]>
Commit | Line | Data |
---|---|---|
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 | ||
26 | #include "qemu/osdep.h" | |
27 | #include "qemu-common.h" | |
28 | #include "tap_int.h" | |
29 | #include "tap-linux.h" | |
30 | #include "net/tap.h" | |
31 | ||
32 | #include <net/if.h> | |
33 | #include <sys/ioctl.h> | |
34 | ||
35 | #include "sysemu/sysemu.h" | |
36 | #include "qapi/error.h" | |
37 | #include "qemu/error-report.h" | |
38 | #include "qemu/cutils.h" | |
39 | ||
40 | #define PATH_NET_TUN "/dev/net/tun" | |
41 | ||
42 | int tap_open(char *ifname, int ifname_size, int *vnet_hdr, | |
43 | int vnet_hdr_required, int mq_required, Error **errp) | |
44 | { | |
45 | struct ifreq ifr; | |
46 | int fd, ret; | |
47 | int len = sizeof(struct virtio_net_hdr); | |
48 | unsigned int features; | |
49 | ||
50 | TFR(fd = open(PATH_NET_TUN, O_RDWR)); | |
51 | if (fd < 0) { | |
52 | error_setg_errno(errp, errno, "could not open %s", PATH_NET_TUN); | |
53 | return -1; | |
54 | } | |
55 | memset(&ifr, 0, sizeof(ifr)); | |
56 | ifr.ifr_flags = IFF_TAP | IFF_NO_PI; | |
57 | ||
58 | if (ioctl(fd, TUNGETFEATURES, &features) == -1) { | |
59 | warn_report("TUNGETFEATURES failed: %s", strerror(errno)); | |
60 | features = 0; | |
61 | } | |
62 | ||
63 | if (features & IFF_ONE_QUEUE) { | |
64 | ifr.ifr_flags |= IFF_ONE_QUEUE; | |
65 | } | |
66 | ||
67 | if (*vnet_hdr) { | |
68 | if (features & IFF_VNET_HDR) { | |
69 | *vnet_hdr = 1; | |
70 | ifr.ifr_flags |= IFF_VNET_HDR; | |
71 | } else { | |
72 | *vnet_hdr = 0; | |
73 | } | |
74 | ||
75 | if (vnet_hdr_required && !*vnet_hdr) { | |
76 | error_setg(errp, "vnet_hdr=1 requested, but no kernel " | |
77 | "support for IFF_VNET_HDR available"); | |
78 | close(fd); | |
79 | return -1; | |
80 | } | |
81 | /* | |
82 | * Make sure vnet header size has the default value: for a persistent | |
83 | * tap it might have been modified e.g. by another instance of qemu. | |
84 | * Ignore errors since old kernels do not support this ioctl: in this | |
85 | * case the header size implicitly has the correct value. | |
86 | */ | |
87 | ioctl(fd, TUNSETVNETHDRSZ, &len); | |
88 | } | |
89 | ||
90 | if (mq_required) { | |
91 | if (!(features & IFF_MULTI_QUEUE)) { | |
92 | error_setg(errp, "multiqueue required, but no kernel " | |
93 | "support for IFF_MULTI_QUEUE available"); | |
94 | close(fd); | |
95 | return -1; | |
96 | } else { | |
97 | ifr.ifr_flags |= IFF_MULTI_QUEUE; | |
98 | } | |
99 | } | |
100 | ||
101 | if (ifname[0] != '\0') | |
102 | pstrcpy(ifr.ifr_name, IFNAMSIZ, ifname); | |
103 | else | |
104 | pstrcpy(ifr.ifr_name, IFNAMSIZ, "tap%d"); | |
105 | ret = ioctl(fd, TUNSETIFF, (void *) &ifr); | |
106 | if (ret != 0) { | |
107 | if (ifname[0] != '\0') { | |
108 | error_setg_errno(errp, errno, "could not configure %s (%s)", | |
109 | PATH_NET_TUN, ifr.ifr_name); | |
110 | } else { | |
111 | error_setg_errno(errp, errno, "could not configure %s", | |
112 | PATH_NET_TUN); | |
113 | } | |
