]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap.c
slirp: Clean up includes
[mirror_qemu.git] / net / tap.c
CommitLineData
5281d757
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
1422e32d 26#include "tap_int.h"
5281d757
MM
27
28#include "config-host.h"
29
5281d757
MM
30#include <sys/ioctl.h>
31#include <sys/stat.h>
32#include <sys/wait.h>
71f4effc 33#include <sys/socket.h>
5281d757
MM
34#include <net/if.h>
35
1422e32d 36#include "net/net.h"
a245fc18 37#include "clients.h"
83c9089e 38#include "monitor/monitor.h"
9c17d615 39#include "sysemu/sysemu.h"
5281d757 40#include "qemu-common.h"
1de7afc9 41#include "qemu/error-report.h"
5281d757 42
1422e32d 43#include "net/tap.h"
5281d757 44
0d09e41a 45#include "net/vhost_net.h"
82b0d80e 46
5281d757 47typedef struct TAPState {
4e68f7a0 48 NetClientState nc;
5281d757
MM
49 int fd;
50 char down_script[1024];
51 char down_script_arg[128];
d32fcad3 52 uint8_t buf[NET_BUFSIZE];
ec45f083
JW
53 bool read_poll;
54 bool write_poll;
55 bool using_vnet_hdr;
56 bool has_ufo;
16dbaf90 57 bool enabled;
82b0d80e 58 VHostNetState *vhost_net;
ef4252b1 59 unsigned host_vnet_hdr_len;
5281d757
MM
60} TAPState;
61
ac4fcf56
MA
62static void launch_script(const char *setup_script, const char *ifname,
63 int fd, Error **errp);
5281d757 64
5281d757
MM
65static void tap_send(void *opaque);
66static void tap_writable(void *opaque);
67
68static void tap_update_fd_handler(TAPState *s)
69{
82e1cc4b
FZ
70 qemu_set_fd_handler(s->fd,
71 s->read_poll && s->enabled ? tap_send : NULL,
72 s->write_poll && s->enabled ? tap_writable : NULL,
73 s);
5281d757
MM
74}
75
ec45f083 76static void tap_read_poll(TAPState *s, bool enable)
5281d757 77{
ec45f083 78 s->read_poll = enable;
5281d757
MM
79 tap_update_fd_handler(s);
80}
81
ec45f083 82static void tap_write_poll(TAPState *s, bool enable)
5281d757 83{
ec45f083 84 s->write_poll = enable;
5281d757
MM
85 tap_update_fd_handler(s);
86}
87
88static void tap_writable(void *opaque)
89{
90 TAPState *s = opaque;
91
ec45f083 92 tap_write_poll(s, false);
5281d757 93
3e35ba93 94 qemu_flush_queued_packets(&s->nc);
5281d757
MM
95}
96
97static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
98{
99 ssize_t len;
100
101 do {
102 len = writev(s->fd, iov, iovcnt);
103 } while (len == -1 && errno == EINTR);
104
105 if (len == -1 && errno == EAGAIN) {
ec45f083 106 tap_write_poll(s, true);
5281d757
MM
107 return 0;
108 }
109
110 return len;
111}
112
4e68f7a0 113static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
5281d757
MM
114 int iovcnt)
115{
3e35ba93 116 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
117 const struct iovec *iovp = iov;
118 struct iovec iov_copy[iovcnt + 1];
ef4252b1 119 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 120
ef4252b1 121 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
5281d757 122 iov_copy[0].iov_base = &hdr;
ef4252b1 123 iov_copy[0].iov_len = s->host_vnet_hdr_len;
5281d757
MM
124 memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
125 iovp = iov_copy;
126 iovcnt++;
127 }
128
129 return tap_write_packet(s, iovp, iovcnt);
130}
131
4e68f7a0 132static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 133{
3e35ba93 134 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
135 struct iovec iov[2];
136 int iovcnt = 0;
ef4252b1 137 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 138
ef4252b1 139 if (s->host_vnet_hdr_len) {
5281d757 140 iov[iovcnt].iov_base = &hdr;
