]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap.c
tap: Add USO support to tap device.
[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
2744d920 26#include "qemu/osdep.h"
1422e32d 27#include "tap_int.h"
5281d757 28
5281d757 29
5281d757 30#include <sys/ioctl.h>
5281d757 31#include <sys/wait.h>
71f4effc 32#include <sys/socket.h>
5281d757
MM
33#include <net/if.h>
34
969e50b6 35#include "net/eth.h"
1422e32d 36#include "net/net.h"
a245fc18 37#include "clients.h"
83c9089e 38#include "monitor/monitor.h"
9c17d615 39#include "sysemu/sysemu.h"
da34e65c 40#include "qapi/error.h"
f348b6d1 41#include "qemu/cutils.h"
1de7afc9 42#include "qemu/error-report.h"
db725815 43#include "qemu/main-loop.h"
d542800d 44#include "qemu/sockets.h"
5281d757 45
1422e32d 46#include "net/tap.h"
5281d757 47
0d09e41a 48#include "net/vhost_net.h"
82b0d80e 49
5281d757 50typedef struct TAPState {
4e68f7a0 51 NetClientState nc;
5281d757
MM
52 int fd;
53 char down_script[1024];
54 char down_script_arg[128];
d32fcad3 55 uint8_t buf[NET_BUFSIZE];
ec45f083
JW
56 bool read_poll;
57 bool write_poll;
58 bool using_vnet_hdr;
59 bool has_ufo;
16dbaf90 60 bool enabled;
82b0d80e 61 VHostNetState *vhost_net;
ef4252b1 62 unsigned host_vnet_hdr_len;
9e32ff32 63 Notifier exit;
5281d757
MM
64} TAPState;
65
ac4fcf56
MA
66static void launch_script(const char *setup_script, const char *ifname,
67 int fd, Error **errp);
5281d757 68
5281d757
MM
69static void tap_send(void *opaque);
70static void tap_writable(void *opaque);
71
72static void tap_update_fd_handler(TAPState *s)
73{
82e1cc4b
FZ
74 qemu_set_fd_handler(s->fd,
75 s->read_poll && s->enabled ? tap_send : NULL,
76 s->write_poll && s->enabled ? tap_writable : NULL,
77 s);
5281d757
MM
78}
79
ec45f083 80static void tap_read_poll(TAPState *s, bool enable)
5281d757 81{
ec45f083 82 s->read_poll = enable;
5281d757
MM
83 tap_update_fd_handler(s);
84}
85
ec45f083 86static void tap_write_poll(TAPState *s, bool enable)
5281d757 87{
ec45f083 88 s->write_poll = enable;
5281d757
MM
89 tap_update_fd_handler(s);
90}
91
92static void tap_writable(void *opaque)
93{
94 TAPState *s = opaque;
95
ec45f083 96 tap_write_poll(s, false);
5281d757 97
3e35ba93 98 qemu_flush_queued_packets(&s->nc);
5281d757
MM
99}
100
101static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
102{
103 ssize_t len;
104
37b0b24e 105 len = RETRY_ON_EINTR(writev(s->fd, iov, iovcnt));
5281d757
MM
106
107 if (len == -1 && errno == EAGAIN) {
ec45f083 108 tap_write_poll(s, true);
5281d757
MM
109 return 0;
110 }
111
112 return len;
113}
114
4e68f7a0 115static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
5281d757
MM
116 int iovcnt)
117{
3e35ba93 118 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
119 const struct iovec *iovp = iov;
120 struct iovec iov_copy[iovcnt + 1];
ef4252b1 121 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 122
ef4252b1 123 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
5281d757 124 iov_copy[0].iov_base = &hdr;
ef4252b1 125 iov_copy[0].iov_len = s->host_vnet_hdr_len;
5281d757
MM
126 memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
127 iovp = iov_copy;
128 iovcnt++;
129 }
130
131 return tap_write_packet(s, iovp, iovcnt);
132}
133
4e68f7a0 134static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 135{
3e35ba93 136 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
137 struct iovec iov[2];
138 int iovcnt = 0;
ef4252b1 139 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 140
ef4252b1 141 if (s->host_vnet_hdr_len) {
5281d757 142 iov[iovcnt].iov_base = &hdr;
ef4252b1 143 iov[iovcnt].iov_len = s->host_vnet_hdr_len;
5281d757
MM
144 iovcnt++;
145 }
146
147 iov[iovcnt].iov_base = (char *)buf;
148 iov[iovcnt].iov_len = size;
149 iovcnt++;
150
151 return tap_write_packet(s, iov, iovcnt);
152}
153
4e68f7a0 154static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 155{
3e35ba93 156 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
157 struct iovec iov[1];
158
