blob: 304ff45071dd4368289d1b0c9f7212e31979e925 [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
Markus Armbrusterda34e652016-03-14 09:01:28 +010034#include "qapi/error.h"
Paolo Bonzini1de7afc2012-12-17 18:20:00 +010035#include "qemu/error-report.h"
Veronia Bahaaf348b6d2016-03-20 19:16:19 +020036#include "qemu/cutils.h"
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010037
Michael Tokarev91ca60e2010-06-02 14:33:01 -030038#define PATH_NET_TUN "/dev/net/tun"
39
Jason Wang264986e2013-01-30 19:12:34 +080040int tap_open(char *ifname, int ifname_size, int *vnet_hdr,
Markus Armbruster468dd822015-05-15 13:58:58 +020041 int vnet_hdr_required, int mq_required, Error **errp)
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010042{
43 struct ifreq ifr;
44 int fd, ret;
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020045 int len = sizeof(struct virtio_net_hdr);
Peter Lievend26e4452013-02-25 10:17:08 +010046 unsigned int features;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010047
Michael Tokarev91ca60e2010-06-02 14:33:01 -030048 TFR(fd = open(PATH_NET_TUN, O_RDWR));
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010049 if (fd < 0) {
Markus Armbruster47896e22015-05-15 13:58:59 +020050 error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010051 return -1;
52 }
53 memset(&ifr, 0, sizeof(ifr));
54 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
55
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090056 if (ioctl(fd, TUNGETFEATURES, &features) == -1) {
Alistair Francis3dc6f862017-07-12 06:57:41 -070057 warn_report("TUNGETFEATURES failed: %s", strerror(errno));
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090058 features = 0;
59 }
60
61 if (features & IFF_ONE_QUEUE) {
Peter Lievend26e4452013-02-25 10:17:08 +010062 ifr.ifr_flags |= IFF_ONE_QUEUE;
63 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010064
Peter Lievend26e4452013-02-25 10:17:08 +010065 if (*vnet_hdr) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090066 if (features & IFF_VNET_HDR) {
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010067 *vnet_hdr = 1;
68 ifr.ifr_flags |= IFF_VNET_HDR;
Pierre Riteau6720b352009-11-25 18:49:34 +000069 } else {
70 *vnet_hdr = 0;
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010071 }
72
73 if (vnet_hdr_required && !*vnet_hdr) {
Markus Armbruster47896e22015-05-15 13:58:59 +020074 error_setg(errp, "vnet_hdr=1 requested, but no kernel "
75 "support for IFF_VNET_HDR available");
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010076 close(fd);
77 return -1;
78 }
Michael S. Tsirkin89e6d682012-11-12 09:13:04 +020079 /*
80 * Make sure vnet header size has the default value: for a persistent
81 * tap it might have been modified e.g. by another instance of qemu.
82 * Ignore errors since old kernels do not support this ioctl: in this
83 * case the header size implicitly has the correct value.
