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 "block/block.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"
33 #include <sys/types.h>
34 #include <sys/socket.h>
35 #include <netinet/in.h>
36 #include <netinet/tcp.h>
37 #include <arpa/inet.h>
42 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
43 #define QEMU_NBD_OPT_CACHE 1
44 #define QEMU_NBD_OPT_AIO 2
45 #define QEMU_NBD_OPT_DISCARD 3
46 #define QEMU_NBD_OPT_DETECT_ZEROES 4
48 static NBDExport
*exp
;
51 static char *sockpath
;
52 static int persistent
= 0;
53 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
54 static int shared
= 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"
80 "Kernel NBD client support:\n"
81 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
82 " -d, --disconnect disconnect the specified device\n"
86 "Block device options:\n"
87 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
88 " -r, --read-only export read-only\n"
89 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
90 " file with backing_file=FILE, redirect the write to\n"
91 " the temporary one\n"
92 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
93 " load an internal snapshot inside FILE and export it\n"
94 " as an read-only device, SNAPSHOT_PARAM format is\n"
95 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
97 " -n, --nocache disable host cache\n"
98 " --cache=MODE set cache mode (none, writeback, ...)\n"
99 #ifdef CONFIG_LINUX_AIO
100 " --aio=MODE set AIO mode (native or threads)\n"
102 " --discard=MODE set discard mode (ignore, unmap)\n"
103 " --detect-zeroes=MODE set detect-zeroes mode (off, on, discard)\n"
105 "Report bugs to <qemu-devel@nongnu.org>\n"
106 , name
, NBD_DEFAULT_PORT
, "DEVICE");
109 static void version(const char *name
)
113 "Written by Anthony Liguori.\n"
115 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
116 "This is free software; see the source for copying conditions. There is NO\n"
117 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
121 struct partition_record
125 uint32_t start_cylinder
;
126 uint8_t start_sector
;
129 uint8_t end_cylinder
;
131 uint32_t start_sector_abs
;
132 uint32_t nb_sectors_abs
;
135 static void read_partition(uint8_t *p
, struct partition_record
*r
)
138 r
->start_head
= p
[1];
139 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
140 r
->start_sector
= p
[2] & 0x3f;
143 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
144 r
->end_sector
= p
[6] & 0x3f;
145 r
->start_sector_abs
= p
[8] | p
[9] << 8 | p
[10] << 16 | p
[11] << 24;
146 r
->nb_sectors_abs
= p
[12] | p
[13] << 8 | p
[14] << 16 | p
[15] << 24;
149 static int find_partition(BlockDriverState
*bs
, int partition
,
150 off_t
*offset
, off_t
*size
)
152 struct partition_record mbr
[4];
158 if ((ret
= bdrv_read(bs
, 0, data
, 1)) < 0) {
160 err(EXIT_FAILURE
, "error while reading");
163 if (data
[510] != 0x55 || data
[511] != 0xaa) {
167 for (i
= 0; i
< 4; i
++) {
168 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
170 if (!mbr
[i
].nb_sectors_abs
)
173 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
174 struct partition_record ext
[4];
178 if ((ret
= bdrv_read(bs
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
180 err(EXIT_FAILURE
, "error while reading");
183 for (j
= 0; j
< 4; j
++) {
184 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
185 if (!ext
[j
].nb_sectors_abs
)
188 if ((ext_partnum
+ j
+ 1) == partition
) {
189 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
190 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
195 } else if ((i
+ 1) == partition
) {
196 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
197 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
205 static void termsig_handler(int signum
)
211 static void combine_addr(char *buf
, size_t len
, const char* address
,
214 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
215 if (strstr(address
, ":")) {
216 snprintf(buf
, len
, "[%s]:%u", address
, port
);
218 snprintf(buf
, len
, "%s:%u", address
, port
);
222 static int tcp_socket_incoming(const char *address
, uint16_t port
)
224 char address_and_port
[128];
225 Error
*local_err
= NULL
;
227 combine_addr(address_and_port
, 128, address
, port
);
228 int fd
= inet_listen(address_and_port
, NULL
, 0, SOCK_STREAM
, 0, &local_err
);
230 if (local_err
!= NULL
) {
231 error_report("%s", error_get_pretty(local_err
));
232 error_free(local_err
);
237 static int unix_socket_incoming(const char *path
)
