root / net / tap.c @ 8af8ce4d
History | View | Annotate | Download (12 kB)
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 "net/tap.h" |
27 |
|
28 |
#include "config-host.h" |
29 |
|
30 |
#include <signal.h> |
31 |
#include <sys/ioctl.h> |
32 |
#include <sys/stat.h> |
33 |
#include <sys/wait.h> |
34 |
#include <sys/socket.h> |
35 |
#include <net/if.h> |
36 |
|
37 |
#include "net.h" |
38 |
#include "sysemu.h" |
39 |
#include "qemu-char.h" |
40 |
#include "qemu-common.h" |
41 |
#include "qemu-error.h" |
42 |
|
43 |
#include "net/tap-linux.h" |
44 |
|
45 |
/* Maximum GSO packet size (64k) plus plenty of room for
|
46 |
* the ethernet and virtio_net headers
|
47 |
*/
|
48 |
#define TAP_BUFSIZE (4096 + 65536) |
49 |
|
50 |
typedef struct TAPState { |
51 |
VLANClientState nc; |
52 |
int fd;
|
53 |
char down_script[1024]; |
54 |
char down_script_arg[128]; |
55 |
uint8_t buf[TAP_BUFSIZE]; |
56 |
Notifier exit_notifier; |
57 |
unsigned int read_poll : 1; |
58 |
unsigned int write_poll : 1; |
59 |
unsigned int has_vnet_hdr : 1; |
60 |
unsigned int using_vnet_hdr : 1; |
61 |
unsigned int has_ufo: 1; |
62 |
} TAPState; |
63 |
|
64 |
static int launch_script(const char *setup_script, const char *ifname, int fd); |
65 |
|
66 |
static int tap_can_send(void *opaque); |
67 |
static void tap_send(void *opaque); |
68 |
static void tap_writable(void *opaque); |
69 |
|
70 |
static void tap_update_fd_handler(TAPState *s) |
71 |
{ |
72 |
qemu_set_fd_handler2(s->fd, |
73 |
s->read_poll ? tap_can_send : NULL,
|
74 |
s->read_poll ? tap_send : NULL,
|
75 |
s->write_poll ? tap_writable : NULL,
|
76 |
s); |
77 |
} |
78 |
|
79 |
static void tap_read_poll(TAPState *s, int enable) |
80 |
{ |
81 |
s->read_poll = !!enable; |
82 |
tap_update_fd_handler(s); |
83 |
} |
84 |
|
85 |
static void tap_write_poll(TAPState *s, int enable) |
86 |
{ |
87 |
s->write_poll = !!enable; |
88 |
tap_update_fd_handler(s); |
89 |
} |
90 |
|
91 |
static void tap_writable(void *opaque) |
92 |
{ |
93 |
TAPState *s = opaque; |
94 |
|
95 |
tap_write_poll(s, 0);
|
96 |
|
97 |
qemu_flush_queued_packets(&s->nc); |
98 |
} |
99 |
|
100 |
static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt) |
101 |
{ |
102 |
ssize_t len; |
103 |
|
104 |
do {
|
105 |
len = writev(s->fd, iov, iovcnt); |
106 |
} while (len == -1 && errno == EINTR); |
107 |
|
108 |
if (len == -1 && errno == EAGAIN) { |
109 |
tap_write_poll(s, 1);
|
110 |
return 0; |
111 |
} |
112 |
|
113 |
return len;
|
114 |
} |
115 |
|
116 |
static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov, |
117 |
int iovcnt)
|
118 |
{ |
119 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
120 |
const struct iovec *iovp = iov; |
121 |
struct iovec iov_copy[iovcnt + 1]; |
122 |
struct virtio_net_hdr hdr = { 0, }; |
123 |
|
124 |
if (s->has_vnet_hdr && !s->using_vnet_hdr) {
|
125 |
iov_copy[0].iov_base = &hdr;
|
126 |
iov_copy[0].iov_len = sizeof(hdr); |
127 |
memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov)); |
128 |
iovp = iov_copy; |
129 |
iovcnt++; |
130 |
} |
131 |
|
132 |
return tap_write_packet(s, iovp, iovcnt);
|
133 |
} |
134 |
|
135 |
static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t size) |
136 |
