Merge tag 'artist-cursor-fix-final-pull-request' of https://github.com/hdeller/qemu...
[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 OPTION_SKIP_BROKEN = 277,
86 };
87
88 typedef enum OutputFormat {
89 OFORMAT_JSON,
90 OFORMAT_HUMAN,
91 } OutputFormat;
92
93 /* Default to cache=writeback as data integrity is not important for qemu-img */
94 #define BDRV_DEFAULT_CACHE "writeback"
95
96 static void format_print(void *opaque, const char *name)
97 {
98 printf(" %s", name);
99 }
100
101 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
102 {
103 va_list ap;
104
105 va_start(ap, fmt);
106 error_vreport(fmt, ap);
107 va_end(ap);
108
109 error_printf("Try 'qemu-img --help' for more information\n");
110 exit(EXIT_FAILURE);
111 }
112
113 static void QEMU_NORETURN missing_argument(const char *option)
114 {
115 error_exit("missing argument for option '%s'", option);
116 }
117
118 static void QEMU_NORETURN unrecognized_option(const char *option)
119 {
120 error_exit("unrecognized option '%s'", option);
121 }
122
123 /* Please keep in synch with docs/tools/qemu-img.rst */
124 static void QEMU_NORETURN help(void)
125 {
126 const char *help_msg =
127 QEMU_IMG_VERSION
128 "usage: qemu-img [standard options] command [command options]\n"
129 "QEMU disk image utility\n"
130 "\n"
131 " '-h', '--help' display this help and exit\n"
132 " '-V', '--version' output version information and exit\n"
133 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
134 " specify tracing options\n"
135 "\n"
136 "Command syntax:\n"
137 #define DEF(option, callback, arg_string) \
138 " " arg_string "\n"
139 #include "qemu-img-cmds.h"
140 #undef DEF
141 "\n"
142 "Command parameters:\n"
143 " 'filename' is a disk image filename\n"
144 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
145 " manual page for a description of the object properties. The most common\n"
146 " object type is a 'secret', which is used to supply passwords and/or\n"
147 " encryption keys.\n"
148 " 'fmt' is the disk image format. It is guessed automatically in most cases\n"
149 " 'cache' is the cache mode used to write the output disk image, the valid\n"
150 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
151 " 'directsync' and 'unsafe' (default for convert)\n"
152 " 'src_cache' is the cache mode used to read input disk images, the valid\n"
153 " options are the same as for the 'cache' option\n"
154 " 'size' is the disk image size in bytes. Optional suffixes\n"
155 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
156 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n"
157 " supported. 'b' is ignored.\n"
158 " 'output_filename' is the destination disk image filename\n"
159 " 'output_fmt' is the destination format\n"
160 " 'options' is a comma separated list of format specific options in a\n"
161 " name=value format. Use -o ? for an overview of the options supported by the\n"
162 " used format\n"
163 " 'snapshot_param' is param used for internal snapshot, format\n"
164 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
165 " '[ID_OR_NAME]'\n"
166 " '-c' indicates that target image must be compressed (qcow format only)\n"
167 " '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
168 " new backing file match exactly. The image doesn't need a working\n"
169 " backing file before rebasing in this case (useful for renaming the\n"
170 " backing file). For image creation, allow creating without attempting\n"
171 " to open the backing file.\n"
172 " '-h' with or without a command shows this help and lists the supported formats\n"
173 " '-p' show progress of command (only certain commands)\n"
174 " '-q' use Quiet mode - do not print any output (except errors)\n"
175 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
176 " contain only zeros for qemu-img to create a sparse image during\n"
177 " conversion. If the number of bytes is 0, the source will not be scanned for\n"
178 " unallocated or zero sectors, and the destination image will always be\n"
179 " fully allocated\n"
180 " '--output' takes the format in which the output must be done (human or json)\n"
181 " '-n' skips the target volume creation (useful if the volume is created\n"
182 " prior to running qemu-img)\n"
183 "\n"
184 "Parameters to bitmap subcommand:\n"
185 " 'bitmap' is the name of the bitmap to manipulate, through one or more\n"
186 " actions from '--add', '--remove', '--clear', '--enable', '--disable',\n"
187 " or '--merge source'\n"
188 " '-g granularity' sets the granularity for '--add' actions\n"
189 " '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n"
190 " bitmaps from an alternative file\n"
191 "\n"
192 "Parameters to check subcommand:\n"
193 " '-r' tries to repair any inconsistencies that are found during the check.\n"
194 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
195 " kinds of errors, with a higher risk of choosing the wrong fix or\n"
196 " hiding corruption that has already occurred.\n"
197 "\n"
198 "Parameters to convert subcommand:\n"
199 " '--bitmaps' copies all top-level persistent bitmaps to destination\n"
200 " '-m' specifies how many coroutines work in parallel during the convert\n"
201 " process (defaults to 8)\n"
202 " '-W' allow to write to the target out of order rather than sequential\n"
203 "\n"
204 "Parameters to snapshot subcommand:\n"
205 " 'snapshot' is the name of the snapshot to create, apply or delete\n"
206 " '-a' applies a snapshot (revert disk to saved state)\n"
207 " '-c' creates a snapshot\n"
208 " '-d' deletes a snapshot\n"
209 " '-l' lists all snapshots in the given image\n"
210 "\n"
211 "Parameters to compare subcommand:\n"
212 " '-f' first image format\n"
213 " '-F' second image format\n"
214 " '-s' run in Strict mode - fail on different image size or sector allocation\n"
215 "\n"
216 "Parameters to dd subcommand:\n"
217 " 'bs=BYTES' read and write up to BYTES bytes at a time "
218 "(default: 512)\n"
219 " 'count=N' copy only N input blocks\n"
220 " 'if=FILE' read from FILE\n"
221 " 'of=FILE' write to FILE\n"
222 " 'skip=N' skip N bs-sized blocks at the start of input\n";
223
224 printf("%s\nSupported formats:", help_msg);
225 bdrv_iterate_format(format_print, NULL, false);
226 printf("\n\n" QEMU_HELP_BOTTOM "\n");
227 exit(EXIT_SUCCESS);
228 }
229
230 /*
231 * Is @optarg safe for accumulate_options()?
232 * It is when multiple of them can be joined together separated by ','.
233 * To make that work, @optarg must not start with ',' (or else a
234 * separating ',' preceding it gets escaped), and it must not end with
235 * an odd number of ',' (or else a separating ',' following it gets
236 * escaped), or be empty (or else a separating ',' preceding it can
237 * escape a separating ',' following it).
238 *
239 */
240 static bool is_valid_option_list(const char *optarg)
241 {
242 size_t len = strlen(optarg);
243 size_t i;
244
245 if (!optarg[0] || optarg[0] == ',') {
246 return false;
247 }
248
249 for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
250 }
251 if ((len - i) % 2) {
252 return false;
253 }
254
255 return true;
256 }
257
258 static int accumulate_options(char **options, char *optarg)
259 {
260 char *new_options;
261
262 if (!is_valid_option_list(optarg)) {
263 error_report("Invalid option list: %s", optarg);
264 return -1;
265 }
266
267 if (!*options) {
268 *options = g_strdup(optarg);
269 } else {
270 new_options = g_strdup_printf("%s,%s", *options, optarg);
271 g_free(*options);
272 *options = new_options;
273 }
274 return 0;
275 }
276
277 static QemuOptsList qemu_source_opts = {
278 .name = "source",
279 .implied_opt_name = "file",
280 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
281 .desc = {
282 { }
283 },
284 };
285
286 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
287 {
288 int ret = 0;
289 if (!quiet) {
290 va_list args;
291 va_start(args, fmt);
292 ret = vprintf(fmt, args);
293 va_end(args);
294 }
295 return ret;
296 }
297
298
299 static int print_block_option_help(const char *filename, const char *fmt)
300 {
301 BlockDriver *drv, *proto_drv;
302 QemuOptsList *create_opts = NULL;
303 Error *local_err = NULL;
304
305 /* Find driver and parse its options */
306 drv = bdrv_find_format(fmt);
307 if (!drv) {
308 error_report("Unknown file format '%s'", fmt);
309 return 1;
310 }
311
312 if (!drv->create_opts) {
313 error_report("Format driver '%s' does not support image creation", fmt);
314 return 1;
315 }
316
317 create_opts = qemu_opts_append(create_opts, drv->create_opts);
318 if (filename) {
319 proto_drv = bdrv_find_protocol(filename, true, &local_err);
320 if (!proto_drv) {
321 error_report_err(local_err);
322 qemu_opts_free(create_opts);
323 return 1;
324 }
325 if (!proto_drv->create_opts) {
326 error_report("Protocol driver '%s' does not support image creation",
327 proto_drv->format_name);
328 qemu_opts_free(create_opts);
329 return 1;
330 }
331 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
332 }
333
334 if (filename) {
335 printf("Supported options:\n");
336 } else {
337 printf("Supported %s options:\n", fmt);
338 }
339 qemu_opts_print_help(create_opts, false);
340 qemu_opts_free(create_opts);
341
342 if (!filename) {
343 printf("\n"
344 "The protocol level may support further options.\n"
345 "Specify the target filename to include those options.\n");
346 }
347
348 return 0;
349 }
350
351
352 static BlockBackend *img_open_opts(const char *optstr,
353 QemuOpts *opts, int flags, bool writethrough,
354 bool quiet, bool force_share)
355 {
356 QDict *options;
357 Error *local_err = NULL;
358 BlockBackend *blk;
359 options = qemu_opts_to_qdict(opts, NULL);
360 if (force_share) {
361 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
362 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
363 error_report("--force-share/-U conflicts with image options");
364 qobject_unref(options);
365 return NULL;
366 }
367 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
368 }
369 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
370 if (!blk) {
371 error_reportf_err(local_err, "Could not open '%s': ", optstr);
372 return NULL;
373 }
374 blk_set_enable_write_cache(blk, !writethrough);
375
376 return blk;
377 }
378
379 static BlockBackend *img_open_file(const char *filename,
380 QDict *options,
381 const char *fmt, int flags,
382 bool writethrough, bool quiet,
383 bool force_share)
384 {
385 BlockBackend *blk;
386 Error *local_err = NULL;
387
388 if (!options) {
389 options = qdict_new();
390 }
391 if (fmt) {
392 qdict_put_str(options, "driver", fmt);
393 }
394
395 if (force_share) {
396 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
397 }
398 blk = blk_new_open(filename, NULL, options, flags, &local_err);
399 if (!blk) {
400 error_reportf_err(local_err, "Could not open '%s': ", filename);
401 return NULL;
402 }
403 blk_set_enable_write_cache(blk, !writethrough);
404
405 return blk;
406 }
407
408
409 static int img_add_key_secrets(void *opaque,
410 const char *name, const char *value,
411 Error **errp)
412 {
413 QDict *options = opaque;
414
415 if (g_str_has_suffix(name, "key-secret")) {
416 qdict_put_str(options, name, value);
417 }
418
419 return 0;
420 }
421
422
423 static BlockBackend *img_open(bool image_opts,
424 const char *filename,
425 const char *fmt, int flags, bool writethrough,
426 bool quiet, bool force_share)
427 {
428 BlockBackend *blk;
429 if (image_opts) {
430 QemuOpts *opts;
431 if (fmt) {
432 error_report("--image-opts and --format are mutually exclusive");
433 return NULL;
434 }
435 opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
436 filename, true);
437 if (!opts) {
438 return NULL;
439 }
440 blk = img_open_opts(filename, opts, flags, writethrough, quiet,
441 force_share);
442 } else {
443 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
444 force_share);
445 }
446 return blk;
447 }
448
449
450 static int add_old_style_options(const char *fmt, QemuOpts *opts,
451 const char *base_filename,
452 const char *base_fmt)
453 {
454 if (base_filename) {
455 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
456 NULL)) {
457 error_report("Backing file not supported for file format '%s'",
458 fmt);
459 return -1;
460 }
461 }
462 if (base_fmt) {
463 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
464 error_report("Backing file format not supported for file "
465 "format '%s'", fmt);
466 return -1;
467 }
468 }
469 return 0;
470 }
471
472 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
473 int64_t max)
474 {
475 int err;
476 uint64_t res;
477
478 err = qemu_strtosz(value, NULL, &res);
479 if (err < 0 && err != -ERANGE) {
480 error_report("Invalid %s specified. You may use "
481 "k, M, G, T, P or E suffixes for", name);
482 error_report("kilobytes, megabytes, gigabytes, terabytes, "
483 "petabytes and exabytes.");
484 return err;
485 }
486 if (err == -ERANGE || res > max || res < min) {
487 error_report("Invalid %s specified. Must be between %" PRId64
488 " and %" PRId64 ".", name, min, max);
489 return -ERANGE;
490 }
491 return res;
492 }
493
494 static int64_t cvtnum(const char *name, const char *value)
495 {
496 return cvtnum_full(name, value, 0, INT64_MAX);
497 }
498
499 static int img_create(int argc, char **argv)
500 {
501 int c;
502 uint64_t img_size = -1;
503 const char *fmt = "raw";
504 const char *base_fmt = NULL;
505 const char *filename;
506 const char *base_filename = NULL;
507 char *options = NULL;
508 Error *local_err = NULL;
509 bool quiet = false;
510 int flags = 0;
511
512 for(;;) {
513 static const struct option long_options[] = {
514 {"help", no_argument, 0, 'h'},
515 {"object", required_argument, 0, OPTION_OBJECT},
516 {0, 0, 0, 0}
517 };
518 c = getopt_long(argc, argv, ":F:b:f:ho:qu",
519 long_options, NULL);
520 if (c == -1) {
521 break;
522 }
523 switch(c) {
524 case ':':
525 missing_argument(argv[optind - 1]);
526 break;
527 case '?':
528 unrecognized_option(argv[optind - 1]);
529 break;
530 case 'h':
531 help();
532 break;
533 case 'F':
534 base_fmt = optarg;
535 break;
536 case 'b':
537 base_filename = optarg;
538 break;
539 case 'f':
540 fmt = optarg;
541 break;
542 case 'o':
543 if (accumulate_options(&options, optarg) < 0) {
544 goto fail;
545 }
546 break;
547 case 'q':
548 quiet = true;
549 break;
550 case 'u':
551 flags |= BDRV_O_NO_BACKING;
552 break;
553 case OPTION_OBJECT:
554 user_creatable_process_cmdline(optarg);
555 break;
556 }
557 }
558
559 /* Get the filename */
560 filename = (optind < argc) ? argv[optind] : NULL;
561 if (options && has_help_option(options)) {
562 g_free(options);
563 return print_block_option_help(filename, fmt);
564 }
565
566 if (optind >= argc) {
567 error_exit("Expecting image file name");
568 }
569 optind++;
570
571 /* Get image size, if specified */
572 if (optind < argc) {
