
memory region (this is device-independent),
nr_tx_slots and nr_rx_slots indicates how many buffers are in a
transmit and receive ring,
nr_tx_rings and nr_rx_rings indicates the number of transmit
and receive rings supported by the hardware.
.Pp
If the device does not support netmap, the ioctl returns EINVAL.
.It Dv NIOCREGIF
puts the interface named in nr_name into netmap mode, disconnecting
it from the host stack, and/or defines which rings are controlled
through this file descriptor.
On return, it gives the same info as NIOCGINFO, and nr_ringid
indicates the identity of the rings controlled through the file
descriptor.
.Pp
Possible values for nr_ringid are
.Bl -tag -width XXXXX
.It 0
default, all hardware rings
.It NETMAP_SW_RING
the ``host rings'' connecting to the host stack
.It NETMAP_HW_RING + i
the i-th hardware ring
.El
By default, a
.Nm poll
or
.Nm select
call pushes out any pending packets on the transmit ring, even if
no write events are specified.
The feature can be disabled by or-ing
.Nm NETMAP_NO_TX_SYNC
to nr_ringid.
But normally you should keep this feature unless you are using
separate file descriptors for the send and receive rings, because
otherwise packets are pushed out only if NETMAP_TXSYNC is called,
or the send queue is full.
.Pp
.Pa NIOCREGIF
can be used multiple times to change the association of a
file descriptor to a ring pair, always within the same device.
.It Dv NIOCUNREGIF
brings an interface back to normal mode.
.It Dv NIOCTXSYNC
tells the hardware of new packets to transmit, and updates the
number of slots available for transmission.
.It Dv NIOCRXSYNC
tells the hardware of consumed packets, and asks for newly available
packets.
.El
.Ss SYSTEM CALLS
.Nm
uses
.Nm select
and
.Nm poll
to wake up processes when significant events occur.
.Sh EXAMPLES
The following code implements a traffic generator
.Pp
.Bd -literal -compact
#include <net/netmap.h>
#include <net/netmap_user.h>
struct netmap_if *nifp;
struct netmap_ring *ring;
struct netmap_request nmr;

fd = open("/dev/netmap", O_RDWR);
bzero(&nmr, sizeof(nmr));
strcpy(nmr.nm_name, "ix0");
nmr.nm_version = NETMAP_API;
ioctl(fd, NIOCREG, &nmr);
p = mmap(0, nmr.memsize, fd);
nifp = NETMAP_IF(p, nmr.offset);
ring = NETMAP_TXRING(nifp, 0);
fds.fd = fd;
fds.events = POLLOUT;
for (;;) {
    poll(list, 1, -1);
    while (ring->avail-- > 0) {
        i = ring->cur;
        buf = NETMAP_BUF(ring, ring->slot[i].buf_index);
        ... prepare packet in buf ...
        ring->slot[i].len = ... packet length ...
        ring->cur = NETMAP_RING_NEXT(ring, i);
    }
}
.Ed
.Sh SUPPORTED INTERFACES
.Nm
supports the following interfaces:
.Xr em 4 ,
.Xr ixgbe 4 ,
.Xr re 4 ,
.Sh AUTHORS
The
.Nm
framework has been designed and implemented by
.An Luigi Rizzo
and
.An Matteo Landi
in 2011 at the Universita` di Pisa.