{ |
137 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
138 |
struct iovec iov[2]; |
139 |
int iovcnt = 0; |
140 |
struct virtio_net_hdr hdr = { 0, }; |
141 |
|
142 |
if (s->has_vnet_hdr) {
|
143 |
iov[iovcnt].iov_base = &hdr; |
144 |
iov[iovcnt].iov_len = sizeof(hdr);
|
145 |
iovcnt++; |
146 |
} |
147 |
|
148 |
iov[iovcnt].iov_base = (char *)buf;
|
149 |
iov[iovcnt].iov_len = size; |
150 |
iovcnt++; |
151 |
|
152 |
return tap_write_packet(s, iov, iovcnt);
|
153 |
} |
154 |
|
155 |
static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size) |
156 |
{ |
157 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
158 |
struct iovec iov[1]; |
159 |
|
160 |
if (s->has_vnet_hdr && !s->using_vnet_hdr) {
|
161 |
return tap_receive_raw(nc, buf, size);
|
162 |
} |
163 |
|
164 |
iov[0].iov_base = (char *)buf; |
165 |
iov[0].iov_len = size;
|
166 |
|
167 |
return tap_write_packet(s, iov, 1); |
168 |
} |
169 |
|
170 |
static int tap_can_send(void *opaque) |
171 |
{ |
172 |
TAPState *s = opaque; |
173 |
|
174 |
return qemu_can_send_packet(&s->nc);
|
175 |
} |
176 |
|
177 |
#ifndef __sun__
|
178 |
ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen) |
179 |
{ |
180 |
return read(tapfd, buf, maxlen);
|
181 |
} |
182 |
#endif
|
183 |
|
184 |
static void tap_send_completed(VLANClientState *nc, ssize_t len) |
185 |
{ |
186 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
187 |
tap_read_poll(s, 1);
|
188 |
} |
189 |
|
190 |
static void tap_send(void *opaque) |
191 |
{ |
192 |
TAPState *s = opaque; |
193 |
int size;
|
194 |
|
195 |
do {
|
196 |
uint8_t *buf = s->buf; |
197 |
|
198 |
size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
|
199 |
if (size <= 0) { |
200 |
break;
|
201 |
} |
202 |
|
203 |
if (s->has_vnet_hdr && !s->using_vnet_hdr) {
|
204 |
buf += sizeof(struct virtio_net_hdr); |
205 |
size -= sizeof(struct virtio_net_hdr); |
206 |
} |
207 |
|
208 |
size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed); |
209 |
if (size == 0) { |
210 |
tap_read_poll(s, 0);
|
211 |
} |
212 |
} while (size > 0 && qemu_can_send_packet(&s->nc)); |
213 |
} |
214 |
|
215 |
int tap_has_ufo(VLANClientState *nc)
|
216 |
{ |
217 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
218 |
|
219 |
assert(nc->info->type == NET_CLIENT_TYPE_TAP); |
220 |
|
221 |
return s->has_ufo;
|
222 |
} |
223 |
|
224 |
int tap_has_vnet_hdr(VLANClientState *nc)
|
225 |
{ |
226 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
227 |
|
228 |
assert(nc->info->type == NET_CLIENT_TYPE_TAP); |
229 |
|
230 |
return s->has_vnet_hdr;
|
231 |
} |
232 |
|
233 |
void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr) |
234 |
{ |
235 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
236 |
|
237 |
using_vnet_hdr = using_vnet_hdr != 0;
|
238 |
|
239 |
assert(nc->info->type == NET_CLIENT_TYPE_TAP); |
240 |
assert(s->has_vnet_hdr == using_vnet_hdr); |
241 |
|
242 |
s->using_vnet_hdr = using_vnet_hdr; |
243 |
} |
244 |
|
245 |
void tap_set_offload(VLANClientState *nc, int csum, int tso4, |
246 |
int tso6, int ecn, int ufo) |
247 |
{ |
248 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
249 |
|
250 |
return tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