573 int64_t sval;
574
575 sval = cvtnum("image size", argv[optind++]);
576 if (sval < 0) {
577 goto fail;
578 }
579 img_size = (uint64_t)sval;
580 }
581 if (optind != argc) {
582 error_exit("Unexpected argument: %s", argv[optind]);
583 }
584
585 bdrv_img_create(filename, fmt, base_filename, base_fmt,
586 options, img_size, flags, quiet, &local_err);
587 if (local_err) {
588 error_reportf_err(local_err, "%s: ", filename);
589 goto fail;
590 }
591
592 g_free(options);
593 return 0;
594
595 fail:
596 g_free(options);
597 return 1;
598 }
599
600 static void dump_json_image_check(ImageCheck *check, bool quiet)
601 {
602 GString *str;
603 QObject *obj;
604 Visitor *v = qobject_output_visitor_new(&obj);
605
606 visit_type_ImageCheck(v, NULL, &check, &error_abort);
607 visit_complete(v, &obj);
608 str = qobject_to_json_pretty(obj, true);
609 assert(str != NULL);
610 qprintf(quiet, "%s\n", str->str);
611 qobject_unref(obj);
612 visit_free(v);
613 g_string_free(str, true);
614 }
615
616 static void dump_human_image_check(ImageCheck *check, bool quiet)
617 {
618 if (!(check->corruptions || check->leaks || check->check_errors)) {
619 qprintf(quiet, "No errors were found on the image.\n");
620 } else {
621 if (check->corruptions) {
622 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
623 "Data may be corrupted, or further writes to the image "
624 "may corrupt it.\n",
625 check->corruptions);
626 }
627
628 if (check->leaks) {
629 qprintf(quiet,
630 "\n%" PRId64 " leaked clusters were found on the image.\n"
631 "This means waste of disk space, but no harm to data.\n",
632 check->leaks);
633 }
634
635 if (check->check_errors) {
636 qprintf(quiet,
637 "\n%" PRId64
638 " internal errors have occurred during the check.\n",
639 check->check_errors);
640 }
641 }
642
643 if (check->total_clusters != 0 && check->allocated_clusters != 0) {
644 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
645 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
646 check->allocated_clusters, check->total_clusters,
647 check->allocated_clusters * 100.0 / check->total_clusters,
648 check->fragmented_clusters * 100.0 / check->allocated_clusters,
649 check->compressed_clusters * 100.0 /
650 check->allocated_clusters);
651 }
652
653 if (check->image_end_offset) {
654 qprintf(quiet,
655 "Image end offset: %" PRId64 "\n", check->image_end_offset);
656 }
657 }
658
659 static int collect_image_check(BlockDriverState *bs,
660 ImageCheck *check,
661 const char *filename,
662 const char *fmt,
663 int fix)
664 {
665 int ret;
666 BdrvCheckResult result;
667
668 ret = bdrv_check(bs, &result, fix);
669 if (ret < 0) {
670 return ret;
671 }
672
673 check->filename = g_strdup(filename);
674 check->format = g_strdup(bdrv_get_format_name(bs));
675 check->check_errors = result.check_errors;
676 check->corruptions = result.corruptions;
677 check->has_corruptions = result.corruptions != 0;
678 check->leaks = result.leaks;
679 check->has_leaks = result.leaks != 0;
680 check->corruptions_fixed = result.corruptions_fixed;
681 check->has_corruptions_fixed = result.corruptions_fixed != 0;
682 check->leaks_fixed = result.leaks_fixed;
683 check->has_leaks_fixed = result.leaks_fixed != 0;
684 check->image_end_offset = result.image_end_offset;
685 check->has_image_end_offset = result.image_end_offset != 0;
686 check->total_clusters = result.bfi.total_clusters;
687 check->has_total_clusters = result.bfi.total_clusters != 0;
688 check->allocated_clusters = result.bfi.allocated_clusters;
689 check->has_allocated_clusters = result.bfi.allocated_clusters != 0;
690 check->fragmented_clusters = result.bfi.fragmented_clusters;
691 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0;
692 check->compressed_clusters = result.bfi.compressed_clusters;
693 check->has_compressed_clusters = result.bfi.compressed_clusters != 0;
694
695 return 0;
696 }
697
698 /*
699 * Checks an image for consistency. Exit codes:
700 *
701 * 0 - Check completed, image is good
702 * 1 - Check not completed because of internal errors
703 * 2 - Check completed, image is corrupted
704 * 3 - Check completed, image has leaked clusters, but is good otherwise
705 * 63 - Checks are not supported by the image format
706 */
707 static int img_check(int argc, char **argv)
708 {
709 int c, ret;
710 OutputFormat output_format = OFORMAT_HUMAN;
711 const char *filename, *fmt, *output, *cache;
712 BlockBackend *blk;
713 BlockDriverState *bs;
714 int fix = 0;
715 int flags = BDRV_O_CHECK;
716 bool writethrough;
717 ImageCheck *check;
718 bool quiet = false;
719 bool image_opts = false;
720 bool force_share = false;
721
722 fmt = NULL;
723 output = NULL;
724 cache = BDRV_DEFAULT_CACHE;
725
726 for(;;) {
727 int option_index = 0;
728 static const struct option long_options[] = {
729 {"help", no_argument, 0, 'h'},
730 {"format", required_argument, 0, 'f'},
731 {"repair", required_argument, 0, 'r'},
732 {"output", required_argument, 0, OPTION_OUTPUT},
733 {"object", required_argument, 0, OPTION_OBJECT},
734 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
735 {"force-share", no_argument, 0, 'U'},
736 {0, 0, 0, 0}
737 };
738 c = getopt_long(argc, argv, ":hf:r:T:qU",
739 long_options, &option_index);
740 if (c == -1) {
741 break;
742 }
743 switch(c) {
744 case ':':
745 missing_argument(argv[optind - 1]);
746 break;
747 case '?':
748 unrecognized_option(argv[optind - 1]);
749 break;
750 case 'h':
751 help();
752 break;
753 case 'f':
754 fmt = optarg;
755 break;
756 case 'r':
757 flags |= BDRV_O_RDWR;
758
759 if (!strcmp(optarg, "leaks")) {
760 fix = BDRV_FIX_LEAKS;
761 } else if (!strcmp(optarg, "all")) {
762 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
763 } else {
764 error_exit("Unknown option value for -r "
765 "(expecting 'leaks' or 'all'): %s", optarg);
766 }
767 break;
768 case OPTION_OUTPUT:
769 output = optarg;
770 break;
771 case 'T':
772 cache = optarg;
773 break;
774 case 'q':
775 quiet = true;
776 break;
777 case 'U':
778 force_share = true;
779 break;
780 case OPTION_OBJECT:
781 user_creatable_process_cmdline(optarg);
782 break;
783 case OPTION_IMAGE_OPTS:
784 image_opts = true;
785 break;
786 }
787 }
788 if (optind != argc - 1) {
789 error_exit("Expecting one image file name");
790 }
791 filename = argv[optind++];
792
793 if (output && !strcmp(output, "json")) {
794 output_format = OFORMAT_JSON;
795 } else if (output && !strcmp(output, "human")) {
796 output_format = OFORMAT_HUMAN;
797 } else if (output) {
798 error_report("--output must be used with human or json as argument.");
799 return 1;
800 }
801
802 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
803 if (ret < 0) {
804 error_report("Invalid source cache option: %s", cache);
805 return 1;
806 }
807
808 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
809 force_share);
810 if (!blk) {
811 return 1;
812 }
813 bs = blk_bs(blk);
814
815 check = g_new0(ImageCheck, 1);
816 ret = collect_image_check(bs, check, filename, fmt, fix);
817
818 if (ret == -ENOTSUP) {
819 error_report("This image format does not support checks");
820 ret = 63;
821 goto fail;
822 }
823
824 if (check->corruptions_fixed || check->leaks_fixed) {
825 int corruptions_fixed, leaks_fixed;
826 bool has_leaks_fixed, has_corruptions_fixed;
827
828 leaks_fixed = check->leaks_fixed;
829 has_leaks_fixed = check->has_leaks_fixed;
830 corruptions_fixed = check->corruptions_fixed;
831 has_corruptions_fixed = check->has_corruptions_fixed;
832
833 if (output_format == OFORMAT_HUMAN) {
834 qprintf(quiet,
835 "The following inconsistencies were found and repaired:\n\n"
836 " %" PRId64 " leaked clusters\n"
837 " %" PRId64 " corruptions\n\n"
838 "Double checking the fixed image now...\n",
839 check->leaks_fixed,
840 check->corruptions_fixed);
841 }
842
843 qapi_free_ImageCheck(check);
844 check = g_new0(ImageCheck, 1);
845 ret = collect_image_check(bs, check, filename, fmt, 0);
846
847 check->leaks_fixed = leaks_fixed;
848 check->has_leaks_fixed = has_leaks_fixed;
849 check->corruptions_fixed = corruptions_fixed;
850 check->has_corruptions_fixed = has_corruptions_fixed;
851 }
852
853 if (!ret) {
854 switch (output_format) {
855 case OFORMAT_HUMAN:
856 dump_human_image_check(check, quiet);
857 break;
858 case OFORMAT_JSON:
859 dump_json_image_check(check, quiet);
860 break;
861 }
862 }
863
864 if (ret || check->check_errors) {
865 if (ret) {
866 error_report("Check failed: %s", strerror(-ret));
867 } else {
868 error_report("Check failed");
869 }
870 ret = 1;
871 goto fail;
872 }
873
874 if (check->corruptions) {
875 ret = 2;
876 } else if (check->leaks) {
877 ret = 3;
878 } else {
879 ret = 0;
880 }
881
882 fail:
883 qapi_free_ImageCheck(check);
884 blk_unref(blk);
885 return ret;
886 }
887
888 typedef struct CommonBlockJobCBInfo {
889 BlockDriverState *bs;
890 Error **errp;
891 } CommonBlockJobCBInfo;
892
893 static void common_block_job_cb(void *opaque, int ret)
894 {
895 CommonBlockJobCBInfo *cbi = opaque;
896
897 if (ret < 0) {
898 error_setg_errno(cbi->errp, -ret, "Block job failed");
899 }
900 }
901
902 static void run_block_job(BlockJob *job, Error **errp)
903 {
904 uint64_t progress_current, progress_total;
905 AioContext *aio_context = blk_get_aio_context(job->blk);
906 int ret = 0;
907
908 aio_context_acquire(aio_context);
909 job_ref(&job->job);
910 do {
911 float progress = 0.0f;
912 aio_poll(aio_context, true);
913
914 progress_get_snapshot(&job->job.progress, &progress_current,
915 &progress_total);
916 if (progress_total) {
917 progress = (float)progress_current / progress_total * 100.f;
918 }
919 qemu_progress_print(progress, 0);
920 } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
921
922 if (!job_is_completed(&job->job)) {
923 ret = job_complete_sync(&job->job, errp);
924 } else {
925 ret = job->job.ret;
926 }
927 job_unref(&job->job);
928 aio_context_release(aio_context);
929
930 /* publish completion progress only when success */
931 if (!ret) {
932 qemu_progress_print(100.f, 0);
933 }
934 }
935
936 static int img_commit(int argc, char **argv)
937 {
938 int c, ret, flags;
939 const char *filename, *fmt, *cache, *base;
940 BlockBackend *blk;
941 BlockDriverState *bs, *base_bs;
942 BlockJob *job;
943 bool progress = false, quiet = false, drop = false;
944 bool writethrough;
945 Error *local_err = NULL;
946 CommonBlockJobCBInfo cbi;
947 bool image_opts = false;
948 AioContext *aio_context;
949 int64_t rate_limit = 0;
950
951 fmt = NULL;
952 cache = BDRV_DEFAULT_CACHE;
953 base = NULL;
954 for(;;) {
955 static const struct option long_options[] = {
956 {"help", no_argument, 0, 'h'},
957 {"object", required_argument, 0, OPTION_OBJECT},
958 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
959 {0, 0, 0, 0}
960 };
961 c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
962 long_options, NULL);
963 if (c == -1) {
964 break;
965 }
966 switch(c) {
967 case ':':
968 missing_argument(argv[optind - 1]);
969 break;
970 case '?':
971 unrecognized_option(argv[optind - 1]);
972 break;
973 case 'h':
974 help();
975 break;
976 case 'f':
977 fmt = optarg;
978 break;
979 case 't':
980 cache = optarg;
981 break;
982 case 'b':
983 base = optarg;
984 /* -b implies -d */
985 drop = true;
986 break;
987 case 'd':
988 drop = true;
989 break;
990 case 'p':
991 progress = true;
992 break;
993 case 'q':
994 quiet = true;
995 break;
996 case 'r':
997 rate_limit = cvtnum("rate limit", optarg);
998 if (rate_limit < 0) {
999 return 1;
1000 }
1001 break;
1002 case OPTION_OBJECT:
1003 user_creatable_process_cmdline(optarg);
1004 break;
1005 case OPTION_IMAGE_OPTS:
1006 image_opts = true;
1007 break;
1008 }
1009 }
1010
1011 /* Progress is not shown in Quiet mode */
1012 if (quiet) {
1013 progress = false;
1014 }
1015
1016 if (optind != argc - 1) {
1017 error_exit("Expecting one image file name");
1018 }
1019 filename = argv[optind++];
1020
1021 flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1022 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1023 if (ret < 0) {
1024 error_report("Invalid cache option: %s", cache);
1025 return 1;
1026 }
1027
1028 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1029 false);
1030 if (!blk) {
1031 return 1;
1032 }
1033 bs = blk_bs(blk);
1034
1035 qemu_progress_init(progress, 1.f);
1036 qemu_progress_print(0.f, 100);
1037
1038 if (base) {
1039 base_bs = bdrv_find_backing_image(bs, base);
1040 if (!base_bs) {
1041 error_setg(&local_err,
1042 "Did not find '%s' in the backing chain of '%s'",
1043 base, filename);
1044 goto done;
1045 }
1046 } else {
1047 /* This is different from QMP, which by default uses the deepest file in
1048 * the backing chain (i.e., the very base); however, the traditional
1049 * behavior of qemu-img commit is using the immediate backing file. */
1050 base_bs = bdrv_backing_chain_next(bs);
1051 if (!base_bs) {
1052 error_setg(&local_err, "Image does not have a backing file");
1053 goto done;
1054 }
1055 }
1056
1057 cbi = (CommonBlockJobCBInfo){
1058 .errp = &local_err,
1059 .bs = bs,
1060 };
1061
1062 aio_context = bdrv_get_aio_context(bs);
1063 aio_context_acquire(aio_context);
1064 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1065 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1066 &cbi, false, &local_err);
1067 aio_context_release(aio_context);
1068 if (local_err) {
1069 goto done;
1070 }
1071
1072 /* When the block job completes, the BlockBackend reference will point to
1073 * the old backing file. In order to avoid that the top image is already
1074 * deleted, so we can still empty it afterwards, increment the reference
1075 * counter here preemptively. */
1076 if (!drop) {
1077 bdrv_ref(bs);
1078 }
1079
1080 job = block_job_get("commit");
1081 assert(job);
1082 run_block_job(job, &local_err);
1083 if (local_err) {
1084 goto unref_backing;
1085 }
1086
1087 if (!drop) {
1088 BlockBackend *old_backing_blk;
1089
1090 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1091 &local_err);
1092 if (!old_backing_blk) {
1093 goto unref_backing;
1094 }
1095 ret = blk_make_empty(old_backing_blk, &local_err);
1096 blk_unref(old_backing_blk);
1097 if (ret == -ENOTSUP) {
1098 error_free(local_err);
1099 local_err = NULL;
1100 } else if (ret < 0) {
1101 goto unref_backing;
1102 }
1103 }
1104
1105 unref_backing:
1106 if (!drop) {
1107 bdrv_unref(bs);
1108 }
1109
1110 done:
1111 qemu_progress_end();
1112
1113 blk_unref(blk);
1114
1115 if (local_err) {
1116 error_report_err(local_err);
1117 return 1;
1118 }
1119
1120 qprintf(quiet, "Image committed.\n");
1121 return 0;
1122 }
1123
1124 /*
1125 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1126 * of the first sector boundary within buf where the sector contains a
1127 * non-zero byte. This function is robust to a buffer that is not
1128 * sector-aligned.
