hw/mips: In KVM mode, inject IRQ2 (I/O) interrupts via ioctls
[qemu-kvm.git] / qemu-nbd.c
blobba6043680a31e5408a73bfb435c570d7e0cac0bf
1 /*
2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
4 * Network Block Device
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/nbd.h"
22 #include "qemu/main-loop.h"
23 #include "qemu/sockets.h"
24 #include "qemu/error-report.h"
25 #include "block/snapshot.h"
27 #include <stdarg.h>
28 #include <stdio.h>
29 #include <getopt.h>
30 #include <err.h>
31 #include <sys/types.h>
32 #include <sys/socket.h>
33 #include <netinet/in.h>
34 #include <netinet/tcp.h>
35 #include <arpa/inet.h>
36 #include <signal.h>
37 #include <libgen.h>
38 #include <pthread.h>
40 #define SOCKET_PATH "/var/lock/qemu-nbd-%s"
41 #define QEMU_NBD_OPT_CACHE 1
42 #define QEMU_NBD_OPT_AIO 2
43 #define QEMU_NBD_OPT_DISCARD 3
45 static NBDExport *exp;
46 static int verbose;
47 static char *srcpath;
48 static char *sockpath;
49 static int persistent = 0;
50 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state;
51 static int shared = 1;
52 static int nb_fds;
54 static void usage(const char *name)
56 (printf) (
57 "Usage: %s [OPTIONS] FILE\n"
58 "QEMU Disk Network Block Device Server\n"
59 "\n"
60 " -h, --help display this help and exit\n"
61 " -V, --version output version information and exit\n"
62 "\n"
63 "Connection properties:\n"
64 " -p, --port=PORT port to listen on (default `%d')\n"
65 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n"
66 " -k, --socket=PATH path to the unix socket\n"
67 " (default '"SOCKET_PATH"')\n"
68 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n"
69 " -t, --persistent don't exit on the last connection\n"
70 " -v, --verbose display extra debugging information\n"
71 "\n"
72 "Exposing part of the image:\n"
73 " -o, --offset=OFFSET offset into the image\n"
74 " -P, --partition=NUM only expose partition NUM\n"
75 "\n"
76 #ifdef __linux__
77 "Kernel NBD client support:\n"
78 " -c, --connect=DEV connect FILE to the local NBD device DEV\n"
79 " -d, --disconnect disconnect the specified device\n"
80 "\n"
81 #endif
82 "\n"
83 "Block device options:\n"
84 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n"
85 " -r, --read-only export read-only\n"
86 " -s, --snapshot use FILE as an external snapshot, create a temporary\n"
87 " file with backing_file=FILE, redirect the write to\n"
88 " the temporary one\n"
89 " -l, --load-snapshot=SNAPSHOT_PARAM\n"
90 " load an internal snapshot inside FILE and export it\n"
91 " as an read-only device, SNAPSHOT_PARAM format is\n"
92 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
93 " '[ID_OR_NAME]'\n"
94 " -n, --nocache disable host cache\n"
95 " --cache=MODE set cache mode (none, writeback, ...)\n"
96 #ifdef CONFIG_LINUX_AIO
97 " --aio=MODE set AIO mode (native or threads)\n"
98 #endif
99 "\n"
100 "Report bugs to <qemu-devel@nongnu.org>\n"
101 , name, NBD_DEFAULT_PORT, "DEVICE");
104 static void version(const char *name)
106 printf(
107 "%s version 0.0.1\n"
108 "Written by Anthony Liguori.\n"
109 "\n"
110 "Copyright (C) 2006 Anthony Liguori <anthony@codemonkey.ws>.\n"
111 "This is free software; see the source for copying conditions. There is NO\n"
112 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
113 , name);
116 struct partition_record
118 uint8_t bootable;
119 uint8_t start_head;
120 uint32_t start_cylinder;
121 uint8_t start_sector;
122 uint8_t system;
123 uint8_t end_head;
124 uint8_t end_cylinder;
125 uint8_t end_sector;
126 uint32_t start_sector_abs;
127 uint32_t nb_sectors_abs;
