]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap.c
vhost: Track number of descs in SVQDescState
[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
105 do {
106 len = writev(s->fd, iov, iovcnt);
107 } while (len == -1 && errno == EINTR);
108
109 if (len == -1 && errno == EAGAIN) {
ec45f083 110 tap_write_poll(s, true);
5281d757
MM
111 return 0;
112 }
113
114 return len;
115}
116
4e68f7a0 117static ssize_t tap_receive_iov(NetClientState *nc, const struct iovec *iov,
5281d757
MM
118 int iovcnt)
119{
3e35ba93 120 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
121 const struct iovec *iovp = iov;
122 struct iovec iov_copy[iovcnt + 1];
ef4252b1 123 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 124
ef4252b1 125 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
5281d757 126 iov_copy[0].iov_base = &hdr;
ef4252b1 127 iov_copy[0].iov_len = s->host_vnet_hdr_len;
5281d757
MM
128 memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
129 iovp = iov_copy;
130 iovcnt++;
131 }
132
133 return tap_write_packet(s, iovp, iovcnt);
134}
135
4e68f7a0 136static ssize_t tap_receive_raw(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 137{
3e35ba93 138 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
139 struct iovec iov[2];
140 int iovcnt = 0;
ef4252b1 141 struct virtio_net_hdr_mrg_rxbuf hdr = { };
5281d757 142
ef4252b1 143 if (s->host_vnet_hdr_len) {
5281d757 144 iov[iovcnt].iov_base = &hdr;
ef4252b1 145 iov[iovcnt].iov_len = s->host_vnet_hdr_len;
5281d757
MM
146 iovcnt++;
147 }
148
149 iov[iovcnt].iov_base = (char *)buf;
150 iov[iovcnt].iov_len = size;
151 iovcnt++;
152
153 return tap_write_packet(s, iov, iovcnt);
154}
155
4e68f7a0 156static ssize_t tap_receive(NetClientState *nc, const uint8_t *buf, size_t size)
5281d757 157{
3e35ba93 158 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757
MM
159 struct iovec iov[1];
160
ef4252b1 161 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
3e35ba93 162 return tap_receive_raw(nc, buf, size);
5281d757
MM
163 }
164
165 iov[0].iov_base = (char *)buf;
166 iov[0].iov_len = size;
167
168 return tap_write_packet(s, iov, 1);
169}
170
966ea5ec
MM
171#ifndef __sun__
172ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
5281d757
MM
173{
174 return read(tapfd, buf, maxlen);
175}
176#endif
177
4e68f7a0 178static void tap_send_completed(NetClientState *nc, ssize_t len)
5281d757 179{
3e35ba93 180 TAPState *s = DO_UPCAST(TAPState, nc, nc);
ec45f083 181 tap_read_poll(s, true);
5281d757
MM
182}
183
184static void tap_send(void *opaque)
185{
186 TAPState *s = opaque;
187 int size;
756ae78b 188 int packets = 0;
5281d757 189
a90a7425 190 while (true) {
5819c918 191 uint8_t *buf = s->buf;
969e50b6
BM
192 uint8_t min_pkt[ETH_ZLEN];
193 size_t min_pktsz = sizeof(min_pkt);
5819c918
MM
194
195 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
196 if (size <= 0) {
197 break;
198 }
199
ef4252b1
MT
200 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
201 buf += s->host_vnet_hdr_len;
202 size -= s->host_vnet_hdr_len;
5819c918
MM
203 }
204
bc38e31b 205 if (net_peer_needs_padding(&s->nc)) {
969e50b6
BM
206 if (eth_pad_short_frame(min_pkt, &min_pktsz, buf, size)) {
207 buf = min_pkt;
208 size = min_pktsz;
209 }
210 }
211
3e35ba93 212 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
5819c918 213 if (size == 0) {
ec45f083 214 tap_read_poll(s, false);
68e5ec64
SH
215 break;
216 } else if (size < 0) {
217 break;
5819c918 218 }
756ae78b
WK
219
220 /*
221 * When the host keeps receiving more packets while tap_send() is
222 * running we can hog the QEMU global mutex. Limit the number of
223 * packets that are processed per tap_send() callback to prevent
224 * stalling the guest.
