blob: 5bd9d21898e74313e64d9ee661a4aadcb108ec14 [file] [log] [blame]
Mark McLoughlinc28b1c12009-10-22 17:49:12 +01001/*
2 * QEMU System Emulator
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 * Copyright (c) 2009 Red Hat, Inc.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23 * THE SOFTWARE.
24 */
25
Paolo Bonzini1422e322012-10-24 08:43:34 +020026#include "tap_int.h"
27#include "tap-linux.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010028#include "net/tap.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010029
30#include <net/if.h>
31#include <sys/ioctl.h>
32
Paolo Bonzini9c17d612012-12-17 18:20:04 +010033#include "sysemu/sysemu.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010034#include "qemu-common.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010035#include "qemu/error-report.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010036
Michael Tokarev91ca60e2010-06-02 14:33:01 -030037#define PATH_NET_TUN "/dev/net/tun"
38
Jason Wang264986e2013-01-30 19:12:34 +080039int tap_open(char *ifname, int ifname_size, int *vnet_hdr,
Markus Armbruster468dd822015-05-15 13:58:58 +020040 int vnet_hdr_required, int mq_required, Error **errp)
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010041{
42 struct ifreq ifr;
43 int fd, ret;
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020044 int len = sizeof(struct virtio_net_hdr);
Peter Lievend26e4452013-02-25 10:17:08 +010045 unsigned int features;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010046
Michael Tokarev91ca60e2010-06-02 14:33:01 -030047 TFR(fd = open(PATH_NET_TUN, O_RDWR));
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010048 if (fd < 0) {
Markus Armbruster47896e22015-05-15 13:58:59 +020049 error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010050 return -1;
51 }
52 memset(&ifr, 0, sizeof(ifr));
53 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
54
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090055 if (ioctl(fd, TUNGETFEATURES, &features) == -1) {
56 error_report("warning: TUNGETFEATURES failed: %s", strerror(errno));
57 features = 0;
58 }
59
60 if (features & IFF_ONE_QUEUE) {
Peter Lievend26e4452013-02-25 10:17:08 +010061 ifr.ifr_flags |= IFF_ONE_QUEUE;
62 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010063
Peter Lievend26e4452013-02-25 10:17:08 +010064 if (*vnet_hdr) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090065 if (features & IFF_VNET_HDR) {
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010066 *vnet_hdr = 1;
67 ifr.ifr_flags |= IFF_VNET_HDR;
Pierre Riteau6720b352009-11-25 18:49:34 +000068 } else {
69 *vnet_hdr = 0;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010070 }
71
72 if (vnet_hdr_required && !*vnet_hdr) {
Markus Armbruster47896e22015-05-15 13:58:59 +020073 error_setg(errp, "vnet_hdr=1 requested, but no kernel "
74 "support for IFF_VNET_HDR available");
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010075 close(fd);
76 return -1;
77 }
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020078 /*
79 * Make sure vnet header size has the default value: for a persistent
80 * tap it might have been modified e.g. by another instance of qemu.
81 * Ignore errors since old kernels do not support this ioctl: in this
82 * case the header size implicitly has the correct value.
83 */
84 ioctl(fd, TUNSETVNETHDRSZ, &len);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010085 }
86
Jason Wang94fdc6d2013-01-30 19:12:31 +080087 if (mq_required) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090088 if (!(features & IFF_MULTI_QUEUE)) {
Markus Armbruster47896e22015-05-15 13:58:59 +020089 error_setg(errp, "multiqueue required, but no kernel "
90 "support for IFF_MULTI_QUEUE available");
Jason Wang94fdc6d2013-01-30 19:12:31 +080091 close(fd);
92 return -1;
93 } else {
94 ifr.ifr_flags |= IFF_MULTI_QUEUE;
95 }
96 }
97
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010098 if (ifname[0] != '\0')
99 pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
100 else
101 pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
102 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
103 if (ret != 0) {
Luiz Capitulino93a73202011-10-14 15:05:10 -0300104 if (ifname[0] != '\0') {
Markus Armbruster47896e22015-05-15 13:58:59 +0200105 error_setg_errno(errp, errno, "could not configure %s (%s)",
106 PATH_NET_TUN, ifr.ifr_name);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300107 } else {
Markus Armbruster47896e22015-05-15 13:58:59 +0200108 error_setg_errno(errp, errno, "could not configure %s",
109 PATH_NET_TUN);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300110 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +0100111 close(fd);
112 return -1;
113 }
114 pstrcpy(ifname, ifname_size, ifr.ifr_name);
115 fcntl(fd, F_SETFL, O_NONBLOCK);
116 return fd;
117}
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100118
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200119/* sndbuf implements a kind of flow control for tap.
120 * Unfortunately when it's enabled, and packets are sent
121 * to other guests on the same host, the receiver
122 * can lock up the transmitter indefinitely.
