export iov_send_recv() and use it in iov_send() and iov_recv()
[qemu-kvm.git] / net / tap.c
blob5ac4ba33432d7cabbb6dc14e2d0c03732d3b929e
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 * Copyright (c) 2009 Red Hat, Inc.
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:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
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.
26 #include "net/tap.h"
28 #include "config-host.h"
30 #include <sys/ioctl.h>
31 #include <sys/stat.h>
32 #include <sys/wait.h>
33 #include <sys/socket.h>
34 #include <net/if.h>
36 #include "net.h"
37 #include "sysemu.h"
38 #include "qemu-char.h"
39 #include "qemu-common.h"
40 #include "qemu-error.h"
42 #include "net/tap-linux.h"
44 #include "hw/vhost_net.h"
46 /* Maximum GSO packet size (64k) plus plenty of room for
47 * the ethernet and virtio_net headers
49 #define TAP_BUFSIZE (4096 + 65536)
51 typedef struct TAPState {
52 VLANClientState nc;
53 int fd;
54 char down_script[1024];
55 char down_script_arg[128];
56 uint8_t buf[TAP_BUFSIZE];
57 unsigned int read_poll : 1;
58 unsigned int write_poll : 1;
59 unsigned int using_vnet_hdr : 1;
60 unsigned int has_ufo: 1;
61 VHostNetState *vhost_net;
62 unsigned host_vnet_hdr_len;
63 } TAPState;
65 static int launch_script(const char *setup_script, const char *ifname, int fd);
67 static int tap_can_send(void *opaque);
68 static void tap_send(void *opaque);
69 static void tap_writable(void *opaque);
71 static void tap_update_fd_handler(TAPState *s)
73 qemu_set_fd_handler2(s->fd,
74 s->read_poll ? tap_can_send : NULL,
75 s->read_poll ? tap_send : NULL,
76 s->write_poll ? tap_writable : NULL,
77 s);
80 static void tap_read_poll(TAPState *s, int enable)
82 s->read_poll = !!enable;
83 tap_update_fd_handler(s);
86 static void tap_write_poll(TAPState *s, int enable)
88 s->write_poll = !!enable;
89 tap_update_fd_handler(s);
92 static void tap_writable(void *opaque)
94 TAPState *s = opaque;
96 tap_write_poll(s, 0);
98 qemu_flush_queued_packets(&s->nc);
101 static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
103 ssize_t len;
105 do {
106 len = writev(s->fd, iov, iovcnt);
107 } while (len == -1 && errno == EINTR);
109 if (len == -1 && errno == EAGAIN) {
110 tap_write_poll(s, 1);
111 return 0;
114 return len;
117 static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
118 int iovcnt)
120 TAPState *s = DO_UPCAST(TAPState, nc, nc);
121 const struct iovec *iovp = iov;
122 struct iovec iov_copy[iovcnt + 1];
123 struct virtio_net_hdr_mrg_rxbuf hdr = { };
125 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
126 iov_copy[0].iov_base = &hdr;
127 iov_copy[0].iov_len = s->host_vnet_hdr_len;
128 memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
129 iovp = iov_copy;
130 iovcnt++;
133 return tap_write_packet(s, iovp, iovcnt);
136 static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t size)
138 TAPState *s = DO_UPCAST(TAPState, nc, nc);
139 struct iovec iov[2];
140 int iovcnt = 0;
141 struct virtio_net_hdr_mrg_rxbuf hdr = { };
143 if (s->host_vnet_hdr_len) {
144 iov[iovcnt].iov_base = &hdr;
145 iov[iovcnt].iov_len = s->host_vnet_hdr_len;
146 iovcnt++;
149 iov[iovcnt].iov_base = (char *)buf;
150 iov[iovcnt].iov_len = size;
151 iovcnt++;
153 return tap_write_packet(s, iov, iovcnt);
156 static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size)
158 TAPState *s = DO_UPCAST(TAPState, nc, nc);
159 struct iovec iov[1];
161 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
162 return tap_receive_raw(nc, buf, size);
165 iov[0].iov_base = (char *)buf;
166 iov[0].iov_len = size;
168 return tap_write_packet(s, iov, 1);
171 static int tap_can_send(void *opaque)
173 TAPState *s = opaque;
175 return qemu_can_send_packet(&s->nc);
178 #ifndef __sun__
179 ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
181 return read(tapfd, buf, maxlen);
183 #endif
185 static void tap_send_completed(VLANClientState *nc, ssize_t len)
187 TAPState *s = DO_UPCAST(TAPState, nc, nc);
