Merge remote-tracking branch 'remotes/cohuck-gitlab/tags/s390x-20210121' into staging
[qemu.git] / qemu-img.c
1 /*
2 * QEMU disk image utility
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27
28 #include "qemu-common.h"
29 #include "qemu-version.h"
30 #include "qapi/error.h"
31 #include "qapi/qapi-commands-block-core.h"
32 #include "qapi/qapi-visit-block-core.h"
33 #include "qapi/qobject-output-visitor.h"
34 #include "qapi/qmp/qjson.h"
35 #include "qapi/qmp/qdict.h"
36 #include "qemu/cutils.h"
37 #include "qemu/config-file.h"
38 #include "qemu/option.h"
39 #include "qemu/error-report.h"
40 #include "qemu/log.h"
41 #include "qemu/main-loop.h"
42 #include "qemu/module.h"
43 #include "qemu/sockets.h"
44 #include "qemu/units.h"
45 #include "qom/object_interfaces.h"
46 #include "sysemu/block-backend.h"
47 #include "block/block_int.h"
48 #include "block/blockjob.h"
49 #include "block/qapi.h"
50 #include "crypto/init.h"
51 #include "trace/control.h"
52 #include "qemu/throttle.h"
53 #include "block/throttle-groups.h"
54
55 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
56 "\n" QEMU_COPYRIGHT "\n"
57
58 typedef struct img_cmd_t {
59 const char *name;
60 int (*handler)(int argc, char **argv);
61 } img_cmd_t;
62
63 enum {
64 OPTION_OUTPUT = 256,
65 OPTION_BACKING_CHAIN = 257,
66 OPTION_OBJECT = 258,
67 OPTION_IMAGE_OPTS = 259,
68 OPTION_PATTERN = 260,
69 OPTION_FLUSH_INTERVAL = 261,
70 OPTION_NO_DRAIN = 262,
71 OPTION_TARGET_IMAGE_OPTS = 263,
72 OPTION_SIZE = 264,
73 OPTION_PREALLOCATION = 265,
74 OPTION_SHRINK = 266,
75 OPTION_SALVAGE = 267,
76 OPTION_TARGET_IS_ZERO = 268,
77 OPTION_ADD = 269,
78 OPTION_REMOVE = 270,
79 OPTION_CLEAR = 271,
80 OPTION_ENABLE = 272,
81 OPTION_DISABLE = 273,
82 OPTION_MERGE = 274,
83 OPTION_BITMAPS = 275,
84 OPTION_FORCE = 276,
85 };
86
87 typedef enum OutputFormat {
88 OFORMAT_JSON,
89 OFORMAT_HUMAN,
90 } OutputFormat;
91
92 /* Default to cache=writeback as data integrity is not important for qemu-img */
93 #define BDRV_DEFAULT_CACHE "writeback"
94
95 static void format_print(void *opaque, const char *name)
96 {
97 printf(" %s", name);
98 }
99
100 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
101 {
102 va_list ap;
103
104 va_start(ap, fmt);
105 error_vreport(fmt, ap);
106 va_end(ap);
107
108 error_printf("Try 'qemu-img --help' for more information\n");
109 exit(EXIT_FAILURE);
110 }
111
112 static void QEMU_NORETURN missing_argument(const char *option)
113 {
114 error_exit("missing argument for option '%s'", option);
115 }
116
117 static void QEMU_NORETURN unrecognized_option(const char *option)
118 {
119 error_exit("unrecognized option '%s'", option);
120 }
121
122 /* Please keep in synch with docs/tools/qemu-img.rst */
123 static void QEMU_NORETURN help(void)
124 {
125 const char *help_msg =
126 QEMU_IMG_VERSION
127 "usage: qemu-img [standard options] command [command options]\n"
128 "QEMU disk image utility\n"
129 "\n"
130 " '-h', '--help' display this help and exit\n"
131 " '-V', '--version' output version information and exit\n"
132 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
133 " specify tracing options\n"
134 "\n"
135 "Command syntax:\n"
136 #define DEF(option, callback, arg_string) \
137 " " arg_string "\n"
138 #include "qemu-img-cmds.h"
139 #undef DEF
140 "\n"
141 "Command parameters:\n"
142 " 'filename' is a disk image filename\n"
143 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
144 " manual page for a description of the object properties. The most common\n"
145 " object type is a 'secret', which is used to supply passwords and/or\n"
146 " encryption keys.\n"
147 " 'fmt' is the disk image format. It is guessed automatically in most cases\n"
148 " 'cache' is the cache mode used to write the output disk image, the valid\n"
149 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
150 " 'directsync' and 'unsafe' (default for convert)\n"
151 " 'src_cache' is the cache mode used to read input disk images, the valid\n"
152 " options are the same as for the 'cache' option\n"
153 " 'size' is the disk image size in bytes. Optional suffixes\n"
154 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
155 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n"
156 " supported. 'b' is ignored.\n"
157 " 'output_filename' is the destination disk image filename\n"
158 " 'output_fmt' is the destination format\n"
159 " 'options' is a comma separated list of format specific options in a\n"
160 " name=value format. Use -o ? for an overview of the options supported by the\n"
161 " used format\n"
162 " 'snapshot_param' is param used for internal snapshot, format\n"
163 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
164 " '[ID_OR_NAME]'\n"
165 " '-c' indicates that target image must be compressed (qcow format only)\n"
166 " '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
167 " new backing file match exactly. The image doesn't need a working\n"
168 " backing file before rebasing in this case (useful for renaming the\n"
169 " backing file). For image creation, allow creating without attempting\n"
170 " to open the backing file.\n"
171 " '-h' with or without a command shows this help and lists the supported formats\n"
172 " '-p' show progress of command (only certain commands)\n"
173 " '-q' use Quiet mode - do not print any output (except errors)\n"
174 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
175 " contain only zeros for qemu-img to create a sparse image during\n"
176 " conversion. If the number of bytes is 0, the source will not be scanned for\n"
177 " unallocated or zero sectors, and the destination image will always be\n"
178 " fully allocated\n"
179 " '--output' takes the format in which the output must be done (human or json)\n"
180 " '-n' skips the target volume creation (useful if the volume is created\n"
181 " prior to running qemu-img)\n"
182 "\n"
183 "Parameters to bitmap subcommand:\n"
184 " 'bitmap' is the name of the bitmap to manipulate, through one or more\n"
185 " actions from '--add', '--remove', '--clear', '--enable', '--disable',\n"
186 " or '--merge source'\n"
187 " '-g granularity' sets the granularity for '--add' actions\n"
188 " '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n"
189 " bitmaps from an alternative file\n"
190 "\n"
191 "Parameters to check subcommand:\n"
192 " '-r' tries to repair any inconsistencies that are found during the check.\n"
193 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
194 " kinds of errors, with a higher risk of choosing the wrong fix or\n"
195 " hiding corruption that has already occurred.\n"
196 "\n"
197 "Parameters to convert subcommand:\n"
198 " '--bitmaps' copies all top-level persistent bitmaps to destination\n"
199 " '-m' specifies how many coroutines work in parallel during the convert\n"
200 " process (defaults to 8)\n"
201 " '-W' allow to write to the target out of order rather than sequential\n"
202 "\n"
203 "Parameters to snapshot subcommand:\n"
204 " 'snapshot' is the name of the snapshot to create, apply or delete\n"
205 " '-a' applies a snapshot (revert disk to saved state)\n"
206 " '-c' creates a snapshot\n"
207 " '-d' deletes a snapshot\n"
208 " '-l' lists all snapshots in the given image\n"
209 "\n"
210 "Parameters to compare subcommand:\n"
211 " '-f' first image format\n"
212 " '-F' second image format\n"
213 " '-s' run in Strict mode - fail on different image size or sector allocation\n"
214 "\n"
215 "Parameters to dd subcommand:\n"
216 " 'bs=BYTES' read and write up to BYTES bytes at a time "
217 "(default: 512)\n"
218 " 'count=N' copy only N input blocks\n"
219 " 'if=FILE' read from FILE\n"
220 " 'of=FILE' write to FILE\n"
221 " 'skip=N' skip N bs-sized blocks at the start of input\n";
222
223 printf("%s\nSupported formats:", help_msg);
224 bdrv_iterate_format(format_print, NULL, false);
225 printf("\n\n" QEMU_HELP_BOTTOM "\n");
226 exit(EXIT_SUCCESS);
227 }
228
229 static QemuOptsList qemu_object_opts = {
230 .name = "object",
231 .implied_opt_name = "qom-type",
232 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
233 .desc = {
234 { }
235 },
236 };
237
238 static bool qemu_img_object_print_help(const char *type, QemuOpts *opts)
239 {
240 if (user_creatable_print_help(type, opts)) {
241 exit(0);
242 }
243 return true;
244 }
245
246 /*
247 * Is @optarg safe for accumulate_options()?
248 * It is when multiple of them can be joined together separated by ','.
249 * To make that work, @optarg must not start with ',' (or else a
250 * separating ',' preceding it gets escaped), and it must not end with
251 * an odd number of ',' (or else a separating ',' following it gets
252 * escaped), or be empty (or else a separating ',' preceding it can
253 * escape a separating ',' following it).