225 */
226 packets++;
227 if (packets >= 50) {
228 break;
229 }
68e5ec64 230 }
5281d757
MM
231}
232
3bac80d3 233static bool tap_has_ufo(NetClientState *nc)
5281d757 234{
3e35ba93 235 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 236
f394b2e2 237 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
5281d757
MM
238
239 return s->has_ufo;
240}
241
3bac80d3 242static bool tap_has_vnet_hdr(NetClientState *nc)
5281d757 243{
3e35ba93 244 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 245
f394b2e2 246 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
5281d757 247
ef4252b1 248 return !!s->host_vnet_hdr_len;
5281d757
MM
249}
250
3bac80d3 251static bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
252{
253 TAPState *s = DO_UPCAST(TAPState, nc, nc);
254
f394b2e2 255 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
445d892f 256
e96dfd11 257 return !!tap_probe_vnet_hdr_len(s->fd, len);
445d892f
MT
258}
259
3bac80d3 260static void tap_set_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
261{
262 TAPState *s = DO_UPCAST(TAPState, nc, nc);
263
f394b2e2 264 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
445d892f 265 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
fbbdbdde
YB
266 len == sizeof(struct virtio_net_hdr) ||
267 len == sizeof(struct virtio_net_hdr_v1_hash));
445d892f
MT
268
269 tap_fd_set_vnet_hdr_len(s->fd, len);
270 s->host_vnet_hdr_len = len;
271}
272
3bac80d3 273static void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
5281d757 274{
3e35ba93 275 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 276
f394b2e2 277 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
ef4252b1 278 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
5281d757
MM
279
280 s->using_vnet_hdr = using_vnet_hdr;
281}
282
c80cd6bb
GK
283static int tap_set_vnet_le(NetClientState *nc, bool is_le)
284{
285 TAPState *s = DO_UPCAST(TAPState, nc, nc);
286
287 return tap_fd_set_vnet_le(s->fd, is_le);
288}
289
290static int tap_set_vnet_be(NetClientState *nc, bool is_be)
291{
292 TAPState *s = DO_UPCAST(TAPState, nc, nc);
293
294 return tap_fd_set_vnet_be(s->fd, is_be);
295}
296
3bac80d3 297static void tap_set_offload(NetClientState *nc, int csum, int tso4,
5281d757
MM
298 int tso6, int ecn, int ufo)
299{
3e35ba93 300 TAPState *s = DO_UPCAST(TAPState, nc, nc);
27a6375d
MT
301 if (s->fd < 0) {
302 return;
303 }
5281d757 304
27a6375d 305 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
5281d757
MM
306}
307
9e32ff32
MAL
308static void tap_exit_notify(Notifier *notifier, void *data)
309{
310 TAPState *s = container_of(notifier, TAPState, exit);
311 Error *err = NULL;
312
313 if (s->down_script[0]) {
314 launch_script(s->down_script, s->down_script_arg, s->fd, &err);
315 if (err) {
316 error_report_err(err);
317 }
318 }
319}
320
4e68f7a0 321static void tap_cleanup(NetClientState *nc)
5281d757 322{
3e35ba93 323 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 324
82b0d80e
MT
325 if (s->vhost_net) {
326 vhost_net_cleanup(s->vhost_net);
e6bcb1b6 327 g_free(s->vhost_net);
43849424 328 s->vhost_net = NULL;
82b0d80e
MT
329 }
330
3e35ba93 331 qemu_purge_queued_packets(nc);
5281d757 332
9e32ff32
MAL
333 tap_exit_notify(&s->exit, NULL);
334 qemu_remove_exit_notifier(&s->exit);
5281d757 335
ec45f083
JW
336 tap_read_poll(s, false);
337 tap_write_poll(s, false);
5281d757 338 close(s->fd);
27a6375d 339 s->fd = -1;
5281d757
MM
340}
341
4e68f7a0 342static void tap_poll(NetClientState *nc, bool enable)
ceb69615
MT
343{
