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/osdep.h"
20 #include "qapi/error.h"
21 #include "qemu-common.h"
22 #include "sysemu/block-backend.h"
23 #include "block/block_int.h"
24 #include "block/nbd.h"
25 #include "qemu/main-loop.h"
26 #include "qemu/error-report.h"
27 #include "qemu/config-file.h"
28 #include "block/snapshot.h"
29 #include "qapi/util.h"
30 #include "qapi/qmp/qstring.h"
31 #include "qom/object_interfaces.h"
32 #include "io/channel-socket.h"
38 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
39 #define QEMU_NBD_OPT_CACHE 256
40 #define QEMU_NBD_OPT_AIO 257
41 #define QEMU_NBD_OPT_DISCARD 258
42 #define QEMU_NBD_OPT_DETECT_ZEROES 259
43 #define QEMU_NBD_OPT_OBJECT 260
44 #define QEMU_NBD_OPT_TLSCREDS 261
45 #define QEMU_NBD_OPT_IMAGE_OPTS 262
47 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;
56 static QIOChannelSocket
*server_ioc
;
57 static int server_watch
= -1;
58 static QCryptoTLSCreds
*tlscreds
;
60 static void usage(const char *name
)
63 "Usage: %s [OPTIONS] FILE\n"
64 "QEMU Disk Network Block Device Server\n"
66 " -h, --help display this help and exit\n"
67 " -V, --version output version information and exit\n"
69 "Connection properties:\n"
70 " -p, --port=PORT port to listen on (default `%d')\n"
71 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
72 " -k, --socket=PATH path to the unix socket\n"
73 " (default '"SOCKET_PATH
"')\n"
74 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
75 " -t, --persistent don't exit on the last connection\n"
76 " -v, --verbose display extra debugging information\n"
78 "Exposing part of the image:\n"
79 " -o, --offset=OFFSET offset into the image\n"
80 " -P, --partition=NUM only expose partition NUM\n"
82 "General purpose options:\n"
83 " --object type,id=ID,... define an object such as 'secret' for providing\n"
84 " passwords and/or encryption keys\n"
86 "Kernel NBD client support:\n"
87 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
88 " -d, --disconnect disconnect the specified device\n"
92 "Block device options:\n"
93 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
94 " -r, --read-only export read-only\n"
95 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
96 " file with backing_file=FILE, redirect the write to\n"
97 " the temporary one\n"
98 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
99 " load an internal snapshot inside FILE and export it\n"
100 " as an read-only device, SNAPSHOT_PARAM format is\n"
101 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
103 " -n, --nocache disable host cache\n"
104 " --cache=MODE set cache mode (none, writeback, ...)\n"
105 " --aio=MODE set AIO mode (native or threads)\n"
106 " --discard=MODE set discard mode (ignore, unmap)\n"
107 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
108 " --image-opts treat FILE as a full set of image options\n"
110 "Report bugs to <qemu-devel@nongnu.org>\n"
111 , name
, NBD_DEFAULT_PORT
, "DEVICE");
114 static void version(const char *name
)
118 "Written by Anthony Liguori.\n"
120 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
121 "This is free software; see the source for copying conditions. There is NO\n"
122 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
126 struct partition_record
130 uint32_t start_cylinder
;
131 uint8_t start_sector
;
134 uint8_t end_cylinder
;
136 uint32_t start_sector_abs
;
137 uint32_t nb_sectors_abs
;
140 static void read_partition(uint8_t *p
, struct partition_record
*r
)
143 r
->start_head
= p
[1];
144 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
145 r
->start_sector
= p
[2] & 0x3f;
148 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
149 r
->end_sector
= p
[6] & 0x3f;
151 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
152 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
155 static int find_partition(BlockBackend
*blk
, int partition
,
156 off_t
*offset
, off_t
*size
)
158 struct partition_record mbr
[4];
164 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
165 error_report("error while reading: %s", strerror(-ret
));
169 if (data
[510] != 0x55 || data
[511] != 0xaa) {
173 for (i
= 0; i
< 4; i
++) {
174 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
176 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
180 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
181 struct partition_record ext
[4];
185 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
186 error_report("error while reading: %s", strerror(-ret
));
190 for (j
= 0; j
< 4; j
