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/sockets.h"
26 #include "qemu/error-report.h"
27 #include "block/snapshot.h"
28 #include "qapi/util.h"
29 #include "qapi/qmp/qstring.h"
32 #include <sys/socket.h>
33 #include <netinet/in.h>
34 #include <netinet/tcp.h>
35 #include <arpa/inet.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
45 static NBDExport
*exp
;
48 static SocketAddress
*saddr
;
49 static int persistent
= 0;
50 static enum { RUNNING
, TERMINATE
, TERMINATING
, TERMINATED
} state
;
51 static int shared
= 1;
55 static void usage(const char *name
)
58 "Usage: %s [OPTIONS] FILE\n"
59 "QEMU Disk Network Block Device Server\n"
61 " -h, --help display this help and exit\n"
62 " -V, --version output version information and exit\n"
64 "Connection properties:\n"
65 " -p, --port=PORT port to listen on (default `%d')\n"
66 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
67 " -k, --socket=PATH path to the unix socket\n"
68 " (default '"SOCKET_PATH
"')\n"
69 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
70 " -t, --persistent don't exit on the last connection\n"
71 " -v, --verbose display extra debugging information\n"
73 "Exposing part of the image:\n"
74 " -o, --offset=OFFSET offset into the image\n"
75 " -P, --partition=NUM only expose partition NUM\n"
78 "Kernel NBD client support:\n"
79 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
80 " -d, --disconnect disconnect the specified device\n"
84 "Block device options:\n"
85 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
86 " -r, --read-only export read-only\n"
87 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
88 " file with backing_file=FILE, redirect the write to\n"
89 " the temporary one\n"
90 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
91 " load an internal snapshot inside FILE and export it\n"
92 " as an read-only device, SNAPSHOT_PARAM format is\n"
93 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
95 " -n, --nocache disable host cache\n"
96 " --cache=MODE set cache mode (none, writeback, ...)\n"
97 " --aio=MODE set AIO mode (native or threads)\n"
98 " --discard=MODE set discard mode (ignore, unmap)\n"
99 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
101 "Report bugs to <qemu-devel@nongnu.org>\n"
102 , name
, NBD_DEFAULT_PORT
, "DEVICE");
105 static void version(const char *name
)
109 "Written by Anthony Liguori.\n"
111 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
112 "This is free software; see the source for copying conditions. There is NO\n"
113 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
117 struct partition_record
121 uint32_t start_cylinder
;
122 uint8_t start_sector
;
125 uint8_t end_cylinder
;
127 uint32_t start_sector_abs
;
128 uint32_t nb_sectors_abs
;
131 static void read_partition(uint8_t *p
, struct partition_record
*r
)
134 r
->start_head
= p
[1];
135 r
->start_cylinder
= p
[3] | ((p
[2] << 2) & 0x0300);
136 r
->start_sector
= p
[2] & 0x3f;
139 r
->end_cylinder
= p
[7] | ((p
[6] << 2) & 0x300);
140 r
->end_sector
= p
[6] & 0x3f;
142 r
->start_sector_abs
= le32_to_cpup((uint32_t *)(p
+ 8));
143 r
->nb_sectors_abs
= le32_to_cpup((uint32_t *)(p
+ 12));
146 static int find_partition(BlockBackend
*blk
, int partition
,
147 off_t
*offset
, off_t
*size
)
149 struct partition_record mbr
[4];
155 if ((ret
= blk_read(blk
, 0, data
, 1)) < 0) {
156 error_report("error while reading: %s", strerror(-ret
));
160 if (data
[510] != 0x55 || data
[511] != 0xaa) {
164 for (i
= 0; i
< 4; i
++) {
165 read_partition(&data
[446 + 16 * i
], &mbr
[i
]);
167 if (!mbr
[i
].system
|| !mbr
[i
].nb_sectors_abs
) {
171 if (mbr
[i
].system
== 0xF || mbr
[i
].system
== 0x5) {
172 struct partition_record ext
[4];
176 if ((ret
= blk_read(blk
, mbr
[i
].start_sector_abs
, data1
, 1)) < 0) {
177 error_report("error while reading: %s", strerror(-ret
));
181 for (j
= 0; j
< 4; j
++) {
182 read_partition(&data1
