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 "sysemu/block-backend.h"
21 #include "block/block_int.h"
22 #include "block/nbd.h"
23 #include "qemu/main-loop.h"
24 #include "qemu/sockets.h"
25 #include "qemu/error-report.h"
26 #include "block/snapshot.h"
27 #include "qapi/util.h"
28 #include "qapi/qmp/qstring.h"
33 #include <sys/types.h>
34 #include <sys/socket.h>
35 #include <netinet/in.h>
36 #include <netinet/tcp.h>
37 #include <arpa/inet.h>
42 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
43 #define QEMU_NBD_OPT_CACHE 1
44 #define QEMU_NBD_OPT_AIO 2
45 #define QEMU_NBD_OPT_DISCARD 3
46 #define QEMU_NBD_OPT_DETECT_ZEROES 4
48 static NBDExport
*exp
;
51 static SocketAddress
*saddr
;
52 static int persistent
= 0;
53 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
54 static int shared
= 1;
58 static void usage(const char *name
)
61 "Usage: %s [OPTIONS] FILE\n"
62 "QEMU Disk Network Block Device Server\n"
64 " -h, --help display this help and exit\n"
65 " -V, --version output version information and exit\n"
67 "Connection properties:\n"
68 " -p, --port=PORT port to listen on (default `%d')\n"
69 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
70 " -k, --socket=PATH path to the unix socket\n"
71 " (default '"SOCKET_PATH
"')\n"
72 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
73 " -t, --persistent don't exit on the last connection\n"
74 " -v, --verbose display extra debugging information\n"
76 "Exposing part of the image:\n"
77 " -o, --offset=OFFSET offset into the image\n"
78 " -P, --partition=NUM only expose partition NUM\n"
81 "Kernel NBD client support:\n"
82 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
83 " -d, --disconnect disconnect the specified device\n"
87 "Block device options:\n"
88 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
89 " -r, --read-only export read-only\n"
90 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
91 " file with backing_file=FILE, redirect the write to\n"
92 " the temporary one\n"
93 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
94 " load an internal snapshot inside FILE and export it\n"
95 " as an read-only device, SNAPSHOT_PARAM format is\n"
96 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
98 " -n, --nocache disable host cache\n"
99 " --cache=MODE set cache mode (none, writeback, ...)\n"
100 " --aio=MODE set AIO mode (native or threads)\n"
101 " --discard=MODE set discard mode (ignore, unmap)\n"
102 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
104 "Report bugs to <qemu-devel@nongnu.org>\n"
105 , name
, NBD_DEFAULT_PORT
, "DEVICE");
108 static void version(const char *name
)
112 "Written by Anthony Liguori.\n"
114 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
115 "This is free software; see the source for copying conditions. There is NO\n"
116 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
120 struct partition_record
124 uint32_t start_cylinder
;
125 uint8_t start_sector
;
128 uint8_t end_cylinder
;
130 uint32_t start_sector_abs
;
131 uint32_t nb_sectors_abs
;
134 static void read_partition(uint8_t *p
, struct partition_record
*r
)
137 r
->start_head
= p
[1];
138 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
139 r
->start_sector
= p
[2] & 0x3f;
142 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
143 r
->end_sector
= p
[6] & 0x3f;
145 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
146 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
149 static int find_partition(BlockBackend
*blk
, int partition
,
150 off_t
*offset
, off_t
*size
)
152 struct partition_record mbr
[4];
158 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
160 error_report("error while reading: %s", strerror(errno
));
164 if (data
[510] != 0x55 || data
[511] != 0xaa) {
168 for (i
= 0; i
< 4; i
++) {
169 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
171 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
175 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
176 struct partition_record ext
[4];
180 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
182 error_report("error while reading: %s", strerror(errno
));
186 for (j
= 0; j
< 4; j
++) {
187 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
188 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
192 if ((ext_partnum
+ j
+ 1) == partition
) {
193 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
194 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
199 } else if ((i
+ 1) == partition
) {
200 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
201 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
209 static void termsig_handler(int signum
)
216 static void *show_parts(void *arg
)
221 /* linux just needs an open() to trigger
222 * the partition table update
223 * but remember to load the module with max_part != 0 :
224 * modprobe nbd max_part=63
226 nbd
= open(device
, O_RDWR
);
233 static void *nbd_client_thread(void *arg
)