ef4252b1 159 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
3e35ba93 160 return tap_receive_raw(nc, buf, size);
5281d757
MM
161 }
162
163 iov[0].iov_base = (char *)buf;
164 iov[0].iov_len = size;
165
166 return tap_write_packet(s, iov, 1);
167}
168
966ea5ec
MM
169#ifndef __sun__
170ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
5281d757
MM
171{
172 return read(tapfd, buf, maxlen);
173}
174#endif
175
4e68f7a0 176static void tap_send_completed(NetClientState *nc, ssize_t len)
5281d757 177{
3e35ba93 178 TAPState *s = DO_UPCAST(TAPState, nc, nc);
ec45f083 179 tap_read_poll(s, true);
5281d757
MM
180}
181
182static void tap_send(void *opaque)
183{
184 TAPState *s = opaque;
185 int size;
756ae78b 186 int packets = 0;
5281d757 187
a90a7425 188 while (true) {
5819c918 189 uint8_t *buf = s->buf;
969e50b6
BM
190 uint8_t min_pkt[ETH_ZLEN];
191 size_t min_pktsz = sizeof(min_pkt);
5819c918
MM
192
193 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
194 if (size <= 0) {
195 break;
196 }
197
ef4252b1
MT
198 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
199 buf += s->host_vnet_hdr_len;
200 size -= s->host_vnet_hdr_len;
5819c918
MM
201 }
202
bc38e31b 203 if (net_peer_needs_padding(&s->nc)) {
969e50b6
BM
204 if (eth_pad_short_frame(min_pkt, &min_pktsz, buf, size)) {
205 buf = min_pkt;
206 size = min_pktsz;
207 }
208 }
209
3e35ba93 210 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
5819c918 211 if (size == 0) {
ec45f083 212 tap_read_poll(s, false);
68e5ec64
SH
213 break;
214 } else if (size < 0) {
215 break;
5819c918 216 }
756ae78b
WK
217
218 /*
219 * When the host keeps receiving more packets while tap_send() is
220 * running we can hog the QEMU global mutex. Limit the number of
221 * packets that are processed per tap_send() callback to prevent
222 * stalling the guest.
223 */
224 packets++;
225 if (packets >= 50) {
226 break;
227 }
68e5ec64 228 }
5281d757
MM
229}
230
3bac80d3 231static bool tap_has_ufo(NetClientState *nc)
5281d757 232{
3e35ba93 233 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 234
f394b2e2 235 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
5281d757
MM
236
237 return s->has_ufo;
238}
239
3bac80d3 240static bool tap_has_vnet_hdr(NetClientState *nc)
5281d757 241{
3e35ba93 242 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 243
f394b2e2 244 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
5281d757 245
ef4252b1 246 return !!s->host_vnet_hdr_len;
5281d757
MM
247}
248
3bac80d3 249static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
250{
251 TAPState *s = DO_UPCAST(TAPState, nc, nc);
252
f394b2e2 253 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
445d892f 254
e96dfd11 255 return !!tap_probe_vnet_hdr_len(s->fd, len);
445d892f
MT
256}
257
481c5232
AO
258static int tap_get_vnet_hdr_len(NetClientState *nc)
259{
260 TAPState *s = DO_UPCAST(TAPState, nc, nc);
261
262 return s->host_vnet_hdr_len;
263}
264
3bac80d3 265static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
266{
267 TAPState *s = DO_UPCAST(TAPState, nc, nc);
268
f394b2e2 269 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
445d892f 270 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
fbbdbdde
YB
271 len == sizeof(struct virtio_net_hdr) ||
272 len == sizeof(struct virtio_net_hdr_v1_hash));
445d892f
MT
273
274 tap_fd_set_vnet_hdr_len(s->fd, len);
275 s->host_vnet_hdr_len = len;
276}
277
481c5232
AO
278static bool tap_get_using_vnet_hdr(NetClientState *nc)
279{
280 TAPState *s = DO_UPCAST(TAPState, nc, nc);
281
282 return s->using_vnet_hdr;
283}
284
3bac80d3 285static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
5281d757 286{
3e35ba93 287 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 288
f394b2e2 289 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
ef4252b1 290 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
5281d757
MM
291
292 s->using_vnet_hdr = using_vnet_hdr;
293}
294
c80cd6bb
GK
295static int tap_set_vnet_le(NetClientState *nc, bool is_le)