[446 + 16 * j
], &ext
[j
]);
183 if (!ext
[j
].system
|| !ext
[j
].nb_sectors_abs
) {
187 if ((ext_partnum
+ j
+ 1) == partition
) {
188 *offset
= (uint64_t)ext
[j
].start_sector_abs
<< 9;
189 *size
= (uint64_t)ext
[j
].nb_sectors_abs
<< 9;
194 } else if ((i
+ 1) == partition
) {
195 *offset
= (uint64_t)mbr
[i
].start_sector_abs
<< 9;
196 *size
= (uint64_t)mbr
[i
].nb_sectors_abs
<< 9;
204 static void termsig_handler(int signum
)
211 static void *show_parts(void *arg
)
216 /* linux just needs an open() to trigger
217 * the partition table update
218 * but remember to load the module with max_part != 0 :
219 * modprobe nbd max_part=63
221 nbd
= open(device
, O_RDWR
);
228 static void *nbd_client_thread(void *arg
)
235 pthread_t show_parts_thread
;
236 Error
*local_error
= NULL
;
239 sock
= socket_connect(saddr
, &local_error
, NULL
, NULL
);
241 error_report_err(local_error
);
245 ret
= nbd_receive_negotiate(sock
, NULL
, &nbdflags
,
246 &size
, &local_error
);
249 error_report_err(local_error
);
254 fd
= open(device
, O_RDWR
);
256 /* Linux-only, we can use %m in printf. */
257 error_report("Failed to open %s: %m", device
);
261 ret
= nbd_init(fd
, sock
, nbdflags
, size
);
266 /* update partition table */
267 pthread_create(&show_parts_thread
, NULL
, show_parts
, device
);
270 fprintf(stderr
, "NBD device %s is now connected to %s\n",
273 /* Close stderr so that the qemu-nbd process exits. */
274 dup2(STDOUT_FILENO
, STDERR_FILENO
);
277 ret
= nbd_client(fd
);
282 kill(getpid(), SIGTERM
);
283 return (void *) EXIT_SUCCESS
;
290 kill(getpid(), SIGTERM
);
291 return (void *) EXIT_FAILURE
;
294 static int nbd_can_accept(void)
296 return nb_fds
< shared
;
299 static void nbd_export_closed(NBDExport
*exp
)
301 assert(state
== TERMINATING
);
305 static void nbd_update_server_fd_handler(int fd
);
307 static void nbd_client_closed(NBDClient
*client
)
310 if (nb_fds
== 0 && !persistent
&& state
== RUNNING
) {
313 nbd_update_server_fd_handler(server_fd
);
314 nbd_client_put(client
);
317 static void nbd_accept(void *opaque
)
319 struct sockaddr_in addr
;
320 socklen_t addr_len
= sizeof(addr
);
322 int fd
= accept(server_fd
, (struct sockaddr
*)&addr
, &addr_len
);
328 if (state
>= TERMINATE
) {
334 nbd_update_server_fd_handler(server_fd
);
335 nbd_client_new(exp
, fd
, nbd_client_closed
);
338 static void nbd_update_server_fd_handler(int fd
)
340 if (nbd_can_accept()) {
341 qemu_set_fd_handler(fd
, nbd_accept
, NULL
, (void *)(uintptr_t)fd
);
343 qemu_set_fd_handler(fd
, NULL
, NULL
, NULL
);
348 static SocketAddress
*nbd_build_socket_address(const char *sockpath
,
352 SocketAddress
*saddr
;
354 saddr
= g_new0(SocketAddress
, 1);
356 saddr
->type
= SOCKET_ADDRESS_KIND_UNIX
;
357 saddr
->u
.q_unix
= g_new0(UnixSocketAddress
, 1);
358 saddr
->u
.q_unix
->path
= g_strdup(sockpath
);
360 saddr
->type
= SOCKET_ADDRESS_KIND_INET
;
361 saddr
->u
.inet
= g_new0(InetSocketAddress
, 1);
362 saddr
->u
.inet
->host
= g_strdup(bindto
);
364 saddr
->u
.inet
->port
= g_strdup(port
);
366 saddr
->u
.inet
->port
= g_strdup_printf("%d", NBD_DEFAULT_PORT
);
374 int main(int argc
, char **argv
)
377 BlockDriverState
*bs
;
378 off_t dev_offset
= 0;
379 uint32_t nbdflags
= 0;
380 bool disconnect
= false;
381 const char *bindto
= "0.0.0.0";
382 const char *port
= NULL
;
383 char *sockpath
= NULL
;
386 QemuOpts
*sn_opts
= NULL
;
387 const char *sn_id_or_name
= NULL
;
388 const char *sopt
= "hVb:o:p:rsnP:c:dvk:e:f:tl:";
389 struct option lopt
[] = {
390 { "help", 0, NULL
, 'h' },
391 { "version", 0, NULL
, 'V' },
392 { "bind", 1, NULL
, 'b' },
393 { "port", 1, NULL
, 'p' },
394 { "socket", 1, NULL
, 'k' },
395 { "offset", 1, NULL
, 'o' },
396 { "read-only", 0, NULL
, 'r' },
397 { "partition", 1, NULL
, 'P' },
398 { "connect", 1, NULL
, 'c' },
399 { "disconnect", 0, NULL
, 'd' },
400 { "snapshot", 0, NULL
, 's' },
401 { "load-snapshot", 1, NULL
, 'l' },
402 { "nocache", 0, NULL