188 tap_read_poll(s, 1);
191 static void tap_send(void *opaque)
193 TAPState *s = opaque;
194 int size;
196 do {
197 uint8_t *buf = s->buf;
199 size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
200 if (size <= 0) {
201 break;
204 if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
205 buf += s->host_vnet_hdr_len;
206 size -= s->host_vnet_hdr_len;
209 size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);
210 if (size == 0) {
211 tap_read_poll(s, 0);
213 } while (size > 0 && qemu_can_send_packet(&s->nc));
216 int tap_has_ufo(VLANClientState *nc)
218 TAPState *s = DO_UPCAST(TAPState, nc, nc);
220 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
222 return s->has_ufo;
225 int tap_has_vnet_hdr(VLANClientState *nc)
227 TAPState *s = DO_UPCAST(TAPState, nc, nc);
229 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
231 return !!s->host_vnet_hdr_len;
234 int tap_has_vnet_hdr_len(VLANClientState *nc, int len)
236 TAPState *s = DO_UPCAST(TAPState, nc, nc);
238 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
240 return tap_probe_vnet_hdr_len(s->fd, len);
243 void tap_set_vnet_hdr_len(VLANClientState *nc, int len)
245 TAPState *s = DO_UPCAST(TAPState, nc, nc);
247 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
248 assert(len == sizeof(struct virtio_net_hdr_mrg_rxbuf) ||
249 len == sizeof(struct virtio_net_hdr));
251 tap_fd_set_vnet_hdr_len(s->fd, len);
252 s->host_vnet_hdr_len = len;
255 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
257 TAPState *s = DO_UPCAST(TAPState, nc, nc);
259 using_vnet_hdr = using_vnet_hdr != 0;
261 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
262 assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
264 s->using_vnet_hdr = using_vnet_hdr;
267 void tap_set_offload(VLANClientState *nc, int csum, int tso4,
268 int tso6, int ecn, int ufo)
270 TAPState *s = DO_UPCAST(TAPState, nc, nc);
271 if (s->fd < 0) {
272 return;
275 tap_fd_set_offload(s->fd, csum, tso4, tso6, ecn, ufo);
278 static void tap_cleanup(VLANClientState *nc)
280 TAPState *s = DO_UPCAST(TAPState, nc, nc);
282 if (s->vhost_net) {
283 vhost_net_cleanup(s->vhost_net);
284 s->vhost_net = NULL;
287 qemu_purge_queued_packets(nc);
289 if (s->down_script[0])
290 launch_script(s->down_script, s->down_script_arg, s->fd);
292 tap_read_poll(s, 0);
293 tap_write_poll(s, 0);
294 close(s->fd);
295 s->fd = -1;
298 static void tap_poll(VLANClientState *nc, bool enable)
300 TAPState *s = DO_UPCAST(TAPState, nc, nc);
301 tap_read_poll(s, enable);
302 tap_write_poll(s, enable);
305 int tap_get_fd(VLANClientState *nc)
307 TAPState *s = DO_UPCAST(TAPState, nc, nc);
308 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
309 return s->fd;
312 /* fd support */
314 static NetClientInfo net_tap_info = {
315 .type = NET_CLIENT_TYPE_TAP,
316 .size = sizeof(TAPState),
317 .receive = tap_receive,
318 .receive_raw = tap_receive_raw,
319 .receive_iov = tap_receive_iov,
320 .poll = tap_poll,
321 .cleanup = tap_cleanup,
324 static TAPState *net_tap_fd_init(VLANState *vlan,
325 const char *model,
326 const char *name,
327 int fd,
328 int vnet_hdr)
330 VLANClientState *nc;
331 TAPState *s;
333 nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
335 s = DO_UPCAST(TAPState, nc, nc);
337 s->fd = fd;
338 s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
339 s->using_vnet_hdr = 0;
340 s->has_ufo = tap_probe_has_ufo(s->fd);
341 tap_set_offload(&s->nc, 0, 0, 0, 0, 0);
342 tap_read_poll(s, 1);
343 s->vhost_net = NULL;
344 return s;
347 static int launch_script(const char *setup_script, const char *ifname, int fd)
349 int pid, status;
350 char *args[3];
351 char **parg;
353 /* try to launch network script */
354 pid = fork();
355 if (pid == 0) {
356 int open_max = sysconf(_SC_OPEN_MAX), i;
358 for (i = 0; i < open_max; i++) {
359 if (i != STDIN_FILENO &&
360 i != STDOUT_FILENO &&
361 i != STDERR_FILENO &&
362 i != fd) {
363 close(i);
366 parg = args;
367 *parg++ = (char *)setup_script;
368 *parg++ = (char *)ifname;