239 Error
*local_err
= NULL
;
240 int fd
= unix_listen(path
, NULL
, 0, &local_err
);
242 if (local_err
!= NULL
) {
243 error_report("%s", error_get_pretty(local_err
));
244 error_free(local_err
);
249 static int unix_socket_outgoing(const char *path
)
251 Error
*local_err
= NULL
;
252 int fd
= unix_connect(path
, &local_err
);
254 if (local_err
!= NULL
) {
255 error_report("%s", error_get_pretty(local_err
));
256 error_free(local_err
);
261 static void *show_parts(void *arg
)
266 /* linux just needs an open() to trigger
267 * the partition table update
268 * but remember to load the module with max_part != 0 :
269 * modprobe nbd max_part=63
271 nbd
= open(device
, O_RDWR
);
278 static void *nbd_client_thread(void *arg
)
286 pthread_t show_parts_thread
;
288 sock
= unix_socket_outgoing(sockpath
);
293 ret
= nbd_receive_negotiate(sock
, NULL
, &nbdflags
,
299 fd
= open(device
, O_RDWR
);
301 /* Linux-only, we can use %m in printf. */
302 fprintf(stderr
, "Failed to open %s: %m\n", device
);
306 ret
= nbd_init(fd
, sock
, nbdflags
, size
, blocksize
);
311 /* update partition table */
312 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
315 fprintf(stderr
, "NBD device %s is now connected to %s\n",
318 /* Close stderr so that the qemu-nbd process exits. */
319 dup2(STDOUT_FILENO
, STDERR_FILENO
);
322 ret
= nbd_client(fd
);
327 kill(getpid(), SIGTERM
);
328 return (void *) EXIT_SUCCESS
;
335 kill(getpid(), SIGTERM
);
336 return (void *) EXIT_FAILURE
;
339 static int nbd_can_accept(void *opaque
)
341 return nb_fds
< shared
;
344 static void nbd_export_closed(NBDExport
*exp
)
346 assert(state
== TERMINATING
);
350 static void nbd_client_closed(NBDClient
*client
)
353 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
357 nbd_client_put(client
);
360 static void nbd_accept(void *opaque
)
362 int server_fd
= (uintptr_t) opaque
;
363 struct sockaddr_in addr
;
364 socklen_t addr_len
= sizeof(addr
);
366 int fd
= accept(server_fd
, (struct sockaddr
*)&addr
, &addr_len
);
372 if (state
>= TERMINATE
) {
377 if (nbd_client_new(exp
, fd
, nbd_client_closed
)) {
385 int main(int argc
, char **argv
)
387 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";
394 int port
= NBD_DEFAULT_PORT
;
396 QemuOpts
*sn_opts
= NULL
;
397 const char *sn_id_or_name
= NULL
;
398 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
399 struct option lopt
[] = {
400 { "help", 0, NULL
, 'h' },
401 { "version", 0, NULL
, 'V' },
402 { "bind", 1, NULL
, 'b' },
403 { "port", 1, NULL
, 'p' },
404 { "socket", 1, NULL
, 'k' },
405 { "offset", 1, NULL
, 'o' },
406 { "read-only", 0, NULL
, 'r' },
407 { "partition", 1, NULL
, 'P' },
408 { "connect", 1, NULL
, 'c' },
409 { "disconnect", 0, NULL
, 'd' },
410 { "snapshot", 0, NULL
, 's' },
411 { "load-snapshot", 1, NULL
, 'l' },
412 { "nocache", 0, NULL
, 'n' },
413 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
414 #ifdef CONFIG_LINUX_AIO
415 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
417 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
418 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
419 { "shared", 1, NULL
, 'e' },
420 { "format", 1, NULL
, 'f' },
421 { "persistent", 0, NULL
, 't' },
422 { "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
;
443 /* The client thread uses SIGTERM to interrupt the server. A signal
444 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
446 struct sigaction sa_sigterm
;
447 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
448 sa_sigterm
.sa_handler
= termsig_handler
;
449 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
450 qemu_init_exec_dir(argv
[0]);
452 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
455 flags
|= BDRV_O_SNAPSHOT
;
458 optarg
= (char *) "none";
460 case QEMU_NBD_OPT_CACHE
:
462 errx(EXIT_FAILURE
, "-n and --cache can only be specified once");
465 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
466 errx(EXIT_FAILURE
, "Invalid cache mode `%s'", optarg
);
469 #ifdef CONFIG_LINUX_AIO
470 case QEMU_NBD_OPT_AIO
:
472 errx(EXIT_FAILURE
, "--aio can only be specified once");
475 if (!strcmp(optarg
, "native")) {
476 flags
|= BDRV_O_NATIVE_AIO
;
477 } else if (!strcmp(optarg
, "threads")) {
478 /* this is the default */
480 errx(EXIT_FAILURE
, "invalid aio mode `%s'", optarg
);
484 case QEMU_NBD_OPT_DISCARD
:
486 errx(EXIT_FAILURE
, "--discard can only be specified once");
489 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
490 errx(EXIT_FAILURE