240 pthread_t show_parts_thread
;
241 Error
*local_error
= NULL
;
244 sock
= socket_connect(saddr
, &local_error
, NULL
, NULL
);
246 error_report_err(local_error
);
250 ret
= nbd_receive_negotiate(sock
, NULL
, &nbdflags
,
251 &size
, &local_error
);
254 fprintf(stderr
, "%s\n", error_get_pretty(local_error
));
255 error_free(local_error
);
260 fd
= open(device
, O_RDWR
);
262 /* Linux-only, we can use %m in printf. */
263 fprintf(stderr
, "Failed to open %s: %m\n", device
);
267 ret
= nbd_init(fd
, sock
, nbdflags
, size
);
272 /* update partition table */
273 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
276 fprintf(stderr
, "NBD device %s is now connected to %s\n",
279 /* Close stderr so that the qemu-nbd process exits. */
280 dup2(STDOUT_FILENO
, STDERR_FILENO
);
283 ret
= nbd_client(fd
);
288 kill(getpid(), SIGTERM
);
289 return (void *) EXIT_SUCCESS
;
296 kill(getpid(), SIGTERM
);
297 return (void *) EXIT_FAILURE
;
300 static int nbd_can_accept(void)
302 return nb_fds
< shared
;
305 static void nbd_export_closed(NBDExport
*exp
)
307 assert(state
== TERMINATING
);
311 static void nbd_update_server_fd_handler(int fd
);
313 static void nbd_client_closed(NBDClient
*client
)
316 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
319 nbd_update_server_fd_handler(server_fd
);
320 nbd_client_put(client
);
323 static void nbd_accept(void *opaque
)
325 struct sockaddr_in addr
;
326 socklen_t addr_len
= sizeof(addr
);
328 int fd
= accept(server_fd
, (struct sockaddr
*)&addr
, &addr_len
);
334 if (state
>= TERMINATE
) {
339 if (nbd_client_new(exp
, fd
, nbd_client_closed
)) {
341 nbd_update_server_fd_handler(server_fd
);
348 static void nbd_update_server_fd_handler(int fd
)
350 if (nbd_can_accept()) {
351 qemu_set_fd_handler(fd
, nbd_accept
, NULL
, (void *)(uintptr_t)fd
);
353 qemu_set_fd_handler(fd
, NULL
, NULL
, NULL
);
358 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
362 SocketAddress
*saddr
;
364 saddr
= g_new0(SocketAddress
, 1);
366 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
367 saddr
->u
.q_unix
= g_new0(UnixSocketAddress
, 1);
368 saddr
->u
.q_unix
->path
= g_strdup(sockpath
);
370 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
371 saddr
->u
.inet
= g_new0(InetSocketAddress
, 1);
372 saddr
->u
.inet
->host
= g_strdup(bindto
);
374 saddr
->u
.inet
->port
= g_strdup(port
);
376 saddr
->u
.inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
384 int main(int argc
, char **argv
)
387 BlockDriverState
*bs
;
388 off_t dev_offset
= 0;
389 uint32_t nbdflags
= 0;
390 bool disconnect
= false;
391 const char *bindto
= "0.0.0.0";
392 const char *port
= NULL
;
393 char *sockpath
= NULL
;
396 QemuOpts
*sn_opts
= NULL
;
397 const char *sn_id_or_name
= NULL
;
398 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
399 struct option lopt
[] = {
400 { "help", 0, NULL
, 'h' },
401 { "version", 0, NULL
, 'V' },
402 { "bind", 1, NULL
, 'b' },
403 { "port", 1, NULL
, 'p' },
404 { "socket", 1, NULL
, 'k' },
405 { "offset", 1, NULL
, 'o' },
406 { "read-only", 0, NULL
, 'r' },
407 { "partition", 1, NULL
, 'P' },
408 { "connect", 1, NULL
, 'c' },
409 { "disconnect", 0, NULL
, 'd' },
410 { "snapshot", 0, NULL
, 's' },
411 { "load-snapshot", 1, NULL
, 'l' },
412 { "nocache", 0, NULL
, 'n' },
413 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
414 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
415 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
416 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
417 { "shared", 1, NULL
, 'e' },
418 { "format", 1, NULL
, 'f' },
419 { "persistent", 0, NULL
, 't' },
420 { "verbose", 0, NULL
, 'v' },
426 int flags
= BDRV_O_RDWR
;
430 bool seen_cache
= false;
431 bool seen_discard
= false;
432 bool seen_aio
= false;
433 pthread_t client_thread
;
434 const char *fmt
= NULL
;
435 Error
*local_err
= NULL
;
436 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
437 QDict
*options
= NULL
;
439 /* The client thread uses SIGTERM to interrupt the server. A signal
440 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
442 struct sigaction sa_sigterm
;
443 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
444 sa_sigterm
.sa_handler
= termsig_handler
;
445 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
446 qemu_init_exec_dir(argv
[0]);
448 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
451 flags
|= BDRV_O_SNAPSHOT
;
454 optarg
= (char *) "none";
456 case QEMU_NBD_OPT_CACHE
:
458 error_report("-n and --cache can only be specified once");
462 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
463 error_report("Invalid cache mode `%s'", optarg
);
467 case QEMU_NBD_OPT_AIO
:
469 error_report("--aio can only be specified once");
473 if (!strcmp(optarg
, "native")) {
474 flags
|= BDRV_O_NATIVE_AIO
;
475 } else if (!strcmp(optarg
, "threads")) {
476 /* this is the default */
478 error_report("invalid aio mode `%s'", optarg
);
482 case QEMU_NBD_OPT_DISCARD
:
484 error_report("--discard can only be specified once");
488 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
489 error_report("Invalid discard mode `%s'", optarg
);
493 case QEMU_NBD_OPT_DETECT_ZEROES
:
495 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
497 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
498 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
501 error_report("Failed to parse detect_zeroes mode: %s",
502 error_get_pretty(local_err
));
505 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
506 !(flags
& BDRV_O_UNMAP
)) {
507 error_report("setting detect-zeroes to unmap is not allowed "
508 "without setting discard operation to unmap");
519 dev_offset
= strtoll (optarg
, &end
, 0);
521 error_report("Invalid offset `%s'", optarg
);
524 if (dev_offset
< 0) {
525 error_report("Offset must be positive `%s'", optarg
);
530 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
531 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
534 error_report("Failed in parsing snapshot param `%s'",
539 sn_id_or_name
= optarg
;
543 nbdflags
|= NBD_FLAG_READ_ONLY
;
544 flags
&= ~BDRV_O_RDWR
;
547 partition
= strtol(optarg
, &end
, 0);
549 error_report("Invalid partition `%s'", optarg
);
552 if (partition
< 1 || partition
> 8) {
553 error_report("Invalid partition %d", partition
);
559 if (sockpath
[0] != '/') {
560 error_report("socket path must be absolute\n");
571 shared
= strtol(optarg
, &end
, 0);
573 error_report("Invalid shared device number '%s'", optarg
);
577 error_report("Shared device number must be greater than 0\n");
599 error_report("Try `%s --help' for more information.", argv
[0]);
604 if ((argc
- optind
) != 1) {
605 error_report("Invalid number of argument.\n"
606 "Try `%s --help' for more information.",
612 fd
= open(argv
[optind
], O_RDWR
);
614 error_report("Cannot open %s: %s", argv
[optind
],
622 printf("%s disconnected\n", argv
[optind
]);
627 if (device
&& !verbose
) {
632 if (qemu_pipe(stderr_fd
) < 0) {
633 error_report("Error setting up communication pipe: %s",
638 /* Now daemonize, but keep a communication channel open to
639 * print errors and exit with the proper status code.
643 error_report("Failed to fork: %s", strerror(errno
));
645 } else if (pid
== 0) {
647 ret
= qemu_daemon(1, 0);
649 /* Temporarily redirect stderr to the parent's pipe... */
650 dup2(stderr_fd
[1], STDERR_FILENO
);
652 error_report("Failed to daemonize: %s", strerror(errno
));
656 /* ... close the descriptor we inherited and go on. */
662 /* In the parent. Print error messages from the child until
663 * it closes the pipe.
666 buf
= g_malloc(1024);
667 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
669 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
675 error_report("Cannot read from daemon: %s",
680 /* Usually the daemon should not print any message.
681 * Exit with zero status in that case.
687 if (device
!= NULL
&& sockpath
== NULL
) {
688 sockpath
= g_malloc(128);
689 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
692 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
694 if (qemu_init_main_loop(&local_err
)) {
695 error_report_err(local_err
);
699 atexit(bdrv_close_all
);
702 options
= qdict_new();
703 qdict_put(options
, "driver", qstring_from_str(fmt
));
706 srcpath
= argv
[optind
];
707 blk
= blk_new_open("hda", srcpath
, NULL
, options
, flags
, &local_err
);
709 error_report("Failed to blk_new_open '%s': %s", argv
[optind
],
710 error_get_pretty(local_err
));
716 ret
= bdrv_snapshot_load_tmp(bs
,
717 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
718 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
720 } else if (sn_id_or_name
) {
721 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
726 error_report("Failed to load snapshot: %s: %s",
727 error_get_pretty(local_err
), strerror(errno
));
731 bs
->detect_zeroes
= detect_zeroes
;
732 fd_size
= blk_getlength(blk
);
734 error_report("Failed to determine the image length: %s",
739 if (partition
!= -1) {
740 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
743 error_report("Could not find partition %d: %s", partition
,
749 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
752 error_report_err(local_err
);
756 fd
= socket_listen(saddr
, &local_err
);
758 error_report_err(local_err
);
765 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
767 error_report("Failed to create client thread: %s", strerror(ret
));
771 /* Shut up GCC warnings. */
772 memset(&client_thread
, 0, sizeof(client_thread
));
776 nbd_update_server_fd_handler(fd
);
778 /* now when the initialization is (almost) complete, chdir("/")
779 * to free any busy filesystems */
780 if (chdir("/") < 0) {
781 error_report("Could not chdir to root directory: %s",
788 main_loop_wait(false);
789 if (state
== TERMINATE
) {
791 nbd_export_close(exp
);
795 } while (state
!= TERMINATED
);
802 qemu_opts_del(sn_opts
);
806 pthread_join(client_thread
, &ret
);