296{
297 TAPState *s = DO_UPCAST(TAPState, nc, nc);
298
299 return tap_fd_set_vnet_le(s->fd, is_le);
300}
301
302static int tap_set_vnet_be(NetClientState *nc, bool is_be)
303{
304 TAPState *s = DO_UPCAST(TAPState, nc, nc);
305
306 return tap_fd_set_vnet_be(s->fd, is_be);
307}
308
3bac80d3 309static void tap_set_offload(NetClientState *nc, int csum, int tso4,
2ab0ec31 310 int tso6, int ecn, int ufo, int uso4, int uso6)
5281d757 311{
3e35ba93 312 TAPState *s = DO_UPCAST(TAPState, nc, nc);
27a6375d
MT
313 if (s->fd < 0) {
314 return;
315 }
5281d757 316
2ab0ec31 317 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo, uso4, uso6);
5281d757
MM
318}
319
9e32ff32
MAL
320static void tap_exit_notify(Notifier *notifier, void *data)
321{
322 TAPState *s = container_of(notifier, TAPState, exit);
323 Error *err = NULL;
324
325 if (s->down_script[0]) {
326 launch_script(s->down_script, s->down_script_arg, s->fd, &err);
327 if (err) {
328 error_report_err(err);
329 }
330 }
331}
332
4e68f7a0 333static void tap_cleanup(NetClientState *nc)
5281d757 334{
3e35ba93 335 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 336
82b0d80e
MT
337 if (s->vhost_net) {
338 vhost_net_cleanup(s->vhost_net);
e6bcb1b6 339 g_free(s->vhost_net);
43849424 340 s->vhost_net = NULL;
82b0d80e
MT
341 }
342
3e35ba93 343 qemu_purge_queued_packets(nc);
5281d757 344
9e32ff32
MAL
345 tap_exit_notify(&s->exit, NULL);
346 qemu_remove_exit_notifier(&s->exit);
5281d757 347
ec45f083
JW
348 tap_read_poll(s, false);
349 tap_write_poll(s, false);
5281d757 350 close(s->fd);
27a6375d 351 s->fd = -1;
5281d757
MM
352}
353
4e68f7a0 354static void tap_poll(NetClientState *nc, bool enable)
ceb69615
MT
355{
356 TAPState *s = DO_UPCAST(TAPState, nc, nc);
357 tap_read_poll(s, enable);
358 tap_write_poll(s, enable);
359}
360
8f364e34
AM
361static bool tap_set_steering_ebpf(NetClientState *nc, int prog_fd)
362{
363 TAPState *s = DO_UPCAST(TAPState, nc, nc);
364 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
365
366 return tap_fd_set_steering_ebpf(s->fd, prog_fd) == 0;
367}
368
4e68f7a0 369int tap_get_fd(NetClientState *nc)
95d528a2
MT
370{
371 TAPState *s = DO_UPCAST(TAPState, nc, nc);
f394b2e2 372 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
95d528a2
MT
373 return s->fd;
374}
375
5281d757
MM
376/* fd support */
377
3e35ba93 378static NetClientInfo net_tap_info = {
f394b2e2 379 .type = NET_CLIENT_DRIVER_TAP,
3e35ba93
MM
380 .size = sizeof(TAPState),
381 .receive = tap_receive,
382 .receive_raw = tap_receive_raw,
383 .receive_iov = tap_receive_iov,
ceb69615 384 .poll = tap_poll,
3e35ba93 385 .cleanup = tap_cleanup,
2e753bcc
VM
386 .has_ufo = tap_has_ufo,
387 .has_vnet_hdr = tap_has_vnet_hdr,
388 .has_vnet_hdr_len = tap_has_vnet_hdr_len,
481c5232 389 .get_using_vnet_hdr = tap_get_using_vnet_hdr,
2e753bcc
VM
390 .using_vnet_hdr = tap_using_vnet_hdr,
391 .set_offload = tap_set_offload,
481c5232 392 .get_vnet_hdr_len = tap_get_vnet_hdr_len,
2e753bcc 393 .set_vnet_hdr_len = tap_set_vnet_hdr_len,
c80cd6bb
GK
394 .set_vnet_le = tap_set_vnet_le,
395 .set_vnet_be = tap_set_vnet_be,
8f364e34 396 .set_steering_ebpf = tap_set_steering_ebpf,
3e35ba93
MM
397};
398
4e68f7a0 399static TAPState *net_tap_fd_init(NetClientState *peer,
5281d757
MM
400 const char *model,
401 const char *name,
402 int fd,
403 int vnet_hdr)
404{
4e68f7a0 405 NetClientState *nc;
5281d757 406 TAPState *s;
5281d757 407
ab5f3f84 408 nc = qemu_new_net_client(&net_tap_info, peer, model, name);
3e35ba93
MM
409
410 s = DO_UPCAST(TAPState, nc, nc);
411
5281d757 412 s->fd = fd;
ef4252b1 413 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
ec45f083 414 s->using_vnet_hdr = false;
9c282718 415 s->has_ufo = tap_probe_has_ufo(s->fd);
16dbaf90 416 s->enabled = true;
2ab0ec31 417 tap_set_offload(&s->nc, 0, 0, 0, 0, 0, 0, 0);
58ddcd50
MT
418 /*
419 * Make sure host header length is set correctly in tap:
420 * it might have been modified by another instance of qemu.
