Merge remote-tracking branch 'remotes/elmarco/tags/libslirp-pull-request' into staging
[qemu.git] / qemu-img.c
1 /*
2 * QEMU disk image utility
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27
28 #include "qemu-common.h"
29 #include "qemu-version.h"
30 #include "qapi/error.h"
31 #include "qapi/qapi-commands-block-core.h"
32 #include "qapi/qapi-visit-block-core.h"
33 #include "qapi/qobject-output-visitor.h"
34 #include "qapi/qmp/qjson.h"
35 #include "qapi/qmp/qdict.h"
36 #include "qapi/qmp/qstring.h"
37 #include "qemu/cutils.h"
38 #include "qemu/config-file.h"
39 #include "qemu/option.h"
40 #include "qemu/error-report.h"
41 #include "qemu/log.h"
42 #include "qemu/main-loop.h"
43 #include "qemu/module.h"
44 #include "qemu/sockets.h"
45 #include "qemu/units.h"
46 #include "qom/object_interfaces.h"
47 #include "sysemu/block-backend.h"
48 #include "block/block_int.h"
49 #include "block/blockjob.h"
50 #include "block/qapi.h"
51 #include "crypto/init.h"
52 #include "trace/control.h"
53 #include "qemu/throttle.h"
54 #include "block/throttle-groups.h"
55
56 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
57 "\n" QEMU_COPYRIGHT "\n"
58
59 typedef struct img_cmd_t {
60 const char *name;
61 int (*handler)(int argc, char **argv);
62 } img_cmd_t;
63
64 enum {
65 OPTION_OUTPUT = 256,
66 OPTION_BACKING_CHAIN = 257,
67 OPTION_OBJECT = 258,
68 OPTION_IMAGE_OPTS = 259,
69 OPTION_PATTERN = 260,
70 OPTION_FLUSH_INTERVAL = 261,
71 OPTION_NO_DRAIN = 262,
72 OPTION_TARGET_IMAGE_OPTS = 263,
73 OPTION_SIZE = 264,
74 OPTION_PREALLOCATION = 265,
75 OPTION_SHRINK = 266,
76 OPTION_SALVAGE = 267,
77 OPTION_TARGET_IS_ZERO = 268,
78 OPTION_ADD = 269,
79 OPTION_REMOVE = 270,
80 OPTION_CLEAR = 271,
81 OPTION_ENABLE = 272,
82 OPTION_DISABLE = 273,
83 OPTION_MERGE = 274,
84 OPTION_BITMAPS = 275,
85 OPTION_FORCE = 276,
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 static QemuOptsList qemu_object_opts = {
231 .name = "object",
232 .implied_opt_name = "qom-type",
233 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
234 .desc = {
235 { }
236 },
237 };
238
239 static bool qemu_img_object_print_help(const char *type, QemuOpts *opts)
240 {
241 if (user_creatable_print_help(type, opts)) {
242 exit(0);
243 }
244 return true;
245 }
246
247 /*
248 * Is @optarg safe for accumulate_options()?
249 * It is when multiple of them can be joined together separated by ','.
250 * To make that work, @optarg must not start with ',' (or else a
251 * separating ',' preceding it gets escaped), and it must not end with
252 * an odd number of ',' (or else a separating ',' following it gets
253 * escaped), or be empty (or else a separating ',' preceding it can
254 * escape a separating ',' following it).
255 *
256 */
257 static bool is_valid_option_list(const char *optarg)
258 {
259 size_t len = strlen(optarg);
260 size_t i;
261
262 if (!optarg[0] || optarg[0] == ',') {
263 return false;
264 }
265
266 for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
267 }
268 if ((len - i) % 2) {
269 return false;
270 }
271
272 return true;
273 }
274
275 static int accumulate_options(char **options, char *optarg)
276 {
277 char *new_options;
278
279 if (!is_valid_option_list(optarg)) {
280 error_report("Invalid option list: %s", optarg);
281 return -1;
282 }
283
284 if (!*options) {
285 *options = g_strdup(optarg);
286 } else {
287 new_options = g_strdup_printf("%s,%s", *options, optarg);
288 g_free(*options);
289 *options = new_options;
290 }
291 return 0;
292 }
293
294 static QemuOptsList qemu_source_opts = {
295 .name = "source",
296 .implied_opt_name = "file",
297 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
298 .desc = {
299 { }
300 },
301 };
302
303 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
304 {
305 int ret = 0;
306 if (!quiet) {
307 va_list args;
308 va_start(args, fmt);
309 ret = vprintf(fmt, args);
310 va_end(args);
311 }
312 return ret;
313 }
314
315
316 static int print_block_option_help(const char *filename, const char *fmt)
317 {
318 BlockDriver *drv, *proto_drv;
319 QemuOptsList *create_opts = NULL;
320 Error *local_err = NULL;
321
322 /* Find driver and parse its options */
323 drv = bdrv_find_format(fmt);
324 if (!drv) {
325 error_report("Unknown file format '%s'", fmt);
326 return 1;
327 }
328
329 if (!drv->create_opts) {
330 error_report("Format driver '%s' does not support image creation", fmt);
331 return 1;
332 }
333
334 create_opts = qemu_opts_append(create_opts, drv->create_opts);
335 if (filename) {
336 proto_drv = bdrv_find_protocol(filename, true, &local_err);
337 if (!proto_drv) {
338 error_report_err(local_err);
339 qemu_opts_free(create_opts);
340 return 1;
341 }
342 if (!proto_drv->create_opts) {
343 error_report("Protocol driver '%s' does not support image creation",
344 proto_drv->format_name);
345 qemu_opts_free(create_opts);
346 return 1;
347 }
348 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
349 }
350
351 if (filename) {
352 printf("Supported options:\n");
353 } else {
354 printf("Supported %s options:\n", fmt);
355 }
356 qemu_opts_print_help(create_opts, false);
357 qemu_opts_free(create_opts);
358
359 if (!filename) {
360 printf("\n"
361 "The protocol level may support further options.\n"
362 "Specify the target filename to include those options.\n");
363 }
364
365 return 0;
366 }
367
368
369 static BlockBackend *img_open_opts(const char *optstr,
370 QemuOpts *opts, int flags, bool writethrough,
371 bool quiet, bool force_share)
372 {
373 QDict *options;
374 Error *local_err = NULL;
375 BlockBackend *blk;
376 options = qemu_opts_to_qdict(opts, NULL);
377 if (force_share) {
378 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
379 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
380 error_report("--force-share/-U conflicts with image options");
381 qobject_unref(options);
382 return NULL;
383 }
384 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
385 }
386 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
387 if (!blk) {
388 error_reportf_err(local_err, "Could not open '%s': ", optstr);
389 return NULL;
390 }
391 blk_set_enable_write_cache(blk, !writethrough);
392
393 return blk;
394 }
395
396 static BlockBackend *img_open_file(const char *filename,
397 QDict *options,
398 const char *fmt, int flags,
399 bool writethrough, bool quiet,
400 bool force_share)
401 {
402 BlockBackend *blk;
403 Error *local_err = NULL;
404
405 if (!options) {
406 options = qdict_new();
407 }
408 if (fmt) {
409 qdict_put_str(options, "driver", fmt);
410 }
411
412 if (force_share) {
413 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
414 }
415 blk = blk_new_open(filename, NULL, options, flags, &local_err);
416 if (!blk) {
417 error_reportf_err(local_err, "Could not open '%s': ", filename);
418 return NULL;
419 }
420 blk_set_enable_write_cache(blk, !writethrough);
421
422 return blk;
423 }
424
425
426 static int img_add_key_secrets(void *opaque,
427 const char *name, const char *value,
428 Error **errp)
429 {
430 QDict *options = opaque;
431
432 if (g_str_has_suffix(name, "key-secret")) {
433 qdict_put_str(options, name, value);
434 }
435
436 return 0;
437 }
438
439
440 static BlockBackend *img_open(bool image_opts,
441 const char *filename,
442 const char *fmt, int flags, bool writethrough,
443 bool quiet, bool force_share)
444 {
445 BlockBackend *blk;
446 if (image_opts) {
447 QemuOpts *opts;
448 if (fmt) {
449 error_report("--image-opts and --format are mutually exclusive");
450 return NULL;
451 }
452 opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
453 filename, true);
454 if (!opts) {
455 return NULL;
456 }
457 blk = img_open_opts(filename, opts, flags, writethrough, quiet,
458 force_share);
459 } else {
460 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
461 force_share);
462 }
463 return blk;
464 }
465
466
467 static int add_old_style_options(const char *fmt, QemuOpts *opts,
468 const char *base_filename,
469 const char *base_fmt)
470 {
471 if (base_filename) {
472 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
473 NULL)) {
474 error_report("Backing file not supported for file format '%s'",
475 fmt);
476 return -1;
477 }
478 }
479 if (base_fmt) {
480 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
481 error_report("Backing file format not supported for file "
482 "format '%s'", fmt);
483 return -1;
484 }
485 }
486 return 0;
487 }
488
489 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
490 int64_t max)
491 {
492 int err;
493 uint64_t res;
494
495 err = qemu_strtosz(value, NULL, &res);
496 if (err < 0 && err != -ERANGE) {
497 error_report("Invalid %s specified. You may use "
498 "k, M, G, T, P or E suffixes for", name);
499 error_report("kilobytes, megabytes, gigabytes, terabytes, "
500 "petabytes and exabytes.");
501 return err;
502 }
503 if (err == -ERANGE || res > max || res < min) {
504 error_report("Invalid %s specified. Must be between %" PRId64
505 " and %" PRId64 ".", name, min, max);
506 return -ERANGE;
507 }
508 return res;
509 }
510
511 static int64_t cvtnum(const char *name, const char *value)
512 {
513 return cvtnum_full(name, value, 0, INT64_MAX);
514 }
515
516 static int img_create(int argc, char **argv)
517 {
518 int c;
519 uint64_t img_size = -1;
520 const char *fmt = "raw";
521 const char *base_fmt = NULL;
522 const char *filename;
523 const char *base_filename = NULL;
524 char *options = NULL;
525 Error *local_err = NULL;
526 bool quiet = false;
527 int flags = 0;
528
529 for(;;) {
530 static const struct option long_options[] = {
531 {"help", no_argument, 0, 'h'},
532 {"object", required_argument, 0, OPTION_OBJECT},
533 {0, 0, 0, 0}
534 };
535 c = getopt_long(argc, argv, ":F:b:f:ho:qu",
536 long_options, NULL);
537 if (c == -1) {
538 break;
539 }
540 switch(c) {