254 *
255 */
256 static bool is_valid_option_list(const char *optarg)
257 {
258 size_t len = strlen(optarg);
259 size_t i;
260
261 if (!optarg[0] || optarg[0] == ',') {
262 return false;
263 }
264
265 for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
266 }
267 if ((len - i) % 2) {
268 return false;
269 }
270
271 return true;
272 }
273
274 static int accumulate_options(char **options, char *optarg)
275 {
276 char *new_options;
277
278 if (!is_valid_option_list(optarg)) {
279 error_report("Invalid option list: %s", optarg);
280 return -1;
281 }
282
283 if (!*options) {
284 *options = g_strdup(optarg);
285 } else {
286 new_options = g_strdup_printf("%s,%s", *options, optarg);
287 g_free(*options);
288 *options = new_options;
289 }
290 return 0;
291 }
292
293 static QemuOptsList qemu_source_opts = {
294 .name = "source",
295 .implied_opt_name = "file",
296 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
297 .desc = {
298 { }
299 },
300 };
301
302 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
303 {
304 int ret = 0;
305 if (!quiet) {
306 va_list args;
307 va_start(args, fmt);
308 ret = vprintf(fmt, args);
309 va_end(args);
310 }
311 return ret;
312 }
313
314
315 static int print_block_option_help(const char *filename, const char *fmt)
316 {
317 BlockDriver *drv, *proto_drv;
318 QemuOptsList *create_opts = NULL;
319 Error *local_err = NULL;
320
321 /* Find driver and parse its options */
322 drv = bdrv_find_format(fmt);
323 if (!drv) {
324 error_report("Unknown file format '%s'", fmt);
325 return 1;
326 }
327
328 if (!drv->create_opts) {
329 error_report("Format driver '%s' does not support image creation", fmt);
330 return 1;
331 }
332
333 create_opts = qemu_opts_append(create_opts, drv->create_opts);
334 if (filename) {
335 proto_drv = bdrv_find_protocol(filename, true, &local_err);
336 if (!proto_drv) {
337 error_report_err(local_err);
338 qemu_opts_free(create_opts);
339 return 1;
340 }
341 if (!proto_drv->create_opts) {
342 error_report("Protocol driver '%s' does not support image creation",
343 proto_drv->format_name);
344 qemu_opts_free(create_opts);
345 return 1;
346 }
347 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
348 }
349
350 if (filename) {
351 printf("Supported options:\n");
352 } else {
353 printf("Supported %s options:\n", fmt);
354 }
355 qemu_opts_print_help(create_opts, false);
356 qemu_opts_free(create_opts);
357
358 if (!filename) {
359 printf("\n"
360 "The protocol level may support further options.\n"
361 "Specify the target filename to include those options.\n");
362 }
363
364 return 0;
365 }
366
367
368 static BlockBackend *img_open_opts(const char *optstr,
369 QemuOpts *opts, int flags, bool writethrough,
370 bool quiet, bool force_share)
371 {
372 QDict *options;
373 Error *local_err = NULL;
374 BlockBackend *blk;
375 options = qemu_opts_to_qdict(opts, NULL);
376 if (force_share) {
377 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
378 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
379 error_report("--force-share/-U conflicts with image options");
380 qobject_unref(options);
381 return NULL;
382 }
383 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
384 }
385 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
386 if (!blk) {
387 error_reportf_err(local_err, "Could not open '%s': ", optstr);
388 return NULL;
389 }
390 blk_set_enable_write_cache(blk, !writethrough);
391
392 return blk;
393 }
394
395 static BlockBackend *img_open_file(const char *filename,
396 QDict *options,
397 const char *fmt, int flags,
398 bool writethrough, bool quiet,
399 bool force_share)
400 {
401 BlockBackend *blk;
402 Error *local_err = NULL;
403
404 if (!options) {
405 options = qdict_new();
406 }
407 if (fmt) {
408 qdict_put_str(options, "driver", fmt);
409 }
410
411 if (force_share) {
412 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
413 }
414 blk = blk_new_open(filename, NULL, options, flags, &local_err);
415 if (!blk) {
416 error_reportf_err(local_err, "Could not open '%s': ", filename);
417 return NULL;
418 }
419 blk_set_enable_write_cache(blk, !writethrough);
420
421 return blk;
422 }
423
424
425 static int img_add_key_secrets(void *opaque,
426 const char *name, const char *value,
427 Error **errp)
428 {
429 QDict *options = opaque;
430
431 if (g_str_has_suffix(name, "key-secret")) {
432 qdict_put_str(options, name, value);
433 }
434
435 return 0;
436 }
437
438
439 static BlockBackend *img_open(bool image_opts,
440 const char *filename,
441 const char *fmt, int flags, bool writethrough,
442 bool quiet, bool force_share)
443 {
444 BlockBackend *blk;
445 if (image_opts) {
446 QemuOpts *opts;
447 if (fmt) {
448 error_report("--image-opts and --format are mutually exclusive");
449 return NULL;
450 }
451 opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
452 filename, true);
453 if (!opts) {
454 return NULL;
455 }
456 blk = img_open_opts(filename, opts, flags, writethrough, quiet,
457 force_share);
458 } else {
459 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
460 force_share);
461 }
462 return blk;
463 }
464
465
466 static int add_old_style_options(const char *fmt, QemuOpts *opts,
467 const char *base_filename,
468 const char *base_fmt)
469 {
470 if (base_filename) {
471 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
472 NULL)) {
473 error_report("Backing file not supported for file format '%s'",
474 fmt);
475 return -1;
476 }
477 }
478 if (base_fmt) {
479 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
480 error_report("Backing file format not supported for file "
481 "format '%s'", fmt);
482 return -1;
483 }
484 }
485 return 0;
486 }
487
488 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
489 int64_t max)
490 {
491 int err;
492 uint64_t res;
493
494 err = qemu_strtosz(value, NULL, &res);
495 if (err < 0 && err != -ERANGE) {
496 error_report("Invalid %s specified. You may use "
497 "k, M, G, T, P or E suffixes for", name);
498 error_report("kilobytes, megabytes, gigabytes, terabytes, "
499 "petabytes and exabytes.");
500 return err;
501 }
502 if (err == -ERANGE || res > max || res < min) {
503 error_report("Invalid %s specified. Must be between %" PRId64
504 " and %" PRId64 ".", name, min, max);
505 return -ERANGE;
506 }
507 return res;
508 }
509
510 static int64_t cvtnum(const char *name, const char *value)
511 {
512 return cvtnum_full(name, value, 0, INT64_MAX);
513 }
514
515 static int img_create(int argc, char **argv)
516 {
517 int c;
518 uint64_t img_size = -1;
519 const char *fmt = "raw";
520 const char *base_fmt = NULL;
521 const char *filename;
522 const char *base_filename = NULL;
523 char *options = NULL;
524 Error *local_err = NULL;
525 bool quiet = false;
526 int flags = 0;
527
528 for(;;) {
529 static const struct option long_options[] = {
530 {"help", no_argument, 0, 'h'},
531 {"object", required_argument, 0, OPTION_OBJECT},
532 {0, 0, 0, 0}
533 };
534 c = getopt_long(argc, argv, ":F:b:f:ho:qu",
535 long_options, NULL);
536 if (c == -1) {
537 break;
538 }
539 switch(c) {
540 case ':':
541 missing_argument(argv[optind - 1]);
542 break;
543 case '?':
544 unrecognized_option(argv[optind - 1]);
545 break;
546 case 'h':
547 help();
548 break;
549 case 'F':
550 base_fmt = optarg;
551 break;
552 case 'b':
553 base_filename = optarg;
554 break;
555 case 'f':
556 fmt = optarg;
557 break;
558 case 'o':
559 if (accumulate_options(&options, optarg) < 0) {
560 goto fail;
561 }
562 break;
563 case 'q':
564 quiet = true;
565 break;
566 case 'u':
567 flags |= BDRV_O_NO_BACKING;
568 break;
569 case OPTION_OBJECT: {
570 QemuOpts *opts;
571 opts = qemu_opts_parse_noisily(&qemu_object_opts,
572 optarg, true);
573 if (!opts) {
574 goto fail;
575 }
576 } break;
577 }
578 }
579
580 /* Get the filename */
581 filename = (optind < argc) ? argv[optind] : NULL;
582 if (options && has_help_option(options)) {
583 g_free(options);
584 return print_block_option_help(filename, fmt);
585 }
586
587 if (optind >= argc) {
588 error_exit("Expecting image file name");
589 }
590 optind++;
591
592 if (qemu_opts_foreach(&qemu_object_opts,
593 user_creatable_add_opts_foreach,
594 qemu_img_object_print_help, &error_fatal)) {
595 goto fail;
596 }
597
598 /* Get image size, if specified */
599 if (optind < argc) {
600 int64_t sval;
601
602 sval = cvtnum("image size", argv[optind++]);
603 if (sval < 0) {
604 goto fail;
605 }
606 img_size = (uint64_t)sval;
607 }
608 if (optind != argc) {
609 error_exit("Unexpected argument: %s", argv[optind]);
610 }
611
612 bdrv_img_create(filename, fmt, base_filename, base_fmt,
613 options, img_size, flags, quiet, &local_err);
614 if (local_err) {
615 error_reportf_err(local_err, "%s: ", filename);
616 goto fail;
617 }
618
619 g_free(options);
620 return 0;
621
622 fail:
623 g_free(options);
624 return 1;
625 }
626
627 static void dump_json_image_check(ImageCheck *check, bool quiet)
628 {
629 GString *str;
630 QObject *obj;
631 Visitor *v = qobject_output_visitor_new(&obj);
632
633 visit_type_ImageCheck(v, NULL, &check, &error_abort);
634 visit_complete(v, &obj);
635 str = qobject_to_json_pretty(obj, true);
636 assert(str != NULL);
637 qprintf(quiet, "%s\n", str->str);
638 qobject_unref(obj);
639 visit_free(v);
640 g_string_free(str, true);
641 }
642
643 static void dump_human_image_check(ImageCheck *check, bool quiet)
644 {
645 if (!(check->corruptions || check->leaks || check->check_errors)) {
646 qprintf(quiet, "No errors were found on the image.\n");
647 } else {
648 if (check->corruptions) {
649 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
650 "Data may be corrupted, or further writes to the image "
651 "may corrupt it.\n",
652 check->corruptions);
653 }
654
655 if (check->leaks) {
656 qprintf(quiet,
657 "\n%" PRId64 " leaked clusters were found on the image.\n"
658 "This means waste of disk space, but no harm to data.\n",
659 check->leaks);
660 }
661
662 if (check->check_errors) {
663 qprintf(quiet,
664 "\n%" PRId64
665 " internal errors have occurred during the check.\n",
666 check->check_errors);
667 }
668 }
669
670 if (check->total_clusters != 0 && check->allocated_clusters != 0) {
671 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
672 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
673 check->allocated_clusters, check->total_clusters,
674 check->allocated_clusters * 100.0 / check->total_clusters,
675 check->fragmented_clusters * 100.0 / check->allocated_clusters,
676 check->compressed_clusters * 100.0 /
677 check->allocated_clusters);
678 }
679
680 if (check->image_end_offset) {
681 qprintf(quiet,
682 "Image end offset: %" PRId64 "\n", check->image_end_offset);
683 }
684 }
685
686 static int collect_image_check(BlockDriverState *bs,
687 ImageCheck *check,
688 const char *filename,
689 const char *fmt,
690 int fix)
691 {
692 int ret;
693 BdrvCheckResult result;
694
695 ret = bdrv_check(bs, &result, fix);
696 if (ret < 0) {
697 return ret;
698 }
699
700 check->filename = g_strdup(filename);
701 check->format = g_strdup(bdrv_get_format_name(bs));
702 check->check_errors = result.check_errors;
703 check->corruptions = result.corruptions;
704 check->has_corruptions = result.corruptions != 0;
705 check->leaks = result.leaks;
706 check->has_leaks = result.leaks != 0;
707 check->corruptions_fixed = result.corruptions_fixed;
708 check->has_corruptions_fixed = result.corruptions_fixed != 0;
709 check->leaks_fixed = result.leaks_fixed;
710 check->has_leaks_fixed = result.leaks_fixed != 0;
711 check->image_end_offset = result.image_end_offset;
712 check->has_image_end_offset = result.image_end_offset != 0;
713 check->total_clusters = result.bfi.total_clusters;
714 check->has_total_clusters = result.bfi.total_clusters != 0;
715 check->allocated_clusters = result.bfi.allocated_clusters;
716 check->has_allocated_clusters = result.bfi.allocated_clusters != 0;
717 check->fragmented_clusters = result.bfi.fragmented_clusters;
718 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0;
719 check->compressed_clusters = result.bfi.compressed_clusters;
720 check->has_compressed_clusters = result.bfi.compressed_clusters != 0;
721
722 return 0;
723 }
724
725 /*
726 * Checks an image for consistency. Exit codes:
727 *
728 * 0 - Check completed, image is good
729 * 1 - Check not completed because of internal errors
730 * 2 - Check completed, image is corrupted
731 * 3 - Check completed, image has leaked clusters, but is good otherwise
732 * 63 - Checks are not supported by the image format
733 */
734 static int img_check(int argc, char **argv)
735 {
736 int c, ret;
737 OutputFormat output_format = OFORMAT_HUMAN;
738 const char *filename, *fmt, *output, *cache;
739 BlockBackend *blk;
740 BlockDriverState *bs;
741 int fix = 0;
