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
46 static NBDExport
*exp
;
49 static SocketAddress
*saddr
;
50 static int persistent
= 0;
51 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
52 static int shared
= 1;
54 static QIOChannelSocket
*server_ioc
;
55 static int server_watch
= -1;
57 static void usage(const char *name
)
60 "Usage: %s [OPTIONS] FILE\n"
61 "QEMU Disk Network Block Device Server\n"
63 " -h, --help display this help and exit\n"
64 " -V, --version output version information and exit\n"
66 "Connection properties:\n"
67 " -p, --port=PORT port to listen on (default `%d')\n"
68 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
69 " -k, --socket=PATH path to the unix socket\n"
70 " (default '"SOCKET_PATH
"')\n"
71 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
72 " -t, --persistent don't exit on the last connection\n"
73 " -v, --verbose display extra debugging information\n"
75 "Exposing part of the image:\n"
76 " -o, --offset=OFFSET offset into the image\n"
77 " -P, --partition=NUM only expose partition NUM\n"
79 "General purpose options:\n"
80 " --object type,id=ID,... define an object such as 'secret' for providing\n"
81 " passwords and/or encryption keys\n"
83 "Kernel NBD client support:\n"
84 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
85 " -d, --disconnect disconnect the specified device\n"
89 "Block device options:\n"
90 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
91 " -r, --read-only export read-only\n"
92 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
93 " file with backing_file=FILE, redirect the write to\n"
94 " the temporary one\n"
95 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
96 " load an internal snapshot inside FILE and export it\n"
97 " as an read-only device, SNAPSHOT_PARAM format is\n"
98 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
100 " -n, --nocache disable host cache\n"
101 " --cache=MODE set cache mode (none, writeback, ...)\n"
102 " --aio=MODE set AIO mode (native or threads)\n"
103 " --discard=MODE set discard mode (ignore, unmap)\n"
104 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
106 "Report bugs to <qemu-devel@nongnu.org>\n"
107 , name
, NBD_DEFAULT_PORT
, "DEVICE");
110 static void version(const char *name
)
114 "Written by Anthony Liguori.\n"
116 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
117 "This is free software; see the source for copying conditions. There is NO\n"
118 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
122 struct partition_record
126 uint32_t start_cylinder
;
127 uint8_t start_sector
;
130 uint8_t end_cylinder
;
132 uint32_t start_sector_abs
;
133 uint32_t nb_sectors_abs
;
136 static void read_partition(uint8_t *p
, struct partition_record
*r
)
139 r
->start_head
= p
[1];
140 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
141 r
->start_sector
= p
[2] & 0x3f;
144 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
145 r
->end_sector
= p
[6] & 0x3f;
147 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
148 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
151 static int find_partition(BlockBackend
*blk
, int partition
,
152 off_t
*offset
, off_t
*size
)
154 struct partition_record mbr
[4];
160 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
161 error_report("error while reading: %s", strerror(-ret
));
165 if (data
[510] != 0x55 || data
[511] != 0xaa) {
169 for (i
= 0; i
< 4; i
++) {
170 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
172 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
176 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
177 struct partition_record ext
[4];
181 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
182 error_report("error while reading: %s", strerror(-ret
));
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
)
238 QIOChannelSocket
*sioc
;
241 pthread_t show_parts_thread
;
242 Error
*local_error
= NULL
;
244 sioc
= qio_channel_socket_new();
245 if (qio_channel_socket_connect_sync(sioc
,
248 error_report_err(local_error
);
252 ret
= nbd_receive_negotiate(QIO_CHANNEL(sioc
), NULL
, &nbdflags
,
253 &size
, &local_error
);
256 error_report_err(local_error
);
261 fd
= open(device
, O_RDWR
);
263 /* Linux-only, we can use %m in printf. */
264 error_report("Failed to open %s: %m", device
);
268 ret
= nbd_init(fd
, sioc
, nbdflags
, size
);
273 /* update partition table */
274 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
277 fprintf(stderr
, "NBD device %s is now connected to %s\n",
280 /* Close stderr so that the qemu-nbd process exits. */
281 dup2(STDOUT_FILENO
, STDERR_FILENO
);
284 ret
= nbd_client(fd
);
289 object_unref(OBJECT(sioc
));
290 kill(getpid(), SIGTERM
);
291 return (void *) EXIT_SUCCESS
;
296 object_unref(OBJECT(sioc
));
298 kill(getpid(), SIGTERM
