2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
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.
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.
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/>.
19 #include "qemu-common.h"
20 #include "block/block.h"
21 #include "block/nbd.h"
22 #include "qemu/main-loop.h"
23 #include "qemu/sockets.h"
24 #include "qemu/error-report.h"
25 #include "block/snapshot.h"
31 #include <sys/types.h>
32 #include <sys/socket.h>
33 #include <netinet/in.h>
34 #include <netinet/tcp.h>
35 #include <arpa/inet.h>
40 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
41 #define QEMU_NBD_OPT_CACHE 1
42 #define QEMU_NBD_OPT_AIO 2
43 #define QEMU_NBD_OPT_DISCARD 3
45 static NBDExport
*exp
;
48 static char *sockpath
;
49 static int persistent
= 0;
50 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
51 static int shared
= 1;
54 static void usage(const char *name
)
57 "Usage: %s [OPTIONS] FILE\n"
58 "QEMU Disk Network Block Device Server\n"
60 " -h, --help display this help and exit\n"
61 " -V, --version output version information and exit\n"
63 "Connection properties:\n"
64 " -p, --port=PORT port to listen on (default `%d')\n"
65 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
66 " -k, --socket=PATH path to the unix socket\n"
67 " (default '"SOCKET_PATH
"')\n"
68 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
69 " -t, --persistent don't exit on the last connection\n"
70 " -v, --verbose display extra debugging information\n"
72 "Exposing part of the image:\n"
73 " -o, --offset=OFFSET offset into the image\n"
74 " -P, --partition=NUM only expose partition NUM\n"
77 "Kernel NBD client support:\n"
78 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
79 " -d, --disconnect disconnect the specified device\n"
83 "Block device options:\n"
84 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
85 " -r, --read-only export read-only\n"
86 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
87 " file with backing_file=FILE, redirect the write to\n"
88 " the temporary one\n"
89 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
90 " load an internal snapshot inside FILE and export it\n"
91 " as an read-only device, SNAPSHOT_PARAM format is\n"
92 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
94 " -n, --nocache disable host cache\n"
95 " --cache=MODE set cache mode (none, writeback, ...)\n"
96 #ifdef CONFIG_LINUX_AIO
97 " --aio=MODE set AIO mode (native or threads)\n"
100 "Report bugs to <qemu-devel@nongnu.org>\n"
101 , name
, NBD_DEFAULT_PORT
, "DEVICE");
104 static void version(const char *name
)
108 "Written by Anthony Liguori.\n"
110 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
111 "This is free software; see the source for copying conditions. There is NO\n"
112 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
116 struct partition_record
120 uint32_t start_cylinder
;
121 uint8_t start_sector
;
124 uint8_t end_cylinder
;
126 uint32_t start_sector_abs
;
127 uint32_t nb_sectors_abs
;
130 static void read_partition(uint8_t *p
, struct partition_record
*r
)
133 r
->start_head
= p
[1];
134 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
135 r
->start_sector
= p
[2] & 0x3f;
138 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
139 r
->end_sector
= p
[6] & 0x3f;
140 r
->start_sector_abs
= p
[8] | p
[9] << 8 | p
[10] << 16 | p
[11] << 24;
141 r
->nb_sectors_abs
= p
[12] | p
[13] << 8 | p
[14] << 16 | p
[15] << 24;
144 static int find_partition(BlockDriverState
*bs
, int partition
,
145 off_t
*offset
, off_t
*size
)
147 struct partition_record mbr
[4];
153 if ((ret
= bdrv_read(bs
, 0, data
, 1)) < 0) {
155 err(EXIT_FAILURE
, "error while reading");
158 if (data
[510] != 0x55 || data
[511] != 0xaa) {
162 for (i
= 0; i
< 4; i
++) {
163 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
165 if (!mbr
[i
].nb_sectors_abs
)
168 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