421 */
422 if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
423 tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
424 }
ec45f083 425 tap_read_poll(s, true);
82b0d80e 426 s->vhost_net = NULL;
9e32ff32
MAL
427
428 s->exit.notify = tap_exit_notify;
429 qemu_add_exit_notifier(&s->exit);
430
5281d757
MM
431 return s;
432}
433
ac4fcf56
MA
434static void launch_script(const char *setup_script, const char *ifname,
435 int fd, Error **errp)
5281d757 436{
5281d757
MM
437 int pid, status;
438 char *args[3];
439 char **parg;
440
5281d757
MM
441 /* try to launch network script */
442 pid = fork();
ac4fcf56
MA
443 if (pid < 0) {
444 error_setg_errno(errp, errno, "could not launch network script %s",
445 setup_script);
446 return;
447 }
5281d757
MM
448 if (pid == 0) {
449 int open_max = sysconf(_SC_OPEN_MAX), i;
450
13a12f86
PG
451 for (i = 3; i < open_max; i++) {
452 if (i != fd) {
5281d757
MM
453 close(i);
454 }
455 }
456 parg = args;
457 *parg++ = (char *)setup_script;
458 *parg++ = (char *)ifname;
9678d950 459 *parg = NULL;
5281d757
MM
460 execv(setup_script, args);
461 _exit(1);
ac4fcf56 462 } else {
5281d757
MM
463 while (waitpid(pid, &status, 0) != pid) {
464 /* loop */
465 }
5281d757
MM
466
467 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
ac4fcf56 468 return;
5281d757 469 }
ac4fcf56
MA
470 error_setg(errp, "network script %s failed with status %d",
471 setup_script, status);
5281d757 472 }
5281d757
MM
473}
474
a7c36ee4
CB
475static int recv_fd(int c)
476{
477 int fd;
478 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
479 struct msghdr msg = {
480 .msg_control = msgbuf,
481 .msg_controllen = sizeof(msgbuf),
482 };
483 struct cmsghdr *cmsg;
484 struct iovec iov;
485 uint8_t req[1];
486 ssize_t len;
487
488 cmsg = CMSG_FIRSTHDR(&msg);
489 cmsg->cmsg_level = SOL_SOCKET;
490 cmsg->cmsg_type = SCM_RIGHTS;
491 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
492 msg.msg_controllen = cmsg->cmsg_len;
493
494 iov.iov_base = req;
495 iov.iov_len = sizeof(req);
496
497 msg.msg_iov = &iov;
498 msg.msg_iovlen = 1;
499
500 len = recvmsg(c, &msg, 0);
501 if (len > 0) {
502 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
503 return fd;
504 }
505
506 return len;
507}
508
a8a21be9
MA
509static int net_bridge_run_helper(const char *helper, const char *bridge,
510 Error **errp)
a7c36ee4
CB
511{
512 sigset_t oldmask, mask;
63c4db4c 513 g_autofree char *default_helper = NULL;
a7c36ee4
CB
514 int pid, status;
515 char *args[5];
516 char **parg;
517 int sv[2];
518
519 sigemptyset(&mask);
520 sigaddset(&mask, SIGCHLD);
521 sigprocmask(SIG_BLOCK, &mask, &oldmask);
522
63c4db4c
PB
523 if (!helper) {
524 helper = default_helper = get_relocated_path(DEFAULT_BRIDGE_HELPER);
525 }
526
a7c36ee4 527 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
a8a21be9 528 error_setg_errno(errp, errno, "socketpair() failed");
a7c36ee4
CB
529 return -1;
530 }
531
532 /* try to launch bridge helper */
533 pid = fork();
a8a21be9
MA
534 if (pid < 0) {
535 error_setg_errno(errp, errno, "Can't fork bridge helper");
536 return -1;
537 }
a7c36ee4
CB
538 if (pid == 0) {
539 int open_max = sysconf(_SC_OPEN_MAX), i;
389abe1d
PP
540 char *fd_buf = NULL;
541 char *br_buf = NULL;
542 char *helper_cmd = NULL;
a7c36ee4 543
13a12f86
PG
544 for (i = 3; i < open_max; i++) {
545 if (i != sv[1]) {
a7c36ee4
CB
546 close(i);
547 }
548 }
549
389abe1d 550 fd_buf = g_strdup_printf("%s%d", "--fd=", sv[1]);
a7c36ee4
CB
551
552 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
553 /* assume helper is a command */
554
555 if (strstr(helper, "--br=") == NULL) {
389abe1d 556 br_buf = g_strdup_printf("%s%s", "--br=", bridge);
a7c36ee4
CB
557 }
558
389abe1d
PP
559 helper_cmd = g_strdup_printf("%s %s %s %s", helper,
560 "--use-vnet", fd_buf, br_buf ? br_buf : "");
a7c36ee4
CB
561
562 parg = args;
563 *parg++ = (char *)"sh";
564 *parg++ = (char *)"-c";
565 *parg++ = helper_cmd;
566 *parg++ = NULL;