344 TAPState *s = DO_UPCAST(TAPState, nc, nc);
345 tap_read_poll(s, enable);
346 tap_write_poll(s, enable);
347}
348
8f364e34
AM
349static bool tap_set_steering_ebpf(NetClientState *nc, int prog_fd)
350{
351 TAPState *s = DO_UPCAST(TAPState, nc, nc);
352 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
353
354 return tap_fd_set_steering_ebpf(s->fd, prog_fd) == 0;
355}
356
4e68f7a0 357int tap_get_fd(NetClientState *nc)
95d528a2
MT
358{
359 TAPState *s = DO_UPCAST(TAPState, nc, nc);
f394b2e2 360 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
95d528a2
MT
361 return s->fd;
362}
363
5281d757
MM
364/* fd support */
365
3e35ba93 366static NetClientInfo net_tap_info = {
f394b2e2 367 .type = NET_CLIENT_DRIVER_TAP,
3e35ba93
MM
368 .size = sizeof(TAPState),
369 .receive = tap_receive,
370 .receive_raw = tap_receive_raw,
371 .receive_iov = tap_receive_iov,
ceb69615 372 .poll = tap_poll,
3e35ba93 373 .cleanup = tap_cleanup,
2e753bcc
VM
374 .has_ufo = tap_has_ufo,
375 .has_vnet_hdr = tap_has_vnet_hdr,
376 .has_vnet_hdr_len = tap_has_vnet_hdr_len,
377 .using_vnet_hdr = tap_using_vnet_hdr,
378 .set_offload = tap_set_offload,
379 .set_vnet_hdr_len = tap_set_vnet_hdr_len,
c80cd6bb
GK
380 .set_vnet_le = tap_set_vnet_le,
381 .set_vnet_be = tap_set_vnet_be,
8f364e34 382 .set_steering_ebpf = tap_set_steering_ebpf,
3e35ba93
MM
383};
384
4e68f7a0 385static TAPState *net_tap_fd_init(NetClientState *peer,
5281d757
MM
386 const char *model,
387 const char *name,
388 int fd,
389 int vnet_hdr)
390{
4e68f7a0 391 NetClientState *nc;
5281d757 392 TAPState *s;
5281d757 393
ab5f3f84 394 nc = qemu_new_net_client(&net_tap_info, peer, model, name);
3e35ba93
MM
395
396 s = DO_UPCAST(TAPState, nc, nc);
397
5281d757 398 s->fd = fd;
ef4252b1 399 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
ec45f083 400 s->using_vnet_hdr = false;
9c282718 401 s->has_ufo = tap_probe_has_ufo(s->fd);
16dbaf90 402 s->enabled = true;
3e35ba93 403 tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
58ddcd50
MT
404 /*
405 * Make sure host header length is set correctly in tap:
406 * it might have been modified by another instance of qemu.
407 */
408 if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
409 tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
410 }
ec45f083 411 tap_read_poll(s, true);
82b0d80e 412 s->vhost_net = NULL;
9e32ff32
MAL
413
414 s->exit.notify = tap_exit_notify;
415 qemu_add_exit_notifier(&s->exit);
416
5281d757
MM
417 return s;
418}
419
ac4fcf56
MA
420static void launch_script(const char *setup_script, const char *ifname,
421 int fd, Error **errp)
5281d757 422{
5281d757
MM
423 int pid, status;
424 char *args[3];
425 char **parg;
426
5281d757
MM
427 /* try to launch network script */
428 pid = fork();
ac4fcf56
MA
429 if (pid < 0) {
430 error_setg_errno(errp, errno, "could not launch network script %s",
431 setup_script);
432 return;
433 }
5281d757
MM
434 if (pid == 0) {
435 int open_max = sysconf(_SC_OPEN_MAX), i;
436
13a12f86
PG
437 for (i = 3; i < open_max; i++) {
438 if (i != fd) {
5281d757
MM
439 close(i);
440 }
441 }
442 parg = args;
443 *parg++ = (char *)setup_script;
444 *parg++ = (char *)ifname;
9678d950 445 *parg = NULL;
5281d757
MM
446 execv(setup_script, args);
447 _exit(1);
ac4fcf56 448 } else {
5281d757
MM
449 while (waitpid(pid, &status, 0) != pid) {
450 /* loop */
451 }
5281d757
MM
452
453 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
ac4fcf56 454 return;
5281d757 455 }
ac4fcf56
MA
456 error_setg(errp, "network script %s failed with status %d",
