root / net / tap.c @ 0d09e41a
History | View | Annotate | Download (22.9 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 "tap_int.h" |
27 |
|
28 |
#include "config-host.h" |
29 |
|
30 |
#include <sys/ioctl.h> |
31 |
#include <sys/stat.h> |
32 |
#include <sys/wait.h> |
33 |
#include <sys/socket.h> |
34 |
#include <net/if.h> |
35 |
|
36 |
#include "net/net.h" |
37 |
#include "clients.h" |
38 |
#include "monitor/monitor.h" |
39 |
#include "sysemu/sysemu.h" |
40 |
#include "qemu-common.h" |
41 |
#include "qemu/error-report.h" |
42 |
|
43 |
#include "net/tap.h" |
44 |
|
45 |
#include "net/vhost_net.h" |
46 |
|
47 |
typedef struct TAPState { |
48 |
NetClientState nc; |
49 |
int fd;
|
50 |
char down_script[1024]; |
51 |
char down_script_arg[128]; |
52 |
uint8_t buf[NET_BUFSIZE]; |
53 |
bool read_poll;
|
54 |
bool write_poll;
|
55 |
bool using_vnet_hdr;
|
56 |
bool has_ufo;
|
57 |
bool enabled;
|
58 |
VHostNetState *vhost_net; |
59 |
unsigned host_vnet_hdr_len;
|
60 |
} TAPState; |
61 |
|
62 |
static int launch_script(const char *setup_script, const char *ifname, int fd); |
63 |
|
64 |
static int tap_can_send(void *opaque); |
65 |
static void tap_send(void *opaque); |
66 |
static void tap_writable(void *opaque); |
67 |
|
68 |
static void tap_update_fd_handler(TAPState *s) |
69 |
{ |
70 |
qemu_set_fd_handler2(s->fd, |
71 |
s->read_poll && s->enabled ? tap_can_send : NULL,
|
72 |
s->read_poll && s->enabled ? tap_send : NULL,
|
73 |
s->write_poll && s->enabled ? tap_writable : NULL,
|
74 |
s); |
75 |
} |
76 |
|
77 |
static void tap_read_poll(TAPState *s, bool enable) |
78 |
{ |
79 |
s->read_poll = enable; |
80 |
tap_update_fd_handler(s); |
81 |
} |
82 |
|
83 |
static void tap_write_poll(TAPState *s, bool enable) |
84 |
{ |
85 |
s->write_poll = enable; |
86 |
tap_update_fd_handler(s); |
87 |
} |
88 |
|
89 |
static void tap_writable(void *opaque) |
90 |
{ |
91 |
TAPState *s = opaque; |
92 |
|
93 |
tap_write_poll(s, false);
|
94 |
|
95 |
qemu_flush_queued_packets(&s->nc); |
96 |
} |
97 |
|
98 |
static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt) |
99 |
{ |
100 |
ssize_t len; |
101 |
|
102 |
do {
|
103 |
len = writev(s->fd, iov, iovcnt); |
104 |
} while (len == -1 && errno == EINTR); |
105 |
|
106 |
if (len == -1 && errno == EAGAIN) { |
107 |
tap_write_poll(s, true);
|
108 |
return 0; |
109 |
} |
110 |
|
111 |
return len;
|
112 |
} |
113 |
|
114 |
static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov, |
115 |
int iovcnt)
|
116 |
{ |
117 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
118 |
const struct iovec *iovp = iov; |
119 |
struct iovec iov_copy[iovcnt + 1]; |
120 |
struct virtio_net_hdr_mrg_rxbuf hdr = { };
|
121 |
|
122 |
if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
|
123 |
iov_copy[0].iov_base = &hdr;
|
124 |
iov_copy[0].iov_len = s->host_vnet_hdr_len;
|
125 |
memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov)); |
126 |
iovp = iov_copy; |
127 |
iovcnt++; |
128 |
} |
129 |
|
130 |
return tap_write_packet(s, iovp, iovcnt);
|
131 |
} |
132 |
|
133 |
static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size) |
134 |
{ |
135 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
136 |
struct iovec iov[2]; |
137 |
int iovcnt = 0; |
138 |
struct virtio_net_hdr_mrg_rxbuf hdr = { };
|
139 |
|
140 |
if (s->host_vnet_hdr_len) {
|
141 |
iov[iovcnt].iov_base = &hdr; |
142 |
iov[iovcnt].iov_len = s->host_vnet_hdr_len; |
143 |
iovcnt++; |