++) {
191 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
192 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
196 if ((ext_partnum
+ j
+ 1) == partition
) {
197 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
198 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
203 } else if ((i
+ 1) == partition
) {
204 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
205 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
213 static void termsig_handler(int signum
)
220 static void *show_parts(void *arg
)
225 /* linux just needs an open() to trigger
226 * the partition table update
227 * but remember to load the module with max_part != 0 :
228 * modprobe nbd max_part=63
230 nbd
= open(device
, O_RDWR
);
237 static void *nbd_client_thread(void *arg
)
242 QIOChannelSocket
*sioc
;
245 pthread_t show_parts_thread
;
246 Error
*local_error
= NULL
;
248 sioc
= qio_channel_socket_new();
249 if (qio_channel_socket_connect_sync(sioc
,
252 error_report_err(local_error
);
256 ret
= nbd_receive_negotiate(QIO_CHANNEL(sioc
), NULL
, &nbdflags
,
258 &size
, &local_error
);
261 error_report_err(local_error
);
266 fd
= open(device
, O_RDWR
);
268 /* Linux-only, we can use %m in printf. */
269 error_report("Failed to open %s: %m", device
);
273 ret
= nbd_init(fd
, sioc
, nbdflags
, size
);
278 /* update partition table */
279 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
282 fprintf(stderr
, "NBD device %s is now connected to %s\n",
285 /* Close stderr so that the qemu-nbd process exits. */
286 dup2(STDOUT_FILENO
, STDERR_FILENO
);
289 ret
= nbd_client(fd
);
294 object_unref(OBJECT(sioc
));
295 kill(getpid(), SIGTERM
);
296 return (void *) EXIT_SUCCESS
;
301 object_unref(OBJECT(sioc
));
303 kill(getpid(), SIGTERM
);
304 return (void *) EXIT_FAILURE
;
307 static int nbd_can_accept(void)
309 return nb_fds
< shared
;
312 static void nbd_export_closed(NBDExport
*exp
)
314 assert(state
== TERMINATING
);
318 static void nbd_update_server_watch(void);
320 static void nbd_client_closed(NBDClient
*client
)
323 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
326 nbd_update_server_watch();
327 nbd_client_put(client
);
330 static gboolean
nbd_accept(QIOChannel
*ioc
, GIOCondition cond
, gpointer opaque
)
332 QIOChannelSocket
*cioc
;
334 cioc
= qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc
),
340 if (state
>= TERMINATE
) {
341 object_unref(OBJECT(cioc
));
346 nbd_update_server_watch();
347 nbd_client_new(newproto
? NULL
: exp
, cioc
,
348 tlscreds
, NULL
, nbd_client_closed
);
349 object_unref(OBJECT(cioc
));
354 static void nbd_update_server_watch(void)
356 if (nbd_can_accept()) {
357 if (server_watch
== -1) {
358 server_watch
= qio_channel_add_watch(QIO_CHANNEL(server_ioc
),
364 if (server_watch
!= -1) {
365 g_source_remove(server_watch
);
372 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
376 SocketAddress
*saddr
;
378 saddr
= g_new0(SocketAddress
, 1);
380 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
381 saddr
->u
.q_unix
.data
= g_new0(UnixSocketAddress
, 1);
382 saddr
->u
.q_unix
.data
->path
= g_strdup(sockpath
);
384 InetSocketAddress
*inet
;
385 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
386 inet
= saddr
->u
.inet
.data
= g_new0(InetSocketAddress
, 1);
387 inet
->host
= g_strdup(bindto
);
389 inet
->port
= g_strdup(port
);
391 inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
399 static QemuOptsList file_opts
= {
401 .implied_opt_name
= "file",
402 .head
= QTAILQ_HEAD_INITIALIZER(file_opts
.head
),
404 /* no elements => accept any params */
405 { /* end of list */ }
409 static QemuOptsList qemu_object_opts
= {
411 .implied_opt_name
= "qom-type",
412 .head
= QTAILQ_HEAD_INITIALIZER(qemu_object_opts
.head
),
420 static QCryptoTLSCreds
*nbd_get_tls_creds(const char *id
, Error
**errp
)
423 QCryptoTLSCreds
*creds
;
425 obj
= object_resolve_path_component(
426 object_get_objects_root(), id
);
428 error_setg(errp
, "No TLS credentials with id '%s'",
432 creds
= (QCryptoTLSCreds
*)
433 object_dynamic_cast(obj
, TYPE_QCRYPTO_TLS_CREDS
);
435 error_setg(errp
, "Object with id '%s' is not TLS credentials",
440 if (creds
->endpoint
!= QCRYPTO_TLS_CREDS_ENDPOINT_SERVER
) {
442 "Expecting TLS credentials with a server endpoint");
450 int main(int argc
, char **argv
)
453 BlockDriverState
*bs
;
454 off_t dev_offset
= 0;
455 uint32_t nbdflags
= 0;
456 bool disconnect
= false;
457 const char *bindto
= "0.0.0.0";
458 const char *port
= NULL
;
459 char *sockpath
= NULL
;
462 QemuOpts