130 static void read_partition(uint8_t *p, struct partition_record *r)
132 r->bootable = p[0];
133 r->start_head = p[1];
134 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300);
135 r->start_sector = p[2] & 0x3f;
136 r->system = p[4];
137 r->end_head = p[5];
138 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300);
139 r->end_sector = p[6] & 0x3f;
140 r->start_sector_abs = p[8] | p[9] << 8 | p[10] << 16 | p[11] << 24;
141 r->nb_sectors_abs = p[12] | p[13] << 8 | p[14] << 16 | p[15] << 24;
144 static int find_partition(BlockDriverState *bs, int partition,
145 off_t *offset, off_t *size)
147 struct partition_record mbr[4];
148 uint8_t data[512];
149 int i;
150 int ext_partnum = 4;
151 int ret;
153 if ((ret = bdrv_read(bs, 0, data, 1)) < 0) {
154 errno = -ret;
155 err(EXIT_FAILURE, "error while reading");
158 if (data[510] != 0x55 || data[511] != 0xaa) {
159 return -EINVAL;
162 for (i = 0; i < 4; i++) {
163 read_partition(&data[446 + 16 * i], &mbr[i]);
165 if (!mbr[i].nb_sectors_abs)
166 continue;
168 if (mbr[i].system == 0xF || mbr[i].system == 0x5) {
169 struct partition_record ext[4];
170 uint8_t data1[512];
171 int j;
173 if ((ret = bdrv_read(bs, mbr[i].start_sector_abs, data1, 1)) < 0) {
174 errno = -ret;
175 err(EXIT_FAILURE, "error while reading");
178 for (j = 0; j < 4; j++) {
179 read_partition(&data1[446 + 16 * j], &ext[j]);
180 if (!ext[j].nb_sectors_abs)
181 continue;
183 if ((ext_partnum + j + 1) == partition) {
184 *offset = (uint64_t)ext[j].start_sector_abs << 9;
185 *size = (uint64_t)ext[j].nb_sectors_abs << 9;
186 return 0;
189 ext_partnum += 4;
190 } else if ((i + 1) == partition) {
191 *offset = (uint64_t)mbr[i].start_sector_abs << 9;
192 *size = (uint64_t)mbr[i].nb_sectors_abs << 9;
193 return 0;
197 return -ENOENT;
200 static void termsig_handler(int signum)
202 state = TERMINATE;
203 qemu_notify_event();
206 static void combine_addr(char *buf, size_t len, const char* address,
207 uint16_t port)
209 /* If the address-part contains a colon, it's an IPv6 IP so needs [] */
210 if (strstr(address, ":")) {
211 snprintf(buf, len, "[%s]:%u", address, port);
212 } else {
213 snprintf(buf, len, "%s:%u", address, port);
217 static int tcp_socket_incoming(const char *address, uint16_t port)
219 char address_and_port[128];
220 Error *local_err = NULL;
222 combine_addr(address_and_port, 128, address, port);
223 int fd = inet_listen(address_and_port, NULL, 0, SOCK_STREAM, 0, &local_err);
225 if (local_err != NULL) {
226 error_report("%s", error_get_pretty(local_err));
227 error_free(local_err);
229 return fd;
232 static int unix_socket_incoming(const char *path)
234 Error *local_err = NULL;
235 int fd = unix_listen(path, NULL, 0, &local_err);
237 if (local_err != NULL) {
238 error_report("%s", error_get_pretty(local_err));
239 error_free(local_err);
241 return fd;
244 static int unix_socket_outgoing(const char *path)
246 Error *local_err = NULL;
247 int fd = unix_connect(path, &local_err);
249 if (local_err != NULL) {
250 error_report("%s", error_get_pretty(local_err));
251 error_free(local_err);
253 return fd;
256 static void *show_parts(void *arg)
258 char *device = arg;
259 int nbd;
261 /* linux just needs an open() to trigger
262 * the partition table update
263 * but remember to load the module with max_part != 0 :
264 * modprobe nbd max_part=63
266 nbd = open(device, O_RDWR);
267 if (nbd >= 0) {
268 close(nbd);
270 return NULL;
273 static void *nbd_client_thread(void *arg)
275 char *device = arg;
276 off_t size;
277 size_t blocksize;
278 uint32_t nbdflags;
279 int fd, sock;
280 int ret;
281 pthread_t show_parts_thread;
