]> git.proxmox.com Git - mirror_qemu.git/blame - net/tap.c
net: virtio-net and vmxnet3 use offloading API
[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;
192
5819c918
MM
193 do {
194 uint8_t *buf = s->buf;
195
196 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
197 if (size <= 0) {
198 break;
199 }
200
ef4252b1
MT
201 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
202 buf += s->host_vnet_hdr_len;
203 size -= s->host_vnet_hdr_len;
5819c918
MM
204 }
205
3e35ba93 206 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
5819c918 207 if (size == 0) {
ec45f083 208 tap_read_poll(s, false);
5819c918 209 }
3e35ba93 210 } while (size > 0 && qemu_can_send_packet(&s->nc));
5281d757
MM
211}
212
ec45f083 213bool tap_has_ufo(NetClientState *nc)
5281d757 214{
3e35ba93 215 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 216
2be64a68 217 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757
MM
218
219 return s->has_ufo;
220}
221
e96dfd11 222bool tap_has_vnet_hdr(NetClientState *nc)
5281d757 223{
3e35ba93 224 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 225
2be64a68 226 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
5281d757 227
ef4252b1 228 return !!s->host_vnet_hdr_len;
5281d757
MM
229}
230
e96dfd11 231bool tap_has_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
232{
233 TAPState *s = DO_UPCAST(TAPState, nc, nc);
234
2be64a68 235 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f 236
e96dfd11 237 return !!tap_probe_vnet_hdr_len(s->fd, len);
445d892f
MT
238}
239
4e68f7a0 240void tap_set_vnet_hdr_len(NetClientState *nc, int len)
445d892f
MT
241{
242 TAPState *s = DO_UPCAST(TAPState, nc, nc);
243
2be64a68 244 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
445d892f
MT
245 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
246 len == sizeof(struct virtio_net_hdr));
247
248 tap_fd_set_vnet_hdr_len(s->fd, len);
249 s->host_vnet_hdr_len = len;
250}
251
ec45f083 252void tap_using_vnet_hdr(NetClientState *nc, bool using_vnet_hdr)
5281d757 253{
3e35ba93 254 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 255
2be64a68 256 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
ef4252b1 257 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
5281d757
MM
258
259 s->using_vnet_hdr = using_vnet_hdr;
260}
261
4e68f7a0 262void tap_set_offload(NetClientState *nc, int csum, int tso4,
5281d757
MM
263 int tso6, int ecn, int ufo)
264{
3e35ba93 265 TAPState *s = DO_UPCAST(TAPState, nc, nc);
27a6375d
MT
266 if (s->fd < 0) {
267 return;
268 }
5281d757 269
27a6375d 270 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
5281d757
MM
271}
272
4e68f7a0 273static void tap_cleanup(NetClientState *nc)
5281d757 274{
3e35ba93 275 TAPState *s = DO_UPCAST(TAPState, nc, nc);
5281d757 276
82b0d80e
MT
277 if (s->vhost_net) {
278 vhost_net_cleanup(s->vhost_net);
43849424 279 s->vhost_net = NULL;
82b0d80e
MT
280 }
281
3e35ba93 282 qemu_purge_queued_packets(nc);
5281d757
MM
283
284 if (s->down_script[0])
285 launch_script(s->down_script, s->down_script_arg, s->fd);
286
ec45f083
JW
287 tap_read_poll(s, false);
288 tap_write_poll(s, false);
5281d757 289 close(s->fd);
27a6375d 290 s->fd = -1;
5281d757
MM
291}
292
4e68f7a0 293static void tap_poll(NetClientState *nc, bool enable)
ceb69615
MT
294{
295 TAPState *s = DO_UPCAST(TAPState, nc, nc);
296 tap_read_poll(s, enable);
297 tap_write_poll(s, enable);
298}
299
4e68f7a0 300int tap_get_fd(NetClientState *nc)
95d528a2
MT
301{