|
251 |
} |
252 |
|
253 |
static void tap_cleanup(VLANClientState *nc) |
254 |
{ |
255 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
256 |
|
257 |
qemu_purge_queued_packets(nc); |
258 |
|
259 |
if (s->down_script[0]) |
260 |
launch_script(s->down_script, s->down_script_arg, s->fd); |
261 |
|
262 |
tap_read_poll(s, 0);
|
263 |
tap_write_poll(s, 0);
|
264 |
close(s->fd); |
265 |
exit_notifier_remove(&s->exit_notifier); |
266 |
} |
267 |
|
268 |
/* Instead of exiting gracefully, we're exiting because someone called
|
269 |
* exit(), make sure to invoke down script at least
|
270 |
*/
|
271 |
static void tap_cleanup_at_exit(Notifier *notifier) |
272 |
{ |
273 |
TAPState *s = container_of(notifier, TAPState, exit_notifier); |
274 |
|
275 |
if (s->down_script[0]) { |
276 |
launch_script(s->down_script, s->down_script_arg, s->fd); |
277 |
} |
278 |
} |
279 |
|
280 |
static void tap_poll(VLANClientState *nc, bool enable) |
281 |
{ |
282 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
283 |
tap_read_poll(s, enable); |
284 |
tap_write_poll(s, enable); |
285 |
} |
286 |
|
287 |
/* fd support */
|
288 |
|
289 |
static NetClientInfo net_tap_info = {
|
290 |
.type = NET_CLIENT_TYPE_TAP, |
291 |
.size = sizeof(TAPState),
|
292 |
.receive = tap_receive, |
293 |
.receive_raw = tap_receive_raw, |
294 |
.receive_iov = tap_receive_iov, |
295 |
.poll = tap_poll, |
296 |
.cleanup = tap_cleanup, |
297 |
}; |
298 |
|
299 |
static TAPState *net_tap_fd_init(VLANState *vlan,
|
300 |
const char *model, |
301 |
const char *name, |
302 |
int fd,
|
303 |
int vnet_hdr)
|
304 |
{ |
305 |
VLANClientState *nc; |
306 |
TAPState *s; |
307 |
|
308 |
nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
|
309 |
|
310 |
s = DO_UPCAST(TAPState, nc, nc); |
311 |
|
312 |
s->fd = fd; |
313 |
s->has_vnet_hdr = vnet_hdr != 0;
|
314 |
s->using_vnet_hdr = 0;
|
315 |
s->has_ufo = tap_probe_has_ufo(s->fd); |
316 |
s->exit_notifier.notify = tap_cleanup_at_exit; |
317 |
exit_notifier_add(&s->exit_notifier); |
318 |
tap_set_offload(&s->nc, 0, 0, 0, 0, 0); |
319 |
tap_read_poll(s, 1);
|
320 |
return s;
|
321 |
} |
322 |
|
323 |
static int launch_script(const char *setup_script, const char *ifname, int fd) |
324 |
{ |
325 |
sigset_t oldmask, mask; |
326 |
int pid, status;
|
327 |
char *args[3]; |
328 |
char **parg;
|
329 |
|
330 |
sigemptyset(&mask); |
331 |
sigaddset(&mask, SIGCHLD); |
332 |
sigprocmask(SIG_BLOCK, &mask, &oldmask); |
333 |
|
334 |
/* try to launch network script */
|
335 |
pid = fork(); |
336 |
if (pid == 0) { |
337 |
int open_max = sysconf(_SC_OPEN_MAX), i;
|
338 |
|
339 |
for (i = 0; i < open_max; i++) { |
340 |
if (i != STDIN_FILENO &&
|
341 |
i != STDOUT_FILENO && |
342 |
i != STDERR_FILENO && |
343 |
i != fd) { |
344 |
close(i); |
345 |
} |
346 |
} |
347 |
parg = args; |
348 |
*parg++ = (char *)setup_script;
|
349 |
*parg++ = (char *)ifname;
|
350 |
*parg++ = NULL;
|
351 |
execv(setup_script, args); |
352 |
_exit(1);
|
353 |
} else if (pid > 0) { |
354 |
while (waitpid(pid, &status, 0) != pid) { |
355 |
/* loop */
|
356 |
} |
357 |
sigprocmask(SIG_SETMASK, &oldmask, NULL);
|
358 |
|
359 |
if (WIFEXITED(status) && WEXITSTATUS(status) == 0) { |