*sn_opts
= NULL
;
463 const char *sn_id_or_name
= NULL
;
464 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:x:";
465 struct option lopt
[] = {
466 { "help", no_argument
, NULL
, 'h' },
467 { "version", no_argument
, NULL
, 'V' },
468 { "bind", required_argument
, NULL
, 'b' },
469 { "port", required_argument
, NULL
, 'p' },
470 { "socket", required_argument
, NULL
, 'k' },
471 { "offset", required_argument
, NULL
, 'o' },
472 { "read-only", no_argument
, NULL
, 'r' },
473 { "partition", required_argument
, NULL
, 'P' },
474 { "connect", required_argument
, NULL
, 'c' },
475 { "disconnect", no_argument
, NULL
, 'd' },
476 { "snapshot", no_argument
, NULL
, 's' },
477 { "load-snapshot", required_argument
, NULL
, 'l' },
478 { "nocache", no_argument
, NULL
, 'n' },
479 { "cache", required_argument
, NULL
, QEMU_NBD_OPT_CACHE
},
480 { "aio", required_argument
, NULL
, QEMU_NBD_OPT_AIO
},
481 { "discard", required_argument
, NULL
, QEMU_NBD_OPT_DISCARD
},
482 { "detect-zeroes", required_argument
, NULL
,
483 QEMU_NBD_OPT_DETECT_ZEROES
},
484 { "shared", required_argument
, NULL
, 'e' },
485 { "format", required_argument
, NULL
, 'f' },
486 { "persistent", no_argument
, NULL
, 't' },
487 { "verbose", no_argument
, NULL
, 'v' },
488 { "object", required_argument
, NULL
, QEMU_NBD_OPT_OBJECT
},
489 { "export-name", required_argument
, NULL
, 'x' },
490 { "tls-creds", required_argument
, NULL
, QEMU_NBD_OPT_TLSCREDS
},
491 { "image-opts", no_argument
, NULL
, QEMU_NBD_OPT_IMAGE_OPTS
},
497 int flags
= BDRV_O_RDWR
;
500 bool seen_cache
= false;
501 bool seen_discard
= false;
502 bool seen_aio
= false;
503 pthread_t client_thread
;
504 const char *fmt
= NULL
;
505 Error
*local_err
= NULL
;
506 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
507 QDict
*options
= NULL
;
508 const char *export_name
= NULL
;
509 const char *tlscredsid
= NULL
;
510 bool imageOpts
= false;
512 /* The client thread uses SIGTERM to interrupt the server. A signal
513 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
515 struct sigaction sa_sigterm
;
516 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
517 sa_sigterm
.sa_handler
= termsig_handler
;
518 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
519 module_call_init(MODULE_INIT_QOM
);
520 qemu_add_opts(&qemu_object_opts
);
521 qemu_init_exec_dir(argv
[0]);
523 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
526 flags
|= BDRV_O_SNAPSHOT
;
529 optarg
= (char *) "none";
531 case QEMU_NBD_OPT_CACHE
:
533 error_report("-n and --cache can only be specified once");
537 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
538 error_report("Invalid cache mode `%s'", optarg
);
542 case QEMU_NBD_OPT_AIO
:
544 error_report("--aio can only be specified once");
548 if (!strcmp(optarg
, "native")) {
549 flags
|= BDRV_O_NATIVE_AIO
;
550 } else if (!strcmp(optarg
, "threads")) {
551 /* this is the default */
553 error_report("invalid aio mode `%s'", optarg
);
557 case QEMU_NBD_OPT_DISCARD
:
559 error_report("--discard can only be specified once");
563 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
564 error_report("Invalid discard mode `%s'", optarg
);
568 case QEMU_NBD_OPT_DETECT_ZEROES
:
570 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
572 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
573 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
576 error_reportf_err(local_err
,
577 "Failed to parse detect_zeroes mode: ");
580 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
581 !(flags
& BDRV_O_UNMAP
)) {
582 error_report("setting detect-zeroes to unmap is not allowed "
583 "without setting discard operation to unmap");
594 dev_offset
= strtoll (optarg
, &end
, 0);
596 error_report("Invalid offset `%s'", optarg
);
599 if (dev_offset
< 0) {
600 error_report("Offset must be positive `%s'", optarg
);
605 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
606 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
609 error_report("Failed in parsing snapshot param `%s'",
614 sn_id_or_name
= optarg
;
618 nbdflags
|= NBD_FLAG_READ_ONLY
;
619 flags
&= ~BDRV_O_RDWR
;
622 partition
= strtol(optarg
, &end
, 0);
624 error_report("Invalid partition `%s'", optarg
);
627 if (partition
< 1 || partition
> 8) {
628 error_report("Invalid partition %d", partition
);
634 if (sockpath
[0] != '/') {
635 error_report("socket path must be absolute");
646 shared
= strtol(optarg
, &end
, 0);
648 error_report("Invalid shared device number '%s'", optarg
);
