1 .\" Copyright (c) 2011 Matteo Landi, Luigi Rizzo, Universita` di Pisa
2 .\" All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions and the following disclaimer.
9 .\" 2. Redistributions in binary form must reproduce the above copyright
10 .\" notice, this list of conditions and the following disclaimer in the
11 .\" documentation and/or other materials provided with the distribution.
13 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 .\" This document is derived in part from the enet man page (enet.4)
26 .\" distributed with 4.3BSD Unix.
29 .\" $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 $
36 .Nd a framework for fast packet I/O
41 is a framework for fast and safe access to network devices
42 (reaching 14.88 Mpps at less than 1 GHz).
44 uses memory mapped buffers and metadata
45 (buffer indexes and lengths) to communicate with the kernel,
46 which is in charge of validating information through
51 can exploit the parallelism in multiqueue devices and
56 requires explicit support in device drivers.
57 For a list of supported devices, see the end of this manual page.
60 clients must first open the
61 .Pa open("/dev/netmap") ,
63 .Pa ioctl(...,NIOCREGIF,...)
64 to bind the file descriptor to a network device.
66 When a device is put in
68 mode, its data path is disconnected from the host stack.
69 The processes owning the file descriptor
70 can exchange packets with the device, or with the host stack,
71 through an mmapped memory region that contains pre-allocated
74 Non blocking I/O is done with special
76 whereas the file descriptor can be passed to
78 to be notified about incoming packet or available transmit buffers.
80 All data structures for all devices in
83 region shared by the kernel and all processes
86 (NOTE: visibility may be restricted in future implementations).
87 All references between the shared data structure
88 are relative (offsets or indexes). Some macros help converting
89 them into actual pointers.
91 The data structures in shared memory are the following:
94 .It Dv struct netmap_if (one per interface)
95 indicates the number of rings supported by an interface, their
96 sizes, and the offsets of the
98 associated to the interface.
101 in the shared memory region is indicated by the
103 field in the structure returned by the
108 char ni_name[IFNAMSIZ]; /* name of the interface. */
109 const u_int ni_num_queues; /* number of hw ring pairs */
110 const ssize_t ring_ofs[]; /* offset of tx and rx rings */
113 .It Dv struct netmap_ring (one per ring)
114 contains the index of the current read or write slot (cur),
115 the number of slots available for reception or transmission (avail),
118 describing the buffers.
119 There is one ring pair for each of the N hardware ring pairs
120 supported by the card (numbered 0..N-1), plus
121 one ring pair (numbered N) for packets from/to the host stack.
124 const ssize_t buf_ofs;
125 const uint32_t num_slots; /* number of slots in the ring. */
126 uint32_t avail; /* number of usable slots */
127 uint32_t cur; /* 'current' index for the user side */
129 const uint16_t nr_buf_size;
131 struct netmap_slot slot[0]; /* array of slots. */
134 .It Dv struct netmap_slot (one per packet)
135 contains the metadata for a packet: a buffer index (buf_idx),
136 a buffer length (len), and some flags.
139 uint32_t buf_idx; /* buffer index */
140 uint16_t len; /* packet length */
141 uint16_t flags; /* buf changed, etc. */
142 #define NS_BUF_CHANGED 0x0001 /* must resync, buffer changed */
143 #define NS_REPORT 0x0002 /* tell hw to report results
144 * e.g. by generating an interrupt
148 .It Dv packet buffers
149 are fixed size (approximately 2k) buffers allocated by the kernel
150 that contain packet data. Buffers addresses are computed through
154 Some macros support the access to objects in the shared memory
155 region. In particular:
157 struct netmap_if *nifp;
158 struct netmap_ring *txring = NETMAP_TXRING(nifp, i);
159 struct netmap_ring *rxring = NETMAP_RXRING(nifp, i);
160 int i = txring->slot[txring->cur].buf_idx;
161 char *buf = NETMAP_BUF(txring, i);
166 supports some ioctl() to synchronize the state of the rings
167 between the kernel and the user processes, plus some
168 to query and configure the interface.
169 The former do not require any argument, whereas the latter
171 .Pa struct netmap_req
175 char nr_name[IFNAMSIZ];
176 uint32_t nr_offset; /* nifp offset in the shared region */
177 uint32_t nr_memsize; /* size of the shared region */
178 uint32_t nr_numdescs; /* descriptors per queue */
179 uint16_t nr_numqueues;
180 uint16_t nr_ringid; /* ring(s) we care about */
181 #define NETMAP_HW_RING 0x4000 /* low bits indicate one hw ring */
182 #define NETMAP_SW_RING 0x2000 /* we process the sw ring */
183 #define NETMAP_NO_TX_POLL 0x1000 /* no gratuitous txsync on poll */
184 #define NETMAP_RING_MASK 0xfff /* the actual ring number */
188 A device descriptor obtained through
190 also supports the ioctl supported by network devices.
194 command codes below are defined in
199 returns information about the interface named in nr_name.
200 On return, nr_memsize indicates the size of the shared netmap
201 memory region (this is device-independent),
202 nr_numslots indicates how many buffers are in a ring,
203 nr_numrings indicates the number of rings supported by the hardware.
205 If the device does not support netmap, the ioctl returns EINVAL.
207 puts the interface named in nr_name into netmap mode, disconnecting
208 it from the host stack, and/or defines which rings are controlled
209 through this file descriptor.
210 On return, it gives the same info as NIOCGINFO, and nr_ringid
211 indicates the identity of the rings controlled through the file
214 Possible values for nr_ringid are
215 .Bl -tag -width XXXXX
217 default, all hardware rings
219 the ``host rings'' connecting to the host stack
220 .It NETMAP_HW_RING + i
221 the i-th hardware ring
227 call pushes out any pending packets on the transmit ring, even if
228 no write events are specified.
229 The feature can be disabled by or-ing
230 .Nm NETMAP_NO_TX_SYNC
232 But normally you should keep this feature unless you are using
233 separate file descriptors for the send and receive rings, because
234 otherwise packets are pushed out only if NETMAP_TXSYNC is called,
235 or the send queue is full.
238 can be used multiple times to change the association of a
239 file descriptor to a ring pair, always within the same device.
241 brings an interface back to normal mode.
243 tells the hardware of new packets to transmit, and updates the
244 number of slots available for transmission.
246 tells the hardware of consumed packets, and asks for newly available
255 to wake up processes when significant events occur.
257 The following code implements a traffic generator
259 .Bd -literal -compact
260 #include <net/netmap.h>
261 #include <net/netmap_user.h>
262 struct netmap_if *nifp;
263 struct netmap_ring *ring;
264 struct netmap_request nmr;
266 fd = open("/dev/netmap", O_RDWR);
267 bzero(&nmr, sizeof(nmr));
268 strcpy(nmr.nm_name, "ix0");
269 ioctl(fd, NIOCREG, &nmr);
270 p = mmap(0, nmr.memsize, fd);
271 nifp = NETMAP_IF(p, nmr.offset);
272 ring = NETMAP_TXRING(nifp, 0);
274 fds.events = POLLOUT;
277 while (ring->avail-- > 0) {
279 buf = NETMAP_BUF(ring, ring->slot[i].buf_index);
280 ... prepare packet in buf ...
281 ring->slot[i].len = ... packet length ...
282 ring->cur = NETMAP_RING_NEXT(ring, i);
286 .Sh SUPPORTED INTERFACES
288 supports the following interfaces:
295 framework has been designed and implemented by
299 in 2011 at the Universita` di Pisa.