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"
34 #include <sys/types.h>
35 #include <sys/socket.h>
36 #include <netinet/in.h>
37 #include <netinet/tcp.h>
38 #include <arpa/inet.h>
43 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
44 #define QEMU_NBD_OPT_CACHE 1
45 #define QEMU_NBD_OPT_AIO 2
46 #define QEMU_NBD_OPT_DISCARD 3
47 #define QEMU_NBD_OPT_DETECT_ZEROES 4
49 static NBDExport
*exp
;
52 static SocketAddress
*saddr
;
53 static int persistent
= 0;
54 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
55 static int shared
= 1;
59 static void usage(const char *name
)
62 "Usage: %s [OPTIONS] FILE\n"
63 "QEMU Disk Network Block Device Server\n"
65 " -h, --help display this help and exit\n"
66 " -V, --version output version information and exit\n"
68 "Connection properties:\n"
69 " -p, --port=PORT port to listen on (default `%d')\n"
70 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
71 " -k, --socket=PATH path to the unix socket\n"
72 " (default '"SOCKET_PATH
"')\n"
73 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
74 " -t, --persistent don't exit on the last connection\n"
75 " -v, --verbose display extra debugging information\n"
77 "Exposing part of the image:\n"
78 " -o, --offset=OFFSET offset into the image\n"
79 " -P, --partition=NUM only expose partition NUM\n"
82 "Kernel NBD client support:\n"
83 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
84 " -d, --disconnect disconnect the specified device\n"
88 "Block device options:\n"
89 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
90 " -r, --read-only export read-only\n"
91 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
92 " file with backing_file=FILE, redirect the write to\n"
93 " the temporary one\n"
94 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
95 " load an internal snapshot inside FILE and export it\n"
96 " as an read-only device, SNAPSHOT_PARAM format is\n"
97 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
99 " -n, --nocache disable host cache\n"
100 " --cache=MODE set cache mode (none, writeback, ...)\n"
101 #ifdef CONFIG_LINUX_AIO
102 " --aio=MODE set AIO mode (native or threads)\n"
104 " --discard=MODE set discard mode (ignore, unmap)\n"
105 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
107 "Report bugs to <qemu-devel@nongnu.org>\n"
108 , name
, NBD_DEFAULT_PORT
, "DEVICE");
111 static void version(const char *name
)
115 "Written by Anthony Liguori.\n"
117 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
118 "This is free software; see the source for copying conditions. There is NO\n"
119 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
123 struct partition_record
127 uint32_t start_cylinder
;
128 uint8_t start_sector
;
131 uint8_t end_cylinder
;
133 uint32_t start_sector_abs
;
134 uint32_t nb_sectors_abs
;
137 static void read_partition(uint8_t *p
, struct partition_record
*r
)
140 r
->start_head
= p
[1];
141 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
142 r
->start_sector
= p
[2] & 0x3f;
145 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
146 r
->end_sector
= p
[6] & 0x3f;
148 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
149 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
152 static int find_partition(BlockBackend
*blk
, int partition
,
153 off_t
*offset
, off_t
*size
)
155 struct partition_record mbr
[4];
161 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
163 err(EXIT_FAILURE
, "error while reading");
166 if (data
[510] != 0x55 || data
[511] != 0xaa) {
170 for (i
= 0; i
< 4; i
++) {
171 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
173 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
177 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
178 struct partition_record ext
[4];
182 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
184 err(EXIT_FAILURE
, "error while reading");
187 for (j
= 0; j
< 4; j
++) {
188 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
189 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
193 if ((ext_partnum
+ j
+ 1) == partition
) {
194 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
195 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
200 } else if ((i
+ 1) == partition
) {
201 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
202 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
210 static void termsig_handler(int signum
)
217 static void *show_parts(void *arg
)
222 /* linux just needs an open() to trigger
223 * the partition table update
224 * but remember to load the module with max_part != 0 :
225 * modprobe nbd max_part=63
227 nbd
= open(device
, O_RDWR
);
234 static void *nbd_client_thread(void *arg
)
241 pthread_t show_parts_thread
;
242 Error
*local_error
= NULL
;
245 sock
= socket_connect(saddr
, &local_error
, NULL
, NULL
);
247 error_report_err(local_error