652 error_report("Shared device number must be greater than 0");
663 export_name
= optarg
;
677 error_report("Try `%s --help' for more information.", argv
[0]);
679 case QEMU_NBD_OPT_OBJECT
: {
681 opts
= qemu_opts_parse_noisily(&qemu_object_opts
,
687 case QEMU_NBD_OPT_TLSCREDS
:
690 case QEMU_NBD_OPT_IMAGE_OPTS
:
696 if ((argc
- optind
) != 1) {
697 error_report("Invalid number of arguments");
698 error_printf("Try `%s --help' for more information.\n", argv
[0]);
702 if (qemu_opts_foreach(&qemu_object_opts
,
703 user_creatable_add_opts_foreach
,
705 error_report_err(local_err
);
711 error_report("TLS is only supported with IPv4/IPv6");
715 error_report("TLS is not supported with a host device");
719 /* Set the default NBD protocol export name, since
720 * we *must* use new style protocol for TLS */
723 tlscreds
= nbd_get_tls_creds(tlscredsid
, &local_err
);
725 error_report("Failed to get TLS creds %s",
726 error_get_pretty(local_err
));
732 int nbdfd
= open(argv
[optind
], O_RDWR
);
734 error_report("Cannot open %s: %s", argv
[optind
],
738 nbd_disconnect(nbdfd
);
742 printf("%s disconnected\n", argv
[optind
]);
747 if (device
&& !verbose
) {
752 if (qemu_pipe(stderr_fd
) < 0) {
753 error_report("Error setting up communication pipe: %s",
758 /* Now daemonize, but keep a communication channel open to
759 * print errors and exit with the proper status code.
763 error_report("Failed to fork: %s", strerror(errno
));
765 } else if (pid
== 0) {
767 ret
= qemu_daemon(1, 0);
769 /* Temporarily redirect stderr to the parent's pipe... */
770 dup2(stderr_fd
[1], STDERR_FILENO
);
772 error_report("Failed to daemonize: %s", strerror(errno
));
776 /* ... close the descriptor we inherited and go on. */
782 /* In the parent. Print error messages from the child until
783 * it closes the pipe.
786 buf
= g_malloc(1024);
787 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
789 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
795 error_report("Cannot read from daemon: %s",
800 /* Usually the daemon should not print any message.
801 * Exit with zero status in that case.
807 if (device
!= NULL
&& sockpath
== NULL
) {
808 sockpath
= g_malloc(128);
809 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
812 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
814 if (qemu_init_main_loop(&local_err
)) {
815 error_report_err(local_err
);
819 atexit(bdrv_close_all
);
821 srcpath
= argv
[optind
];
825 error_report("--image-opts and -f are mutually exclusive");
828 opts
= qemu_opts_parse_noisily(&file_opts
, srcpath
, true);
830 qemu_opts_reset(&file_opts
);
833 options
= qemu_opts_to_qdict(opts
, NULL
);
834 qemu_opts_reset(&file_opts
);
835 blk
= blk_new_open(NULL
, NULL
, options
, flags
, &local_err
);
838 options
= qdict_new();
839 qdict_put(options
, "driver", qstring_from_str(fmt
));
841 blk
= blk_new_open(srcpath
, NULL
, options
, flags
, &local_err
);
845 error_reportf_err(local_err
, "Failed to blk_new_open '%s': ",
852 ret
= bdrv_snapshot_load_tmp(bs
,
853 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
854 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
856 } else if (sn_id_or_name
) {
857 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
861 error_reportf_err(local_err
, "Failed to load snapshot: ");
865 bs
->detect_zeroes
= detect_zeroes
;
866 fd_size
= blk_getlength(blk
);
868 error_report("Failed to determine the image length: %s",
873 if (partition
!= -1) {
874 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
876 error_report("Could not find partition %d: %s", partition
,
882 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
885 error_report_err(local_err
);
889 nbd_export_set_name(exp
, export_name
);
893 server_ioc
= qio_channel_socket_new();
894 if (qio_channel_socket_listen_sync(server_ioc
, saddr
, &local_err
) < 0) {
895 object_unref(OBJECT(server_ioc
));
896 error_report_err(local_err
);
903 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
905 error_report("Failed to create client thread: %s", strerror(ret
));
909 /* Shut up GCC warnings. */
910 memset(&client_thread
, 0, sizeof(client_thread
));
913 nbd_update_server_watch();
915 /* now when the initialization is (almost) complete, chdir("/")
916 * to free any busy filesystems */
917 if (chdir("/") < 0) {
918 error_report("Could not chdir to root directory: %s",
925 main_loop_wait(false);
926 if (state
== TERMINATE
) {
928 nbd_export_close(exp
);
932 } while (state
!= TERMINATED
);
939 qemu_opts_del(sn_opts
);
943 pthread_join(client_thread
, &ret
);