302 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 303 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
95d528a2
MT
304 return s->fd;
305}
306
5281d757
MM
307/* fd support */
308
3e35ba93 309static NetClientInfo net_tap_info = {
2be64a68 310 .type = NET_CLIENT_OPTIONS_KIND_TAP,
3e35ba93
MM
311 .size = sizeof(TAPState),
312 .receive = tap_receive,
313 .receive_raw = tap_receive_raw,
314 .receive_iov = tap_receive_iov,
ceb69615 315 .poll = tap_poll,
3e35ba93 316 .cleanup = tap_cleanup,
2e753bcc
VM
317 .has_ufo = tap_has_ufo,
318 .has_vnet_hdr = tap_has_vnet_hdr,
319 .has_vnet_hdr_len = tap_has_vnet_hdr_len,
320 .using_vnet_hdr = tap_using_vnet_hdr,
321 .set_offload = tap_set_offload,
322 .set_vnet_hdr_len = tap_set_vnet_hdr_len,
3e35ba93
MM
323};
324
4e68f7a0 325static TAPState *net_tap_fd_init(NetClientState *peer,
5281d757
MM
326 const char *model,
327 const char *name,
328 int fd,
329 int vnet_hdr)
330{
4e68f7a0 331 NetClientState *nc;
5281d757 332 TAPState *s;
5281d757 333
ab5f3f84 334 nc = qemu_new_net_client(&net_tap_info, peer, model, name);
3e35ba93
MM
335
336 s = DO_UPCAST(TAPState, nc, nc);
337
5281d757 338 s->fd = fd;
ef4252b1 339 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
ec45f083 340 s->using_vnet_hdr = false;
9c282718 341 s->has_ufo = tap_probe_has_ufo(s->fd);
16dbaf90 342 s->enabled = true;
3e35ba93 343 tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
58ddcd50
MT
344 /*
345 * Make sure host header length is set correctly in tap:
346 * it might have been modified by another instance of qemu.
347 */
348 if (tap_probe_vnet_hdr_len(s->fd, s->host_vnet_hdr_len)) {
349 tap_fd_set_vnet_hdr_len(s->fd, s->host_vnet_hdr_len);
350 }
ec45f083 351 tap_read_poll(s, true);
82b0d80e 352 s->vhost_net = NULL;
5281d757
MM
353 return s;
354}
355
5281d757
MM
356static int launch_script(const char *setup_script, const char *ifname, int fd)
357{
5281d757
MM
358 int pid, status;
359 char *args[3];
360 char **parg;
361
5281d757
MM
362 /* try to launch network script */
363 pid = fork();
364 if (pid == 0) {
365 int open_max = sysconf(_SC_OPEN_MAX), i;
366
367 for (i = 0; i < open_max; i++) {
368 if (i != STDIN_FILENO &&
369 i != STDOUT_FILENO &&
370 i != STDERR_FILENO &&
371 i != fd) {
372 close(i);
373 }
374 }
375 parg = args;
376 *parg++ = (char *)setup_script;
377 *parg++ = (char *)ifname;
9678d950 378 *parg = NULL;
5281d757
MM
379 execv(setup_script, args);
380 _exit(1);
381 } else if (pid > 0) {
382 while (waitpid(pid, &status, 0) != pid) {
383 /* loop */
384 }
5281d757
MM
385
386 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
387 return 0;
388 }
389 }
390 fprintf(stderr, "%s: could not launch network script\n", setup_script);
391 return -1;
392}
393
a7c36ee4
CB
394static int recv_fd(int c)
395{
396 int fd;
397 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
398 struct msghdr msg = {
399 .msg_control = msgbuf,
400 .msg_controllen = sizeof(msgbuf),
401 };
402 struct cmsghdr *cmsg;
403 struct iovec iov;
404 uint8_t req[1];
405 ssize_t len;
406
407 cmsg = CMSG_FIRSTHDR(&msg);
408 cmsg->cmsg_level = SOL_SOCKET;
409 cmsg->cmsg_type = SCM_RIGHTS;
410 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
411 msg.msg_controllen = cmsg->cmsg_len;
412
413 iov.iov_base = req;
414 iov.iov_len = sizeof(req);
415
416 msg.msg_iov = &iov;
417 msg.msg_iovlen = 1;
418
419 len = recvmsg(c, &msg, 0);
420 if (len > 0) {
421 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
422 return fd;
423 }
424
425 return len;
426}
427
428static int net_bridge_run_helper(const char *helper, const char *bridge)
