]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap.c
tap: Convert net_init_tap_one() to Error
[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
62static int launch_script(const char *setup_script, const char *ifname, int fd);
63
64static int tap_can_send(void *opaque);
65static void tap_send(void *opaque);
66static void tap_writable(void *opaque);
67
68static void tap_update_fd_handler(TAPState *s)
69{
70 qemu_set_fd_handler2(s->fd,
16dbaf90
JW
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,
5281d757
MM
74 s);
75}
76
ec45f083 77static void tap_read_poll(TAPState *s, bool enable)
5281d757 78{
ec45f083 79 s->read_poll = enable;
5281d757
MM
80 tap_update_fd_handler(s);
81}
82
ec45f083 83static void tap_write_poll(TAPState *s, bool enable)
5281d757 84{
ec45f083 85 s->write_poll = enable;
5281d757
MM
86 tap_update_fd_handler(s);
87}
88
89static void tap_writable(void *opaque)
90{
91 TAPState *s = opaque;
92
ec45f083 93 tap_write_poll(s, false);
5281d757 94
3e35ba93 95 qemu_flush_queued_packets(&s->nc);
5281d757
MM
96}
97
98static 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) {
ec45f083 107 tap_write_poll(s, true);
5281d757
MM
108 return 0;
109 }
110
111 return len;
112}
113
4e68f7a0 114static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
5281d757
MM
115 int iovcnt)
116{
3e35ba93 117 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
118 const struct iovec *iovp = iov;
119 struct iovec iov_copy[iovcnt + 1];
ef4252b1 120 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 121
ef4252b1 122 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
5281d757 123 iov_copy[0].iov_base = &hdr;
ef4252b1 124 iov_copy[0].iov_len = s->host_vnet_hdr_len;
5281d757
MM
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
4e68f7a0 133static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 134{
3e35ba93 135 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
136 struct iovec iov[2];
137 int iovcnt = 0;
ef4252b1 138 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 139
ef4252b1 140 if (s->host_vnet_hdr_len) {
5281d757 141 iov[iovcnt].iov_base = &hdr;
ef4252b1 142 iov[iovcnt].iov_len = s->host_vnet_hdr_len;
5281d757
MM
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
4e68f7a0 153static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 154{
3e35ba93 155 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
156 struct iovec iov[1];
157
ef4252b1 158 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
3e35ba93 159 return tap_receive_raw(nc, buf, size);
5281d757
MM
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
168static int tap_can_send(void *opaque)
169{
170 TAPState *s = opaque;
171
3e35ba93 172 return qemu_can_send_packet(&s->nc);
5281d757
MM
173}
174
966ea5ec
MM
175#ifndef __sun__
176ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
5281d757
MM
177{
178 return read(tapfd, buf, maxlen);
179}
180#endif
181
4e68f7a0 182static void tap_send_completed(NetClientState *nc, ssize_t len)
5281d757 183{
3e35ba93 184 TAPState *s = DO_UPCAST(TAPState, nc, nc);
ec45f083 185 tap_read_poll(s, true);
5281d757
MM
186}
187
188static void tap_send(void *opaque)
189{
190 TAPState *s = opaque;
191 int size;
756ae78b 192 int packets = 0;
5281d757 193
68e5ec64 194 while (qemu_can_send_packet(&s->nc)) {
5819c918
MM
195 uint8_t *buf = s->buf;
196
197 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
198 if (size <= 0) {
199 break;
200 }
201
ef4252b1
MT
202 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
203 buf += s->host_vnet_hdr_len;
204 size -= s->host_vnet_hdr_len;
5819c918
MM
205 }
206
3e35ba93 207 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
5819c918 208 if (size == 0) {
ec45f083 209 tap_read_poll(s, false);
68e5ec64
SH
210 break;
211 } else if (size < 0) {
212 break;
5819c918 213 }
756ae78b
WK
214
215 /*
216 * When the host keeps receiving more packets while tap_send() is
217 * running we can hog the QEMU global mutex. Limit the number of
218 * packets that are processed per tap_send() callback to prevent
219 * stalling the guest.