123 *
124 * To avoid packet loss, sndbuf should be set to a value lower than the tx
125 * queue capacity of any destination network interface.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100126 * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200127 * a good value, given a 1500 byte MTU.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100128 */
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200129#define TAP_DEFAULT_SNDBUF 0
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100130
Markus Armbruster80b832c2015-05-15 13:58:55 +0200131void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp)
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100132{
133 int sndbuf;
134
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200135 sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF :
136 tap->sndbuf > INT_MAX ? INT_MAX :
137 tap->sndbuf;
138
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100139 if (!sndbuf) {
140 sndbuf = INT_MAX;
141 }
142
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200143 if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
Markus Armbruster80b832c2015-05-15 13:58:55 +0200144 error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed");
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100145 }
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100146}
Mark McLoughlindc690042009-10-22 17:49:14 +0100147
148int tap_probe_vnet_hdr(int fd)
149{
150 struct ifreq ifr;
151
152 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
Markus Armbruster1ecda022010-02-18 17:25:24 +0100153 error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
Mark McLoughlindc690042009-10-22 17:49:14 +0100154 return 0;
155 }
156
157 return ifr.ifr_flags & IFF_VNET_HDR;
158}
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100159
Mark McLoughlin9c282712009-10-22 17:49:16 +0100160int tap_probe_has_ufo(int fd)
161{
162 unsigned offload;
163
164 offload = TUN_F_CSUM | TUN_F_UFO;
165
166 if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
167 return 0;
168
169 return 1;
170}
171
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300172/* Verify that we can assign given length */
173int tap_probe_vnet_hdr_len(int fd, int len)
174{
175 int orig;
176 if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
177 return 0;
178 }
179 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
180 return 0;
181 }
182 /* Restore original length: we can't handle failure. */
183 if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
184 fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
185 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800186 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300187 return -errno;
188 }
189 return 1;
190}
191
192void tap_fd_set_vnet_hdr_len(int fd, int len)
193{
194 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
195 fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
196 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800197 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300198 }
199}
200
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200201int tap_fd_set_vnet_le(int fd, int is_le)
202{
203 int arg = is_le ? 1 : 0;
204
205 if (!ioctl(fd, TUNSETVNETLE, &arg)) {
206 return 0;
207 }
208
209 /* Check if our kernel supports TUNSETVNETLE */
210 if (errno == EINVAL) {
211 return -errno;
212 }
213
John Snow594fd212015-06-29 16:56:26 -0400214 error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200215 abort();
216}
217
218int tap_fd_set_vnet_be(int fd, int is_be)
219{
220 int arg = is_be ? 1 : 0;
221
222 if (!ioctl(fd, TUNSETVNETBE, &arg)) {
223 return 0;
224 }
225
226 /* Check if our kernel supports TUNSETVNETBE */
227 if (errno == EINVAL) {
228 return -errno;
229 }
230
John Snow594fd212015-06-29 16:56:26 -0400231 error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200232 abort();
233}
234
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100235void tap_fd_set_offload(int fd, int csum, int tso4,
236 int tso6, int ecn, int ufo)
237{
238 unsigned int offload = 0;
239
Pierre Riteau2e503262009-11-25 18:49:35 +0000240 /* Check if our kernel supports TUNSETOFFLOAD */
241 if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
242 return;
243 }
244
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100245 if (csum) {
246 offload |= TUN_F_CSUM;
247 if (tso4)
248 offload |= TUN_F_TSO4;
249 if (tso6)
250 offload |= TUN_F_TSO6;
251 if ((tso4 || tso6) && ecn)
252 offload |= TUN_F_TSO_ECN;
253 if (ufo)
254 offload |= TUN_F_UFO;
255 }
256
257 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
258 offload &= ~TUN_F_UFO;
259 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
260 fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
261 strerror(errno));
262 }
263 }
264}
Jason Wang94fdc6d2013-01-30 19:12:31 +0800265
266/* Enable a specific queue of tap. */
267int tap_fd_enable(int fd)
268{
269 struct ifreq ifr;
270 int ret;
271
272 memset(&ifr, 0, sizeof(ifr));
273
274 ifr.ifr_flags = IFF_ATTACH_QUEUE;
275 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
276
277 if (ret != 0) {
278 error_report("could not enable queue");
279 }
280
281 return ret;
282}
283
284/* Disable a specific queue of tap/ */
285int tap_fd_disable(int fd)
286{
287 struct ifreq ifr;
288 int ret;
289
290 memset(&ifr, 0, sizeof(ifr));
291
292 ifr.ifr_flags = IFF_DETACH_QUEUE;
293 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
294
295 if (ret != 0) {
296 error_report("could not disable queue");
297 }
298
299 return ret;
300}
301
Jason Wange5dc0b42013-01-30 19:12:33 +0800302int tap_fd_get_ifname(int fd, char *ifname)
303{
304 struct ifreq ifr;
305
306 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
307 error_report("TUNGETIFF ioctl() failed: %s",
308 strerror(errno));
309 return -1;
310 }
311
312 pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name);
313 return 0;
314}