429{
430 sigset_t oldmask, mask;
431 int pid, status;
432 char *args[5];
433 char **parg;
434 int sv[2];
435
436 sigemptyset(&mask);
437 sigaddset(&mask, SIGCHLD);
438 sigprocmask(SIG_BLOCK, &mask, &oldmask);
439
440 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
441 return -1;
442 }
443
444 /* try to launch bridge helper */
445 pid = fork();
446 if (pid == 0) {
447 int open_max = sysconf(_SC_OPEN_MAX), i;
448 char fd_buf[6+10];
449 char br_buf[6+IFNAMSIZ] = {0};
450 char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
451
452 for (i = 0; i < open_max; i++) {
453 if (i != STDIN_FILENO &&
454 i != STDOUT_FILENO &&
455 i != STDERR_FILENO &&
456 i != sv[1]) {
457 close(i);
458 }
459 }
460
461 snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
462
463 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
464 /* assume helper is a command */
465
466 if (strstr(helper, "--br=") == NULL) {
467 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
468 }
469
470 snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
471 helper, "--use-vnet", fd_buf, br_buf);
472
473 parg = args;
474 *parg++ = (char *)"sh";
475 *parg++ = (char *)"-c";
476 *parg++ = helper_cmd;
477 *parg++ = NULL;
478
479 execv("/bin/sh", args);
480 } else {
481 /* assume helper is just the executable path name */
482
483 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
484
485 parg = args;
486 *parg++ = (char *)helper;
487 *parg++ = (char *)"--use-vnet";
488 *parg++ = fd_buf;
489 *parg++ = br_buf;
490 *parg++ = NULL;
491
492 execv(helper, args);
493 }
494 _exit(1);
495
496 } else if (pid > 0) {
497 int fd;
498
499 close(sv[1]);
500
501 do {
502 fd = recv_fd(sv[0]);
503 } while (fd == -1 && errno == EINTR);
504
505 close(sv[0]);
506
507 while (waitpid(pid, &status, 0) != pid) {
508 /* loop */
509 }
510 sigprocmask(SIG_SETMASK, &oldmask, NULL);
511 if (fd < 0) {
512 fprintf(stderr, "failed to recv file descriptor\n");
513 return -1;
514 }
515
516 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
517 return fd;
518 }
519 }
520 fprintf(stderr, "failed to launch bridge helper\n");
521 return -1;
522}
523
1a0c0958 524int net_init_bridge(const NetClientOptions *opts, const char *name,
4e68f7a0 525 NetClientState *peer)
a7c36ee4 526{
f79b51b0
LE
527 const NetdevBridgeOptions *bridge;
528 const char *helper, *br;
529
a7c36ee4
CB
530 TAPState *s;
531 int fd, vnet_hdr;
532
f79b51b0
LE
533 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_BRIDGE);
534 bridge = opts->bridge;
535
536 helper = bridge->has_helper ? bridge->helper : DEFAULT_BRIDGE_HELPER;
537 br = bridge->has_br ? bridge->br : DEFAULT_BRIDGE_INTERFACE;
a7c36ee4 538
f79b51b0 539 fd = net_bridge_run_helper(helper, br);
a7c36ee4
CB
540 if (fd == -1) {
541 return -1;
542 }
543
544 fcntl(fd, F_SETFL, O_NONBLOCK);
545
546 vnet_hdr = tap_probe_vnet_hdr(fd);
547
d33d93b2 548 s = net_tap_fd_init(peer, "bridge", name, fd, vnet_hdr);
a7c36ee4
CB
549 if (!s) {
550 close(fd);
551 return -1;
552 }
553
f79b51b0
LE
554 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s", helper,
555 br);
a7c36ee4
CB
556
557 return 0;
558}
559
08c573a8
LE
560static int net_tap_init(const NetdevTapOptions *tap, int *vnet_hdr,
561 const char *setup_script, char *ifname,
264986e2 562 size_t ifname_sz, int mq_required)
5281d757
MM
563{
564 int fd, vnet_hdr_required;
5281d757 565
08c573a8
LE
566 if (tap->has_vnet_hdr) {
567 *vnet_hdr = tap->vnet_hdr;
5281d757
MM
568 vnet_hdr_required = *vnet_hdr;
569 } else {