220 */
221 packets++;
222 if (packets >= 50) {
223 break;
224 }
68e5ec64 225 }
5281d757
MM
226}
227
3bac80d3 228static bool tap_has_ufo(NetClientState *nc)
5281d757 229{
3e35ba93 230 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 231
2be64a68 232 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757
MM
233
234 return s->has_ufo;
235}
236
3bac80d3 237static bool tap_has_vnet_hdr(NetClientState *nc)
5281d757 238{
3e35ba93 239 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 240
2be64a68 241 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757 242
ef4252b1 243 return !!s->host_vnet_hdr_len;
5281d757
MM
244}
245
3bac80d3 246static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
247{
248 TAPState *s = DO_UPCAST(TAPState, nc, nc);
249
2be64a68 250 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f 251
e96dfd11 252 return !!tap_probe_vnet_hdr_len(s->fd, len);
445d892f
MT
253}
254
3bac80d3 255static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
256{
257 TAPState *s = DO_UPCAST(TAPState, nc, nc);
258
2be64a68 259 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f
MT
260 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
261 len == sizeof(struct virtio_net_hdr));
262
263 tap_fd_set_vnet_hdr_len(s->fd, len);
264 s->host_vnet_hdr_len = len;
265}
266
3bac80d3 267static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
5281d757 268{
3e35ba93 269 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 270
2be64a68 271 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
ef4252b1 272 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
5281d757
MM
273
274 s->using_vnet_hdr = using_vnet_hdr;
275}
276
3bac80d3 277static void tap_set_offload(NetClientState *nc, int csum, int tso4,
5281d757
MM
278 int tso6, int ecn, int ufo)
279{
3e35ba93 280 TAPState *s = DO_UPCAST(TAPState, nc, nc);
27a6375d
MT
281 if (s->fd < 0) {
282 return;
283 }
5281d757 284
27a6375d 285 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
5281d757
MM
286}
287
4e68f7a0 288static void tap_cleanup(NetClientState *nc)
5281d757 289{
3e35ba93 290 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 291
82b0d80e
MT
292 if (s->vhost_net) {
293 vhost_net_cleanup(s->vhost_net);
43849424 294 s->vhost_net = NULL;
82b0d80e
MT
295 }
296
3e35ba93 297 qemu_purge_queued_packets(nc);
5281d757
MM
298
299 if (s->down_script[0])
300 launch_script(s->down_script, s->down_script_arg, s->fd);
301
ec45f083
JW
302 tap_read_poll(s, false);
303 tap_write_poll(s, false);
5281d757 304 close(s->fd);
27a6375d 305 s->fd = -1;
5281d757
MM
306}
307
4e68f7a0 308static void tap_poll(NetClientState *nc, bool enable)
ceb69615
MT
309{
310 TAPState *s = DO_UPCAST(TAPState, nc, nc);
311 tap_read_poll(s, enable);
312 tap_write_poll(s, enable);
313}
314
4e68f7a0 315int tap_get_fd(NetClientState *nc)
95d528a2
MT
316{
317 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 318 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
95d528a2
MT
319 return s->fd;
320}
321
5281d757
MM
322/* fd support */
323
3e35ba93 324static NetClientInfo net_tap_info = {
2be64a68 325 .type = NET_CLIENT_OPTIONS_KIND_TAP,
3e35ba93
MM
326 .size = sizeof(TAPState),
327 .receive = tap_receive,
328 .receive_raw = tap_receive_raw,
329 .receive_iov = tap_receive_iov,
ceb69615 330 .poll = tap_poll,
3e35ba93 331 .cleanup = tap_cleanup,
2e753bcc
VM
332 .has_ufo = tap_has_ufo,
333 .has_vnet_hdr = tap_has_vnet_hdr,
334 .has_vnet_hdr_len = tap_has_vnet_hdr_len,
335 .using_vnet_hdr = tap_using_vnet_hdr,
336 .set_offload = tap_set_offload,
337 .set_vnet_hdr_len = tap_set_vnet_hdr_len,
3e35ba93
MM
338};
339
4e68f7a0 340static TAPState *net_tap_fd_init(NetClientState *peer,
5281d757
MM
341 const char *model,
342 const char *name,
343 int fd,
344 int vnet_hdr)
345{
4e68f7a0 346 NetClientState *nc;
5281d757 347 TAPState *s;
5281d757 348
ab5f3f84 349 nc = qemu_new_net_client(&net_tap_info, peer, model, name);
3e35ba93
MM
350
351 s = DO_UPCAST(TAPState, nc, nc);
352
5281d757 353 s->fd = fd;
ef4252b1 354 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
ec45f083 355 s->using_vnet_hdr = false;
9c282718 356 s->has_ufo = tap_probe_has_ufo(s->fd);
16dbaf90 357 s->enabled = true;
3e35ba93 358 tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
58ddcd50
MT
359 /*
360 * Make sure host header length is set correctly in tap:
361 * it might have been modified by another instance of qemu.