1129 */
1130 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1131 {
1132 int64_t i;
1133 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1134
1135 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1136 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1137 return i;
1138 }
1139 }
1140 if (i < n && !buffer_is_zero(buf + i, n - end)) {
1141 return i;
1142 }
1143 return -1;
1144 }
1145
1146 /*
1147 * Returns true iff the first sector pointed to by 'buf' contains at least
1148 * a non-NUL byte.
1149 *
1150 * 'pnum' is set to the number of sectors (including and immediately following
1151 * the first one) that are known to be in the same allocated/unallocated state.
1152 * The function will try to align the end offset to alignment boundaries so
1153 * that the request will at least end aligned and consecutive requests will
1154 * also start at an aligned offset.
1155 */
1156 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1157 int64_t sector_num, int alignment)
1158 {
1159 bool is_zero;
1160 int i, tail;
1161
1162 if (n <= 0) {
1163 *pnum = 0;
1164 return 0;
1165 }
1166 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1167 for(i = 1; i < n; i++) {
1168 buf += BDRV_SECTOR_SIZE;
1169 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1170 break;
1171 }
1172 }
1173
1174 tail = (sector_num + i) & (alignment - 1);
1175 if (tail) {
1176 if (is_zero && i <= tail) {
1177 /* treat unallocated areas which only consist
1178 * of a small tail as allocated. */
1179 is_zero = false;
1180 }
1181 if (!is_zero) {
1182 /* align up end offset of allocated areas. */
1183 i += alignment - tail;
1184 i = MIN(i, n);
1185 } else {
1186 /* align down end offset of zero areas. */
1187 i -= tail;
1188 }
1189 }
1190 *pnum = i;
1191 return !is_zero;
1192 }
1193
1194 /*
1195 * Like is_allocated_sectors, but if the buffer starts with a used sector,
1196 * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1197 * breaking up write requests for only small sparse areas.
1198 */
1199 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1200 int min, int64_t sector_num, int alignment)
1201 {
1202 int ret;
1203 int num_checked, num_used;
1204
1205 if (n < min) {
1206 min = n;
1207 }
1208
1209 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1210 if (!ret) {
1211 return ret;
1212 }
1213
1214 num_used = *pnum;
1215 buf += BDRV_SECTOR_SIZE * *pnum;
1216 n -= *pnum;
1217 sector_num += *pnum;
1218 num_checked = num_used;
1219
1220 while (n > 0) {
1221 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1222
1223 buf += BDRV_SECTOR_SIZE * *pnum;
1224 n -= *pnum;
1225 sector_num += *pnum;
1226 num_checked += *pnum;
1227 if (ret) {
1228 num_used = num_checked;
1229 } else if (*pnum >= min) {
1230 break;
1231 }
1232 }
1233
1234 *pnum = num_used;
1235 return 1;
1236 }
1237
1238 /*
1239 * Compares two buffers sector by sector. Returns 0 if the first
1240 * sector of each buffer matches, non-zero otherwise.
1241 *
1242 * pnum is set to the sector-aligned size of the buffer prefix that
1243 * has the same matching status as the first sector.
1244 */
1245 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1246 int64_t bytes, int64_t *pnum)
1247 {
1248 bool res;
1249 int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1250
1251 assert(bytes > 0);
1252
1253 res = !!memcmp(buf1, buf2, i);
1254 while (i < bytes) {
1255 int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1256
1257 if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1258 break;
1259 }
1260 i += len;
1261 }
1262
1263 *pnum = i;
1264 return res;
1265 }
1266
1267 #define IO_BUF_SIZE (2 * MiB)
1268
1269 /*
1270 * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1271 *
1272 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1273 * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1274 * failure), and 4 on error (the exit status for read errors), after emitting
1275 * an error message.
1276 *
1277 * @param blk: BlockBackend for the image
1278 * @param offset: Starting offset to check
1279 * @param bytes: Number of bytes to check
1280 * @param filename: Name of disk file we are checking (logging purpose)
1281 * @param buffer: Allocated buffer for storing read data
1282 * @param quiet: Flag for quiet mode
1283 */
1284 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1285 int64_t bytes, const char *filename,
1286 uint8_t *buffer, bool quiet)
1287 {
1288 int ret = 0;
1289 int64_t idx;
1290
1291 ret = blk_pread(blk, offset, buffer, bytes);
1292 if (ret < 0) {
1293 error_report("Error while reading offset %" PRId64 " of %s: %s",
1294 offset, filename, strerror(-ret));
1295 return 4;
1296 }
1297 idx = find_nonzero(buffer, bytes);
1298 if (idx >= 0) {
1299 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1300 offset + idx);
1301 return 1;
1302 }
1303
1304 return 0;
1305 }
1306
1307 /*
1308 * Compares two images. Exit codes:
1309 *
1310 * 0 - Images are identical or the requested help was printed
1311 * 1 - Images differ
1312 * >1 - Error occurred
1313 */
1314 static int img_compare(int argc, char **argv)
1315 {
1316 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1317 BlockBackend *blk1, *blk2;
1318 BlockDriverState *bs1, *bs2;
1319 int64_t total_size1, total_size2;
1320 uint8_t *buf1 = NULL, *buf2 = NULL;
1321 int64_t pnum1, pnum2;
1322 int allocated1, allocated2;
1323 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1324 bool progress = false, quiet = false, strict = false;
1325 int flags;
1326 bool writethrough;
1327 int64_t total_size;
1328 int64_t offset = 0;
1329 int64_t chunk;
1330 int c;
1331 uint64_t progress_base;
1332 bool image_opts = false;
1333 bool force_share = false;
1334
1335 cache = BDRV_DEFAULT_CACHE;
1336 for (;;) {
1337 static const struct option long_options[] = {
1338 {"help", no_argument, 0, 'h'},
1339 {"object", required_argument, 0, OPTION_OBJECT},
1340 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1341 {"force-share", no_argument, 0, 'U'},
1342 {0, 0, 0, 0}
1343 };
1344 c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1345 long_options, NULL);
1346 if (c == -1) {
1347 break;
1348 }
1349 switch (c) {
1350 case ':':
1351 missing_argument(argv[optind - 1]);
1352 break;
1353 case '?':
1354 unrecognized_option(argv[optind - 1]);
1355 break;
1356 case 'h':
1357 help();
1358 break;
1359 case 'f':
1360 fmt1 = optarg;
1361 break;
1362 case 'F':
1363 fmt2 = optarg;
1364 break;
1365 case 'T':
1366 cache = optarg;
1367 break;
1368 case 'p':
1369 progress = true;
1370 break;
1371 case 'q':
1372 quiet = true;
1373 break;
1374 case 's':
1375 strict = true;
1376 break;
1377 case 'U':
1378 force_share = true;
1379 break;
1380 case OPTION_OBJECT:
1381 {
1382 Error *local_err = NULL;
1383
1384 if (!user_creatable_add_from_str(optarg, &local_err)) {
1385 if (local_err) {
1386 error_report_err(local_err);
1387 exit(2);
1388 } else {
1389 /* Help was printed */
1390 exit(EXIT_SUCCESS);
1391 }
1392 }
1393 break;
1394 }
1395 case OPTION_IMAGE_OPTS:
1396 image_opts = true;
1397 break;
1398 }
1399 }
1400
1401 /* Progress is not shown in Quiet mode */
1402 if (quiet) {
1403 progress = false;
1404 }
1405
1406
1407 if (optind != argc - 2) {
1408 error_exit("Expecting two image file names");
1409 }
1410 filename1 = argv[optind++];
1411 filename2 = argv[optind++];
1412
1413 /* Initialize before goto out */
1414 qemu_progress_init(progress, 2.0);
1415
1416 flags = 0;
1417 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1418 if (ret < 0) {
1419 error_report("Invalid source cache option: %s", cache);
1420 ret = 2;
1421 goto out3;
1422 }
1423
1424 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1425 force_share);
1426 if (!blk1) {
1427 ret = 2;
1428 goto out3;
1429 }
1430
1431 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1432 force_share);
1433 if (!blk2) {
1434 ret = 2;
1435 goto out2;
1436 }
1437 bs1 = blk_bs(blk1);
1438 bs2 = blk_bs(blk2);
1439
1440 buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1441 buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1442 total_size1 = blk_getlength(blk1);
1443 if (total_size1 < 0) {
1444 error_report("Can't get size of %s: %s",
1445 filename1, strerror(-total_size1));
1446 ret = 4;
1447 goto out;
1448 }
1449 total_size2 = blk_getlength(blk2);
1450 if (total_size2 < 0) {
1451 error_report("Can't get size of %s: %s",
1452 filename2, strerror(-total_size2));
1453 ret = 4;
1454 goto out;
1455 }
1456 total_size = MIN(total_size1, total_size2);
1457 progress_base = MAX(total_size1, total_size2);
1458
1459 qemu_progress_print(0, 100);
1460
1461 if (strict && total_size1 != total_size2) {
1462 ret = 1;
1463 qprintf(quiet, "Strict mode: Image size mismatch!\n");
1464 goto out;
1465 }
1466
1467 while (offset < total_size) {
1468 int status1, status2;
1469
1470 status1 = bdrv_block_status_above(bs1, NULL, offset,
1471 total_size1 - offset, &pnum1, NULL,
1472 NULL);
1473 if (status1 < 0) {
1474 ret = 3;
1475 error_report("Sector allocation test failed for %s", filename1);
1476 goto out;
1477 }
1478 allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1479
1480 status2 = bdrv_block_status_above(bs2, NULL, offset,
1481 total_size2 - offset, &pnum2, NULL,
1482 NULL);
1483 if (status2 < 0) {
1484 ret = 3;
1485 error_report("Sector allocation test failed for %s", filename2);
1486 goto out;
1487 }
1488 allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1489
1490 assert(pnum1 && pnum2);
1491 chunk = MIN(pnum1, pnum2);
1492
1493 if (strict) {
1494 if (status1 != status2) {
1495 ret = 1;
1496 qprintf(quiet, "Strict mode: Offset %" PRId64
1497 " block status mismatch!\n", offset);
1498 goto out;
1499 }
1500 }
1501 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1502 /* nothing to do */
1503 } else if (allocated1 == allocated2) {
1504 if (allocated1) {
1505 int64_t pnum;
1506
1507 chunk = MIN(chunk, IO_BUF_SIZE);
1508 ret = blk_pread(blk1, offset, buf1, chunk);
1509 if (ret < 0) {
1510 error_report("Error while reading offset %" PRId64
1511 " of %s: %s",
1512 offset, filename1, strerror(-ret));
1513 ret = 4;
1514 goto out;
1515 }
1516 ret = blk_pread(blk2, offset, buf2, chunk);
1517 if (ret < 0) {
1518 error_report("Error while reading offset %" PRId64
1519 " of %s: %s",
1520 offset, filename2, strerror(-ret));
1521 ret = 4;
1522 goto out;
1523 }
1524 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1525 if (ret || pnum != chunk) {
1526 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1527 offset + (ret ? 0 : pnum));
1528 ret = 1;
1529 goto out;
1530 }
1531 }
1532 } else {
1533 chunk = MIN(chunk, IO_BUF_SIZE);
1534 if (allocated1) {
1535 ret = check_empty_sectors(blk1, offset, chunk,
1536 filename1, buf1, quiet);
1537 } else {
1538 ret = check_empty_sectors(blk2, offset, chunk,
1539 filename2, buf1, quiet);
1540 }
1541 if (ret) {
1542 goto out;
1543 }
1544 }
1545 offset += chunk;
1546 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1547 }
1548
1549 if (total_size1 != total_size2) {
1550 BlockBackend *blk_over;
1551 const char *filename_over;
1552
1553 qprintf(quiet, "Warning: Image size mismatch!\n");
1554 if (total_size1 > total_size2) {
1555 blk_over = blk1;
1556 filename_over = filename1;
1557 } else {
1558 blk_over = blk2;
1559 filename_over = filename2;
1560 }
1561
1562 while (offset < progress_base) {
1563 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1564 progress_base - offset, &chunk,
1565 NULL, NULL);
1566 if (ret < 0) {
1567 ret = 3;
1568 error_report("Sector allocation test failed for %s",
1569 filename_over);
1570 goto out;
1571
1572 }
1573 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1574 chunk = MIN(chunk, IO_BUF_SIZE);
1575 ret = check_empty_sectors(blk_over, offset, chunk,
1576 filename_over, buf1, quiet);
1577 if (ret) {
1578 goto out;
1579 }
1580 }
1581 offset += chunk;
1582 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1583 }
1584 }
1585
1586 qprintf(quiet, "Images are identical.\n");
1587 ret = 0;
1588
1589 out:
1590 qemu_vfree(buf1);
1591 qemu_vfree(buf2);
1592 blk_unref(blk2);
1593 out2:
1594 blk_unref(blk1);
1595 out3:
1596 qemu_progress_end();
1597 return ret;
1598 }