169 struct partition_record ext
[4];
173 if ((ret
= bdrv_read(bs
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
175 err(EXIT_FAILURE
, "error while reading");
178 for (j
= 0; j
< 4; j
++) {
179 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
180 if (!ext
[j
].nb_sectors_abs
)
183 if ((ext_partnum
+ j
+ 1) == partition
) {
184 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
185 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
190 } else if ((i
+ 1) == partition
) {
191 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
192 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
200 static void termsig_handler(int signum
)
206 static void combine_addr(char *buf
, size_t len
, const char* address
,
209 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
210 if (strstr(address
, ":")) {
211 snprintf(buf
, len
, "[%s]:%u", address
, port
);
213 snprintf(buf
, len
, "%s:%u", address
, port
);
217 static int tcp_socket_incoming(const char *address
, uint16_t port
)
219 char address_and_port
[128];
220 Error
*local_err
= NULL
;
222 combine_addr(address_and_port
, 128, address
, port
);
223 int fd
= inet_listen(address_and_port
, NULL
, 0, SOCK_STREAM
, 0, &local_err
);
225 if (local_err
!= NULL
) {
226 qerror_report_err(local_err
);
227 error_free(local_err
);
232 static int unix_socket_incoming(const char *path
)
234 Error
*local_err
= NULL
;
235 int fd
= unix_listen(path
, NULL
, 0, &local_err
);
237 if (local_err
!= NULL
) {
238 qerror_report_err(local_err
);
239 error_free(local_err
);
244 static int unix_socket_outgoing(const char *path
)
246 Error
*local_err
= NULL
;
247 int fd
= unix_connect(path
, &local_err
);
249 if (local_err
!= NULL
) {
250 qerror_report_err(local_err
);
251 error_free(local_err
);
256 static void *show_parts(void *arg
)
261 /* linux just needs an open() to trigger
262 * the partition table update
263 * but remember to load the module with max_part != 0 :
264 * modprobe nbd max_part=63
266 nbd
= open(device
, O_RDWR
);
273 static void *nbd_client_thread(void *arg
)
281 pthread_t show_parts_thread
;
283 sock
= unix_socket_outgoing(sockpath
);
288 ret
= nbd_receive_negotiate(sock
, NULL
, &nbdflags
,
294 fd
= open(device
, O_RDWR
);
296 /* Linux-only, we can use %m in printf. */
297 fprintf(stderr
, "Failed to open %s: %m", device
);
301 ret
= nbd_init(fd
, sock
, nbdflags
, size
, blocksize
);
306 /* update partition table */
307 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
310 fprintf(stderr
, "NBD device %s is now connected to %s\n",
313 /* Close stderr so that the qemu-nbd process exits. */
314 dup2(STDOUT_FILENO
, STDERR_FILENO
);
317 ret
= nbd_client(fd
);
322 kill(getpid(), SIGTERM
);
323 return (void *) EXIT_SUCCESS
;
326 kill(getpid(), SIGTERM
);
327 return (void *) EXIT_FAILURE
;
330 static int nbd_can_accept(void *opaque
)
332 return nb_fds
< shared
;
335 static void nbd_export_closed(NBDExport
*exp
)
337 assert(state
== TERMINATING
);
341 static void nbd_client_closed(NBDClient
*client
)
344 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
348 nbd_client_put(client
);
351 static void nbd_accept(void *opaque
)
353 int server_fd
= (uintptr_t) opaque
;
354 struct sockaddr_in addr
;
355 socklen_t addr_len
= sizeof(addr
);
357 int fd
= accept(server_fd
, (struct sockaddr
*)&addr
, &addr_len
);
358 if (state
>= TERMINATE
) {
363 if (fd
>= 0 && nbd_client_new(exp
, fd
, nbd_client_closed
)) {
368 int main(int argc
, char **argv
)
370 BlockDriverState
*bs
;
372 off_t dev_offset
= 0;
373 uint32_t nbdflags
= 0;
374 bool disconnect
= false;
375 const char *bindto
= "0.0.0.0";
377 int port
= NBD_DEFAULT_PORT
;
379 QemuOpts
*sn_opts
= NULL
;
380 const char *sn_id_or_name
= NULL
;
381 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
382 struct option lopt
[] = {
383 { "help", 0, NULL
, 'h' },
384 { "version", 0, NULL
, 'V' },
385 { "bind", 1, NULL