362 */
363 if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
364 tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
365 }
ec45f083 366 tap_read_poll(s, true);
82b0d80e 367 s->vhost_net = NULL;
5281d757
MM
368 return s;
369}
370
5281d757
MM
371static int launch_script(const char *setup_script, const char *ifname, int fd)
372{
5281d757
MM
373 int pid, status;
374 char *args[3];
375 char **parg;
376
5281d757
MM
377 /* try to launch network script */
378 pid = fork();
379 if (pid == 0) {
380 int open_max = sysconf(_SC_OPEN_MAX), i;
381
13a12f86
PG
382 for (i = 3; i < open_max; i++) {
383 if (i != fd) {
5281d757
MM
384 close(i);
385 }
386 }
387 parg = args;
388 *parg++ = (char *)setup_script;
389 *parg++ = (char *)ifname;
9678d950 390 *parg = NULL;
5281d757
MM
391 execv(setup_script, args);
392 _exit(1);
393 } else if (pid > 0) {
394 while (waitpid(pid, &status, 0) != pid) {
395 /* loop */
396 }
5281d757
MM
397
398 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
399 return 0;
400 }
401 }
402 fprintf(stderr, "%s: could not launch network script\n", setup_script);
403 return -1;
404}
405
a7c36ee4
CB
406static int recv_fd(int c)
407{
408 int fd;
409 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
410 struct msghdr msg = {
411 .msg_control = msgbuf,
412 .msg_controllen = sizeof(msgbuf),
413 };
414 struct cmsghdr *cmsg;
415 struct iovec iov;
416 uint8_t req[1];
417 ssize_t len;
418
419 cmsg = CMSG_FIRSTHDR(&msg);
420 cmsg->cmsg_level = SOL_SOCKET;
421 cmsg->cmsg_type = SCM_RIGHTS;
422 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
423 msg.msg_controllen = cmsg->cmsg_len;
424
425 iov.iov_base = req;
426 iov.iov_len = sizeof(req);
427
428 msg.msg_iov = &iov;
429 msg.msg_iovlen = 1;
430
431 len = recvmsg(c, &msg, 0);
432 if (len > 0) {
433 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
434 return fd;
435 }
436
437 return len;
438}
439
a8a21be9
MA
440static int net_bridge_run_helper(const char *helper, const char *bridge,
441 Error **errp)
a7c36ee4
CB
442{
443 sigset_t oldmask, mask;
444 int pid, status;
445 char *args[5];
446 char **parg;
447 int sv[2];
448
449 sigemptyset(&mask);
450 sigaddset(&mask, SIGCHLD);
451 sigprocmask(SIG_BLOCK, &mask, &oldmask);
452
453 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
a8a21be9 454 error_setg_errno(errp, errno, "socketpair() failed");
a7c36ee4
CB
455 return -1;
456 }
457
458 /* try to launch bridge helper */
459 pid = fork();
a8a21be9
MA
460 if (pid < 0) {
461 error_setg_errno(errp, errno, "Can't fork bridge helper");
462 return -1;
463 }
a7c36ee4
CB
464 if (pid == 0) {
465 int open_max = sysconf(_SC_OPEN_MAX), i;
466 char fd_buf[6+10];
467 char br_buf[6+IFNAMSIZ] = {0};
468 char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
469
13a12f86
PG
470 for (i = 3; i < open_max; i++) {
471 if (i != sv[1]) {
a7c36ee4
CB
472 close(i);
473 }
474 }
475
476 snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
477
478 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
479 /* assume helper is a command */
480
481 if (strstr(helper, "--br=") == NULL) {
482 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
483 }
484
485 snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
486 helper, "--use-vnet", fd_buf, br_buf);
487
488 parg = args;
489 *parg++ = (char *)"sh";
490 *parg++ = (char *)"-c";
491 *parg++ = helper_cmd;
492 *parg++ = NULL;
493
494 execv("/bin/sh", args);
495 } else {
496 /* assume helper is just the executable path name */
497
498 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
499
500 parg = args;
501 *parg++ = (char *)helper;
502 *parg++ = (char *)"--use-vnet";
503 *parg++ = fd_buf;
504 *parg++ = br_buf;
505 *parg++ = NULL;
506
507 execv(helper, args);
508 }
509 _exit(1);
510
a8a21be9 511 } else {
a7c36ee4 512 int fd;
a8a21be9 513 int saved_errno;
a7c36ee4
CB
514
515 close(sv[1]);
516