1599
1600 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1601 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1602 const char *src_node, const char *src_name,
1603 Error **errp)
1604 {
1605 BlockDirtyBitmapMergeSource *merge_src;
1606 BlockDirtyBitmapMergeSourceList *list = NULL;
1607
1608 merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1609 merge_src->type = QTYPE_QDICT;
1610 merge_src->u.external.node = g_strdup(src_node);
1611 merge_src->u.external.name = g_strdup(src_name);
1612 QAPI_LIST_PREPEND(list, merge_src);
1613 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1614 qapi_free_BlockDirtyBitmapMergeSourceList(list);
1615 }
1616
1617 enum ImgConvertBlockStatus {
1618 BLK_DATA,
1619 BLK_ZERO,
1620 BLK_BACKING_FILE,
1621 };
1622
1623 #define MAX_COROUTINES 16
1624 #define CONVERT_THROTTLE_GROUP "img_convert"
1625
1626 typedef struct ImgConvertState {
1627 BlockBackend **src;
1628 int64_t *src_sectors;
1629 int *src_alignment;
1630 int src_num;
1631 int64_t total_sectors;
1632 int64_t allocated_sectors;
1633 int64_t allocated_done;
1634 int64_t sector_num;
1635 int64_t wr_offs;
1636 enum ImgConvertBlockStatus status;
1637 int64_t sector_next_status;
1638 BlockBackend *target;
1639 bool has_zero_init;
1640 bool compressed;
1641 bool target_is_new;
1642 bool target_has_backing;
1643 int64_t target_backing_sectors; /* negative if unknown */
1644 bool wr_in_order;
1645 bool copy_range;
1646 bool salvage;
1647 bool quiet;
1648 int min_sparse;
1649 int alignment;
1650 size_t cluster_sectors;
1651 size_t buf_sectors;
1652 long num_coroutines;
1653 int running_coroutines;
1654 Coroutine *co[MAX_COROUTINES];
1655 int64_t wait_sector_num[MAX_COROUTINES];
1656 CoMutex lock;
1657 int ret;
1658 } ImgConvertState;
1659
1660 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1661 int *src_cur, int64_t *src_cur_offset)
1662 {
1663 *src_cur = 0;
1664 *src_cur_offset = 0;
1665 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1666 *src_cur_offset += s->src_sectors[*src_cur];
1667 (*src_cur)++;
1668 assert(*src_cur < s->src_num);
1669 }
1670 }
1671
1672 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1673 {
1674 int64_t src_cur_offset;
1675 int ret, n, src_cur;
1676 bool post_backing_zero = false;
1677
1678 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1679
1680 assert(s->total_sectors > sector_num);
1681 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1682
1683 if (s->target_backing_sectors >= 0) {
1684 if (sector_num >= s->target_backing_sectors) {
1685 post_backing_zero = true;
1686 } else if (sector_num + n > s->target_backing_sectors) {
1687 /* Split requests around target_backing_sectors (because
1688 * starting from there, zeros are handled differently) */
1689 n = s->target_backing_sectors - sector_num;
1690 }
1691 }
1692
1693 if (s->sector_next_status <= sector_num) {
1694 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1695 int64_t count;
1696 int tail;
1697 BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1698 BlockDriverState *base;
1699
1700 if (s->target_has_backing) {
1701 base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1702 } else {
1703 base = NULL;
1704 }
1705
1706 do {
1707 count = n * BDRV_SECTOR_SIZE;
1708
1709 ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1710 NULL, NULL);
1711
1712 if (ret < 0) {
1713 if (s->salvage) {
1714 if (n == 1) {
1715 if (!s->quiet) {
1716 warn_report("error while reading block status at "
1717 "offset %" PRIu64 ": %s", offset,
1718 strerror(-ret));
1719 }
1720 /* Just try to read the data, then */
1721 ret = BDRV_BLOCK_DATA;
1722 count = BDRV_SECTOR_SIZE;
1723 } else {
1724 /* Retry on a shorter range */
1725 n = DIV_ROUND_UP(n, 4);
1726 }
1727 } else {
1728 error_report("error while reading block status at offset "
1729 "%" PRIu64 ": %s", offset, strerror(-ret));
1730 return ret;
1731 }
1732 }
1733 } while (ret < 0);
1734
1735 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1736
1737 /*
1738 * Avoid that s->sector_next_status becomes unaligned to the source
1739 * request alignment and/or cluster size to avoid unnecessary read
1740 * cycles.
1741 */
1742 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1743 if (n > tail) {
1744 n -= tail;
1745 }
1746
1747 if (ret & BDRV_BLOCK_ZERO) {
1748 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1749 } else if (ret & BDRV_BLOCK_DATA) {
1750 s->status = BLK_DATA;
1751 } else {
1752 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1753 }
1754
1755 s->sector_next_status = sector_num + n;
1756 }
1757
1758 n = MIN(n, s->sector_next_status - sector_num);
1759 if (s->status == BLK_DATA) {
1760 n = MIN(n, s->buf_sectors);
1761 }
1762
1763 /* We need to write complete clusters for compressed images, so if an
1764 * unallocated area is shorter than that, we must consider the whole
1765 * cluster allocated. */
1766 if (s->compressed) {
1767 if (n < s->cluster_sectors) {
1768 n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1769 s->status = BLK_DATA;
1770 } else {
1771 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1772 }
1773 }
1774
1775 return n;
1776 }
1777
1778 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1779 int nb_sectors, uint8_t *buf)
1780 {
1781 uint64_t single_read_until = 0;
1782 int n, ret;
1783
1784 assert(nb_sectors <= s->buf_sectors);
1785 while (nb_sectors > 0) {
1786 BlockBackend *blk;
1787 int src_cur;
1788 int64_t bs_sectors, src_cur_offset;
1789 uint64_t offset;
1790
1791 /* In the case of compression with multiple source files, we can get a
1792 * nb_sectors that spreads into the next part. So we must be able to
1793 * read across multiple BDSes for one convert_read() call. */
1794 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1795 blk = s->src[src_cur];
1796 bs_sectors = s->src_sectors[src_cur];
1797
1798 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1799
1800 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1801 if (single_read_until > offset) {
1802 n = 1;
1803 }
1804
1805 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1806 if (ret < 0) {
1807 if (s->salvage) {
1808 if (n > 1) {
1809 single_read_until = offset + (n << BDRV_SECTOR_BITS);
1810 continue;
1811 } else {
1812 if (!s->quiet) {
1813 warn_report("error while reading offset %" PRIu64
1814 ": %s", offset, strerror(-ret));
1815 }
1816 memset(buf, 0, BDRV_SECTOR_SIZE);
1817 }
1818 } else {
1819 return ret;
1820 }
1821 }
1822
1823 sector_num += n;
1824 nb_sectors -= n;
1825 buf += n * BDRV_SECTOR_SIZE;
1826 }
1827
1828 return 0;
1829 }
1830
1831
1832 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1833 int nb_sectors, uint8_t *buf,
1834 enum ImgConvertBlockStatus status)
1835 {
1836 int ret;
1837
1838 while (nb_sectors > 0) {
1839 int n = nb_sectors;
1840 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1841
1842 switch (status) {
1843 case BLK_BACKING_FILE:
1844 /* If we have a backing file, leave clusters unallocated that are
1845 * unallocated in the source image, so that the backing file is
1846 * visible at the respective offset. */
1847 assert(s->target_has_backing);
1848 break;
1849
1850 case BLK_DATA:
1851 /* If we're told to keep the target fully allocated (-S 0) or there
1852 * is real non-zero data, we must write it. Otherwise we can treat
1853 * it as zero sectors.
1854 * Compressed clusters need to be written as a whole, so in that
1855 * case we can only save the write if the buffer is completely
1856 * zeroed. */
1857 if (!s->min_sparse ||
1858 (!s->compressed &&
1859 is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1860 sector_num, s->alignment)) ||
1861 (s->compressed &&
1862 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1863 {
1864 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1865 n << BDRV_SECTOR_BITS, buf, flags);
1866 if (ret < 0) {
1867 return ret;
1868 }
1869 break;
1870 }
1871 /* fall-through */
1872
1873 case BLK_ZERO:
1874 if (s->has_zero_init) {
1875 assert(!s->target_has_backing);
1876 break;
1877 }
1878 ret = blk_co_pwrite_zeroes(s->target,
1879 sector_num << BDRV_SECTOR_BITS,
1880 n << BDRV_SECTOR_BITS,
1881 BDRV_REQ_MAY_UNMAP);
1882 if (ret < 0) {
1883 return ret;
1884 }
1885 break;
1886 }
1887
1888 sector_num += n;
1889 nb_sectors -= n;
1890 buf += n * BDRV_SECTOR_SIZE;
1891 }
1892
1893 return 0;
1894 }
1895
1896 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1897 int nb_sectors)
1898 {
1899 int n, ret;
1900
1901 while (nb_sectors > 0) {
1902 BlockBackend *blk;
1903 int src_cur;
1904 int64_t bs_sectors, src_cur_offset;
1905 int64_t offset;
1906
1907 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1908 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1909 blk = s->src[src_cur];
1910 bs_sectors = s->src_sectors[src_cur];
1911
1912 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1913
1914 ret = blk_co_copy_range(blk, offset, s->target,
1915 sector_num << BDRV_SECTOR_BITS,
1916 n << BDRV_SECTOR_BITS, 0, 0);
1917 if (ret < 0) {
1918 return ret;
1919 }
1920
1921 sector_num += n;
1922 nb_sectors -= n;
1923 }
1924 return 0;
1925 }
1926
1927 static void coroutine_fn convert_co_do_copy(void *opaque)
1928 {
1929 ImgConvertState *s = opaque;
1930 uint8_t *buf = NULL;
1931 int ret, i;
1932 int index = -1;
1933
1934 for (i = 0; i < s->num_coroutines; i++) {
1935 if (s->co[i] == qemu_coroutine_self()) {
1936 index = i;
1937 break;
1938 }
1939 }
1940 assert(index >= 0);
1941
1942 s->running_coroutines++;
1943 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1944
1945 while (1) {
1946 int n;
1947 int64_t sector_num;
1948 enum ImgConvertBlockStatus status;
1949 bool copy_range;
1950
1951 qemu_co_mutex_lock(&s->lock);
1952 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1953 qemu_co_mutex_unlock(&s->lock);
1954 break;
1955 }
1956 n = convert_iteration_sectors(s, s->sector_num);
1957 if (n < 0) {
1958 qemu_co_mutex_unlock(&s->lock);
1959 s->ret = n;
1960 break;
1961 }
1962 /* save current sector and allocation status to local variables */
1963 sector_num = s->sector_num;
1964 status = s->status;
1965 if (!s->min_sparse && s->status == BLK_ZERO) {
1966 n = MIN(n, s->buf_sectors);
1967 }
1968 /* increment global sector counter so that other coroutines can
1969 * already continue reading beyond this request */
1970 s->sector_num += n;
1971 qemu_co_mutex_unlock(&s->lock);
1972
1973 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1974 s->allocated_done += n;
1975 qemu_progress_print(100.0 * s->allocated_done /
1976 s->allocated_sectors, 0);
1977 }
1978
1979 retry:
1980 copy_range = s->copy_range && s->status == BLK_DATA;
1981 if (status == BLK_DATA && !copy_range) {
1982 ret = convert_co_read(s, sector_num, n, buf);
1983 if (ret < 0) {
1984 error_report("error while reading at byte %lld: %s",
1985 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
1986 s->ret = ret;
1987 }
1988 } else if (!s->min_sparse && status == BLK_ZERO) {
1989 status = BLK_DATA;
1990 memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1991 }
1992
1993 if (s->wr_in_order) {
1994 /* keep writes in order */
1995 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1996 s->wait_sector_num[index] = sector_num;
1997 qemu_coroutine_yield();
1998 }
1999 s->wait_sector_num[index] = -1;
2000 }
2001
2002 if (s->ret == -EINPROGRESS) {
2003 if (copy_range) {
2004 ret = convert_co_copy_range(s, sector_num, n);
2005 if (ret) {
2006 s->copy_range = false;
2007 goto retry;
2008 }
2009 } else {
2010 ret = convert_co_write(s, sector_num, n, buf, status);
2011 }
2012 if (ret < 0) {
2013 error_report("error while writing at byte %lld: %s",
2014 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2015 s->ret = ret;
2016 }
2017 }
2018
2019 if (s->wr_in_order) {
2020 /* reenter the coroutine that might have waited
2021 * for this write to complete */
2022 s->wr_offs = sector_num + n;
2023 for (i = 0; i < s->num_coroutines; i++) {
2024 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2025 /*
2026 * A -> B -> A cannot occur because A has
2027 * s->wait_sector_num[i] == -1 during A -> B. Therefore
2028 * B will never enter A during this time window.