541 case ':':
542 missing_argument(argv[optind - 1]);
543 break;
544 case '?':
545 unrecognized_option(argv[optind - 1]);
546 break;
547 case 'h':
548 help();
549 break;
550 case 'F':
551 base_fmt = optarg;
552 break;
553 case 'b':
554 base_filename = optarg;
555 break;
556 case 'f':
557 fmt = optarg;
558 break;
559 case 'o':
560 if (accumulate_options(&options, optarg) < 0) {
561 goto fail;
562 }
563 break;
564 case 'q':
565 quiet = true;
566 break;
567 case 'u':
568 flags |= BDRV_O_NO_BACKING;
569 break;
570 case OPTION_OBJECT: {
571 QemuOpts *opts;
572 opts = qemu_opts_parse_noisily(&qemu_object_opts,
573 optarg, true);
574 if (!opts) {
575 goto fail;
576 }
577 } break;
578 }
579 }
580
581 /* Get the filename */
582 filename = (optind < argc) ? argv[optind] : NULL;
583 if (options && has_help_option(options)) {
584 g_free(options);
585 return print_block_option_help(filename, fmt);
586 }
587
588 if (optind >= argc) {
589 error_exit("Expecting image file name");
590 }
591 optind++;
592
593 if (qemu_opts_foreach(&qemu_object_opts,
594 user_creatable_add_opts_foreach,
595 qemu_img_object_print_help, &error_fatal)) {
596 goto fail;
597 }
598
599 /* Get image size, if specified */
600 if (optind < argc) {
601 int64_t sval;
602
603 sval = cvtnum("image size", argv[optind++]);
604 if (sval < 0) {
605 goto fail;
606 }
607 img_size = (uint64_t)sval;
608 }
609 if (optind != argc) {
610 error_exit("Unexpected argument: %s", argv[optind]);
611 }
612
613 bdrv_img_create(filename, fmt, base_filename, base_fmt,
614 options, img_size, flags, quiet, &local_err);
615 if (local_err) {
616 error_reportf_err(local_err, "%s: ", filename);
617 goto fail;
618 }
619
620 g_free(options);
621 return 0;
622
623 fail:
624 g_free(options);
625 return 1;
626 }
627
628 static void dump_json_image_check(ImageCheck *check, bool quiet)
629 {
630 QString *str;
631 QObject *obj;
632 Visitor *v = qobject_output_visitor_new(&obj);
633
634 visit_type_ImageCheck(v, NULL, &check, &error_abort);
635 visit_complete(v, &obj);
636 str = qobject_to_json_pretty(obj);
637 assert(str != NULL);
638 qprintf(quiet, "%s\n", qstring_get_str(str));
639 qobject_unref(obj);
640 visit_free(v);
641 qobject_unref(str);
642 }
643
644 static void dump_human_image_check(ImageCheck *check, bool quiet)
645 {
646 if (!(check->corruptions || check->leaks || check->check_errors)) {
647 qprintf(quiet, "No errors were found on the image.\n");
648 } else {
649 if (check->corruptions) {
650 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
651 "Data may be corrupted, or further writes to the image "
652 "may corrupt it.\n",
653 check->corruptions);
654 }
655
656 if (check->leaks) {
657 qprintf(quiet,
658 "\n%" PRId64 " leaked clusters were found on the image.\n"
659 "This means waste of disk space, but no harm to data.\n",
660 check->leaks);
661 }
662
663 if (check->check_errors) {
664 qprintf(quiet,
665 "\n%" PRId64
666 " internal errors have occurred during the check.\n",
667 check->check_errors);
668 }
669 }
670
671 if (check->total_clusters != 0 && check->allocated_clusters != 0) {
672 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
673 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
674 check->allocated_clusters, check->total_clusters,
675 check->allocated_clusters * 100.0 / check->total_clusters,
676 check->fragmented_clusters * 100.0 / check->allocated_clusters,
677 check->compressed_clusters * 100.0 /
678 check->allocated_clusters);
679 }
680
681 if (check->image_end_offset) {
682 qprintf(quiet,
683 "Image end offset: %" PRId64 "\n", check->image_end_offset);
684 }
685 }
686
687 static int collect_image_check(BlockDriverState *bs,
688 ImageCheck *check,
689 const char *filename,
690 const char *fmt,
691 int fix)
692 {
693 int ret;
694 BdrvCheckResult result;
695
696 ret = bdrv_check(bs, &result, fix);
697 if (ret < 0) {
698 return ret;
699 }
700
701 check->filename = g_strdup(filename);
702 check->format = g_strdup(bdrv_get_format_name(bs));
703 check->check_errors = result.check_errors;
704 check->corruptions = result.corruptions;
705 check->has_corruptions = result.corruptions != 0;
706 check->leaks = result.leaks;
707 check->has_leaks = result.leaks != 0;
708 check->corruptions_fixed = result.corruptions_fixed;
709 check->has_corruptions_fixed = result.corruptions_fixed != 0;
710 check->leaks_fixed = result.leaks_fixed;
711 check->has_leaks_fixed = result.leaks_fixed != 0;
712 check->image_end_offset = result.image_end_offset;
713 check->has_image_end_offset = result.image_end_offset != 0;
714 check->total_clusters = result.bfi.total_clusters;
715 check->has_total_clusters = result.bfi.total_clusters != 0;
716 check->allocated_clusters = result.bfi.allocated_clusters;
717 check->has_allocated_clusters = result.bfi.allocated_clusters != 0;
718 check->fragmented_clusters = result.bfi.fragmented_clusters;
719 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0;
720 check->compressed_clusters = result.bfi.compressed_clusters;
721 check->has_compressed_clusters = result.bfi.compressed_clusters != 0;
722
723 return 0;
724 }
725
726 /*
727 * Checks an image for consistency. Exit codes:
728 *
729 * 0 - Check completed, image is good
730 * 1 - Check not completed because of internal errors
731 * 2 - Check completed, image is corrupted
732 * 3 - Check completed, image has leaked clusters, but is good otherwise
733 * 63 - Checks are not supported by the image format
734 */
735 static int img_check(int argc, char **argv)
736 {
737 int c, ret;
738 OutputFormat output_format = OFORMAT_HUMAN;
739 const char *filename, *fmt, *output, *cache;
740 BlockBackend *blk;
741 BlockDriverState *bs;
742 int fix = 0;
743 int flags = BDRV_O_CHECK;
744 bool writethrough;
745 ImageCheck *check;
746 bool quiet = false;
747 bool image_opts = false;
748 bool force_share = false;
749
750 fmt = NULL;
751 output = NULL;
752 cache = BDRV_DEFAULT_CACHE;
753
754 for(;;) {
755 int option_index = 0;
756 static const struct option long_options[] = {
757 {"help", no_argument, 0, 'h'},
758 {"format", required_argument, 0, 'f'},
759 {"repair", required_argument, 0, 'r'},
760 {"output", required_argument, 0, OPTION_OUTPUT},
761 {"object", required_argument, 0, OPTION_OBJECT},
762 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
763 {"force-share", no_argument, 0, 'U'},
764 {0, 0, 0, 0}
765 };
766 c = getopt_long(argc, argv, ":hf:r:T:qU",
767 long_options, &option_index);
768 if (c == -1) {
769 break;
770 }
771 switch(c) {
772 case ':':
773 missing_argument(argv[optind - 1]);
774 break;
775 case '?':
776 unrecognized_option(argv[optind - 1]);
777 break;
778 case 'h':
779 help();
780 break;
781 case 'f':
782 fmt = optarg;
783 break;
784 case 'r':
785 flags |= BDRV_O_RDWR;
786
787 if (!strcmp(optarg, "leaks")) {
788 fix = BDRV_FIX_LEAKS;
789 } else if (!strcmp(optarg, "all")) {
790 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
791 } else {
792 error_exit("Unknown option value for -r "
793 "(expecting 'leaks' or 'all'): %s", optarg);
794 }
795 break;
796 case OPTION_OUTPUT:
797 output = optarg;
798 break;
799 case 'T':
800 cache = optarg;
801 break;
802 case 'q':
803 quiet = true;
804 break;
805 case 'U':
806 force_share = true;
807 break;
808 case OPTION_OBJECT: {
809 QemuOpts *opts;
810 opts = qemu_opts_parse_noisily(&qemu_object_opts,
811 optarg, true);
812 if (!opts) {
813 return 1;
814 }
815 } break;
816 case OPTION_IMAGE_OPTS:
817 image_opts = true;
818 break;
819 }
820 }
821 if (optind != argc - 1) {
822 error_exit("Expecting one image file name");
823 }
824 filename = argv[optind++];
825
826 if (output && !strcmp(output, "json")) {
827 output_format = OFORMAT_JSON;
828 } else if (output && !strcmp(output, "human")) {
829 output_format = OFORMAT_HUMAN;
830 } else if (output) {
831 error_report("--output must be used with human or json as argument.");
832 return 1;
833 }
834
835 if (qemu_opts_foreach(&qemu_object_opts,
836 user_creatable_add_opts_foreach,
837 qemu_img_object_print_help, &error_fatal)) {
838 return 1;
839 }
840
841 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
842 if (ret < 0) {
843 error_report("Invalid source cache option: %s", cache);
844 return 1;
845 }
846
847 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
848 force_share);
849 if (!blk) {
850 return 1;
851 }
852 bs = blk_bs(blk);
853
854 check = g_new0(ImageCheck, 1);
855 ret = collect_image_check(bs, check, filename, fmt, fix);
856
857 if (ret == -ENOTSUP) {
858 error_report("This image format does not support checks");
859 ret = 63;
860 goto fail;
861 }
862
863 if (check->corruptions_fixed || check->leaks_fixed) {
864 int corruptions_fixed, leaks_fixed;
865 bool has_leaks_fixed, has_corruptions_fixed;
866
867 leaks_fixed = check->leaks_fixed;
868 has_leaks_fixed = check->has_leaks_fixed;
869 corruptions_fixed = check->corruptions_fixed;
870 has_corruptions_fixed = check->has_corruptions_fixed;
871
872 if (output_format == OFORMAT_HUMAN) {
873 qprintf(quiet,
874 "The following inconsistencies were found and repaired:\n\n"
875 " %" PRId64 " leaked clusters\n"
876 " %" PRId64 " corruptions\n\n"
877 "Double checking the fixed image now...\n",
878 check->leaks_fixed,
879 check->corruptions_fixed);
880 }
881
882 qapi_free_ImageCheck(check);
883 check = g_new0(ImageCheck, 1);
884 ret = collect_image_check(bs, check, filename, fmt, 0);
885
886 check->leaks_fixed = leaks_fixed;
887 check->has_leaks_fixed = has_leaks_fixed;
888 check->corruptions_fixed = corruptions_fixed;
889 check->has_corruptions_fixed = has_corruptions_fixed;
890 }
891
892 if (!ret) {
893 switch (output_format) {
894 case OFORMAT_HUMAN:
895 dump_human_image_check(check, quiet);
896 break;
897 case OFORMAT_JSON:
898 dump_json_image_check(check, quiet);
