xref: /illumos-kvm-cmd/qemu-nbd.c (revision 68396ea9)
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 verbose;
40 
usage(const char * name)41 static void usage(const char *name)
42 {
43     printf(
44 "Usage: %s [OPTIONS] FILE\n"
45 "QEMU Disk Network Block Device Server\n"
46 "\n"
47 "  -p, --port=PORT      port to listen on (default `%d')\n"
48 "  -o, --offset=OFFSET  offset into the image\n"
49 "  -b, --bind=IFACE     interface to bind to (default `0.0.0.0')\n"
50 "  -k, --socket=PATH    path to the unix socket\n"
51 "                       (default '"SOCKET_PATH"')\n"
52 "  -r, --read-only      export read-only\n"
53 "  -P, --partition=NUM  only expose partition NUM\n"
54 "  -s, --snapshot       use snapshot file\n"
55 "  -n, --nocache        disable host cache\n"
56 "  -c, --connect=DEV    connect FILE to the local NBD device DEV\n"
57 "  -d, --disconnect     disconnect the specified device\n"
58 "  -e, --shared=NUM     device can be shared by NUM clients (default '1')\n"
59 "  -t, --persistent     don't exit on the last connection\n"
60 "  -v, --verbose        display extra debugging information\n"
61 "  -h, --help           display this help and exit\n"
62 "  -V, --version        output version information and exit\n"
63 "\n"
64 "Report bugs to <anthony@codemonkey.ws>\n"
65     , name, NBD_DEFAULT_PORT, "DEVICE");
66 }
67 
version(const char * name)68 static void version(const char *name)
69 {
70     printf(
71 "%s version 0.0.1\n"
72 "Written by Anthony Liguori.\n"
73 "\n"
74 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
75 "This is free software; see the source for copying conditions.  There is NO\n"
76 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
77     , name);
78 }
79 
80 struct partition_record
81 {
82     uint8_t bootable;
83     uint8_t start_head;
84     uint32_t start_cylinder;
85     uint8_t start_sector;
86     uint8_t system;
87     uint8_t end_head;
88     uint8_t end_cylinder;
89     uint8_t end_sector;
90     uint32_t start_sector_abs;
91     uint32_t nb_sectors_abs;
92 };
93 
read_partition(uint8_t * p,struct partition_record * r)94 static void read_partition(uint8_t *p, struct partition_record *r)
95 {
96     r->bootable = p[0];
97     r->start_head = p[1];
98     r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
99     r->start_sector = p[2] & 0x3f;
100     r->system = p[4];
101     r->end_head = p[5];
102     r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
103     r->end_sector = p[6] & 0x3f;
104     r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
105     r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
106 }
107 
find_partition(BlockDriverState * bs,int partition,off_t * offset,off_t * size)108 static int find_partition(BlockDriverState *bs, int partition,
109                           off_t *offset, off_t *size)
110 {
111     struct partition_record mbr[4];
112     uint8_t data[512];
113     int i;
114     int ext_partnum = 4;
115     int ret;
116 
117     if ((ret = bdrv_read(bs, 0, data, 1)) < 0) {
118         errno = -ret;
119         err(EXIT_FAILURE, "error while reading");
120     }
121 
122     if (data[510] != 0x55 || data[511] != 0xaa) {
123         errno = -EINVAL;
124         return -1;
125     }
126 
127     for (i = 0; i < 4; i++) {
128         read_partition(&data[446 + 16 * i], &mbr[i]);
129 
130         if (!mbr[i].nb_sectors_abs)
131             continue;
132 
133         if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
134             struct partition_record ext[4];
135             uint8_t data1[512];
136             int j;
137 
138             if ((ret = bdrv_read(bs, mbr[i].start_sector_abs, data1, 1)) < 0) {
139                 errno = -ret;
140                 err(EXIT_FAILURE, "error while reading");
141             }
142 
143             for (j = 0; j < 4; j++) {
144                 read_partition(&data1[446 + 16 * j], &ext[j]);
145                 if (!ext[j].nb_sectors_abs)
146                     continue;
147 
148                 if ((ext_partnum + j + 1) == partition) {
149                     *offset = (uint64_t)ext[j].start_sector_abs << 9;
150                     *size = (uint64_t)ext[j].nb_sectors_abs << 9;
151                     return 0;
152                 }
153             }
154             ext_partnum += 4;
155         } else if ((i + 1) == partition) {
156             *offset = (uint64_t)mbr[i].start_sector_abs << 9;
157             *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
158             return 0;
159         }
160     }
161 
162     errno = -ENOENT;
163     return -1;
164 }
165 
show_parts(const char * device)166 static void show_parts(const char *device)
167 {
168     if (fork() == 0) {
169         int nbd;
170 
171         /* linux just needs an open() to trigger
172          * the partition table update
173          * but remember to load the module with max_part != 0 :
174          *     modprobe nbd max_part=63
175          */
176         nbd = open(device, O_RDWR);
177         if (nbd != -1)
