blob: 535b1ddb61007aad516a0337eecefdc8c1cc1fd9 [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
Peter Maydell2744d922016-01-29 17:50:00 +000026#include "qemu/osdep.h"
Paolo Bonzini1422e322012-10-24 08:43:34 +020027#include "tap_int.h"
28#include "tap-linux.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010029#include "net/tap.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010030
31#include <net/if.h>
32#include <sys/ioctl.h>
33
Paolo Bonzini9c17d612012-12-17 18:20:04 +010034#include "sysemu/sysemu.h"
Markus Armbrusterda34e652016-03-14 09:01:28 +010035#include "qapi/error.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010036#include "qemu/error-report.h"
Veronia Bahaaf348b6d2016-03-20 19:16:19 +020037#include "qemu/cutils.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010038
Michael Tokarev91ca60e2010-06-02 14:33:01 -030039#define PATH_NET_TUN "/dev/net/tun"
40
Jason Wang264986e2013-01-30 19:12:34 +080041int tap_open(char *ifname, int ifname_size, int *vnet_hdr,
Markus Armbruster468dd822015-05-15 13:58:58 +020042 int vnet_hdr_required, int mq_required, Error **errp)
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010043{
44 struct ifreq ifr;
45 int fd, ret;
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020046 int len = sizeof(struct virtio_net_hdr);
Peter Lievend26e4452013-02-25 10:17:08 +010047 unsigned int features;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010048
Michael Tokarev91ca60e2010-06-02 14:33:01 -030049 TFR(fd = open(PATH_NET_TUN, O_RDWR));
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010050 if (fd < 0) {
Markus Armbruster47896e22015-05-15 13:58:59 +020051 error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010052 return -1;
53 }
54 memset(&ifr, 0, sizeof(ifr));
55 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
56
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090057 if (ioctl(fd, TUNGETFEATURES, &features) == -1) {
Alistair Francis3dc6f862017-07-12 06:57:41 -070058 warn_report("TUNGETFEATURES failed: %s", strerror(errno));
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090059 features = 0;
60 }
61
62 if (features & IFF_ONE_QUEUE) {
Peter Lievend26e4452013-02-25 10:17:08 +010063 ifr.ifr_flags |= IFF_ONE_QUEUE;
64 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010065
Peter Lievend26e4452013-02-25 10:17:08 +010066 if (*vnet_hdr) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090067 if (features & IFF_VNET_HDR) {
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010068 *vnet_hdr = 1;
69 ifr.ifr_flags |= IFF_VNET_HDR;
Pierre Riteau6720b352009-11-25 18:49:34 +000070 } else {
71 *vnet_hdr = 0;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010072 }
73
74 if (vnet_hdr_required && !*vnet_hdr) {
Markus Armbruster47896e22015-05-15 13:58:59 +020075 error_setg(errp, "vnet_hdr=1 requested, but no kernel "
76 "support for IFF_VNET_HDR available");
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010077 close(fd);
78 return -1;
79 }
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020080 /*
81 * Make sure vnet header size has the default value: for a persistent
82 * tap it might have been modified e.g. by another instance of qemu.
83 * Ignore errors since old kernels do not support this ioctl: in this
84 * case the header size implicitly has the correct value.
85 */
86 ioctl(fd, TUNSETVNETHDRSZ, &len);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010087 }
88
Jason Wang94fdc6d2013-01-30 19:12:31 +080089 if (mq_required) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090090 if (!(features & IFF_MULTI_QUEUE)) {
Markus Armbruster47896e22015-05-15 13:58:59 +020091 error_setg(errp, "multiqueue required, but no kernel "
92 "support for IFF_MULTI_QUEUE available");
Jason Wang94fdc6d2013-01-30 19:12:31 +080093 close(fd);
94 return -1;
95 } else {
96 ifr.ifr_flags |= IFF_MULTI_QUEUE;
97 }
98 }
99
Mark McLoughlinc28b1c12009-10-22 17:49:12 +0100100 if (ifname[0] != '\0')
101 pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
102 else
103 pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
104 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
105 if (ret != 0) {
Luiz Capitulino93a73202011-10-14 15:05:10 -0300106 if (ifname[0] != '\0') {
Markus Armbruster47896e22015-05-15 13:58:59 +0200107 error_setg_errno(errp, errno, "could not configure %s (%s)",
108 PATH_NET_TUN, ifr.ifr_name);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300109 } else {
Markus Armbruster47896e22015-05-15 13:58:59 +0200110 error_setg_errno(errp, errno, "could not configure %s",
111 PATH_NET_TUN);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300112 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +0100113 close(fd);
114 return -1;
115 }
116 pstrcpy(ifname, ifname_size, ifr.ifr_name);
117 fcntl(fd, F_SETFL, O_NONBLOCK);
118 return fd;
119}
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100120
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200121/* sndbuf implements a kind of flow control for tap.
122 * Unfortunately when it's enabled, and packets are sent
123 * to other guests on the same host, the receiver
124 * can lock up the transmitter indefinitely.