283 sock = unix_socket_outgoing(sockpath);
284 if (sock < 0) {
285 goto out;
288 ret = nbd_receive_negotiate(sock, NULL, &nbdflags,
289 &size, &blocksize);
290 if (ret < 0) {
291 goto out_socket;
294 fd = open(device, O_RDWR);
295 if (fd < 0) {
296 /* Linux-only, we can use %m in printf. */
297 fprintf(stderr, "Failed to open %s: %m\n", device);
298 goto out_socket;
301 ret = nbd_init(fd, sock, nbdflags, size, blocksize);
302 if (ret < 0) {
303 goto out_fd;
306 /* update partition table */
307 pthread_create(&show_parts_thread, NULL, show_parts, device);
309 if (verbose) {
310 fprintf(stderr, "NBD device %s is now connected to %s\n",
311 device, srcpath);
312 } else {
313 /* Close stderr so that the qemu-nbd process exits. */
314 dup2(STDOUT_FILENO, STDERR_FILENO);
317 ret = nbd_client(fd);
318 if (ret) {
319 goto out_fd;
321 close(fd);
322 kill(getpid(), SIGTERM);
323 return (void *) EXIT_SUCCESS;
325 out_fd:
326 close(fd);
327 out_socket:
328 closesocket(sock);
329 out:
330 kill(getpid(), SIGTERM);
331 return (void *) EXIT_FAILURE;
334 static int nbd_can_accept(void *opaque)
336 return nb_fds < shared;
339 static void nbd_export_closed(NBDExport *exp)
341 assert(state == TERMINATING);
342 state = TERMINATED;
345 static void nbd_client_closed(NBDClient *client)
347 nb_fds--;
348 if (nb_fds == 0 && !persistent && state == RUNNING) {
349 state = TERMINATE;
351 qemu_notify_event();
352 nbd_client_put(client);
355 static void nbd_accept(void *opaque)
357 int server_fd = (uintptr_t) opaque;
358 struct sockaddr_in addr;
359 socklen_t addr_len = sizeof(addr);
361 int fd = accept(server_fd, (struct sockaddr *)&addr, &addr_len);
362 if (fd < 0) {
363 perror("accept");
364 return;
367 if (state >= TERMINATE) {
368 close(fd);
369 return;
372 if (nbd_client_new(exp, fd, nbd_client_closed)) {
373 nb_fds++;
374 } else {
375 close(fd);
379 int main(int argc, char **argv)
381 BlockDriverState *bs;
382 BlockDriver *drv;
383 off_t dev_offset = 0;
384 uint32_t nbdflags = 0;
385 bool disconnect = false;
386 const char *bindto = "0.0.0.0";
387 char *device = NULL;
388 int port = NBD_DEFAULT_PORT;
389 off_t fd_size;
390 QemuOpts *sn_opts = NULL;
391 const char *sn_id_or_name = NULL;
392 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:";
393 struct option lopt[] = {
394 { "help", 0, NULL, 'h' },
395 { "version", 0, NULL, 'V' },
396 { "bind", 1, NULL, 'b' },
397 { "port", 1, NULL, 'p' },
398 { "socket", 1, NULL, 'k' },
399 { "offset", 1, NULL, 'o' },
400 { "read-only", 0, NULL, 'r' },
401 { "partition", 1, NULL, 'P' },
402 { "connect", 1, NULL, 'c' },
403 { "disconnect", 0, NULL, 'd' },
404 { "snapshot", 0, NULL, 's' },
405 { "load-snapshot", 1, NULL, 'l' },
406 { "nocache", 0, NULL, 'n' },
407 { "cache", 1, NULL, QEMU_NBD_OPT_CACHE },
408 #ifdef CONFIG_LINUX_AIO
409 { "aio", 1, NULL, QEMU_NBD_OPT_AIO },
410 #endif
411 { "discard", 1, NULL, QEMU_NBD_OPT_DISCARD },
412 { "shared", 1, NULL, 'e' },
413 { "format", 1, NULL, 'f' },
414 { "persistent", 0, NULL, 't' },
415 { "verbose", 0, NULL, 'v' },
416 { NULL, 0, NULL, 0 }
418 int ch;
419 int opt_ind = 0;
420 int li;
421 char *end;
422 int flags = BDRV_O_RDWR;
423 int partition = -1;
424 int ret;
425 int fd;
426 bool seen_cache = false;
427 bool seen_discard = false;
428 #ifdef CONFIG_LINUX_AIO
429 bool seen_aio = false;
430 #endif
431 pthread_t client_thread;
432 const char *fmt = NULL;
433 Error *local_err = NULL;
435 /* The client thread uses SIGTERM to interrupt the server. A signal
436 * handler ensures that "qemu-nbd -v -c" exits with a nice status code.