144 |
} |
145 |
|
146 |
iov[iovcnt].iov_base = (char *)buf;
|
147 |
iov[iovcnt].iov_len = size; |
148 |
iovcnt++; |
149 |
|
150 |
return tap_write_packet(s, iov, iovcnt);
|
151 |
} |
152 |
|
153 |
static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size) |
154 |
{ |
155 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
156 |
struct iovec iov[1]; |
157 |
|
158 |
if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
|
159 |
return tap_receive_raw(nc, buf, size);
|
160 |
} |
161 |
|
162 |
iov[0].iov_base = (char *)buf; |
163 |
iov[0].iov_len = size;
|
164 |
|
165 |
return tap_write_packet(s, iov, 1); |
166 |
} |
167 |
|
168 |
static int tap_can_send(void *opaque) |
169 |
{ |
170 |
TAPState *s = opaque; |
171 |
|
172 |
return qemu_can_send_packet(&s->nc);
|
173 |
} |
174 |
|
175 |
#ifndef __sun__
|
176 |
ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen) |
177 |
{ |
178 |
return read(tapfd, buf, maxlen);
|
179 |
} |
180 |
#endif
|
181 |
|
182 |
static void tap_send_completed(NetClientState *nc, ssize_t len) |
183 |
{ |
184 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
185 |
tap_read_poll(s, true);
|
186 |
} |
187 |
|
188 |
static void tap_send(void *opaque) |
189 |
{ |
190 |
TAPState *s = opaque; |
191 |
int size;
|
192 |
|
193 |
do {
|
194 |
uint8_t *buf = s->buf; |
195 |
|
196 |
size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
|
197 |
if (size <= 0) { |
198 |
break;
|
199 |
} |
200 |
|
201 |
if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
|
202 |
buf += s->host_vnet_hdr_len; |
203 |
size -= s->host_vnet_hdr_len; |
204 |
} |
205 |
|
206 |
size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed); |
207 |
if (size == 0) { |
208 |
tap_read_poll(s, false);
|
209 |
} |
210 |
} while (size > 0 && qemu_can_send_packet(&s->nc)); |
211 |
} |
212 |
|
213 |
bool tap_has_ufo(NetClientState *nc)
|
214 |
{ |
215 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
216 |
|
217 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
218 |
|
219 |
return s->has_ufo;
|
220 |
} |
221 |
|
222 |
int tap_has_vnet_hdr(NetClientState *nc)
|
223 |
{ |
224 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
225 |
|
226 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
227 |
|
228 |
return !!s->host_vnet_hdr_len;
|
229 |
} |
230 |
|
231 |
int tap_has_vnet_hdr_len(NetClientState *nc, int len) |
232 |
{ |
233 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
234 |
|
235 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
236 |
|
237 |
return tap_probe_vnet_hdr_len(s->fd, len);
|
238 |
} |
239 |
|
240 |
void tap_set_vnet_hdr_len(NetClientState *nc, int len) |
241 |
{ |
242 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
243 |
|
244 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
245 |
assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) || |
246 |
len == sizeof(struct virtio_net_hdr)); |
247 |
|
248 |
tap_fd_set_vnet_hdr_len(s->fd, len); |
249 |
s->host_vnet_hdr_len = len; |
250 |
} |
251 |
|
252 |
void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr) |
253 |
{ |
254 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
255 |
|
256 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
257 |
assert(!!s->host_vnet_hdr_len == using_vnet_hdr); |
258 |
|
259 |
s->using_vnet_hdr = using_vnet_hdr; |
260 |
} |
261 |
|
262 |
void tap_set_offload(NetClientState *nc, int csum, int tso4, |
263 |
int tso6, int ecn, int ufo) |
264 |
{ |