ef4252b1 141 iov[iovcnt].iov_len = s->host_vnet_hdr_len;
5281d757
MM
142 iovcnt++;
143 }
144
145 iov[iovcnt].iov_base = (char *)buf;
146 iov[iovcnt].iov_len = size;
147 iovcnt++;
148
149 return tap_write_packet(s, iov, iovcnt);
150}
151
4e68f7a0 152static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 153{
3e35ba93 154 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
155 struct iovec iov[1];
156
ef4252b1 157 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
3e35ba93 158 return tap_receive_raw(nc, buf, size);
5281d757
MM
159 }
160
161 iov[0].iov_base = (char *)buf;
162 iov[0].iov_len = size;
163
164 return tap_write_packet(s, iov, 1);
165}
166
966ea5ec
MM
167#ifndef __sun__
168ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
5281d757
MM
169{
170 return read(tapfd, buf, maxlen);
171}
172#endif
173
4e68f7a0 174static void tap_send_completed(NetClientState *nc, ssize_t len)
5281d757 175{
3e35ba93 176 TAPState *s = DO_UPCAST(TAPState, nc, nc);
ec45f083 177 tap_read_poll(s, true);
5281d757
MM
178}
179
180static void tap_send(void *opaque)
181{
182 TAPState *s = opaque;
183 int size;
756ae78b 184 int packets = 0;
5281d757 185
a90a7425 186 while (true) {
5819c918
MM
187 uint8_t *buf = s->buf;
188
189 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
190 if (size <= 0) {
191 break;
192 }
193
ef4252b1
MT
194 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
195 buf += s->host_vnet_hdr_len;
196 size -= s->host_vnet_hdr_len;
5819c918
MM
197 }
198
3e35ba93 199 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
5819c918 200 if (size == 0) {
ec45f083 201 tap_read_poll(s, false);
68e5ec64
SH
202 break;
203 } else if (size < 0) {
204 break;
5819c918 205 }
756ae78b
WK
206
207 /*
208 * When the host keeps receiving more packets while tap_send() is
209 * running we can hog the QEMU global mutex. Limit the number of
210 * packets that are processed per tap_send() callback to prevent
211 * stalling the guest.
212 */
213 packets++;
214 if (packets >= 50) {
215 break;
216 }
68e5ec64 217 }
5281d757
MM
218}
219
3bac80d3 220static bool tap_has_ufo(NetClientState *nc)
5281d757 221{
3e35ba93 222 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 223
2be64a68 224 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757
MM
225
226 return s->has_ufo;
227}
228
3bac80d3 229static bool tap_has_vnet_hdr(NetClientState *nc)
5281d757 230{
3e35ba93 231 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 232
2be64a68 233 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757 234
ef4252b1 235 return !!s->host_vnet_hdr_len;
5281d757
MM
236}
237
3bac80d3 238static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
239{
240 TAPState *s = DO_UPCAST(TAPState, nc, nc);
241
2be64a68 242 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f 243
e96dfd11 244 return !!tap_probe_vnet_hdr_len(s->fd, len);
445d892f
MT
245}
246
3bac80d3 247static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
248{
249 TAPState *s = DO_UPCAST(TAPState, nc, nc);
250
2be64a68 251 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f
MT
252 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
253 len == sizeof(struct virtio_net_hdr));
254
255 tap_fd_set_vnet_hdr_len(s->fd, len);
256 s->host_vnet_hdr_len = len;
257}
258
3bac80d3 259static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
5281d757 260{
3e35ba93 261 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 262
2be64a68 263 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