457 setup_script, status);
5281d757 458 }
5281d757
MM
459}
460
a7c36ee4
CB
461static int recv_fd(int c)
462{
463 int fd;
464 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
465 struct msghdr msg = {
466 .msg_control = msgbuf,
467 .msg_controllen = sizeof(msgbuf),
468 };
469 struct cmsghdr *cmsg;
470 struct iovec iov;
471 uint8_t req[1];
472 ssize_t len;
473
474 cmsg = CMSG_FIRSTHDR(&msg);
475 cmsg->cmsg_level = SOL_SOCKET;
476 cmsg->cmsg_type = SCM_RIGHTS;
477 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
478 msg.msg_controllen = cmsg->cmsg_len;
479
480 iov.iov_base = req;
481 iov.iov_len = sizeof(req);
482
483 msg.msg_iov = &iov;
484 msg.msg_iovlen = 1;
485
486 len = recvmsg(c, &msg, 0);
487 if (len > 0) {
488 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
489 return fd;
490 }
491
492 return len;
493}
494
a8a21be9
MA
495static int net_bridge_run_helper(const char *helper, const char *bridge,
496 Error **errp)
a7c36ee4
CB
497{
498 sigset_t oldmask, mask;
63c4db4c 499 g_autofree char *default_helper = NULL;
a7c36ee4
CB
500 int pid, status;
501 char *args[5];
502 char **parg;
503 int sv[2];
504
505 sigemptyset(&mask);
506 sigaddset(&mask, SIGCHLD);
507 sigprocmask(SIG_BLOCK, &mask, &oldmask);
508
63c4db4c
PB
509 if (!helper) {
510 helper = default_helper = get_relocated_path(DEFAULT_BRIDGE_HELPER);
511 }
512
a7c36ee4 513 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
a8a21be9 514 error_setg_errno(errp, errno, "socketpair() failed");
a7c36ee4
CB
515 return -1;
516 }
517
518 /* try to launch bridge helper */
519 pid = fork();
a8a21be9
MA
520 if (pid < 0) {
521 error_setg_errno(errp, errno, "Can't fork bridge helper");
522 return -1;
523 }
a7c36ee4
CB
524 if (pid == 0) {
525 int open_max = sysconf(_SC_OPEN_MAX), i;
389abe1d
PP
526 char *fd_buf = NULL;
527 char *br_buf = NULL;
528 char *helper_cmd = NULL;
a7c36ee4 529
13a12f86
PG
530 for (i = 3; i < open_max; i++) {
531 if (i != sv[1]) {
a7c36ee4
CB
532 close(i);
533 }
534 }
535
389abe1d 536 fd_buf = g_strdup_printf("%s%d", "--fd=", sv[1]);
a7c36ee4
CB
537
538 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
539 /* assume helper is a command */
540
541 if (strstr(helper, "--br=") == NULL) {
389abe1d 542 br_buf = g_strdup_printf("%s%s", "--br=", bridge);
a7c36ee4
CB
543 }
544
389abe1d
PP
545 helper_cmd = g_strdup_printf("%s %s %s %s", helper,
546 "--use-vnet", fd_buf, br_buf ? br_buf : "");
a7c36ee4
CB
547
548 parg = args;
549 *parg++ = (char *)"sh";
550 *parg++ = (char *)"-c";
551 *parg++ = helper_cmd;
552 *parg++ = NULL;
553
554 execv("/bin/sh", args);
389abe1d 555 g_free(helper_cmd);
a7c36ee4
CB
556 } else {
557 /* assume helper is just the executable path name */
558
389abe1d 559 br_buf = g_strdup_printf("%s%s", "--br=", bridge);
a7c36ee4
CB
560
561 parg = args;
562 *parg++ = (char *)helper;
563 *parg++ = (char *)"--use-vnet";
564 *parg++ = fd_buf;
565 *parg++ = br_buf;
566 *parg++ = NULL;
567
568 execv(helper, args);
569 }
389abe1d
PP
570 g_free(fd_buf);
571 g_free(br_buf);
a7c36ee4
CB
572 _exit(1);
573
a8a21be9 574 } else {
a7c36ee4 575 int fd;
a8a21be9 576 int saved_errno;
a7c36ee4
CB
577
578 close(sv[1]);
579
580 do {
581 fd = recv_fd(sv[0]);
582 } while (fd == -1 && errno == EINTR);
a8a21be9 583 saved_errno = errno;
a7c36ee4
CB
584
585 close(sv[0]);
586
587 while (waitpid(pid, &status, 0) != pid) {
588 /* loop */
589 }
590 sigprocmask(SIG_SETMASK, &oldmask, NULL);
591 if (fd < 0) {
a8a21be9
MA