84 */
85 ioctl(fd, TUNSETVNETHDRSZ, &len);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010086 }
87
Jason Wang94fdc6d2013-01-30 19:12:31 +080088 if (mq_required) {
Kusanagi Kouichi1f149e72014-01-18 14:38:45 +090089 if (!(features & IFF_MULTI_QUEUE)) {
Markus Armbruster47896e22015-05-15 13:58:59 +020090 error_setg(errp, "multiqueue required, but no kernel "
91 "support for IFF_MULTI_QUEUE available");
Jason Wang94fdc6d2013-01-30 19:12:31 +080092 close(fd);
93 return -1;
94 } else {
95 ifr.ifr_flags |= IFF_MULTI_QUEUE;
96 }
97 }
98
Mark McLoughlinc28b1c12009-10-22 17:49:12 +010099 if (ifname[0] != '\0')
100 pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname);
101 else
102 pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d");
103 ret = ioctl(fd, TUNSETIFF, (void *) &ifr);
104 if (ret != 0) {
Luiz Capitulino93a73202011-10-14 15:05:10 -0300105 if (ifname[0] != '\0') {
Markus Armbruster47896e22015-05-15 13:58:59 +0200106 error_setg_errno(errp, errno, "could not configure %s (%s)",
107 PATH_NET_TUN, ifr.ifr_name);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300108 } else {
Markus Armbruster47896e22015-05-15 13:58:59 +0200109 error_setg_errno(errp, errno, "could not configure %s",
110 PATH_NET_TUN);
Luiz Capitulino93a73202011-10-14 15:05:10 -0300111 }
Mark McLoughlinc28b1c12009-10-22 17:49:12 +0100112 close(fd);
113 return -1;
114 }
115 pstrcpy(ifname, ifname_size, ifr.ifr_name);
Marc-André Lureau22e135f2022-04-25 11:56:42 +0400116 g_unix_set_fd_nonblocking(fd, true, NULL);
Mark McLoughlinc28b1c12009-10-22 17:49:12 +0100117 return fd;
118}
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100119
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200120/* sndbuf implements a kind of flow control for tap.
121 * Unfortunately when it's enabled, and packets are sent
122 * to other guests on the same host, the receiver
123 * can lock up the transmitter indefinitely.
124 *
125 * To avoid packet loss, sndbuf should be set to a value lower than the tx
126 * queue capacity of any destination network interface.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100127 * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200128 * a good value, given a 1500 byte MTU.
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100129 */
Michael S. Tsirkinf157ed22011-02-01 14:25:40 +0200130#define TAP_DEFAULT_SNDBUF 0
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100131
Markus Armbruster80b832c2015-05-15 13:58:55 +0200132void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp)
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100133{
134 int sndbuf;
135
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200136 sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF :
137 tap->sndbuf > INT_MAX ? INT_MAX :
138 tap->sndbuf;
139
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100140 if (!sndbuf) {
141 sndbuf = INT_MAX;
142 }
143
Laszlo Ersek08c573a2012-07-17 16:17:19 +0200144 if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
Markus Armbruster80b832c2015-05-15 13:58:55 +0200145 error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed");
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100146 }
Mark McLoughlin15ac9132009-10-22 17:49:13 +0100147}
Mark McLoughlindc690042009-10-22 17:49:14 +0100148
Daniel P. Berrangee7b347d2020-07-07 20:45:15 +0200149int tap_probe_vnet_hdr(int fd, Error **errp)
Mark McLoughlindc690042009-10-22 17:49:14 +0100150{
151 struct ifreq ifr;
Peter Foleye29919c2022-01-14 13:08:59 +0800152 memset(&ifr, 0, sizeof(ifr));
Mark McLoughlindc690042009-10-22 17:49:14 +0100153
154 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
Daniel P. Berrangee7b347d2020-07-07 20:45:15 +0200155 /* TUNGETIFF is available since kernel v2.6.27 */
156 error_setg_errno(errp, errno,
157 "Unable to query TUNGETIFF on FD %d", fd);
158 return -1;
Mark McLoughlindc690042009-10-22 17:49:14 +0100159 }
160
161 return ifr.ifr_flags & IFF_VNET_HDR;
162}
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100163
Mark McLoughlin9c282712009-10-22 17:49:16 +0100164int tap_probe_has_ufo(int fd)
165{
166 unsigned offload;
167
168 offload = TUN_F_CSUM | TUN_F_UFO;
169
170 if (ioctl(fd, TUNSETOFFLOAD, offload) < 0)
171 return 0;
172