ef4252b1 264 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
5281d757
MM
265
266 s->using_vnet_hdr = using_vnet_hdr;
267}
268
c80cd6bb
GK
269static int tap_set_vnet_le(NetClientState *nc, bool is_le)
270{
271 TAPState *s = DO_UPCAST(TAPState, nc, nc);
272
273 return tap_fd_set_vnet_le(s->fd, is_le);
274}
275
276static int tap_set_vnet_be(NetClientState *nc, bool is_be)
277{
278 TAPState *s = DO_UPCAST(TAPState, nc, nc);
279
280 return tap_fd_set_vnet_be(s->fd, is_be);
281}
282
3bac80d3 283static void tap_set_offload(NetClientState *nc, int csum, int tso4,
5281d757
MM
284 int tso6, int ecn, int ufo)
285{
3e35ba93 286 TAPState *s = DO_UPCAST(TAPState, nc, nc);
27a6375d
MT
287 if (s->fd < 0) {
288 return;
289 }
5281d757 290
27a6375d 291 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
5281d757
MM
292}
293
4e68f7a0 294static void tap_cleanup(NetClientState *nc)
5281d757 295{
3e35ba93 296 TAPState *s = DO_UPCAST(TAPState, nc, nc);
ac4fcf56 297 Error *err = NULL;
5281d757 298
82b0d80e
MT
299 if (s->vhost_net) {
300 vhost_net_cleanup(s->vhost_net);
43849424 301 s->vhost_net = NULL;
82b0d80e
MT
302 }
303
3e35ba93 304 qemu_purge_queued_packets(nc);
5281d757 305
ac4fcf56
MA
306 if (s->down_script[0]) {
307 launch_script(s->down_script, s->down_script_arg, s->fd, &err);
308 if (err) {
309 error_report_err(err);
310 }
311 }
5281d757 312
ec45f083
JW
313 tap_read_poll(s, false);
314 tap_write_poll(s, false);
5281d757 315 close(s->fd);
27a6375d 316 s->fd = -1;
5281d757
MM
317}
318
4e68f7a0 319static void tap_poll(NetClientState *nc, bool enable)
ceb69615
MT
320{
321 TAPState *s = DO_UPCAST(TAPState, nc, nc);
322 tap_read_poll(s, enable);
323 tap_write_poll(s, enable);
324}
325
4e68f7a0 326int tap_get_fd(NetClientState *nc)
95d528a2
MT
327{
328 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 329 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
95d528a2
MT
330 return s->fd;
331}
332
5281d757
MM
333/* fd support */
334
3e35ba93 335static NetClientInfo net_tap_info = {
2be64a68 336 .type = NET_CLIENT_OPTIONS_KIND_TAP,
3e35ba93
MM
337 .size = sizeof(TAPState),
338 .receive = tap_receive,
339 .receive_raw = tap_receive_raw,
340 .receive_iov = tap_receive_iov,
ceb69615 341 .poll = tap_poll,
3e35ba93 342 .cleanup = tap_cleanup,
2e753bcc
VM
343 .has_ufo = tap_has_ufo,
344 .has_vnet_hdr = tap_has_vnet_hdr,
345 .has_vnet_hdr_len = tap_has_vnet_hdr_len,
346 .using_vnet_hdr = tap_using_vnet_hdr,
347 .set_offload = tap_set_offload,
348 .set_vnet_hdr_len = tap_set_vnet_hdr_len,
c80cd6bb
GK
349 .set_vnet_le = tap_set_vnet_le,
350 .set_vnet_be = tap_set_vnet_be,
3e35ba93
MM
351};
352
4e68f7a0 353static TAPState *net_tap_fd_init(NetClientState *peer,
5281d757
MM
354 const char *model,
355 const char *name,
356 int fd,
357 int vnet_hdr)
358{
4e68f7a0 359 NetClientState *nc;
5281d757 360 TAPState *s;
5281d757 361
ab5f3f84 362 nc = qemu_new_net_client(&net_tap_info, peer, model, name);
3e35ba93
MM
363
364 s = DO_UPCAST(TAPState, nc, nc);
365
5281d757 366 s->fd = fd;
ef4252b1 367 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
ec45f083 368 s->using_vnet_hdr = false;
9c282718 369 s->has_ufo = tap_probe_has_ufo(s->fd);
16dbaf90 370 s->enabled = true;
3e35ba93 371 tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
58ddcd50
MT
372 /*
373 * Make sure host header length is set correctly in tap:
374 * it might have been modified by another instance of qemu.