178               close(nbd);
179         exit(0);
180     }
181 }
182 
main(int argc,char ** argv)183 int main(int argc, char **argv)
184 {
185     BlockDriverState *bs;
186     off_t dev_offset = 0;
187     off_t offset = 0;
188     bool readonly = false;
189     bool disconnect = false;
190     const char *bindto = "0.0.0.0";
191     int port = NBD_DEFAULT_PORT;
192     struct sockaddr_in addr;
193     socklen_t addr_len = sizeof(addr);
194     off_t fd_size;
195     char *device = NULL;
196     char *socket = NULL;
197     char sockpath[128];
198     const char *sopt = "hVb:o:p:rsnP:c:dvk:e:t";
199     struct option lopt[] = {
200         { "help", 0, NULL, 'h' },
201         { "version", 0, NULL, 'V' },
202         { "bind", 1, NULL, 'b' },
203         { "port", 1, NULL, 'p' },
204         { "socket", 1, NULL, 'k' },
205         { "offset", 1, NULL, 'o' },
206         { "read-only", 0, NULL, 'r' },
207         { "partition", 1, NULL, 'P' },
208         { "connect", 1, NULL, 'c' },
209         { "disconnect", 0, NULL, 'd' },
210         { "snapshot", 0, NULL, 's' },
211         { "nocache", 0, NULL, 'n' },
212         { "shared", 1, NULL, 'e' },
213         { "persistent", 0, NULL, 't' },
214         { "verbose", 0, NULL, 'v' },
215         { NULL, 0, NULL, 0 }
216     };
217     int ch;
218     int opt_ind = 0;
219     int li;
220     char *end;
221     int flags = BDRV_O_RDWR;
222     int partition = -1;
223     int ret;
224     int shared = 1;
225     uint8_t *data;
226     fd_set fds;
227     int *sharing_fds;
228     int fd;
229     int i;
230     int nb_fds = 0;
231     int max_fd;
232     int persistent = 0;
233     uint32_t nbdflags;
234 
235     while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
236         switch (ch) {
237         case 's':
238             flags |= BDRV_O_SNAPSHOT;
239             break;
240         case 'n':
241             flags |= BDRV_O_NOCACHE;
242             break;
243         case 'b':
244             bindto = optarg;
245             break;
246         case 'p':
247             li = strtol(optarg, &end, 0);
248             if (*end) {
249                 errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
250             }
251             if (li < 1 || li > 65535) {
252                 errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
253             }
254             port = (uint16_t)li;
255             break;
256         case 'o':
257                 dev_offset = strtoll (optarg, &end, 0);
258             if (*end) {
259                 errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
260             }
261             if (dev_offset < 0) {
262                 errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
263             }
264             break;
265         case 'r':
266             readonly = true;
267             flags &= ~BDRV_O_RDWR;
268             break;
269         case 'P':
270             partition = strtol(optarg, &end, 0);
271             if (*end)
272                 errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
273             if (partition < 1 || partition > 8)
274                 errx(EXIT_FAILURE, "Invalid partition %d", partition);
275             break;
276         case 'k':
277             socket = optarg;
278             if (socket[0] != '/')
279                 errx(EXIT_FAILURE, "socket path must be absolute\n");
280             break;
281         case 'd':
282             disconnect = true;
283             break;
284         case 'c':
285             device = optarg;
286             break;
287         case 'e':
288             shared = strtol(optarg, &end, 0);
289             if (*end) {
290                 errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
291             }
292             if (shared < 1) {
293                 errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
294             }
295             break;
296 	case 't':
297 	    persistent = 1;
298 	    break;
299         case 'v':
300             verbose = 1;
301             break;
302         case 'V':
303             version(argv[0]);
304             exit(0);
305             break;
306         case 'h':
307             usage(argv[0]);
308             exit(0);
309             break;
310         case '?':
311             errx(EXIT_FAILURE, "Try `%s --help' for more information.",
312                  argv[0]);
313         }
314     }
315 
316     if ((argc - optind) != 1) {
317         errx(EXIT_FAILURE, "Invalid number of argument.\n"
318              "Try `%s --help' for more information.",
319              argv[0]);
320     }
321 
322     if (disconnect) {
323         fd = open(argv[optind], O_RDWR);
324         if (fd == -1)
325             err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
326 
327         nbd_disconnect(fd);
328 
329         close(fd);
330 
331         printf("%s disconnected\n", argv[optind]);
332 
333 	return 0;
334     }
335 
336     bdrv_init();
337 
338     bs = bdrv_new("hda");
339 