742 int flags = BDRV_O_CHECK;
743 bool writethrough;
744 ImageCheck *check;
745 bool quiet = false;
746 bool image_opts = false;
747 bool force_share = false;
748
749 fmt = NULL;
750 output = NULL;
751 cache = BDRV_DEFAULT_CACHE;
752
753 for(;;) {
754 int option_index = 0;
755 static const struct option long_options[] = {
756 {"help", no_argument, 0, 'h'},
757 {"format", required_argument, 0, 'f'},
758 {"repair", required_argument, 0, 'r'},
759 {"output", required_argument, 0, OPTION_OUTPUT},
760 {"object", required_argument, 0, OPTION_OBJECT},
761 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
762 {"force-share", no_argument, 0, 'U'},
763 {0, 0, 0, 0}
764 };
765 c = getopt_long(argc, argv, ":hf:r:T:qU",
766 long_options, &option_index);
767 if (c == -1) {
768 break;
769 }
770 switch(c) {
771 case ':':
772 missing_argument(argv[optind - 1]);
773 break;
774 case '?':
775 unrecognized_option(argv[optind - 1]);
776 break;
777 case 'h':
778 help();
779 break;
780 case 'f':
781 fmt = optarg;
782 break;
783 case 'r':
784 flags |= BDRV_O_RDWR;
785
786 if (!strcmp(optarg, "leaks")) {
787 fix = BDRV_FIX_LEAKS;
788 } else if (!strcmp(optarg, "all")) {
789 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
790 } else {
791 error_exit("Unknown option value for -r "
792 "(expecting 'leaks' or 'all'): %s", optarg);
793 }
794 break;
795 case OPTION_OUTPUT:
796 output = optarg;
797 break;
798 case 'T':
799 cache = optarg;
800 break;
801 case 'q':
802 quiet = true;
803 break;
804 case 'U':
805 force_share = true;
806 break;
807 case OPTION_OBJECT: {
808 QemuOpts *opts;
809 opts = qemu_opts_parse_noisily(&qemu_object_opts,
810 optarg, true);
811 if (!opts) {
812 return 1;
813 }
814 } break;
815 case OPTION_IMAGE_OPTS:
816 image_opts = true;
817 break;
818 }
819 }
820 if (optind != argc - 1) {
821 error_exit("Expecting one image file name");
822 }
823 filename = argv[optind++];
824
825 if (output && !strcmp(output, "json")) {
826 output_format = OFORMAT_JSON;
827 } else if (output && !strcmp(output, "human")) {
828 output_format = OFORMAT_HUMAN;
829 } else if (output) {
830 error_report("--output must be used with human or json as argument.");
831 return 1;
832 }
833
834 if (qemu_opts_foreach(&qemu_object_opts,
835 user_creatable_add_opts_foreach,
836 qemu_img_object_print_help, &error_fatal)) {
837 return 1;
838 }
839
840 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
841 if (ret < 0) {
842 error_report("Invalid source cache option: %s", cache);
843 return 1;
844 }
845
846 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
847 force_share);
848 if (!blk) {
849 return 1;
850 }
851 bs = blk_bs(blk);
852
853 check = g_new0(ImageCheck, 1);
854 ret = collect_image_check(bs, check, filename, fmt, fix);
855
856 if (ret == -ENOTSUP) {
857 error_report("This image format does not support checks");
858 ret = 63;
859 goto fail;
860 }
861
862 if (check->corruptions_fixed || check->leaks_fixed) {
863 int corruptions_fixed, leaks_fixed;
864 bool has_leaks_fixed, has_corruptions_fixed;
865
866 leaks_fixed = check->leaks_fixed;
867 has_leaks_fixed = check->has_leaks_fixed;
868 corruptions_fixed = check->corruptions_fixed;
869 has_corruptions_fixed = check->has_corruptions_fixed;
870
871 if (output_format == OFORMAT_HUMAN) {
872 qprintf(quiet,
873 "The following inconsistencies were found and repaired:\n\n"
874 " %" PRId64 " leaked clusters\n"
875 " %" PRId64 " corruptions\n\n"
876 "Double checking the fixed image now...\n",
877 check->leaks_fixed,
878 check->corruptions_fixed);
879 }
880
881 qapi_free_ImageCheck(check);
882 check = g_new0(ImageCheck, 1);
883 ret = collect_image_check(bs, check, filename, fmt, 0);
884
885 check->leaks_fixed = leaks_fixed;
886 check->has_leaks_fixed = has_leaks_fixed;
887 check->corruptions_fixed = corruptions_fixed;
888 check->has_corruptions_fixed = has_corruptions_fixed;
889 }
890
891 if (!ret) {
892 switch (output_format) {
893 case OFORMAT_HUMAN:
894 dump_human_image_check(check, quiet);
895 break;
896 case OFORMAT_JSON:
897 dump_json_image_check(check, quiet);
898 break;
899 }
900 }
901
902 if (ret || check->check_errors) {
903 if (ret) {
904 error_report("Check failed: %s", strerror(-ret));
905 } else {
906 error_report("Check failed");
907 }
908 ret = 1;
909 goto fail;
910 }
911
912 if (check->corruptions) {
913 ret = 2;
914 } else if (check->leaks) {
915 ret = 3;
916 } else {
917 ret = 0;
918 }
919
920 fail:
921 qapi_free_ImageCheck(check);
922 blk_unref(blk);
923 return ret;
924 }
925
926 typedef struct CommonBlockJobCBInfo {
927 BlockDriverState *bs;
928 Error **errp;
929 } CommonBlockJobCBInfo;
930
931 static void common_block_job_cb(void *opaque, int ret)
932 {
933 CommonBlockJobCBInfo *cbi = opaque;
934
935 if (ret < 0) {
936 error_setg_errno(cbi->errp, -ret, "Block job failed");
937 }
938 }
939
940 static void run_block_job(BlockJob *job, Error **errp)
941 {
942 AioContext *aio_context = blk_get_aio_context(job->blk);
943 int ret = 0;
944
945 aio_context_acquire(aio_context);
946 job_ref(&job->job);
947 do {
948 float progress = 0.0f;
949 aio_poll(aio_context, true);
950 if (job->job.progress.total) {
951 progress = (float)job->job.progress.current /
952 job->job.progress.total * 100.f;
953 }
954 qemu_progress_print(progress, 0);
955 } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
956
957 if (!job_is_completed(&job->job)) {
958 ret = job_complete_sync(&job->job, errp);
959 } else {
960 ret = job->job.ret;
961 }
962 job_unref(&job->job);
963 aio_context_release(aio_context);
964
965 /* publish completion progress only when success */
966 if (!ret) {
967 qemu_progress_print(100.f, 0);
968 }
969 }
970
971 static int img_commit(int argc, char **argv)
972 {
973 int c, ret, flags;
974 const char *filename, *fmt, *cache, *base;
975 BlockBackend *blk;
976 BlockDriverState *bs, *base_bs;
977 BlockJob *job;
978 bool progress = false, quiet = false, drop = false;
979 bool writethrough;
980 Error *local_err = NULL;
981 CommonBlockJobCBInfo cbi;
982 bool image_opts = false;
983 AioContext *aio_context;
984 int64_t rate_limit = 0;
985
986 fmt = NULL;
987 cache = BDRV_DEFAULT_CACHE;
988 base = NULL;
989 for(;;) {
990 static const struct option long_options[] = {
991 {"help", no_argument, 0, 'h'},
992 {"object", required_argument, 0, OPTION_OBJECT},
993 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
994 {0, 0, 0, 0}
995 };
996 c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
997 long_options, NULL);
998 if (c == -1) {
999 break;
1000 }
1001 switch(c) {
1002 case ':':
1003 missing_argument(argv[optind - 1]);
1004 break;
1005 case '?':
1006 unrecognized_option(argv[optind - 1]);
1007 break;
1008 case 'h':
1009 help();
1010 break;
1011 case 'f':
1012 fmt = optarg;
1013 break;
1014 case 't':
1015 cache = optarg;
1016 break;
1017 case 'b':
1018 base = optarg;
1019 /* -b implies -d */
1020 drop = true;
1021 break;
1022 case 'd':
1023 drop = true;
1024 break;
1025 case 'p':
1026 progress = true;
1027 break;
1028 case 'q':
1029 quiet = true;
1030 break;
1031 case 'r':
1032 rate_limit = cvtnum("rate limit", optarg);
1033 if (rate_limit < 0) {
1034 return 1;
1035 }
1036 break;
1037 case OPTION_OBJECT: {
1038 QemuOpts *opts;
1039 opts = qemu_opts_parse_noisily(&qemu_object_opts,
1040 optarg, true);
1041 if (!opts) {
1042 return 1;
1043 }
1044 } break;
1045 case OPTION_IMAGE_OPTS:
1046 image_opts = true;
1047 break;
1048 }
1049 }
1050
1051 /* Progress is not shown in Quiet mode */
1052 if (quiet) {
1053 progress = false;
1054 }
1055
1056 if (optind != argc - 1) {
1057 error_exit("Expecting one image file name");
1058 }
1059 filename = argv[optind++];
1060
1061 if (qemu_opts_foreach(&qemu_object_opts,
1062 user_creatable_add_opts_foreach,
1063 qemu_img_object_print_help, &error_fatal)) {
1064 return 1;
1065 }
1066
1067 flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1068 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1069 if (ret < 0) {
1070 error_report("Invalid cache option: %s", cache);
1071 return 1;
1072 }
1073
1074 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1075 false);
1076 if (!blk) {
1077 return 1;
1078 }
1079 bs = blk_bs(blk);
1080
1081 qemu_progress_init(progress, 1.f);
1082 qemu_progress_print(0.f, 100);
1083
1084 if (base) {
1085 base_bs = bdrv_find_backing_image(bs, base);
1086 if (!base_bs) {
1087 error_setg(&local_err,
1088 "Did not find '%s' in the backing chain of '%s'",
1089 base, filename);
1090 goto done;
1091 }
1092 } else {
1093 /* This is different from QMP, which by default uses the deepest file in
1094 * the backing chain (i.e., the very base); however, the traditional
1095 * behavior of qemu-img commit is using the immediate backing file. */
1096 base_bs = bdrv_backing_chain_next(bs);
1097 if (!base_bs) {
1098 error_setg(&local_err, "Image does not have a backing file");
1099 goto done;
1100 }
1101 }
1102
1103 cbi = (CommonBlockJobCBInfo){
1104 .errp = &local_err,
1105 .bs = bs,
1106 };
1107
1108 aio_context = bdrv_get_aio_context(bs);
1109 aio_context_acquire(aio_context);
1110 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1111 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1112 &cbi, false, &local_err);
1113 aio_context_release(aio_context);
1114 if (local_err) {
1115 goto done;
1116 }
1117
1118 /* When the block job completes, the BlockBackend reference will point to
1119 * the old backing file. In order to avoid that the top image is already
1120 * deleted, so we can still empty it afterwards, increment the reference
1121 * counter here preemptively. */
1122 if (!drop) {
1123 bdrv_ref(bs);
1124 }
1125
1126 job = block_job_get("commit");
1127 assert(job);
1128 run_block_job(job, &local_err);
1129 if (local_err) {
1130 goto unref_backing;
1131 }
1132
1133 if (!drop) {
1134 BlockBackend *old_backing_blk;
1135
1136 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1137 &local_err);
1138 if (!old_backing_blk) {
1139 goto unref_backing;
1140 }
1141 ret = blk_make_empty(old_backing_blk, &local_err);
1142 blk_unref(old_backing_blk);
1143 if (ret == -ENOTSUP) {
1144 error_free(local_err);
1145 local_err = NULL;
1146 } else if (ret < 0) {
1147 goto unref_backing;
1148 }
1149 }
1150
1151 unref_backing:
1152 if (!drop) {
1153 bdrv_unref(bs);
1154 }
1155
1156 done:
1157 qemu_progress_end();
1158
1159 blk_unref(blk);
1160
1161 if (local_err) {
1162 error_report_err(local_err);
1163 return 1;
1164 }
1165
1166 qprintf(quiet, "Image committed.\n");
1167 return 0;
1168 }
1169
1170 /*
1171 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1172 * of the first sector boundary within buf where the sector contains a
1173 * non-zero byte. This function is robust to a buffer that is not
1174 * sector-aligned.
1175 */
1176 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1177 {
1178 int64_t i;
1179 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1180
1181 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1182 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1183 return i;
1184 }
1185 }
1186 if (i < n && !buffer_is_zero(buf + i, n - end)) {
1187 return i;
1188 }
1189 return -1;
1190 }
1191
1192 /*
1193 * Returns true iff the first sector pointed to by 'buf' contains at least
1194 * a non-NUL byte.
1195 *
1196 * 'pnum' is set to the number of sectors (including and immediately following
1197 * the first one) that are known to be in the same allocated/unallocated state.
1198 * The function will try to align the end offset to alignment boundaries so
1199 * that the request will at least end aligned and consecutive requests will
1200 * also start at an aligned offset.
1201 */
1202 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1203 int64_t sector_num, int alignment)
1204 {
1205 bool is_zero;
1206 int i, tail;
1207
1208 if (n <= 0) {
1209 *pnum = 0;
1210 return 0;
1211 }
1212 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1213 for(i = 1; i < n; i++) {
1214 buf += BDRV_SECTOR_SIZE;
1215 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1216 break;
1217 }
1218 }
1219
1220 tail = (sector_num + i) & (alignment - 1);
1221 if (tail) {
1222 if (is_zero && i <= tail) {
1223 /* treat unallocated areas which only consist
1224 * of a small tail as allocated. */
1225 is_zero = false;
1226 }
1227 if (!is_zero) {
1228 /* align up end offset of allocated areas. */
1229 i += alignment - tail;
1230 i = MIN(i, n);
1231 } else {
1232 /* align down end offset of zero areas. */
1233 i -= tail;
1234 }
1235 }
1236 *pnum = i;
1237 return !is_zero;
1238 }
1239
1240 /*
1241 * Like is_allocated_sectors, but if the buffer starts with a used sector,
1242 * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1243 * breaking up write requests for only small sparse areas.