360 |
return 0; |
361 |
} |
362 |
} |
363 |
fprintf(stderr, "%s: could not launch network script\n", setup_script);
|
364 |
return -1; |
365 |
} |
366 |
|
367 |
static int net_tap_init(QemuOpts *opts, int *vnet_hdr) |
368 |
{ |
369 |
int fd, vnet_hdr_required;
|
370 |
char ifname[128] = {0,}; |
371 |
const char *setup_script; |
372 |
|
373 |
if (qemu_opt_get(opts, "ifname")) { |
374 |
pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname")); |
375 |
} |
376 |
|
377 |
*vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1); |
378 |
if (qemu_opt_get(opts, "vnet_hdr")) { |
379 |
vnet_hdr_required = *vnet_hdr; |
380 |
} else {
|
381 |
vnet_hdr_required = 0;
|
382 |
} |
383 |
|
384 |
TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
|
385 |
if (fd < 0) { |
386 |
return -1; |
387 |
} |
388 |
|
389 |
setup_script = qemu_opt_get(opts, "script");
|
390 |
if (setup_script &&
|
391 |
setup_script[0] != '\0' && |
392 |
strcmp(setup_script, "no") != 0 && |
393 |
launch_script(setup_script, ifname, fd)) { |
394 |
close(fd); |
395 |
return -1; |
396 |
} |
397 |
|
398 |
qemu_opt_set(opts, "ifname", ifname);
|
399 |
|
400 |
return fd;
|
401 |
} |
402 |
|
403 |
int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan) |
404 |
{ |
405 |
TAPState *s; |
406 |
int fd, vnet_hdr = 0; |
407 |
|
408 |
if (qemu_opt_get(opts, "fd")) { |
409 |
if (qemu_opt_get(opts, "ifname") || |
410 |
qemu_opt_get(opts, "script") ||
|
411 |
qemu_opt_get(opts, "downscript") ||
|
412 |
qemu_opt_get(opts, "vnet_hdr")) {
|
413 |
error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
|
414 |
return -1; |
415 |
} |
416 |
|
417 |
fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
|
418 |
if (fd == -1) { |
419 |
return -1; |
420 |
} |
421 |
|
422 |
fcntl(fd, F_SETFL, O_NONBLOCK); |
423 |
|
424 |
vnet_hdr = tap_probe_vnet_hdr(fd); |
425 |
} else {
|
426 |
if (!qemu_opt_get(opts, "script")) { |
427 |
qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
|
428 |
} |
429 |
|
430 |
if (!qemu_opt_get(opts, "downscript")) { |
431 |
qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
|
432 |
} |
433 |
|
434 |
fd = net_tap_init(opts, &vnet_hdr); |
435 |
if (fd == -1) { |
436 |
return -1; |
437 |
} |
438 |
} |
439 |
|
440 |
s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
|
441 |
if (!s) {
|
442 |
close(fd); |
443 |
return -1; |
444 |
} |
445 |
|
446 |
if (tap_set_sndbuf(s->fd, opts) < 0) { |
447 |
return -1; |
448 |
} |
449 |
|
450 |
if (qemu_opt_get(opts, "fd")) { |
451 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd); |
452 |
} else {
|
453 |
const char *ifname, *script, *downscript; |
454 |
|
455 |
ifname = qemu_opt_get(opts, "ifname");
|
456 |
script = qemu_opt_get(opts, "script");
|
457 |
downscript = qemu_opt_get(opts, "downscript");
|
458 |
|
459 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str),
|
460 |
"ifname=%s,script=%s,downscript=%s",
|
461 |
ifname, script, downscript); |
462 |
|
463 |
if (strcmp(downscript, "no") != 0) { |
464 |
snprintf(s->down_script, sizeof(s->down_script), "%s", downscript); |
465 |
snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname); |
466 |
} |
467 |
} |
468 |
|
469 |
return 0; |
470 |
} |