438 struct sigaction sa_sigterm;
439 memset(&sa_sigterm, 0, sizeof(sa_sigterm));
440 sa_sigterm.sa_handler = termsig_handler;
441 sigaction(SIGTERM, &sa_sigterm, NULL);
442 qemu_init_exec_dir(argv[0]);
444 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
445 switch (ch) {
446 case 's':
447 flags |= BDRV_O_SNAPSHOT;
448 break;
449 case 'n':
450 optarg = (char *) "none";
451 /* fallthrough */
452 case QEMU_NBD_OPT_CACHE:
453 if (seen_cache) {
454 errx(EXIT_FAILURE, "-n and --cache can only be specified once");
456 seen_cache = true;
457 if (bdrv_parse_cache_flags(optarg, &flags) == -1) {
458 errx(EXIT_FAILURE, "Invalid cache mode `%s'", optarg);
460 break;
461 #ifdef CONFIG_LINUX_AIO
462 case QEMU_NBD_OPT_AIO:
463 if (seen_aio) {
464 errx(EXIT_FAILURE, "--aio can only be specified once");
466 seen_aio = true;
467 if (!strcmp(optarg, "native")) {
468 flags |= BDRV_O_NATIVE_AIO;
469 } else if (!strcmp(optarg, "threads")) {
470 /* this is the default */
471 } else {
472 errx(EXIT_FAILURE, "invalid aio mode `%s'", optarg);
474 break;
475 #endif
476 case QEMU_NBD_OPT_DISCARD:
477 if (seen_discard) {
478 errx(EXIT_FAILURE, "--discard can only be specified once");
480 seen_discard = true;
481 if (bdrv_parse_discard_flags(optarg, &flags) == -1) {
482 errx(EXIT_FAILURE, "Invalid discard mode `%s'", optarg);
484 break;
485 case 'b':
486 bindto = optarg;
487 break;
488 case 'p':
489 li = strtol(optarg, &end, 0);
490 if (*end) {
491 errx(EXIT_FAILURE, "Invalid port `%s'", optarg);
493 if (li < 1 || li > 65535) {
494 errx(EXIT_FAILURE, "Port out of range `%s'", optarg);
496 port = (uint16_t)li;
497 break;
498 case 'o':
499 dev_offset = strtoll (optarg, &end, 0);
500 if (*end) {
501 errx(EXIT_FAILURE, "Invalid offset `%s'", optarg);
503 if (dev_offset < 0) {
504 errx(EXIT_FAILURE, "Offset must be positive `%s'", optarg);
506 break;
507 case 'l':
508 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
509 sn_opts = qemu_opts_parse(&internal_snapshot_opts, optarg, 0);
510 if (!sn_opts) {
511 errx(EXIT_FAILURE, "Failed in parsing snapshot param `%s'",
512 optarg);
514 } else {
515 sn_id_or_name = optarg;
517 /* fall through */
518 case 'r':
519 nbdflags |= NBD_FLAG_READ_ONLY;
520 flags &= ~BDRV_O_RDWR;
521 break;
522 case 'P':
523 partition = strtol(optarg, &end, 0);
524 if (*end)
525 errx(EXIT_FAILURE, "Invalid partition `%s'", optarg);
526 if (partition < 1 || partition > 8)
527 errx(EXIT_FAILURE, "Invalid partition %d", partition);
528 break;
529 case 'k':
530 sockpath = optarg;
531 if (sockpath[0] != '/')
532 errx(EXIT_FAILURE, "socket path must be absolute\n");
533 break;
534 case 'd':
535 disconnect = true;
536 break;
537 case 'c':
538 device = optarg;
539 break;
540 case 'e':
541 shared = strtol(optarg, &end, 0);
542 if (*end) {
543 errx(EXIT_FAILURE, "Invalid shared device number '%s'", optarg);
545 if (shared < 1) {
546 errx(EXIT_FAILURE, "Shared device number must be greater than 0\n");
548 break;
549 case 'f':
550 fmt = optarg;
551 break;
552 case 't':
553 persistent = 1;
554 break;
555 case 'v':
556 verbose = 1;
557 break;
558 case 'V':
559 version(argv[0]);
560 exit(0);
561 break;
562 case 'h':
563 usage(argv[0]);
564 exit(0);
565 break;
566 case '?':
567 errx(EXIT_FAILURE, "Try `%s --help' for more information.",
568 argv[0]);
572 if ((argc - optind) != 1) {
573 errx(EXIT_FAILURE, "Invalid number of argument.\n"
574 "Try `%s --help' for more information.",
575 argv[0]);
578 if (disconnect) {
579 fd = open(argv[optind], O_RDWR);
580 if (fd < 0) {
581 err(EXIT_FAILURE, "Cannot open %s", argv[optind]);
583 nbd_disconnect(fd);
585 close(fd);
587 printf("%s disconnected\n", argv[optind]);
589 return 0;
592 if (device && !verbose) {
593 int stderr_fd[2];
594 pid_t pid;
595 int ret;
597 if (qemu_pipe(stderr_fd) < 0) {
598 err(EXIT_FAILURE, "Error setting up communication pipe");
601 /* Now daemonize, but keep a communication channel open to
602 * print errors and exit with the proper status code.