1244 */
1245 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1246 int min, int64_t sector_num, int alignment)
1247 {
1248 int ret;
1249 int num_checked, num_used;
1250
1251 if (n < min) {
1252 min = n;
1253 }
1254
1255 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1256 if (!ret) {
1257 return ret;
1258 }
1259
1260 num_used = *pnum;
1261 buf += BDRV_SECTOR_SIZE * *pnum;
1262 n -= *pnum;
1263 sector_num += *pnum;
1264 num_checked = num_used;
1265
1266 while (n > 0) {
1267 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1268
1269 buf += BDRV_SECTOR_SIZE * *pnum;
1270 n -= *pnum;
1271 sector_num += *pnum;
1272 num_checked += *pnum;
1273 if (ret) {
1274 num_used = num_checked;
1275 } else if (*pnum >= min) {
1276 break;
1277 }
1278 }
1279
1280 *pnum = num_used;
1281 return 1;
1282 }
1283
1284 /*
1285 * Compares two buffers sector by sector. Returns 0 if the first
1286 * sector of each buffer matches, non-zero otherwise.
1287 *
1288 * pnum is set to the sector-aligned size of the buffer prefix that
1289 * has the same matching status as the first sector.
1290 */
1291 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1292 int64_t bytes, int64_t *pnum)
1293 {
1294 bool res;
1295 int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1296
1297 assert(bytes > 0);
1298
1299 res = !!memcmp(buf1, buf2, i);
1300 while (i < bytes) {
1301 int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1302
1303 if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1304 break;
1305 }
1306 i += len;
1307 }
1308
1309 *pnum = i;
1310 return res;
1311 }
1312
1313 #define IO_BUF_SIZE (2 * MiB)
1314
1315 /*
1316 * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1317 *
1318 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1319 * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1320 * failure), and 4 on error (the exit status for read errors), after emitting
1321 * an error message.
1322 *
1323 * @param blk: BlockBackend for the image
1324 * @param offset: Starting offset to check
1325 * @param bytes: Number of bytes to check
1326 * @param filename: Name of disk file we are checking (logging purpose)
1327 * @param buffer: Allocated buffer for storing read data
1328 * @param quiet: Flag for quiet mode
1329 */
1330 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1331 int64_t bytes, const char *filename,
1332 uint8_t *buffer, bool quiet)
1333 {
1334 int ret = 0;
1335 int64_t idx;
1336
1337 ret = blk_pread(blk, offset, buffer, bytes);
1338 if (ret < 0) {
1339 error_report("Error while reading offset %" PRId64 " of %s: %s",
1340 offset, filename, strerror(-ret));
1341 return 4;
1342 }
1343 idx = find_nonzero(buffer, bytes);
1344 if (idx >= 0) {
1345 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1346 offset + idx);
1347 return 1;
1348 }
1349
1350 return 0;
1351 }
1352
1353 /*
1354 * Compares two images. Exit codes:
1355 *
1356 * 0 - Images are identical
1357 * 1 - Images differ
1358 * >1 - Error occurred
1359 */
1360 static int img_compare(int argc, char **argv)
1361 {
1362 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1363 BlockBackend *blk1, *blk2;
1364 BlockDriverState *bs1, *bs2;
1365 int64_t total_size1, total_size2;
1366 uint8_t *buf1 = NULL, *buf2 = NULL;
1367 int64_t pnum1, pnum2;
1368 int allocated1, allocated2;
1369 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1370 bool progress = false, quiet = false, strict = false;
1371 int flags;
1372 bool writethrough;
1373 int64_t total_size;
1374 int64_t offset = 0;
1375 int64_t chunk;
1376 int c;
1377 uint64_t progress_base;
1378 bool image_opts = false;
1379 bool force_share = false;
1380
1381 cache = BDRV_DEFAULT_CACHE;
1382 for (;;) {
1383 static const struct option long_options[] = {
1384 {"help", no_argument, 0, 'h'},
1385 {"object", required_argument, 0, OPTION_OBJECT},
1386 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1387 {"force-share", no_argument, 0, 'U'},
1388 {0, 0, 0, 0}
1389 };
1390 c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1391 long_options, NULL);
1392 if (c == -1) {
1393 break;
1394 }
1395 switch (c) {
1396 case ':':
1397 missing_argument(argv[optind - 1]);
1398 break;
1399 case '?':
1400 unrecognized_option(argv[optind - 1]);
1401 break;
1402 case 'h':
1403 help();
1404 break;
1405 case 'f':
1406 fmt1 = optarg;
1407 break;
1408 case 'F':
1409 fmt2 = optarg;
1410 break;
1411 case 'T':
1412 cache = optarg;
1413 break;
1414 case 'p':
1415 progress = true;
1416 break;
1417 case 'q':
1418 quiet = true;
1419 break;
1420 case 's':
1421 strict = true;
1422 break;
1423 case 'U':
1424 force_share = true;
1425 break;
1426 case OPTION_OBJECT: {
1427 QemuOpts *opts;
1428 opts = qemu_opts_parse_noisily(&qemu_object_opts,
1429 optarg, true);
1430 if (!opts) {
1431 ret = 2;
1432 goto out4;
1433 }
1434 } break;
1435 case OPTION_IMAGE_OPTS:
1436 image_opts = true;
1437 break;
1438 }
1439 }
1440
1441 /* Progress is not shown in Quiet mode */
1442 if (quiet) {
1443 progress = false;
1444 }
1445
1446
1447 if (optind != argc - 2) {
1448 error_exit("Expecting two image file names");
1449 }
1450 filename1 = argv[optind++];
1451 filename2 = argv[optind++];
1452
1453 if (qemu_opts_foreach(&qemu_object_opts,
1454 user_creatable_add_opts_foreach,
1455 qemu_img_object_print_help, &error_fatal)) {
1456 ret = 2;
1457 goto out4;
1458 }
1459
1460 /* Initialize before goto out */
1461 qemu_progress_init(progress, 2.0);
1462
1463 flags = 0;
1464 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1465 if (ret < 0) {
1466 error_report("Invalid source cache option: %s", cache);
1467 ret = 2;
1468 goto out3;
1469 }
1470
1471 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1472 force_share);
1473 if (!blk1) {
1474 ret = 2;
1475 goto out3;
1476 }
1477
1478 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1479 force_share);
1480 if (!blk2) {
1481 ret = 2;
1482 goto out2;
1483 }
1484 bs1 = blk_bs(blk1);
1485 bs2 = blk_bs(blk2);
1486
1487 buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1488 buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1489 total_size1 = blk_getlength(blk1);
1490 if (total_size1 < 0) {
1491 error_report("Can't get size of %s: %s",
1492 filename1, strerror(-total_size1));
1493 ret = 4;
1494 goto out;
1495 }
1496 total_size2 = blk_getlength(blk2);
1497 if (total_size2 < 0) {
1498 error_report("Can't get size of %s: %s",
1499 filename2, strerror(-total_size2));
1500 ret = 4;
1501 goto out;
1502 }
1503 total_size = MIN(total_size1, total_size2);
1504 progress_base = MAX(total_size1, total_size2);
1505
1506 qemu_progress_print(0, 100);
1507
1508 if (strict && total_size1 != total_size2) {
1509 ret = 1;
1510 qprintf(quiet, "Strict mode: Image size mismatch!\n");
1511 goto out;
1512 }
1513
1514 while (offset < total_size) {
1515 int status1, status2;
1516
1517 status1 = bdrv_block_status_above(bs1, NULL, offset,
1518 total_size1 - offset, &pnum1, NULL,
1519 NULL);
1520 if (status1 < 0) {
1521 ret = 3;
1522 error_report("Sector allocation test failed for %s", filename1);
1523 goto out;
1524 }
1525 allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1526
1527 status2 = bdrv_block_status_above(bs2, NULL, offset,
1528 total_size2 - offset, &pnum2, NULL,
1529 NULL);
1530 if (status2 < 0) {
1531 ret = 3;
1532 error_report("Sector allocation test failed for %s", filename2);
1533 goto out;
1534 }
1535 allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1536
1537 assert(pnum1 && pnum2);
1538 chunk = MIN(pnum1, pnum2);
1539
1540 if (strict) {
1541 if (status1 != status2) {
1542 ret = 1;
1543 qprintf(quiet, "Strict mode: Offset %" PRId64
1544 " block status mismatch!\n", offset);
1545 goto out;
1546 }
1547 }
1548 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1549 /* nothing to do */
1550 } else if (allocated1 == allocated2) {
1551 if (allocated1) {
1552 int64_t pnum;
1553
1554 chunk = MIN(chunk, IO_BUF_SIZE);
1555 ret = blk_pread(blk1, offset, buf1, chunk);
1556 if (ret < 0) {
1557 error_report("Error while reading offset %" PRId64
1558 " of %s: %s",
1559 offset, filename1, strerror(-ret));
1560 ret = 4;
1561 goto out;
1562 }
1563 ret = blk_pread(blk2, offset, buf2, chunk);
1564 if (ret < 0) {
1565 error_report("Error while reading offset %" PRId64
1566 " of %s: %s",
1567 offset, filename2, strerror(-ret));
1568 ret = 4;
1569 goto out;
1570 }
1571 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1572 if (ret || pnum != chunk) {
1573 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1574 offset + (ret ? 0 : pnum));
1575 ret = 1;
1576 goto out;
1577 }
1578 }
1579 } else {
1580 chunk = MIN(chunk, IO_BUF_SIZE);
1581 if (allocated1) {
1582 ret = check_empty_sectors(blk1, offset, chunk,
1583 filename1, buf1, quiet);
1584 } else {
1585 ret = check_empty_sectors(blk2, offset, chunk,
1586 filename2, buf1, quiet);
1587 }
1588 if (ret) {
1589 goto out;
1590 }
1591 }
1592 offset += chunk;
1593 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1594 }
1595
1596 if (total_size1 != total_size2) {
1597 BlockBackend *blk_over;
1598 const char *filename_over;
1599
1600 qprintf(quiet, "Warning: Image size mismatch!\n");
1601 if (total_size1 > total_size2) {
1602 blk_over = blk1;
1603 filename_over = filename1;
1604 } else {
1605 blk_over = blk2;
1606 filename_over = filename2;
1607 }
1608
1609 while (offset < progress_base) {
1610 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1611 progress_base - offset, &chunk,
1612 NULL, NULL);
1613 if (ret < 0) {
1614 ret = 3;
1615 error_report("Sector allocation test failed for %s",
1616 filename_over);
1617 goto out;
1618
1619 }
1620 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1621 chunk = MIN(chunk, IO_BUF_SIZE);
1622 ret = check_empty_sectors(blk_over, offset, chunk,
1623 filename_over, buf1, quiet);
1624 if (ret) {
1625 goto out;
1626 }
1627 }
1628 offset += chunk;
1629 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1630 }
1631 }
1632
1633 qprintf(quiet, "Images are identical.\n");
1634 ret = 0;
1635
1636 out:
1637 qemu_vfree(buf1);
1638 qemu_vfree(buf2);
1639 blk_unref(blk2);
1640 out2:
1641 blk_unref(blk1);
1642 out3:
1643 qemu_progress_end();
1644 out4:
1645 return ret;
1646 }
1647
1648 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1649 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1650 const char *src_node, const char *src_name,
1651 Error **errp)
1652 {
1653 BlockDirtyBitmapMergeSource *merge_src;
1654 BlockDirtyBitmapMergeSourceList *list = NULL;
1655
1656 merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1657 merge_src->type = QTYPE_QDICT;
1658 merge_src->u.external.node = g_strdup(src_node);
1659 merge_src->u.external.name = g_strdup(src_name);
1660 QAPI_LIST_PREPEND(list, merge_src);
1661 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1662 qapi_free_BlockDirtyBitmapMergeSourceList(list);
1663 }
1664
1665 enum ImgConvertBlockStatus {
1666 BLK_DATA,
1667 BLK_ZERO,
1668 BLK_BACKING_FILE,
1669 };
1670
1671 #define MAX_COROUTINES 16
1672 #define CONVERT_THROTTLE_GROUP "img_convert"
1673
1674 typedef struct ImgConvertState {
1675 BlockBackend **src;
1676 int64_t *src_sectors;
1677 int *src_alignment;
1678 int src_num;
1679 int64_t total_sectors;
1680 int64_t allocated_sectors;
1681 int64_t allocated_done;
1682 int64_t sector_num;
1683 int64_t wr_offs;
1684 enum ImgConvertBlockStatus status;
1685 int64_t sector_next_status;
1686 BlockBackend *target;
1687 bool has_zero_init;
1688 bool compressed;
1689 bool target_is_new;
1690 bool target_has_backing;
1691 int64_t target_backing_sectors; /* negative if unknown */
1692 bool wr_in_order;
1693 bool copy_range;
1694 bool salvage;
1695 bool quiet;
1696 int min_sparse;
1697 int alignment;
1698 size_t cluster_sectors;
1699 size_t buf_sectors;
1700 long num_coroutines;
1701 int running_coroutines;
1702 Coroutine *co[MAX_COROUTINES];
1703 int64_t wait_sector_num[MAX_COROUTINES];
1704 CoMutex lock;
1705 int ret;
1706 } ImgConvertState;
1707
1708 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1709 int *src_cur, int64_t *src_cur_offset)
1710 {
1711 *src_cur = 0;
1712 *src_cur_offset = 0;
1713 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1714 *src_cur_offset += s->src_sectors[*src_cur];
1715 (*src_cur)++;
1716 assert(*src_cur < s->src_num);
1717 }
1718 }
1719
1720 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1721 {
1722 int64_t src_cur_offset;
1723 int ret, n, src_cur;
1724 bool post_backing_zero = false;
1725
1726 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1727
1728 assert(s->total_sectors > sector_num);
1729 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1730
1731 if (s->target_backing_sectors >= 0) {
1732 if (sector_num >= s->target_backing_sectors) {
1733 post_backing_zero = true;
1734 } else if (sector_num + n > s->target_backing_sectors) {
1735 /* Split requests around target_backing_sectors (because
1736 * starting from there, zeros are handled differently) */
1737 n = s->target_backing_sectors - sector_num;
1738 }
1739 }
1740
1741 if (s->sector_next_status <= sector_num) {
1742 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1743 int64_t count;
1744 int tail;
1745 BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1746 BlockDriverState *base;
1747
1748 if (s->target_has_backing) {
1749 base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1750 } else {
1751 base = NULL;
1752 }
1753
1754 do {
1755 count = n * BDRV_SECTOR_SIZE;
1756
1757 ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1758 NULL, NULL);
1759
1760 if (ret < 0) {
1761 if (s->salvage) {
1762 if (n == 1) {
1763 if (!s->quiet) {
1764 warn_report("error while reading block status at "
1765 "offset %" PRIu64 ": %s", offset,
1766 strerror(-ret));
1767 }
1768 /* Just try to read the data, then */
1769 ret = BDRV_BLOCK_DATA;
1770 count = BDRV_SECTOR_SIZE;
1771 } else {
1772 /* Retry on a shorter range */
1773 n = DIV_ROUND_UP(n, 4);
1774 }
1775 } else {
1776 error_report("error while reading block status at offset "
1777 "%" PRIu64 ": %s", offset, strerror(-ret));
1778 return ret;
1779 }
1780 }
1781 } while (ret < 0);
1782
1783 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1784
1785 /*
1786 * Avoid that s->sector_next_status becomes unaligned to the source
1787 * request alignment and/or cluster size to avoid unnecessary read
1788 * cycles.