265 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
266 |
if (s->fd < 0) { |
267 |
return;
|
268 |
} |
269 |
|
270 |
tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo); |
271 |
} |
272 |
|
273 |
static void tap_cleanup(NetClientState *nc) |
274 |
{ |
275 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
276 |
|
277 |
if (s->vhost_net) {
|
278 |
vhost_net_cleanup(s->vhost_net); |
279 |
s->vhost_net = NULL;
|
280 |
} |
281 |
|
282 |
qemu_purge_queued_packets(nc); |
283 |
|
284 |
if (s->down_script[0]) |
285 |
launch_script(s->down_script, s->down_script_arg, s->fd); |
286 |
|
287 |
tap_read_poll(s, false);
|
288 |
tap_write_poll(s, false);
|
289 |
close(s->fd); |
290 |
s->fd = -1;
|
291 |
} |
292 |
|
293 |
static void tap_poll(NetClientState *nc, bool enable) |
294 |
{ |
295 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
296 |
tap_read_poll(s, enable); |
297 |
tap_write_poll(s, enable); |
298 |
} |
299 |
|
300 |
int tap_get_fd(NetClientState *nc)
|
301 |
{ |
302 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
303 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
304 |
return s->fd;
|
305 |
} |
306 |
|
307 |
/* fd support */
|
308 |
|
309 |
static NetClientInfo net_tap_info = {
|
310 |
.type = NET_CLIENT_OPTIONS_KIND_TAP, |
311 |
.size = sizeof(TAPState),
|
312 |
.receive = tap_receive, |
313 |
.receive_raw = tap_receive_raw, |
314 |
.receive_iov = tap_receive_iov, |
315 |
.poll = tap_poll, |
316 |
.cleanup = tap_cleanup, |
317 |
}; |
318 |
|
319 |
static TAPState *net_tap_fd_init(NetClientState *peer,
|
320 |
const char *model, |
321 |
const char *name, |
322 |
int fd,
|
323 |
int vnet_hdr)
|
324 |
{ |
325 |
NetClientState *nc; |
326 |
TAPState *s; |
327 |
|
328 |
nc = qemu_new_net_client(&net_tap_info, peer, model, name); |
329 |
|
330 |
s = DO_UPCAST(TAPState, nc, nc); |
331 |
|
332 |
s->fd = fd; |
333 |
s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0; |
334 |
s->using_vnet_hdr = false;
|
335 |
s->has_ufo = tap_probe_has_ufo(s->fd); |
336 |
s->enabled = true;
|
337 |
tap_set_offload(&s->nc, 0, 0, 0, 0, 0); |
338 |
/*
|
339 |
* Make sure host header length is set correctly in tap:
|
340 |
* it might have been modified by another instance of qemu.
|
341 |
*/
|
342 |
if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
|
343 |
tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len); |
344 |
} |
345 |
tap_read_poll(s, true);
|
346 |
s->vhost_net = NULL;
|
347 |
return s;
|
348 |
} |
349 |
|
350 |
static int launch_script(const char *setup_script, const char *ifname, int fd) |
351 |
{ |
352 |
int pid, status;
|
353 |
char *args[3]; |
354 |
char **parg;
|
355 |
|
356 |
/* try to launch network script */
|
357 |
pid = fork(); |
358 |
if (pid == 0) { |
359 |
int open_max = sysconf(_SC_OPEN_MAX), i;
|
360 |
|
361 |
for (i = 0; i < open_max; i++) { |
362 |
if (i != STDIN_FILENO &&
|
363 |
i != STDOUT_FILENO && |
364 |
i != STDERR_FILENO && |
365 |
i != fd) { |
366 |
close(i); |
367 |
} |
368 |
} |
369 |
parg = args; |
370 |
*parg++ = (char *)setup_script;
|
371 |
*parg++ = (char *)ifname;
|
372 |
*parg = NULL;
|
373 |
execv(setup_script, args); |
374 |
_exit(1);
|
375 |
} else if (pid > 0) { |
376 |
while (waitpid(pid, &status, 0) != pid) { |
377 |
/* loop */
|
378 |
} |
379 |
|
380 |
if (WIFEXITED(status) && WEXITSTATUS(status) == 0) { |
381 |
return 0; |
382 |
} |
383 |
} |
384 |
fprintf(stderr, "%s: could not launch network script\n", setup_script);