173 return 1;
174}
175
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300176/* Verify that we can assign given length */
177int tap_probe_vnet_hdr_len(int fd, int len)
178{
179 int orig;
180 if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
181 return 0;
182 }
183 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
184 return 0;
185 }
186 /* Restore original length: we can't handle failure. */
187 if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
188 fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
189 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800190 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300191 return -errno;
192 }
193 return 1;
194}
195
196void tap_fd_set_vnet_hdr_len(int fd, int len)
197{
198 if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
199 fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
200 strerror(errno));
Jason Wang28a65892013-01-30 19:12:21 +0800201 abort();
Michael S. Tsirkin445d8922010-07-16 11:16:06 +0300202 }
203}
204
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200205int tap_fd_set_vnet_le(int fd, int is_le)
206{
207 int arg = is_le ? 1 : 0;
208
209 if (!ioctl(fd, TUNSETVNETLE, &arg)) {
210 return 0;
211 }
212
213 /* Check if our kernel supports TUNSETVNETLE */
214 if (errno == EINVAL) {
215 return -errno;
216 }
217
John Snow594fd212015-06-29 16:56:26 -0400218 error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200219 abort();
220}
221
222int tap_fd_set_vnet_be(int fd, int is_be)
223{
224 int arg = is_be ? 1 : 0;
225
226 if (!ioctl(fd, TUNSETVNETBE, &arg)) {
227 return 0;
228 }
229
230 /* Check if our kernel supports TUNSETVNETBE */
231 if (errno == EINVAL) {
232 return -errno;
233 }
234
John Snow594fd212015-06-29 16:56:26 -0400235 error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno));
Greg Kurzc80cd6b2015-06-17 15:23:44 +0200236 abort();
237}
238
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100239void tap_fd_set_offload(int fd, int csum, int tso4,
240 int tso6, int ecn, int ufo)
241{
242 unsigned int offload = 0;
243
Pierre Riteau2e503262009-11-25 18:49:35 +0000244 /* Check if our kernel supports TUNSETOFFLOAD */
245 if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
246 return;
247 }
248
Mark McLoughlin1faac1f2009-10-22 17:49:15 +0100249 if (csum) {
250 offload |= TUN_F_CSUM;
251 if (tso4)
252 offload |= TUN_F_TSO4;
253 if (tso6)
254 offload |= TUN_F_TSO6;
255 if ((tso4 || tso6) && ecn)
256 offload |= TUN_F_TSO_ECN;
257 if (ufo)
258 offload |= TUN_F_UFO;
259 }
260
261 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
262 offload &= ~TUN_F_UFO;
263 if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) {
264 fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n",
265 strerror(errno));
266 }
267 }
268}
Jason Wang94fdc6d2013-01-30 19:12:31 +0800269
270/* Enable a specific queue of tap. */
271int tap_fd_enable(int fd)
272{
273 struct ifreq ifr;
274 int ret;
275
276 memset(&ifr, 0, sizeof(ifr));
277
278 ifr.ifr_flags = IFF_ATTACH_QUEUE;
279 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
280
281 if (ret != 0) {
282 error_report("could not enable queue");
283 }
284
285 return ret;
286}
287
288/* Disable a specific queue of tap/ */
289int tap_fd_disable(int fd)
290{
291 struct ifreq ifr;
292 int ret;
293
294 memset(&ifr, 0, sizeof(ifr));
295
296 ifr.ifr_flags = IFF_DETACH_QUEUE;
297 ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr);
298
299 if (ret != 0) {
300 error_report("could not disable queue");
301 }
302
303 return ret;
304}
305
Jason Wange5dc0b42013-01-30 19:12:33 +0800306int tap_fd_get_ifname(int fd, char *ifname)
307{
308 struct ifreq ifr;
309
310 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
311 error_report("TUNGETIFF ioctl() failed: %s",
312 strerror(errno));
313 return -1;
314 }
315
316 pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name);
317 return 0;
318}
Andrew Melnychenko8f364e32021-05-14 14:48:30 +0300319
320int tap_fd_set_steering_ebpf(int fd, int prog_fd)
321{
322 if (ioctl(fd, TUNSETSTEERINGEBPF, (void *) &prog_fd) != 0) {
323 error_report("Issue while setting TUNSETSTEERINGEBPF:"
324 " %s with fd: %d, prog_fd: %d",
325 strerror(errno), fd, prog_fd);
326
327 return -1;
328 }
329
330 return 0;
331}