899 break;
900 }
901 }
902
903 if (ret || check->check_errors) {
904 if (ret) {
905 error_report("Check failed: %s", strerror(-ret));
906 } else {
907 error_report("Check failed");
908 }
909 ret = 1;
910 goto fail;
911 }
912
913 if (check->corruptions) {
914 ret = 2;
915 } else if (check->leaks) {
916 ret = 3;
917 } else {
918 ret = 0;
919 }
920
921 fail:
922 qapi_free_ImageCheck(check);
923 blk_unref(blk);
924 return ret;
925 }
926
927 typedef struct CommonBlockJobCBInfo {
928 BlockDriverState *bs;
929 Error **errp;
930 } CommonBlockJobCBInfo;
931
932 static void common_block_job_cb(void *opaque, int ret)
933 {
934 CommonBlockJobCBInfo *cbi = opaque;
935
936 if (ret < 0) {
937 error_setg_errno(cbi->errp, -ret, "Block job failed");
938 }
939 }
940
941 static void run_block_job(BlockJob *job, Error **errp)
942 {
943 AioContext *aio_context = blk_get_aio_context(job->blk);
944 int ret = 0;
945
946 aio_context_acquire(aio_context);
947 job_ref(&job->job);
948 do {
949 float progress = 0.0f;
950 aio_poll(aio_context, true);
951 if (job->job.progress.total) {
952 progress = (float)job->job.progress.current /
953 job->job.progress.total * 100.f;
954 }
955 qemu_progress_print(progress, 0);
956 } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
957
958 if (!job_is_completed(&job->job)) {
959 ret = job_complete_sync(&job->job, errp);
960 } else {
961 ret = job->job.ret;
962 }
963 job_unref(&job->job);
964 aio_context_release(aio_context);
965
966 /* publish completion progress only when success */
967 if (!ret) {
968 qemu_progress_print(100.f, 0);
969 }
970 }
971
972 static int img_commit(int argc, char **argv)
973 {
974 int c, ret, flags;
975 const char *filename, *fmt, *cache, *base;
976 BlockBackend *blk;
977 BlockDriverState *bs, *base_bs;
978 BlockJob *job;
979 bool progress = false, quiet = false, drop = false;
980 bool writethrough;
981 Error *local_err = NULL;
982 CommonBlockJobCBInfo cbi;
983 bool image_opts = false;
984 AioContext *aio_context;
985 int64_t rate_limit = 0;
986
987 fmt = NULL;
988 cache = BDRV_DEFAULT_CACHE;
989 base = NULL;
990 for(;;) {
991 static const struct option long_options[] = {
992 {"help", no_argument, 0, 'h'},
993 {"object", required_argument, 0, OPTION_OBJECT},
994 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
995 {0, 0, 0, 0}
996 };
997 c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
998 long_options, NULL);
999 if (c == -1) {
1000 break;
1001 }
1002 switch(c) {
1003 case ':':
1004 missing_argument(argv[optind - 1]);
1005 break;
1006 case '?':
1007 unrecognized_option(argv[optind - 1]);
1008 break;
1009 case 'h':
1010 help();
1011 break;
1012 case 'f':
1013 fmt = optarg;
1014 break;
1015 case 't':
1016 cache = optarg;
1017 break;
1018 case 'b':
1019 base = optarg;
1020 /* -b implies -d */
1021 drop = true;
1022 break;
1023 case 'd':
1024 drop = true;
1025 break;
1026 case 'p':
1027 progress = true;
1028 break;
1029 case 'q':
1030 quiet = true;
1031 break;
1032 case 'r':
1033 rate_limit = cvtnum("rate limit", optarg);
1034 if (rate_limit < 0) {
1035 return 1;
1036 }
1037 break;
1038 case OPTION_OBJECT: {
1039 QemuOpts *opts;
1040 opts = qemu_opts_parse_noisily(&qemu_object_opts,
1041 optarg, true);
1042 if (!opts) {
1043 return 1;
1044 }
1045 } break;
1046 case OPTION_IMAGE_OPTS:
1047 image_opts = true;
1048 break;
1049 }
1050 }
1051
1052 /* Progress is not shown in Quiet mode */
1053 if (quiet) {
1054 progress = false;
1055 }
1056
1057 if (optind != argc - 1) {
1058 error_exit("Expecting one image file name");
1059 }
1060 filename = argv[optind++];
1061
1062 if (qemu_opts_foreach(&qemu_object_opts,
1063 user_creatable_add_opts_foreach,
1064 qemu_img_object_print_help, &error_fatal)) {
1065 return 1;
1066 }
1067
1068 flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1069 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1070 if (ret < 0) {
1071 error_report("Invalid cache option: %s", cache);
1072 return 1;
1073 }
1074
1075 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1076 false);
1077 if (!blk) {
1078 return 1;
1079 }
1080 bs = blk_bs(blk);
1081
1082 qemu_progress_init(progress, 1.f);
1083 qemu_progress_print(0.f, 100);
1084
1085 if (base) {
1086 base_bs = bdrv_find_backing_image(bs, base);
1087 if (!base_bs) {
1088 error_setg(&local_err,
1089 "Did not find '%s' in the backing chain of '%s'",
1090 base, filename);
1091 goto done;
1092 }
1093 } else {
1094 /* This is different from QMP, which by default uses the deepest file in
1095 * the backing chain (i.e., the very base); however, the traditional
1096 * behavior of qemu-img commit is using the immediate backing file. */
1097 base_bs = bdrv_backing_chain_next(bs);
1098 if (!base_bs) {
1099 error_setg(&local_err, "Image does not have a backing file");
1100 goto done;
1101 }
1102 }
1103
1104 cbi = (CommonBlockJobCBInfo){
1105 .errp = &local_err,
1106 .bs = bs,
1107 };
1108
1109 aio_context = bdrv_get_aio_context(bs);
1110 aio_context_acquire(aio_context);
1111 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1112 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1113 &cbi, false, &local_err);
1114 aio_context_release(aio_context);
1115 if (local_err) {
1116 goto done;
1117 }
1118
1119 /* When the block job completes, the BlockBackend reference will point to
1120 * the old backing file. In order to avoid that the top image is already
1121 * deleted, so we can still empty it afterwards, increment the reference
1122 * counter here preemptively. */
1123 if (!drop) {
1124 bdrv_ref(bs);
1125 }
1126
1127 job = block_job_get("commit");
1128 assert(job);
1129 run_block_job(job, &local_err);
1130 if (local_err) {
1131 goto unref_backing;
1132 }
1133
1134 if (!drop) {
1135 BlockBackend *old_backing_blk;
1136
1137 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1138 &local_err);
1139 if (!old_backing_blk) {
1140 goto unref_backing;
1141 }
1142 ret = blk_make_empty(old_backing_blk, &local_err);
1143 blk_unref(old_backing_blk);
1144 if (ret == -ENOTSUP) {
1145 error_free(local_err);
1146 local_err = NULL;
1147 } else if (ret < 0) {
1148 goto unref_backing;
1149 }
1150 }
1151
1152 unref_backing:
1153 if (!drop) {
1154 bdrv_unref(bs);
1155 }
1156
1157 done:
1158 qemu_progress_end();
1159
1160 blk_unref(blk);
1161
1162 if (local_err) {
1163 error_report_err(local_err);
1164 return 1;
1165 }
1166
1167 qprintf(quiet, "Image committed.\n");
1168 return 0;
1169 }
1170
1171 /*
1172 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1173 * of the first sector boundary within buf where the sector contains a
1174 * non-zero byte. This function is robust to a buffer that is not
1175 * sector-aligned.
1176 */
1177 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1178 {
1179 int64_t i;
1180 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1181
1182 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1183 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1184 return i;
1185 }
1186 }
1187 if (i < n && !buffer_is_zero(buf + i, n - end)) {
1188 return i;
1189 }
1190 return -1;
1191 }
1192
1193 /*
1194 * Returns true iff the first sector pointed to by 'buf' contains at least
1195 * a non-NUL byte.
1196 *
1197 * 'pnum' is set to the number of sectors (including and immediately following
1198 * the first one) that are known to be in the same allocated/unallocated state.
1199 * The function will try to align the end offset to alignment boundaries so
1200 * that the request will at least end aligned and consecutive requests will
1201 * also start at an aligned offset.
1202 */
1203 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1204 int64_t sector_num, int alignment)
1205 {
1206 bool is_zero;
1207 int i, tail;
1208
1209 if (n <= 0) {
1210 *pnum = 0;
1211 return 0;
1212 }
1213 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1214 for(i = 1; i < n; i++) {
1215 buf += BDRV_SECTOR_SIZE;
1216 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1217 break;
1218 }
1219 }
1220
1221 tail = (sector_num + i) & (alignment - 1);
1222 if (tail) {
1223 if (is_zero && i <= tail) {
1224 /* treat unallocated areas which only consist
1225 * of a small tail as allocated. */
1226 is_zero = false;
1227 }
1228 if (!is_zero) {
1229 /* align up end offset of allocated areas. */
1230 i += alignment - tail;
1231 i = MIN(i, n);
1232 } else {
1233 /* align down end offset of zero areas. */
1234 i -= tail;
1235 }
1236 }
1237 *pnum = i;
1238 return !is_zero;
1239 }
1240
1241 /*
1242 * Like is_allocated_sectors, but if the buffer starts with a used sector,
1243 * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1244 * breaking up write requests for only small sparse areas.
1245 */
1246 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1247 int min, int64_t sector_num, int alignment)
1248 {
1249 int ret;
1250 int num_checked, num_used;
1251
1252 if (n < min) {
1253 min = n;
1254 }
1255
1256 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1257 if (!ret) {
1258 return ret;
1259 }
1260
1261 num_used = *pnum;
1262 buf += BDRV_SECTOR_SIZE * *pnum;
1263 n -= *pnum;
1264 sector_num += *pnum;
1265 num_checked = num_used;
1266
1267 while (n > 0) {
1268 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1269
1270 buf += BDRV_SECTOR_SIZE * *pnum;
1271 n -= *pnum;
1272 sector_num += *pnum;
1273 num_checked += *pnum;
1274 if (ret) {
1275 num_used = num_checked;
1276 } else if (*pnum >= min) {
1277 break;
1278 }
1279 }
1280
1281 *pnum = num_used;
1282 return 1;
1283 }
1284
1285 /*
1286 * Compares two buffers sector by sector. Returns 0 if the first
1287 * sector of each buffer matches, non-zero otherwise.
1288 *
1289 * pnum is set to the sector-aligned size of the buffer prefix that
1290 * has the same matching status as the first sector.