2029 */
2030 qemu_coroutine_enter(s->co[i]);
2031 break;
2032 }
2033 }
2034 }
2035 }
2036
2037 qemu_vfree(buf);
2038 s->co[index] = NULL;
2039 s->running_coroutines--;
2040 if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2041 /* the convert job finished successfully */
2042 s->ret = 0;
2043 }
2044 }
2045
2046 static int convert_do_copy(ImgConvertState *s)
2047 {
2048 int ret, i, n;
2049 int64_t sector_num = 0;
2050
2051 /* Check whether we have zero initialisation or can get it efficiently */
2052 if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2053 !s->target_has_backing) {
2054 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2055 }
2056
2057 /* Allocate buffer for copied data. For compressed images, only one cluster
2058 * can be copied at a time. */
2059 if (s->compressed) {
2060 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2061 error_report("invalid cluster size");
2062 return -EINVAL;
2063 }
2064 s->buf_sectors = s->cluster_sectors;
2065 }
2066
2067 while (sector_num < s->total_sectors) {
2068 n = convert_iteration_sectors(s, sector_num);
2069 if (n < 0) {
2070 return n;
2071 }
2072 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2073 {
2074 s->allocated_sectors += n;
2075 }
2076 sector_num += n;
2077 }
2078
2079 /* Do the copy */
2080 s->sector_next_status = 0;
2081 s->ret = -EINPROGRESS;
2082
2083 qemu_co_mutex_init(&s->lock);
2084 for (i = 0; i < s->num_coroutines; i++) {
2085 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2086 s->wait_sector_num[i] = -1;
2087 qemu_coroutine_enter(s->co[i]);
2088 }
2089
2090 while (s->running_coroutines) {
2091 main_loop_wait(false);
2092 }
2093
2094 if (s->compressed && !s->ret) {
2095 /* signal EOF to align */
2096 ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2097 if (ret < 0) {
2098 return ret;
2099 }
2100 }
2101
2102 return s->ret;
2103 }
2104
2105 /* Check that bitmaps can be copied, or output an error */
2106 static int convert_check_bitmaps(BlockDriverState *src, bool skip_broken)
2107 {
2108 BdrvDirtyBitmap *bm;
2109
2110 if (!bdrv_supports_persistent_dirty_bitmap(src)) {
2111 error_report("Source lacks bitmap support");
2112 return -1;
2113 }
2114 FOR_EACH_DIRTY_BITMAP(src, bm) {
2115 if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2116 continue;
2117 }
2118 if (!skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) {
2119 error_report("Cannot copy inconsistent bitmap '%s'",
2120 bdrv_dirty_bitmap_name(bm));
2121 error_printf("Try --skip-broken-bitmaps, or "
2122 "use 'qemu-img bitmap --remove' to delete it\n");
2123 return -1;
2124 }
2125 }
2126 return 0;
2127 }
2128
2129 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst,
2130 bool skip_broken)
2131 {
2132 BdrvDirtyBitmap *bm;
2133 Error *err = NULL;
2134
2135 FOR_EACH_DIRTY_BITMAP(src, bm) {
2136 const char *name;
2137
2138 if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2139 continue;
2140 }
2141 name = bdrv_dirty_bitmap_name(bm);
2142 if (skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) {
2143 warn_report("Skipping inconsistent bitmap '%s'", name);
2144 continue;
2145 }
2146 qmp_block_dirty_bitmap_add(dst->node_name, name,
2147 true, bdrv_dirty_bitmap_granularity(bm),
2148 true, true,
2149 true, !bdrv_dirty_bitmap_enabled(bm),
2150 &err);
2151 if (err) {
2152 error_reportf_err(err, "Failed to create bitmap %s: ", name);
2153 return -1;
2154 }
2155
2156 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2157 &err);
2158 if (err) {
2159 error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2160 qmp_block_dirty_bitmap_remove(dst->node_name, name, NULL);
2161 return -1;
2162 }
2163 }
2164
2165 return 0;
2166 }
2167
2168 #define MAX_BUF_SECTORS 32768
2169
2170 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2171 {
2172 ThrottleConfig cfg;
2173
2174 throttle_config_init(&cfg);
2175 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2176
2177 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2178 blk_set_io_limits(blk, &cfg);
2179 }
2180
2181 static int img_convert(int argc, char **argv)
2182 {
2183 int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE;
2184 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2185 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2186 *out_filename, *out_baseimg_param, *snapshot_name = NULL,
2187 *backing_fmt = NULL;
2188 BlockDriver *drv = NULL, *proto_drv = NULL;
2189 BlockDriverInfo bdi;
2190 BlockDriverState *out_bs;
2191 QemuOpts *opts = NULL, *sn_opts = NULL;
2192 QemuOptsList *create_opts = NULL;
2193 QDict *open_opts = NULL;
2194 char *options = NULL;
2195 Error *local_err = NULL;
2196 bool writethrough, src_writethrough, image_opts = false,
2197 skip_create = false, progress = false, tgt_image_opts = false;
2198 int64_t ret = -EINVAL;
2199 bool force_share = false;
2200 bool explict_min_sparse = false;
2201 bool bitmaps = false;
2202 bool skip_broken = false;
2203 int64_t rate_limit = 0;
2204
2205 ImgConvertState s = (ImgConvertState) {
2206 /* Need at least 4k of zeros for sparse detection */
2207 .min_sparse = 8,
2208 .copy_range = false,
2209 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2210 .wr_in_order = true,
2211 .num_coroutines = 8,
2212 };
2213
2214 for(;;) {
2215 static const struct option long_options[] = {
2216 {"help", no_argument, 0, 'h'},
2217 {"object", required_argument, 0, OPTION_OBJECT},
2218 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2219 {"force-share", no_argument, 0, 'U'},
2220 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2221 {"salvage", no_argument, 0, OPTION_SALVAGE},
2222 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2223 {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2224 {"skip-broken-bitmaps", no_argument, 0, OPTION_SKIP_BROKEN},
2225 {0, 0, 0, 0}
2226 };
2227 c = getopt_long(argc, argv, ":hf:O:B:CcF:o:l:S:pt:T:qnm:WUr:",
2228 long_options, NULL);
2229 if (c == -1) {
2230 break;
2231 }
2232 switch(c) {
2233 case ':':
2234 missing_argument(argv[optind - 1]);
2235 break;
2236 case '?':
2237 unrecognized_option(argv[optind - 1]);
2238 break;
2239 case 'h':
2240 help();
2241 break;
2242 case 'f':
2243 fmt = optarg;
2244 break;
2245 case 'O':
2246 out_fmt = optarg;
2247 break;
2248 case 'B':
2249 out_baseimg = optarg;
2250 break;
2251 case 'C':
2252 s.copy_range = true;
2253 break;
2254 case 'c':
2255 s.compressed = true;
2256 break;
2257 case 'F':
2258 backing_fmt = optarg;
2259 break;
2260 case 'o':
2261 if (accumulate_options(&options, optarg) < 0) {
2262 goto fail_getopt;
2263 }
2264 break;
2265 case 'l':
2266 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2267 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2268 optarg, false);
2269 if (!sn_opts) {
2270 error_report("Failed in parsing snapshot param '%s'",
2271 optarg);
2272 goto fail_getopt;
2273 }
2274 } else {
2275 snapshot_name = optarg;
2276 }
2277 break;
2278 case 'S':
2279 {
2280 int64_t sval;
2281
2282 sval = cvtnum("buffer size for sparse output", optarg);
2283 if (sval < 0) {
2284 goto fail_getopt;
2285 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2286 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2287 error_report("Invalid buffer size for sparse output specified. "
2288 "Valid sizes are multiples of %llu up to %llu. Select "
2289 "0 to disable sparse detection (fully allocates output).",
2290 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2291 goto fail_getopt;
2292 }
2293
2294 s.min_sparse = sval / BDRV_SECTOR_SIZE;
2295 explict_min_sparse = true;
2296 break;
2297 }
2298 case 'p':
2299 progress = true;
2300 break;
2301 case 't':
2302 cache = optarg;
2303 break;
2304 case 'T':
2305 src_cache = optarg;
2306 break;
2307 case 'q':
2308 s.quiet = true;
2309 break;
2310 case 'n':
2311 skip_create = true;
2312 break;
2313 case 'm':
2314 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2315 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2316 error_report("Invalid number of coroutines. Allowed number of"
2317 " coroutines is between 1 and %d", MAX_COROUTINES);
2318 goto fail_getopt;
2319 }
2320 break;
2321 case 'W':
2322 s.wr_in_order = false;
2323 break;
2324 case 'U':
2325 force_share = true;
2326 break;
2327 case 'r':
2328 rate_limit = cvtnum("rate limit", optarg);
2329 if (rate_limit < 0) {
2330 goto fail_getopt;
2331 }
2332 break;
2333 case OPTION_OBJECT:
2334 user_creatable_process_cmdline(optarg);
2335 break;
2336 case OPTION_IMAGE_OPTS:
2337 image_opts = true;
2338 break;
2339 case OPTION_SALVAGE:
2340 s.salvage = true;
2341 break;
2342 case OPTION_TARGET_IMAGE_OPTS:
2343 tgt_image_opts = true;
2344 break;
2345 case OPTION_TARGET_IS_ZERO:
2346 /*
2347 * The user asserting that the target is blank has the
2348 * same effect as the target driver supporting zero
2349 * initialisation.