517 do {
518 fd = recv_fd(sv[0]);
519 } while (fd == -1 && errno == EINTR);
a8a21be9 520 saved_errno = errno;
a7c36ee4
CB
521
522 close(sv[0]);
523
524 while (waitpid(pid, &status, 0) != pid) {
525 /* loop */
526 }
527 sigprocmask(SIG_SETMASK, &oldmask, NULL);
528 if (fd < 0) {
a8a21be9
MA
529 error_setg_errno(errp, saved_errno,
530 "failed to recv file descriptor");
a7c36ee4
CB
531 return -1;
532 }
a8a21be9
MA
533 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
534 error_setg(errp, "bridge helper failed");
535 return -1;
a7c36ee4 536 }
a8a21be9 537 return fd;
a7c36ee4 538 }
a7c36ee4
CB
539}
540
1a0c0958 541int net_init_bridge(const NetClientOptions *opts, const char *name,
a30ecde6 542 NetClientState *peer, Error **errp)
a7c36ee4 543{
f79b51b0
LE
544 const NetdevBridgeOptions *bridge;
545 const char *helper, *br;
a7c36ee4
CB
546 TAPState *s;
547 int fd, vnet_hdr;
548
f79b51b0
LE
549 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
550 bridge = opts->bridge;
551
552 helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
553 br = bridge->has_br ? bridge->br : DEFAULT_BRIDGE_INTERFACE;
a7c36ee4 554
a8a21be9 555 fd = net_bridge_run_helper(helper, br, errp);
a7c36ee4
CB
556 if (fd == -1) {
557 return -1;
558 }
559
560 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4 561 vnet_hdr = tap_probe_vnet_hdr(fd);
d33d93b2 562 s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
a7c36ee4 563
f79b51b0
LE
564 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
565 br);
a7c36ee4
CB
566
567 return 0;
568}
569
08c573a8
LE
570static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
571 const char *setup_script, char *ifname,
264986e2 572 size_t ifname_sz, int mq_required)
5281d757
MM
573{
574 int fd, vnet_hdr_required;
5281d757 575
08c573a8
LE
576 if (tap->has_vnet_hdr) {
577 *vnet_hdr = tap->vnet_hdr;
5281d757
MM
578 vnet_hdr_required = *vnet_hdr;
579 } else {
08c573a8 580 *vnet_hdr = 1;
5281d757
MM
581 vnet_hdr_required = 0;
582 }
583
264986e2
JW
584 TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
585 mq_required));
5281d757
MM
586 if (fd < 0) {
587 return -1;
588 }
589
5281d757
MM
590 if (setup_script &&
591 setup_script[0] != '\0' &&
592 strcmp(setup_script, "no") != 0 &&
593 launch_script(setup_script, ifname, fd)) {
594 close(fd);
595 return -1;
596 }
597
5281d757
MM
598 return fd;
599}
600
264986e2
JW
601#define MAX_TAP_QUEUES 1024
602
445f116c
MA
603static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
604 const char *model, const char *name,
605 const char *ifname, const char *script,
606 const char *downscript, const char *vhostfdname,
607 int vnet_hdr, int fd, Error **errp)
5193e5fb 608{
1677f4c6 609 Error *err = NULL;
da4a4eac 610 TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
81647a65 611 int vhostfd;
5193e5fb 612
80b832c3
MA
613 tap_set_sndbuf(s->fd, tap, &err);
614 if (err) {
445f116c
MA
615 error_propagate(errp, err);
616 return;
5193e5fb
JW
617 }
618
264986e2 619 if (tap->has_fd || tap->has_fds) {
5193e5fb
JW
620 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
621 } else if (tap->has_helper) {
622 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
623 tap->helper);
624 } else {
5193e5fb
JW
625 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
626 "ifname=%s,script=%s,downscript=%s", ifname, script,
627 downscript);
628
629 if (strcmp(downscript, "no") != 0) {
630 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
631 snprintf(s->down_script_arg, sizeof(s->down_script_arg),
632 "%s", ifname);
633 }
634 }
635
636 if (tap->has_vhost ? tap->vhost :
637 vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
81647a65
NN
638 VhostNetOptions options;
639
1a1bfac9 640 options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
81647a65
NN
641 options.net_backend = &s->nc;
642 options.force = tap->has_vhostforce && tap->vhostforce;