375 */
376 if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
377 tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
378 }
ec45f083 379 tap_read_poll(s, true);
82b0d80e 380 s->vhost_net = NULL;
5281d757
MM
381 return s;
382}
383
ac4fcf56
MA
384static void launch_script(const char *setup_script, const char *ifname,
385 int fd, Error **errp)
5281d757 386{
5281d757
MM
387 int pid, status;
388 char *args[3];
389 char **parg;
390
5281d757
MM
391 /* try to launch network script */
392 pid = fork();
ac4fcf56
MA
393 if (pid < 0) {
394 error_setg_errno(errp, errno, "could not launch network script %s",
395 setup_script);
396 return;
397 }
5281d757
MM
398 if (pid == 0) {
399 int open_max = sysconf(_SC_OPEN_MAX), i;
400
13a12f86
PG
401 for (i = 3; i < open_max; i++) {
402 if (i != fd) {
5281d757
MM
403 close(i);
404 }
405 }
406 parg = args;
407 *parg++ = (char *)setup_script;
408 *parg++ = (char *)ifname;
9678d950 409 *parg = NULL;
5281d757
MM
410 execv(setup_script, args);
411 _exit(1);
ac4fcf56 412 } else {
5281d757
MM
413 while (waitpid(pid, &status, 0) != pid) {
414 /* loop */
415 }
5281d757
MM
416
417 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
ac4fcf56 418 return;
5281d757 419 }
ac4fcf56
MA
420 error_setg(errp, "network script %s failed with status %d",
421 setup_script, status);
5281d757 422 }
5281d757
MM
423}
424
a7c36ee4
CB
425static int recv_fd(int c)
426{
427 int fd;
428 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
429 struct msghdr msg = {
430 .msg_control = msgbuf,
431 .msg_controllen = sizeof(msgbuf),
432 };
433 struct cmsghdr *cmsg;
434 struct iovec iov;
435 uint8_t req[1];
436 ssize_t len;
437
438 cmsg = CMSG_FIRSTHDR(&msg);
439 cmsg->cmsg_level = SOL_SOCKET;
440 cmsg->cmsg_type = SCM_RIGHTS;
441 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
442 msg.msg_controllen = cmsg->cmsg_len;
443
444 iov.iov_base = req;
445 iov.iov_len = sizeof(req);
446
447 msg.msg_iov = &iov;
448 msg.msg_iovlen = 1;
449
450 len = recvmsg(c, &msg, 0);
451 if (len > 0) {
452 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
453 return fd;
454 }
455
456 return len;
457}
458
a8a21be9
MA
459static int net_bridge_run_helper(const char *helper, const char *bridge,
460 Error **errp)
a7c36ee4
CB
461{
462 sigset_t oldmask, mask;
463 int pid, status;
464 char *args[5];
465 char **parg;
466 int sv[2];
467
468 sigemptyset(&mask);
469 sigaddset(&mask, SIGCHLD);
470 sigprocmask(SIG_BLOCK, &mask, &oldmask);
471
472 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
a8a21be9 473 error_setg_errno(errp, errno, "socketpair() failed");
a7c36ee4
CB
474 return -1;
475 }
476
477 /* try to launch bridge helper */
478 pid = fork();
a8a21be9
MA
479 if (pid < 0) {
480 error_setg_errno(errp, errno, "Can't fork bridge helper");
481 return -1;
482 }
a7c36ee4
CB
483 if (pid == 0) {
484 int open_max = sysconf(_SC_OPEN_MAX), i;
485 char fd_buf[6+10];
486 char br_buf[6+IFNAMSIZ] = {0};
487 char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
488
13a12f86
PG
489 for (i = 3; i < open_max; i++) {
490 if (i != sv[1]) {
a7c36ee4
CB
491 close(i);
492 }
493 }
494
495 snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
496
497 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
498 /* assume helper is a command */
499
500 if (strstr(helper, "--br=") == NULL) {
501 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
502 }
503
504 snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
505 helper, "--use-vnet", fd_buf, br_buf);
506
507 parg = args;
508 *parg++ = (char *)"sh";
509 *parg++ = (char *)"-c";
510 *parg++ = helper_cmd;
511 *parg++ = NULL;
512
513 execv("/bin/sh", args);
514 } else {
515 /* assume helper is just the executable path name */
516
517 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
518
519 parg = args;
520 *parg++ = (char *)helper;
521 *parg++ = (char *)"--use-vnet";
522 *parg++ = fd_buf;
523 *parg++ = br_buf;
524 *parg++ = NULL;
525
526 execv(helper, args);
527 }
528 _exit(1);