369 *parg = NULL;
370 execv(setup_script, args);
371 _exit(1);
372 } else if (pid > 0) {
373 while (waitpid(pid, &status, 0) != pid) {
374 /* loop */
377 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
378 return 0;
381 fprintf(stderr, "%s: could not launch network script\n", setup_script);
382 return -1;
385 static int recv_fd(int c)
387 int fd;
388 uint8_t msgbuf[CMSG_SPACE(sizeof(fd))];
389 struct msghdr msg = {
390 .msg_control = msgbuf,
391 .msg_controllen = sizeof(msgbuf),
393 struct cmsghdr *cmsg;
394 struct iovec iov;
395 uint8_t req[1];
396 ssize_t len;
398 cmsg = CMSG_FIRSTHDR(&msg);
399 cmsg->cmsg_level = SOL_SOCKET;
400 cmsg->cmsg_type = SCM_RIGHTS;
401 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
402 msg.msg_controllen = cmsg->cmsg_len;
404 iov.iov_base = req;
405 iov.iov_len = sizeof(req);
407 msg.msg_iov = &iov;
408 msg.msg_iovlen = 1;
410 len = recvmsg(c, &msg, 0);
411 if (len > 0) {
412 memcpy(&fd, CMSG_DATA(cmsg), sizeof(fd));
413 return fd;
416 return len;
419 static int net_bridge_run_helper(const char *helper, const char *bridge)
421 sigset_t oldmask, mask;
422 int pid, status;
423 char *args[5];
424 char **parg;
425 int sv[2];
427 sigemptyset(&mask);
428 sigaddset(&mask, SIGCHLD);
429 sigprocmask(SIG_BLOCK, &mask, &oldmask);
431 if (socketpair(PF_UNIX, SOCK_STREAM, 0, sv) == -1) {
432 return -1;
435 /* try to launch bridge helper */
436 pid = fork();
437 if (pid == 0) {
438 int open_max = sysconf(_SC_OPEN_MAX), i;
439 char fd_buf[6+10];
440 char br_buf[6+IFNAMSIZ] = {0};
441 char helper_cmd[PATH_MAX + sizeof(fd_buf) + sizeof(br_buf) + 15];
443 for (i = 0; i < open_max; i++) {
444 if (i != STDIN_FILENO &&
445 i != STDOUT_FILENO &&
446 i != STDERR_FILENO &&
447 i != sv[1]) {
448 close(i);
452 snprintf(fd_buf, sizeof(fd_buf), "%s%d", "--fd=", sv[1]);
454 if (strrchr(helper, ' ') || strrchr(helper, '\t')) {
455 /* assume helper is a command */
457 if (strstr(helper, "--br=") == NULL) {
458 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
461 snprintf(helper_cmd, sizeof(helper_cmd), "%s %s %s %s",
462 helper, "--use-vnet", fd_buf, br_buf);
464 parg = args;
465 *parg++ = (char *)"sh";
466 *parg++ = (char *)"-c";
467 *parg++ = helper_cmd;
468 *parg++ = NULL;
470 execv("/bin/sh", args);
471 } else {
472 /* assume helper is just the executable path name */
474 snprintf(br_buf, sizeof(br_buf), "%s%s", "--br=", bridge);
476 parg = args;
477 *parg++ = (char *)helper;
478 *parg++ = (char *)"--use-vnet";
479 *parg++ = fd_buf;
480 *parg++ = br_buf;
481 *parg++ = NULL;
483 execv(helper, args);
485 _exit(1);
487 } else if (pid > 0) {
488 int fd;
490 close(sv[1]);
492 do {
493 fd = recv_fd(sv[0]);
494 } while (fd == -1 && errno == EINTR);
496 close(sv[0]);
498 while (waitpid(pid, &status, 0) != pid) {
499 /* loop */
501 sigprocmask(SIG_SETMASK, &oldmask, NULL);
502 if (fd < 0) {
503 fprintf(stderr, "failed to recv file descriptor\n");
504 return -1;
507 if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
508 return fd;
511 fprintf(stderr, "failed to launch bridge helper\n");
512 return -1;
515 int net_init_bridge(QemuOpts *opts, const char *name, VLANState *vlan)
517 TAPState *s;
518 int fd, vnet_hdr;
520 if (!qemu_opt_get(opts, "br")) {
521 qemu_opt_set(opts, "br", DEFAULT_BRIDGE_INTERFACE);
523 if (!qemu_opt_get(opts, "helper")) {
524 qemu_opt_set(opts, "helper", DEFAULT_BRIDGE_HELPER);
527 fd = net_bridge_run_helper(qemu_opt_get(opts, "helper"),
528 qemu_opt_get(opts, "br"));
529 if (fd == -1) {
530 return -1;
533 fcntl(fd, F_SETFL, O_NONBLOCK);
535 vnet_hdr = tap_probe_vnet_hdr(fd);
537 s = net_tap_fd_init(vlan, "bridge", name, fd, vnet_hdr);
538 if (!s) {
539 close(fd);
540 return -1;
543 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "helper=%s,br=%s",
544 qemu_opt_get(opts, "helper"), qemu_opt_get(opts, "br"));
546 return 0;
549 static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
551 int fd, vnet_hdr_required;
552 char ifname[128] = {0,};