2350 */
2351 s.has_zero_init = true;
2352 break;
2353 case OPTION_BITMAPS:
2354 bitmaps = true;
2355 break;
2356 case OPTION_SKIP_BROKEN:
2357 skip_broken = true;
2358 break;
2359 }
2360 }
2361
2362 if (!out_fmt && !tgt_image_opts) {
2363 out_fmt = "raw";
2364 }
2365
2366 if (skip_broken && !bitmaps) {
2367 error_report("Use of --skip-broken-bitmaps requires --bitmaps");
2368 goto fail_getopt;
2369 }
2370
2371 if (s.compressed && s.copy_range) {
2372 error_report("Cannot enable copy offloading when -c is used");
2373 goto fail_getopt;
2374 }
2375
2376 if (explict_min_sparse && s.copy_range) {
2377 error_report("Cannot enable copy offloading when -S is used");
2378 goto fail_getopt;
2379 }
2380
2381 if (s.copy_range && s.salvage) {
2382 error_report("Cannot use copy offloading in salvaging mode");
2383 goto fail_getopt;
2384 }
2385
2386 if (tgt_image_opts && !skip_create) {
2387 error_report("--target-image-opts requires use of -n flag");
2388 goto fail_getopt;
2389 }
2390
2391 if (skip_create && options) {
2392 error_report("-o has no effect when skipping image creation");
2393 goto fail_getopt;
2394 }
2395
2396 if (s.has_zero_init && !skip_create) {
2397 error_report("--target-is-zero requires use of -n flag");
2398 goto fail_getopt;
2399 }
2400
2401 s.src_num = argc - optind - 1;
2402 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2403
2404 if (options && has_help_option(options)) {
2405 if (out_fmt) {
2406 ret = print_block_option_help(out_filename, out_fmt);
2407 goto fail_getopt;
2408 } else {
2409 error_report("Option help requires a format be specified");
2410 goto fail_getopt;
2411 }
2412 }
2413
2414 if (s.src_num < 1) {
2415 error_report("Must specify image file name");
2416 goto fail_getopt;
2417 }
2418
2419 /* ret is still -EINVAL until here */
2420 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2421 if (ret < 0) {
2422 error_report("Invalid source cache option: %s", src_cache);
2423 goto fail_getopt;
2424 }
2425
2426 /* Initialize before goto out */
2427 if (s.quiet) {
2428 progress = false;
2429 }
2430 qemu_progress_init(progress, 1.0);
2431 qemu_progress_print(0, 100);
2432
2433 s.src = g_new0(BlockBackend *, s.src_num);
2434 s.src_sectors = g_new(int64_t, s.src_num);
2435 s.src_alignment = g_new(int, s.src_num);
2436
2437 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2438 BlockDriverState *src_bs;
2439 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2440 fmt, src_flags, src_writethrough, s.quiet,
2441 force_share);
2442 if (!s.src[bs_i]) {
2443 ret = -1;
2444 goto out;
2445 }
2446 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2447 if (s.src_sectors[bs_i] < 0) {
2448 error_report("Could not get size of %s: %s",
2449 argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2450 ret = -1;
2451 goto out;
2452 }
2453 src_bs = blk_bs(s.src[bs_i]);
2454 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2455 BDRV_SECTOR_SIZE);
2456 if (!bdrv_get_info(src_bs, &bdi)) {
2457 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2458 bdi.cluster_size / BDRV_SECTOR_SIZE);
2459 }
2460 s.total_sectors += s.src_sectors[bs_i];
2461 }
2462
2463 if (sn_opts) {
2464 bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2465 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2466 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2467 &local_err);
2468 } else if (snapshot_name != NULL) {
2469 if (s.src_num > 1) {
2470 error_report("No support for concatenating multiple snapshot");
2471 ret = -1;
2472 goto out;
2473 }
2474
2475 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2476 &local_err);
2477 }
2478 if (local_err) {
2479 error_reportf_err(local_err, "Failed to load snapshot: ");
2480 ret = -1;
2481 goto out;
2482 }
2483
2484 if (!skip_create) {
2485 /* Find driver and parse its options */
2486 drv = bdrv_find_format(out_fmt);
2487 if (!drv) {
2488 error_report("Unknown file format '%s'", out_fmt);
2489 ret = -1;
2490 goto out;
2491 }
2492
2493 proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2494 if (!proto_drv) {
2495 error_report_err(local_err);
2496 ret = -1;
2497 goto out;
2498 }
2499
2500 if (!drv->create_opts) {
2501 error_report("Format driver '%s' does not support image creation",
2502 drv->format_name);
2503 ret = -1;
2504 goto out;
2505 }
2506
2507 if (!proto_drv->create_opts) {
2508 error_report("Protocol driver '%s' does not support image creation",
2509 proto_drv->format_name);
2510 ret = -1;
2511 goto out;
2512 }
2513
2514 create_opts = qemu_opts_append(create_opts, drv->create_opts);
2515 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2516
2517 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2518 if (options) {
2519 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2520 error_report_err(local_err);
2521 ret = -1;
2522 goto out;
2523 }
2524 }
2525
2526 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2527 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2528 ret = add_old_style_options(out_fmt, opts, out_baseimg, backing_fmt);
2529 if (ret < 0) {
2530 goto out;
2531 }
2532 }
2533
2534 /* Get backing file name if -o backing_file was used */
2535 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2536 if (out_baseimg_param) {
2537 out_baseimg = out_baseimg_param;
2538 }
2539 s.target_has_backing = (bool) out_baseimg;
2540
2541 if (s.has_zero_init && s.target_has_backing) {
2542 error_report("Cannot use --target-is-zero when the destination "
2543 "image has a backing file");
2544 goto out;
2545 }
2546
2547 if (s.src_num > 1 && out_baseimg) {
2548 error_report("Having a backing file for the target makes no sense when "
2549 "concatenating multiple input images");
2550 ret = -1;
2551 goto out;
2552 }
2553
2554 if (out_baseimg_param) {
2555 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2556 error_report("Use of backing file requires explicit "
2557 "backing format");
2558 ret = -1;
2559 goto out;
2560 }
2561 }
2562
2563 /* Check if compression is supported */
2564 if (s.compressed) {
2565 bool encryption =
2566 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2567 const char *encryptfmt =
2568 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2569 const char *preallocation =
2570 qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2571
2572 if (drv && !block_driver_can_compress(drv)) {
2573 error_report("Compression not supported for this file format");
2574 ret = -1;
2575 goto out;
2576 }
2577
2578 if (encryption || encryptfmt) {
2579 error_report("Compression and encryption not supported at "
2580 "the same time");
2581 ret = -1;
2582 goto out;
2583 }
2584
2585 if (preallocation
2586 && strcmp(preallocation, "off"))
2587 {
2588 error_report("Compression and preallocation not supported at "
2589 "the same time");
2590 ret = -1;
2591 goto out;
2592 }
2593 }
2594
2595 /* Determine if bitmaps need copying */
2596 if (bitmaps) {
2597 if (s.src_num > 1) {
2598 error_report("Copying bitmaps only possible with single source");
2599 ret = -1;
2600 goto out;
2601 }
2602 ret = convert_check_bitmaps(blk_bs(s.src[0]), skip_broken);
2603 if (ret < 0) {
2604 goto out;
2605 }
2606 }
2607
2608 /*
2609 * The later open call will need any decryption secrets, and
2610 * bdrv_create() will purge "opts", so extract them now before
2611 * they are lost.
2612 */
2613 if (!skip_create) {
2614 open_opts = qdict_new();
2615 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2616
2617 /* Create the new image */
2618 ret = bdrv_create(drv, out_filename, opts, &local_err);
2619 if (ret < 0) {
2620 error_reportf_err(local_err, "%s: error while converting %s: ",
2621 out_filename, out_fmt);
2622 goto out;
2623 }
2624 }
2625
2626 s.target_is_new = !skip_create;
2627
2628 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2629 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2630 if (ret < 0) {
2631 error_report("Invalid cache option: %s", cache);
2632 goto out;
2633 }
2634
2635 if (flags & BDRV_O_NOCACHE) {
2636 /*
2637 * If we open the target with O_DIRECT, it may be necessary to
2638 * extend its size to align to the physical sector size.
2639 */
2640 flags |= BDRV_O_RESIZE;
2641 }
2642
2643 if (skip_create) {
2644 s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2645 flags, writethrough, s.quiet, false);
2646 } else {
2647 /* TODO ultimately we should allow --target-image-opts
2648 * to be used even when -n is not given.
2649 * That has to wait for bdrv_create to be improved
2650 * to allow filenames in option syntax
2651 */
2652 s.target = img_open_file(out_filename, open_opts, out_fmt,
2653 flags, writethrough, s.quiet, false);
2654 open_opts = NULL; /* blk_new_open will have freed it */
2655 }
2656 if (!s.target) {
2657 ret = -1;
2658 goto out;
2659 }
2660 out_bs = blk_bs(s.target);
2661
2662 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2663 error_report("Format driver '%s' does not support bitmaps",
2664 out_bs->drv->format_name);
2665 ret = -1;
2666 goto out;
2667 }
2668
2669 if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2670 error_report("Compression not supported for this file format");
2671 ret = -1;
2672 goto out;
2673 }
2674
2675 /* increase bufsectors from the default 4096 (2M) if opt_transfer
2676 * or discard_alignment of the out_bs is greater. Limit to
2677 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2678 s.buf_sectors = MIN(MAX_BUF_SECTORS,
2679 MAX(s.buf_sectors,
2680 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2681 out_bs->bl.pdiscard_alignment >>
2682 BDRV_SECTOR_BITS)));
2683
2684 /* try to align the write requests to the destination to avoid unnecessary
2685 * RMW cycles. */
2686 s.alignment = MAX(pow2floor(s.min_sparse),
2687 DIV_ROUND_UP(out_bs->bl.request_alignment,
2688 BDRV_SECTOR_SIZE));
2689 assert(is_power_of_2(s.alignment));
2690
2691 if (skip_create) {
2692 int64_t output_sectors = blk_nb_sectors(s.target);
2693 if (output_sectors < 0) {
2694 error_report("unable to get output image length: %s",
2695 strerror(-output_sectors));
2696 ret = -1;
2697 goto out;
2698 } else if (output_sectors < s.total_sectors) {
2699 error_report("output file is smaller than input file");
2700 ret = -1;
2701 goto out;
2702 }
2703 }
2704
2705 if (s.target_has_backing && s.target_is_new) {
2706 /* Errors are treated as "backing length unknown" (which means
2707 * s.target_backing_sectors has to be negative, which it will
2708 * be automatically). The backing file length is used only
2709 * for optimizations, so such a case is not fatal. */
2710 s.target_backing_sectors =
2711 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2712 } else {
2713 s.target_backing_sectors = -1;
2714 }
2715
2716 ret = bdrv_get_info(out_bs, &bdi);
2717 if (ret < 0) {
2718 if (s.compressed) {
2719 error_report("could not get block driver info");
2720 goto out;
2721 }
2722 } else {
2723 s.compressed = s.compressed || bdi.needs_compressed_writes;
2724 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2725 }
2726
2727 if (rate_limit) {
2728 set_rate_limit(s.target, rate_limit);
2729 }
2730
2731 ret = convert_do_copy(&s);
2732
2733 /* Now copy the bitmaps */
2734 if (bitmaps && ret == 0) {
2735 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs, skip_broken);
2736 }
2737
2738 out:
2739 if (!ret) {
2740 qemu_progress_print(100, 0);
2741 }
2742 qemu_progress_end();
2743 qemu_opts_del(opts);
2744 qemu_opts_free(create_opts);
2745 qobject_unref(open_opts);
2746 blk_unref(s.target);
2747 if (s.src) {
2748 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2749 blk_unref(s.src[bs_i]);
2750 }
2751 g_free(s.src);
2752 }
2753 g_free(s.src_sectors);
2754 g_free(s.src_alignment);
2755 fail_getopt:
2756 qemu_opts_del(sn_opts);
2757 g_free(options);
2758
2759 return !!ret;
2760 }
2761
2762
2763 static void dump_snapshots(BlockDriverState *bs)
2764 {
2765 QEMUSnapshotInfo *sn_tab, *sn;
2766 int nb_sns, i;
2767
2768 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2769 if (nb_sns <= 0)
2770 return;
2771 printf("Snapshot list:\n");
2772 bdrv_snapshot_dump(NULL);
2773 printf("\n");
2774 for(i = 0; i < nb_sns; i++) {
2775 sn = &sn_tab[i];
2776 bdrv_snapshot_dump(sn);
2777 printf("\n");
2778 }
2779 g_free(sn_tab);
2780 }
2781
2782 static void dump_json_image_info_list(ImageInfoList *list)
2783 {
2784 GString *str;
2785 QObject *obj;
2786 Visitor *v = qobject_output_visitor_new(&obj);
2787
2788 visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2789 visit_complete(v, &obj);
2790 str = qobject_to_json_pretty(obj, true);
2791 assert(str != NULL);
2792 printf("%s\n", str->str);
2793 qobject_unref(obj);
2794 visit_free(v);
2795 g_string_free(str, true);
2796 }
2797
2798 static void dump_json_image_info(ImageInfo *info)
2799 {
2800 GString *str;
2801 QObject *obj;
2802 Visitor *v = qobject_output_visitor_new(&obj);
2803
2804 visit_type_ImageInfo(v, NULL, &info, &error_abort);
2805 visit_complete(v, &obj);
2806 str = qobject_to_json_pretty(obj, true);
2807 assert(str != NULL);
2808 printf("%s\n", str->str);
2809 qobject_unref(obj);
2810 visit_free(v);
2811 g_string_free(str, true);
2812 }
2813
2814 static void dump_human_image_info_list(ImageInfoList *list)
2815 {
2816 ImageInfoList *elem;
2817 bool delim = false;
2818
2819 for (elem = list; elem; elem = elem->next) {
2820 if (delim) {
2821 printf("\n");
2822 }
2823 delim = true;
2824
2825 bdrv_image_info_dump(elem->value);
2826 }
2827 }
2828
2829 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2830 {
2831 return strcmp(a, b) == 0;
2832 }
2833
2834 /**
2835 * Open an image file chain and return an ImageInfoList
2836 *
2837 * @filename: topmost image filename
2838 * @fmt: topmost image format (may be NULL to autodetect)
2839 * @chain: true - enumerate entire backing file chain
2840 * false - only topmost image file
2841 *
2842 * Returns a list of ImageInfo objects or NULL if there was an error opening an
2843 * image file. If there was an error a message will have been printed to
2844 * stderr.