08c573a8 570 *vnet_hdr = 1;
5281d757
MM
571 vnet_hdr_required = 0;
572 }
573
264986e2
JW
574 TFR(fd = tap_open(ifname, ifname_sz, vnet_hdr, vnet_hdr_required,
575 mq_required));
5281d757
MM
576 if (fd < 0) {
577 return -1;
578 }
579
5281d757
MM
580 if (setup_script &&
581 setup_script[0] != '\0' &&
582 strcmp(setup_script, "no") != 0 &&
583 launch_script(setup_script, ifname, fd)) {
584 close(fd);
585 return -1;
586 }
587
5281d757
MM
588 return fd;
589}
590
264986e2
JW
591#define MAX_TAP_QUEUES 1024
592
5193e5fb
JW
593static int net_init_tap_one(const NetdevTapOptions *tap, NetClientState *peer,
594 const char *model, const char *name,
595 const char *ifname, const char *script,
596 const char *downscript, const char *vhostfdname,
597 int vnet_hdr, int fd)
598{
599 TAPState *s;
600
601 s = net_tap_fd_init(peer, model, name, fd, vnet_hdr);
602 if (!s) {
603 close(fd);
604 return -1;
605 }
606
607 if (tap_set_sndbuf(s->fd, tap) < 0) {
608 return -1;
609 }
610
264986e2 611 if (tap->has_fd || tap->has_fds) {
5193e5fb
JW
612 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
613 } else if (tap->has_helper) {
614 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s",
615 tap->helper);
616 } else {
5193e5fb
JW
617 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
618 "ifname=%s,script=%s,downscript=%s", ifname, script,
619 downscript);
620
621 if (strcmp(downscript, "no") != 0) {
622 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
623 snprintf(s->down_script_arg, sizeof(s->down_script_arg),
624 "%s", ifname);
625 }
626 }
627
628 if (tap->has_vhost ? tap->vhost :
629 vhostfdname || (tap->has_vhostforce && tap->vhostforce)) {
630 int vhostfd;
631
7873df40 632 if (tap->has_vhostfd || tap->has_vhostfds) {
5193e5fb
JW
633 vhostfd = monitor_handle_fd_param(cur_mon, vhostfdname);
634 if (vhostfd == -1) {
635 return -1;
636 }
637 } else {
638 vhostfd = -1;
639 }
640
641 s->vhost_net = vhost_net_init(&s->nc, vhostfd,
642 tap->has_vhostforce && tap->vhostforce);
643 if (!s->vhost_net) {
644 error_report("vhost-net requested but could not be initialized");
645 return -1;
646 }
264986e2 647 } else if (tap->has_vhostfd || tap->has_vhostfds) {
5193e5fb
JW
648 error_report("vhostfd= is not valid without vhost");
649 return -1;
650 }
651
652 return 0;
653}
654
264986e2
JW
655static int get_fds(char *str, char *fds[], int max)
656{
657 char *ptr = str, *this;
658 size_t len = strlen(str);
659 int i = 0;
660
661 while (i < max && ptr < str + len) {
662 this = strchr(ptr, ':');
663
664 if (this == NULL) {
665 fds[i] = g_strdup(ptr);
666 } else {
667 fds[i] = g_strndup(ptr, this - ptr);
668 }
669
670 i++;
671 if (this == NULL) {
672 break;
673 } else {
674 ptr = this + 1;
675 }
676 }
677
678 return i;
679}
680
1a0c0958 681int net_init_tap(const NetClientOptions *opts, const char *name,
4e68f7a0 682 NetClientState *peer)
5281d757 683{
08c573a8 684 const NetdevTapOptions *tap;
264986e2 685 int fd, vnet_hdr = 0, i = 0, queues;
08c573a8
LE
686 /* for the no-fd, no-helper case */
687 const char *script = NULL; /* suppress wrong "uninit'd use" gcc warning */
5193e5fb 688 const char *downscript = NULL;
264986e2 689 const char *vhostfdname;
08c573a8
LE
690 char ifname[128];
691
692 assert(opts->kind == NET_CLIENT_OPTIONS_KIND_TAP);
693 tap = opts->tap;
264986e2
JW
694 queues = tap->has_queues ? tap->queues : 1;
695 vhostfdname = tap->has_vhostfd ? tap->vhostfd : NULL;
5281d757 696
ce675a75
JW
697 /* QEMU vlans does not support multiqueue tap, in this case peer is set.