5193e5fb 643
7873df40 644 if (tap->has_vhostfd || tap->has_vhostfds) {
1677f4c6 645 vhostfd = monitor_fd_param(cur_mon, vhostfdname, &err);
5193e5fb 646 if (vhostfd == -1) {
445f116c
MA
647 error_propagate(errp, err);
648 return;
5193e5fb
JW
649 }
650 } else {
81647a65
NN
651 vhostfd = open("/dev/vhost-net", O_RDWR);
652 if (vhostfd < 0) {
445f116c
MA
653 error_setg_errno(errp, errno,
654 "tap: open vhost char device failed");
655 return;
81647a65 656 }
5193e5fb 657 }
81647a65 658 options.opaque = (void *)(uintptr_t)vhostfd;
5193e5fb 659
81647a65 660 s->vhost_net = vhost_net_init(&options);
5193e5fb 661 if (!s->vhost_net) {
445f116c
MA
662 error_setg(errp,
663 "vhost-net requested but could not be initialized");
664 return;
5193e5fb 665 }
264986e2 666 } else if (tap->has_vhostfd || tap->has_vhostfds) {
445f116c 667 error_setg(errp, "vhostfd= is not valid without vhost");
5193e5fb 668 }
5193e5fb
JW
669}
670
264986e2
JW
671static int get_fds(char *str, char *fds[], int max)
672{
673 char *ptr = str, *this;
674 size_t len = strlen(str);
675 int i = 0;
676
677 while (i < max && ptr < str + len) {
678 this = strchr(ptr, ':');
679
680 if (this == NULL) {
681 fds[i] = g_strdup(ptr);
682 } else {
683 fds[i] = g_strndup(ptr, this - ptr);
684 }
685
686 i++;
687 if (this == NULL) {
688 break;
689 } else {
690 ptr = this + 1;
691 }
692 }
693
694 return i;
695}
696
1a0c0958 697int net_init_tap(const NetClientOptions *opts, const char *name,
a30ecde6 698 NetClientState *peer, Error **errp)
5281d757 699{
a30ecde6 700 /* FIXME error_setg(errp, ...) on failure */
08c573a8 701 const NetdevTapOptions *tap;
264986e2 702 int fd, vnet_hdr = 0, i = 0, queues;
08c573a8
LE
703 /* for the no-fd, no-helper case */
704 const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
5193e5fb 705 const char *downscript = NULL;
1677f4c6 706 Error *err = NULL;
264986e2 707 const char *vhostfdname;
08c573a8
LE
708 char ifname[128];
709
710 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
711 tap = opts->tap;
264986e2
JW
712 queues = tap->has_queues ? tap->queues : 1;
713 vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
5281d757 714
ce675a75
JW
715 /* QEMU vlans does not support multiqueue tap, in this case peer is set.
716 * For -netdev, peer is always NULL. */
717 if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
085d8134 718 error_report("Multiqueue tap cannot be used with QEMU vlans");
ce675a75
JW
719 return -1;
720 }
721
08c573a8
LE
722 if (tap->has_fd) {
723 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
264986e2 724 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 725 tap->has_fds || tap->has_vhostfds) {
a7c36ee4 726 error_report("ifname=, script=, downscript=, vnet_hdr=, "
c87826a8
JW
727 "helper=, queues=, fds=, and vhostfds= "
728 "are invalid with fd=");
5281d757
MM
729 return -1;
730 }
731
1677f4c6 732 fd = monitor_fd_param(cur_mon, tap->fd, &err);
5281d757 733 if (fd == -1) {
1677f4c6 734 error_report_err(err);
5281d757
MM
735 return -1;
736 }
737
738 fcntl(fd, F_SETFL, O_NONBLOCK);
739
740 vnet_hdr = tap_probe_vnet_hdr(fd);
a7c36ee4 741
445f116c
MA
742 net_init_tap_one(tap, peer, "tap", name, NULL,
743 script, downscript,
744 vhostfdname, vnet_hdr, fd, &err);
745 if (err) {
746 error_report_err(err);
264986e2
JW
747 return -1;
748 }
749 } else if (tap->has_fds) {
750 char *fds[MAX_TAP_QUEUES];
751 char *vhost_fds[MAX_TAP_QUEUES];
752 int nfds, nvhosts;
753
754 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
755 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 756 tap->has_vhostfd) {
264986e2 757 error_report("ifname=, script=, downscript=, vnet_hdr=, "
c87826a8
JW
758 "helper=, queues=, and vhostfd= "
759 "are invalid with fds=");
264986e2
JW
760 return -1;
761 }
762
763 nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
764 if (tap->has_vhostfds) {