1291 */
1292 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1293 int64_t bytes, int64_t *pnum)
1294 {
1295 bool res;
1296 int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1297
1298 assert(bytes > 0);
1299
1300 res = !!memcmp(buf1, buf2, i);
1301 while (i < bytes) {
1302 int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1303
1304 if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1305 break;
1306 }
1307 i += len;
1308 }
1309
1310 *pnum = i;
1311 return res;
1312 }
1313
1314 #define IO_BUF_SIZE (2 * MiB)
1315
1316 /*
1317 * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1318 *
1319 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1320 * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1321 * failure), and 4 on error (the exit status for read errors), after emitting
1322 * an error message.
1323 *
1324 * @param blk: BlockBackend for the image
1325 * @param offset: Starting offset to check
1326 * @param bytes: Number of bytes to check
1327 * @param filename: Name of disk file we are checking (logging purpose)
1328 * @param buffer: Allocated buffer for storing read data
1329 * @param quiet: Flag for quiet mode
1330 */
1331 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1332 int64_t bytes, const char *filename,
1333 uint8_t *buffer, bool quiet)
1334 {
1335 int ret = 0;
1336 int64_t idx;
1337
1338 ret = blk_pread(blk, offset, buffer, bytes);
1339 if (ret < 0) {
1340 error_report("Error while reading offset %" PRId64 " of %s: %s",
1341 offset, filename, strerror(-ret));
1342 return 4;
1343 }
1344 idx = find_nonzero(buffer, bytes);
1345 if (idx >= 0) {
1346 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1347 offset + idx);
1348 return 1;
1349 }
1350
1351 return 0;
1352 }
1353
1354 /*
1355 * Compares two images. Exit codes:
1356 *
1357 * 0 - Images are identical
1358 * 1 - Images differ
1359 * >1 - Error occurred
1360 */
1361 static int img_compare(int argc, char **argv)
1362 {
1363 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1364 BlockBackend *blk1, *blk2;
1365 BlockDriverState *bs1, *bs2;
1366 int64_t total_size1, total_size2;
1367 uint8_t *buf1 = NULL, *buf2 = NULL;
1368 int64_t pnum1, pnum2;
1369 int allocated1, allocated2;
1370 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1371 bool progress = false, quiet = false, strict = false;
1372 int flags;
1373 bool writethrough;
1374 int64_t total_size;
1375 int64_t offset = 0;
1376 int64_t chunk;
1377 int c;
1378 uint64_t progress_base;
1379 bool image_opts = false;
1380 bool force_share = false;
1381
1382 cache = BDRV_DEFAULT_CACHE;
1383 for (;;) {
1384 static const struct option long_options[] = {
1385 {"help", no_argument, 0, 'h'},
1386 {"object", required_argument, 0, OPTION_OBJECT},
1387 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1388 {"force-share", no_argument, 0, 'U'},
1389 {0, 0, 0, 0}
1390 };
1391 c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1392 long_options, NULL);
1393 if (c == -1) {
1394 break;
1395 }
1396 switch (c) {
1397 case ':':
1398 missing_argument(argv[optind - 1]);
1399 break;
1400 case '?':
1401 unrecognized_option(argv[optind - 1]);
1402 break;
1403 case 'h':
1404 help();
1405 break;
1406 case 'f':
1407 fmt1 = optarg;
1408 break;
1409 case 'F':
1410 fmt2 = optarg;
1411 break;
1412 case 'T':
1413 cache = optarg;
1414 break;
1415 case 'p':
1416 progress = true;
1417 break;
1418 case 'q':
1419 quiet = true;
1420 break;
1421 case 's':
1422 strict = true;
1423 break;
1424 case 'U':
1425 force_share = true;
1426 break;
1427 case OPTION_OBJECT: {
1428 QemuOpts *opts;
1429 opts = qemu_opts_parse_noisily(&qemu_object_opts,
1430 optarg, true);
1431 if (!opts) {
1432 ret = 2;
1433 goto out4;
1434 }
1435 } break;
1436 case OPTION_IMAGE_OPTS:
1437 image_opts = true;
1438 break;
1439 }
1440 }
1441
1442 /* Progress is not shown in Quiet mode */
1443 if (quiet) {
1444 progress = false;
1445 }
1446
1447
1448 if (optind != argc - 2) {
1449 error_exit("Expecting two image file names");
1450 }
1451 filename1 = argv[optind++];
1452 filename2 = argv[optind++];
1453
1454 if (qemu_opts_foreach(&qemu_object_opts,
1455 user_creatable_add_opts_foreach,
1456 qemu_img_object_print_help, &error_fatal)) {
1457 ret = 2;
1458 goto out4;
1459 }
1460
1461 /* Initialize before goto out */
1462 qemu_progress_init(progress, 2.0);
1463
1464 flags = 0;
1465 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1466 if (ret < 0) {
1467 error_report("Invalid source cache option: %s", cache);
1468 ret = 2;
1469 goto out3;
1470 }
1471
1472 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1473 force_share);
1474 if (!blk1) {
1475 ret = 2;
1476 goto out3;
1477 }
1478
1479 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1480 force_share);
1481 if (!blk2) {
1482 ret = 2;
1483 goto out2;
1484 }
1485 bs1 = blk_bs(blk1);
1486 bs2 = blk_bs(blk2);
1487
1488 buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1489 buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1490 total_size1 = blk_getlength(blk1);
1491 if (total_size1 < 0) {
1492 error_report("Can't get size of %s: %s",
1493 filename1, strerror(-total_size1));
1494 ret = 4;
1495 goto out;
1496 }
1497 total_size2 = blk_getlength(blk2);
1498 if (total_size2 < 0) {
1499 error_report("Can't get size of %s: %s",
1500 filename2, strerror(-total_size2));
1501 ret = 4;
1502 goto out;
1503 }
1504 total_size = MIN(total_size1, total_size2);
1505 progress_base = MAX(total_size1, total_size2);
1506
1507 qemu_progress_print(0, 100);
1508
1509 if (strict && total_size1 != total_size2) {
1510 ret = 1;
1511 qprintf(quiet, "Strict mode: Image size mismatch!\n");
1512 goto out;
1513 }
1514
1515 while (offset < total_size) {
1516 int status1, status2;
1517
1518 status1 = bdrv_block_status_above(bs1, NULL, offset,
1519 total_size1 - offset, &pnum1, NULL,
1520 NULL);
1521 if (status1 < 0) {
1522 ret = 3;
1523 error_report("Sector allocation test failed for %s", filename1);
1524 goto out;
1525 }
1526 allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1527
1528 status2 = bdrv_block_status_above(bs2, NULL, offset,
1529 total_size2 - offset, &pnum2, NULL,
1530 NULL);
1531 if (status2 < 0) {
1532 ret = 3;
1533 error_report("Sector allocation test failed for %s", filename2);
1534 goto out;
1535 }
1536 allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1537
1538 assert(pnum1 && pnum2);
1539 chunk = MIN(pnum1, pnum2);
1540
1541 if (strict) {
1542 if (status1 != status2) {
1543 ret = 1;
1544 qprintf(quiet, "Strict mode: Offset %" PRId64
1545 " block status mismatch!\n", offset);
1546 goto out;
1547 }
1548 }
1549 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1550 /* nothing to do */
1551 } else if (allocated1 == allocated2) {
1552 if (allocated1) {
1553 int64_t pnum;
1554
1555 chunk = MIN(chunk, IO_BUF_SIZE);
1556 ret = blk_pread(blk1, offset, buf1, chunk);
1557 if (ret < 0) {
1558 error_report("Error while reading offset %" PRId64
1559 " of %s: %s",
1560 offset, filename1, strerror(-ret));
1561 ret = 4;
1562 goto out;
1563 }
1564 ret = blk_pread(blk2, offset, buf2, chunk);
1565 if (ret < 0) {
1566 error_report("Error while reading offset %" PRId64
1567 " of %s: %s",
1568 offset, filename2, strerror(-ret));
1569 ret = 4;
1570 goto out;
1571 }
1572 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1573 if (ret || pnum != chunk) {
1574 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1575 offset + (ret ? 0 : pnum));
1576 ret = 1;
1577 goto out;
1578 }
1579 }
1580 } else {
1581 chunk = MIN(chunk, IO_BUF_SIZE);
1582 if (allocated1) {
1583 ret = check_empty_sectors(blk1, offset, chunk,
1584 filename1, buf1, quiet);
1585 } else {
1586 ret = check_empty_sectors(blk2, offset, chunk,
1587 filename2, buf1, quiet);
1588 }
1589 if (ret) {
1590 goto out;
1591 }
1592 }
1593 offset += chunk;
1594 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1595 }
1596
1597 if (total_size1 != total_size2) {
1598 BlockBackend *blk_over;
1599 const char *filename_over;
1600
1601 qprintf(quiet, "Warning: Image size mismatch!\n");
1602 if (total_size1 > total_size2) {
1603 blk_over = blk1;
1604 filename_over = filename1;
1605 } else {
1606 blk_over = blk2;
1607 filename_over = filename2;
1608 }
1609
1610 while (offset < progress_base) {
1611 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1612 progress_base - offset, &chunk,
1613 NULL, NULL);
1614 if (ret < 0) {
1615 ret = 3;
1616 error_report("Sector allocation test failed for %s",
1617 filename_over);
1618 goto out;
1619
1620 }
1621 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1622 chunk = MIN(chunk, IO_BUF_SIZE);
1623 ret = check_empty_sectors(blk_over, offset, chunk,
1624 filename_over, buf1, quiet);
1625 if (ret) {
1626 goto out;
1627 }
1628 }
1629 offset += chunk;
1630 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1631 }
1632 }
1633
1634 qprintf(quiet, "Images are identical.\n");
1635 ret = 0;
1636
1637 out:
1638 qemu_vfree(buf1);
1639 qemu_vfree(buf2);
1640 blk_unref(blk2);
1641 out2:
1642 blk_unref(blk1);
1643 out3:
1644 qemu_progress_end();
1645 out4:
1646 return ret;
1647 }
1648
1649 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1650 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1651 const char *src_node, const char *src_name,
1652 Error **errp)
1653 {
1654 BlockDirtyBitmapMergeSource *merge_src;
1655 BlockDirtyBitmapMergeSourceList *list;
1656
1657 merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1658 merge_src->type = QTYPE_QDICT;
1659 merge_src->u.external.node = g_strdup(src_node);
1660 merge_src->u.external.name = g_strdup(src_name);
1661 list = g_new0(BlockDirtyBitmapMergeSourceList, 1);
1662 list->value = merge_src;
1663 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1664 qapi_free_BlockDirtyBitmapMergeSourceList(list);
1665 }
1666
1667 enum ImgConvertBlockStatus {
1668 BLK_DATA,
1669 BLK_ZERO,
1670 BLK_BACKING_FILE,
1671 };
1672
1673 #define MAX_COROUTINES 16
1674 #define CONVERT_THROTTLE_GROUP "img_convert"
1675
1676 typedef struct ImgConvertState {
1677 BlockBackend **src;
1678 int64_t *src_sectors;
1679 int *src_alignment;
1680 int src_num;
1681 int64_t total_sectors;
1682 int64_t allocated_sectors;
1683 int64_t allocated_done;
1684 int64_t sector_num;
1685 int64_t wr_offs;
1686 enum ImgConvertBlockStatus status;
1687 int64_t sector_next_status;
1688 BlockBackend *target;
1689 bool has_zero_init;
1690 bool compressed;
1691 bool target_is_new;
1692 bool target_has_backing;
1693 int64_t target_backing_sectors; /* negative if unknown */
1694 bool wr_in_order;
1695 bool copy_range;
1696 bool salvage;
1697 bool quiet;
1698 int min_sparse;