2845 */
2846 static ImageInfoList *collect_image_info_list(bool image_opts,
2847 const char *filename,
2848 const char *fmt,
2849 bool chain, bool force_share)
2850 {
2851 ImageInfoList *head = NULL;
2852 ImageInfoList **tail = &head;
2853 GHashTable *filenames;
2854 Error *err = NULL;
2855
2856 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2857
2858 while (filename) {
2859 BlockBackend *blk;
2860 BlockDriverState *bs;
2861 ImageInfo *info;
2862
2863 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2864 error_report("Backing file '%s' creates an infinite loop.",
2865 filename);
2866 goto err;
2867 }
2868 g_hash_table_insert(filenames, (gpointer)filename, NULL);
2869
2870 blk = img_open(image_opts, filename, fmt,
2871 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2872 force_share);
2873 if (!blk) {
2874 goto err;
2875 }
2876 bs = blk_bs(blk);
2877
2878 bdrv_query_image_info(bs, &info, &err);
2879 if (err) {
2880 error_report_err(err);
2881 blk_unref(blk);
2882 goto err;
2883 }
2884
2885 QAPI_LIST_APPEND(tail, info);
2886
2887 blk_unref(blk);
2888
2889 /* Clear parameters that only apply to the topmost image */
2890 filename = fmt = NULL;
2891 image_opts = false;
2892
2893 if (chain) {
2894 if (info->has_full_backing_filename) {
2895 filename = info->full_backing_filename;
2896 } else if (info->has_backing_filename) {
2897 error_report("Could not determine absolute backing filename,"
2898 " but backing filename '%s' present",
2899 info->backing_filename);
2900 goto err;
2901 }
2902 if (info->has_backing_filename_format) {
2903 fmt = info->backing_filename_format;
2904 }
2905 }
2906 }
2907 g_hash_table_destroy(filenames);
2908 return head;
2909
2910 err:
2911 qapi_free_ImageInfoList(head);
2912 g_hash_table_destroy(filenames);
2913 return NULL;
2914 }
2915
2916 static int img_info(int argc, char **argv)
2917 {
2918 int c;
2919 OutputFormat output_format = OFORMAT_HUMAN;
2920 bool chain = false;
2921 const char *filename, *fmt, *output;
2922 ImageInfoList *list;
2923 bool image_opts = false;
2924 bool force_share = false;
2925
2926 fmt = NULL;
2927 output = NULL;
2928 for(;;) {
2929 int option_index = 0;
2930 static const struct option long_options[] = {
2931 {"help", no_argument, 0, 'h'},
2932 {"format", required_argument, 0, 'f'},
2933 {"output", required_argument, 0, OPTION_OUTPUT},
2934 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2935 {"object", required_argument, 0, OPTION_OBJECT},
2936 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2937 {"force-share", no_argument, 0, 'U'},
2938 {0, 0, 0, 0}
2939 };
2940 c = getopt_long(argc, argv, ":f:hU",
2941 long_options, &option_index);
2942 if (c == -1) {
2943 break;
2944 }
2945 switch(c) {
2946 case ':':
2947 missing_argument(argv[optind - 1]);
2948 break;
2949 case '?':
2950 unrecognized_option(argv[optind - 1]);
2951 break;
2952 case 'h':
2953 help();
2954 break;
2955 case 'f':
2956 fmt = optarg;
2957 break;
2958 case 'U':
2959 force_share = true;
2960 break;
2961 case OPTION_OUTPUT:
2962 output = optarg;
2963 break;
2964 case OPTION_BACKING_CHAIN:
2965 chain = true;
2966 break;
2967 case OPTION_OBJECT:
2968 user_creatable_process_cmdline(optarg);
2969 break;
2970 case OPTION_IMAGE_OPTS:
2971 image_opts = true;
2972 break;
2973 }
2974 }
2975 if (optind != argc - 1) {
2976 error_exit("Expecting one image file name");
2977 }
2978 filename = argv[optind++];
2979
2980 if (output && !strcmp(output, "json")) {
2981 output_format = OFORMAT_JSON;
2982 } else if (output && !strcmp(output, "human")) {
2983 output_format = OFORMAT_HUMAN;
2984 } else if (output) {
2985 error_report("--output must be used with human or json as argument.");
2986 return 1;
2987 }
2988
2989 list = collect_image_info_list(image_opts, filename, fmt, chain,
2990 force_share);
2991 if (!list) {
2992 return 1;
2993 }
2994
2995 switch (output_format) {
2996 case OFORMAT_HUMAN:
2997 dump_human_image_info_list(list);
2998 break;
2999 case OFORMAT_JSON:
3000 if (chain) {
3001 dump_json_image_info_list(list);
3002 } else {
3003 dump_json_image_info(list->value);
3004 }
3005 break;
3006 }
3007
3008 qapi_free_ImageInfoList(list);
3009 return 0;
3010 }
3011
3012 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
3013 MapEntry *next)
3014 {
3015 switch (output_format) {
3016 case OFORMAT_HUMAN:
3017 if (e->data && !e->has_offset) {
3018 error_report("File contains external, encrypted or compressed clusters.");
3019 return -1;
3020 }
3021 if (e->data && !e->zero) {
3022 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
3023 e->start, e->length,
3024 e->has_offset ? e->offset : 0,
3025 e->has_filename ? e->filename : "");
3026 }
3027 /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
3028 * Modify the flags here to allow more coalescing.
3029 */
3030 if (next && (!next->data || next->zero)) {
3031 next->data = false;
3032 next->zero = true;
3033 }
3034 break;
3035 case OFORMAT_JSON:
3036 printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
3037 " \"depth\": %"PRId64", \"present\": %s, \"zero\": %s,"
3038 " \"data\": %s", e->start, e->length, e->depth,
3039 e->present ? "true" : "false",
3040 e->zero ? "true" : "false",
3041 e->data ? "true" : "false");
3042 if (e->has_offset) {
3043 printf(", \"offset\": %"PRId64"", e->offset);
3044 }
3045 putchar('}');
3046
3047 if (next) {
3048 puts(",");
3049 }
3050 break;
3051 }
3052 return 0;
3053 }
3054
3055 static int get_block_status(BlockDriverState *bs, int64_t offset,
3056 int64_t bytes, MapEntry *e)
3057 {
3058 int ret;
3059 int depth;
3060 BlockDriverState *file;
3061 bool has_offset;
3062 int64_t map;
3063 char *filename = NULL;
3064
3065 /* As an optimization, we could cache the current range of unallocated
3066 * clusters in each file of the chain, and avoid querying the same
3067 * range repeatedly.
3068 */
3069
3070 depth = 0;
3071 for (;;) {
3072 bs = bdrv_skip_filters(bs);
3073 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3074 if (ret < 0) {
3075 return ret;
3076 }
3077 assert(bytes);
3078 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3079 break;
3080 }
3081 bs = bdrv_cow_bs(bs);
3082 if (bs == NULL) {
3083 ret = 0;
3084 break;
3085 }
3086
3087 depth++;
3088 }
3089
3090 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3091
3092 if (file && has_offset) {
3093 bdrv_refresh_filename(file);
3094 filename = file->filename;
3095 }
3096
3097 *e = (MapEntry) {
3098 .start = offset,
3099 .length = bytes,
3100 .data = !!(ret & BDRV_BLOCK_DATA),
3101 .zero = !!(ret & BDRV_BLOCK_ZERO),
3102 .offset = map,
3103 .has_offset = has_offset,
3104 .depth = depth,
3105 .present = !!(ret & BDRV_BLOCK_ALLOCATED),
3106 .has_filename = filename,
3107 .filename = filename,
3108 };
3109
3110 return 0;
3111 }
3112
3113 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3114 {
3115 if (curr->length == 0) {
3116 return false;
3117 }
3118 if (curr->zero != next->zero ||
3119 curr->data != next->data ||
3120 curr->depth != next->depth ||
3121 curr->present != next->present ||
3122 curr->has_filename != next->has_filename ||
3123 curr->has_offset != next->has_offset) {
3124 return false;
3125 }
3126 if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3127 return false;
3128 }
3129 if (curr->has_offset && curr->offset + curr->length != next->offset) {
3130 return false;
3131 }
3132 return true;
3133 }
3134
3135 static int img_map(int argc, char **argv)
3136 {
3137 int c;
3138 OutputFormat output_format = OFORMAT_HUMAN;
3139 BlockBackend *blk;
3140 BlockDriverState *bs;
3141 const char *filename, *fmt, *output;
3142 int64_t length;
3143 MapEntry curr = { .length = 0 }, next;
3144 int ret = 0;
3145 bool image_opts = false;
3146 bool force_share = false;
3147 int64_t start_offset = 0;
3148 int64_t max_length = -1;
3149
3150 fmt = NULL;
3151 output = NULL;
3152 for (;;) {
3153 int option_index = 0;
3154 static const struct option long_options[] = {
3155 {"help", no_argument, 0, 'h'},
3156 {"format", required_argument, 0, 'f'},
3157 {"output", required_argument, 0, OPTION_OUTPUT},
3158 {"object", required_argument, 0, OPTION_OBJECT},
3159 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3160 {"force-share", no_argument, 0, 'U'},
3161 {"start-offset", required_argument, 0, 's'},
3162 {"max-length", required_argument, 0, 'l'},
3163 {0, 0, 0, 0}
3164 };
3165 c = getopt_long(argc, argv, ":f:s:l:hU",
3166 long_options, &option_index);
3167 if (c == -1) {
3168 break;
3169 }
3170 switch (c) {
3171 case ':':
3172 missing_argument(argv[optind - 1]);
3173 break;
3174 case '?':
3175 unrecognized_option(argv[optind - 1]);
3176 break;
3177 case 'h':
3178 help();
3179 break;
3180 case 'f':
3181 fmt = optarg;
3182 break;
3183 case 'U':
3184 force_share = true;
3185 break;
3186 case OPTION_OUTPUT:
3187 output = optarg;
3188 break;
3189 case 's':
3190 start_offset = cvtnum("start offset", optarg);
3191 if (start_offset < 0) {
3192 return 1;
3193 }
3194 break;
3195 case 'l':
3196 max_length = cvtnum("max length", optarg);
3197 if (max_length < 0) {
3198 return 1;
3199 }
3200 break;
3201 case OPTION_OBJECT:
3202 user_creatable_process_cmdline(optarg);
3203 break;
3204 case OPTION_IMAGE_OPTS:
3205 image_opts = true;
3206 break;
3207 }
3208 }
3209 if (optind != argc - 1) {
3210 error_exit("Expecting one image file name");
3211 }
3212 filename = argv[optind];
3213
3214 if (output && !strcmp(output, "json")) {
3215 output_format = OFORMAT_JSON;
3216 } else if (output && !strcmp(output, "human")) {
3217 output_format = OFORMAT_HUMAN;
3218 } else if (output) {
3219 error_report("--output must be used with human or json as argument.");
3220 return 1;
3221 }
3222
3223 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3224 if (!blk) {
3225 return 1;
3226 }
3227 bs = blk_bs(blk);
3228
3229 if (output_format == OFORMAT_HUMAN) {
3230 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3231 } else if (output_format == OFORMAT_JSON) {
3232 putchar('[');
3233 }
3234
3235 length = blk_getlength(blk);
3236 if (length < 0) {
3237 error_report("Failed to get size for '%s'", filename);
3238 return 1;
3239 }
3240 if (max_length != -1) {
3241 length = MIN(start_offset + max_length, length);
3242 }
3243
3244 curr.start = start_offset;
3245 while (curr.start + curr.length < length) {
3246 int64_t offset = curr.start + curr.length;
3247 int64_t n = length - offset;
3248
3249 ret = get_block_status(bs, offset, n, &next);
3250 if (ret < 0) {
3251 error_report("Could not read file metadata: %s", strerror(-ret));
3252 goto out;
3253 }
3254
3255 if (entry_mergeable(&curr, &next)) {
3256 curr.length += next.length;
3257 continue;
3258 }
3259
3260 if (curr.length > 0) {
3261 ret = dump_map_entry(output_format, &curr, &next);
3262 if (ret < 0) {
3263 goto out;
3264 }
3265 }
3266 curr = next;
3267 }
3268
3269 ret = dump_map_entry(output_format, &curr, NULL);
3270 if (output_format == OFORMAT_JSON) {
3271 puts("]");
3272 }
3273
3274 out:
3275 blk_unref(blk);
3276 return ret < 0;
3277 }
3278
3279 #define SNAPSHOT_LIST 1
3280 #define SNAPSHOT_CREATE 2
3281 #define SNAPSHOT_APPLY 3
3282 #define SNAPSHOT_DELETE 4
3283
3284 static int img_snapshot(int argc, char **argv)
3285 {
3286 BlockBackend *blk;
3287 BlockDriverState *bs;
3288 QEMUSnapshotInfo sn;
3289 char *filename, *snapshot_name = NULL;
3290 int c, ret = 0, bdrv_oflags;
3291 int action = 0;
3292 qemu_timeval tv;
3293 bool quiet = false;
3294 Error *err = NULL;
3295 bool image_opts = false;
3296 bool force_share = false;
3297
3298 bdrv_oflags = BDRV_O_RDWR;
3299 /* Parse commandline parameters */
3300 for(;;) {
3301 static const struct option long_options[] = {
3302 {"help", no_argument, 0, 'h'},
3303 {"object", required_argument, 0, OPTION_OBJECT},
3304 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3305 {"force-share", no_argument, 0, 'U'},
3306 {0, 0, 0, 0}
3307 };
3308 c = getopt_long(argc, argv, ":la:c:d:hqU",
3309 long_options, NULL);
3310 if (c == -1) {
3311 break;
3312 }
3313 switch(c) {
3314 case ':':
3315 missing_argument(argv[optind - 1]);
3316 break;
3317 case '?':
3318 unrecognized_option(argv[optind - 1]);
3319 break;
3320 case 'h':
3321 help();
3322 return 0;
3323 case 'l':
3324 if (action) {
3325 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3326 return 0;
3327 }
3328 action = SNAPSHOT_LIST;
3329 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3330 break;
3331 case 'a':
3332 if (action) {
3333 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3334 return 0;
3335 }
3336 action = SNAPSHOT_APPLY;
3337 snapshot_name = optarg;
3338 break;
3339 case 'c':
3340 if (action) {
3341 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3342 return 0;
3343 }
3344 action = SNAPSHOT_CREATE;
3345 snapshot_name = optarg;
3346 break;
3347 case 'd':
3348 if (action) {
3349 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3350 return 0;
3351 }
3352 action = SNAPSHOT_DELETE;
3353 snapshot_name = optarg;
3354 break;
3355 case 'q':
3356 quiet = true;
3357 break;
3358 case 'U':
3359 force_share = true;
3360 break;
3361 case OPTION_OBJECT:
3362 user_creatable_process_cmdline(optarg);
3363 break;
3364 case OPTION_IMAGE_OPTS:
3365 image_opts = true;
3366 break;
3367 }
3368 }
3369
3370 if (optind != argc - 1) {