567
568 execv("/bin/sh", args);
389abe1d 569 g_free(helper_cmd);
a7c36ee4
CB
570 } else {
571 /* assume helper is just the executable path name */
572
389abe1d 573 br_buf = g_strdup_printf("%s%s", "--br=", bridge);
a7c36ee4
CB
574
575 parg = args;
576 *parg++ = (char *)helper;
577 *parg++ = (char *)"--use-vnet";
578 *parg++ = fd_buf;
579 *parg++ = br_buf;
580 *parg++ = NULL;
581
582 execv(helper, args);
583 }
389abe1d
PP
584 g_free(fd_buf);
585 g_free(br_buf);
a7c36ee4
CB
586 _exit(1);
587
a8a21be9 588 } else {
a7c36ee4 589 int fd;
a8a21be9 590 int saved_errno;
a7c36ee4
CB
591
592 close(sv[1]);
593
37b0b24e 594 fd = RETRY_ON_EINTR(recv_fd(sv[0]));
a8a21be9 595 saved_errno = errno;
a7c36ee4
CB
596
597 close(sv[0]);
598
599 while (waitpid(pid, &status, 0) != pid) {
600 /* loop */
601 }
602 sigprocmask(SIG_SETMASK, &oldmask, NULL);
603 if (fd < 0) {
a8a21be9
MA
604 error_setg_errno(errp, saved_errno,
605 "failed to recv file descriptor");
a7c36ee4
CB
606 return -1;
607 }
a8a21be9
MA
608 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
609 error_setg(errp, "bridge helper failed");
610 return -1;
a7c36ee4 611 }
a8a21be9 612 return fd;
a7c36ee4 613 }
a7c36ee4
CB
614}
615
cebea510 616int net_init_bridge(const Netdev *netdev, const char *name,
a30ecde6 617 NetClientState *peer, Error **errp)
a7c36ee4 618{
f79b51b0
LE
619 const NetdevBridgeOptions *bridge;
620 const char *helper, *br;
a7c36ee4
CB
621 TAPState *s;
622 int fd, vnet_hdr;
623
f394b2e2
EB
624 assert(netdev->type == NET_CLIENT_DRIVER_BRIDGE);
625 bridge = &netdev->u.bridge;
7480874a
MA
626 helper = bridge->helper;
627 br = bridge->br ?: DEFAULT_BRIDGE_INTERFACE;
a7c36ee4 628
a8a21be9 629 fd = net_bridge_run_helper(helper, br, errp);
a7c36ee4
CB
630 if (fd == -1) {
631 return -1;
632 }
633
a8208626
MAL
634 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
635 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
636 return -1;
637 }
e7b347d0
DB
638 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
639 if (vnet_hdr < 0) {
640 close(fd);
641 return -1;
642 }
d33d93b2 643 s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
a7c36ee4 644
53b85d95 645 qemu_set_info_str(&s->nc, "helper=%s,br=%s", helper, br);
d89b4f83 646
a7c36ee4
CB
647 return 0;
648}
649
08c573a8
LE
650static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
651 const char *setup_script, char *ifname,
468dd824 652 size_t ifname_sz, int mq_required, Error **errp)
5281d757 653{
ac4fcf56 654 Error *err = NULL;
5281d757 655 int fd, vnet_hdr_required;
5281d757 656
08c573a8
LE
657 if (tap->has_vnet_hdr) {
658 *vnet_hdr = tap->vnet_hdr;
5281d757
MM
659 vnet_hdr_required = *vnet_hdr;
660 } else {
08c573a8 661 *vnet_hdr = 1;
5281d757
MM
662 vnet_hdr_required = 0;
663 }
664
8b6aa693 665 fd = RETRY_ON_EINTR(tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
468dd824 666 mq_required, errp));
5281d757
MM
667 if (fd < 0) {
668 return -1;
669 }
670
5281d757
MM
671 if (setup_script &&
672 setup_script[0] != '\0' &&
ac4fcf56
MA
673 strcmp(setup_script, "no") != 0) {
674 launch_script(setup_script, ifname, fd, &err);
675 if (err) {
468dd824 676 error_propagate(errp, err);
ac4fcf56
MA
677 close(fd);
678 return -1;
679 }
5281d757
MM
680 }
681
5281d757
MM
682 return fd;
683}
684
264986e2
JW
685#define MAX_TAP_QUEUES 1024
686
445f116c
MA
687static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
688 const char *model, const char *name,
689 const char *ifname, const char *script,
690 const char *downscript, const char *vhostfdname,
f9bb0c1f 691 int vnet_hdr, int fd, Error **errp)
5193e5fb 692{
1677f4c6 693 Error *err = NULL;
da4a4eac 694 TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
81647a65 695 int vhostfd;
5193e5fb 696
80b832c3
MA
697 tap_set_sndbuf(s->fd, tap, &err);
698 if (err) {
445f116c 699 error_propagate(errp, err);
bf769f74 700 goto failed;
5193e5fb
JW
701 }
702
7480874a 703 if (tap->fd || tap->fds) {