553 const char *setup_script;
555 if (qemu_opt_get(opts, "ifname")) {
556 pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
559 *vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
560 if (qemu_opt_get(opts, "vnet_hdr")) {
561 vnet_hdr_required = *vnet_hdr;
562 } else {
563 vnet_hdr_required = 0;
566 TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
567 if (fd < 0) {
568 return -1;
571 setup_script = qemu_opt_get(opts, "script");
572 if (setup_script &&
573 setup_script[0] != '\0' &&
574 strcmp(setup_script, "no") != 0 &&
575 launch_script(setup_script, ifname, fd)) {
576 close(fd);
577 return -1;
580 qemu_opt_set(opts, "ifname", ifname);
582 return fd;
585 int net_init_tap(QemuOpts *opts, const char *name, VLANState *vlan)
587 TAPState *s;
588 int fd, vnet_hdr = 0;
589 const char *model;
591 if (qemu_opt_get(opts, "fd")) {
592 if (qemu_opt_get(opts, "ifname") ||
593 qemu_opt_get(opts, "script") ||
594 qemu_opt_get(opts, "downscript") ||
595 qemu_opt_get(opts, "vnet_hdr") ||
596 qemu_opt_get(opts, "helper")) {
597 error_report("ifname=, script=, downscript=, vnet_hdr=, "
598 "and helper= are invalid with fd=");
599 return -1;
602 fd = net_handle_fd_param(cur_mon, qemu_opt_get(opts, "fd"));
603 if (fd == -1) {
604 return -1;
607 fcntl(fd, F_SETFL, O_NONBLOCK);
609 vnet_hdr = tap_probe_vnet_hdr(fd);
611 model = "tap";
613 } else if (qemu_opt_get(opts, "helper")) {
614 if (qemu_opt_get(opts, "ifname") ||
615 qemu_opt_get(opts, "script") ||
616 qemu_opt_get(opts, "downscript") ||
617 qemu_opt_get(opts, "vnet_hdr")) {
618 error_report("ifname=, script=, downscript=, and vnet_hdr= "
619 "are invalid with helper=");
620 return -1;
623 fd = net_bridge_run_helper(qemu_opt_get(opts, "helper"),
624 DEFAULT_BRIDGE_INTERFACE);
625 if (fd == -1) {
626 return -1;
629 fcntl(fd, F_SETFL, O_NONBLOCK);
631 vnet_hdr = tap_probe_vnet_hdr(fd);
633 model = "bridge";
635 } else {
636 if (!qemu_opt_get(opts, "script")) {
637 qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
640 if (!qemu_opt_get(opts, "downscript")) {
641 qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
644 fd = net_tap_init(opts, &vnet_hdr);
645 if (fd == -1) {
646 return -1;
649 model = "tap";
652 s = net_tap_fd_init(vlan, model, name, fd, vnet_hdr);
653 if (!s) {
654 close(fd);
655 return -1;
658 if (tap_set_sndbuf(s->fd, opts) < 0) {
659 return -1;
662 if (qemu_opt_get(opts, "fd")) {
663 snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
664 } else if (qemu_opt_get(opts, "helper")) {
665 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
666 "helper=%s", qemu_opt_get(opts, "helper"));
667 } else {
668 const char *ifname, *script, *downscript;
670 ifname = qemu_opt_get(opts, "ifname");
671 script = qemu_opt_get(opts, "script");
672 downscript = qemu_opt_get(opts, "downscript");
674 snprintf(s->nc.info_str, sizeof(s->nc.info_str),
675 "ifname=%s,script=%s,downscript=%s",
676 ifname, script, downscript);
678 if (strcmp(downscript, "no") != 0) {
679 snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
680 snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
684 if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd") ||
685 qemu_opt_get_bool(opts, "vhostforce", false))) {
686 int vhostfd, r;
687 bool force = qemu_opt_get_bool(opts, "vhostforce", false);
688 if (qemu_opt_get(opts, "vhostfd")) {
689 r = net_handle_fd_param(cur_mon, qemu_opt_get(opts, "vhostfd"));
690 if (r == -1) {
691 return -1;
693 vhostfd = r;
694 } else {
695 vhostfd = -1;
697 s->vhost_net = vhost_net_init(&s->nc, vhostfd, force);
698 if (!s->vhost_net) {
699 error_report("vhost-net requested but could not be initialized");
700 return -1;
702 } else if (qemu_opt_get(opts, "vhostfd")) {
703 error_report("vhostfd= is not valid without vhost");
704 return -1;
707 return 0;
710 VHostNetState *tap_get_vhost_net(VLANClientState *nc)
712 TAPState *s = DO_UPCAST(TAPState, nc, nc);
713 assert(nc->info->type == NET_CLIENT_TYPE_TAP);
714 return s->vhost_net;