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 "qemu/cutils.h"
23 #include "sysemu/block-backend.h"
24 #include "block/block_int.h"
25 #include "block/nbd.h"
26 #include "qemu/main-loop.h"
27 #include "qemu/error-report.h"
28 #include "qemu/config-file.h"
29 #include "qemu/bswap.h"
31 #include "block/snapshot.h"
32 #include "qapi/util.h"
33 #include "qapi/qmp/qstring.h"
34 #include "qom/object_interfaces.h"
35 #include "io/channel-socket.h"
36 #include "crypto/init.h"
37 #include "trace/control.h"
43 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
44 #define QEMU_NBD_OPT_CACHE 256
45 #define QEMU_NBD_OPT_AIO 257
46 #define QEMU_NBD_OPT_DISCARD 258
47 #define QEMU_NBD_OPT_DETECT_ZEROES 259
48 #define QEMU_NBD_OPT_OBJECT 260
49 #define QEMU_NBD_OPT_TLSCREDS 261
50 #define QEMU_NBD_OPT_IMAGE_OPTS 262
54 static NBDExport
*exp
;
58 static SocketAddress
*saddr
;
59 static int persistent
= 0;
60 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
61 static int shared
= 1;
63 static QIOChannelSocket
*server_ioc
;
64 static int server_watch
= -1;
65 static QCryptoTLSCreds
*tlscreds
;
67 static void usage(const char *name
)
70 "Usage: %s [OPTIONS] FILE\n"
71 "QEMU Disk Network Block Device Server\n"
73 " -h, --help display this help and exit\n"
74 " -V, --version output version information and exit\n"
76 "Connection properties:\n"
77 " -p, --port=PORT port to listen on (default `%d')\n"
78 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
79 " -k, --socket=PATH path to the unix socket\n"
80 " (default '"SOCKET_PATH
"')\n"
81 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
82 " -t, --persistent don't exit on the last connection\n"
83 " -v, --verbose display extra debugging information\n"
84 " -x, --export-name=NAME expose export by name\n"
86 "Exposing part of the image:\n"
87 " -o, --offset=OFFSET offset into the image\n"
88 " -P, --partition=NUM only expose partition NUM\n"
90 "General purpose options:\n"
91 " --object type,id=ID,... define an object such as 'secret' for providing\n"
92 " passwords and/or encryption keys\n"
93 " -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
94 " specify tracing options\n"
96 "Kernel NBD client support:\n"
97 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
98 " -d, --disconnect disconnect the specified device\n"
102 "Block device options:\n"
103 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
104 " -r, --read-only export read-only\n"
105 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
106 " file with backing_file=FILE, redirect the write to\n"
107 " the temporary one\n"
108 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
109 " load an internal snapshot inside FILE and export it\n"
110 " as an read-only device, SNAPSHOT_PARAM format is\n"
111 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
113 " -n, --nocache disable host cache\n"
114 " --cache=MODE set cache mode (none, writeback, ...)\n"
115 " --aio=MODE set AIO mode (native or threads)\n"
116 " --discard=MODE set discard mode (ignore, unmap)\n"
117 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
118 " --image-opts treat FILE as a full set of image options\n"
120 "Report bugs to <qemu-devel@nongnu.org>\n"
121 , name
, NBD_DEFAULT_PORT
, "DEVICE");
124 static void version(const char *name
)
128 "Written by Anthony Liguori.\n"
130 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
131 "This is free software; see the source for copying conditions. There is NO\n"
132 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
136 struct partition_record
140 uint32_t start_cylinder
;
141 uint8_t start_sector
;
144 uint8_t end_cylinder
;
146 uint32_t start_sector_abs
;
147 uint32_t nb_sectors_abs
;
150 static void read_partition(uint8_t *p
, struct partition_record
*r
)
153 r
->start_head
= p
[1];
154 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
155 r
->start_sector
= p
[2] & 0x3f;