53b85d95 704 qemu_set_info_str(&s->nc, "fd=%d", fd);
7480874a 705 } else if (tap->helper) {
53b85d95 706 qemu_set_info_str(&s->nc, "helper=%s", tap->helper);
5193e5fb 707 } else {
53b85d95
LV
708 qemu_set_info_str(&s->nc, "ifname=%s,script=%s,downscript=%s", ifname,
709 script, downscript);
d89b4f83 710
5193e5fb
JW
711 if (strcmp(downscript, "no") != 0) {
712 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
713 snprintf(s->down_script_arg, sizeof(s->down_script_arg),
714 "%s", ifname);
715 }
716 }
717
718 if (tap->has_vhost ? tap->vhost :
719 vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
81647a65
NN
720 VhostNetOptions options;
721
1a1bfac9 722 options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
81647a65 723 options.net_backend = &s->nc;
69e87b32
JW
724 if (tap->has_poll_us) {
725 options.busyloop_timeout = tap->poll_us;
726 } else {
727 options.busyloop_timeout = 0;
728 }
5193e5fb 729
3a2d44f6 730 if (vhostfdname) {
947e4744 731 vhostfd = monitor_fd_param(monitor_cur(), vhostfdname, &err);
5193e5fb 732 if (vhostfd == -1) {
46d4d36d
JZ
733 if (tap->has_vhostforce && tap->vhostforce) {
734 error_propagate(errp, err);
735 } else {
736 warn_report_err(err);
737 }
bf769f74 738 goto failed;
5193e5fb 739 }
a8208626
MAL
740 if (!g_unix_set_fd_nonblocking(vhostfd, true, NULL)) {
741 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6 742 name, fd);
bf769f74 743 goto failed;
894022e6 744 }
5193e5fb 745 } else {
81647a65
NN
746 vhostfd = open("/dev/vhost-net", O_RDWR);
747 if (vhostfd < 0) {
46d4d36d
JZ
748 if (tap->has_vhostforce && tap->vhostforce) {
749 error_setg_errno(errp, errno,
750 "tap: open vhost char device failed");
751 } else {
752 warn_report("tap: open vhost char device failed: %s",
753 strerror(errno));
754 }
bf769f74 755 goto failed;
81647a65 756 }
a8208626
MAL
757 if (!g_unix_set_fd_nonblocking(vhostfd, true, NULL)) {
758 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
bf769f74 759 goto failed;
a8208626 760 }
5193e5fb 761 }
81647a65 762 options.opaque = (void *)(uintptr_t)vhostfd;
6a756d14 763 options.nvqs = 2;
5193e5fb 764
81647a65 765 s->vhost_net = vhost_net_init(&options);
5193e5fb 766 if (!s->vhost_net) {
46d4d36d
JZ
767 if (tap->has_vhostforce && tap->vhostforce) {
768 error_setg(errp, VHOST_NET_INIT_FAILED);
769 } else {
770 warn_report(VHOST_NET_INIT_FAILED);
771 }
bf769f74 772 goto failed;
5193e5fb 773 }
3a2d44f6 774 } else if (vhostfdname) {
69e87b32 775 error_setg(errp, "vhostfd(s)= is not valid without vhost");
bf769f74 776 goto failed;
5193e5fb 777 }
bf769f74 778
779 return;
780
781failed:
782 qemu_del_net_client(&s->nc);
5193e5fb
JW
783}
784
264986e2
JW
785static int get_fds(char *str, char *fds[], int max)
786{
787 char *ptr = str, *this;
788 size_t len = strlen(str);
789 int i = 0;
790
791 while (i < max && ptr < str + len) {
792 this = strchr(ptr, ':');
793
794 if (this == NULL) {
795 fds[i] = g_strdup(ptr);
796 } else {
797 fds[i] = g_strndup(ptr, this - ptr);
798 }
799
800 i++;
801 if (this == NULL) {
802 break;
803 } else {
804 ptr = this + 1;
805 }
806 }
807
808 return i;
809}
810
cebea510 811int net_init_tap(const Netdev *netdev, const char *name,
a30ecde6 812 NetClientState *peer, Error **errp)
5281d757 813{
08c573a8 814 const NetdevTapOptions *tap;
264986e2 815 int fd, vnet_hdr = 0, i = 0, queues;
08c573a8 816 /* for the no-fd, no-helper case */
63c4db4c
PB
817 const char *script;
818 const char *downscript;
1677f4c6 819 Error *err = NULL;
264986e2 820 const char *vhostfdname;
08c573a8 821 char ifname[128];
894022e6 822 int ret = 0;
08c573a8 823
f394b2e2
EB
824 assert(netdev->type == NET_CLIENT_DRIVER_TAP);
825 tap = &netdev->u.tap;
264986e2 826 queues = tap->has_queues ? tap->queues : 1;
7480874a
MA
827 vhostfdname = tap->vhostfd;
828 script = tap->script;
829 downscript = tap->downscript;
5281d757 830
442da403 831 /* QEMU hubs do not support multiqueue tap, in this case peer is set.