340     if ((ret = bdrv_open(bs, argv[optind], flags, NULL)) < 0) {
341         errno = -ret;
342         err(EXIT_FAILURE, "Failed to bdrv_open '%s'", argv[optind]);
343     }
344 
345     fd_size = bs->total_sectors * 512;
346 
347     if (partition != -1 &&
348         find_partition(bs, partition, &dev_offset, &fd_size))
349         err(EXIT_FAILURE, "Could not find partition %d", partition);
350 
351     if (device) {
352         pid_t pid;
353         int sock;
354 
355         /* want to fail before daemonizing */
356         if (access(device, R_OK|W_OK) == -1) {
357             err(EXIT_FAILURE, "Could not access '%s'", device);
358         }
359 
360         if (!verbose) {
361             /* detach client and server */
362             if (daemon(0, 0) == -1) {
363                 err(EXIT_FAILURE, "Failed to daemonize");
364             }
365         }
366 
367         if (socket == NULL) {
368             snprintf(sockpath, sizeof(sockpath), SOCKET_PATH,
369                      basename(device));
370             socket = sockpath;
371         }
372 
373         pid = fork();
374         if (pid < 0)
375             return 1;
376         if (pid != 0) {
377             off_t size;
378             size_t blocksize;
379 
380             ret = 0;
381             bdrv_close(bs);
382 
383             do {
384                 sock = unix_socket_outgoing(socket);
385                 if (sock == -1) {
386                     if (errno != ENOENT && errno != ECONNREFUSED) {
387                         ret = 1;
388                         goto out;
389                     }
390                     sleep(1);	/* wait children */
391                 }
392             } while (sock == -1);
393 
394             fd = open(device, O_RDWR);
395             if (fd == -1) {
396                 ret = 1;
397                 goto out;
398             }
399 
400             ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
401 					&size, &blocksize);
402             if (ret == -1) {
403                 ret = 1;
404                 goto out;
405             }
406 
407             ret = nbd_init(fd, sock, size, blocksize);
408             if (ret == -1) {
409                 ret = 1;
410                 goto out;
411             }
412 
413             printf("NBD device %s is now connected to file %s\n",
414                     device, argv[optind]);
415 
416 	    /* update partition table */
417 
418             show_parts(device);
419 
420             ret = nbd_client(fd);
421             if (ret) {
422                 ret = 1;
423             }
424             close(fd);
425  out:
426             kill(pid, SIGTERM);
427             unlink(socket);
428 
429             return ret;
430         }
431         /* children */
432     }
433 
434     sharing_fds = qemu_malloc((shared + 1) * sizeof(int));
435 
436     if (socket) {
437         sharing_fds[0] = unix_socket_incoming(socket);
438     } else {
439         sharing_fds[0] = tcp_socket_incoming(bindto, port);
440     }
441 
442     if (sharing_fds[0] == -1)
443         return 1;
444     max_fd = sharing_fds[0];
445     nb_fds++;
446 
447     data = qemu_blockalign(bs, NBD_BUFFER_SIZE);
448     if (data == NULL)
449         errx(EXIT_FAILURE, "Cannot allocate data buffer");
450 
451     do {
452 
453         FD_ZERO(&fds);
454         for (i = 0; i < nb_fds; i++)
455             FD_SET(sharing_fds[i], &fds);
456 
457         ret = select(max_fd + 1, &fds, NULL, NULL, NULL);
458         if (ret == -1)
459             break;
460 
461         if (FD_ISSET(sharing_fds[0], &fds))
462             ret--;
463         for (i = 1; i < nb_fds && ret; i++) {
464             if (FD_ISSET(sharing_fds[i], &fds)) {
465                 if (nbd_trip(bs, sharing_fds[i], fd_size, dev_offset,
466                     &offset, readonly, data, NBD_BUFFER_SIZE) != 0) {
467                     close(sharing_fds[i]);
468                     nb_fds--;
469                     sharing_fds[i] = sharing_fds[nb_fds];
470                     i--;
471                 }
472                 ret--;
473             }
474         }
475         /* new connection ? */
476         if (FD_ISSET(sharing_fds[0], &fds)) {
477             if (nb_fds < shared + 1) {
478                 sharing_fds[nb_fds] = accept(sharing_fds[0],
479                                              (struct sockaddr *)&addr,
480                                              &addr_len);
481                 if (sharing_fds[nb_fds] != -1 &&
482                     nbd_negotiate(sharing_fds[nb_fds], fd_size) != -1) {
483                         if (sharing_fds[nb_fds] > max_fd)
484                             max_fd = sharing_fds[nb_fds];
485                         nb_fds++;
486                 }
487             }
488         }
489     } while (persistent || nb_fds > 1);
490     qemu_vfree(data);
491 
492     close(sharing_fds[0]);
493     bdrv_close(bs);
494     qemu_free(sharing_fds);
495     if (socket)
496         unlink(socket);
497 
498     return 0;
499 }
500