698 * For -netdev, peer is always NULL. */
699 if (peer && (tap->has_queues || tap->has_fds || tap->has_vhostfds)) {
085d8134 700 error_report("Multiqueue tap cannot be used with QEMU vlans");
ce675a75
JW
701 return -1;
702 }
703
08c573a8
LE
704 if (tap->has_fd) {
705 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
264986e2 706 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 707 tap->has_fds || tap->has_vhostfds) {
a7c36ee4 708 error_report("ifname=, script=, downscript=, vnet_hdr=, "
c87826a8
JW
709 "helper=, queues=, fds=, and vhostfds= "
710 "are invalid with fd=");
5281d757
MM
711 return -1;
712 }
713
a96ed02f 714 fd = monitor_handle_fd_param(cur_mon, tap->fd);
5281d757
MM
715 if (fd == -1) {
716 return -1;
717 }
718
719 fcntl(fd, F_SETFL, O_NONBLOCK);
720
721 vnet_hdr = tap_probe_vnet_hdr(fd);
a7c36ee4 722
02cd8090 723 if (net_init_tap_one(tap, peer, "tap", name, NULL,
264986e2
JW
724 script, downscript,
725 vhostfdname, vnet_hdr, fd)) {
726 return -1;
727 }
728 } else if (tap->has_fds) {
729 char *fds[MAX_TAP_QUEUES];
730 char *vhost_fds[MAX_TAP_QUEUES];
731 int nfds, nvhosts;
732
733 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
734 tap->has_vnet_hdr || tap->has_helper || tap->has_queues ||
c87826a8 735 tap->has_vhostfd) {
264986e2 736 error_report("ifname=, script=, downscript=, vnet_hdr=, "
c87826a8
JW
737 "helper=, queues=, and vhostfd= "
738 "are invalid with fds=");
264986e2
JW
739 return -1;
740 }
741
742 nfds = get_fds(tap->fds, fds, MAX_TAP_QUEUES);
743 if (tap->has_vhostfds) {
744 nvhosts = get_fds(tap->vhostfds, vhost_fds, MAX_TAP_QUEUES);
745 if (nfds != nvhosts) {
746 error_report("The number of fds passed does not match the "
747 "number of vhostfds passed");
748 return -1;
749 }
750 }
751
752 for (i = 0; i < nfds; i++) {
753 fd = monitor_handle_fd_param(cur_mon, fds[i]);
754 if (fd == -1) {
755 return -1;
756 }
757
758 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4 759
264986e2
JW
760 if (i == 0) {
761 vnet_hdr = tap_probe_vnet_hdr(fd);
762 } else if (vnet_hdr != tap_probe_vnet_hdr(fd)) {
763 error_report("vnet_hdr not consistent across given tap fds");
764 return -1;
765 }
766
767 if (net_init_tap_one(tap, peer, "tap", name, ifname,
768 script, downscript,
769 tap->has_vhostfds ? vhost_fds[i] : NULL,
770 vnet_hdr, fd)) {
771 return -1;
772 }
773 }
08c573a8
LE
774 } else if (tap->has_helper) {
775 if (tap->has_ifname || tap->has_script || tap->has_downscript ||
c87826a8 776 tap->has_vnet_hdr || tap->has_queues || tap->has_vhostfds) {