|
385 |
return -1; |
386 |
} |
387 |
|
388 |
static int recv_fd(int c) |
389 |
{ |
390 |
int fd;
|
391 |
uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
|
392 |
struct msghdr msg = {
|
393 |
.msg_control = msgbuf, |
394 |
.msg_controllen = sizeof(msgbuf),
|
395 |
}; |
396 |
struct cmsghdr *cmsg;
|
397 |
struct iovec iov;
|
398 |
uint8_t req[1];
|
399 |
ssize_t len; |
400 |
|
401 |
cmsg = CMSG_FIRSTHDR(&msg); |
402 |
cmsg->cmsg_level = SOL_SOCKET; |
403 |
cmsg->cmsg_type = SCM_RIGHTS; |
404 |
cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
|
405 |
msg.msg_controllen = cmsg->cmsg_len; |
406 |
|
407 |
iov.iov_base = req; |
408 |
iov.iov_len = sizeof(req);
|
409 |
|
410 |
msg.msg_iov = &iov; |
411 |
msg.msg_iovlen = 1;
|
412 |
|
413 |
len = recvmsg(c, &msg, 0);
|
414 |
if (len > 0) { |
415 |
memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
|
416 |
return fd;
|
417 |
} |
418 |
|
419 |
return len;
|
420 |
} |
421 |
|
422 |
static int net_bridge_run_helper(const char *helper, const char *bridge) |
423 |
{ |
424 |
sigset_t oldmask, mask; |
425 |
int pid, status;
|
426 |
char *args[5]; |
427 |
char **parg;
|
428 |
int sv[2]; |
429 |
|
430 |
sigemptyset(&mask); |
431 |
sigaddset(&mask, SIGCHLD); |
432 |
sigprocmask(SIG_BLOCK, &mask, &oldmask); |
433 |
|
434 |
if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) { |
435 |
return -1; |
436 |
} |
437 |
|
438 |
/* try to launch bridge helper */
|
439 |
pid = fork(); |
440 |
if (pid == 0) { |
441 |
int open_max = sysconf(_SC_OPEN_MAX), i;
|
442 |
char fd_buf[6+10]; |
443 |
char br_buf[6+IFNAMSIZ] = {0}; |
444 |
char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15]; |
445 |
|
446 |
for (i = 0; i < open_max; i++) { |
447 |
if (i != STDIN_FILENO &&
|
448 |
i != STDOUT_FILENO && |
449 |
i != STDERR_FILENO && |
450 |
i != sv[1]) {
|
451 |
close(i); |
452 |
} |
453 |
} |
454 |
|
455 |
snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]); |
456 |
|
457 |
if (strrchr(helper, ' ') || strrchr(helper, '\t')) { |
458 |
/* assume helper is a command */
|
459 |
|
460 |
if (strstr(helper, "--br=") == NULL) { |
461 |
snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge); |
462 |
} |
463 |
|
464 |
snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s", |
465 |
helper, "--use-vnet", fd_buf, br_buf);
|
466 |
|
467 |
parg = args; |
468 |
*parg++ = (char *)"sh"; |
469 |
*parg++ = (char *)"-c"; |
470 |
*parg++ = helper_cmd; |
471 |
*parg++ = NULL;
|
472 |
|
473 |
execv("/bin/sh", args);
|
474 |
} else {
|
475 |
/* assume helper is just the executable path name */
|
476 |
|
477 |
snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge); |
478 |
|
479 |
parg = args; |
480 |
*parg++ = (char *)helper;
|
481 |
*parg++ = (char *)"--use-vnet"; |
482 |
*parg++ = fd_buf; |
483 |
*parg++ = br_buf; |
484 |
*parg++ = NULL;
|
485 |
|
486 |
execv(helper, args); |
487 |
} |
488 |
_exit(1);
|
489 |
|
490 |
} else if (pid > 0) { |
491 |
int fd;
|
492 |
|
493 |
close(sv[1]);
|
494 |
|
495 |
do {
|
496 |
fd = recv_fd(sv[0]);
|
497 |
} while (fd == -1 && errno == EINTR); |
498 |
|
499 |
close(sv[0]);
|
500 |
|
501 |
while (waitpid(pid, &status, 0) != pid) { |
502 |
/* loop */
|
503 |
} |
504 |
sigprocmask(SIG_SETMASK, &oldmask, NULL);
|
505 |
if (fd < 0) { |
506 |
fprintf(stderr, "failed to recv file descriptor\n");
|
507 |
return -1; |
508 |
} |
509 |
|
510 |
if (WIFEXITED(status) && WEXITSTATUS(status) == 0) { |
