root / qemu-nbd.c @ bb345110
History | View | Annotate | Download (14.5 kB)
1 |
/*
|
---|---|
2 |
* Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
|
3 |
*
|
4 |
* Network Block Device
|
5 |
*
|
6 |
* This program is free software; you can redistribute it and/or modify
|
7 |
* it under the terms of the GNU General Public License as published by
|
8 |
* the Free Software Foundation; under version 2 of the License.
|
9 |
*
|
10 |
* This program is distributed in the hope that it will be useful,
|
11 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
12 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
13 |
* GNU General Public License for more details.
|
14 |
*
|
15 |
* You should have received a copy of the GNU General Public License
|
16 |
* along with this program; if not, see <http://www.gnu.org/licenses/>.
|
17 |
*/
|
18 |
|
19 |
#include "qemu-common.h" |
20 |
#include "block_int.h" |
21 |
#include "nbd.h" |
22 |
|
23 |
#include <stdarg.h> |
24 |
#include <stdio.h> |
25 |
#include <getopt.h> |
26 |
#include <err.h> |
27 |
#include <sys/types.h> |
28 |
#include <sys/socket.h> |
29 |
#include <netinet/in.h> |
30 |
#include <netinet/tcp.h> |
31 |
#include <arpa/inet.h> |
32 |
#include <signal.h> |
33 |
#include <libgen.h> |
34 |
|
35 |
#define SOCKET_PATH "/var/lock/qemu-nbd-%s" |
36 |
|
37 |
#define NBD_BUFFER_SIZE (1024*1024) |
38 |
|
39 |
static int sigterm_wfd; |
40 |
static int verbose; |
41 |
|
42 |
static void usage(const char *name) |
43 |
{ |
44 |
printf( |
45 |
"Usage: %s [OPTIONS] FILE\n"
|
46 |
"QEMU Disk Network Block Device Server\n"
|
47 |
"\n"
|
48 |
" -p, --port=PORT port to listen on (default `%d')\n"
|
49 |
" -o, --offset=OFFSET offset into the image\n"
|
50 |
" -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
|
51 |
" -k, --socket=PATH path to the unix socket\n"
|
52 |
" (default '"SOCKET_PATH"')\n" |
53 |
" -r, --read-only export read-only\n"
|
54 |
" -P, --partition=NUM only expose partition NUM\n"
|
55 |
" -s, --snapshot use snapshot file\n"
|
56 |
" -n, --nocache disable host cache\n"
|
57 |
" -c, --connect=DEV connect FILE to the local NBD device DEV\n"
|
58 |
" -d, --disconnect disconnect the specified device\n"
|
59 |
" -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
|
60 |
" -t, --persistent don't exit on the last connection\n"
|
61 |
" -v, --verbose display extra debugging information\n"
|
62 |
" -h, --help display this help and exit\n"
|
63 |
" -V, --version output version information and exit\n"
|
64 |
"\n"
|
65 |
"Report bugs to <anthony@codemonkey.ws>\n"
|
66 |
, name, NBD_DEFAULT_PORT, "DEVICE");
|
67 |
} |
68 |
|
69 |
static void version(const char *name) |
70 |
{ |
71 |
printf( |
72 |
"%s version 0.0.1\n"
|
73 |
"Written by Anthony Liguori.\n"
|
74 |
"\n"
|
75 |
"Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
|
76 |
"This is free software; see the source for copying conditions. There is NO\n"
|
77 |
"warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
|
78 |
, name); |
79 |
} |
80 |
|
81 |
struct partition_record
