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
28
#include "config-host.h"
30
#include <sys/ioctl.h>
33
#include <sys/socket.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 {
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;
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,
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)
98
qemu_flush_queued_packets(&s->nc);
101
static ssize_t tap_write_packet(TAPState *s, const struct iovec *iov, int iovcnt)
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);
117
static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
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));
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);
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;
149
iov[iovcnt].iov_base = (char *)buf;
150
iov[iovcnt].iov_len = size;
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);
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);
179
ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen)
181
return read(tapfd, buf, maxlen);
185
static void tap_send_completed(VLANClientState *nc, ssize_t len)
187
TAPState *s = DO_UPCAST(TAPState, nc, nc);
191
static void tap_send(void *opaque)
193
TAPState *s = opaque;
197
uint8_t *buf = s->buf;
199
size = tap_read_packet(s->fd, s->buf, sizeof(s->buf));
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);
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);
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);
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);
283
vhost_net_cleanup(s->vhost_net);
287
qemu_purge_queued_packets(nc);
289
if (s->down_script[0])
290
launch_script(s->down_script, s->down_script_arg, s->fd);
293
tap_write_poll(s, 0);
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);
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,
321
.cleanup = tap_cleanup,
324
static TAPState *net_tap_fd_init(VLANState *vlan,
333
nc = qemu_new_net_client(&net_tap_info, vlan, NULL, model, name);
335
s = DO_UPCAST(TAPState, nc, nc);
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);
347
static int launch_script(const char *setup_script, const char *ifname, int fd)
349
sigset_t oldmask, mask;
355
sigaddset(&mask, SIGCHLD);
356
sigprocmask(SIG_BLOCK, &mask, &oldmask);
358
/* try to launch network script */
361
int open_max = sysconf(_SC_OPEN_MAX), i;
363
for (i = 0; i < open_max; i++) {
364
if (i != STDIN_FILENO &&
365
i != STDOUT_FILENO &&
366
i != STDERR_FILENO &&
372
*parg++ = (char *)setup_script;
373
*parg++ = (char *)ifname;
375
execv(setup_script, args);
377
} else if (pid > 0) {
378
while (waitpid(pid, &status, 0) != pid) {
381
sigprocmask(SIG_SETMASK, &oldmask, NULL);
383
if (WIFEXITED(status) && WEXITSTATUS(status) == 0) {
387
fprintf(stderr, "%s: could not launch network script\n", setup_script);
391
static int net_tap_init(QemuOpts *opts, int *vnet_hdr)
393
int fd, vnet_hdr_required;
394
char ifname[128] = {0,};
395
const char *setup_script;
397
if (qemu_opt_get(opts, "ifname")) {
398
pstrcpy(ifname, sizeof(ifname), qemu_opt_get(opts, "ifname"));
401
*vnet_hdr = qemu_opt_get_bool(opts, "vnet_hdr", 1);
402
if (qemu_opt_get(opts, "vnet_hdr")) {
403
vnet_hdr_required = *vnet_hdr;
405
vnet_hdr_required = 0;
408
TFR(fd = tap_open(ifname, sizeof(ifname), vnet_hdr, vnet_hdr_required));
413
setup_script = qemu_opt_get(opts, "script");
415
setup_script[0] != '\0' &&
416
strcmp(setup_script, "no") != 0 &&
417
launch_script(setup_script, ifname, fd)) {
422
qemu_opt_set(opts, "ifname", ifname);
427
int net_init_tap(QemuOpts *opts, Monitor *mon, const char *name, VLANState *vlan)
430
int fd, vnet_hdr = 0;
432
if (qemu_opt_get(opts, "fd")) {
433
if (qemu_opt_get(opts, "ifname") ||
434
qemu_opt_get(opts, "script") ||
435
qemu_opt_get(opts, "downscript") ||
436
qemu_opt_get(opts, "vnet_hdr")) {
437
error_report("ifname=, script=, downscript= and vnet_hdr= is invalid with fd=");
441
fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
446
fcntl(fd, F_SETFL, O_NONBLOCK);
448
vnet_hdr = tap_probe_vnet_hdr(fd);
450
if (!qemu_opt_get(opts, "script")) {
451
qemu_opt_set(opts, "script", DEFAULT_NETWORK_SCRIPT);
454
if (!qemu_opt_get(opts, "downscript")) {
455
qemu_opt_set(opts, "downscript", DEFAULT_NETWORK_DOWN_SCRIPT);
458
fd = net_tap_init(opts, &vnet_hdr);
464
s = net_tap_fd_init(vlan, "tap", name, fd, vnet_hdr);
470
if (tap_set_sndbuf(s->fd, opts) < 0) {
474
if (qemu_opt_get(opts, "fd")) {
475
snprintf(s->nc.info_str, sizeof(s->nc.info_str), "fd=%d", fd);
477
const char *ifname, *script, *downscript;
479
ifname = qemu_opt_get(opts, "ifname");
480
script = qemu_opt_get(opts, "script");
481
downscript = qemu_opt_get(opts, "downscript");
483
snprintf(s->nc.info_str, sizeof(s->nc.info_str),
484
"ifname=%s,script=%s,downscript=%s",
485
ifname, script, downscript);
487
if (strcmp(downscript, "no") != 0) {
488
snprintf(s->down_script, sizeof(s->down_script), "%s", downscript);
489
snprintf(s->down_script_arg, sizeof(s->down_script_arg), "%s", ifname);
493
if (qemu_opt_get_bool(opts, "vhost", !!qemu_opt_get(opts, "vhostfd") ||
494
qemu_opt_get_bool(opts, "vhostforce", false))) {
496
bool force = qemu_opt_get_bool(opts, "vhostforce", false);
497
if (qemu_opt_get(opts, "vhostfd")) {
498
r = net_handle_fd_param(mon, qemu_opt_get(opts, "vhostfd"));
506
s->vhost_net = vhost_net_init(&s->nc, vhostfd, force);
508
error_report("vhost-net requested but could not be initialized");
511
} else if (qemu_opt_get(opts, "vhostfd")) {
512
error_report("vhostfd= is not valid without vhost");
519
VHostNetState *tap_get_vhost_net(VLANClientState *nc)
521
TAPState *s = DO_UPCAST(TAPState, nc, nc);
522
assert(nc->info->type == NET_CLIENT_TYPE_TAP);