511 |
return fd;
|
512 |
} |
513 |
} |
514 |
fprintf(stderr, "failed to launch bridge helper\n");
|
515 |
return -1; |
516 |
} |
517 |
|
518 |
int net_init_bridge(const NetClientOptions *opts, const char *name, |
519 |
NetClientState *peer) |
520 |
{ |
521 |
const NetdevBridgeOptions *bridge;
|
522 |
const char *helper, *br; |
523 |
|
524 |
TAPState *s; |
525 |
int fd, vnet_hdr;
|
526 |
|
527 |
assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE); |
528 |
bridge = opts->bridge; |
529 |
|
530 |
helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER; |
531 |
br = bridge->has_br ? bridge->br : DEFAULT_BRIDGE_INTERFACE; |
532 |
|
533 |
fd = net_bridge_run_helper(helper, br); |
534 |
if (fd == -1) { |
535 |
return -1; |
536 |
} |
537 |
|
538 |
fcntl(fd, F_SETFL, O_NONBLOCK); |
539 |
|
540 |
vnet_hdr = tap_probe_vnet_hdr(fd); |
541 |
|
542 |
s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
|
543 |
if (!s) {
|
544 |
close(fd); |
545 |
return -1; |
546 |
} |
547 |
|
548 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper, |
549 |
br); |
550 |
|
551 |
return 0; |
552 |
} |
553 |
|
554 |
static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr, |
555 |
const char *setup_script, char *ifname, |
556 |
size_t ifname_sz, int mq_required)
|
557 |
{ |
558 |
int fd, vnet_hdr_required;
|
559 |
|
560 |
if (tap->has_vnet_hdr) {
|
561 |
*vnet_hdr = tap->vnet_hdr; |
562 |
vnet_hdr_required = *vnet_hdr; |
563 |
} else {
|
564 |
*vnet_hdr = 1;
|
565 |
vnet_hdr_required = 0;
|
566 |
} |
567 |
|
568 |
TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required, |
569 |
mq_required)); |
570 |
if (fd < 0) { |
571 |
return -1; |
572 |
} |
573 |
|
574 |
if (setup_script &&
|
575 |
setup_script[0] != '\0' && |
576 |
strcmp(setup_script, "no") != 0 && |
577 |
launch_script(setup_script, ifname, fd)) { |
578 |
close(fd); |
579 |
return -1; |
580 |
} |
581 |
|
582 |
return fd;
|
583 |
} |
584 |
|
585 |
#define MAX_TAP_QUEUES 1024 |
586 |
|
587 |
static int net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer, |
588 |
const char *model, const char *name, |
589 |
const char *ifname, const char *script, |
590 |
const char *downscript, const char *vhostfdname, |
591 |
int vnet_hdr, int fd) |
592 |
{ |
593 |
TAPState *s; |
594 |
|
595 |
s = net_tap_fd_init(peer, model, name, fd, vnet_hdr); |
596 |
if (!s) {
|
597 |
close(fd); |
598 |
return -1; |
599 |
} |
600 |
|
601 |
if (tap_set_sndbuf(s->fd, tap) < 0) { |
602 |
return -1; |
603 |
} |
604 |
|
605 |
if (tap->has_fd || tap->has_fds) {
|
606 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd); |
607 |
} else if (tap->has_helper) { |
608 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s", |
609 |
tap->helper); |
610 |
} else {
|
611 |
snprintf(s->nc.info_str, sizeof(s->nc.info_str),
|
612 |
"ifname=%s,script=%s,downscript=%s", ifname, script,
|
613 |
downscript); |
614 |
|
615 |
if (strcmp(downscript, "no") != 0) { |
616 |
snprintf(s->down_script, sizeof(s->down_script), "%s", downscript); |
617 |
snprintf(s->down_script_arg, sizeof(s->down_script_arg),
|
618 |
"%s", ifname);
|
619 |
} |
620 |
} |
621 |
|
622 |
if (tap->has_vhost ? tap->vhost :
|
623 |
vhostfdname || (tap->has_vhostforce && tap->vhostforce)) { |
624 |
int vhostfd;
|
625 |
|
626 |
if (tap->has_vhostfd) {
|
627 |
vhostfd = monitor_handle_fd_param(cur_mon, vhostfdname); |
628 |
if (vhostfd == -1) { |
629 |
return -1; |
630 |
} |
631 |
} else {
|
632 |
