]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap-linux.c
io: replace pipe() with g_unix_open_pipe(CLOEXEC)
[mirror_qemu.git] / net / tap-linux.c
CommitLineData
c28b1c10
MM
1/*
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
2744d920 26#include "qemu/osdep.h"
1422e32d
PB
27#include "tap_int.h"
28#include "tap-linux.h"
c28b1c10 29#include "net/tap.h"
c28b1c10
MM
30
31#include <net/if.h>
32#include <sys/ioctl.h>
33
da34e65c 34#include "qapi/error.h"
1de7afc9 35#include "qemu/error-report.h"
f348b6d1 36#include "qemu/cutils.h"
c28b1c10 37
91ca60e0
MT
38#define PATH_NET_TUN "/dev/net/tun"
39
264986e2 40int tap_open(char *ifname, int ifname_size, int *vnet_hdr,
468dd824 41 int vnet_hdr_required, int mq_required, Error **errp)
c28b1c10
MM
42{
43 struct ifreq ifr;
44 int fd, ret;
89e6d68e 45 int len = sizeof(struct virtio_net_hdr);
d26e445c 46 unsigned int features;
c28b1c10 47
91ca60e0 48 TFR(fd = open(PATH_NET_TUN, O_RDWR));
c28b1c10 49 if (fd < 0) {
47896e2f 50 error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN);
c28b1c10
MM
51 return -1;
52 }
53 memset(&ifr, 0, sizeof(ifr));
54 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
55
1f149e72 56 if (ioctl(fd, TUNGETFEATURES, &features) == -1) {
3dc6f869 57 warn_report("TUNGETFEATURES failed: %s", strerror(errno));
1f149e72
KK
58 features = 0;
59 }
60
61 if (features & IFF_ONE_QUEUE) {
d26e445c
PL
62 ifr.ifr_flags |= IFF_ONE_QUEUE;
63 }
c28b1c10 64
d26e445c 65 if (*vnet_hdr) {
1f149e72 66 if (features & IFF_VNET_HDR) {
c28b1c10
MM
67 *vnet_hdr = 1;
68 ifr.ifr_flags |= IFF_VNET_HDR;
6720b35b
PR
69 } else {
70 *vnet_hdr = 0;
c28b1c10
MM
71 }
72
73 if (vnet_hdr_required && !*vnet_hdr) {
47896e2f
MA
74 error_setg(errp, "vnet_hdr=1 requested, but no kernel "
75 "support for IFF_VNET_HDR available");
c28b1c10
MM
76 close(fd);
77 return -1;
78 }
89e6d68e
MT
79 /*
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);
c28b1c10
MM
86 }
87
94fdc6d0 88 if (mq_required) {
1f149e72 89 if (!(features & IFF_MULTI_QUEUE)) {
47896e2f
MA
90 error_setg(errp, "multiqueue required, but no kernel "
91 "support for IFF_MULTI_QUEUE available");
94fdc6d0
JW
92 close(fd);
93 return -1;
94 } else {
95 ifr.ifr_flags |= IFF_MULTI_QUEUE;
96 }
97 }
98
c28b1c10
MM
99 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) {
93a7320e 105 if (ifname[0] != '\0') {
47896e2f
MA
106 error_setg_errno(errp, errno, "could not configure %s (%s)",
107 PATH_NET_TUN, ifr.ifr_name);
93a7320e 108 } else {
47896e2f
MA
109 error_setg_errno(errp, errno, "could not configure %s",
110 PATH_NET_TUN);
93a7320e 111 }
c28b1c10
MM
112 close(fd);
113 return -1;
114 }
115 pstrcpy(ifname, ifname_size, ifr.ifr_name);
116 fcntl(fd, F_SETFL, O_NONBLOCK);
117 return fd;
118}
15ac913b 119
f157ed20
MT
120/* 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.
15ac913b 127 * Ethernet NICs generally have txqueuelen=1000, so 1Mb is
f157ed20 128 * a good value, given a 1500 byte MTU.
15ac913b 129 */
f157ed20 130#define TAP_DEFAULT_SNDBUF 0
15ac913b 131
80b832c3 132void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp)
15ac913b
MM
133{
134 int sndbuf;
135
08c573a8
LE
136 sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF :
137 tap->sndbuf > INT_MAX ? INT_MAX :
138 tap->sndbuf;
139
15ac913b
MM
140 if (!sndbuf) {
141 sndbuf = INT_MAX;
142 }
143
08c573a8 144 if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) {
80b832c3 145 error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed");
15ac913b 146 }
15ac913b 147}
dc69004c 148
e7b347d0 149int tap_probe_vnet_hdr(int fd, Error **errp)
dc69004c
MM
150{
151 struct ifreq ifr;
e29919c9 152 memset(&ifr, 0, sizeof(ifr));
dc69004c
MM
153
154 if (ioctl(fd, TUNGETIFF, &ifr) != 0) {
e7b347d0
DB
155 /* 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;
dc69004c
MM
159 }
160
161 return ifr.ifr_flags & IFF_VNET_HDR;
162}
1faac1f7 163
9c282718
MM
164int 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
445d892f
MT
176/* 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));
28a65891 190 abort();
445d892f
MT
191 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));
28a65891 201 abort();
445d892f
MT
202 }
203}
204
c80cd6bb
GK
205int 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
594fd211 218 error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno));
c80cd6bb
GK
219 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
594fd211 235 error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno));
c80cd6bb
GK
236 abort();
237}
238
1faac1f7
MM
239void tap_fd_set_offload(int fd, int csum, int tso4,
240 int tso6, int ecn, int ufo)
241{
242 unsigned int offload = 0;
243
2e50326c
PR
244 /* Check if our kernel supports TUNSETOFFLOAD */
245 if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) {
246 return;
247 }
248
1faac1f7
MM
249 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}
94fdc6d0
JW
269
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
e5dc0b40
JW
306int 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}
8f364e34
AM
319
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}