1789 */
1790 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1791 if (n > tail) {
1792 n -= tail;
1793 }
1794
1795 if (ret & BDRV_BLOCK_ZERO) {
1796 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1797 } else if (ret & BDRV_BLOCK_DATA) {
1798 s->status = BLK_DATA;
1799 } else {
1800 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1801 }
1802
1803 s->sector_next_status = sector_num + n;
1804 }
1805
1806 n = MIN(n, s->sector_next_status - sector_num);
1807 if (s->status == BLK_DATA) {
1808 n = MIN(n, s->buf_sectors);
1809 }
1810
1811 /* We need to write complete clusters for compressed images, so if an
1812 * unallocated area is shorter than that, we must consider the whole
1813 * cluster allocated. */
1814 if (s->compressed) {
1815 if (n < s->cluster_sectors) {
1816 n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1817 s->status = BLK_DATA;
1818 } else {
1819 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1820 }
1821 }
1822
1823 return n;
1824 }
1825
1826 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1827 int nb_sectors, uint8_t *buf)
1828 {
1829 uint64_t single_read_until = 0;
1830 int n, ret;
1831
1832 assert(nb_sectors <= s->buf_sectors);
1833 while (nb_sectors > 0) {
1834 BlockBackend *blk;
1835 int src_cur;
1836 int64_t bs_sectors, src_cur_offset;
1837 uint64_t offset;
1838
1839 /* In the case of compression with multiple source files, we can get a
1840 * nb_sectors that spreads into the next part. So we must be able to
1841 * read across multiple BDSes for one convert_read() call. */
1842 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1843 blk = s->src[src_cur];
1844 bs_sectors = s->src_sectors[src_cur];
1845
1846 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1847
1848 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1849 if (single_read_until > offset) {
1850 n = 1;
1851 }
1852
1853 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1854 if (ret < 0) {
1855 if (s->salvage) {
1856 if (n > 1) {
1857 single_read_until = offset + (n << BDRV_SECTOR_BITS);
1858 continue;
1859 } else {
1860 if (!s->quiet) {
1861 warn_report("error while reading offset %" PRIu64
1862 ": %s", offset, strerror(-ret));
1863 }
1864 memset(buf, 0, BDRV_SECTOR_SIZE);
1865 }
1866 } else {
1867 return ret;
1868 }
1869 }
1870
1871 sector_num += n;
1872 nb_sectors -= n;
1873 buf += n * BDRV_SECTOR_SIZE;
1874 }
1875
1876 return 0;
1877 }
1878
1879
1880 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1881 int nb_sectors, uint8_t *buf,
1882 enum ImgConvertBlockStatus status)
1883 {
1884 int ret;
1885
1886 while (nb_sectors > 0) {
1887 int n = nb_sectors;
1888 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1889
1890 switch (status) {
1891 case BLK_BACKING_FILE:
1892 /* If we have a backing file, leave clusters unallocated that are
1893 * unallocated in the source image, so that the backing file is
1894 * visible at the respective offset. */
1895 assert(s->target_has_backing);
1896 break;
1897
1898 case BLK_DATA:
1899 /* If we're told to keep the target fully allocated (-S 0) or there
1900 * is real non-zero data, we must write it. Otherwise we can treat
1901 * it as zero sectors.
1902 * Compressed clusters need to be written as a whole, so in that
1903 * case we can only save the write if the buffer is completely
1904 * zeroed. */
1905 if (!s->min_sparse ||
1906 (!s->compressed &&
1907 is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1908 sector_num, s->alignment)) ||
1909 (s->compressed &&
1910 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1911 {
1912 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1913 n << BDRV_SECTOR_BITS, buf, flags);
1914 if (ret < 0) {
1915 return ret;
1916 }
1917 break;
1918 }
1919 /* fall-through */
1920
1921 case BLK_ZERO:
1922 if (s->has_zero_init) {
1923 assert(!s->target_has_backing);
1924 break;
1925 }
1926 ret = blk_co_pwrite_zeroes(s->target,
1927 sector_num << BDRV_SECTOR_BITS,
1928 n << BDRV_SECTOR_BITS,
1929 BDRV_REQ_MAY_UNMAP);
1930 if (ret < 0) {
1931 return ret;
1932 }
1933 break;
1934 }
1935
1936 sector_num += n;
1937 nb_sectors -= n;
1938 buf += n * BDRV_SECTOR_SIZE;
1939 }
1940
1941 return 0;
1942 }
1943
1944 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1945 int nb_sectors)
1946 {
1947 int n, ret;
1948
1949 while (nb_sectors > 0) {
1950 BlockBackend *blk;
1951 int src_cur;
1952 int64_t bs_sectors, src_cur_offset;
1953 int64_t offset;
1954
1955 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1956 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1957 blk = s->src[src_cur];
1958 bs_sectors = s->src_sectors[src_cur];
1959
1960 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1961
1962 ret = blk_co_copy_range(blk, offset, s->target,
1963 sector_num << BDRV_SECTOR_BITS,
1964 n << BDRV_SECTOR_BITS, 0, 0);
1965 if (ret < 0) {
1966 return ret;
1967 }
1968
1969 sector_num += n;
1970 nb_sectors -= n;
1971 }
1972 return 0;
1973 }
1974
1975 static void coroutine_fn convert_co_do_copy(void *opaque)
1976 {
1977 ImgConvertState *s = opaque;
1978 uint8_t *buf = NULL;
1979 int ret, i;
1980 int index = -1;
1981
1982 for (i = 0; i < s->num_coroutines; i++) {
1983 if (s->co[i] == qemu_coroutine_self()) {
1984 index = i;
1985 break;
1986 }
1987 }
1988 assert(index >= 0);
1989
1990 s->running_coroutines++;
1991 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1992
1993 while (1) {
1994 int n;
1995 int64_t sector_num;
1996 enum ImgConvertBlockStatus status;
1997 bool copy_range;
1998
1999 qemu_co_mutex_lock(&s->lock);
2000 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
2001 qemu_co_mutex_unlock(&s->lock);
2002 break;
2003 }
2004 n = convert_iteration_sectors(s, s->sector_num);
2005 if (n < 0) {
2006 qemu_co_mutex_unlock(&s->lock);
2007 s->ret = n;
2008 break;
2009 }
2010 /* save current sector and allocation status to local variables */
2011 sector_num = s->sector_num;
2012 status = s->status;
2013 if (!s->min_sparse && s->status == BLK_ZERO) {
2014 n = MIN(n, s->buf_sectors);
2015 }
2016 /* increment global sector counter so that other coroutines can
2017 * already continue reading beyond this request */
2018 s->sector_num += n;
2019 qemu_co_mutex_unlock(&s->lock);
2020
2021 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
2022 s->allocated_done += n;
2023 qemu_progress_print(100.0 * s->allocated_done /
2024 s->allocated_sectors, 0);
2025 }
2026
2027 retry:
2028 copy_range = s->copy_range && s->status == BLK_DATA;
2029 if (status == BLK_DATA && !copy_range) {
2030 ret = convert_co_read(s, sector_num, n, buf);
2031 if (ret < 0) {
2032 error_report("error while reading at byte %lld: %s",
2033 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2034 s->ret = ret;
2035 }
2036 } else if (!s->min_sparse && status == BLK_ZERO) {
2037 status = BLK_DATA;
2038 memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
2039 }
2040
2041 if (s->wr_in_order) {
2042 /* keep writes in order */
2043 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
2044 s->wait_sector_num[index] = sector_num;
2045 qemu_coroutine_yield();
2046 }
2047 s->wait_sector_num[index] = -1;
2048 }
2049
2050 if (s->ret == -EINPROGRESS) {
2051 if (copy_range) {
2052 ret = convert_co_copy_range(s, sector_num, n);
2053 if (ret) {
2054 s->copy_range = false;
2055 goto retry;
2056 }
2057 } else {
2058 ret = convert_co_write(s, sector_num, n, buf, status);
2059 }
2060 if (ret < 0) {
2061 error_report("error while writing at byte %lld: %s",
2062 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2063 s->ret = ret;
2064 }
2065 }
2066
2067 if (s->wr_in_order) {
2068 /* reenter the coroutine that might have waited
2069 * for this write to complete */
2070 s->wr_offs = sector_num + n;
2071 for (i = 0; i < s->num_coroutines; i++) {
2072 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2073 /*
2074 * A -> B -> A cannot occur because A has
2075 * s->wait_sector_num[i] == -1 during A -> B. Therefore
2076 * B will never enter A during this time window.