158 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
159 r
->end_sector
= p
[6] & 0x3f;
161 r
->start_sector_abs
= ldl_le_p(p
+ 8);
162 r
->nb_sectors_abs
= ldl_le_p(p
+ 12);
165 static int find_partition(BlockBackend
*blk
, int partition
,
166 off_t
*offset
, off_t
*size
)
168 struct partition_record mbr
[4];
169 uint8_t data
[MBR_SIZE
];
174 ret
= blk_pread(blk
, 0, data
, sizeof(data
));
176 error_report("error while reading: %s", strerror(-ret
));
180 if (data
[510] != 0x55 || data
[511] != 0xaa) {
184 for (i
= 0; i
< 4; i
++) {
185 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
187 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
191 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
192 struct partition_record ext
[4];
193 uint8_t data1
[MBR_SIZE
];
196 ret
= blk_pread(blk
, mbr
[i
].start_sector_abs
* MBR_SIZE
,
197 data1
, sizeof(data1
));
199 error_report("error while reading: %s", strerror(-ret
));
203 for (j
= 0; j
< 4; j
++) {
204 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
205 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
209 if ((ext_partnum
+ j
+ 1) == partition
) {
210 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
211 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
216 } else if ((i
+ 1) == partition
) {
217 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
218 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
226 static void termsig_handler(int signum
)
228 atomic_cmpxchg(&state
, RUNNING
, TERMINATE
);
233 static void *show_parts(void *arg
)
238 /* linux just needs an open() to trigger
239 * the partition table update
240 * but remember to load the module with max_part != 0 :
241 * modprobe nbd max_part=63
243 nbd
= open(device
, O_RDWR
);
250 static void *nbd_client_thread(void *arg
)
255 QIOChannelSocket
*sioc
;
258 pthread_t show_parts_thread
;
259 Error
*local_error
= NULL
;
261 sioc
= qio_channel_socket_new();
262 if (qio_channel_socket_connect_sync(sioc
,
265 error_report_err(local_error
);
269 ret
= nbd_receive_negotiate(QIO_CHANNEL(sioc
), NULL
, &nbdflags
,
271 &size
, &local_error
);
274 error_report_err(local_error
);
279 fd
= open(device
, O_RDWR
);
281 /* Linux-only, we can use %m in printf. */
282 error_report("Failed to open %s: %m", device
);
286 ret
= nbd_init(fd
, sioc
, nbdflags
, size
);
291 /* update partition table */
292 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
295 fprintf(stderr
, "NBD device %s is now connected to %s\n",
298 /* Close stderr so that the qemu-nbd process exits. */
299 dup2(STDOUT_FILENO
, STDERR_FILENO
);
302 ret
= nbd_client(fd
);
307 object_unref(OBJECT(sioc
));
308 kill(getpid(), SIGTERM
);
309 return (void *) EXIT_SUCCESS
;
314 object_unref(OBJECT(sioc
));
316 kill(getpid(), SIGTERM
);
317 return (void *) EXIT_FAILURE
;
320 static int nbd_can_accept(void)
322 return nb_fds
< shared
;
325 static void nbd_export_closed(NBDExport
*exp
)
327 assert(state
== TERMINATING
);
331 static void nbd_update_server_watch(void);
333 static void nbd_client_closed(NBDClient
*client
)
336 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
339 nbd_update_server_watch();
340 nbd_client_put(client
);
343 static gboolean
nbd_accept(QIOChannel
*ioc
, GIOCondition cond
, gpointer opaque
)
345 QIOChannelSocket
*cioc
;
347 cioc
= qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc
),
353 if (state
>= TERMINATE
) {
354 object_unref(OBJECT(cioc
));
359 nbd_update_server_watch();
360 nbd_client_new(newproto
? NULL
: exp
, cioc
,
361 tlscreds
, NULL
, nbd_client_closed
);
362 object_unref(OBJECT(cioc
));
367 static void nbd_update_server_watch(void)
369 if (nbd_can_accept()) {
370 if (server_watch
== -1) {
371 server_watch
= qio_channel_add_watch(QIO_CHANNEL(server_ioc
),
377 if (server_watch
!= -1) {
378 g_source_remove(server_watch
);
385 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
389 SocketAddress