vhostfd = -1;
|
633 |
} |
634 |
|
635 |
s->vhost_net = vhost_net_init(&s->nc, vhostfd, |
636 |
tap->has_vhostforce && tap->vhostforce); |
637 |
if (!s->vhost_net) {
|
638 |
error_report("vhost-net requested but could not be initialized");
|
639 |
return -1; |
640 |
} |
641 |
} else if (tap->has_vhostfd || tap->has_vhostfds) { |
642 |
error_report("vhostfd= is not valid without vhost");
|
643 |
return -1; |
644 |
} |
645 |
|
646 |
return 0; |
647 |
} |
648 |
|
649 |
static int get_fds(char *str, char *fds[], int max) |
650 |
{ |
651 |
char *ptr = str, *this;
|
652 |
size_t len = strlen(str); |
653 |
int i = 0; |
654 |
|
655 |
while (i < max && ptr < str + len) {
|
656 |
this = strchr(ptr, ':');
|
657 |
|
658 |
if (this == NULL) { |
659 |
fds[i] = g_strdup(ptr); |
660 |
} else {
|
661 |
fds[i] = g_strndup(ptr, this - ptr); |
662 |
} |
663 |
|
664 |
i++; |
665 |
if (this == NULL) { |
666 |
break;
|
667 |
} else {
|
668 |
ptr = this + 1;
|
669 |
} |
670 |
} |
671 |
|
672 |
return i;
|
673 |
} |
674 |
|
675 |
int net_init_tap(const NetClientOptions *opts, const char *name, |
676 |
NetClientState *peer) |
677 |
{ |
678 |
const NetdevTapOptions *tap;
|
679 |
int fd, vnet_hdr = 0, i = 0, queues; |
680 |
/* for the no-fd, no-helper case */
|
681 |
const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */ |
682 |
const char *downscript = NULL; |
683 |
const char *vhostfdname; |
684 |
char ifname[128]; |
685 |
|
686 |
assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP); |
687 |
tap = opts->tap; |
688 |
queues = tap->has_queues ? tap->queues : 1;
|
689 |
vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
|
690 |
|
691 |
/* QEMU vlans does not support multiqueue tap, in this case peer is set.
|
692 |
* For -netdev, peer is always NULL. */
|
693 |
if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
|
694 |
error_report("Multiqueue tap cannot be used with QEMU vlans");
|
695 |
return -1; |
696 |
} |
697 |
|
698 |
if (tap->has_fd) {
|
699 |
if (tap->has_ifname || tap->has_script || tap->has_downscript ||
|
700 |
tap->has_vnet_hdr || tap->has_helper || tap->has_queues || |
701 |
tap->has_fds) { |
702 |
error_report("ifname=, script=, downscript=, vnet_hdr=, "
|
703 |
"helper=, queues=, and fds= are invalid with fd=");
|
704 |
return -1; |
705 |
} |
706 |
|
707 |
fd = monitor_handle_fd_param(cur_mon, tap->fd); |
708 |
if (fd == -1) { |
709 |
return -1; |
710 |
} |
711 |
|
712 |
fcntl(fd, F_SETFL, O_NONBLOCK); |
713 |
|
714 |
vnet_hdr = tap_probe_vnet_hdr(fd); |
715 |
|
716 |
if (net_init_tap_one(tap, peer, "tap", name, NULL, |
717 |
script, downscript, |
718 |
vhostfdname, vnet_hdr, fd)) { |
719 |
return -1; |
720 |
} |
721 |
} else if (tap->has_fds) { |
722 |
char *fds[MAX_TAP_QUEUES];
|
723 |
char *vhost_fds[MAX_TAP_QUEUES];
|
724 |
int nfds, nvhosts;
|
725 |
|
726 |
if (tap->has_ifname || tap->has_script || tap->has_downscript ||
|
727 |
tap->has_vnet_hdr || tap->has_helper || tap->has_queues || |
728 |
tap->has_fd) { |
729 |
error_report("ifname=, script=, downscript=, vnet_hdr=, "
|
730 |
"helper=, queues=, and fd= are invalid with fds=");
|
731 |
return -1; |
732 |
} |
733 |
|
734 |
nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES); |
735 |
if (tap->has_vhostfds) {
|
736 |
nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES); |
737 |
if (nfds != nvhosts) {
|
738 |
error_report("The number of fds passed does not match the "
|
739 |
"number of vhostfds passed");