529
a8a21be9 530 } else {
a7c36ee4 531 int fd;
a8a21be9 532 int saved_errno;
a7c36ee4
CB
533
534 close(sv[1]);
535
536 do {
537 fd = recv_fd(sv[0]);
538 } while (fd == -1 && errno == EINTR);
a8a21be9 539 saved_errno = errno;
a7c36ee4
CB
540
541 close(sv[0]);
542
543 while (waitpid(pid, &status, 0) != pid) {
544 /* loop */
545 }
546 sigprocmask(SIG_SETMASK, &oldmask, NULL);
547 if (fd < 0) {
a8a21be9
MA
548 error_setg_errno(errp, saved_errno,
549 "failed to recv file descriptor");
a7c36ee4
CB
550 return -1;
551 }
a8a21be9
MA
552 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
553 error_setg(errp, "bridge helper failed");
554 return -1;
a7c36ee4 555 }
a8a21be9 556 return fd;
a7c36ee4 557 }
a7c36ee4
CB
558}
559
1a0c0958 560int net_init_bridge(const NetClientOptions *opts, const char *name,
a30ecde6 561 NetClientState *peer, Error **errp)
a7c36ee4 562{
f79b51b0
LE
563 const NetdevBridgeOptions *bridge;
564 const char *helper, *br;
a7c36ee4
CB
565 TAPState *s;
566 int fd, vnet_hdr;
567
8d0bcba8
EB
568 assert(opts->type == NET_CLIENT_OPTIONS_KIND_BRIDGE);
569 bridge = opts->u.bridge;
f79b51b0
LE
570
571 helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
572 br = bridge->has_br ? bridge->br : DEFAULT_BRIDGE_INTERFACE;
a7c36ee4 573
a8a21be9 574 fd = net_bridge_run_helper(helper, br, errp);
a7c36ee4
CB
575 if (fd == -1) {
576 return -1;
577 }
578
579 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4 580 vnet_hdr = tap_probe_vnet_hdr(fd);
d33d93b2 581 s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
a7c36ee4 582
f79b51b0
LE
583 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
584 br);
a7c36ee4
CB
585
586 return 0;
587}
588
08c573a8
LE
589static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
590 const char *setup_script, char *ifname,
468dd824 591 size_t ifname_sz, int mq_required, Error **errp)
5281d757 592{
ac4fcf56 593 Error *err = NULL;
5281d757 594 int fd, vnet_hdr_required;
5281d757 595
08c573a8
LE
596 if (tap->has_vnet_hdr) {
597 *vnet_hdr = tap->vnet_hdr;
5281d757
MM
598 vnet_hdr_required = *vnet_hdr;
599 } else {
08c573a8 600 *vnet_hdr = 1;
5281d757
MM
601 vnet_hdr_required = 0;
602 }
603
264986e2 604 TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
468dd824 605 mq_required, errp));
5281d757
MM
606 if (fd < 0) {
607 return -1;
608 }
609
5281d757
MM
610 if (setup_script &&
611 setup_script[0] != '\0' &&
ac4fcf56
MA
612 strcmp(setup_script, "no") != 0) {
613 launch_script(setup_script, ifname, fd, &err);
614 if (err) {
468dd824 615 error_propagate(errp, err);
ac4fcf56
MA
616 close(fd);
617 return -1;
618 }
5281d757
MM
619 }
620
5281d757
MM
621 return fd;
622}
623
264986e2
JW
624#define MAX_TAP_QUEUES 1024
625
445f116c
MA
626static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
627 const char *model, const char *name,
628 const char *ifname, const char *script,
629 const char *downscript, const char *vhostfdname,
630 int vnet_hdr, int fd, Error **errp)
5193e5fb 631{
1677f4c6 632 Error *err = NULL;
da4a4eac 633 TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
81647a65 634 int vhostfd;
5193e5fb 635
80b832c3
MA
636 tap_set_sndbuf(s->fd, tap, &err);
637 if (err) {
445f116c
MA
638 error_propagate(errp, err);
639 return;
5193e5fb
JW
640 }
641
264986e2 642 if (tap->has_fd || tap->has_fds) {
5193e5fb
JW
643 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
644 } else if (tap->has_helper) {
645 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
646 tap->helper);
647 } else {
5193e5fb
JW
648 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
649 "ifname=%s,script=%s,downscript=%s", ifname, script,
650 downscript);
651
652 if (strcmp(downscript, "no") != 0) {
653 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
654 snprintf(s->down_script_arg, sizeof(s->down_script_arg),
655 "%s", ifname);
656 }
657 }
658
659 if (tap->has_vhost ? tap->vhost :
660 vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
81647a65
NN
661 VhostNetOptions options;