, 'n' },
403 { "cache", 1, NULL
, QEMU_NBD_OPT_CACHE
},
404 { "aio", 1, NULL
, QEMU_NBD_OPT_AIO
},
405 { "discard", 1, NULL
, QEMU_NBD_OPT_DISCARD
},
406 { "detect-zeroes", 1, NULL
, QEMU_NBD_OPT_DETECT_ZEROES
},
407 { "shared", 1, NULL
, 'e' },
408 { "format", 1, NULL
, 'f' },
409 { "persistent", 0, NULL
, 't' },
410 { "verbose", 0, NULL
, 'v' },
416 int flags
= BDRV_O_RDWR
;
420 bool seen_cache
= false;
421 bool seen_discard
= false;
422 bool seen_aio
= false;
423 pthread_t client_thread
;
424 const char *fmt
= NULL
;
425 Error
*local_err
= NULL
;
426 BlockdevDetectZeroesOptions detect_zeroes
= BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
427 QDict
*options
= NULL
;
429 /* The client thread uses SIGTERM to interrupt the server. A signal
430 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
432 struct sigaction sa_sigterm
;
433 memset(&sa_sigterm
, 0, sizeof(sa_sigterm
));
434 sa_sigterm
.sa_handler
= termsig_handler
;
435 sigaction(SIGTERM
, &sa_sigterm
, NULL
);
436 qemu_init_exec_dir(argv
[0]);
438 while ((ch
= getopt_long(argc
, argv
, sopt
, lopt
, &opt_ind
)) != -1) {
441 flags
|= BDRV_O_SNAPSHOT
;
444 optarg
= (char *) "none";
446 case QEMU_NBD_OPT_CACHE
:
448 error_report("-n and --cache can only be specified once");
452 if (bdrv_parse_cache_flags(optarg
, &flags
) == -1) {
453 error_report("Invalid cache mode `%s'", optarg
);
457 case QEMU_NBD_OPT_AIO
:
459 error_report("--aio can only be specified once");
463 if (!strcmp(optarg
, "native")) {
464 flags
|= BDRV_O_NATIVE_AIO
;
465 } else if (!strcmp(optarg
, "threads")) {
466 /* this is the default */
468 error_report("invalid aio mode `%s'", optarg
);
472 case QEMU_NBD_OPT_DISCARD
:
474 error_report("--discard can only be specified once");
478 if (bdrv_parse_discard_flags(optarg
, &flags
) == -1) {
479 error_report("Invalid discard mode `%s'", optarg
);
483 case QEMU_NBD_OPT_DETECT_ZEROES
:
485 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup
,
487 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX
,
488 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
491 error_reportf_err(local_err
,
492 "Failed to parse detect_zeroes mode: ");
495 if (detect_zeroes
== BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP
&&
496 !(flags
& BDRV_O_UNMAP
)) {
497 error_report("setting detect-zeroes to unmap is not allowed "
498 "without setting discard operation to unmap");
509 dev_offset
= strtoll (optarg
, &end
, 0);
511 error_report("Invalid offset `%s'", optarg
);
514 if (dev_offset
< 0) {
515 error_report("Offset must be positive `%s'", optarg
);
520 if (strstart(optarg
, SNAPSHOT_OPT_BASE
, NULL
)) {
521 sn_opts
= qemu_opts_parse_noisily(&internal_snapshot_opts
,
524 error_report("Failed in parsing snapshot param `%s'",
529 sn_id_or_name
= optarg
;
533 nbdflags
|= NBD_FLAG_READ_ONLY
;
534 flags
&= ~BDRV_O_RDWR
;
537 partition
= strtol(optarg
, &end
, 0);
539 error_report("Invalid partition `%s'", optarg
);
542 if (partition
< 1 || partition
> 8) {
543 error_report("Invalid partition %d", partition
);
549 if (sockpath
[0] != '/') {
550 error_report("socket path must be absolute");
561 shared
= strtol(optarg
, &end
, 0);
563 error_report("Invalid shared device number '%s'", optarg
);
567 error_report("Shared device number must be greater than 0");
589 error_report("Try `%s --help' for more information.", argv
[0]);
594 if ((argc
- optind
) != 1) {
595 error_report("Invalid number of arguments");
596 error_printf("Try `%s --help' for more information.\n", argv
[0]);
601 fd
= open(argv
[optind
], O_RDWR
);
603 error_report("Cannot open %s: %s", argv
[optind
],
611 printf("%s disconnected\n", argv
[optind
]);
616 if (device
&& !verbose
) {
621 if (qemu_pipe(stderr_fd
) < 0) {
622 error_report("Error setting up communication pipe: %s",
627 /* Now daemonize, but keep a communication channel open to
628 * print errors and exit with the proper status code.