);
251 ret
= nbd_receive_negotiate(sock
, NULL
, &nbdflags
,
252 &size
, &local_error
);
255 fprintf(stderr
, "%s\n", error_get_pretty(local_error
));
256 error_free(local_error
);
261 fd
= open(device
, O_RDWR
);
263 /* Linux-only, we can use %m in printf. */
264 fprintf(stderr
, "Failed to open %s: %m\n", device
);
268 ret
= nbd_init(fd
, sock
, 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 kill(getpid(), SIGTERM
);
290 return (void *) EXIT_SUCCESS
;
297 kill(getpid(), SIGTERM
);
298 return (void *) EXIT_FAILURE
;
301 static int nbd_can_accept(void)
303 return nb_fds
< shared
;
306 static void nbd_export_closed(NBDExport
*exp
)
308 assert(state
== TERMINATING
);
312 static void nbd_update_server_fd_handler(int fd
);
314 static void nbd_client_closed(NBDClient
*client
)
317 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
320 nbd_update_server_fd_handler(server_fd
);
321 nbd_client_put(client
);
324 static void nbd_accept(void *opaque
)
326 struct sockaddr_in addr
;
327 socklen_t addr_len
= sizeof(addr
);
329 int fd
= accept(server_fd
, (struct sockaddr
*)&addr
, &addr_len
);
335 if (state
>= TERMINATE
) {
340 if (nbd_client_new(exp
, fd
, nbd_client_closed
)) {
342 nbd_update_server_fd_handler(server_fd
);
349 static void nbd_update_server_fd_handler(int fd
)
351 if (nbd_can_accept()) {
352 qemu_set_fd_handler(fd
, nbd_accept
, NULL
, (void *)(uintptr_t)fd
);
354 qemu_set_fd_handler(fd
, NULL
, NULL
, NULL
);
359 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
363 SocketAddress
*saddr
;
365 saddr
= g_new0(SocketAddress
, 1);
367 saddr
->kind
= SOCKET_ADDRESS_KIND_UNIX
;
368 saddr
->q_unix
= g_new0(UnixSocketAddress
, 1);
369 saddr
->q_unix
->path
= g_strdup(sockpath
);
371 saddr
->kind
= SOCKET_ADDRESS_KIND_INET
;
372 saddr
->inet
= g_new0(InetSocketAddress
, 1);
373 saddr
->inet
->host
= g_strdup(bindto
);
375 saddr
->inet
->port
= g_strdup(port
);
377 saddr
->inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
385 int main(int argc
, char **argv
)
388 BlockDriverState
*bs
;
389 off_t dev_offset
= 0;
390 uint32_t nbdflags
= 0;
391 bool disconnect
= false;
392 const char *bindto
= "0.0.0.0";
393 const char *port
= NULL
;
394 char *sockpath
= NULL
;
397 QemuOpts
*sn_opts
= NULL
;
398 const char *sn_id_or_name
= NULL
;
399 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
400 struct option lopt
[] = {
401 { "help", 0, NULL
, 'h' },
402 { "version", 0, NULL
, 'V' },
403 { "bind", 1, NULL
, 'b' },
404 { "port", 1, NULL
, 'p' },
405 { "socket", 1, NULL
, 'k' },
406 { "offset", 1, NULL
, 'o' },
407 { "read-only", 0, NULL
, 'r' },
408 { "partition", 1, NULL
, 'P' },
409 { "connect", 1, NULL
, 'c' },
410 { "disconnect", 0, NULL
, 'd' },
411 { "snapshot", 0, NULL
, 's' },
412 { "load-snapshot", 1, NULL
, 'l' },
413 { "nocache", 0, NULL
, 'n' },
414 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
415 #ifdef CONFIG_LINUX_AIO
416 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
418 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
419 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
420 { "shared", 1, NULL
, 'e' },
421 { "format", 1, NULL
, 'f' },
422 { "persistent", 0, NULL
, 't' },
423 { "verbose", 0, NULL
, 'v' },
429 int flags
= BDRV_O_RDWR
;
433 bool seen_cache
= false;
434 bool seen_discard
= false;
435 #ifdef CONFIG_LINUX_AIO
436 bool seen_aio
= false;
438 pthread_t client_thread
;
439 const char *fmt
= NULL
;
440 Error
*local_err
= NULL
;
441 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
442 QDict
*options
= NULL
;
444 /* The client thread uses SIGTERM to interrupt the server. A signal
445 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
447 struct sigaction sa_sigterm
;
448 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
449 sa_sigterm
.sa_handler
= termsig_handler
;
450 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
451 qemu_init_exec_dir(argv
[0]);
453 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
456 flags
|= BDRV_O_SNAPSHOT
;
459 optarg
= (char *) "none";
461 case QEMU_NBD_OPT_CACHE
:
463 errx(EXIT_FAILURE
, "-n and --cache can only be specified once");
466 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
467 errx(EXIT_FAILURE
, "Invalid cache mode `%s'", optarg
);
470 #ifdef CONFIG_LINUX_AIO
471 case QEMU_NBD_OPT_AIO
:
473 errx(EXIT_FAILURE
, "--aio can only be specified once");
476 if (!strcmp(optarg
, "native")) {
477 flags
|= BDRV_O_NATIVE_AIO
;
478 } else if (!strcmp(optarg
, "threads")) {
479 /* this is the default */
481 errx(EXIT_FAILURE
, "invalid aio mode `%s'", optarg
);
485 case QEMU_NBD_OPT_DISCARD
:
487 errx(EXIT_FAILURE
, "--discard can only be specified once");
490 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
491 errx(EXIT_FAILURE
, "Invalid discard mode `%s'", optarg
);