662
1a1bfac9 663 options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
81647a65 664 options.net_backend = &s->nc;
5193e5fb 665
7873df40 666 if (tap->has_vhostfd || tap->has_vhostfds) {
1677f4c6 667 vhostfd = monitor_fd_param(cur_mon, vhostfdname, &err);
5193e5fb 668 if (vhostfd == -1) {
445f116c
MA
669 error_propagate(errp, err);
670 return;
5193e5fb
JW
671 }
672 } else {
81647a65
NN
673 vhostfd = open("/dev/vhost-net", O_RDWR);
674 if (vhostfd < 0) {
445f116c
MA
675 error_setg_errno(errp, errno,
676 "tap: open vhost char device failed");
677 return;
81647a65 678 }
5193e5fb 679 }
81647a65 680 options.opaque = (void *)(uintptr_t)vhostfd;
5193e5fb 681
81647a65 682 s->vhost_net = vhost_net_init(&options);
5193e5fb 683 if (!s->vhost_net) {
445f116c
MA
684 error_setg(errp,
685 "vhost-net requested but could not be initialized");
686 return;
5193e5fb 687 }
264986e2 688 } else if (tap->has_vhostfd || tap->has_vhostfds) {
445f116c 689 error_setg(errp, "vhostfd= is not valid without vhost");
5193e5fb 690 }
5193e5fb
JW
691}
692
264986e2
JW
693static int get_fds(char *str, char *fds[], int max)
694{
695 char *ptr = str, *this;
696 size_t len = strlen(str);
697 int i = 0;
698
699 while (i < max && ptr < str + len) {
700 this = strchr(ptr, ':');
701
702 if (this == NULL) {
703 fds[i] = g_strdup(ptr);
704 } else {
705 fds[i] = g_strndup(ptr, this - ptr);
706 }
707
708 i++;
709 if (this == NULL) {
710 break;
711 } else {
712 ptr = this + 1;
713 }
714 }
715
716 return i;
717}
718
1a0c0958 719int net_init_tap(const NetClientOptions *opts, const char *name,
a30ecde6 720 NetClientState *peer, Error **errp)
5281d757 721{
08c573a8 722 const NetdevTapOptions *tap;
264986e2 723 int fd, vnet_hdr = 0, i = 0, queues;
08c573a8
LE
724 /* for the no-fd, no-helper case */
725 const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
5193e5fb 726 const char *downscript = NULL;
1677f4c6 727 Error *err = NULL;
264986e2 728 const char *vhostfdname;
08c573a8
LE
729 char ifname[128];
730
8d0bcba8
EB
731 assert(opts->type == NET_CLIENT_OPTIONS_KIND_TAP);
732 tap = opts->u.tap;
264986e2
JW
733 queues = tap->has_queues ? tap->queues : 1;
734 vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
5281d757 735
ce675a75
JW
736 /* QEMU vlans does not support multiqueue tap, in this case peer is set.
737 * For -netdev, peer is always NULL. */
738 if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
a3088177 739 error_setg(errp, "Multiqueue tap cannot be used with QEMU vlans");
ce675a75
JW
740 return -1;
741 }
742
08c573a8
LE
743 if (tap->has_fd) {
744 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
264986e2 745 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 746 tap->has_fds || tap->has_vhostfds) {
a3088177
MA
747 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
748 "helper=, queues=, fds=, and vhostfds= "
749 "are invalid with fd=");
5281d757
MM
750 return -1;
751 }
752
1677f4c6 753 fd = monitor_fd_param(cur_mon, tap->fd, &err);
5281d757 754 if (fd == -1) {
a3088177 755 error_propagate(errp, err);
5281d757
MM
756 return -1;
757 }
758
759 fcntl(fd, F_SETFL, O_NONBLOCK);
760
761 vnet_hdr = tap_probe_vnet_hdr(fd);
a7c36ee4 762
445f116c
MA
763 net_init_tap_one(tap, peer, "tap", name, NULL,
764 script, downscript,
765 vhostfdname, vnet_hdr, fd, &err);
766 if (err) {
a3088177 767 error_propagate(errp, err);
264986e2
JW
768 return -1;
769 }
770 } else if (tap->has_fds) {
771 char *fds[MAX_TAP_QUEUES];
772 char *vhost_fds[MAX_TAP_QUEUES];
773 int nfds, nvhosts;
774
775 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
776 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 777 tap->has_vhostfd) {
a3088177
MA
778 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
779 "helper=, queues=, and vhostfd= "
780 "are invalid with fds=");
264986e2
JW
781 return -1;
782 }
783
784 nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
785 if (tap->has_vhostfds) {
786 nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