114 | close(fd); | |
115 | return -1; | |
116 | } | |
117 | pstrcpy(ifname, ifname_size, ifr.ifr_name); | |
118 | fcntl(fd, F_SETFL, O_NONBLOCK); | |
119 | return fd; | |
120 | } | |
121 | ||
122 | /* sndbuf implements a kind of flow control for tap. | |
123 | * Unfortunately when it's enabled, and packets are sent | |
124 | * to other guests on the same host, the receiver | |
125 | * can lock up the transmitter indefinitely. | |
126 | * | |
127 | * To avoid packet loss, sndbuf should be set to a value lower than the tx | |
128 | * queue capacity of any destination network interface. | |
129 | * Ethernet NICs generally have txqueuelen=1000, so 1Mb is | |
130 | * a good value, given a 1500 byte MTU. | |
131 | */ | |
132 | #define TAP_DEFAULT_SNDBUF 0 | |
133 | ||
134 | void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp) | |
135 | { | |
136 | int sndbuf; | |
137 | ||
138 | sndbuf = !tap->has_sndbuf ? TAP_DEFAULT_SNDBUF : | |
139 | tap->sndbuf > INT_MAX ? INT_MAX : | |
140 | tap->sndbuf; | |
141 | ||
142 | if (!sndbuf) { | |
143 | sndbuf = INT_MAX; | |
144 | } | |
145 | ||
146 | if (ioctl(fd, TUNSETSNDBUF, &sndbuf) == -1 && tap->has_sndbuf) { | |
147 | error_setg_errno(errp, errno, "TUNSETSNDBUF ioctl failed"); | |
148 | } | |
149 | } | |
150 | ||
151 | int tap_probe_vnet_hdr(int fd) | |
152 | { | |
153 | struct ifreq ifr; | |
154 | ||
155 | if (ioctl(fd, TUNGETIFF, &ifr) != 0) { | |
156 | error_report("TUNGETIFF ioctl() failed: %s", strerror(errno)); | |
157 | return 0; | |
158 | } | |
159 | ||
160 | return ifr.ifr_flags & IFF_VNET_HDR; | |
161 | } | |
162 | ||
163 | int tap_probe_has_ufo(int fd) | |
164 | { | |
165 | unsigned offload; | |
166 | ||
167 | offload = TUN_F_CSUM | TUN_F_UFO; | |
168 | ||
169 | if (ioctl(fd, TUNSETOFFLOAD, offload) < 0) | |
170 | return 0; | |
171 | ||
172 | return 1; | |
173 | } | |
174 | ||
175 | /* Verify that we can assign given length */ | |
176 | int tap_probe_vnet_hdr_len(int fd, int len) | |
177 | { | |
178 | int orig; | |
179 | if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) { | |
180 | return 0; | |
181 | } | |
182 | if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { | |
183 | return 0; | |
184 | } | |
185 | /* Restore original length: we can't handle failure. */ | |
186 | if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) { | |
187 | fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n", | |
188 | strerror(errno)); | |
189 | abort(); | |
190 | return -errno; | |
191 | } | |
192 | return 1; | |
193 | } | |
194 | ||
195 | void tap_fd_set_vnet_hdr_len(int fd, int len) | |
196 | { | |
197 | if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) { | |
198 | fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n", | |
199 | strerror(errno)); | |
200 | abort(); | |
201 | } | |
202 | } | |
203 | ||
204 | int tap_fd_set_vnet_le(int fd, int is_le) | |
205 | { | |
206 | int arg = is_le ? 1 : 0; | |
207 | ||
208 | if (!ioctl(fd, TUNSETVNETLE, &arg)) { | |