1699 int alignment;
1700 size_t cluster_sectors;
1701 size_t buf_sectors;
1702 long num_coroutines;
1703 int running_coroutines;
1704 Coroutine *co[MAX_COROUTINES];
1705 int64_t wait_sector_num[MAX_COROUTINES];
1706 CoMutex lock;
1707 int ret;
1708 } ImgConvertState;
1709
1710 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1711 int *src_cur, int64_t *src_cur_offset)
1712 {
1713 *src_cur = 0;
1714 *src_cur_offset = 0;
1715 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1716 *src_cur_offset += s->src_sectors[*src_cur];
1717 (*src_cur)++;
1718 assert(*src_cur < s->src_num);
1719 }
1720 }
1721
1722 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1723 {
1724 int64_t src_cur_offset;
1725 int ret, n, src_cur;
1726 bool post_backing_zero = false;
1727
1728 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1729
1730 assert(s->total_sectors > sector_num);
1731 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1732
1733 if (s->target_backing_sectors >= 0) {
1734 if (sector_num >= s->target_backing_sectors) {
1735 post_backing_zero = true;
1736 } else if (sector_num + n > s->target_backing_sectors) {
1737 /* Split requests around target_backing_sectors (because
1738 * starting from there, zeros are handled differently) */
1739 n = s->target_backing_sectors - sector_num;
1740 }
1741 }
1742
1743 if (s->sector_next_status <= sector_num) {
1744 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1745 int64_t count;
1746 int tail;
1747 BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1748 BlockDriverState *base;
1749
1750 if (s->target_has_backing) {
1751 base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1752 } else {
1753 base = NULL;
1754 }
1755
1756 do {
1757 count = n * BDRV_SECTOR_SIZE;
1758
1759 ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1760 NULL, NULL);
1761
1762 if (ret < 0) {
1763 if (s->salvage) {
1764 if (n == 1) {
1765 if (!s->quiet) {
1766 warn_report("error while reading block status at "
1767 "offset %" PRIu64 ": %s", offset,
1768 strerror(-ret));
1769 }
1770 /* Just try to read the data, then */
1771 ret = BDRV_BLOCK_DATA;
1772 count = BDRV_SECTOR_SIZE;
1773 } else {
1774 /* Retry on a shorter range */
1775 n = DIV_ROUND_UP(n, 4);
1776 }
1777 } else {
1778 error_report("error while reading block status at offset "
1779 "%" PRIu64 ": %s", offset, strerror(-ret));
1780 return ret;
1781 }
1782 }
1783 } while (ret < 0);
1784
1785 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1786
1787 /*
1788 * Avoid that s->sector_next_status becomes unaligned to the source
1789 * request alignment and/or cluster size to avoid unnecessary read
1790 * cycles.
1791 */
1792 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1793 if (n > tail) {
1794 n -= tail;
1795 }
1796
1797 if (ret & BDRV_BLOCK_ZERO) {
1798 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1799 } else if (ret & BDRV_BLOCK_DATA) {
1800 s->status = BLK_DATA;
1801 } else {
1802 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1803 }
1804
1805 s->sector_next_status = sector_num + n;
1806 }
1807
1808 n = MIN(n, s->sector_next_status - sector_num);
1809 if (s->status == BLK_DATA) {
1810 n = MIN(n, s->buf_sectors);
1811 }
1812
1813 /* We need to write complete clusters for compressed images, so if an
1814 * unallocated area is shorter than that, we must consider the whole
1815 * cluster allocated. */
1816 if (s->compressed) {
1817 if (n < s->cluster_sectors) {
1818 n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1819 s->status = BLK_DATA;
1820 } else {
1821 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1822 }
1823 }
1824
1825 return n;
1826 }
1827
1828 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1829 int nb_sectors, uint8_t *buf)
1830 {
1831 uint64_t single_read_until = 0;
1832 int n, ret;
1833
1834 assert(nb_sectors <= s->buf_sectors);
1835 while (nb_sectors > 0) {
1836 BlockBackend *blk;
1837 int src_cur;
1838 int64_t bs_sectors, src_cur_offset;
1839 uint64_t offset;
1840
1841 /* In the case of compression with multiple source files, we can get a
1842 * nb_sectors that spreads into the next part. So we must be able to
1843 * read across multiple BDSes for one convert_read() call. */
1844 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1845 blk = s->src[src_cur];
1846 bs_sectors = s->src_sectors[src_cur];
1847
1848 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1849
1850 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1851 if (single_read_until > offset) {
1852 n = 1;
1853 }
1854
1855 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1856 if (ret < 0) {
1857 if (s->salvage) {
1858 if (n > 1) {
1859 single_read_until = offset + (n << BDRV_SECTOR_BITS);
1860 continue;
1861 } else {
1862 if (!s->quiet) {
1863 warn_report("error while reading offset %" PRIu64
1864 ": %s", offset, strerror(-ret));
1865 }
1866 memset(buf, 0, BDRV_SECTOR_SIZE);
1867 }
1868 } else {
1869 return ret;
1870 }
1871 }
1872
1873 sector_num += n;
1874 nb_sectors -= n;
1875 buf += n * BDRV_SECTOR_SIZE;
1876 }
1877
1878 return 0;
1879 }
1880
1881
1882 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1883 int nb_sectors, uint8_t *buf,
1884 enum ImgConvertBlockStatus status)
1885 {
1886 int ret;
1887
1888 while (nb_sectors > 0) {
1889 int n = nb_sectors;
1890 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1891
1892 switch (status) {
1893 case BLK_BACKING_FILE:
1894 /* If we have a backing file, leave clusters unallocated that are
1895 * unallocated in the source image, so that the backing file is
1896 * visible at the respective offset. */
1897 assert(s->target_has_backing);
1898 break;
1899
1900 case BLK_DATA:
1901 /* If we're told to keep the target fully allocated (-S 0) or there
1902 * is real non-zero data, we must write it. Otherwise we can treat
1903 * it as zero sectors.
1904 * Compressed clusters need to be written as a whole, so in that
1905 * case we can only save the write if the buffer is completely
1906 * zeroed. */
1907 if (!s->min_sparse ||
1908 (!s->compressed &&
1909 is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1910 sector_num, s->alignment)) ||
1911 (s->compressed &&
1912 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1913 {
1914 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1915 n << BDRV_SECTOR_BITS, buf, flags);
1916 if (ret < 0) {
1917 return ret;
1918 }
1919 break;
1920 }
1921 /* fall-through */
1922
1923 case BLK_ZERO:
1924 if (s->has_zero_init) {
1925 assert(!s->target_has_backing);
1926 break;
1927 }
1928 ret = blk_co_pwrite_zeroes(s->target,
1929 sector_num << BDRV_SECTOR_BITS,
1930 n << BDRV_SECTOR_BITS,
1931 BDRV_REQ_MAY_UNMAP);
1932 if (ret < 0) {
1933 return ret;
1934 }
1935 break;
1936 }
1937
1938 sector_num += n;
1939 nb_sectors -= n;
1940 buf += n * BDRV_SECTOR_SIZE;
1941 }
1942
1943 return 0;
1944 }
1945
1946 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1947 int nb_sectors)
1948 {
1949 int n, ret;
1950
1951 while (nb_sectors > 0) {
1952 BlockBackend *blk;
1953 int src_cur;
1954 int64_t bs_sectors, src_cur_offset;
1955 int64_t offset;
1956
1957 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1958 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1959 blk = s->src[src_cur];
1960 bs_sectors = s->src_sectors[src_cur];
1961
1962 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1963
1964 ret = blk_co_copy_range(blk, offset, s->target,
1965 sector_num << BDRV_SECTOR_BITS,
1966 n << BDRV_SECTOR_BITS, 0, 0);
1967 if (ret < 0) {
1968 return ret;
1969 }
1970
1971 sector_num += n;
1972 nb_sectors -= n;
1973 }
1974 return 0;
1975 }
1976
1977 static void coroutine_fn convert_co_do_copy(void *opaque)
1978 {
1979 ImgConvertState *s = opaque;
1980 uint8_t *buf = NULL;
1981 int ret, i;
1982 int index = -1;
1983
1984 for (i = 0; i < s->num_coroutines; i++) {
1985 if (s->co[i] == qemu_coroutine_self()) {
1986 index = i;
1987 break;
1988 }
1989 }
1990 assert(index >= 0);
1991
1992 s->running_coroutines++;
1993 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1994
1995 while (1) {
1996 int n;
1997 int64_t sector_num;
1998 enum ImgConvertBlockStatus status;
1999 bool copy_range;
2000
2001 qemu_co_mutex_lock(&s->lock);
2002 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
2003 qemu_co_mutex_unlock(&s->lock);
2004 break;
2005 }
2006 n = convert_iteration_sectors(s, s->sector_num);
2007 if (n < 0) {
2008 qemu_co_mutex_unlock(&s->lock);
2009 s->ret = n;
2010 break;
2011 }
2012 /* save current sector and allocation status to local variables */
2013 sector_num = s->sector_num;
2014 status = s->status;
2015 if (!s->min_sparse && s->status == BLK_ZERO) {
2016 n = MIN(n, s->buf_sectors);
2017 }
2018 /* increment global sector counter so that other coroutines can
2019 * already continue reading beyond this request */
2020 s->sector_num += n;
2021 qemu_co_mutex_unlock(&s->lock);
2022
2023 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
2024 s->allocated_done += n;
2025 qemu_progress_print(100.0 * s->allocated_done /
2026 s->allocated_sectors, 0);
2027 }
2028
2029 retry:
2030 copy_range = s->copy_range && s->status == BLK_DATA;
2031 if (status == BLK_DATA && !copy_range) {
2032 ret = convert_co_read(s, sector_num, n, buf);
2033 if (ret < 0) {
2034 error_report("error while reading at byte %lld: %s",
2035 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2036 s->ret = ret;
2037 }
2038 } else if (!s->min_sparse && status == BLK_ZERO) {
2039 status = BLK_DATA;
2040 memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
2041 }
2042
2043 if (s->wr_in_order) {
2044 /* keep writes in order */
2045 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
2046 s->wait_sector_num[index] = sector_num;
2047 qemu_coroutine_yield();
2048 }
2049 s->wait_sector_num[index] = -1;
2050 }
2051
2052 if (s->ret == -EINPROGRESS) {
2053 if (copy_range) {
2054 ret = convert_co_copy_range(s, sector_num, n);
2055 if (ret) {
2056 s->copy_range = false;
2057 goto retry;
2058 }
2059 } else {
2060 ret = convert_co_write(s, sector_num, n, buf, status);
2061 }
2062 if (ret < 0) {
2063 error_report("error while writing at byte %lld: %s",
2064 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2065 s->ret = ret;
2066 }
2067 }
2068
2069 if (s->wr_in_order) {
2070 /* reenter the coroutine that might have waited
2071 * for this write to complete */
2072 s->wr_offs = sector_num + n;
2073 for (i = 0; i < s->num_coroutines; i++) {
2074 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2075 /*
2076 * A -> B -> A cannot occur because A has
2077 * s->wait_sector_num[i] == -1 during A -> B. Therefore
2078 * B will never enter A during this time window.