3371 error_exit("Expecting one image file name");
3372 }
3373 filename = argv[optind++];
3374
3375 /* Open the image */
3376 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3377 force_share);
3378 if (!blk) {
3379 return 1;
3380 }
3381 bs = blk_bs(blk);
3382
3383 /* Perform the requested action */
3384 switch(action) {
3385 case SNAPSHOT_LIST:
3386 dump_snapshots(bs);
3387 break;
3388
3389 case SNAPSHOT_CREATE:
3390 memset(&sn, 0, sizeof(sn));
3391 pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3392
3393 qemu_gettimeofday(&tv);
3394 sn.date_sec = tv.tv_sec;
3395 sn.date_nsec = tv.tv_usec * 1000;
3396
3397 ret = bdrv_snapshot_create(bs, &sn);
3398 if (ret) {
3399 error_report("Could not create snapshot '%s': %d (%s)",
3400 snapshot_name, ret, strerror(-ret));
3401 }
3402 break;
3403
3404 case SNAPSHOT_APPLY:
3405 ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3406 if (ret) {
3407 error_reportf_err(err, "Could not apply snapshot '%s': ",
3408 snapshot_name);
3409 }
3410 break;
3411
3412 case SNAPSHOT_DELETE:
3413 ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3414 if (ret < 0) {
3415 error_report("Could not delete snapshot '%s': snapshot not "
3416 "found", snapshot_name);
3417 ret = 1;
3418 } else {
3419 ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3420 if (ret < 0) {
3421 error_reportf_err(err, "Could not delete snapshot '%s': ",
3422 snapshot_name);
3423 ret = 1;
3424 }
3425 }
3426 break;
3427 }
3428
3429 /* Cleanup */
3430 blk_unref(blk);
3431 if (ret) {
3432 return 1;
3433 }
3434 return 0;
3435 }
3436
3437 static int img_rebase(int argc, char **argv)
3438 {
3439 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3440 uint8_t *buf_old = NULL;
3441 uint8_t *buf_new = NULL;
3442 BlockDriverState *bs = NULL, *prefix_chain_bs = NULL;
3443 BlockDriverState *unfiltered_bs;
3444 char *filename;
3445 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3446 int c, flags, src_flags, ret;
3447 bool writethrough, src_writethrough;
3448 int unsafe = 0;
3449 bool force_share = false;
3450 int progress = 0;
3451 bool quiet = false;
3452 Error *local_err = NULL;
3453 bool image_opts = false;
3454
3455 /* Parse commandline parameters */
3456 fmt = NULL;
3457 cache = BDRV_DEFAULT_CACHE;
3458 src_cache = BDRV_DEFAULT_CACHE;
3459 out_baseimg = NULL;
3460 out_basefmt = NULL;
3461 for(;;) {
3462 static const struct option long_options[] = {
3463 {"help", no_argument, 0, 'h'},
3464 {"object", required_argument, 0, OPTION_OBJECT},
3465 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3466 {"force-share", no_argument, 0, 'U'},
3467 {0, 0, 0, 0}
3468 };
3469 c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3470 long_options, NULL);
3471 if (c == -1) {
3472 break;
3473 }
3474 switch(c) {
3475 case ':':
3476 missing_argument(argv[optind - 1]);
3477 break;
3478 case '?':
3479 unrecognized_option(argv[optind - 1]);
3480 break;
3481 case 'h':
3482 help();
3483 return 0;
3484 case 'f':
3485 fmt = optarg;
3486 break;
3487 case 'F':
3488 out_basefmt = optarg;
3489 break;
3490 case 'b':
3491 out_baseimg = optarg;
3492 break;
3493 case 'u':
3494 unsafe = 1;
3495 break;
3496 case 'p':
3497 progress = 1;
3498 break;
3499 case 't':
3500 cache = optarg;
3501 break;
3502 case 'T':
3503 src_cache = optarg;
3504 break;
3505 case 'q':
3506 quiet = true;
3507 break;
3508 case OPTION_OBJECT:
3509 user_creatable_process_cmdline(optarg);
3510 break;
3511 case OPTION_IMAGE_OPTS:
3512 image_opts = true;
3513 break;
3514 case 'U':
3515 force_share = true;
3516 break;
3517 }
3518 }
3519
3520 if (quiet) {
3521 progress = 0;
3522 }
3523
3524 if (optind != argc - 1) {
3525 error_exit("Expecting one image file name");
3526 }
3527 if (!unsafe && !out_baseimg) {
3528 error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3529 }
3530 filename = argv[optind++];
3531
3532 qemu_progress_init(progress, 2.0);
3533 qemu_progress_print(0, 100);
3534
3535 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3536 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3537 if (ret < 0) {
3538 error_report("Invalid cache option: %s", cache);
3539 goto out;
3540 }
3541
3542 src_flags = 0;
3543 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3544 if (ret < 0) {
3545 error_report("Invalid source cache option: %s", src_cache);
3546 goto out;
3547 }
3548
3549 /* The source files are opened read-only, don't care about WCE */
3550 assert((src_flags & BDRV_O_RDWR) == 0);
3551 (void) src_writethrough;
3552
3553 /*
3554 * Open the images.
3555 *
3556 * Ignore the old backing file for unsafe rebase in case we want to correct
3557 * the reference to a renamed or moved backing file.
3558 */
3559 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3560 false);
3561 if (!blk) {
3562 ret = -1;
3563 goto out;
3564 }
3565 bs = blk_bs(blk);
3566
3567 unfiltered_bs = bdrv_skip_filters(bs);
3568
3569 if (out_basefmt != NULL) {
3570 if (bdrv_find_format(out_basefmt) == NULL) {
3571 error_report("Invalid format name: '%s'", out_basefmt);
3572 ret = -1;
3573 goto out;
3574 }
3575 }
3576
3577 /* For safe rebasing we need to compare old and new backing file */
3578 if (!unsafe) {
3579 QDict *options = NULL;
3580 BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs);
3581
3582 if (base_bs) {
3583 blk_old_backing = blk_new(qemu_get_aio_context(),
3584 BLK_PERM_CONSISTENT_READ,
3585 BLK_PERM_ALL);
3586 ret = blk_insert_bs(blk_old_backing, base_bs,
3587 &local_err);
3588 if (ret < 0) {
3589 error_reportf_err(local_err,
3590 "Could not reuse old backing file '%s': ",
3591 base_bs->filename);
3592 goto out;
3593 }
3594 } else {
3595 blk_old_backing = NULL;
3596 }
3597
3598 if (out_baseimg[0]) {
3599 const char *overlay_filename;
3600 char *out_real_path;
3601
3602 options = qdict_new();
3603 if (out_basefmt) {
3604 qdict_put_str(options, "driver", out_basefmt);
3605 }
3606 if (force_share) {
3607 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3608 }
3609
3610 bdrv_refresh_filename(bs);
3611 overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3612 : bs->filename;
3613 out_real_path =
3614 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3615 out_baseimg,
3616 &local_err);
3617 if (local_err) {
3618 qobject_unref(options);
3619 error_reportf_err(local_err,
3620 "Could not resolve backing filename: ");
3621 ret = -1;
3622 goto out;
3623 }
3624
3625 /*
3626 * Find out whether we rebase an image on top of a previous image
3627 * in its chain.
3628 */
3629 prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path);
3630 if (prefix_chain_bs) {
3631 qobject_unref(options);
3632 g_free(out_real_path);
3633
3634 blk_new_backing = blk_new(qemu_get_aio_context(),
3635 BLK_PERM_CONSISTENT_READ,
3636 BLK_PERM_ALL);
3637 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs,
3638 &local_err);
3639 if (ret < 0) {
3640 error_reportf_err(local_err,
3641 "Could not reuse backing file '%s': ",
3642 out_baseimg);
3643 goto out;
3644 }
3645 } else {
3646 blk_new_backing = blk_new_open(out_real_path, NULL,
3647 options, src_flags, &local_err);
3648 g_free(out_real_path);
3649 if (!blk_new_backing) {
3650 error_reportf_err(local_err,
3651 "Could not open new backing file '%s': ",
3652 out_baseimg);
3653 ret = -1;
3654 goto out;
3655 }
3656 }
3657 }
3658 }
3659
3660 /*
3661 * Check each unallocated cluster in the COW file. If it is unallocated,
3662 * accesses go to the backing file. We must therefore compare this cluster
3663 * in the old and new backing file, and if they differ we need to copy it
3664 * from the old backing file into the COW file.
3665 *
3666 * If qemu-img crashes during this step, no harm is done. The content of
3667 * the image is the same as the original one at any time.
3668 */
3669 if (!unsafe) {
3670 int64_t size;
3671 int64_t old_backing_size = 0;
3672 int64_t new_backing_size = 0;
3673 uint64_t offset;
3674 int64_t n;
3675 float local_progress = 0;
3676
3677 buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3678 buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3679
3680 size = blk_getlength(blk);
3681 if (size < 0) {
3682 error_report("Could not get size of '%s': %s",
3683 filename, strerror(-size));
3684 ret = -1;
3685 goto out;
3686 }
3687 if (blk_old_backing) {
3688 old_backing_size = blk_getlength(blk_old_backing);
3689 if (old_backing_size < 0) {
3690 char backing_name[PATH_MAX];
3691
3692 bdrv_get_backing_filename(bs, backing_name,
3693 sizeof(backing_name));
3694 error_report("Could not get size of '%s': %s",
3695 backing_name, strerror(-old_backing_size));
3696 ret = -1;
3697 goto out;
3698 }
3699 }
3700 if (blk_new_backing) {
3701 new_backing_size = blk_getlength(blk_new_backing);
3702 if (new_backing_size < 0) {
3703 error_report("Could not get size of '%s': %s",
3704 out_baseimg, strerror(-new_backing_size));
3705 ret = -1;
3706 goto out;
3707 }
3708 }
3709
3710 if (size != 0) {
3711 local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3712 }
3713
3714 for (offset = 0; offset < size; offset += n) {
3715 bool buf_old_is_zero = false;
3716
3717 /* How many bytes can we handle with the next read? */
3718 n = MIN(IO_BUF_SIZE, size - offset);
3719
3720 /* If the cluster is allocated, we don't need to take action */
3721 ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n);
3722 if (ret < 0) {
3723 error_report("error while reading image metadata: %s",
3724 strerror(-ret));
3725 goto out;
3726 }
3727 if (ret) {
3728 continue;
3729 }
3730
3731 if (prefix_chain_bs) {
3732 /*
3733 * If cluster wasn't changed since prefix_chain, we don't need
3734 * to take action
3735 */
3736 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs),
3737 prefix_chain_bs, false,
3738 offset, n, &n);
3739 if (ret < 0) {
3740 error_report("error while reading image metadata: %s",
3741 strerror(-ret));
3742 goto out;
3743 }
3744 if (!ret) {
3745 continue;
3746 }
3747 }
3748
3749 /*
3750 * Read old and new backing file and take into consideration that
3751 * backing files may be smaller than the COW image.
3752 */
3753 if (offset >= old_backing_size) {
3754 memset(buf_old, 0, n);
3755 buf_old_is_zero = true;
3756 } else {
3757 if (offset + n > old_backing_size) {
3758 n = old_backing_size - offset;
3759 }
3760
3761 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3762 if (ret < 0) {
3763 error_report("error while reading from old backing file");
3764 goto out;
3765 }
3766 }
3767
3768 if (offset >= new_backing_size || !blk_new_backing) {
3769 memset(buf_new, 0, n);
3770 } else {
3771 if (offset + n > new_backing_size) {
3772 n = new_backing_size - offset;
3773 }
3774
3775 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3776 if (ret < 0) {
3777 error_report("error while reading from new backing file");
3778 goto out;
3779 }
3780 }
3781
3782 /* If they differ, we need to write to the COW file */
3783 uint64_t written = 0;
3784
3785 while (written < n) {
3786 int64_t pnum;
3787
3788 if (compare_buffers(buf_old + written, buf_new + written,
3789 n - written, &pnum))
3790 {
3791 if (buf_old_is_zero) {
3792 ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3793 } else {
3794 ret = blk_pwrite(blk, offset + written,
3795 buf_old + written, pnum, 0);
3796 }
3797 if (ret < 0) {
3798 error_report("Error while writing to COW image: %s",
3799 strerror(-ret));
3800 goto out;
3801 }
3802 }
3803
3804 written += pnum;
3805 }
3806 qemu_progress_print(local_progress, 100);
3807 }
3808 }
3809
3810 /*
3811 * Change the backing file. All clusters that are different from the old
3812 * backing file are overwritten in the COW file now, so the visible content
3813 * doesn't change when we switch the backing file.
3814 */
3815 if (out_baseimg && *out_baseimg) {
3816 ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt,
3817 true);
3818 } else {
3819 ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false);
3820 }
3821
3822 if (ret == -ENOSPC) {
3823 error_report("Could not change the backing file to '%s': No "
3824 "space left in the file header", out_baseimg);
3825 } else if (ret == -EINVAL && out_baseimg && !out_basefmt) {
3826 error_report("Could not change the backing file to '%s': backing "
3827 "format must be specified", out_baseimg);
3828 } else if (ret < 0) {
3829 error_report("Could not change the backing file to '%s': %s",
3830 out_baseimg, strerror(-ret));
3831 }
3832
3833 qemu_progress_print(100, 0);
3834 /*
3835 * TODO At this point it is possible to check if any clusters that are
3836 * allocated in the COW file are the same in the backing file. If so, they
3837 * could be dropped from the COW file. Don't do this before switching the
3838 * backing file, in case of a crash this would lead to corruption.
3839 */
3840 out:
3841 qemu_progress_end();
3842 /* Cleanup */
3843 if (!unsafe) {
3844 blk_unref(blk_old_backing);
3845 blk_unref(blk_new_backing);
3846 }
3847 qemu_vfree(buf_old);
3848 qemu_vfree(buf_new);
3849
3850 blk_unref(blk);
3851 if (ret) {
3852 return 1;
3853 }
3854 return 0;
3855 }
3856
3857 static int img_resize(int argc, char **argv)
3858 {
3859 Error *err = NULL;
3860 int c, ret, relative;
3861 const char *filename, *fmt, *size;
3862 int64_t n, total_size, current_size;
3863 bool quiet = false;
3864 BlockBackend *blk = NULL;