|
82 |
{ |
83 |
uint8_t bootable; |
84 |
uint8_t start_head; |
85 |
uint32_t start_cylinder; |
86 |
uint8_t start_sector; |
87 |
uint8_t system; |
88 |
uint8_t end_head; |
89 |
uint8_t end_cylinder; |
90 |
uint8_t end_sector; |
91 |
uint32_t start_sector_abs; |
92 |
uint32_t nb_sectors_abs; |
93 |
}; |
94 |
|
95 |
static void read_partition(uint8_t *p, struct partition_record *r) |
96 |
{ |
97 |
r->bootable = p[0];
|
98 |
r->start_head = p[1];
|
99 |
r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300); |
100 |
r->start_sector = p[2] & 0x3f; |
101 |
r->system = p[4];
|
102 |
r->end_head = p[5];
|
103 |
r->end_cylinder = p[7] | ((p[6] << 2) & 0x300); |
104 |
r->end_sector = p[6] & 0x3f; |
105 |
r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24; |
106 |
r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24; |
107 |
} |
108 |
|
109 |
static int find_partition(BlockDriverState *bs, int partition, |
110 |
off_t *offset, off_t *size) |
111 |
{ |
112 |
struct partition_record mbr[4]; |
113 |
uint8_t data[512];
|
114 |
int i;
|
115 |
int ext_partnum = 4; |
116 |
int ret;
|
117 |
|
118 |
if ((ret = bdrv_read(bs, 0, data, 1)) < 0) { |
119 |
errno = -ret; |
120 |
err(EXIT_FAILURE, "error while reading");
|
121 |
} |
122 |
|
123 |
if (data[510] != 0x55 || data[511] != 0xaa) { |
124 |
errno = -EINVAL; |
125 |
return -1; |
126 |
} |
127 |
|
128 |
for (i = 0; i < 4; i++) { |
129 |
read_partition(&data[446 + 16 * i], &mbr[i]); |
130 |
|
131 |
if (!mbr[i].nb_sectors_abs)
|
132 |
continue;
|
133 |
|
134 |
if (mbr[i].system == 0xF || mbr[i].system == 0x5) { |
135 |
struct partition_record ext[4]; |
136 |
uint8_t data1[512];
|
137 |
int j;
|
138 |
|
139 |
if ((ret = bdrv_read(bs, mbr[i].start_sector_abs, data1, 1)) < 0) { |
140 |
errno = -ret; |
141 |
err(EXIT_FAILURE, "error while reading");
|
142 |
} |
143 |
|
144 |
for (j = 0; j < 4; j++) { |
145 |
read_partition(&data1[446 + 16 * j], &ext[j]); |
146 |
if (!ext[j].nb_sectors_abs)
|
147 |
continue;
|
148 |
|
149 |
if ((ext_partnum + j + 1) == partition) { |
150 |
*offset = (uint64_t)ext[j].start_sector_abs << 9;
|
151 |
*size = (uint64_t)ext[j].nb_sectors_abs << 9;
|
152 |
return 0; |
153 |
} |
154 |
} |
155 |
ext_partnum += 4;
|
156 |
} else if ((i + 1) == partition) { |
157 |
*offset = (uint64_t)mbr[i].start_sector_abs << 9;
|
158 |
*size = (uint64_t)mbr[i].nb_sectors_abs << 9;
|
159 |
return 0; |
160 |
} |
161 |
} |
162 |
|
163 |
errno = -ENOENT; |
164 |
return -1; |
165 |
} |
166 |
|
167 |
static void termsig_handler(int signum) |
168 |
{ |
169 |
static int sigterm_reported; |
170 |
if (!sigterm_reported) {
|
171 |
sigterm_reported = (write(sigterm_wfd, "", 1) == 1); |
172 |
} |
173 |
} |
174 |
|
175 |
static void show_parts(const char *device) |
176 |
{ |
177 |
if (fork() == 0) { |
178 |
int nbd;
|
179 |
|
180 |
/* linux just needs an open() to trigger
|
181 |
* the partition table update
|
182 |
* but remember to load the module with max_part != 0 :
|
183 |
* modprobe nbd max_part=63
|
184 |
*/
|
185 |
