Statistics
| Branch: | Revision:

root / qemu-nbd.c @ 3b05a8e9

History | View | Annotate | Download (12.8 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, write to the Free Software
17
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
18
 */
19

    
20
#include <qemu-common.h>
21
#include "block_int.h"
22
#include "nbd.h"
23

    
24
#include <stdarg.h>
25
#include <stdio.h>
26
#include <getopt.h>
27
#include <err.h>
28
#include <sys/types.h>
29
#include <sys/socket.h>
30
#include <netinet/in.h>
31
#include <netinet/tcp.h>
32
#include <arpa/inet.h>
33
#include <signal.h>
34

    
35
#define SOCKET_PATH    "/var/lock/qemu-nbd-%s"
36

    
37
#define NBD_BUFFER_SIZE (1024*1024)
38

    
39
int verbose;
40

    
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 `1024')\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
"  -v, --verbose        display extra debugging information\n"
60
"  -h, --help           display this help and exit\n"
61
"  -V, --version        output version information and exit\n"
62
"\n"
63
"Report bugs to <anthony@codemonkey.ws>\n"
64
    , name, "DEVICE");
65
}
66

    
67
static void version(const char *name)
68
{
69
    printf(
70
"qemu-nbd version 0.0.1\n"
71
"Written by Anthony Liguori.\n"
72
"\n"
73
"Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
74
"This is free software; see the source for copying conditions.  There is NO\n"
75
"warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
76
    );
77
}
78

    
79
struct partition_record
80
{
81
    uint8_t bootable;
82
    uint8_t start_head;
83
    uint32_t start_cylinder;
84
    uint8_t start_sector;
85
    uint8_t system;
86
    uint8_t end_head;
87
    uint8_t end_cylinder;
88
    uint8_t end_sector;
89
    uint32_t start_sector_abs;
90
    uint32_t nb_sectors_abs;
91
};
92

    
93
static void read_partition(uint8_t *p, struct partition_record *r)
94
{
95
    r->bootable = p[0];
96
    r->start_head = p[1];
97
    r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
98
    r->start_sector = p[2] & 0x3f;
99
    r->system = p[4];
100
    r->end_head = p[5];
101
    r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
102
    r->end_sector = p[6] & 0x3f;
103
    r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
104
    r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
105
}
106

    
107
static int find_partition(BlockDriverState *bs, int partition,
108
                          off_t *offset, off_t *size)
109
{
110
    struct partition_record mbr[4];
111
    uint8_t data[512];
112
    int i;
113
    int ext_partnum = 4;
114

    
115
    if (bdrv_read(bs, 0, data, 1))
116
        errx(EINVAL, "error while reading");
117

    
118
    if (data[510] != 0x55 || data[511] != 0xaa) {
119
        errno = -EINVAL;
120
        return -1;
121
    }
122

    
123
    for (i = 0; i < 4; i++) {
124
        read_partition(&data[446 + 16 * i], &mbr[i]);
125

    
126
        if (!mbr[i].nb_sectors_abs)
127
            continue;
128

    
129
        if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
130
            struct partition_record ext[4];
131
            uint8_t data1[512];
132
            int j;
133

    
134
            if (bdrv_read(bs, mbr[i].start_sector_abs, data1, 1))
135
                errx(EINVAL, "error while reading");
136

    
137
            for (j = 0; j < 4; j++) {
138
                read_partition(&data1[446 + 16 * j], &ext[j]);
139
                if (!ext[j].nb_sectors_abs)
140
                    continue;
141

    
142
                if ((ext_partnum + j + 1) == partition) {
143
                    *offset = (uint64_t)ext[j].start_sector_abs << 9;
144
                    *size = (uint64_t)ext[j].nb_sectors_abs << 9;
145
                    return 0;
146
                }
147
            }
148
            ext_partnum += 4;
149
        } else if ((i + 1) == partition) {
150
            *offset = (uint64_t)mbr[i].start_sector_abs << 9;
151
            *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
152
            return 0;
153
        }
154
    }
155

    
156
    errno = -ENOENT;
157
    return -1;
158
}
159

    
160
static void show_parts(const char *device)
161
{
162
    if (fork() == 0) {
163
        int nbd;
164

    
165
        /* linux just needs an open() to trigger
166
         * the partition table update
167
         * but remember to load the module with max_part != 0 :
168
         *     modprobe nbd max_part=63
169
         */
170
        nbd = open(device, O_RDWR);
171
        if (nbd != -1)
172
              close(nbd);
173
        exit(0);
174
    }
175
}
176