592 error_setg_errno(errp, saved_errno,
593 "failed to recv file descriptor");
a7c36ee4
CB
594 return -1;
595 }
a8a21be9
MA
596 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
597 error_setg(errp, "bridge helper failed");
598 return -1;
a7c36ee4 599 }
a8a21be9 600 return fd;
a7c36ee4 601 }
a7c36ee4
CB
602}
603
cebea510 604int net_init_bridge(const Netdev *netdev, const char *name,
a30ecde6 605 NetClientState *peer, Error **errp)
a7c36ee4 606{
f79b51b0
LE
607 const NetdevBridgeOptions *bridge;
608 const char *helper, *br;
a7c36ee4
CB
609 TAPState *s;
610 int fd, vnet_hdr;
611
f394b2e2
EB
612 assert(netdev->type == NET_CLIENT_DRIVER_BRIDGE);
613 bridge = &netdev->u.bridge;
63c4db4c 614 helper = bridge->has_helper ? bridge->helper : NULL;
f79b51b0 615 br = bridge->has_br ? bridge->br : DEFAULT_BRIDGE_INTERFACE;
a7c36ee4 616
a8a21be9 617 fd = net_bridge_run_helper(helper, br, errp);
a7c36ee4
CB
618 if (fd == -1) {
619 return -1;
620 }
621
a8208626
MAL
622 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
623 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
624 return -1;
625 }
e7b347d0
DB
626 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
627 if (vnet_hdr < 0) {
628 close(fd);
629 return -1;
630 }
d33d93b2 631 s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
a7c36ee4 632
56e6f594
JW
633 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
634 br);
d89b4f83 635
a7c36ee4
CB
636 return 0;
637}
638
08c573a8
LE
639static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
640 const char *setup_script, char *ifname,
468dd824 641 size_t ifname_sz, int mq_required, Error **errp)
5281d757 642{
ac4fcf56 643 Error *err = NULL;
5281d757 644 int fd, vnet_hdr_required;
5281d757 645
08c573a8
LE
646 if (tap->has_vnet_hdr) {
647 *vnet_hdr = tap->vnet_hdr;
5281d757
MM
648 vnet_hdr_required = *vnet_hdr;
649 } else {
08c573a8 650 *vnet_hdr = 1;
5281d757
MM
651 vnet_hdr_required = 0;
652 }
653
264986e2 654 TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
468dd824 655 mq_required, errp));
5281d757
MM
656 if (fd < 0) {
657 return -1;
658 }
659
5281d757
MM
660 if (setup_script &&
661 setup_script[0] != '\0' &&
ac4fcf56
MA
662 strcmp(setup_script, "no") != 0) {
663 launch_script(setup_script, ifname, fd, &err);
664 if (err) {
468dd824 665 error_propagate(errp, err);
ac4fcf56
MA
666 close(fd);
667 return -1;
668 }
5281d757
MM
669 }
670
5281d757
MM
671 return fd;
672}
673
264986e2
JW
674#define MAX_TAP_QUEUES 1024
675
445f116c
MA
676static void net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
677 const char *model, const char *name,
678 const char *ifname, const char *script,
679 const char *downscript, const char *vhostfdname,
f9bb0c1f 680 int vnet_hdr, int fd, Error **errp)
5193e5fb 681{
1677f4c6 682 Error *err = NULL;
da4a4eac 683 TAPState *s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
81647a65 684 int vhostfd;
5193e5fb 685
80b832c3
MA
686 tap_set_sndbuf(s->fd, tap, &err);
687 if (err) {
445f116c
MA
688 error_propagate(errp, err);
689 return;
5193e5fb
JW
690 }
691
264986e2 692 if (tap->has_fd || tap->has_fds) {
56e6f594 693 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
5193e5fb 694 } else if (tap->has_helper) {
56e6f594
JW
695 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
696 tap->helper);
5193e5fb 697 } else {
56e6f594
JW
698 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
699 "ifname=%s,script=%s,downscript=%s", ifname, script,
700 downscript);
d89b4f83 701
5193e5fb
JW
702 if (strcmp(downscript, "no") != 0) {