nbd = open(device, O_RDWR); |
186 |
if (nbd != -1) |
187 |
close(nbd); |
188 |
exit(0);
|
189 |
} |
190 |
} |
191 |
|
192 |
int main(int argc, char **argv) |
193 |
{ |
194 |
BlockDriverState *bs; |
195 |
off_t dev_offset = 0;
|
196 |
off_t offset = 0;
|
197 |
uint32_t nbdflags = 0;
|
198 |
bool disconnect = false; |
199 |
const char *bindto = "0.0.0.0"; |
200 |
int port = NBD_DEFAULT_PORT;
|
201 |
struct sockaddr_in addr;
|
202 |
socklen_t addr_len = sizeof(addr);
|
203 |
off_t fd_size; |
204 |
char *device = NULL; |
205 |
char *socket = NULL; |
206 |
char sockpath[128]; |
207 |
const char *sopt = "hVb:o:p:rsnP:c:dvk:e:t"; |
208 |
struct option lopt[] = {
|
209 |
{ "help", 0, NULL, 'h' }, |
210 |
{ "version", 0, NULL, 'V' }, |
211 |
{ "bind", 1, NULL, 'b' }, |
212 |
{ "port", 1, NULL, 'p' }, |
213 |
{ "socket", 1, NULL, 'k' }, |
214 |
{ "offset", 1, NULL, 'o' }, |
215 |
{ "read-only", 0, NULL, 'r' }, |
216 |
{ "partition", 1, NULL, 'P' }, |
217 |
{ "connect", 1, NULL, 'c' }, |
218 |
{ "disconnect", 0, NULL, 'd' }, |
219 |
{ "snapshot", 0, NULL, 's' }, |
220 |
{ "nocache", 0, NULL, 'n' }, |
221 |
{ "shared", 1, NULL, 'e' }, |
222 |
{ "persistent", 0, NULL, 't' }, |
223 |
{ "verbose", 0, NULL, 'v' }, |
224 |
{ NULL, 0, NULL, 0 } |
225 |
}; |
226 |
int ch;
|
227 |
int opt_ind = 0; |
228 |
int li;
|
229 |
char *end;
|
230 |
int flags = BDRV_O_RDWR;
|
231 |
int partition = -1; |
232 |
int ret;
|
233 |
int shared = 1; |
234 |
uint8_t *data; |
235 |
fd_set fds; |
236 |
int *sharing_fds;
|
237 |
int fd;
|
238 |
int i;
|
239 |
int nb_fds = 0; |
240 |
int max_fd;
|
241 |
int persistent = 0; |
242 |
|
243 |
/* Set up a SIGTERM handler so that we exit with a nice status code. */
|
244 |
struct sigaction sa_sigterm;
|
245 |
int sigterm_fd[2]; |
246 |
if (qemu_pipe(sigterm_fd) == -1) { |
247 |
err(EXIT_FAILURE, "Error setting up communication pipe");
|
248 |
} |
249 |
|
250 |
sigterm_wfd = sigterm_fd[1];
|
251 |
memset(&sa_sigterm, 0, sizeof(sa_sigterm)); |
252 |
sa_sigterm.sa_handler = termsig_handler; |
253 |
sigaction(SIGTERM, &sa_sigterm, NULL);
|
254 |
|
255 |
while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) { |
256 |
switch (ch) {
|
257 |
case 's': |
258 |
flags |= BDRV_O_SNAPSHOT; |
259 |
break;
|
260 |
case 'n': |
261 |
flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; |
262 |
break;
|
263 |
case 'b': |
264 |
bindto = optarg; |
265 |
break;
|
266 |
case 'p': |
267 |
li = strtol(optarg, &end, 0);
|
268 |
if (*end) {
|
269 |
errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
|
270 |
} |
271 |
if (li < 1 || li > 65535) { |
272 |
errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
|
273 |
} |
274 |
port = (uint16_t)li; |
275 |
break;
|
276 |
case 'o': |
277 |
dev_offset = strtoll (optarg, &end, 0);
|
278 |
if (*end) {
|
279 |
errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
|
280 |
} |
281 |
if (dev_offset < 0) { |
282 |
errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
|
283 |
} |
284 |
break;
|
285 |
case 'r': |
286 |
nbdflags |= NBD_FLAG_READ_ONLY; |