);
299 return (void *) EXIT_FAILURE
;
302 static int nbd_can_accept(void)
304 return nb_fds
< shared
;
307 static void nbd_export_closed(NBDExport
*exp
)
309 assert(state
== TERMINATING
);
313 static void nbd_update_server_watch(void);
315 static void nbd_client_closed(NBDClient
*client
)
318 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
321 nbd_update_server_watch();
322 nbd_client_put(client
);
325 static gboolean
nbd_accept(QIOChannel
*ioc
, GIOCondition cond
, gpointer opaque
)
327 QIOChannelSocket
*cioc
;
329 cioc
= qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc
),
335 if (state
>= TERMINATE
) {
336 object_unref(OBJECT(cioc
));
341 nbd_update_server_watch();
342 nbd_client_new(exp
, cioc
, nbd_client_closed
);
343 object_unref(OBJECT(cioc
));
348 static void nbd_update_server_watch(void)
350 if (nbd_can_accept()) {
351 if (server_watch
== -1) {
352 server_watch
= qio_channel_add_watch(QIO_CHANNEL(server_ioc
),
358 if (server_watch
!= -1) {
359 g_source_remove(server_watch
);
366 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
370 SocketAddress
*saddr
;
372 saddr
= g_new0(SocketAddress
, 1);
374 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
375 saddr
->u
.q_unix
= g_new0(UnixSocketAddress
, 1);
376 saddr
->u
.q_unix
->path
= g_strdup(sockpath
);
378 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
379 saddr
->u
.inet
= g_new0(InetSocketAddress
, 1);
380 saddr
->u
.inet
->host
= g_strdup(bindto
);
382 saddr
->u
.inet
->port
= g_strdup(port
);
384 saddr
->u
.inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
392 static QemuOptsList qemu_object_opts
= {
394 .implied_opt_name
= "qom-type",
395 .head
= QTAILQ_HEAD_INITIALIZER(qemu_object_opts
.head
),
402 int main(int argc
, char **argv
)
405 BlockDriverState
*bs
;
406 off_t dev_offset
= 0;
407 uint32_t nbdflags
= 0;
408 bool disconnect
= false;
409 const char *bindto
= "0.0.0.0";
410 const char *port
= NULL
;
411 char *sockpath
= NULL
;
414 QemuOpts
*sn_opts
= NULL
;
415 const char *sn_id_or_name
= NULL
;
416 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
417 struct option lopt
[] = {
418 { "help", 0, NULL
, 'h' },
419 { "version", 0, NULL
, 'V' },
420 { "bind", 1, NULL
, 'b' },
421 { "port", 1, NULL
, 'p' },
422 { "socket", 1, NULL
, 'k' },
423 { "offset", 1, NULL
, 'o' },
424 { "read-only", 0, NULL
, 'r' },
425 { "partition", 1, NULL
, 'P' },
426 { "connect", 1, NULL
, 'c' },
427 { "disconnect", 0, NULL
, 'd' },
428 { "snapshot", 0, NULL
, 's' },
429 { "load-snapshot", 1, NULL
, 'l' },
430 { "nocache", 0, NULL
, 'n' },
431 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
432 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
433 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
434 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
435 { "shared", 1, NULL
, 'e' },
436 { "format", 1, NULL
, 'f' },
437 { "persistent", 0, NULL
, 't' },
438 { "verbose", 0, NULL
, 'v' },
439 { "object", 1, NULL
, QEMU_NBD_OPT_OBJECT
},
445 int flags
= BDRV_O_RDWR
;
448 bool seen_cache
= false;
449 bool seen_discard
= false;
450 bool seen_aio
= false;
451 pthread_t client_thread
;
452 const char *fmt
= NULL
;
453 Error
*local_err
= NULL
;
454 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
455 QDict
*options
= NULL
;
457 /* The client thread uses SIGTERM to interrupt the server. A signal
458 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
460 struct sigaction sa_sigterm
;
461 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
462 sa_sigterm
.sa_handler
= termsig_handler
;
463 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
464 module_call_init(MODULE_INIT_QOM
);
465 qemu_add_opts(&qemu_object_opts
);
466 qemu_init_exec_dir(argv
[0]);
468 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
471 flags
|= BDRV_O_SNAPSHOT
;
474 optarg
= (char *) "none";
476 case QEMU_NBD_OPT_CACHE
:
478 error_report("-n and --cache can only be specified once");
482 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
483 error_report("Invalid cache mode `%s'", optarg
);
487 case QEMU_NBD_OPT_AIO
:
489 error_report("--aio can only be specified once");
493 if (!strcmp(optarg
, "native")) {
494 flags
|= BDRV_O_NATIVE_AIO
;
495 } else if (!strcmp(optarg
, "threads")) {
496 /* this is the default */
498 error_report("invalid aio mode `%s'", optarg
);
502 case QEMU_NBD_OPT_DISCARD
:
504 error_report("--discard can only be specified once");
508 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
509 error_report("Invalid discard mode `%s'", optarg
);
513 case QEMU_NBD_OPT_DETECT_ZEROES
:
515 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
517 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
518 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
521 error_reportf_err(local_err
,
