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 "qemu-common.h"
21 #include "sysemu/block-backend.h"
22 #include "block/block_int.h"
23 #include "block/nbd.h"
24 #include "qemu/main-loop.h"
25 #include "qemu/error-report.h"
26 #include "qemu/config-file.h"
27 #include "block/snapshot.h"
28 #include "qapi/util.h"
29 #include "qapi/qmp/qstring.h"
30 #include "qom/object_interfaces.h"
31 #include "io/channel-socket.h"
34 #include <sys/types.h>
39 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
40 #define QEMU_NBD_OPT_CACHE 1
41 #define QEMU_NBD_OPT_AIO 2
42 #define QEMU_NBD_OPT_DISCARD 3
43 #define QEMU_NBD_OPT_DETECT_ZEROES 4
44 #define QEMU_NBD_OPT_OBJECT 5
45 #define QEMU_NBD_OPT_TLSCREDS 6
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"
109 "Report bugs to <qemu-devel@nongnu.org>\n"
110 , name
, NBD_DEFAULT_PORT
, "DEVICE");
113 static void version(const char *name
)
117 "Written by Anthony Liguori.\n"
119 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
120 "This is free software; see the source for copying conditions. There is NO\n"
121 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
125 struct partition_record
129 uint32_t start_cylinder
;
130 uint8_t start_sector
;
133 uint8_t end_cylinder
;
135 uint32_t start_sector_abs
;
136 uint32_t nb_sectors_abs
;
139 static void read_partition(uint8_t *p
, struct partition_record
*r
)
142 r
->start_head
= p
[1];
143 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
144 r
->start_sector
= p
[2] & 0x3f;
147 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
148 r
->end_sector
= p
[6] & 0x3f;
150 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
151 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
154 static int find_partition(BlockBackend
*blk
, int partition
,
155 off_t
*offset
, off_t
*size
)
157 struct partition_record mbr
[4];
163 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
164 error_report("error while reading: %s", strerror(-ret
));
168 if (data
[510] != 0x55 || data
[511] != 0xaa) {
172 for (i
= 0; i
< 4; i
++) {
173 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
175 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
179 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
180 struct partition_record ext
[4];
184 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
185 error_report("error while reading: %s", strerror(-ret
));
189 for (j
= 0; j
< 4; j
++) {
190 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
191 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
195 if ((ext_partnum
+ j
+ 1) == partition
) {
196 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
197 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
202 } else if ((i
+ 1) == partition
) {
203 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
204 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
212 static void termsig_handler(int signum
)
219 static void *show_parts(void *arg
)
224 /* linux just needs an open() to trigger
225 * the partition table update
226 * but remember to load the module with max_part != 0 :
227 * modprobe nbd max_part=63
229 nbd
= open(device
, O_RDWR
);
236 static void *nbd_client_thread(void *arg
)
241 QIOChannelSocket
*sioc
;
244 pthread_t show_parts_thread
;
245 Error
*local_error
= NULL
;
247 sioc
= qio_channel_socket_new();
248 if (qio_channel_socket_connect_sync(sioc
,
251 error_report_err(local_error
);
255 ret
= nbd_receive_negotiate(QIO_CHANNEL(sioc
), NULL
, &nbdflags
,
257 &size
, &local_error
);
260 error_report_err(local_error
);
265 fd
= open(device
, O_RDWR
);
267 /* Linux-only, we can use %m in printf. */
268 error_report("Failed to open %s: %m", device
);
272 ret
= nbd_init(fd
, sioc
, nbdflags
, size
);
277 /* update partition table */
278 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
281 fprintf(stderr
, "NBD device %s is now connected to %s\n",
284 /* Close stderr so that the qemu-nbd process exits. */
285 dup2(STDOUT_FILENO
, STDERR_FILENO
);
288 ret
= nbd_client(fd
);
293 object_unref(OBJECT(sioc