, "Invalid discard mode `%s'", optarg
);
493 case QEMU_NBD_OPT_DETECT_ZEROES
:
495 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
497 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX
,
498 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
501 errx(EXIT_FAILURE
, "Failed to parse detect_zeroes mode: %s",
502 error_get_pretty(local_err
));
504 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
505 !(flags
& BDRV_O_UNMAP
)) {
506 errx(EXIT_FAILURE
, "setting detect-zeroes to unmap is not allowed "
507 "without setting discard operation to unmap");
514 li
= strtol(optarg
, &end
, 0);
516 errx(EXIT_FAILURE
, "Invalid port `%s'", optarg
);
518 if (li
< 1 || li
> 65535) {
519 errx(EXIT_FAILURE
, "Port out of range `%s'", optarg
);
524 dev_offset
= strtoll (optarg
, &end
, 0);
526 errx(EXIT_FAILURE
, "Invalid offset `%s'", optarg
);
528 if (dev_offset
< 0) {
529 errx(EXIT_FAILURE
, "Offset must be positive `%s'", optarg
);
533 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
534 sn_opts
= qemu_opts_parse(&internal_snapshot_opts
, optarg
, 0);
536 errx(EXIT_FAILURE
, "Failed in parsing snapshot param `%s'",
540 sn_id_or_name
= optarg
;
544 nbdflags
|= NBD_FLAG_READ_ONLY
;
545 flags
&= ~BDRV_O_RDWR
;
548 partition
= strtol(optarg
, &end
, 0);
550 errx(EXIT_FAILURE
, "Invalid partition `%s'", optarg
);
552 if (partition
< 1 || partition
> 8) {
553 errx(EXIT_FAILURE
, "Invalid partition %d", partition
);
558 if (sockpath
[0] != '/') {
559 errx(EXIT_FAILURE
, "socket path must be absolute\n");
569 shared
= strtol(optarg
, &end
, 0);
571 errx(EXIT_FAILURE
, "Invalid shared device number '%s'", optarg
);
574 errx(EXIT_FAILURE
, "Shared device number must be greater than 0\n");
595 errx(EXIT_FAILURE
, "Try `%s --help' for more information.",
600 if ((argc
- optind
) != 1) {
601 errx(EXIT_FAILURE
, "Invalid number of argument.\n"
602 "Try `%s --help' for more information.",
607 fd
= open(argv
[optind
], O_RDWR
);
609 err(EXIT_FAILURE
, "Cannot open %s", argv
[optind
]);
615 printf("%s disconnected\n", argv
[optind
]);
620 if (device
&& !verbose
) {
625 if (qemu_pipe(stderr_fd
) < 0) {
626 err(EXIT_FAILURE
, "Error setting up communication pipe");
629 /* Now daemonize, but keep a communication channel open to
630 * print errors and exit with the proper status code.
635 ret
= qemu_daemon(1, 0);
637 /* Temporarily redirect stderr to the parent's pipe... */
638 dup2(stderr_fd
[1], STDERR_FILENO
);
640 err(EXIT_FAILURE
, "Failed to daemonize");
643 /* ... close the descriptor we inherited and go on. */
649 /* In the parent. Print error messages from the child until
650 * it closes the pipe.
653 buf
= g_malloc(1024);
654 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
656 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
662 err(EXIT_FAILURE
, "Cannot read from daemon");
665 /* Usually the daemon should not print any message.
666 * Exit with zero status in that case.
672 if (device
!= NULL
&& sockpath
== NULL
) {
673 sockpath
= g_malloc(128);
674 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
677 qemu_init_main_loop();
679 atexit(bdrv_close_all
);
682 drv
= bdrv_find_format(fmt
);
684 errx(EXIT_FAILURE
, "Unknown file format '%s'", fmt
);
690 bs
= bdrv_new("hda", &error_abort
);
692 srcpath
= argv
[optind
];
693 ret
= bdrv_open(&bs
, srcpath
, NULL
, NULL
, flags
, drv
, &local_err
);
696 err(EXIT_FAILURE
, "Failed to bdrv_open '%s': %s", argv
[optind
],
697 error_get_pretty(local_err
));
701 ret
= bdrv_snapshot_load_tmp(bs
,
702 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
703 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
705 } else if (sn_id_or_name
) {
706 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
712 "Failed to load snapshot: %s",
713 error_get_pretty(local_err
));
716 bs
->detect_zeroes
= detect_zeroes
;
717 fd_size
= bdrv_getlength(bs
);
719 if (partition
!= -1) {
720 ret
= find_partition(bs
, partition
, &dev_offset
, &fd_size
);
723 err(EXIT_FAILURE
, "Could not find partition %d", partition
);
727 exp
= nbd_export_new(bs
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
);
730 fd
= unix_socket_incoming(sockpath
);
732 fd
= tcp_socket_incoming(bindto
, port
);
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
));
752 qemu_set_fd_handler2(fd
, nbd_can_accept
, nbd_accept
, NULL
,
753 (void *)(uintptr_t)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
);
778 qemu_opts_del(sn_opts
);
783 pthread_join(client_thread
, &ret
);