, 'b' },
386 { "port", 1, NULL
, 'p' },
387 { "socket", 1, NULL
, 'k' },
388 { "offset", 1, NULL
, 'o' },
389 { "read-only", 0, NULL
, 'r' },
390 { "partition", 1, NULL
, 'P' },
391 { "connect", 1, NULL
, 'c' },
392 { "disconnect", 0, NULL
, 'd' },
393 { "snapshot", 0, NULL
, 's' },
394 { "load-snapshot", 1, NULL
, 'l' },
395 { "nocache", 0, NULL
, 'n' },
396 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
397 #ifdef CONFIG_LINUX_AIO
398 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
400 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
401 { "shared", 1, NULL
, 'e' },
402 { "format", 1, NULL
, 'f' },
403 { "persistent", 0, NULL
, 't' },
404 { "verbose", 0, NULL
, 'v' },
411 int flags
= BDRV_O_RDWR
;
415 bool seen_cache
= false;
416 bool seen_discard
= false;
417 #ifdef CONFIG_LINUX_AIO
418 bool seen_aio
= false;
420 pthread_t client_thread
;
421 const char *fmt
= NULL
;
422 Error
*local_err
= NULL
;
424 /* The client thread uses SIGTERM to interrupt the server. A signal
425 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
427 struct sigaction sa_sigterm
;
428 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
429 sa_sigterm
.sa_handler
= termsig_handler
;
430 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
431 qemu_init_exec_dir(argv
[0]);
433 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
436 flags
|= BDRV_O_SNAPSHOT
;
439 optarg
= (char *) "none";
441 case QEMU_NBD_OPT_CACHE
:
443 errx(EXIT_FAILURE
, "-n and --cache can only be specified once");
446 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
447 errx(EXIT_FAILURE
, "Invalid cache mode `%s'", optarg
);
450 #ifdef CONFIG_LINUX_AIO
451 case QEMU_NBD_OPT_AIO
:
453 errx(EXIT_FAILURE
, "--aio can only be specified once");
456 if (!strcmp(optarg
, "native")) {
457 flags
|= BDRV_O_NATIVE_AIO
;
458 } else if (!strcmp(optarg
, "threads")) {
459 /* this is the default */
461 errx(EXIT_FAILURE
, "invalid aio mode `%s'", optarg
);
465 case QEMU_NBD_OPT_DISCARD
:
467 errx(EXIT_FAILURE
, "--discard can only be specified once");
470 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
471 errx(EXIT_FAILURE
, "Invalid discard mode `%s'", optarg
);
478 li
= strtol(optarg
, &end
, 0);
480 errx(EXIT_FAILURE
, "Invalid port `%s'", optarg
);
482 if (li
< 1 || li
> 65535) {
483 errx(EXIT_FAILURE
, "Port out of range `%s'", optarg
);
488 dev_offset
= strtoll (optarg
, &end
, 0);
490 errx(EXIT_FAILURE
, "Invalid offset `%s'", optarg
);
492 if (dev_offset
< 0) {
493 errx(EXIT_FAILURE
, "Offset must be positive `%s'", optarg
);
497 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
498 sn_opts
= qemu_opts_parse(&internal_snapshot_opts
, optarg
, 0);
500 errx(EXIT_FAILURE
, "Failed in parsing snapshot param `%s'",
504 sn_id_or_name
= optarg
;
508 nbdflags
|= NBD_FLAG_READ_ONLY
;
509 flags
&= ~BDRV_O_RDWR
;
512 partition
= strtol(optarg
, &end
, 0);
514 errx(EXIT_FAILURE
, "Invalid partition `%s'", optarg
);
515 if (partition
< 1 || partition
> 8)
516 errx(EXIT_FAILURE
, "Invalid partition %d", partition
);
520 if (sockpath
[0] != '/')
521 errx(EXIT_FAILURE
, "socket path must be absolute\n");
530 shared
= strtol(optarg
, &end
, 0);
532 errx(EXIT_FAILURE
, "Invalid shared device number '%s'", optarg
);
535 errx(EXIT_FAILURE
, "Shared device number must be greater than 0\n");
556 errx(EXIT_FAILURE
, "Try `%s --help' for more information.",
561 if ((argc
- optind
) != 1) {
562 errx(EXIT_FAILURE
, "Invalid number of argument.\n"
563 "Try `%s --help' for more information.",
568 fd
= open(argv
[optind
], O_RDWR
);
570 err(EXIT_FAILURE
, "Cannot open %s", argv
[optind
]);
576 printf("%s disconnected\n", argv
[optind
]);
581 if (device
&& !verbose
) {
586 if (qemu_pipe(stderr_fd
) < 0) {
587 err(EXIT_FAILURE
, "Error setting up communication pipe");
590 /* Now daemonize, but keep a communication channel open to
591 * print errors and exit with the proper status code.