2079 */
2080 qemu_coroutine_enter(s->co[i]);
2081 break;
2082 }
2083 }
2084 }
2085 }
2086
2087 qemu_vfree(buf);
2088 s->co[index] = NULL;
2089 s->running_coroutines--;
2090 if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2091 /* the convert job finished successfully */
2092 s->ret = 0;
2093 }
2094 }
2095
2096 static int convert_do_copy(ImgConvertState *s)
2097 {
2098 int ret, i, n;
2099 int64_t sector_num = 0;
2100
2101 /* Check whether we have zero initialisation or can get it efficiently */
2102 if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2103 !s->target_has_backing) {
2104 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2105 }
2106
2107 /* Allocate buffer for copied data. For compressed images, only one cluster
2108 * can be copied at a time. */
2109 if (s->compressed) {
2110 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2111 error_report("invalid cluster size");
2112 return -EINVAL;
2113 }
2114 s->buf_sectors = s->cluster_sectors;
2115 }
2116
2117 while (sector_num < s->total_sectors) {
2118 n = convert_iteration_sectors(s, sector_num);
2119 if (n < 0) {
2120 return n;
2121 }
2122 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2123 {
2124 s->allocated_sectors += n;
2125 }
2126 sector_num += n;
2127 }
2128
2129 /* Do the copy */
2130 s->sector_next_status = 0;
2131 s->ret = -EINPROGRESS;
2132
2133 qemu_co_mutex_init(&s->lock);
2134 for (i = 0; i < s->num_coroutines; i++) {
2135 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2136 s->wait_sector_num[i] = -1;
2137 qemu_coroutine_enter(s->co[i]);
2138 }
2139
2140 while (s->running_coroutines) {
2141 main_loop_wait(false);
2142 }
2143
2144 if (s->compressed && !s->ret) {
2145 /* signal EOF to align */
2146 ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2147 if (ret < 0) {
2148 return ret;
2149 }
2150 }
2151
2152 return s->ret;
2153 }
2154
2155 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst)
2156 {
2157 BdrvDirtyBitmap *bm;
2158 Error *err = NULL;
2159
2160 FOR_EACH_DIRTY_BITMAP(src, bm) {
2161 const char *name;
2162
2163 if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2164 continue;
2165 }
2166 name = bdrv_dirty_bitmap_name(bm);
2167 qmp_block_dirty_bitmap_add(dst->node_name, name,
2168 true, bdrv_dirty_bitmap_granularity(bm),
2169 true, true,
2170 true, !bdrv_dirty_bitmap_enabled(bm),
2171 &err);
2172 if (err) {
2173 error_reportf_err(err, "Failed to create bitmap %s: ", name);
2174 return -1;
2175 }
2176
2177 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2178 &err);
2179 if (err) {
2180 error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2181 return -1;
2182 }
2183 }
2184
2185 return 0;
2186 }
2187
2188 #define MAX_BUF_SECTORS 32768
2189
2190 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2191 {
2192 ThrottleConfig cfg;
2193
2194 throttle_config_init(&cfg);
2195 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2196
2197 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2198 blk_set_io_limits(blk, &cfg);
2199 }
2200
2201 static int img_convert(int argc, char **argv)
2202 {
2203 int c, bs_i, flags, src_flags = 0;
2204 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2205 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2206 *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2207 BlockDriver *drv = NULL, *proto_drv = NULL;
2208 BlockDriverInfo bdi;
2209 BlockDriverState *out_bs;
2210 QemuOpts *opts = NULL, *sn_opts = NULL;
2211 QemuOptsList *create_opts = NULL;
2212 QDict *open_opts = NULL;
2213 char *options = NULL;
2214 Error *local_err = NULL;
2215 bool writethrough, src_writethrough, image_opts = false,
2216 skip_create = false, progress = false, tgt_image_opts = false;
2217 int64_t ret = -EINVAL;
2218 bool force_share = false;
2219 bool explict_min_sparse = false;
2220 bool bitmaps = false;
2221 int64_t rate_limit = 0;
2222
2223 ImgConvertState s = (ImgConvertState) {
2224 /* Need at least 4k of zeros for sparse detection */
2225 .min_sparse = 8,
2226 .copy_range = false,
2227 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2228 .wr_in_order = true,
2229 .num_coroutines = 8,
2230 };
2231
2232 for(;;) {
2233 static const struct option long_options[] = {
2234 {"help", no_argument, 0, 'h'},
2235 {"object", required_argument, 0, OPTION_OBJECT},
2236 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2237 {"force-share", no_argument, 0, 'U'},
2238 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2239 {"salvage", no_argument, 0, OPTION_SALVAGE},
2240 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2241 {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2242 {0, 0, 0, 0}
2243 };
2244 c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WUr:",
2245 long_options, NULL);
2246 if (c == -1) {
2247 break;
2248 }
2249 switch(c) {
2250 case ':':
2251 missing_argument(argv[optind - 1]);
2252 break;
2253 case '?':
2254 unrecognized_option(argv[optind - 1]);
2255 break;
2256 case 'h':
2257 help();
2258 break;
2259 case 'f':
2260 fmt = optarg;
2261 break;
2262 case 'O':
2263 out_fmt = optarg;
2264 break;
2265 case 'B':
2266 out_baseimg = optarg;
2267 break;
2268 case 'C':
2269 s.copy_range = true;
2270 break;
2271 case 'c':
2272 s.compressed = true;
2273 break;
2274 case 'o':
2275 if (accumulate_options(&options, optarg) < 0) {
2276 goto fail_getopt;
2277 }
2278 break;
2279 case 'l':
2280 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2281 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2282 optarg, false);
2283 if (!sn_opts) {
2284 error_report("Failed in parsing snapshot param '%s'",
2285 optarg);
2286 goto fail_getopt;
2287 }
2288 } else {
2289 snapshot_name = optarg;
2290 }
2291 break;
2292 case 'S':
2293 {
2294 int64_t sval;
2295
2296 sval = cvtnum("buffer size for sparse output", optarg);
2297 if (sval < 0) {
2298 goto fail_getopt;
2299 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2300 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2301 error_report("Invalid buffer size for sparse output specified. "
2302 "Valid sizes are multiples of %llu up to %llu. Select "
2303 "0 to disable sparse detection (fully allocates output).",
2304 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2305 goto fail_getopt;
2306 }
2307
2308 s.min_sparse = sval / BDRV_SECTOR_SIZE;
2309 explict_min_sparse = true;
2310 break;
2311 }
2312 case 'p':
2313 progress = true;
2314 break;
2315 case 't':
2316 cache = optarg;
2317 break;
2318 case 'T':
2319 src_cache = optarg;
2320 break;
2321 case 'q':
2322 s.quiet = true;
2323 break;
2324 case 'n':
2325 skip_create = true;
2326 break;
2327 case 'm':
2328 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2329 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2330 error_report("Invalid number of coroutines. Allowed number of"
2331 " coroutines is between 1 and %d", MAX_COROUTINES);
2332 goto fail_getopt;
2333 }
2334 break;
2335 case 'W':
2336 s.wr_in_order = false;
2337 break;
2338 case 'U':
2339 force_share = true;
2340 break;
2341 case 'r':
2342 rate_limit = cvtnum("rate limit", optarg);
2343 if (rate_limit < 0) {
2344 goto fail_getopt;
2345 }
2346 break;
2347 case OPTION_OBJECT: {
2348 QemuOpts *object_opts;
2349 object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2350 optarg, true);
2351 if (!object_opts) {
2352 goto fail_getopt;
2353 }
2354 break;
2355 }
2356 case OPTION_IMAGE_OPTS:
2357 image_opts = true;
2358 break;
2359 case OPTION_SALVAGE:
2360 s.salvage = true;
2361 break;
2362 case OPTION_TARGET_IMAGE_OPTS:
2363 tgt_image_opts = true;
2364 break;
2365 case OPTION_TARGET_IS_ZERO:
2366 /*
2367 * The user asserting that the target is blank has the
2368 * same effect as the target driver supporting zero
2369 * initialisation.