a7c36ee4 777 error_report("ifname=, script=, downscript=, and vnet_hdr= "
c87826a8 778 "queues=, and vhostfds= are invalid with helper=");
a7c36ee4
CB
779 return -1;
780 }
781
08c573a8 782 fd = net_bridge_run_helper(tap->helper, DEFAULT_BRIDGE_INTERFACE);
a7c36ee4
CB
783 if (fd == -1) {
784 return -1;
785 }
786
787 fcntl(fd, F_SETFL, O_NONBLOCK);
a7c36ee4
CB
788 vnet_hdr = tap_probe_vnet_hdr(fd);
789
264986e2
JW
790 if (net_init_tap_one(tap, peer, "bridge", name, ifname,
791 script, downscript, vhostfdname,
792 vnet_hdr, fd)) {
793 return -1;
794 }
5281d757 795 } else {
c87826a8
JW
796 if (tap->has_vhostfds) {
797 error_report("vhostfds= is invalid if fds= wasn't specified");
798 return -1;
799 }
08c573a8 800 script = tap->has_script ? tap->script : DEFAULT_NETWORK_SCRIPT;
5193e5fb
JW
801 downscript = tap->has_downscript ? tap->downscript :
802 DEFAULT_NETWORK_DOWN_SCRIPT;
264986e2
JW
803
804 if (tap->has_ifname) {
805 pstrcpy(ifname, sizeof ifname, tap->ifname);
806 } else {
807 ifname[0] = '\0';
929fe497 808 }
a7c36ee4 809
264986e2
JW
810 for (i = 0; i < queues; i++) {
811 fd = net_tap_init(tap, &vnet_hdr, i >= 1 ? "no" : script,
812 ifname, sizeof ifname, queues > 1);
813 if (fd == -1) {
814 return -1;
815 }
816
817 if (queues > 1 && i == 0 && !tap->has_ifname) {
818 if (tap_fd_get_ifname(fd, ifname)) {
819 error_report("Fail to get ifname");
820 return -1;
821 }
822 }
823
824 if (net_init_tap_one(tap, peer, "tap", name, ifname,
825 i >= 1 ? "no" : script,
826 i >= 1 ? "no" : downscript,
827 vhostfdname, vnet_hdr, fd)) {
828 return -1;
829 }
830 }
5281d757
MM
831 }
832
264986e2 833 return 0;
5281d757 834}
b202554c 835
4e68f7a0 836VHostNetState *tap_get_vhost_net(NetClientState *nc)
b202554c
MT
837{
838 TAPState *s = DO_UPCAST(TAPState, nc, nc);
2be64a68 839 assert(nc->info->type == NET_CLIENT_OPTIONS_KIND_TAP);
b202554c
MT
840 return s->vhost_net;
841}
16dbaf90
JW
842
843int tap_enable(NetClientState *nc)
844{
845 TAPState *s = DO_UPCAST(TAPState, nc, nc);
846 int ret;
847
848 if (s->enabled) {
849 return 0;
850 } else {
851 ret = tap_fd_enable(s->fd);
852 if (ret == 0) {
853 s->enabled = true;
854 tap_update_fd_handler(s);
855 }
856 return ret;
857 }
858}
859
860int tap_disable(NetClientState *nc)
861{
862 TAPState *s = DO_UPCAST(TAPState, nc, nc);
863 int ret;
864
865 if (s->enabled == 0) {
866 return 0;
867 } else {
868 ret = tap_fd_disable(s->fd);
869 if (ret == 0) {
870 qemu_purge_queued_packets(nc);
871 s->enabled = false;
872 tap_update_fd_handler(s);
873 }
874 return ret;
875 }
876}