703 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
704 snprintf(s->down_script_arg, sizeof(s->down_script_arg),
705 "%s", ifname);
706 }
707 }
708
709 if (tap->has_vhost ? tap->vhost :
710 vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
81647a65
NN
711 VhostNetOptions options;
712
1a1bfac9 713 options.backend_type = VHOST_BACKEND_TYPE_KERNEL;
81647a65 714 options.net_backend = &s->nc;
69e87b32
JW
715 if (tap->has_poll_us) {
716 options.busyloop_timeout = tap->poll_us;
717 } else {
718 options.busyloop_timeout = 0;
719 }
5193e5fb 720
3a2d44f6 721 if (vhostfdname) {
947e4744 722 vhostfd = monitor_fd_param(monitor_cur(), vhostfdname, &err);
5193e5fb 723 if (vhostfd == -1) {
46d4d36d
JZ
724 if (tap->has_vhostforce && tap->vhostforce) {
725 error_propagate(errp, err);
726 } else {
727 warn_report_err(err);
728 }
445f116c 729 return;
5193e5fb 730 }
a8208626
MAL
731 if (!g_unix_set_fd_nonblocking(vhostfd, true, NULL)) {
732 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6
LV
733 name, fd);
734 return;
735 }
5193e5fb 736 } else {
81647a65
NN
737 vhostfd = open("/dev/vhost-net", O_RDWR);
738 if (vhostfd < 0) {
46d4d36d
JZ
739 if (tap->has_vhostforce && tap->vhostforce) {
740 error_setg_errno(errp, errno,
741 "tap: open vhost char device failed");
742 } else {
743 warn_report("tap: open vhost char device failed: %s",
744 strerror(errno));
745 }
445f116c 746 return;
81647a65 747 }
a8208626
MAL
748 if (!g_unix_set_fd_nonblocking(vhostfd, true, NULL)) {
749 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
750 return;
751 }
5193e5fb 752 }
81647a65 753 options.opaque = (void *)(uintptr_t)vhostfd;
6a756d14 754 options.nvqs = 2;
5193e5fb 755
81647a65 756 s->vhost_net = vhost_net_init(&options);
5193e5fb 757 if (!s->vhost_net) {
46d4d36d
JZ
758 if (tap->has_vhostforce && tap->vhostforce) {
759 error_setg(errp, VHOST_NET_INIT_FAILED);
760 } else {
761 warn_report(VHOST_NET_INIT_FAILED);
762 }
445f116c 763 return;
5193e5fb 764 }
3a2d44f6 765 } else if (vhostfdname) {
69e87b32 766 error_setg(errp, "vhostfd(s)= is not valid without vhost");
5193e5fb 767 }
5193e5fb
JW
768}
769
264986e2
JW
770static int get_fds(char *str, char *fds[], int max)
771{
772 char *ptr = str, *this;
773 size_t len = strlen(str);
774 int i = 0;
775
776 while (i < max && ptr < str + len) {
777 this = strchr(ptr, ':');
778
779 if (this == NULL) {
780 fds[i] = g_strdup(ptr);
781 } else {
782 fds[i] = g_strndup(ptr, this - ptr);
783 }
784
785 i++;
786 if (this == NULL) {
787 break;
788 } else {
789 ptr = this + 1;
790 }
791 }
792
793 return i;
794}
795
cebea510 796int net_init_tap(const Netdev *netdev, const char *name,
a30ecde6 797 NetClientState *peer, Error **errp)
5281d757 798{
08c573a8 799 const NetdevTapOptions *tap;
264986e2 800 int fd, vnet_hdr = 0, i = 0, queues;
08c573a8 801 /* for the no-fd, no-helper case */
63c4db4c
PB
802 const char *script;
803 const char *downscript;
1677f4c6 804 Error *err = NULL;
264986e2 805 const char *vhostfdname;
08c573a8 806 char ifname[128];
894022e6 807 int ret = 0;
08c573a8 808
f394b2e2
EB
809 assert(netdev->type == NET_CLIENT_DRIVER_TAP);
810 tap = &netdev->u.tap;
264986e2
JW
811 queues = tap->has_queues ? tap->queues : 1;
812 vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
63c4db4c
PB
813 script = tap->has_script ? tap->script : NULL;
814 downscript = tap->has_downscript ? tap->downscript : NULL;
5281d757 815
442da403 816 /* QEMU hubs do not support multiqueue tap, in this case peer is set.