));
294 kill(getpid(), SIGTERM
);
295 return (void *) EXIT_SUCCESS
;
300 object_unref(OBJECT(sioc
));
302 kill(getpid(), SIGTERM
);
303 return (void *) EXIT_FAILURE
;
306 static int nbd_can_accept(void)
308 return nb_fds
< shared
;
311 static void nbd_export_closed(NBDExport
*exp
)
313 assert(state
== TERMINATING
);
317 static void nbd_update_server_watch(void);
319 static void nbd_client_closed(NBDClient
*client
)
322 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
325 nbd_update_server_watch();
326 nbd_client_put(client
);
329 static gboolean
nbd_accept(QIOChannel
*ioc
, GIOCondition cond
, gpointer opaque
)
331 QIOChannelSocket
*cioc
;
333 cioc
= qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc
),
339 if (state
>= TERMINATE
) {
340 object_unref(OBJECT(cioc
));
345 nbd_update_server_watch();
346 nbd_client_new(newproto
? NULL
: exp
, cioc
,
347 tlscreds
, NULL
, nbd_client_closed
);
348 object_unref(OBJECT(cioc
));
353 static void nbd_update_server_watch(void)
355 if (nbd_can_accept()) {
356 if (server_watch
== -1) {
357 server_watch
= qio_channel_add_watch(QIO_CHANNEL(server_ioc
),
363 if (server_watch
!= -1) {
364 g_source_remove(server_watch
);
371 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
375 SocketAddress
*saddr
;
377 saddr
= g_new0(SocketAddress
, 1);
379 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
380 saddr
->u
.q_unix
= g_new0(UnixSocketAddress
, 1);
381 saddr
->u
.q_unix
->path
= g_strdup(sockpath
);
383 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
384 saddr
->u
.inet
= g_new0(InetSocketAddress
, 1);
385 saddr
->u
.inet
->host
= g_strdup(bindto
);
387 saddr
->u
.inet
->port
= g_strdup(port
);
389 saddr
->u
.inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
397 static QemuOptsList qemu_object_opts
= {
399 .implied_opt_name
= "qom-type",
400 .head
= QTAILQ_HEAD_INITIALIZER(qemu_object_opts
.head
),
408 static QCryptoTLSCreds
*nbd_get_tls_creds(const char *id
, Error
**errp
)
411 QCryptoTLSCreds
*creds
;
413 obj
= object_resolve_path_component(
414 object_get_objects_root(), id
);
416 error_setg(errp
, "No TLS credentials with id '%s'",
420 creds
= (QCryptoTLSCreds
*)
421 object_dynamic_cast(obj
, TYPE_QCRYPTO_TLS_CREDS
);
423 error_setg(errp
, "Object with id '%s' is not TLS credentials",
428 if (creds
->endpoint
!= QCRYPTO_TLS_CREDS_ENDPOINT_SERVER
) {
430 "Expecting TLS credentials with a server endpoint");
438 int main(int argc
, char **argv
)
441 BlockDriverState
*bs
;
442 off_t dev_offset
= 0;
443 uint32_t nbdflags
= 0;
444 bool disconnect
= false;
445 const char *bindto
= "0.0.0.0";
446 const char *port
= NULL
;
447 char *sockpath
= NULL
;
450 QemuOpts
*sn_opts
= NULL
;
451 const char *sn_id_or_name
= NULL
;
452 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:x:";
453 struct option lopt
[] = {
454 { "help", 0, NULL
, 'h' },
455 { "version", 0, NULL
, 'V' },
456 { "bind", 1, NULL
, 'b' },
457 { "port", 1, NULL
, 'p' },
458 { "socket", 1, NULL
, 'k' },
459 { "offset", 1, NULL
, 'o' },
460 { "read-only", 0, NULL
, 'r' },
461 { "partition", 1, NULL
, 'P' },
462 { "connect", 1, NULL
, 'c' },
463 { "disconnect", 0, NULL
, 'd' },
464 { "snapshot", 0, NULL
, 's' },
465 { "load-snapshot", 1, NULL
, 'l' },
466 { "nocache", 0, NULL
, 'n' },
467 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
468 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
469 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
470 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
471 { "shared", 1, NULL
, 'e' },
472 { "format", 1, NULL
, 'f' },
473 { "persistent", 0, NULL
, 't' },
474 { "verbose", 0, NULL
, 'v' },
475 { "object", 1, NULL
, QEMU_NBD_OPT_OBJECT
},
476 { "export-name", 1, NULL
, 'x' },
477 { "tls-creds", 1, NULL
, QEMU_NBD_OPT_TLSCREDS
},
483 int flags
= BDRV_O_RDWR
;
486 bool seen_cache
= false;
487 bool seen_discard
= false;
488 bool seen_aio
= false;
489 pthread_t client_thread
;
490 const char *fmt
= NULL
;
491 Error
*local_err
= NULL
;
492 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