632 error_report("Failed to fork: %s", strerror(errno
));
634 } else if (pid
== 0) {
636 ret
= qemu_daemon(1, 0);
638 /* Temporarily redirect stderr to the parent's pipe... */
639 dup2(stderr_fd
[1], STDERR_FILENO
);
641 error_report("Failed to daemonize: %s", strerror(errno
));
645 /* ... close the descriptor we inherited and go on. */
651 /* In the parent. Print error messages from the child until
652 * it closes the pipe.
655 buf
= g_malloc(1024);
656 while ((ret
= read(stderr_fd
[0], buf
, 1024)) > 0) {
658 ret
= qemu_write_full(STDERR_FILENO
, buf
, ret
);
664 error_report("Cannot read from daemon: %s",
669 /* Usually the daemon should not print any message.
670 * Exit with zero status in that case.
676 if (device
!= NULL
&& sockpath
== NULL
) {
677 sockpath
= g_malloc(128);
678 snprintf(sockpath
, 128, SOCKET_PATH
, basename(device
));
681 saddr
= nbd_build_socket_address(sockpath
, bindto
, port
);
683 if (qemu_init_main_loop(&local_err
)) {
684 error_report_err(local_err
);
688 atexit(bdrv_close_all
);
691 options
= qdict_new();
692 qdict_put(options
, "driver", qstring_from_str(fmt
));
695 srcpath
= argv
[optind
];
696 blk
= blk_new_open("hda", srcpath
, NULL
, options
, flags
, &local_err
);
698 error_reportf_err(local_err
, "Failed to blk_new_open '%s': ",
705 ret
= bdrv_snapshot_load_tmp(bs
,
706 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_ID
),
707 qemu_opt_get(sn_opts
, SNAPSHOT_OPT_NAME
),
709 } else if (sn_id_or_name
) {
710 ret
= bdrv_snapshot_load_tmp_by_id_or_name(bs
, sn_id_or_name
,
714 error_reportf_err(local_err
, "Failed to load snapshot: ");
718 bs
->detect_zeroes
= detect_zeroes
;
719 fd_size
= blk_getlength(blk
);
721 error_report("Failed to determine the image length: %s",
726 if (partition
!= -1) {
727 ret
= find_partition(blk
, partition
, &dev_offset
, &fd_size
);
729 error_report("Could not find partition %d: %s", partition
,
735 exp
= nbd_export_new(blk
, dev_offset
, fd_size
, nbdflags
, nbd_export_closed
,
738 error_report_err(local_err
);
742 fd
= socket_listen(saddr
, &local_err
);
744 error_report_err(local_err
);
751 ret
= pthread_create(&client_thread
, NULL
, nbd_client_thread
, device
);
753 error_report("Failed to create client thread: %s", strerror(ret
));
757 /* Shut up GCC warnings. */
758 memset(&client_thread
, 0, sizeof(client_thread
));
762 nbd_update_server_fd_handler(fd
);
764 /* now when the initialization is (almost) complete, chdir("/")
765 * to free any busy filesystems */
766 if (chdir("/") < 0) {
767 error_report("Could not chdir to root directory: %s",
774 main_loop_wait(false);
775 if (state
== TERMINATE
) {
777 nbd_export_close(exp
);
781 } while (state
!= TERMINATED
);
788 qemu_opts_del(sn_opts
);
792 pthread_join(client_thread
, &ret
);