*saddr
;
391 saddr
= g_new0(SocketAddress
, 1);
393 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
394 saddr
->u
.q_unix
.data
= g_new0(UnixSocketAddress
, 1);
395 saddr
->u
.q_unix
.data
->path
= g_strdup(sockpath
);
397 InetSocketAddress
*inet
;
398 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
399 inet
= saddr
->u
.inet
.data
= g_new0(InetSocketAddress
, 1);
400 inet
->host
= g_strdup(bindto
);
402 inet
->port
= g_strdup(port
);
404 inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
412 static QemuOptsList file_opts
= {
414 .implied_opt_name
= "file",
415 .head
= QTAILQ_HEAD_INITIALIZER(file_opts
.head
),
417 /* no elements => accept any params */
418 { /* end of list */ }
422 static QemuOptsList qemu_object_opts
= {
424 .implied_opt_name
= "qom-type",
425 .head
= QTAILQ_HEAD_INITIALIZER(qemu_object_opts
.head
),
433 static QCryptoTLSCreds
*nbd_get_tls_creds(const char *id
, Error
**errp
)
436 QCryptoTLSCreds
*creds
;
438 obj
= object_resolve_path_component(
439 object_get_objects_root(), id
);
441 error_setg(errp
, "No TLS credentials with id '%s'",
445 creds
= (QCryptoTLSCreds
*)
446 object_dynamic_cast(obj
, TYPE_QCRYPTO_TLS_CREDS
);
448 error_setg(errp
, "Object with id '%s' is not TLS credentials",
453 if (creds
->endpoint
!= QCRYPTO_TLS_CREDS_ENDPOINT_SERVER
) {
455 "Expecting TLS credentials with a server endpoint");
463 int main(int argc
, char **argv
)
466 BlockDriverState
*bs
;
467 off_t dev_offset
= 0;
468 uint32_t nbdflags
= 0;
469 bool disconnect
= false;
470 const char *bindto
= "0.0.0.0";
471 const char *port
= NULL
;
472 char *sockpath
= NULL
;
475 QemuOpts
*sn_opts
= NULL
;
476 const char *sn_id_or_name
= NULL
;
477 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:";
478 struct option lopt
[] = {
479 { "help", no_argument
, NULL
, 'h' },
480 { "version", no_argument
, NULL
, 'V' },
481 { "bind", required_argument
, NULL
, 'b' },
482 { "port", required_argument
, NULL
, 'p' },
483 { "socket", required_argument
, NULL
, 'k' },
484 { "offset", required_argument
, NULL
, 'o' },
485 { "read-only", no_argument
, NULL
, 'r' },
486 { "partition", required_argument
, NULL
, 'P' },
487 { "connect", required_argument
, NULL
, 'c' },
488 { "disconnect", no_argument
, NULL
, 'd' },
489 { "snapshot", no_argument
, NULL
, 's' },
490 { "load-snapshot", required_argument
, NULL
, 'l' },
491 { "nocache", no_argument
, NULL
, 'n' },
492 { "cache", required_argument
, NULL
, QEMU_NBD_OPT_CACHE
},
493 { "aio", required_argument
, NULL
, QEMU_NBD_OPT_AIO
},
494 { "discard", required_argument
, NULL
, QEMU_NBD_OPT_DISCARD
},
495 { "detect-zeroes", required_argument
, NULL
,
496 QEMU_NBD_OPT_DETECT_ZEROES
},
497 { "shared", required_argument
, NULL
, 'e' },
498 { "format", required_argument
, NULL
, 'f' },
499 { "persistent", no_argument
, NULL
, 't' },
500 { "verbose", no_argument
, NULL
, 'v' },
501 { "object", required_argument
, NULL
, QEMU_NBD_OPT_OBJECT
},
502 { "export-name", required_argument
, NULL
, 'x' },
503 { "tls-creds", required_argument
, NULL
, QEMU_NBD_OPT_TLSCREDS
},
504 { "image-opts", no_argument
, NULL
, QEMU_NBD_OPT_IMAGE_OPTS
},
505 { "trace", required_argument
, NULL
, 'T' },
511 int flags
= BDRV_O_RDWR
;
514 bool seen_cache
= false;
515 bool seen_discard
= false;
516 bool seen_aio
= false;
517 pthread_t client_thread
;
518 const char *fmt
= NULL
;
519 Error
*local_err
= NULL
;
520 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
521 QDict
*options
= NULL
;
522 const char *export_name
= NULL
;
523 const char *tlscredsid
= NULL
;
524 bool imageOpts
= false;
525 bool writethrough
= true;
526 char *trace_file
= NULL
;
528 /* The client thread uses SIGTERM to interrupt the server. A signal
529 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
531 struct sigaction sa_sigterm
;
532 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
533 sa_sigterm
.sa_handler