2077 */
2078 qemu_coroutine_enter(s->co[i]);
2079 break;
2080 }
2081 }
2082 }
2083 }
2084
2085 qemu_vfree(buf);
2086 s->co[index] = NULL;
2087 s->running_coroutines--;
2088 if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2089 /* the convert job finished successfully */
2090 s->ret = 0;
2091 }
2092 }
2093
2094 static int convert_do_copy(ImgConvertState *s)
2095 {
2096 int ret, i, n;
2097 int64_t sector_num = 0;
2098
2099 /* Check whether we have zero initialisation or can get it efficiently */
2100 if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2101 !s->target_has_backing) {
2102 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2103 }
2104
2105 /* Allocate buffer for copied data. For compressed images, only one cluster
2106 * can be copied at a time. */
2107 if (s->compressed) {
2108 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2109 error_report("invalid cluster size");
2110 return -EINVAL;
2111 }
2112 s->buf_sectors = s->cluster_sectors;
2113 }
2114
2115 while (sector_num < s->total_sectors) {
2116 n = convert_iteration_sectors(s, sector_num);
2117 if (n < 0) {
2118 return n;
2119 }
2120 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2121 {
2122 s->allocated_sectors += n;
2123 }
2124 sector_num += n;
2125 }
2126
2127 /* Do the copy */
2128 s->sector_next_status = 0;
2129 s->ret = -EINPROGRESS;
2130
2131 qemu_co_mutex_init(&s->lock);
2132 for (i = 0; i < s->num_coroutines; i++) {
2133 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2134 s->wait_sector_num[i] = -1;
2135 qemu_coroutine_enter(s->co[i]);
2136 }
2137
2138 while (s->running_coroutines) {
2139 main_loop_wait(false);
2140 }
2141
2142 if (s->compressed && !s->ret) {
2143 /* signal EOF to align */
2144 ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2145 if (ret < 0) {
2146 return ret;
2147 }
2148 }
2149
2150 return s->ret;
2151 }
2152
2153 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst)
2154 {
2155 BdrvDirtyBitmap *bm;
2156 Error *err = NULL;
2157
2158 FOR_EACH_DIRTY_BITMAP(src, bm) {
2159 const char *name;
2160
2161 if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2162 continue;
2163 }
2164 name = bdrv_dirty_bitmap_name(bm);
2165 qmp_block_dirty_bitmap_add(dst->node_name, name,
2166 true, bdrv_dirty_bitmap_granularity(bm),
2167 true, true,
2168 true, !bdrv_dirty_bitmap_enabled(bm),
2169 &err);
2170 if (err) {
2171 error_reportf_err(err, "Failed to create bitmap %s: ", name);
2172 return -1;
2173 }
2174
2175 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2176 &err);
2177 if (err) {
2178 error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2179 return -1;
2180 }
2181 }
2182
2183 return 0;
2184 }
2185
2186 #define MAX_BUF_SECTORS 32768
2187
2188 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2189 {
2190 ThrottleConfig cfg;
2191
2192 throttle_config_init(&cfg);
2193 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2194
2195 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2196 blk_set_io_limits(blk, &cfg);
2197 }
2198
2199 static int img_convert(int argc, char **argv)
2200 {
2201 int c, bs_i, flags, src_flags = 0;
2202 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2203 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2204 *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2205 BlockDriver *drv = NULL, *proto_drv = NULL;
2206 BlockDriverInfo bdi;
2207 BlockDriverState *out_bs;
2208 QemuOpts *opts = NULL, *sn_opts = NULL;
2209 QemuOptsList *create_opts = NULL;
2210 QDict *open_opts = NULL;
2211 char *options = NULL;
2212 Error *local_err = NULL;
2213 bool writethrough, src_writethrough, image_opts = false,
2214 skip_create = false, progress = false, tgt_image_opts = false;
2215 int64_t ret = -EINVAL;
2216 bool force_share = false;
2217 bool explict_min_sparse = false;
2218 bool bitmaps = false;
2219 int64_t rate_limit = 0;
2220
2221 ImgConvertState s = (ImgConvertState) {
2222 /* Need at least 4k of zeros for sparse detection */
2223 .min_sparse = 8,
2224 .copy_range = false,
2225 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2226 .wr_in_order = true,
2227 .num_coroutines = 8,
2228 };
2229
2230 for(;;) {
2231 static const struct option long_options[] = {
2232 {"help", no_argument, 0, 'h'},
2233 {"object", required_argument, 0, OPTION_OBJECT},
2234 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2235 {"force-share", no_argument, 0, 'U'},
2236 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2237 {"salvage", no_argument, 0, OPTION_SALVAGE},
2238 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2239 {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2240 {0, 0, 0, 0}
2241 };
2242 c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WUr:",
2243 long_options, NULL);
2244 if (c == -1) {
2245 break;
2246 }
2247 switch(c) {
2248 case ':':
2249 missing_argument(argv[optind - 1]);
2250 break;
2251 case '?':
2252 unrecognized_option(argv[optind - 1]);
2253 break;
2254 case 'h':
2255 help();
2256 break;
2257 case 'f':
2258 fmt = optarg;
2259 break;
2260 case 'O':
2261 out_fmt = optarg;
2262 break;
2263 case 'B':
2264 out_baseimg = optarg;
2265 break;
2266 case 'C':
2267 s.copy_range = true;
2268 break;
2269 case 'c':
2270 s.compressed = true;
2271 break;
2272 case 'o':
2273 if (accumulate_options(&options, optarg) < 0) {
2274 goto fail_getopt;
2275 }
2276 break;
2277 case 'l':
2278 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2279 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2280 optarg, false);
2281 if (!sn_opts) {
2282 error_report("Failed in parsing snapshot param '%s'",
2283 optarg);
2284 goto fail_getopt;
2285 }
2286 } else {
2287 snapshot_name = optarg;
2288 }
2289 break;
2290 case 'S':
2291 {
2292 int64_t sval;
2293
2294 sval = cvtnum("buffer size for sparse output", optarg);
2295 if (sval < 0) {
2296 goto fail_getopt;
2297 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2298 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2299 error_report("Invalid buffer size for sparse output specified. "
2300 "Valid sizes are multiples of %llu up to %llu. Select "
2301 "0 to disable sparse detection (fully allocates output).",
2302 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2303 goto fail_getopt;
2304 }
2305
2306 s.min_sparse = sval / BDRV_SECTOR_SIZE;
2307 explict_min_sparse = true;
2308 break;
2309 }
2310 case 'p':
2311 progress = true;
2312 break;
2313 case 't':
2314 cache = optarg;
2315 break;
2316 case 'T':
2317 src_cache = optarg;
2318 break;
2319 case 'q':
2320 s.quiet = true;
2321 break;
2322 case 'n':
2323 skip_create = true;
2324 break;
2325 case 'm':
2326 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2327 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2328 error_report("Invalid number of coroutines. Allowed number of"
2329 " coroutines is between 1 and %d", MAX_COROUTINES);
2330 goto fail_getopt;
2331 }
2332 break;
2333 case 'W':
2334 s.wr_in_order = false;
2335 break;
2336 case 'U':
2337 force_share = true;
2338 break;
2339 case 'r':
2340 rate_limit = cvtnum("rate limit", optarg);
2341 if (rate_limit < 0) {
2342 goto fail_getopt;
2343 }
2344 break;
2345 case OPTION_OBJECT: {
2346 QemuOpts *object_opts;
2347 object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2348 optarg, true);
2349 if (!object_opts) {
2350 goto fail_getopt;
2351 }
2352 break;
2353 }
2354 case OPTION_IMAGE_OPTS:
2355 image_opts = true;
2356 break;
2357 case OPTION_SALVAGE:
2358 s.salvage = true;
2359 break;
2360 case OPTION_TARGET_IMAGE_OPTS:
2361 tgt_image_opts = true;
2362 break;
2363 case OPTION_TARGET_IS_ZERO:
2364 /*
2365 * The user asserting that the target is blank has the
2366 * same effect as the target driver supporting zero
2367 * initialisation.
2368 */
2369 s.has_zero_init = true;
2370 break;
2371 case OPTION_BITMAPS:
2372 bitmaps = true;
2373 break;
2374 }
2375 }
2376
2377 if (!out_fmt && !tgt_image_opts) {
2378 out_fmt = "raw";
2379 }
2380
2381 if (qemu_opts_foreach(&qemu_object_opts,
2382 user_creatable_add_opts_foreach,
2383 qemu_img_object_print_help, &error_fatal)) {
2384 goto fail_getopt;
2385 }
2386
2387 if (s.compressed && s.copy_range) {
2388 error_report("Cannot enable copy offloading when -c is used");
2389 goto fail_getopt;
2390 }
2391
2392 if (explict_min_sparse && s.copy_range) {
2393 error_report("Cannot enable copy offloading when -S is used");
2394 goto fail_getopt;
2395 }
2396
2397 if (s.copy_range && s.salvage) {
2398 error_report("Cannot use copy offloading in salvaging mode");
2399 goto fail_getopt;
2400 }
2401
2402 if (tgt_image_opts && !skip_create) {
2403 error_report("--target-image-opts requires use of -n flag");
2404 goto fail_getopt;
2405 }
2406
2407 if (skip_create && options) {
2408 error_report("-o has no effect when skipping image creation");
2409 goto fail_getopt;
2410 }
2411
2412 if (s.has_zero_init && !skip_create) {
2413 error_report("--target-is-zero requires use of -n flag");
2414 goto fail_getopt;
2415 }
2416
2417 s.src_num = argc - optind - 1;
2418 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2419
2420 if (options && has_help_option(options)) {
2421 if (out_fmt) {
2422 ret = print_block_option_help(out_filename, out_fmt);
2423 goto fail_getopt;
2424 } else {
2425 error_report("Option help requires a format be specified");
2426 goto fail_getopt;
2427 }
2428 }
2429
2430 if (s.src_num < 1) {
2431 error_report("Must specify image file name");
2432 goto fail_getopt;
2433 }
2434
2435 /* ret is still -EINVAL until here */
2436 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2437 if (ret < 0) {
2438 error_report("Invalid source cache option: %s", src_cache);
2439 goto fail_getopt;
2440 }
2441
2442 /* Initialize before goto out */
2443 if (s.quiet) {
2444 progress = false;
2445 }
2446 qemu_progress_init(progress, 1.0);
2447 qemu_progress_print(0, 100);
2448
2449 s.src = g_new0(BlockBackend *, s.src_num);
2450 s.src_sectors = g_new(int64_t, s.src_num);
2451 s.src_alignment = g_new(int, s.src_num);
2452
2453 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2454 BlockDriverState *src_bs;
2455 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2456 fmt, src_flags, src_writethrough, s.quiet,
2457 force_share);
2458 if (!s.src[bs_i]) {
2459 ret = -1;
2460 goto out;
2461 }
2462 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2463 if (s.src_sectors[bs_i] < 0) {
2464 error_report("Could not get size of %s: %s",
2465 argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2466 ret = -1;
2467 goto out;
2468 }
2469 src_bs = blk_bs(s.src[bs_i]);
2470 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2471 BDRV_SECTOR_SIZE);
2472 if (!bdrv_get_info(src_bs, &bdi)) {
2473 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2474 bdi.cluster_size / BDRV_SECTOR_SIZE);
2475 }
2476 s.total_sectors += s.src_sectors[bs_i];
2477 }
2478
2479 if (sn_opts) {
2480 bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2481 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2482 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2483 &local_err);
2484 } else if (snapshot_name != NULL) {
2485 if (s.src_num > 1) {
2486 error_report("No support for concatenating multiple snapshot");
2487 ret = -1;
2488 goto out;
2489 }
2490
2491 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2492 &local_err);
2493 }
2494 if (local_err) {
2495 error_reportf_err(local_err, "Failed to load snapshot: ");
2496 ret = -1;
2497 goto out;
2498 }
2499
2500 if (!skip_create) {
2501 /* Find driver and parse its options */
2502 drv = bdrv_find_format(out_fmt);
2503 if (!drv) {
2504 error_report("Unknown file format '%s'", out_fmt);
2505 ret = -1;
2506 goto out;
2507 }
2508
2509 proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2510 if (!proto_drv) {
2511 error_report_err(local_err);
2512 ret = -1;
2513 goto out;
2514 }
2515
2516 if (!drv->create_opts) {
2517 error_report("Format driver '%s' does not support image creation",
2518 drv->format_name);
2519 ret = -1;
2520 goto out;
2521 }
2522
2523 if (!proto_drv->create_opts) {
2524 error_report("Protocol driver '%s' does not support image creation",
2525 proto_drv->format_name);
2526 ret = -1;
2527 goto out;
2528 }
2529
2530 create_opts = qemu_opts_append(create_opts, drv->create_opts);
2531 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2532
2533 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2534 if (options) {
2535 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2536 error_report_err(local_err);
2537 ret = -1;
2538 goto out;
2539 }
2540 }
2541
2542 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2543 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2544 ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2545 if (ret < 0) {
2546 goto out;
2547 }
2548 }
2549
2550 /* Get backing file name if -o backing_file was used */
2551 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2552 if (out_baseimg_param) {
2553 out_baseimg = out_baseimg_param;
2554 }
2555 s.target_has_backing = (bool) out_baseimg;
2556
2557 if (s.has_zero_init && s.target_has_backing) {
2558 error_report("Cannot use --target-is-zero when the destination "
2559 "image has a backing file");
2560 goto out;
2561 }
2562
2563 if (s.src_num > 1 && out_baseimg) {
2564 error_report("Having a backing file for the target makes no sense when "
2565 "concatenating multiple input images");
2566 ret = -1;
2567 goto out;
2568 }
2569
2570 if (out_baseimg_param) {
2571 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2572 warn_report("Deprecated use of backing file without explicit "
2573 "backing format");
2574 }
2575 }
2576
2577 /* Check if compression is supported */
2578 if (s.compressed) {
2579 bool encryption =
2580 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2581 const char *encryptfmt =
2582 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2583 const char *preallocation =
2584 qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2585
2586 if (drv && !block_driver_can_compress(drv)) {
2587 error_report("Compression not supported for this file format");
2588 ret = -1;
2589 goto out;
2590 }
2591
2592 if (encryption || encryptfmt) {
2593 error_report("Compression and encryption not supported at "
2594 "the same time");
2595 ret = -1;
2596 goto out;
2597 }
2598
2599 if (preallocation
2600 && strcmp(preallocation, "off"))
2601 {
2602 error_report("Compression and preallocation not supported at "
2603 "the same time");
2604 ret = -1;
2605 goto out;
2606 }
2607 }
2608
2609 /* Determine if bitmaps need copying */
2610 if (bitmaps) {
2611 if (s.src_num > 1) {
2612 error_report("Copying bitmaps only possible with single source");
2613 ret = -1;
2614 goto out;
2615 }
2616 if (!bdrv_supports_persistent_dirty_bitmap(blk_bs(s.src[0]))) {
2617 error_report("Source lacks bitmap support");
2618 ret = -1;
2619 goto out;
2620 }
2621 }
2622
2623 /*
2624 * The later open call will need any decryption secrets, and
2625 * bdrv_create() will purge "opts", so extract them now before
2626 * they are lost.