596 ret
= qemu_daemon(1, 0);
598 /* Temporarily redirect stderr to the parent's pipe... */
599 dup2(stderr_fd
[1], STDERR_FILENO
);
601 err(EXIT_FAILURE
, "Failed to daemonize");
604 /* ... close the descriptor we inherited and go on. */
610 /* In the parent. Print error messages from the child until
611 * it closes the pipe.
614 buf
= g_malloc(1024);
615 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
617 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
623 err(EXIT_FAILURE
, "Cannot read from daemon");
626 /* Usually the daemon should not print any message.
627 * Exit with zero status in that case.
633 if (device
!= NULL
&& sockpath
== NULL
) {
634 sockpath
= g_malloc(128);
635 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
638 qemu_init_main_loop();
640 atexit(bdrv_close_all
);
643 drv
= bdrv_find_format(fmt
);
645 errx(EXIT_FAILURE
, "Unknown file format '%s'", fmt
);
651 bs
= bdrv_new("hda");
652 srcpath
= argv
[optind
];
653 ret
= bdrv_open(&bs
, srcpath
, NULL
, NULL
, flags
, drv
, &local_err
);
656 err(EXIT_FAILURE
, "Failed to bdrv_open '%s': %s", argv
[optind
],
657 error_get_pretty(local_err
));
661 ret
= bdrv_snapshot_load_tmp(bs
,
662 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
663 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
665 } else if (sn_id_or_name
) {
666 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
672 "Failed to load snapshot: %s",
673 error_get_pretty(local_err
));
676 fd_size
= bdrv_getlength(bs
);
678 if (partition
!= -1) {
679 ret
= find_partition(bs
, partition
, &dev_offset
, &fd_size
);
682 err(EXIT_FAILURE
, "Could not find partition %d", partition
);
686 exp
= nbd_export_new(bs
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
);
689 fd
= unix_socket_incoming(sockpath
);
691 fd
= tcp_socket_incoming(bindto
, port
);
701 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
703 errx(EXIT_FAILURE
, "Failed to create client thread: %s",
707 /* Shut up GCC warnings. */
708 memset(&client_thread
, 0, sizeof(client_thread
));
711 qemu_set_fd_handler2(fd
, nbd_can_accept
, nbd_accept
, NULL
,
712 (void *)(uintptr_t)fd
);
714 /* now when the initialization is (almost) complete, chdir("/")
715 * to free any busy filesystems */
716 if (chdir("/") < 0) {
717 err(EXIT_FAILURE
, "Could not chdir to root directory");
722 main_loop_wait(false);
723 if (state
== TERMINATE
) {
725 nbd_export_close(exp
);
729 } while (state
!= TERMINATED
);
737 qemu_opts_del(sn_opts
);
742 pthread_join(client_thread
, &ret
);