= termsig_handler
;
534 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
536 qcrypto_init(&error_fatal
);
538 module_call_init(MODULE_INIT_QOM
);
539 qemu_add_opts(&qemu_object_opts
);
540 qemu_add_opts(&qemu_trace_opts
);
541 qemu_init_exec_dir(argv
[0]);
543 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
546 flags
|= BDRV_O_SNAPSHOT
;
549 optarg
= (char *) "none";
551 case QEMU_NBD_OPT_CACHE
:
553 error_report("-n and --cache can only be specified once");
557 if (bdrv_parse_cache_mode(optarg
, &flags
, &writethrough
) == -1) {
558 error_report("Invalid cache mode `%s'", optarg
);
562 case QEMU_NBD_OPT_AIO
:
564 error_report("--aio can only be specified once");
568 if (!strcmp(optarg
, "native")) {
569 flags
|= BDRV_O_NATIVE_AIO
;
570 } else if (!strcmp(optarg
, "threads")) {
571 /* this is the default */
573 error_report("invalid aio mode `%s'", optarg
);
577 case QEMU_NBD_OPT_DISCARD
:
579 error_report("--discard can only be specified once");
583 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
584 error_report("Invalid discard mode `%s'", optarg
);
588 case QEMU_NBD_OPT_DETECT_ZEROES
:
590 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
592 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
593 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
596 error_reportf_err(local_err
,
597 "Failed to parse detect_zeroes mode: ");
600 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
601 !(flags
& BDRV_O_UNMAP
)) {
602 error_report("setting detect-zeroes to unmap is not allowed "
603 "without setting discard operation to unmap");
614 dev_offset
= strtoll (optarg
, &end
, 0);
616 error_report("Invalid offset `%s'", optarg
);
619 if (dev_offset
< 0) {
620 error_report("Offset must be positive `%s'", optarg
);
625 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
626 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
629 error_report("Failed in parsing snapshot param `%s'",
634 sn_id_or_name
= optarg
;
638 nbdflags
|= NBD_FLAG_READ_ONLY
;
639 flags
&= ~BDRV_O_RDWR
;
642 partition
= strtol(optarg
, &end
, 0);
644 error_report("Invalid partition `%s'", optarg
);
647 if (partition
< 1 || partition
> 8) {
648 error_report("Invalid partition %d", partition
);
654 if (sockpath
[0] != '/') {
655 error_report("socket path must be absolute");
666 shared
= strtol(optarg
, &end
, 0);
668 error_report("Invalid shared device number '%s'", optarg
);
672 error_report("Shared device number must be greater than 0");
683 export_name
= optarg
;
697 error_report("Try `%s --help' for more information.", argv
[0]);
699 case QEMU_NBD_OPT_OBJECT
: {
701 opts
= qemu_opts_parse_noisily(&qemu_object_opts
,
707 case QEMU_NBD_OPT_TLSCREDS
:
710 case QEMU_NBD_OPT_IMAGE_OPTS
:
715 trace_file
= trace_opt_parse(optarg
);
720 if ((argc
- optind
) != 1) {
721 error_report("Invalid number of arguments");
722 error_printf("Try `%s --help' for more information.\n", argv
[0]);
726 if (qemu_opts_foreach(&qemu_object_opts
,
727 user_creatable_add_opts_foreach
,
732 if (!trace_init_backends()) {
735 trace_init_file(trace_file
);
736 qemu_set_log(LOG_TRACE
);
740 error_report("TLS is only supported with IPv4/IPv6");
744 error_report("TLS is not supported with a host device");
748 /* Set the default NBD protocol export name, since
749 * we *must* use new style protocol for TLS */
752 tlscreds
= nbd_get_tls_creds(tlscredsid
, &local_err
);
754 error_report("Failed to get TLS creds %s",
755 error_get_pretty(local_err
));
761 int nbdfd
= open(argv
[optind
], O_RDWR
);
763 error_report("Cannot open %s: %s", argv
[optind
],
767 nbd_disconnect(nbdfd
);
771 printf("%s disconnected\n", argv
[optind
]);
776 if (device
&& !verbose
) {
781 if (qemu_pipe(stderr_fd
) < 0) {
782 error_report("Error setting up communication pipe: %s",
787 /* Now daemonize, but keep a communication channel open to
788 * print errors and exit with the proper status code.