209 | return 0; | |
210 | } | |
211 | ||
212 | /* Check if our kernel supports TUNSETVNETLE */ | |
213 | if (errno == EINVAL) { | |
214 | return -errno; | |
215 | } | |
216 | ||
217 | error_report("TUNSETVNETLE ioctl() failed: %s.", strerror(errno)); | |
218 | abort(); | |
219 | } | |
220 | ||
221 | int tap_fd_set_vnet_be(int fd, int is_be) | |
222 | { | |
223 | int arg = is_be ? 1 : 0; | |
224 | ||
225 | if (!ioctl(fd, TUNSETVNETBE, &arg)) { | |
226 | return 0; | |
227 | } | |
228 | ||
229 | /* Check if our kernel supports TUNSETVNETBE */ | |
230 | if (errno == EINVAL) { | |
231 | return -errno; | |
232 | } | |
233 | ||
234 | error_report("TUNSETVNETBE ioctl() failed: %s.", strerror(errno)); | |
235 | abort(); | |
236 | } | |
237 | ||
238 | void tap_fd_set_offload(int fd, int csum, int tso4, | |
239 | int tso6, int ecn, int ufo) | |
240 | { | |
241 | unsigned int offload = 0; | |
242 | ||
243 | /* Check if our kernel supports TUNSETOFFLOAD */ | |
244 | if (ioctl(fd, TUNSETOFFLOAD, 0) != 0 && errno == EINVAL) { | |
245 | return; | |
246 | } | |
247 | ||
248 | if (csum) { | |
249 | offload |= TUN_F_CSUM; | |
250 | if (tso4) | |
251 | offload |= TUN_F_TSO4; | |
252 | if (tso6) | |
253 | offload |= TUN_F_TSO6; | |
254 | if ((tso4 || tso6) && ecn) | |
255 | offload |= TUN_F_TSO_ECN; | |
256 | if (ufo) | |
257 | offload |= TUN_F_UFO; | |
258 | } | |
259 | ||
260 | if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { | |
261 | offload &= ~TUN_F_UFO; | |
262 | if (ioctl(fd, TUNSETOFFLOAD, offload) != 0) { | |
263 | fprintf(stderr, "TUNSETOFFLOAD ioctl() failed: %s\n", | |
264 | strerror(errno)); | |
265 | } | |
266 | } | |
267 | } | |
268 | ||
269 | /* Enable a specific queue of tap. */ | |
270 | int tap_fd_enable(int fd) | |
271 | { | |
272 | struct ifreq ifr; | |
273 | int ret; | |
274 | ||
275 | memset(&ifr, 0, sizeof(ifr)); | |
276 | ||
277 | ifr.ifr_flags = IFF_ATTACH_QUEUE; | |
278 | ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); | |
279 | ||
280 | if (ret != 0) { | |
281 | error_report("could not enable queue"); | |
282 | } | |
283 | ||
284 | return ret; | |
285 | } | |
286 | ||
287 | /* Disable a specific queue of tap/ */ | |
288 | int tap_fd_disable(int fd) | |
289 | { | |
290 | struct ifreq ifr; | |
291 | int ret; | |
292 | ||
293 | memset(&ifr, 0, sizeof(ifr)); | |
294 | ||
295 | ifr.ifr_flags = IFF_DETACH_QUEUE; | |
296 | ret = ioctl(fd, TUNSETQUEUE, (void *) &ifr); | |
297 | ||
298 | if (ret != 0) { | |
299 | error_report("could not disable queue"); | |
300 | } | |
301 | ||
302 | return ret; | |
303 | } | |
304 | ||
305 | int tap_fd_get_ifname(int fd, char *ifname) | |
306 | { | |
307 | struct ifreq ifr; | |
308 | ||
309 | if (ioctl(fd, TUNGETIFF, &ifr) != 0) { | |
310 | error_report("TUNGETIFF ioctl() failed: %s", | |
311 | strerror(errno)); | |
312 | return -1; | |
313 | } | |
314 | ||
315 | pstrcpy(ifname, sizeof(ifr.ifr_name), ifr.ifr_name); | |
316 | return 0; | |
317 | } |