125 *
126 * To avoid packet loss, sndbuf should be set to a value lower than the tx
127 * queue capacity of any destination network interface.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100128 * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200129 * a good value, given a 1500 byte MTU.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100130 */
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200131#define TAP_DEFAULT_SNDBUF 0
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100132
Markus Armbruster80b832c2015-05-15 13:58:55 +0200133void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp)
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100134{
135 int sndbuf;
136
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200137 sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF :
138 tap->sndbuf > INT_MAX ? INT_MAX :
139 tap->sndbuf;
140
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100141 if (!sndbuf) {
142 sndbuf = INT_MAX;
143 }
144
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200145 if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
Markus Armbruster80b832c2015-05-15 13:58:55 +0200146 error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed");
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100147 }
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100148}
Mark McLoughlindc690042009-10-22 17:49:14 +0100149
150int tap_probe_vnet_hdr(int fd)
151{
152 struct ifreq ifr;
153
154 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
Markus Armbruster1ecda022010-02-18 17:25:24 +0100155 error_report("TUNGETIFF ioctl() failed: %s", strerror(errno));
Mark McLoughlindc690042009-10-22 17:49:14 +0100156 return 0;
157 }
158
159 return ifr.ifr_flags & IFF_VNET_HDR;
160}
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100161
Mark McLoughlin9c282712009-10-22 17:49:16 +0100162int tap_probe_has_ufo(int fd)
163{
164 unsigned offload;
165
166 offload = TUN_F_CSUM | TUN_F_UFO;
167
168 if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
169 return 0;
170
171 return 1;
172}
173
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300174/* Verify that we can assign given length */
175int tap_probe_vnet_hdr_len(int fd, int len)
176{
177 int orig;
178 if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
179 return 0;
180 }
181 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
182 return 0;
183 }
184 /* Restore original length: we can't handle failure. */
185 if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
186 fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
187 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800188 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300189 return -errno;
190 }
191 return 1;
192}
193
194void tap_fd_set_vnet_hdr_len(int fd, int len)
195{
196 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
197 fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
198 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800199 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300200 }
201}
202
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200203int tap_fd_set_vnet_le(int fd, int is_le)
204{
205 int arg = is_le ? 1 : 0;
206
207 if (!ioctl(fd, TUNSETVNETLE, &arg)) {
208 return 0;
209 }
210
211 /* Check if our kernel supports TUNSETVNETLE */
212 if (errno == EINVAL) {
213 return -errno;
214 }
215
John Snow594fd212015-06-29 16:56:26 -0400216 error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200217 abort();
218}
219
220int tap_fd_set_vnet_be(int fd, int is_be)
221{
222 int arg = is_be ? 1 : 0;
223
224 if (!ioctl(fd, TUNSETVNETBE, &arg)) {
225 return 0;
226 }
227
228 /* Check if our kernel supports TUNSETVNETBE */
229 if (errno == EINVAL) {
230 return -errno;
231 }
232
John Snow594fd212015-06-29 16:56:26 -0400233 error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200234 abort();
235}
236
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100237void tap_fd_set_offload(int fd, int csum, int tso4,
238 int tso6, int ecn, int ufo)
239{
240 unsigned int offload = 0;
241
Pierre Riteau2e503262009-11-25 18:49:35 +0000242 /* Check if our kernel supports TUNSETOFFLOAD */
243 if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
244 return;
245 }
246
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100247 if (csum) {
248 offload |= TUN_F_CSUM;
249 if (tso4)
250 offload |= TUN_F_TSO4;
251 if (tso6)
252 offload |= TUN_F_TSO6;
253 if ((tso4 || tso6) && ecn)
254 offload |= TUN_F_TSO_ECN;
255 if (ufo)
256 offload |= TUN_F_UFO;
257 }
258
259 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
260 offload &= ~TUN_F_UFO;
261 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
262 fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
263 strerror(errno));
264 }
265 }
266}
Jason Wang94fdc6d2013-01-30 19:12:31 +0800267
268/* Enable a specific queue of tap. */
269int tap_fd_enable(int fd)
270{
271 struct ifreq ifr;
272 int ret;
273
274 memset(&ifr, 0, sizeof(ifr));
275
276 ifr.ifr_flags = IFF_ATTACH_QUEUE;
277 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
278
279 if (ret != 0) {
280 error_report("could not enable queue");
281 }
282
283 return ret;
284}
285
286/* Disable a specific queue of tap/ */
287int tap_fd_disable(int fd)
288{
289 struct ifreq ifr;
290 int ret;
291
292 memset(&ifr, 0, sizeof(ifr));
293
294 ifr.ifr_flags = IFF_DETACH_QUEUE;
295 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
296
297 if (ret != 0) {
298 error_report("could not disable queue");
299 }
300
301 return ret;
302}
303
Jason Wange5dc0b42013-01-30 19:12:33 +0800304int tap_fd_get_ifname(int fd, char *ifname)
305{
306 struct ifreq ifr;
307
308 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
309 error_report("TUNGETIFF ioctl() failed: %s",
310 strerror(errno));
311 return -1;
312 }
313
314 pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name);
315 return 0;
316}