ce675a75 832 * For -netdev, peer is always NULL. */
7480874a 833 if (peer && (tap->has_queues || tap->fds || tap->vhostfds)) {
442da403 834 error_setg(errp, "Multiqueue tap cannot be used with hubs");
ce675a75
JW
835 return -1;
836 }
837
7480874a
MA
838 if (tap->fd) {
839 if (tap->ifname || tap->script || tap->downscript ||
840 tap->has_vnet_hdr || tap->helper || tap->has_queues ||
841 tap->fds || tap->vhostfds) {
a3088177
MA
842 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
843 "helper=, queues=, fds=, and vhostfds= "
844 "are invalid with fd=");
5281d757
MM
845 return -1;
846 }
847
947e4744 848 fd = monitor_fd_param(monitor_cur(), tap->fd, errp);
5281d757
MM
849 if (fd == -1) {
850 return -1;
851 }
852
a8208626
MAL
853 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
854 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6 855 name, fd);
f012bec8 856 close(fd);
894022e6
LV
857 return -1;
858 }
5281d757 859
e7b347d0
DB
860 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
861 if (vnet_hdr < 0) {
862 close(fd);
863 return -1;
864 }
a7c36ee4 865
445f116c
MA
866 net_init_tap_one(tap, peer, "tap", name, NULL,
867 script, downscript,
f9bb0c1f 868 vhostfdname, vnet_hdr, fd, &err);
445f116c 869 if (err) {
a3088177 870 error_propagate(errp, err);
f012bec8 871 close(fd);
264986e2
JW
872 return -1;
873 }
7480874a 874 } else if (tap->fds) {
fac7d7b1
PM
875 char **fds;
876 char **vhost_fds;
323e7c11 877 int nfds = 0, nvhosts = 0;
264986e2 878
7480874a
MA
879 if (tap->ifname || tap->script || tap->downscript ||
880 tap->has_vnet_hdr || tap->helper || tap->has_queues ||
881 tap->vhostfd) {
a3088177
MA
882 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
883 "helper=, queues=, and vhostfd= "
884 "are invalid with fds=");
264986e2
JW
885 return -1;
886 }
887
fac7d7b1
PM
888 fds = g_new0(char *, MAX_TAP_QUEUES);
889 vhost_fds = g_new0(char *, MAX_TAP_QUEUES);
890
264986e2 891 nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
7480874a 892 if (tap->vhostfds) {
264986e2
JW
893 nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
894 if (nfds != nvhosts) {
a3088177
MA
895 error_setg(errp, "The number of fds passed does not match "
896 "the number of vhostfds passed");
323e7c11 897 ret = -1;
091a6b2a 898 goto free_fail;
264986e2
JW
899 }
900 }
901
902 for (i = 0; i < nfds; i++) {
947e4744 903 fd = monitor_fd_param(monitor_cur(), fds[i], errp);
264986e2 904 if (fd == -1) {
323e7c11 905 ret = -1;
091a6b2a 906 goto free_fail;
264986e2
JW
907 }
908
a8208626
MAL
909 ret = g_unix_set_fd_nonblocking(fd, true, NULL);
910 if (!ret) {
911 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6
LV
912 name, fd);
913 goto free_fail;
914 }
a7c36ee4 915
264986e2 916 if (i == 0) {
e7b347d0
DB
917 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
918 if (vnet_hdr < 0) {
41bcea7b 919 ret = -1;
e7b347d0
DB
920 goto free_fail;
921 }
922 } else if (vnet_hdr != tap_probe_vnet_hdr(fd, NULL)) {
a3088177
MA
923 error_setg(errp,
924 "vnet_hdr not consistent across given tap fds");
323e7c11 925 ret = -1;
091a6b2a 926 goto free_fail;
264986e2
JW
927 }
928
445f116c
MA
929 net_init_tap_one(tap, peer, "tap", name, ifname,
930 script, downscript,
7480874a 931 tap->vhostfds ? vhost_fds[i] : NULL,
f9bb0c1f 932 vnet_hdr, fd, &err);
445f116c 933 if (err) {
a3088177 934 error_propagate(errp, err);
323e7c11 935 ret = -1;
091a6b2a 936 goto free_fail;
264986e2
JW
937 }
938 }
091a6b2a
PB
939
940free_fail:
323e7c11
YW
941 for (i = 0; i < nvhosts; i++) {
942 g_free(vhost_fds[i]);
943 }
091a6b2a
PB
944 for (i = 0; i < nfds; i++) {
945 g_free(fds[i]);
091a6b2a
PB
946 }