ce675a75
JW
817 * For -netdev, peer is always NULL. */
818 if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
442da403 819 error_setg(errp, "Multiqueue tap cannot be used with hubs");
ce675a75
JW
820 return -1;
821 }
822
08c573a8
LE
823 if (tap->has_fd) {
824 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
264986e2 825 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 826 tap->has_fds || tap->has_vhostfds) {
a3088177
MA
827 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
828 "helper=, queues=, fds=, and vhostfds= "
829 "are invalid with fd=");
5281d757
MM
830 return -1;
831 }
832
947e4744 833 fd = monitor_fd_param(monitor_cur(), tap->fd, errp);
5281d757
MM
834 if (fd == -1) {
835 return -1;
836 }
837
a8208626
MAL
838 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
839 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6 840 name, fd);
f012bec8 841 close(fd);
894022e6
LV
842 return -1;
843 }
5281d757 844
e7b347d0
DB
845 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
846 if (vnet_hdr < 0) {
847 close(fd);
848 return -1;
849 }
a7c36ee4 850
445f116c
MA
851 net_init_tap_one(tap, peer, "tap", name, NULL,
852 script, downscript,
f9bb0c1f 853 vhostfdname, vnet_hdr, fd, &err);
445f116c 854 if (err) {
a3088177 855 error_propagate(errp, err);
f012bec8 856 close(fd);
264986e2
JW
857 return -1;
858 }
859 } else if (tap->has_fds) {
fac7d7b1
PM
860 char **fds;
861 char **vhost_fds;
323e7c11 862 int nfds = 0, nvhosts = 0;
264986e2
JW
863
864 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
865 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 866 tap->has_vhostfd) {
a3088177
MA
867 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
868 "helper=, queues=, and vhostfd= "
869 "are invalid with fds=");
264986e2
JW
870 return -1;
871 }
872
fac7d7b1
PM
873 fds = g_new0(char *, MAX_TAP_QUEUES);
874 vhost_fds = g_new0(char *, MAX_TAP_QUEUES);
875
264986e2
JW
876 nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
877 if (tap->has_vhostfds) {
878 nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
879 if (nfds != nvhosts) {
a3088177
MA
880 error_setg(errp, "The number of fds passed does not match "
881 "the number of vhostfds passed");
323e7c11 882 ret = -1;
091a6b2a 883 goto free_fail;
264986e2
JW
884 }
885 }
886
887 for (i = 0; i < nfds; i++) {
947e4744 888 fd = monitor_fd_param(monitor_cur(), fds[i], errp);
264986e2 889 if (fd == -1) {
323e7c11 890 ret = -1;
091a6b2a 891 goto free_fail;
264986e2
JW
892 }
893
a8208626
MAL
894 ret = g_unix_set_fd_nonblocking(fd, true, NULL);
895 if (!ret) {
896 error_setg_errno(errp, errno, "%s: Can't use file descriptor %d",
894022e6
LV
897 name, fd);
898 goto free_fail;
899 }
a7c36ee4 900
264986e2 901 if (i == 0) {
e7b347d0
DB
902 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
903 if (vnet_hdr < 0) {
41bcea7b 904 ret = -1;
e7b347d0
DB
905 goto free_fail;
906 }
907 } else if (vnet_hdr != tap_probe_vnet_hdr(fd, NULL)) {
a3088177
MA
908 error_setg(errp,
909 "vnet_hdr not consistent across given tap fds");
323e7c11 910 ret = -1;
091a6b2a 911 goto free_fail;
264986e2
JW
912 }
913
445f116c
MA
914 net_init_tap_one(tap, peer, "tap", name, ifname,
915 script, downscript,
916 tap->has_vhostfds ? vhost_fds[i] : NULL,
f9bb0c1f 917 vnet_hdr, fd, &err);
445f116c 918 if (err) {
a3088177 919 error_propagate(errp, err);
323e7c11 920 ret = -1;
091a6b2a 921 goto free_fail;
264986e2
JW
922 }
923 }
091a6b2a
PB
924
925free_fail:
323e7c11
YW
926 for (i = 0; i < nvhosts; i++) {
927 g_free(vhost_fds[i]);
928 }
091a6b2a
PB
929 for (i = 0; i < nfds; i++) {
930 g_free(fds[i]);
091a6b2a
PB
931 }
932 g_free(fds);
933 g_free(vhost_fds);
323e7c11 934 return ret;
08c573a8
LE
935 } else if (tap->has_helper) {
936 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
c87826a8 937 tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
a3088177
MA
938 error_setg(errp, "ifname=, script=, downscript=, vnet_hdr=, "
939 "queues=, and vhostfds= are invalid with helper=");
a7c36ee4
CB
940 return -1;
941 }
942
584613ea
AK
943 fd = net_bridge_run_helper(tap->helper,
944 tap->has_br ?