|
740 |
return -1; |
741 |
} |
742 |
} |
743 |
|
744 |
for (i = 0; i < nfds; i++) { |
745 |
fd = monitor_handle_fd_param(cur_mon, fds[i]); |
746 |
if (fd == -1) { |
747 |
return -1; |
748 |
} |
749 |
|
750 |
fcntl(fd, F_SETFL, O_NONBLOCK); |
751 |
|
752 |
if (i == 0) { |
753 |
vnet_hdr = tap_probe_vnet_hdr(fd); |
754 |
} else if (vnet_hdr != tap_probe_vnet_hdr(fd)) { |
755 |
error_report("vnet_hdr not consistent across given tap fds");
|
756 |
return -1; |
757 |
} |
758 |
|
759 |
if (net_init_tap_one(tap, peer, "tap", name, ifname, |
760 |
script, downscript, |
761 |
tap->has_vhostfds ? vhost_fds[i] : NULL,
|
762 |
vnet_hdr, fd)) { |
763 |
return -1; |
764 |
} |
765 |
} |
766 |
} else if (tap->has_helper) { |
767 |
if (tap->has_ifname || tap->has_script || tap->has_downscript ||
|
768 |
tap->has_vnet_hdr || tap->has_queues || tap->has_fds) { |
769 |
error_report("ifname=, script=, downscript=, and vnet_hdr= "
|
770 |
"queues=, and fds= are invalid with helper=");
|
771 |
return -1; |
772 |
} |
773 |
|
774 |
fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE); |
775 |
if (fd == -1) { |
776 |
return -1; |
777 |
} |
778 |
|
779 |
fcntl(fd, F_SETFL, O_NONBLOCK); |
780 |
vnet_hdr = tap_probe_vnet_hdr(fd); |
781 |
|
782 |
if (net_init_tap_one(tap, peer, "bridge", name, ifname, |
783 |
script, downscript, vhostfdname, |
784 |
vnet_hdr, fd)) { |
785 |
return -1; |
786 |
} |
787 |
} else {
|
788 |
script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT; |
789 |
downscript = tap->has_downscript ? tap->downscript : |
790 |
DEFAULT_NETWORK_DOWN_SCRIPT; |
791 |
|
792 |
if (tap->has_ifname) {
|
793 |
pstrcpy(ifname, sizeof ifname, tap->ifname);
|
794 |
} else {
|
795 |
ifname[0] = '\0'; |
796 |
} |
797 |
|
798 |
for (i = 0; i < queues; i++) { |
799 |
fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script, |
800 |
ifname, sizeof ifname, queues > 1); |
801 |
if (fd == -1) { |
802 |
return -1; |
803 |
} |
804 |
|
805 |
if (queues > 1 && i == 0 && !tap->has_ifname) { |
806 |
if (tap_fd_get_ifname(fd, ifname)) {
|
807 |
error_report("Fail to get ifname");
|
808 |
return -1; |
809 |
} |
810 |
} |
811 |
|
812 |
if (net_init_tap_one(tap, peer, "tap", name, ifname, |
813 |
i >= 1 ? "no" : script, |
814 |
i >= 1 ? "no" : downscript, |
815 |
vhostfdname, vnet_hdr, fd)) { |
816 |
return -1; |
817 |
} |
818 |
} |
819 |
} |
820 |
|
821 |
return 0; |
822 |
} |
823 |
|
824 |
VHostNetState *tap_get_vhost_net(NetClientState *nc) |
825 |
{ |
826 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
827 |
assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP); |
828 |
return s->vhost_net;
|
829 |
} |
830 |
|
831 |
int tap_enable(NetClientState *nc)
|
832 |
{ |
833 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
834 |
int ret;
|
835 |
|
836 |
if (s->enabled) {
|
837 |
return 0; |
838 |
} else {
|
839 |
ret = tap_fd_enable(s->fd); |
840 |
if (ret == 0) { |
841 |
s->enabled = true;
|
842 |
tap_update_fd_handler(s); |
843 |
} |
844 |
return ret;
|
845 |
} |
846 |
} |
847 |
|
848 |
int tap_disable(NetClientState *nc)
|
849 |
{ |
850 |
TAPState *s = DO_UPCAST(TAPState, nc, nc); |
851 |
int ret;
|
852 |
|
853 |
if (s->enabled == 0) { |
854 |
return 0; |
855 |
} else {
|
856 |
ret = tap_fd_disable(s->fd); |
857 |
if (ret == 0) { |
858 |
qemu_purge_queued_packets(nc); |
859 |
s->enabled = false;
|
860 |
tap_update_fd_handler(s); |
861 |
} |
862 |
return ret;
|
863 |
} |
864 |
} |