493 QDict
*options
= NULL
;
494 const char *export_name
= NULL
;
495 const char *tlscredsid
= NULL
;
497 /* The client thread uses SIGTERM to interrupt the server. A signal
498 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
500 struct sigaction sa_sigterm
;
501 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
502 sa_sigterm
.sa_handler
= termsig_handler
;
503 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
504 module_call_init(MODULE_INIT_QOM
);
505 qemu_add_opts(&qemu_object_opts
);
506 qemu_init_exec_dir(argv
[0]);
508 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
511 flags
|= BDRV_O_SNAPSHOT
;
514 optarg
= (char *) "none";
516 case QEMU_NBD_OPT_CACHE
:
518 error_report("-n and --cache can only be specified once");
522 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
523 error_report("Invalid cache mode `%s'", optarg
);
527 case QEMU_NBD_OPT_AIO
:
529 error_report("--aio can only be specified once");
533 if (!strcmp(optarg
, "native")) {
534 flags
|= BDRV_O_NATIVE_AIO
;
535 } else if (!strcmp(optarg
, "threads")) {
536 /* this is the default */
538 error_report("invalid aio mode `%s'", optarg
);
542 case QEMU_NBD_OPT_DISCARD
:
544 error_report("--discard can only be specified once");
548 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
549 error_report("Invalid discard mode `%s'", optarg
);
553 case QEMU_NBD_OPT_DETECT_ZEROES
:
555 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
557 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
558 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
561 error_reportf_err(local_err
,
562 "Failed to parse detect_zeroes mode: ");
565 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
566 !(flags
& BDRV_O_UNMAP
)) {
567 error_report("setting detect-zeroes to unmap is not allowed "
568 "without setting discard operation to unmap");
579 dev_offset
= strtoll (optarg
, &end
, 0);
581 error_report("Invalid offset `%s'", optarg
);
584 if (dev_offset
< 0) {
585 error_report("Offset must be positive `%s'", optarg
);
590 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
591 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
594 error_report("Failed in parsing snapshot param `%s'",
599 sn_id_or_name
= optarg
;
603 nbdflags
|= NBD_FLAG_READ_ONLY
;
604 flags
&= ~BDRV_O_RDWR
;
607 partition
= strtol(optarg
, &end
, 0);
609 error_report("Invalid partition `%s'", optarg
);
612 if (partition
< 1 || partition
> 8) {
613 error_report("Invalid partition %d", partition
);
619 if (sockpath
[0] != '/') {
620 error_report("socket path must be absolute");
631 shared
= strtol(optarg
, &end
, 0);
633 error_report("Invalid shared device number '%s'", optarg
);
637 error_report("Shared device number must be greater than 0");
648 export_name
= optarg
;
662 error_report("Try `%s --help' for more information.", argv
[0]);
664 case QEMU_NBD_OPT_OBJECT
: {
666 opts
= qemu_opts_parse_noisily(&qemu_object_opts
,
672 case QEMU_NBD_OPT_TLSCREDS
:
678 if ((argc
- optind
) != 1) {
679 error_report("Invalid number of arguments");
680 error_printf("Try `%s --help' for more information.\n", argv
[0]);
684 if (qemu_opts_foreach(&qemu_object_opts
,
685 user_creatable_add_opts_foreach
,
687 error_report_err(local_err
);
693 error_report("TLS is only supported with IPv4/IPv6");
697 error_report("TLS is not supported with a host device");
701 /* Set the default NBD protocol export name, since
702 * we *must* use new style protocol for TLS */
705 tlscreds
= nbd_get_tls_creds(tlscredsid
, &local_err
);
707 error_report("Failed to get TLS creds %s",
708 error_get_pretty(local_err
));
714 int nbdfd
= open(argv
[optind
], O_RDWR
);
716 error_report("Cannot open %s: %s", argv
[optind
],
720 nbd_disconnect(nbdfd
);
724 printf("%s disconnected\n", argv
[optind
]);
729 if (device
&& !verbose
) {
734 if (qemu_pipe(stderr_fd
) < 0) {
735 error_report("Error setting up communication pipe: %s",
740 /* Now daemonize, but keep a communication channel open to
741 * print errors and exit with the proper status code.