287 |
flags &= ~BDRV_O_RDWR; |
288 |
break;
|
289 |
case 'P': |
290 |
partition = strtol(optarg, &end, 0);
|
291 |
if (*end)
|
292 |
errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
|
293 |
if (partition < 1 || partition > 8) |
294 |
errx(EXIT_FAILURE, "Invalid partition %d", partition);
|
295 |
break;
|
296 |
case 'k': |
297 |
socket = optarg; |
298 |
if (socket[0] != '/') |
299 |
errx(EXIT_FAILURE, "socket path must be absolute\n");
|
300 |
break;
|
301 |
case 'd': |
302 |
disconnect = true;
|
303 |
break;
|
304 |
case 'c': |
305 |
device = optarg; |
306 |
break;
|
307 |
case 'e': |
308 |
shared = strtol(optarg, &end, 0);
|
309 |
if (*end) {
|
310 |
errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
|
311 |
} |
312 |
if (shared < 1) { |
313 |
errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
|
314 |
} |
315 |
break;
|
316 |
case 't': |
317 |
persistent = 1;
|
318 |
break;
|
319 |
case 'v': |
320 |
verbose = 1;
|
321 |
break;
|
322 |
case 'V': |
323 |
version(argv[0]);
|
324 |
exit(0);
|
325 |
break;
|
326 |
case 'h': |
327 |
usage(argv[0]);
|
328 |
exit(0);
|
329 |
break;
|
330 |
case '?': |
331 |
errx(EXIT_FAILURE, "Try `%s --help' for more information.",
|
332 |
argv[0]);
|
333 |
} |
334 |
} |
335 |
|
336 |
if ((argc - optind) != 1) { |
337 |
errx(EXIT_FAILURE, "Invalid number of argument.\n"
|
338 |
"Try `%s --help' for more information.",
|
339 |
argv[0]);
|
340 |
} |
341 |
|
342 |
if (disconnect) {
|
343 |
fd = open(argv[optind], O_RDWR); |
344 |
if (fd == -1) |
345 |
err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
|
346 |
|
347 |
nbd_disconnect(fd); |
348 |
|
349 |
close(fd); |
350 |
|
351 |
printf("%s disconnected\n", argv[optind]);
|
352 |
|
353 |
return 0; |
354 |
} |
355 |
|
356 |
bdrv_init(); |
357 |
|
358 |
bs = bdrv_new("hda");
|
359 |
|
360 |
if ((ret = bdrv_open(bs, argv[optind], flags, NULL)) < 0) { |
361 |
errno = -ret; |
362 |
err(EXIT_FAILURE, "Failed to bdrv_open '%s'", argv[optind]);
|
363 |
} |
364 |
|
365 |
fd_size = bs->total_sectors * 512;
|
366 |
|
367 |
if (partition != -1 && |
368 |
find_partition(bs, partition, &dev_offset, &fd_size)) |
369 |
err(EXIT_FAILURE, "Could not find partition %d", partition);
|
370 |
|
371 |
if (device) {
|
372 |
pid_t pid; |
373 |
int sock;
|
374 |
|
375 |
/* want to fail before daemonizing */
|
376 |
if (access(device, R_OK|W_OK) == -1) { |
377 |
err(EXIT_FAILURE, "Could not access '%s'", device);
|
378 |
} |
379 |
|
380 |
if (!verbose) {
|
381 |
/* detach client and server */
|
382 |
if (qemu_daemon(0, 0) == -1) { |
383 |
err(EXIT_FAILURE, "Failed to daemonize");
|
384 |
} |
385 |
} |
386 |
|
387 |
if (socket == NULL) { |
388 |
snprintf(sockpath, sizeof(sockpath), SOCKET_PATH,
|
389 |
basename(device)); |
390 |
socket = sockpath; |
391 |
} |
392 |
|
393 |
pid = fork(); |
394 |
if (pid < 0) |
395 |
return 1; |
396 |
if (pid != 0) { |
397 |
off_t size; |
398 |
size_t blocksize; |
399 |
|
400 |
ret = 0;
|
401 |
bdrv_close(bs); |
402 |
|
403 |
do {
|
404 |
sock = unix_socket_outgoing(socket); |