945 tap->br : DEFAULT_BRIDGE_INTERFACE,
a8a21be9 946 errp);
a7c36ee4
CB
947 if (fd == -1) {
948 return -1;
949 }
950
a8208626
MAL
951 if (!g_unix_set_fd_nonblocking(fd, true, NULL)) {
952 error_setg_errno(errp, errno, "Failed to set FD nonblocking");
953 return -1;
954 }
e7b347d0
DB
955 vnet_hdr = tap_probe_vnet_hdr(fd, errp);
956 if (vnet_hdr < 0) {
957 close(fd);
958 return -1;
959 }
a7c36ee4 960
445f116c
MA
961 net_init_tap_one(tap, peer, "bridge", name, ifname,
962 script, downscript, vhostfdname,
f9bb0c1f 963 vnet_hdr, fd, &err);
445f116c 964 if (err) {
a3088177 965 error_propagate(errp, err);
84f8f3da 966 close(fd);
264986e2
JW
967 return -1;
968 }
5281d757 969 } else {
63c4db4c
PB
970 g_autofree char *default_script = NULL;
971 g_autofree char *default_downscript = NULL;
c87826a8 972 if (tap->has_vhostfds) {
a3088177 973 error_setg(errp, "vhostfds= is invalid if fds= wasn't specified");
c87826a8
JW
974 return -1;
975 }
63c4db4c
PB
976
977 if (!script) {
978 script = default_script = get_relocated_path(DEFAULT_NETWORK_SCRIPT);
979 }
980 if (!downscript) {
9925990d
KZ
981 downscript = default_downscript =
982 get_relocated_path(DEFAULT_NETWORK_DOWN_SCRIPT);
63c4db4c 983 }
264986e2
JW
984
985 if (tap->has_ifname) {
986 pstrcpy(ifname, sizeof ifname, tap->ifname);
987 } else {
988 ifname[0] = '\0';
929fe497 989 }
a7c36ee4 990
264986e2
JW
991 for (i = 0; i < queues; i++) {
992 fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
a3088177 993 ifname, sizeof ifname, queues > 1, errp);
264986e2
JW
994 if (fd == -1) {
995 return -1;
996 }
997
998 if (queues > 1 && i == 0 && !tap->has_ifname) {
999 if (tap_fd_get_ifname(fd, ifname)) {
a3088177 1000 error_setg(errp, "Fail to get ifname");
84f8f3da 1001 close(fd);
264986e2
JW
1002 return -1;
1003 }
1004 }
1005
445f116c
MA
1006 net_init_tap_one(tap, peer, "tap", name, ifname,
1007 i >= 1 ? "no" : script,
1008 i >= 1 ? "no" : downscript,
f9bb0c1f 1009 vhostfdname, vnet_hdr, fd, &err);
445f116c 1010 if (err) {
a3088177 1011 error_propagate(errp, err);
84f8f3da 1012 close(fd);
264986e2
JW
1013 return -1;
1014 }
1015 }
5281d757
MM
1016 }
1017
264986e2 1018 return 0;
5281d757 1019}
b202554c 1020
4e68f7a0 1021VHostNetState *tap_get_vhost_net(NetClientState *nc)
b202554c
MT
1022{
1023 TAPState *s = DO_UPCAST(TAPState, nc, nc);
f394b2e2 1024 assert(nc->info->type == NET_CLIENT_DRIVER_TAP);
b202554c
MT
1025 return s->vhost_net;
1026}
16dbaf90
JW
1027
1028int tap_enable(NetClientState *nc)
1029{
1030 TAPState *s = DO_UPCAST(TAPState, nc, nc);
1031 int ret;
1032
1033 if (s->enabled) {
1034 return 0;
1035 } else {
1036 ret = tap_fd_enable(s->fd);
1037 if (ret == 0) {
1038 s->enabled = true;
1039 tap_update_fd_handler(s);
1040 }
1041 return ret;
1042 }
1043}
1044
1045int tap_disable(NetClientState *nc)
1046{
1047 TAPState *s = DO_UPCAST(TAPState, nc, nc);
1048 int ret;
1049
1050 if (s->enabled == 0) {
1051 return 0;
1052 } else {
1053 ret = tap_fd_disable(s->fd);
1054 if (ret == 0) {
1055 qemu_purge_queued_packets(nc);
1056 s->enabled = false;
1057 tap_update_fd_handler(s);
1058 }
1059 return ret;
1060 }
1061}