745 error_report("Failed to fork: %s", strerror(errno
));
747 } else if (pid
== 0) {
749 ret
= qemu_daemon(1, 0);
751 /* Temporarily redirect stderr to the parent's pipe... */
752 dup2(stderr_fd
[1], STDERR_FILENO
);
754 error_report("Failed to daemonize: %s", strerror(errno
));
758 /* ... close the descriptor we inherited and go on. */
764 /* In the parent. Print error messages from the child until
765 * it closes the pipe.
768 buf
= g_malloc(1024);
769 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
771 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
777 error_report("Cannot read from daemon: %s",
782 /* Usually the daemon should not print any message.
783 * Exit with zero status in that case.
789 if (device
!= NULL
&& sockpath
== NULL
) {
790 sockpath
= g_malloc(128);
791 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
794 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
796 if (qemu_init_main_loop(&local_err
)) {
797 error_report_err(local_err
);
801 atexit(bdrv_close_all
);
804 options
= qdict_new();
805 qdict_put(options
, "driver", qstring_from_str(fmt
));
808 srcpath
= argv
[optind
];
809 blk
= blk_new_open("hda", srcpath
, NULL
, options
, flags
, &local_err
);
811 error_reportf_err(local_err
, "Failed to blk_new_open '%s': ",
818 ret
= bdrv_snapshot_load_tmp(bs
,
819 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
820 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
822 } else if (sn_id_or_name
) {
823 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
827 error_reportf_err(local_err
, "Failed to load snapshot: ");
831 bs
->detect_zeroes
= detect_zeroes
;
832 fd_size
= blk_getlength(blk
);
834 error_report("Failed to determine the image length: %s",
839 if (partition
!= -1) {
840 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
842 error_report("Could not find partition %d: %s", partition
,
848 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
851 error_report_err(local_err
);
855 nbd_export_set_name(exp
, export_name
);
859 server_ioc
= qio_channel_socket_new();
860 if (qio_channel_socket_listen_sync(server_ioc
, saddr
, &local_err
) < 0) {
861 object_unref(OBJECT(server_ioc
));
862 error_report_err(local_err
);
869 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
871 error_report("Failed to create client thread: %s", strerror(ret
));
875 /* Shut up GCC warnings. */
876 memset(&client_thread
, 0, sizeof(client_thread
));
879 nbd_update_server_watch();
881 /* now when the initialization is (almost) complete, chdir("/")
882 * to free any busy filesystems */
883 if (chdir("/") < 0) {
884 error_report("Could not chdir to root directory: %s",
891 main_loop_wait(false);
892 if (state
== TERMINATE
) {
894 nbd_export_close(exp
);
898 } while (state
!= TERMINATED
);
905 qemu_opts_del(sn_opts
);
909 pthread_join(client_thread
, &ret
);