2627 */
2628 if (!skip_create) {
2629 open_opts = qdict_new();
2630 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2631
2632 /* Create the new image */
2633 ret = bdrv_create(drv, out_filename, opts, &local_err);
2634 if (ret < 0) {
2635 error_reportf_err(local_err, "%s: error while converting %s: ",
2636 out_filename, out_fmt);
2637 goto out;
2638 }
2639 }
2640
2641 s.target_is_new = !skip_create;
2642
2643 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2644 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2645 if (ret < 0) {
2646 error_report("Invalid cache option: %s", cache);
2647 goto out;
2648 }
2649
2650 if (skip_create) {
2651 s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2652 flags, writethrough, s.quiet, false);
2653 } else {
2654 /* TODO ultimately we should allow --target-image-opts
2655 * to be used even when -n is not given.
2656 * That has to wait for bdrv_create to be improved
2657 * to allow filenames in option syntax
2658 */
2659 s.target = img_open_file(out_filename, open_opts, out_fmt,
2660 flags, writethrough, s.quiet, false);
2661 open_opts = NULL; /* blk_new_open will have freed it */
2662 }
2663 if (!s.target) {
2664 ret = -1;
2665 goto out;
2666 }
2667 out_bs = blk_bs(s.target);
2668
2669 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2670 error_report("Format driver '%s' does not support bitmaps",
2671 out_bs->drv->format_name);
2672 ret = -1;
2673 goto out;
2674 }
2675
2676 if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2677 error_report("Compression not supported for this file format");
2678 ret = -1;
2679 goto out;
2680 }
2681
2682 /* increase bufsectors from the default 4096 (2M) if opt_transfer
2683 * or discard_alignment of the out_bs is greater. Limit to
2684 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2685 s.buf_sectors = MIN(MAX_BUF_SECTORS,
2686 MAX(s.buf_sectors,
2687 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2688 out_bs->bl.pdiscard_alignment >>
2689 BDRV_SECTOR_BITS)));
2690
2691 /* try to align the write requests to the destination to avoid unnecessary
2692 * RMW cycles. */
2693 s.alignment = MAX(pow2floor(s.min_sparse),
2694 DIV_ROUND_UP(out_bs->bl.request_alignment,
2695 BDRV_SECTOR_SIZE));
2696 assert(is_power_of_2(s.alignment));
2697
2698 if (skip_create) {
2699 int64_t output_sectors = blk_nb_sectors(s.target);
2700 if (output_sectors < 0) {
2701 error_report("unable to get output image length: %s",
2702 strerror(-output_sectors));
2703 ret = -1;
2704 goto out;
2705 } else if (output_sectors < s.total_sectors) {
2706 error_report("output file is smaller than input file");
2707 ret = -1;
2708 goto out;
2709 }
2710 }
2711
2712 if (s.target_has_backing && s.target_is_new) {
2713 /* Errors are treated as "backing length unknown" (which means
2714 * s.target_backing_sectors has to be negative, which it will
2715 * be automatically). The backing file length is used only
2716 * for optimizations, so such a case is not fatal. */
2717 s.target_backing_sectors =
2718 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2719 } else {
2720 s.target_backing_sectors = -1;
2721 }
2722
2723 ret = bdrv_get_info(out_bs, &bdi);
2724 if (ret < 0) {
2725 if (s.compressed) {
2726 error_report("could not get block driver info");
2727 goto out;
2728 }
2729 } else {
2730 s.compressed = s.compressed || bdi.needs_compressed_writes;
2731 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2732 }
2733
2734 if (rate_limit) {
2735 set_rate_limit(s.target, rate_limit);
2736 }
2737
2738 ret = convert_do_copy(&s);
2739
2740 /* Now copy the bitmaps */
2741 if (bitmaps && ret == 0) {
2742 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs);
2743 }
2744
2745 out:
2746 if (!ret) {
2747 qemu_progress_print(100, 0);
2748 }
2749 qemu_progress_end();
2750 qemu_opts_del(opts);
2751 qemu_opts_free(create_opts);
2752 qobject_unref(open_opts);
2753 blk_unref(s.target);
2754 if (s.src) {
2755 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2756 blk_unref(s.src[bs_i]);
2757 }
2758 g_free(s.src);
2759 }
2760 g_free(s.src_sectors);
2761 g_free(s.src_alignment);
2762 fail_getopt:
2763 qemu_opts_del(sn_opts);
2764 g_free(options);
2765
2766 return !!ret;
2767 }
2768
2769
2770 static void dump_snapshots(BlockDriverState *bs)
2771 {
2772 QEMUSnapshotInfo *sn_tab, *sn;
2773 int nb_sns, i;
2774
2775 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2776 if (nb_sns <= 0)
2777 return;
2778 printf("Snapshot list:\n");
2779 bdrv_snapshot_dump(NULL);
2780 printf("\n");
2781 for(i = 0; i < nb_sns; i++) {
2782 sn = &sn_tab[i];
2783 bdrv_snapshot_dump(sn);
2784 printf("\n");
2785 }
2786 g_free(sn_tab);
2787 }
2788
2789 static void dump_json_image_info_list(ImageInfoList *list)
2790 {
2791 GString *str;
2792 QObject *obj;
2793 Visitor *v = qobject_output_visitor_new(&obj);
2794
2795 visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2796 visit_complete(v, &obj);
2797 str = qobject_to_json_pretty(obj, true);
2798 assert(str != NULL);
2799 printf("%s\n", str->str);
2800 qobject_unref(obj);
2801 visit_free(v);
2802 g_string_free(str, true);
2803 }
2804
2805 static void dump_json_image_info(ImageInfo *info)
2806 {
2807 GString *str;
2808 QObject *obj;
2809 Visitor *v = qobject_output_visitor_new(&obj);
2810
2811 visit_type_ImageInfo(v, NULL, &info, &error_abort);
2812 visit_complete(v, &obj);
2813 str = qobject_to_json_pretty(obj, true);
2814 assert(str != NULL);
2815 printf("%s\n", str->str);
2816 qobject_unref(obj);
2817 visit_free(v);
2818 g_string_free(str, true);
2819 }
2820
2821 static void dump_human_image_info_list(ImageInfoList *list)
2822 {
2823 ImageInfoList *elem;
2824 bool delim = false;
2825
2826 for (elem = list; elem; elem = elem->next) {
2827 if (delim) {
2828 printf("\n");
2829 }
2830 delim = true;
2831
2832 bdrv_image_info_dump(elem->value);
2833 }
2834 }
2835
2836 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2837 {
2838 return strcmp(a, b) == 0;
2839 }
2840
2841 /**
2842 * Open an image file chain and return an ImageInfoList
2843 *
2844 * @filename: topmost image filename
2845 * @fmt: topmost image format (may be NULL to autodetect)
2846 * @chain: true - enumerate entire backing file chain
2847 * false - only topmost image file
2848 *
2849 * Returns a list of ImageInfo objects or NULL if there was an error opening an
2850 * image file. If there was an error a message will have been printed to
2851 * stderr.
2852 */
2853 static ImageInfoList *collect_image_info_list(bool image_opts,
2854 const char *filename,
2855 const char *fmt,
2856 bool chain, bool force_share)
2857 {
2858 ImageInfoList *head = NULL;
2859 ImageInfoList **last = &head;
2860 GHashTable *filenames;
2861 Error *err = NULL;
2862
2863 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2864
2865 while (filename) {
2866 BlockBackend *blk;
2867 BlockDriverState *bs;
2868 ImageInfo *info;
2869 ImageInfoList *elem;
2870
2871 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2872 error_report("Backing file '%s' creates an infinite loop.",
2873 filename);
2874 goto err;
2875 }
2876 g_hash_table_insert(filenames, (gpointer)filename, NULL);
2877
2878 blk = img_open(image_opts, filename, fmt,
2879 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2880 force_share);
2881 if (!blk) {
2882 goto err;
2883 }
2884 bs = blk_bs(blk);
2885
2886 bdrv_query_image_info(bs, &info, &err);
2887 if (err) {
2888 error_report_err(err);
2889 blk_unref(blk);
2890 goto err;
2891 }
2892
2893 elem = g_new0(ImageInfoList, 1);
2894 elem->value = info;
2895 *last = elem;
2896 last = &elem->next;
2897
2898 blk_unref(blk);
2899
2900 /* Clear parameters that only apply to the topmost image */
2901 filename = fmt = NULL;
2902 image_opts = false;
2903
2904 if (chain) {
2905 if (info->has_full_backing_filename) {
2906 filename = info->full_backing_filename;
2907 } else if (info->has_backing_filename) {
2908 error_report("Could not determine absolute backing filename,"
2909 " but backing filename '%s' present",
2910 info->backing_filename);
2911 goto err;
2912 }
2913 if (info->has_backing_filename_format) {
2914 fmt = info->backing_filename_format;
2915 }
2916 }
2917 }
2918 g_hash_table_destroy(filenames);
2919 return head;
2920
2921 err:
2922 qapi_free_ImageInfoList(head);
2923 g_hash_table_destroy(filenames);
2924 return NULL;
2925 }
2926
2927 static int img_info(int argc, char **argv)
2928 {
2929 int c;
2930 OutputFormat output_format = OFORMAT_HUMAN;
2931 bool chain = false;
2932 const char *filename, *fmt, *output;
2933 ImageInfoList *list;
2934 bool image_opts = false;
2935 bool force_share = false;
2936
2937 fmt = NULL;
2938 output = NULL;
2939 for(;;) {
2940 int option_index = 0;
2941 static const struct option long_options[] = {
2942 {"help", no_argument, 0, 'h'},
2943 {"format", required_argument, 0, 'f'},
2944 {"output", required_argument, 0, OPTION_OUTPUT},
2945 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2946 {"object", required_argument, 0, OPTION_OBJECT},
2947 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2948 {"force-share", no_argument, 0, 'U'},
2949 {0, 0, 0, 0}
2950 };
2951 c = getopt_long(argc, argv, ":f:hU",
2952 long_options, &option_index);
2953 if (c == -1) {
2954 break;
2955 }
2956 switch(c) {
2957 case ':':
2958 missing_argument(argv[optind - 1]);
2959 break;
2960 case '?':
2961 unrecognized_option(argv[optind - 1]);
2962 break;
2963 case 'h':
2964 help();
2965 break;
2966 case 'f':
2967 fmt = optarg;
2968 break;
2969 case 'U':
2970 force_share = true;
2971 break;
2972 case OPTION_OUTPUT:
2973 output = optarg;
2974 break;
2975 case OPTION_BACKING_CHAIN:
2976 chain = true;
2977 break;
2978 case OPTION_OBJECT: {
2979 QemuOpts *opts;
2980 opts = qemu_opts_parse_noisily(&qemu_object_opts,
2981 optarg, true);
2982 if (!opts) {
2983 return 1;
2984 }
2985 } break;
2986 case OPTION_IMAGE_OPTS:
2987 image_opts = true;
2988 break;
2989 }
2990 }
2991 if (optind != argc - 1) {
2992 error_exit("Expecting one image file name");
2993 }
2994 filename = argv[optind++];
2995
2996 if (output && !strcmp(output, "json")) {
2997 output_format = OFORMAT_JSON;
2998 } else if (output && !strcmp(output, "human")) {
2999 output_format = OFORMAT_HUMAN;
3000 } else if (output) {
3001 error_report("--output must be used with human or json as argument.");
3002 return 1;
3003 }
3004
3005 if (qemu_opts_foreach(&qemu_object_opts,
3006 user_creatable_add_opts_foreach,
3007 qemu_img_object_print_help, &error_fatal)) {
3008 return 1;
3009 }
3010
3011 list = collect_image_info_list(image_opts, filename, fmt, chain,
3012 force_share);
3013 if (!list) {
3014 return 1;
3015 }
3016
3017 switch (output_format) {
3018 case OFORMAT_HUMAN:
3019 dump_human_image_info_list(list);
3020 break;
3021 case OFORMAT_JSON:
3022 if (chain) {
3023 dump_json_image_info_list(list);
3024 } else {
3025 dump_json_image_info(list->value);
3026 }
3027 break;
3028 }
3029
3030 qapi_free_ImageInfoList(list);
3031 return 0;
3032 }
3033
3034 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
3035 MapEntry *next)
3036 {
3037 switch (output_format) {
3038 case OFORMAT_HUMAN:
3039 if (e->data && !e->has_offset) {
3040 error_report("File contains external, encrypted or compressed clusters.");
3041 return -1;
3042 }
3043 if (e->data && !e->zero) {
3044 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
3045 e->start, e->length,
3046 e->has_offset ? e->offset : 0,
3047 e->has_filename ? e->filename : "");
3048 }
3049 /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
3050 * Modify the flags here to allow more coalescing.