2370 */
2371 s.has_zero_init = true;
2372 break;
2373 case OPTION_BITMAPS:
2374 bitmaps = true;
2375 break;
2376 }
2377 }
2378
2379 if (!out_fmt && !tgt_image_opts) {
2380 out_fmt = "raw";
2381 }
2382
2383 if (qemu_opts_foreach(&qemu_object_opts,
2384 user_creatable_add_opts_foreach,
2385 qemu_img_object_print_help, &error_fatal)) {
2386 goto fail_getopt;
2387 }
2388
2389 if (s.compressed && s.copy_range) {
2390 error_report("Cannot enable copy offloading when -c is used");
2391 goto fail_getopt;
2392 }
2393
2394 if (explict_min_sparse && s.copy_range) {
2395 error_report("Cannot enable copy offloading when -S is used");
2396 goto fail_getopt;
2397 }
2398
2399 if (s.copy_range && s.salvage) {
2400 error_report("Cannot use copy offloading in salvaging mode");
2401 goto fail_getopt;
2402 }
2403
2404 if (tgt_image_opts && !skip_create) {
2405 error_report("--target-image-opts requires use of -n flag");
2406 goto fail_getopt;
2407 }
2408
2409 if (skip_create && options) {
2410 error_report("-o has no effect when skipping image creation");
2411 goto fail_getopt;
2412 }
2413
2414 if (s.has_zero_init && !skip_create) {
2415 error_report("--target-is-zero requires use of -n flag");
2416 goto fail_getopt;
2417 }
2418
2419 s.src_num = argc - optind - 1;
2420 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2421
2422 if (options && has_help_option(options)) {
2423 if (out_fmt) {
2424 ret = print_block_option_help(out_filename, out_fmt);
2425 goto fail_getopt;
2426 } else {
2427 error_report("Option help requires a format be specified");
2428 goto fail_getopt;
2429 }
2430 }
2431
2432 if (s.src_num < 1) {
2433 error_report("Must specify image file name");
2434 goto fail_getopt;
2435 }
2436
2437 /* ret is still -EINVAL until here */
2438 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2439 if (ret < 0) {
2440 error_report("Invalid source cache option: %s", src_cache);
2441 goto fail_getopt;
2442 }
2443
2444 /* Initialize before goto out */
2445 if (s.quiet) {
2446 progress = false;
2447 }
2448 qemu_progress_init(progress, 1.0);
2449 qemu_progress_print(0, 100);
2450
2451 s.src = g_new0(BlockBackend *, s.src_num);
2452 s.src_sectors = g_new(int64_t, s.src_num);
2453 s.src_alignment = g_new(int, s.src_num);
2454
2455 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2456 BlockDriverState *src_bs;
2457 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2458 fmt, src_flags, src_writethrough, s.quiet,
2459 force_share);
2460 if (!s.src[bs_i]) {
2461 ret = -1;
2462 goto out;
2463 }
2464 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2465 if (s.src_sectors[bs_i] < 0) {
2466 error_report("Could not get size of %s: %s",
2467 argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2468 ret = -1;
2469 goto out;
2470 }
2471 src_bs = blk_bs(s.src[bs_i]);
2472 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2473 BDRV_SECTOR_SIZE);
2474 if (!bdrv_get_info(src_bs, &bdi)) {
2475 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2476 bdi.cluster_size / BDRV_SECTOR_SIZE);
2477 }
2478 s.total_sectors += s.src_sectors[bs_i];
2479 }
2480
2481 if (sn_opts) {
2482 bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2483 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2484 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2485 &local_err);
2486 } else if (snapshot_name != NULL) {
2487 if (s.src_num > 1) {
2488 error_report("No support for concatenating multiple snapshot");
2489 ret = -1;
2490 goto out;
2491 }
2492
2493 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2494 &local_err);
2495 }
2496 if (local_err) {
2497 error_reportf_err(local_err, "Failed to load snapshot: ");
2498 ret = -1;
2499 goto out;
2500 }
2501
2502 if (!skip_create) {
2503 /* Find driver and parse its options */
2504 drv = bdrv_find_format(out_fmt);
2505 if (!drv) {
2506 error_report("Unknown file format '%s'", out_fmt);
2507 ret = -1;
2508 goto out;
2509 }
2510
2511 proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2512 if (!proto_drv) {
2513 error_report_err(local_err);
2514 ret = -1;
2515 goto out;
2516 }
2517
2518 if (!drv->create_opts) {
2519 error_report("Format driver '%s' does not support image creation",
2520 drv->format_name);
2521 ret = -1;
2522 goto out;
2523 }
2524
2525 if (!proto_drv->create_opts) {
2526 error_report("Protocol driver '%s' does not support image creation",
2527 proto_drv->format_name);
2528 ret = -1;
2529 goto out;
2530 }
2531
2532 create_opts = qemu_opts_append(create_opts, drv->create_opts);
2533 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2534
2535 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2536 if (options) {
2537 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2538 error_report_err(local_err);
2539 ret = -1;
2540 goto out;
2541 }
2542 }
2543
2544 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2545 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2546 ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2547 if (ret < 0) {
2548 goto out;
2549 }
2550 }
2551
2552 /* Get backing file name if -o backing_file was used */
2553 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2554 if (out_baseimg_param) {
2555 out_baseimg = out_baseimg_param;
2556 }
2557 s.target_has_backing = (bool) out_baseimg;
2558
2559 if (s.has_zero_init && s.target_has_backing) {
2560 error_report("Cannot use --target-is-zero when the destination "
2561 "image has a backing file");
2562 goto out;
2563 }
2564
2565 if (s.src_num > 1 && out_baseimg) {
2566 error_report("Having a backing file for the target makes no sense when "
2567 "concatenating multiple input images");
2568 ret = -1;
2569 goto out;
2570 }
2571
2572 if (out_baseimg_param) {
2573 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2574 warn_report("Deprecated use of backing file without explicit "
2575 "backing format");
2576 }
2577 }
2578
2579 /* Check if compression is supported */
2580 if (s.compressed) {
2581 bool encryption =
2582 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2583 const char *encryptfmt =
2584 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2585 const char *preallocation =
2586 qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2587
2588 if (drv && !block_driver_can_compress(drv)) {
2589 error_report("Compression not supported for this file format");
2590 ret = -1;
2591 goto out;
2592 }
2593
2594 if (encryption || encryptfmt) {
2595 error_report("Compression and encryption not supported at "
2596 "the same time");
2597 ret = -1;
2598 goto out;
2599 }
2600
2601 if (preallocation
2602 && strcmp(preallocation, "off"))
2603 {
2604 error_report("Compression and preallocation not supported at "
2605 "the same time");
2606 ret = -1;
2607 goto out;
2608 }
2609 }
2610
2611 /* Determine if bitmaps need copying */
2612 if (bitmaps) {
2613 if (s.src_num > 1) {
2614 error_report("Copying bitmaps only possible with single source");
2615 ret = -1;
2616 goto out;
2617 }
2618 if (!bdrv_supports_persistent_dirty_bitmap(blk_bs(s.src[0]))) {
2619 error_report("Source lacks bitmap support");
2620 ret = -1;
2621 goto out;
2622 }
2623 }
2624
2625 /*
2626 * The later open call will need any decryption secrets, and
2627 * bdrv_create() will purge "opts", so extract them now before
2628 * they are lost.
2629 */
2630 if (!skip_create) {
2631 open_opts = qdict_new();
2632 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2633
2634 /* Create the new image */
2635 ret = bdrv_create(drv, out_filename, opts, &local_err);
2636 if (ret < 0) {
2637 error_reportf_err(local_err, "%s: error while converting %s: ",
2638 out_filename, out_fmt);
2639 goto out;
2640 }
2641 }
2642
2643 s.target_is_new = !skip_create;
2644
2645 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2646 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2647 if (ret < 0) {
2648 error_report("Invalid cache option: %s", cache);
2649 goto out;
2650 }
2651
2652 if (skip_create) {
2653 s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2654 flags, writethrough, s.quiet, false);
2655 } else {
2656 /* TODO ultimately we should allow --target-image-opts
2657 * to be used even when -n is not given.
2658 * That has to wait for bdrv_create to be improved
2659 * to allow filenames in option syntax
2660 */
2661 s.target = img_open_file(out_filename, open_opts, out_fmt,
2662 flags, writethrough, s.quiet, false);
2663 open_opts = NULL; /* blk_new_open will have freed it */
2664 }
2665 if (!s.target) {
2666 ret = -1;
2667 goto out;
2668 }
2669 out_bs = blk_bs(s.target);
2670
2671 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2672 error_report("Format driver '%s' does not support bitmaps",
2673 out_bs->drv->format_name);
2674 ret = -1;
2675 goto out;
2676 }
2677
2678 if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2679 error_report("Compression not supported for this file format");
2680 ret = -1;
2681 goto out;
2682 }
2683
2684 /* increase bufsectors from the default 4096 (2M) if opt_transfer
2685 * or discard_alignment of the out_bs is greater. Limit to
2686 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2687 s.buf_sectors = MIN(MAX_BUF_SECTORS,
2688 MAX(s.buf_sectors,
2689 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2690 out_bs->bl.pdiscard_alignment >>
2691 BDRV_SECTOR_BITS)));
2692
2693 /* try to align the write requests to the destination to avoid unnecessary
2694 * RMW cycles. */
2695 s.alignment = MAX(pow2floor(s.min_sparse),
2696 DIV_ROUND_UP(out_bs->bl.request_alignment,
2697 BDRV_SECTOR_SIZE));
2698 assert(is_power_of_2(s.alignment));
2699
2700 if (skip_create) {
2701 int64_t output_sectors = blk_nb_sectors(s.target);
2702 if (output_sectors < 0) {
2703 error_report("unable to get output image length: %s",
2704 strerror(-output_sectors));
2705 ret = -1;
2706 goto out;
2707 } else if (output_sectors < s.total_sectors) {
2708 error_report("output file is smaller than input file");
2709 ret = -1;
2710 goto out;
2711 }
2712 }
2713
2714 if (s.target_has_backing && s.target_is_new) {
2715 /* Errors are treated as "backing length unknown" (which means
2716 * s.target_backing_sectors has to be negative, which it will
2717 * be automatically). The backing file length is used only
2718 * for optimizations, so such a case is not fatal. */
2719 s.target_backing_sectors =
2720 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2721 } else {
2722 s.target_backing_sectors = -1;
2723 }
2724
2725 ret = bdrv_get_info(out_bs, &bdi);
2726 if (ret < 0) {
2727 if (s.compressed) {
2728 error_report("could not get block driver info");
2729 goto out;
2730 }
2731 } else {
2732 s.compressed = s.compressed || bdi.needs_compressed_writes;
2733 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2734 }
2735
2736 if (rate_limit) {
2737 set_rate_limit(s.target, rate_limit);
2738 }
2739
2740 ret = convert_do_copy(&s);
2741
2742 /* Now copy the bitmaps */
2743 if (bitmaps && ret == 0) {
2744 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs);
2745 }
2746
2747 out:
2748 if (!ret) {
2749 qemu_progress_print(100, 0);
2750 }
2751 qemu_progress_end();
2752 qemu_opts_del(opts);
2753 qemu_opts_free(create_opts);
2754 qobject_unref(open_opts);
2755 blk_unref(s.target);
2756 if (s.src) {
2757 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2758 blk_unref(s.src[bs_i]);
2759 }
2760 g_free(s.src);
2761 }
2762 g_free(s.src_sectors);
2763 g_free(s.src_alignment);
2764 fail_getopt:
2765 qemu_opts_del(sn_opts);
2766 g_free(options);
2767
2768 return !!ret;
2769 }
2770
2771
2772 static void dump_snapshots(BlockDriverState *bs)
2773 {
2774 QEMUSnapshotInfo *sn_tab, *sn;
2775 int nb_sns, i;
2776
2777 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2778 if (nb_sns <= 0)
2779 return;
2780 printf("Snapshot list:\n");
2781 bdrv_snapshot_dump(NULL);
2782 printf("\n");
2783 for(i = 0; i < nb_sns; i++) {
2784 sn = &sn_tab[i];
2785 bdrv_snapshot_dump(sn);
2786 printf("\n");
2787 }
2788 g_free(sn_tab);
2789 }
2790
2791 static void dump_json_image_info_list(ImageInfoList *list)
2792 {
2793 QString *str;
2794 QObject *obj;
2795 Visitor *v = qobject_output_visitor_new(&obj);
2796
2797 visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2798 visit_complete(v, &obj);
2799 str = qobject_to_json_pretty(obj);
2800 assert(str != NULL);
2801 printf("%s\n", qstring_get_str(str));
2802 qobject_unref(obj);
2803 visit_free(v);
2804 qobject_unref(str);
2805 }
2806
2807 static void dump_json_image_info(ImageInfo *info)
2808 {
2809 QString *str;
2810 QObject *obj;
2811 Visitor *v = qobject_output_visitor_new(&obj);
2812
2813 visit_type_ImageInfo(v, NULL, &info, &error_abort);
2814 visit_complete(v, &obj);
2815 str = qobject_to_json_pretty(obj);
2816 assert(str != NULL);
2817 printf("%s\n", qstring_get_str(str));
2818 qobject_unref(obj);
2819 visit_free(v);
2820 qobject_unref(str);
2821 }
2822
2823 static void dump_human_image_info_list(ImageInfoList *list)
2824 {
2825 ImageInfoList *elem;
2826 bool delim = false;
2827
2828 for (elem = list; elem; elem = elem->next) {
2829 if (delim) {
2830 printf("\n");
2831 }
2832 delim = true;
2833
2834 bdrv_image_info_dump(elem->value);
2835 }
2836 }
2837
2838 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2839 {
2840 return strcmp(a, b) == 0;
2841 }
2842
2843 /**
2844 * Open an image file chain and return an ImageInfoList
2845 *
2846 * @filename: topmost image filename
2847 * @fmt: topmost image format (may be NULL to autodetect)
2848 * @chain: true - enumerate entire backing file chain
2849 * false - only topmost image file
2850 *
2851 * Returns a list of ImageInfo objects or NULL if there was an error opening an
2852 * image file. If there was an error a message will have been printed to
2853 * stderr.