494 case QEMU_NBD_OPT_DETECT_ZEROES
:
496 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
498 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX
,
499 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
502 errx(EXIT_FAILURE
, "Failed to parse detect_zeroes mode: %s",
503 error_get_pretty(local_err
));
505 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
506 !(flags
& BDRV_O_UNMAP
)) {
507 errx(EXIT_FAILURE
, "setting detect-zeroes to unmap is not allowed "
508 "without setting discard operation to unmap");
518 dev_offset
= strtoll (optarg
, &end
, 0);
520 errx(EXIT_FAILURE
, "Invalid offset `%s'", optarg
);
522 if (dev_offset
< 0) {
523 errx(EXIT_FAILURE
, "Offset must be positive `%s'", optarg
);
527 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
528 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
531 errx(EXIT_FAILURE
, "Failed in parsing snapshot param `%s'",
535 sn_id_or_name
= optarg
;
539 nbdflags
|= NBD_FLAG_READ_ONLY
;
540 flags
&= ~BDRV_O_RDWR
;
543 partition
= strtol(optarg
, &end
, 0);
545 errx(EXIT_FAILURE
, "Invalid partition `%s'", optarg
);
547 if (partition
< 1 || partition
> 8) {
548 errx(EXIT_FAILURE
, "Invalid partition %d", partition
);
553 if (sockpath
[0] != '/') {
554 errx(EXIT_FAILURE
, "socket path must be absolute\n");
564 shared
= strtol(optarg
, &end
, 0);
566 errx(EXIT_FAILURE
, "Invalid shared device number '%s'", optarg
);
569 errx(EXIT_FAILURE
, "Shared device number must be greater than 0\n");
590 errx(EXIT_FAILURE
, "Try `%s --help' for more information.",
595 if ((argc
- optind
) != 1) {
596 errx(EXIT_FAILURE
, "Invalid number of argument.\n"
597 "Try `%s --help' for more information.",
602 fd
= open(argv
[optind
], O_RDWR
);
604 err(EXIT_FAILURE
, "Cannot open %s", argv
[optind
]);
610 printf("%s disconnected\n", argv
[optind
]);
615 if (device
&& !verbose
) {
620 if (qemu_pipe(stderr_fd
) < 0) {
621 err(EXIT_FAILURE
, "Error setting up communication pipe");
624 /* Now daemonize, but keep a communication channel open to
625 * print errors and exit with the proper status code.
629 err(EXIT_FAILURE
, "Failed to fork");
630 } else if (pid
== 0) {
632 ret
= qemu_daemon(1, 0);
634 /* Temporarily redirect stderr to the parent's pipe... */
635 dup2(stderr_fd
[1], STDERR_FILENO
);
637 err(EXIT_FAILURE
, "Failed to daemonize");
640 /* ... close the descriptor we inherited and go on. */
646 /* In the parent. Print error messages from the child until
647 * it closes the pipe.
650 buf
= g_malloc(1024);
651 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
653 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
659 err(EXIT_FAILURE
, "Cannot read from daemon");
662 /* Usually the daemon should not print any message.
663 * Exit with zero status in that case.
669 if (device
!= NULL
&& sockpath
== NULL
) {
670 sockpath
= g_malloc(128);
671 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
674 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
676 if (qemu_init_main_loop(&local_err
)) {
677 error_report_err(local_err
);
681 atexit(bdrv_close_all
);
684 options
= qdict_new();
685 qdict_put(options
, "driver", qstring_from_str(fmt
));
688 srcpath
= argv
[optind
];
689 blk
= blk_new_open("hda", srcpath
, NULL
, options
, flags
, &local_err
);
691 errx(EXIT_FAILURE
, "Failed to blk_new_open '%s': %s", argv
[optind
],
692 error_get_pretty(local_err
));
697 ret
= bdrv_snapshot_load_tmp(bs
,
698 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
699 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
701 } else if (sn_id_or_name
) {
702 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
708 "Failed to load snapshot: %s",
709 error_get_pretty(local_err
));
712 bs
->detect_zeroes
= detect_zeroes
;
713 fd_size
= blk_getlength(blk
);
715 errx(EXIT_FAILURE
, "Failed to determine the image length: %s",
719 if (partition
!= -1) {
720 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
723 err(EXIT_FAILURE
, "Could not find partition %d", partition
);
727 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
730 errx(EXIT_FAILURE
, "%s", error_get_pretty(local_err
));
733 fd
= socket_listen(saddr
, &local_err
);
735 error_report_err(local_err
);
742 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
744 errx(EXIT_FAILURE
, "Failed to create client thread: %s",
748 /* Shut up GCC warnings. */
749 memset(&client_thread
, 0, sizeof(client_thread
));
753 nbd_update_server_fd_handler(fd
);
755 /* now when the initialization is (almost) complete, chdir("/")
756 * to free any busy filesystems */
757 if (chdir("/") < 0) {
758 err(EXIT_FAILURE
, "Could not chdir to root directory");
763 main_loop_wait(false);
764 if (state
== TERMINATE
) {
766 nbd_export_close(exp
);
770 } while (state
!= TERMINATED
);
777 qemu_opts_del(sn_opts
);
781 pthread_join(client_thread
, &ret
);