792 error_report("Failed to fork: %s", strerror(errno
));
794 } else if (pid
== 0) {
796 ret
= qemu_daemon(1, 0);
798 /* Temporarily redirect stderr to the parent's pipe... */
799 dup2(stderr_fd
[1], STDERR_FILENO
);
801 error_report("Failed to daemonize: %s", strerror(errno
));
805 /* ... close the descriptor we inherited and go on. */
811 /* In the parent. Print error messages from the child until
812 * it closes the pipe.
815 buf
= g_malloc(1024);
816 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
818 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
824 error_report("Cannot read from daemon: %s",
829 /* Usually the daemon should not print any message.
830 * Exit with zero status in that case.
836 if (device
!= NULL
&& sockpath
== NULL
) {
837 sockpath
= g_malloc(128);
838 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
841 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
843 if (qemu_init_main_loop(&local_err
)) {
844 error_report_err(local_err
);
848 atexit(bdrv_close_all
);
850 srcpath
= argv
[optind
];
854 error_report("--image-opts and -f are mutually exclusive");
857 opts
= qemu_opts_parse_noisily(&file_opts
, srcpath
, true);
859 qemu_opts_reset(&file_opts
);
862 options
= qemu_opts_to_qdict(opts
, NULL
);
863 qemu_opts_reset(&file_opts
);
864 blk
= blk_new_open(NULL
, NULL
, options
, flags
, &local_err
);
867 options
= qdict_new();
868 qdict_put(options
, "driver", qstring_from_str(fmt
));
870 blk
= blk_new_open(srcpath
, NULL
, options
, flags
, &local_err
);
874 error_reportf_err(local_err
, "Failed to blk_new_open '%s': ",
880 blk_set_enable_write_cache(blk
, !writethrough
);
883 ret
= bdrv_snapshot_load_tmp(bs
,
884 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
885 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
887 } else if (sn_id_or_name
) {
888 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
892 error_reportf_err(local_err
, "Failed to load snapshot: ");
896 bs
->detect_zeroes
= detect_zeroes
;
897 fd_size
= blk_getlength(blk
);
899 error_report("Failed to determine the image length: %s",
904 if (partition
!= -1) {
905 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
907 error_report("Could not find partition %d: %s", partition
,
913 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
916 error_report_err(local_err
);
920 nbd_export_set_name(exp
, export_name
);
924 server_ioc
= qio_channel_socket_new();
925 if (qio_channel_socket_listen_sync(server_ioc
, saddr
, &local_err
) < 0) {
926 object_unref(OBJECT(server_ioc
));
927 error_report_err(local_err
);
934 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
936 error_report("Failed to create client thread: %s", strerror(ret
));
940 /* Shut up GCC warnings. */
941 memset(&client_thread
, 0, sizeof(client_thread
));
944 nbd_update_server_watch();
946 /* now when the initialization is (almost) complete, chdir("/")
947 * to free any busy filesystems */
948 if (chdir("/") < 0) {
949 error_report("Could not chdir to root directory: %s",
956 main_loop_wait(false);
957 if (state
== TERMINATE
) {
959 nbd_export_close(exp
);
963 } while (state
!= TERMINATED
);
970 qemu_opts_del(sn_opts
);
974 pthread_join(client_thread
, &ret
);