604 pid = fork();
605 if (pid == 0) {
606 close(stderr_fd[0]);
607 ret = qemu_daemon(1, 0);
609 /* Temporarily redirect stderr to the parent's pipe... */
610 dup2(stderr_fd[1], STDERR_FILENO);
611 if (ret < 0) {
612 err(EXIT_FAILURE, "Failed to daemonize");
615 /* ... close the descriptor we inherited and go on. */
616 close(stderr_fd[1]);
617 } else {
618 bool errors = false;
619 char *buf;
621 /* In the parent. Print error messages from the child until
622 * it closes the pipe.
624 close(stderr_fd[1]);
625 buf = g_malloc(1024);
626 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) {
627 errors = true;
628 ret = qemu_write_full(STDERR_FILENO, buf, ret);
629 if (ret < 0) {
630 exit(EXIT_FAILURE);
633 if (ret < 0) {
634 err(EXIT_FAILURE, "Cannot read from daemon");
637 /* Usually the daemon should not print any message.
638 * Exit with zero status in that case.
640 exit(errors);
644 if (device != NULL && sockpath == NULL) {
645 sockpath = g_malloc(128);
646 snprintf(sockpath, 128, SOCKET_PATH, basename(device));
649 qemu_init_main_loop();
650 bdrv_init();
651 atexit(bdrv_close_all);
653 if (fmt) {
654 drv = bdrv_find_format(fmt);
655 if (!drv) {
656 errx(EXIT_FAILURE, "Unknown file format '%s'", fmt);
658 } else {
659 drv = NULL;
662 bs = bdrv_new("hda", &error_abort);
664 srcpath = argv[optind];
665 ret = bdrv_open(&bs, srcpath, NULL, NULL, flags, drv, &local_err);
666 if (ret < 0) {
667 errno = -ret;
668 err(EXIT_FAILURE, "Failed to bdrv_open '%s': %s", argv[optind],
669 error_get_pretty(local_err));
672 if (sn_opts) {
673 ret = bdrv_snapshot_load_tmp(bs,
674 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
675 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
676 &local_err);
677 } else if (sn_id_or_name) {
678 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name,
679 &local_err);
681 if (ret < 0) {
682 errno = -ret;
683 err(EXIT_FAILURE,
684 "Failed to load snapshot: %s",
685 error_get_pretty(local_err));
688 fd_size = bdrv_getlength(bs);
690 if (partition != -1) {
691 ret = find_partition(bs, partition, &dev_offset, &fd_size);
692 if (ret < 0) {
693 errno = -ret;
694 err(EXIT_FAILURE, "Could not find partition %d", partition);
698 exp = nbd_export_new(bs, dev_offset, fd_size, nbdflags, nbd_export_closed);
700 if (sockpath) {
701 fd = unix_socket_incoming(sockpath);
702 } else {
703 fd = tcp_socket_incoming(bindto, port);
706 if (fd < 0) {
707 return 1;
710 if (device) {
711 int ret;
713 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device);
714 if (ret != 0) {
715 errx(EXIT_FAILURE, "Failed to create client thread: %s",
716 strerror(ret));
718 } else {
719 /* Shut up GCC warnings. */
720 memset(&client_thread, 0, sizeof(client_thread));
723 qemu_set_fd_handler2(fd, nbd_can_accept, nbd_accept, NULL,
724 (void *)(uintptr_t)fd);
726 /* now when the initialization is (almost) complete, chdir("/")
727 * to free any busy filesystems */
728 if (chdir("/") < 0) {
729 err(EXIT_FAILURE, "Could not chdir to root directory");
732 state = RUNNING;
733 do {
734 main_loop_wait(false);
735 if (state == TERMINATE) {
736 state = TERMINATING;
737 nbd_export_close(exp);
738 nbd_export_put(exp);
739 exp = NULL;
741 } while (state != TERMINATED);
743 bdrv_close(bs);
744 if (sockpath) {
745 unlink(sockpath);
748 if (sn_opts) {
749 qemu_opts_del(sn_opts);
752 if (device) {
753 void *ret;
754 pthread_join(client_thread, &ret);
755 exit(ret != NULL);
756 } else {
757 exit(EXIT_SUCCESS);