3051 */
3052 if (next && (!next->data || next->zero)) {
3053 next->data = false;
3054 next->zero = true;
3055 }
3056 break;
3057 case OFORMAT_JSON:
3058 printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
3059 " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
3060 e->start, e->length, e->depth,
3061 e->zero ? "true" : "false",
3062 e->data ? "true" : "false");
3063 if (e->has_offset) {
3064 printf(", \"offset\": %"PRId64"", e->offset);
3065 }
3066 putchar('}');
3067
3068 if (next) {
3069 puts(",");
3070 }
3071 break;
3072 }
3073 return 0;
3074 }
3075
3076 static int get_block_status(BlockDriverState *bs, int64_t offset,
3077 int64_t bytes, MapEntry *e)
3078 {
3079 int ret;
3080 int depth;
3081 BlockDriverState *file;
3082 bool has_offset;
3083 int64_t map;
3084 char *filename = NULL;
3085
3086 /* As an optimization, we could cache the current range of unallocated
3087 * clusters in each file of the chain, and avoid querying the same
3088 * range repeatedly.
3089 */
3090
3091 depth = 0;
3092 for (;;) {
3093 bs = bdrv_skip_filters(bs);
3094 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3095 if (ret < 0) {
3096 return ret;
3097 }
3098 assert(bytes);
3099 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3100 break;
3101 }
3102 bs = bdrv_cow_bs(bs);
3103 if (bs == NULL) {
3104 ret = 0;
3105 break;
3106 }
3107
3108 depth++;
3109 }
3110
3111 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3112
3113 if (file && has_offset) {
3114 bdrv_refresh_filename(file);
3115 filename = file->filename;
3116 }
3117
3118 *e = (MapEntry) {
3119 .start = offset,
3120 .length = bytes,
3121 .data = !!(ret & BDRV_BLOCK_DATA),
3122 .zero = !!(ret & BDRV_BLOCK_ZERO),
3123 .offset = map,
3124 .has_offset = has_offset,
3125 .depth = depth,
3126 .has_filename = filename,
3127 .filename = filename,
3128 };
3129
3130 return 0;
3131 }
3132
3133 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3134 {
3135 if (curr->length == 0) {
3136 return false;
3137 }
3138 if (curr->zero != next->zero ||
3139 curr->data != next->data ||
3140 curr->depth != next->depth ||
3141 curr->has_filename != next->has_filename ||
3142 curr->has_offset != next->has_offset) {
3143 return false;
3144 }
3145 if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3146 return false;
3147 }
3148 if (curr->has_offset && curr->offset + curr->length != next->offset) {
3149 return false;
3150 }
3151 return true;
3152 }
3153
3154 static int img_map(int argc, char **argv)
3155 {
3156 int c;
3157 OutputFormat output_format = OFORMAT_HUMAN;
3158 BlockBackend *blk;
3159 BlockDriverState *bs;
3160 const char *filename, *fmt, *output;
3161 int64_t length;
3162 MapEntry curr = { .length = 0 }, next;
3163 int ret = 0;
3164 bool image_opts = false;
3165 bool force_share = false;
3166 int64_t start_offset = 0;
3167 int64_t max_length = -1;
3168
3169 fmt = NULL;
3170 output = NULL;
3171 for (;;) {
3172 int option_index = 0;
3173 static const struct option long_options[] = {
3174 {"help", no_argument, 0, 'h'},
3175 {"format", required_argument, 0, 'f'},
3176 {"output", required_argument, 0, OPTION_OUTPUT},
3177 {"object", required_argument, 0, OPTION_OBJECT},
3178 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3179 {"force-share", no_argument, 0, 'U'},
3180 {"start-offset", required_argument, 0, 's'},
3181 {"max-length", required_argument, 0, 'l'},
3182 {0, 0, 0, 0}
3183 };
3184 c = getopt_long(argc, argv, ":f:s:l:hU",
3185 long_options, &option_index);
3186 if (c == -1) {
3187 break;
3188 }
3189 switch (c) {
3190 case ':':
3191 missing_argument(argv[optind - 1]);
3192 break;
3193 case '?':
3194 unrecognized_option(argv[optind - 1]);
3195 break;
3196 case 'h':
3197 help();
3198 break;
3199 case 'f':
3200 fmt = optarg;
3201 break;
3202 case 'U':
3203 force_share = true;
3204 break;
3205 case OPTION_OUTPUT:
3206 output = optarg;
3207 break;
3208 case 's':
3209 start_offset = cvtnum("start offset", optarg);
3210 if (start_offset < 0) {
3211 return 1;
3212 }
3213 break;
3214 case 'l':
3215 max_length = cvtnum("max length", optarg);
3216 if (max_length < 0) {
3217 return 1;
3218 }
3219 break;
3220 case OPTION_OBJECT: {
3221 QemuOpts *opts;
3222 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3223 optarg, true);
3224 if (!opts) {
3225 return 1;
3226 }
3227 } break;
3228 case OPTION_IMAGE_OPTS:
3229 image_opts = true;
3230 break;
3231 }
3232 }
3233 if (optind != argc - 1) {
3234 error_exit("Expecting one image file name");
3235 }
3236 filename = argv[optind];
3237
3238 if (output && !strcmp(output, "json")) {
3239 output_format = OFORMAT_JSON;
3240 } else if (output && !strcmp(output, "human")) {
3241 output_format = OFORMAT_HUMAN;
3242 } else if (output) {
3243 error_report("--output must be used with human or json as argument.");
3244 return 1;
3245 }
3246
3247 if (qemu_opts_foreach(&qemu_object_opts,
3248 user_creatable_add_opts_foreach,
3249 qemu_img_object_print_help, &error_fatal)) {
3250 return 1;
3251 }
3252
3253 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3254 if (!blk) {
3255 return 1;
3256 }
3257 bs = blk_bs(blk);
3258
3259 if (output_format == OFORMAT_HUMAN) {
3260 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3261 } else if (output_format == OFORMAT_JSON) {
3262 putchar('[');
3263 }
3264
3265 length = blk_getlength(blk);
3266 if (length < 0) {
3267 error_report("Failed to get size for '%s'", filename);
3268 return 1;
3269 }
3270 if (max_length != -1) {
3271 length = MIN(start_offset + max_length, length);
3272 }
3273
3274 curr.start = start_offset;
3275 while (curr.start + curr.length < length) {
3276 int64_t offset = curr.start + curr.length;
3277 int64_t n = length - offset;
3278
3279 ret = get_block_status(bs, offset, n, &next);
3280 if (ret < 0) {
3281 error_report("Could not read file metadata: %s", strerror(-ret));
3282 goto out;
3283 }
3284
3285 if (entry_mergeable(&curr, &next)) {
3286 curr.length += next.length;
3287 continue;
3288 }
3289
3290 if (curr.length > 0) {
3291 ret = dump_map_entry(output_format, &curr, &next);
3292 if (ret < 0) {
3293 goto out;
3294 }
3295 }
3296 curr = next;
3297 }
3298
3299 ret = dump_map_entry(output_format, &curr, NULL);
3300 if (output_format == OFORMAT_JSON) {
3301 puts("]");
3302 }
3303
3304 out:
3305 blk_unref(blk);
3306 return ret < 0;
3307 }
3308
3309 #define SNAPSHOT_LIST 1
3310 #define SNAPSHOT_CREATE 2
3311 #define SNAPSHOT_APPLY 3
3312 #define SNAPSHOT_DELETE 4
3313
3314 static int img_snapshot(int argc, char **argv)
3315 {
3316 BlockBackend *blk;
3317 BlockDriverState *bs;
3318 QEMUSnapshotInfo sn;
3319 char *filename, *snapshot_name = NULL;
3320 int c, ret = 0, bdrv_oflags;
3321 int action = 0;
3322 qemu_timeval tv;
3323 bool quiet = false;
3324 Error *err = NULL;
3325 bool image_opts = false;
3326 bool force_share = false;
3327
3328 bdrv_oflags = BDRV_O_RDWR;
3329 /* Parse commandline parameters */
3330 for(;;) {
3331 static const struct option long_options[] = {
3332 {"help", no_argument, 0, 'h'},
3333 {"object", required_argument, 0, OPTION_OBJECT},
3334 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3335 {"force-share", no_argument, 0, 'U'},
3336 {0, 0, 0, 0}
3337 };
3338 c = getopt_long(argc, argv, ":la:c:d:hqU",
3339 long_options, NULL);
3340 if (c == -1) {
3341 break;
3342 }
3343 switch(c) {
3344 case ':':
3345 missing_argument(argv[optind - 1]);
3346 break;
3347 case '?':
3348 unrecognized_option(argv[optind - 1]);
3349 break;
3350 case 'h':
3351 help();
3352 return 0;
3353 case 'l':
3354 if (action) {
3355 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3356 return 0;
3357 }
3358 action = SNAPSHOT_LIST;
3359 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3360 break;
3361 case 'a':
3362 if (action) {
3363 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3364 return 0;
3365 }
3366 action = SNAPSHOT_APPLY;
3367 snapshot_name = optarg;
3368 break;
3369 case 'c':
3370 if (action) {
3371 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3372 return 0;
3373 }
3374 action = SNAPSHOT_CREATE;
3375 snapshot_name = optarg;
3376 break;
3377 case 'd':
3378 if (action) {
3379 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3380 return 0;
3381 }
3382 action = SNAPSHOT_DELETE;
3383 snapshot_name = optarg;
3384 break;
3385 case 'q':
3386 quiet = true;
3387 break;
3388 case 'U':
3389 force_share = true;
3390 break;
3391 case OPTION_OBJECT: {
3392 QemuOpts *opts;
3393 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3394 optarg, true);
3395 if (!opts) {
3396 return 1;
3397 }
3398 } break;
3399 case OPTION_IMAGE_OPTS:
3400 image_opts = true;
3401 break;
3402 }
3403 }
3404
3405 if (optind != argc - 1) {
3406 error_exit("Expecting one image file name");
3407 }
3408 filename = argv[optind++];
3409
3410 if (qemu_op