    
177
int main(int argc, char **argv)
178
{
179
    BlockDriverState *bs;
180
    off_t dev_offset = 0;
181
    off_t offset = 0;
182
    bool readonly = false;
183
    bool disconnect = false;
184
    const char *bindto = "0.0.0.0";
185
    int port = 1024;
186
    struct sockaddr_in addr;
187
    socklen_t addr_len = sizeof(addr);
188
    off_t fd_size;
189
    char *device = NULL;
190
    char *socket = NULL;
191
    char sockpath[128];
192
    const char *sopt = "hVbo:p:rsnP:c:dvk:e:";
193
    struct option lopt[] = {
194
        { "help", 0, 0, 'h' },
195
        { "version", 0, 0, 'V' },
196
        { "bind", 1, 0, 'b' },
197
        { "port", 1, 0, 'p' },
198
        { "socket", 1, 0, 'k' },
199
        { "offset", 1, 0, 'o' },
200
        { "read-only", 0, 0, 'r' },
201
        { "partition", 1, 0, 'P' },
202
        { "connect", 1, 0, 'c' },
203
        { "disconnect", 0, 0, 'd' },
204
        { "snapshot", 0, 0, 's' },
205
        { "nocache", 0, 0, 'n' },
206
        { "shared", 1, 0, 'e' },
207
        { "verbose", 0, 0, 'v' },
208
        { NULL, 0, 0, 0 }
209
    };
210
    int ch;
211
    int opt_ind = 0;
212
    int li;
213
    char *end;
214
    int flags = 0;
215
    int partition = -1;
216
    int ret;
217
    int shared = 1;
218
    uint8_t *data;
219
    fd_set fds;
220
    int *sharing_fds;
221
    int fd;
222
    int i;
223
    int nb_fds = 0;
224
    int max_fd;
225

    
226
    while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
227
        switch (ch) {
228
        case 's':
229
            flags |= BDRV_O_SNAPSHOT;
230
            break;
231
        case 'n':
232
            flags |= BDRV_O_DIRECT;
233
            break;
234
        case 'b':
235
            bindto = optarg;
236
            break;
237
        case 'p':
238
            li = strtol(optarg, &end, 0);
239
            if (*end) {
240
                errx(EINVAL, "Invalid port `%s'", optarg);
241
            }
242
            if (li < 1 || li > 65535) {
243
                errx(EINVAL, "Port out of range `%s'", optarg);
244
            }
245
            port = (uint16_t)li;
246
            break;
247
        case 'o':
248
                dev_offset = strtoll (optarg, &end, 0);
249
            if (*end) {
250
                errx(EINVAL, "Invalid offset `%s'", optarg);
251
            }
252
            if (dev_offset < 0) {
253
                errx(EINVAL, "Offset must be positive `%s'", optarg);
254
            }
255
            break;
256
        case 'r':
257
            readonly = true;
258
            break;
259
        case 'P':
260
            partition = strtol(optarg, &end, 0);
261
            if (*end)
262
                errx(EINVAL, "Invalid partition `%s'", optarg);
263
            if (partition < 1 || partition > 8)
264
                errx(EINVAL, "Invalid partition %d", partition);
265
            break;
266
        case 'k':
267
            socket = optarg;
268
            if (socket[0] != '/')
269
                errx(EINVAL, "socket path must be absolute\n");
270
            break;
271
        case 'd':
272
            disconnect = true;
273
            break;
274
        case 'c':
275
            device = optarg;
276
            break;
277
        case 'e':
278
            shared = strtol(optarg, &end, 0);
279
            if (*end) {
280
                errx(EINVAL, "Invalid shared device number '%s'", optarg);
281
            }
282
            if (shared < 1) {
283
                errx(EINVAL, "Shared device number must be greater than 0\n");
284
            }
285
            break;
286
        case 'v':
287
            verbose = 1;
288
            break;
289
        case 'V':
290
            version(argv[0]);
291
            exit(0);
292
            break;
293
        case 'h':
294
            usage(argv[0]);
295
            exit(0);
296
            break;
297
        case '?':
298
            errx(EINVAL, "Try `%s --help' for more information.",
299
                 argv[0]);
300
        }
301
    }
302

    
303
    if ((argc - optind) != 1) {
304
        errx(EINVAL, "Invalid number of argument.\n"
305
             "Try `%s --help' for more information.",
306
             argv[0]);
307
    }
308

    
309
    if (disconnect) {
310
        fd = open(argv[optind], O_RDWR);
311
        if (fd == -1)
312
            errx(errno, "Cannot open %s", argv[optind]);
313

    
314
        nbd_disconnect(fd);
315

    
316
        close(fd);
317

    
318
        printf("%s disconnected\n", argv[optind]);
319

    
320
        return 0;
321
    }
322

    
323
    bdrv_init();
324

    
325
    bs = bdrv_new("hda");
326
    if (bs == NULL)
327
        return 1;
328

    
329
    if (bdrv_open(bs, argv[optind], flags) == -1)
330
        return 1;
331

    
332
    fd_size = bs->total_sectors * 512;
333

    
334
    if (partition != -1 &&
335
        find_partition(bs, partition, &dev_offset, &fd_size))
336
        errx(errno, "Could not find partition %d", partition);
337