765 nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
766 if (nfds != nvhosts) {
767 error_report("The number of fds passed does not match the "
768 "number of vhostfds passed");
769 return -1;
770 }
771 }
772
773 for (i = 0; i < nfds; i++) {
1677f4c6 774 fd = monitor_fd_param(cur_mon, fds[i], &err);
264986e2 775 if (fd == -1) {
1677f4c6 776 error_report_err(err);
264986e2
JW
777 return -1;
778 }
779
780 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4 781
264986e2
JW
782 if (i == 0) {
783 vnet_hdr = tap_probe_vnet_hdr(fd);
784 } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
785 error_report("vnet_hdr not consistent across given tap fds");
786 return -1;
787 }
788
445f116c
MA
789 net_init_tap_one(tap, peer, "tap", name, ifname,
790 script, downscript,
791 tap->has_vhostfds ? vhost_fds[i] : NULL,
792 vnet_hdr, fd, &err);
793 if (err) {
794 error_report_err(err);
264986e2
JW
795 return -1;
796 }
797 }
08c573a8
LE
798 } else if (tap->has_helper) {
799 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
c87826a8 800 tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
a7c36ee4 801 error_report("ifname=, script=, downscript=, and vnet_hdr= "
c87826a8 802 "queues=, and vhostfds= are invalid with helper=");
a7c36ee4
CB
803 return -1;
804 }
805
a8a21be9
MA
806 fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE,
807 errp);
a7c36ee4
CB
808 if (fd == -1) {
809 return -1;
810 }
811
812 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4
CB
813 vnet_hdr = tap_probe_vnet_hdr(fd);
814
445f116c
MA
815 net_init_tap_one(tap, peer, "bridge", name, ifname,
816 script, downscript, vhostfdname,
817 vnet_hdr, fd, &err);
818 if (err) {
819 error_report_err(err);
84f8f3da 820 close(fd);
264986e2
JW
821 return -1;
822 }
5281d757 823 } else {
c87826a8
JW
824 if (tap->has_vhostfds) {
825 error_report("vhostfds= is invalid if fds= wasn't specified");
826 return -1;
827 }
08c573a8 828 script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
5193e5fb
JW
829 downscript = tap->has_downscript ? tap->downscript :
830 DEFAULT_NETWORK_DOWN_SCRIPT;
264986e2
JW
831
832 if (tap->has_ifname) {
833 pstrcpy(ifname, sizeof ifname, tap->ifname);
834 } else {
835 ifname[0] = '\0';
929fe497 836 }
a7c36ee4 837
264986e2
JW
838 for (i = 0; i < queues; i++) {
839 fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
840 ifname, sizeof ifname, queues > 1);
841 if (fd == -1) {
842 return -1;
843 }
844
845 if (queues > 1 && i == 0 && !tap->has_ifname) {
846 if (tap_fd_get_ifname(fd, ifname)) {
847 error_report("Fail to get ifname");
84f8f3da 848 close(fd);
264986e2
JW
849 return -1;
850 }
851 }
852
445f116c
MA
853 net_init_tap_one(tap, peer, "tap", name, ifname,
854 i >= 1 ? "no" : script,
855 i >= 1 ? "no" : downscript,
856 vhostfdname, vnet_hdr, fd, &err);
857 if (err) {
858 error_report_err(err);
84f8f3da 859 close(fd);
264986e2
JW
860 return -1;
861 }
862 }
5281d757
MM
863 }
864
264986e2 865 return 0;
5281d757 866}
b202554c 867
4e68f7a0 868VHostNetState *tap_get_vhost_net(NetClientState *nc)
b202554c
MT
869{
870 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 871 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
b202554c
MT
872 return s->vhost_net;
873}
16dbaf90
JW
874
875int tap_enable(NetClientState *nc)
876{
877 TAPState *s = DO_UPCAST(TAPState, nc, nc);
878 int ret;
879
880 if (s->enabled) {
881 return 0;
882 } else {
883 ret = tap_fd_enable(s->fd);
884 if (ret == 0) {
885 s->enabled = true;
886 tap_update_fd_handler(s);
887 }
888 return ret;
889 }
890}
891
892int tap_disable(NetClientState *nc)
893{
894 TAPState *s = DO_UPCAST(TAPState, nc, nc);
895 int ret;
896
897 if (s->enabled == 0) {
898 return 0;
899 } else {
900 ret = tap_fd_disable(s->fd);
901 if (ret == 0) {
902 qemu_purge_queued_packets(nc);
903 s->enabled = false;
904 tap_update_fd_handler(s);
905 }
906 return ret;
907 }
908}