947 g_free(fds);
948 g_free(vhost_fds);
323e7c11 949 return ret;
7480874a
MA
950 } else if (tap->helper) {
951 if (tap->ifname || tap->script || tap->downscript ||
952 tap->has_vnet_hdr || tap->has_queues || tap->vhostfds) {
a3088177
MA
953 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
954 "queues=, and vhostfds= are invalid with helper=");
a7c36ee4
CB
955 return -1;
956 }
957
584613ea 958 fd = net_bridge_run_helper(tap->helper,
7480874a 959 tap->br ?: DEFAULT_BRIDGE_INTERFACE,
a8a21be9 960 errp);
a7c36ee4
CB
961 if (fd == -1) {
962 return -1;
963 }
964
a8208626
MAL
965 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
966 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
967 return -1;
968 }
e7b347d0
DB
969 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
970 if (vnet_hdr < 0) {
971 close(fd);
972 return -1;
973 }
a7c36ee4 974
445f116c
MA
975 net_init_tap_one(tap, peer, "bridge", name, ifname,
976 script, downscript, vhostfdname,
f9bb0c1f 977 vnet_hdr, fd, &err);
445f116c 978 if (err) {
a3088177 979 error_propagate(errp, err);
84f8f3da 980 close(fd);
264986e2
JW
981 return -1;
982 }
5281d757 983 } else {
63c4db4c
PB
984 g_autofree char *default_script = NULL;
985 g_autofree char *default_downscript = NULL;
7480874a 986 if (tap->vhostfds) {
a3088177 987 error_setg(errp, "vhostfds= is invalid if fds= wasn't specified");
c87826a8
JW
988 return -1;
989 }
63c4db4c
PB
990
991 if (!script) {
992 script = default_script = get_relocated_path(DEFAULT_NETWORK_SCRIPT);
993 }
994 if (!downscript) {
9925990d
KZ
995 downscript = default_downscript =
996 get_relocated_path(DEFAULT_NETWORK_DOWN_SCRIPT);
63c4db4c 997 }
264986e2 998
7480874a 999 if (tap->ifname) {
264986e2
JW
1000 pstrcpy(ifname, sizeof ifname, tap->ifname);
1001 } else {
1002 ifname[0] = '\0';
929fe497 1003 }
a7c36ee4 1004
264986e2
JW
1005 for (i = 0; i < queues; i++) {
1006 fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
a3088177 1007 ifname, sizeof ifname, queues > 1, errp);
264986e2
JW
1008 if (fd == -1) {
1009 return -1;
1010 }
1011
7480874a 1012 if (queues > 1 && i == 0 && !tap->ifname) {
264986e2 1013 if (tap_fd_get_ifname(fd, ifname)) {
a3088177 1014 error_setg(errp, "Fail to get ifname");
84f8f3da 1015 close(fd);
264986e2
JW
1016 return -1;
1017 }
1018 }
1019
445f116c
MA
1020 net_init_tap_one(tap, peer, "tap", name, ifname,
1021 i >= 1 ? "no" : script,
1022 i >= 1 ? "no" : downscript,
f9bb0c1f 1023 vhostfdname, vnet_hdr, fd, &err);
445f116c 1024 if (err) {
a3088177 1025 error_propagate(errp, err);
84f8f3da 1026 close(fd);
264986e2
JW
1027 return -1;
1028 }
1029 }
5281d757
MM
1030 }
1031
264986e2 1032 return 0;
5281d757 1033}
b202554c 1034
4e68f7a0 1035VHostNetState *tap_get_vhost_net(NetClientState *nc)
b202554c
MT
1036{
1037 TAPState *s = DO_UPCAST(TAPState, nc, nc);
f394b2e2 1038 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
b202554c
MT
1039 return s->vhost_net;
1040}
16dbaf90
JW
1041
1042int tap_enable(NetClientState *nc)
1043{
1044 TAPState *s = DO_UPCAST(TAPState, nc, nc);
1045 int ret;
1046
1047 if (s->enabled) {
1048 return 0;
1049 } else {
1050 ret = tap_fd_enable(s->fd);
1051 if (ret == 0) {
1052 s->enabled = true;
1053 tap_update_fd_handler(s);
1054 }
1055 return ret;
1056 }
1057}
1058
1059int tap_disable(NetClientState *nc)
1060{
1061 TAPState *s = DO_UPCAST(TAPState, nc, nc);
1062 int ret;
1063
1064 if (s->enabled == 0) {
1065 return 0;
1066 } else {
1067 ret = tap_fd_disable(s->fd);
1068 if (ret == 0) {
1069 qemu_purge_queued_packets(nc);
1070 s->enabled = false;
1071 tap_update_fd_handler(s);
1072 }
1073 return ret;
1074 }
1075}