522 "Failed to parse detect_zeroes mode: ");
525 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
526 !(flags
& BDRV_O_UNMAP
)) {
527 error_report("setting detect-zeroes to unmap is not allowed "
528 "without setting discard operation to unmap");
539 dev_offset
= strtoll (optarg
, &end
, 0);
541 error_report("Invalid offset `%s'", optarg
);
544 if (dev_offset
< 0) {
545 error_report("Offset must be positive `%s'", optarg
);
550 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
551 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
554 error_report("Failed in parsing snapshot param `%s'",
559 sn_id_or_name
= optarg
;
563 nbdflags
|= NBD_FLAG_READ_ONLY
;
564 flags
&= ~BDRV_O_RDWR
;
567 partition
= strtol(optarg
, &end
, 0);
569 error_report("Invalid partition `%s'", optarg
);
572 if (partition
< 1 || partition
> 8) {
573 error_report("Invalid partition %d", partition
);
579 if (sockpath
[0] != '/') {
580 error_report("socket path must be absolute");
591 shared
= strtol(optarg
, &end
, 0);
593 error_report("Invalid shared device number '%s'", optarg
);
597 error_report("Shared device number must be greater than 0");
619 error_report("Try `%s --help' for more information.", argv
[0]);
621 case QEMU_NBD_OPT_OBJECT
: {
623 opts
= qemu_opts_parse_noisily(&qemu_object_opts
,
632 if ((argc
- optind
) != 1) {
633 error_report("Invalid number of arguments");
634 error_printf("Try `%s --help' for more information.\n", argv
[0]);
638 if (qemu_opts_foreach(&qemu_object_opts
,
639 user_creatable_add_opts_foreach
,
641 error_report_err(local_err
);
646 int nbdfd
= open(argv
[optind
], O_RDWR
);
648 error_report("Cannot open %s: %s", argv
[optind
],
652 nbd_disconnect(nbdfd
);
656 printf("%s disconnected\n", argv
[optind
]);
661 if (device
&& !verbose
) {
666 if (qemu_pipe(stderr_fd
) < 0) {
667 error_report("Error setting up communication pipe: %s",
672 /* Now daemonize, but keep a communication channel open to
673 * print errors and exit with the proper status code.
677 error_report("Failed to fork: %s", strerror(errno
));
679 } else if (pid
== 0) {
681 ret
= qemu_daemon(1, 0);
683 /* Temporarily redirect stderr to the parent's pipe... */
684 dup2(stderr_fd
[1], STDERR_FILENO
);
686 error_report("Failed to daemonize: %s", strerror(errno
));
690 /* ... close the descriptor we inherited and go on. */
696 /* In the parent. Print error messages from the child until
697 * it closes the pipe.
700 buf
= g_malloc(1024);
701 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
703 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
709 error_report("Cannot read from daemon: %s",
714 /* Usually the daemon should not print any message.
715 * Exit with zero status in that case.
721 if (device
!= NULL
&& sockpath
== NULL
) {
722 sockpath
= g_malloc(128);
723 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
726 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
728 if (qemu_init_main_loop(&local_err
)) {
729 error_report_err(local_err
);
733 atexit(bdrv_close_all
);
736 options
= qdict_new();
737 qdict_put(options
, "driver", qstring_from_str(fmt
));
740 srcpath
= argv
[optind
];
741 blk
= blk_new_open("hda", srcpath
, NULL
, options
, flags
, &local_err
);
743 error_reportf_err(local_err
, "Failed to blk_new_open '%s': ",
750 ret
= bdrv_snapshot_load_tmp(bs
,
751 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
752 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
754 } else if (sn_id_or_name
) {
755 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
759 error_reportf_err(local_err
, "Failed to load snapshot: ");
763 bs
->detect_zeroes
= detect_zeroes
;
764 fd_size
= blk_getlength(blk
);
766 error_report("Failed to determine the image length: %s",
771 if (partition
!= -1) {
772 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
774 error_report("Could not find partition %d: %s", partition
,
780 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
783 error_report_err(local_err
);
787 server_ioc
= qio_channel_socket_new();
788 if (qio_channel_socket_listen_sync(server_ioc
, saddr
, &local_err
) < 0) {
789 object_unref(OBJECT(server_ioc
));
790 error_report_err(local_err
);
797 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
799 error_report("Failed to create client thread: %s", strerror(ret
));
803 /* Shut up GCC warnings. */
804 memset(&client_thread
, 0, sizeof(client_thread
));
807 nbd_update_server_watch();
809 /* now when the initialization is (almost) complete, chdir("/")
810 * to free any busy filesystems */
811 if (chdir("/") < 0) {
812 error_report("Could not chdir to root directory: %s",
819 main_loop_wait(false);
820 if (state
== TERMINATE
) {
822 nbd_export_close(exp
);
826 } while (state
!= TERMINATED
);
833 qemu_opts_del(sn_opts
);
837 pthread_join(client_thread
, &ret
);