    
338
    if (device) {
339
        pid_t pid;
340
        int sock;
341

    
342
        if (!verbose)
343
            daemon(0, 0);        /* detach client and server */
344

    
345
        if (socket == NULL) {
346
            sprintf(sockpath, SOCKET_PATH, basename(device));
347
            socket = sockpath;
348
        }
349

    
350
        pid = fork();
351
        if (pid < 0)
352
            return 1;
353
        if (pid != 0) {
354
            off_t size;
355
            size_t blocksize;
356

    
357
            ret = 0;
358
            bdrv_close(bs);
359

    
360
            do {
361
                sock = unix_socket_outgoing(socket);
362
                if (sock == -1) {
363
                    if (errno != ENOENT && errno != ECONNREFUSED)
364
                        goto out;
365
                    sleep(1);        /* wait children */
366
                }
367
            } while (sock == -1);
368

    
369
            fd = open(device, O_RDWR);
370
            if (fd == -1) {
371
                ret = 1;
372
                goto out;
373
            }
374

    
375
            ret = nbd_receive_negotiate(sock, &size, &blocksize);
376
            if (ret == -1) {
377
                ret = 1;
378
                goto out;
379
            }
380

    
381
            ret = nbd_init(fd, sock, size, blocksize);
382
            if (ret == -1) {
383
                ret = 1;
384
                goto out;
385
            }
386

    
387
            printf("NBD device %s is now connected to file %s\n",
388
                    device, argv[optind]);
389

    
390
            /* update partition table */
391

    
392
            show_parts(device);
393

    
394
            nbd_client(fd, sock);
395
            close(fd);
396
 out:
397
            kill(pid, SIGTERM);
398
            unlink(socket);
399

    
400
            return ret;
401
        }
402
        /* children */
403
    }
404

    
405
    sharing_fds = qemu_malloc((shared + 1) * sizeof(int));
406
    if (sharing_fds == NULL)
407
        errx(ENOMEM, "Cannot allocate sharing fds");
408

    
409
    if (socket) {
410
        sharing_fds[0] = unix_socket_incoming(socket);
411
    } else {
412
        sharing_fds[0] = tcp_socket_incoming(bindto, port);
413
    }
414

    
415
    if (sharing_fds[0] == -1)
416
        return 1;
417
    max_fd = sharing_fds[0];
418
    nb_fds++;
419

    
420
    data = qemu_memalign(512, NBD_BUFFER_SIZE);
421
    if (data == NULL)
422
        errx(ENOMEM, "Cannot allocate data buffer");
423

    
424
    do {
425

    
426
        FD_ZERO(&fds);
427
        for (i = 0; i < nb_fds; i++)
428
            FD_SET(sharing_fds[i], &fds);
429

    
430
        ret = select(max_fd + 1, &fds, NULL, NULL, NULL);
431
        if (ret == -1)
432
            break;
433

    
434
        if (FD_ISSET(sharing_fds[0], &fds))
435
            ret--;
436
        for (i = 1; i < nb_fds && ret; i++) {
437
            if (FD_ISSET(sharing_fds[i], &fds)) {
438
                if (nbd_trip(bs, sharing_fds[i], fd_size, dev_offset,
439
                    &offset, readonly, data, NBD_BUFFER_SIZE) != 0) {
440
                    close(sharing_fds[i]);
441
                    nb_fds--;
442
                    sharing_fds[i] = sharing_fds[nb_fds];
443
                    i--;
444
                }
445
                ret--;
446
            }
447
        }
448
        /* new connection ? */
449
        if (FD_ISSET(sharing_fds[0], &fds)) {
450
            if (nb_fds < shared + 1) {
451
                sharing_fds[nb_fds] = accept(sharing_fds[0],
452
                                             (struct sockaddr *)&addr,
453
                                             &addr_len);
454
                if (sharing_fds[nb_fds] != -1 &&
455
                    nbd_negotiate(bs, sharing_fds[nb_fds], fd_size) != -1) {
456
                        if (sharing_fds[nb_fds] > max_fd)
457
                            max_fd = sharing_fds[nb_fds];
458
                        nb_fds++;
459
                }
460
            }
461
        }
462
    } while (nb_fds > 1);
463
    qemu_free(data);
464

    
465
    close(sharing_fds[0]);
466
    bdrv_close(bs);
467
    qemu_free(sharing_fds);
468
    if (socket)
469
        unlink(socket);
470

    
471
    return 0;
472
}