405 |
if (sock == -1) { |
406 |
if (errno != ENOENT && errno != ECONNREFUSED) {
|
407 |
ret = 1;
|
408 |
goto out;
|
409 |
} |
410 |
sleep(1); /* wait children */ |
411 |
} |
412 |
} while (sock == -1); |
413 |
|
414 |
fd = open(device, O_RDWR); |
415 |
if (fd == -1) { |
416 |
ret = 1;
|
417 |
goto out;
|
418 |
} |
419 |
|
420 |
ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
|
421 |
&size, &blocksize); |
422 |
if (ret == -1) { |
423 |
ret = 1;
|
424 |
goto out;
|
425 |
} |
426 |
|
427 |
ret = nbd_init(fd, sock, nbdflags, size, blocksize); |
428 |
if (ret == -1) { |
429 |
ret = 1;
|
430 |
goto out;
|
431 |
} |
432 |
|
433 |
printf("NBD device %s is now connected to file %s\n",
|
434 |
device, argv[optind]); |
435 |
|
436 |
/* update partition table */
|
437 |
|
438 |
show_parts(device); |
439 |
|
440 |
ret = nbd_client(fd); |
441 |
if (ret) {
|
442 |
ret = 1;
|
443 |
} |
444 |
close(fd); |
445 |
out:
|
446 |
kill(pid, SIGTERM); |
447 |
|
448 |
return ret;
|
449 |
} |
450 |
/* children */
|
451 |
} |
452 |
|
453 |
sharing_fds = g_malloc((shared + 1) * sizeof(int)); |
454 |
|
455 |
if (socket) {
|
456 |
sharing_fds[0] = unix_socket_incoming(socket);
|
457 |
} else {
|
458 |
sharing_fds[0] = tcp_socket_incoming(bindto, port);
|
459 |
} |
460 |
|
461 |
if (sharing_fds[0] == -1) |
462 |
return 1; |
463 |
max_fd = sharing_fds[0];
|
464 |
nb_fds++; |
465 |
|
466 |
data = qemu_blockalign(bs, NBD_BUFFER_SIZE); |
467 |
if (data == NULL) { |
468 |
errx(EXIT_FAILURE, "Cannot allocate data buffer");
|
469 |
} |
470 |
|
471 |
do {
|
472 |
FD_ZERO(&fds); |
473 |
FD_SET(sigterm_fd[0], &fds);
|
474 |
for (i = 0; i < nb_fds; i++) |
475 |
FD_SET(sharing_fds[i], &fds); |
476 |
|
477 |
do {
|
478 |
ret = select(max_fd + 1, &fds, NULL, NULL, NULL); |
479 |
} while (ret == -1 && errno == EINTR); |
480 |
if (ret == -1 || FD_ISSET(sigterm_fd[0], &fds)) { |
481 |
break;
|
482 |
} |
483 |
|
484 |
if (FD_ISSET(sharing_fds[0], &fds)) |
485 |
ret--; |
486 |
for (i = 1; i < nb_fds && ret; i++) { |
487 |
if (FD_ISSET(sharing_fds[i], &fds)) {
|
488 |
if (nbd_trip(bs, sharing_fds[i], fd_size, dev_offset,
|
489 |
&offset, nbdflags, data, NBD_BUFFER_SIZE) != 0) {
|
490 |
close(sharing_fds[i]); |
491 |
nb_fds--; |
492 |
sharing_fds[i] = sharing_fds[nb_fds]; |
493 |
i--; |
494 |
} |
495 |
ret--; |
496 |
} |
497 |
} |
498 |
/* new connection ? */
|
499 |
if (FD_ISSET(sharing_fds[0], &fds)) { |
500 |
if (nb_fds < shared + 1) { |
501 |
sharing_fds[nb_fds] = accept(sharing_fds[0],
|
502 |
(struct sockaddr *)&addr,
|
503 |
&addr_len); |
504 |
if (sharing_fds[nb_fds] != -1 && |
505 |
nbd_negotiate(sharing_fds[nb_fds], fd_size, nbdflags) != -1) {
|
506 |
if (sharing_fds[nb_fds] > max_fd)
|
507 |
max_fd = sharing_fds[nb_fds]; |
508 |
nb_fds++; |
509 |
} |
510 |
} |
511 |
} |
512 |
} while (persistent || nb_fds > 1); |
513 |
qemu_vfree(data); |
514 |
|
515 |
close(sharing_fds[0]);
|
516 |
bdrv_close(bs); |
517 |
g_free(sharing_fds); |
518 |
if (socket)
|
519 |
unlink(socket); |
520 |
|
521 |
return 0; |
522 |
} |