787 if (nfds != nvhosts) {
a3088177
MA
788 error_setg(errp, "The number of fds passed does not match "
789 "the number of vhostfds passed");
264986e2
JW
790 return -1;
791 }
792 }
793
794 for (i = 0; i < nfds; i++) {
1677f4c6 795 fd = monitor_fd_param(cur_mon, fds[i], &err);
264986e2 796 if (fd == -1) {
a3088177 797 error_propagate(errp, err);
264986e2
JW
798 return -1;
799 }
800
801 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4 802
264986e2
JW
803 if (i == 0) {
804 vnet_hdr = tap_probe_vnet_hdr(fd);
805 } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
a3088177
MA
806 error_setg(errp,
807 "vnet_hdr not consistent across given tap fds");
264986e2
JW
808 return -1;
809 }
810
445f116c
MA
811 net_init_tap_one(tap, peer, "tap", name, ifname,
812 script, downscript,
813 tap->has_vhostfds ? vhost_fds[i] : NULL,
814 vnet_hdr, fd, &err);
815 if (err) {
a3088177 816 error_propagate(errp, err);
264986e2
JW
817 return -1;
818 }
819 }
08c573a8
LE
820 } else if (tap->has_helper) {
821 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
c87826a8 822 tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
a3088177
MA
823 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
824 "queues=, and vhostfds= are invalid with helper=");
a7c36ee4
CB
825 return -1;
826 }
827
a8a21be9
MA
828 fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE,
829 errp);
a7c36ee4
CB
830 if (fd == -1) {
831 return -1;
832 }
833
834 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4
CB
835 vnet_hdr = tap_probe_vnet_hdr(fd);
836
445f116c
MA
837 net_init_tap_one(tap, peer, "bridge", name, ifname,
838 script, downscript, vhostfdname,
839 vnet_hdr, fd, &err);
840 if (err) {
a3088177 841 error_propagate(errp, err);
84f8f3da 842 close(fd);
264986e2
JW
843 return -1;
844 }
5281d757 845 } else {
c87826a8 846 if (tap->has_vhostfds) {
a3088177 847 error_setg(errp, "vhostfds= is invalid if fds= wasn't specified");
c87826a8
JW
848 return -1;
849 }
08c573a8 850 script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
5193e5fb
JW
851 downscript = tap->has_downscript ? tap->downscript :
852 DEFAULT_NETWORK_DOWN_SCRIPT;
264986e2
JW
853
854 if (tap->has_ifname) {
855 pstrcpy(ifname, sizeof ifname, tap->ifname);
856 } else {
857 ifname[0] = '\0';
929fe497 858 }
a7c36ee4 859
264986e2
JW
860 for (i = 0; i < queues; i++) {
861 fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
a3088177 862 ifname, sizeof ifname, queues > 1, errp);
264986e2
JW
863 if (fd == -1) {
864 return -1;
865 }
866
867 if (queues > 1 && i == 0 && !tap->has_ifname) {
868 if (tap_fd_get_ifname(fd, ifname)) {
a3088177 869 error_setg(errp, "Fail to get ifname");
84f8f3da 870 close(fd);
264986e2
JW
871 return -1;
872 }
873 }
874
445f116c
MA
875 net_init_tap_one(tap, peer, "tap", name, ifname,
876 i >= 1 ? "no" : script,
877 i >= 1 ? "no" : downscript,
878 vhostfdname, vnet_hdr, fd, &err);
879 if (err) {
a3088177 880 error_propagate(errp, err);
84f8f3da 881 close(fd);
264986e2
JW
882 return -1;
883 }
884 }
5281d757
MM
885 }
886
264986e2 887 return 0;
5281d757 888}
b202554c 889
4e68f7a0 890VHostNetState *tap_get_vhost_net(NetClientState *nc)
b202554c
MT
891{
892 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 893 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
b202554c
MT
894 return s->vhost_net;
895}
16dbaf90
JW
896
897int tap_enable(NetClientState *nc)
898{
899 TAPState *s = DO_UPCAST(TAPState, nc, nc);
900 int ret;
901
902 if (s->enabled) {
903 return 0;
904 } else {
905 ret = tap_fd_enable(s->fd);
906 if (ret == 0) {
907 s->enabled = true;
908 tap_update_fd_handler(s);
909 }
910 return ret;
911 }
912}
913
914int tap_disable(NetClientState *nc)
915{
916 TAPState *s = DO_UPCAST(TAPState, nc, nc);
917 int ret;
918
919 if (s->enabled == 0) {
920 return 0;
921 } else {
922 ret = tap_fd_disable(s->fd);
923 if (ret == 0) {
924 qemu_purge_queued_packets(nc);
925 s->enabled = false;
926 tap_update_fd_handler(s);
927 }
928 return ret;
929 }
930}