2854 */
2855 static ImageInfoList *collect_image_info_list(bool image_opts,
2856 const char *filename,
2857 const char *fmt,
2858 bool chain, bool force_share)
2859 {
2860 ImageInfoList *head = NULL;
2861 ImageInfoList **last = &head;
2862 GHashTable *filenames;
2863 Error *err = NULL;
2864
2865 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2866
2867 while (filename) {
2868 BlockBackend *blk;
2869 BlockDriverState *bs;
2870 ImageInfo *info;
2871 ImageInfoList *elem;
2872
2873 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2874 error_report("Backing file '%s' creates an infinite loop.",
2875 filename);
2876 goto err;
2877 }
2878 g_hash_table_insert(filenames, (gpointer)filename, NULL);
2879
2880 blk = img_open(image_opts, filename, fmt,
2881 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2882 force_share);
2883 if (!blk) {
2884 goto err;
2885 }
2886 bs = blk_bs(blk);
2887
2888 bdrv_query_image_info(bs, &info, &err);
2889 if (err) {
2890 error_report_err(err);
2891 blk_unref(blk);
2892 goto err;
2893 }
2894
2895 elem = g_new0(ImageInfoList, 1);
2896 elem->value = info;
2897 *last = elem;
2898 last = &elem->next;
2899
2900 blk_unref(blk);
2901
2902 /* Clear parameters that only apply to the topmost image */
2903 filename = fmt = NULL;
2904 image_opts = false;
2905
2906 if (chain) {
2907 if (info->has_full_backing_filename) {
2908 filename = info->full_backing_filename;
2909 } else if (info->has_backing_filename) {
2910 error_report("Could not determine absolute backing filename,"
2911 " but backing filename '%s' present",
2912 info->backing_filename);
2913 goto err;
2914 }
2915 if (info->has_backing_filename_format) {
2916 fmt = info->backing_filename_format;
2917 }
2918 }
2919 }
2920 g_hash_table_destroy(filenames);
2921 return head;
2922
2923 err:
2924 qapi_free_ImageInfoList(head);
2925 g_hash_table_destroy(filenames);
2926 return NULL;
2927 }
2928
2929 static int img_info(int argc, char **argv)
2930 {
2931 int c;
2932 OutputFormat output_format = OFORMAT_HUMAN;
2933 bool chain = false;
2934 const char *filename, *fmt, *output;
2935 ImageInfoList *list;
2936 bool image_opts = false;
2937 bool force_share = false;
2938
2939 fmt = NULL;
2940 output = NULL;
2941 for(;;) {
2942 int option_index = 0;
2943 static const struct option long_options[] = {
2944 {"help", no_argument, 0, 'h'},
2945 {"format", required_argument, 0, 'f'},
2946 {"output", required_argument, 0, OPTION_OUTPUT},
2947 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2948 {"object", required_argument, 0, OPTION_OBJECT},
2949 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2950 {"force-share", no_argument, 0, 'U'},
2951 {0, 0, 0, 0}
2952 };
2953 c = getopt_long(argc, argv, ":f:hU",
2954 long_options, &option_index);
2955 if (c == -1) {
2956 break;
2957 }
2958 switch(c) {
2959 case ':':
2960 missing_argument(argv[optind - 1]);
2961 break;
2962 case '?':
2963 unrecognized_option(argv[optind - 1]);
2964 break;
2965 case 'h':
2966 help();
2967 break;
2968 case 'f':
2969 fmt = optarg;
2970 break;
2971 case 'U':
2972 force_share = true;
2973 break;
2974 case OPTION_OUTPUT:
2975 output = optarg;
2976 break;
2977 case OPTION_BACKING_CHAIN:
2978 chain = true;
2979 break;
2980 case OPTION_OBJECT: {
2981 QemuOpts *opts;
2982 opts = qemu_opts_parse_noisily(&qemu_object_opts,
2983 optarg, true);
2984 if (!opts) {
2985 return 1;
2986 }
2987 } break;
2988 case OPTION_IMAGE_OPTS:
2989 image_opts = true;
2990 break;
2991 }
2992 }
2993 if (optind != argc - 1) {
2994 error_exit("Expecting one image file name");
2995 }
2996 filename = argv[optind++];
2997
2998 if (output && !strcmp(output, "json")) {
2999 output_format = OFORMAT_JSON;
3000 } else if (output && !strcmp(output, "human")) {
3001 output_format = OFORMAT_HUMAN;
3002 } else if (output) {
3003 error_report("--output must be used with human or json as argument.");
3004 return 1;
3005 }
3006
3007 if (qemu_opts_foreach(&qemu_object_opts,
3008 user_creatable_add_opts_foreach,
3009 qemu_img_object_print_help, &error_fatal)) {
3010 return 1;
3011 }
3012
3013 list = collect_image_info_list(image_opts, filename, fmt, chain,
3014 force_share);
3015 if (!list) {
3016 return 1;
3017 }
3018
3019 switch (output_format) {
3020 case OFORMAT_HUMAN:
3021 dump_human_image_info_list(list);
3022 break;
3023 case OFORMAT_JSON:
3024 if (chain) {
3025 dump_json_image_info_list(list);
3026 } else {
3027 dump_json_image_info(list->value);
3028 }
3029 break;
3030 }
3031
3032 qapi_free_ImageInfoList(list);
3033 return 0;
3034 }
3035
3036 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
3037 MapEntry *next)
3038 {
3039 switch (output_format) {
3040 case OFORMAT_HUMAN:
3041 if (e->data && !e->has_offset) {
3042 error_report("File contains external, encrypted or compressed clusters.");
3043 return -1;
3044 }
3045 if (e->data && !e->zero) {
3046 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
3047 e->start, e->length,
3048 e->has_offset ? e->offset : 0,
3049 e->has_filename ? e->filename : "");
3050 }
3051 /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
3052 * Modify the flags here to allow more coalescing.
3053 */
3054 if (next && (!next->data || next->zero)) {
3055 next->data = false;
3056 next->zero = true;
3057 }
3058 break;
3059 case OFORMAT_JSON:
3060 printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
3061 " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
3062 e->start, e->length, e->depth,
3063 e->zero ? "true" : "false",
3064 e->data ? "true" : "false");
3065 if (e->has_offset) {
3066 printf(", \"offset\": %"PRId64"", e->offset);
3067 }
3068 putchar('}');
3069
3070 if (next) {
3071 puts(",");
3072 }
3073 break;
3074 }
3075 return 0;
3076 }
3077
3078 static int get_block_status(BlockDriverState *bs, int64_t offset,
3079 int64_t bytes, MapEntry *e)
3080 {
3081 int ret;
3082 int depth;
3083 BlockDriverState *file;
3084 bool has_offset;
3085 int64_t map;
3086 char *filename = NULL;
3087
3088 /* As an optimization, we could cache the current range of unallocated
3089 * clusters in each file of the chain, and avoid querying the same
3090 * range repeatedly.
3091 */
3092
3093 depth = 0;
3094 for (;;) {
3095 bs = bdrv_skip_filters(bs);
3096 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3097 if (ret < 0) {
3098 return ret;
3099 }
3100 assert(bytes);
3101 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3102 break;
3103 }
3104 bs = bdrv_cow_bs(bs);
3105 if (bs == NULL) {
3106 ret = 0;
3107 break;
3108 }
3109
3110 depth++;
3111 }
3112
3113 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3114
3115 if (file && has_offset) {
3116 bdrv_refresh_filename(file);
3117 filename = file->filename;
3118 }
3119
3120 *e = (MapEntry) {
3121 .start = offset,
3122 .length = bytes,
3123 .data = !!(ret & BDRV_BLOCK_DATA),
3124 .zero = !!(ret & BDRV_BLOCK_ZERO),
3125 .offset = map,
3126 .has_offset = has_offset,
3127 .depth = depth,
3128 .has_filename = filename,
3129 .filename = filename,
3130 };
3131
3132 return 0;
3133 }
3134
3135 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3136 {
3137 if (curr->length == 0) {
3138 return false;
3139 }
3140 if (curr->zero != next->zero ||
3141 curr->data != next->data ||
3142 curr->depth != next->depth ||
3143 curr->has_filename != next->has_filename ||
3144 curr->has_offset != next->has_offset) {
3145 return false;
3146 }
3147 if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3148 return false;
3149 }
3150 if (curr->has_offset && curr->offset + curr->length != next->offset) {
3151 return false;
3152 }
3153 return true;
3154 }
3155
3156 static int img_map(int argc, char **argv)
3157 {
3158 int c;
3159 OutputFormat output_format = OFORMAT_HUMAN;
3160 BlockBackend *blk;
3161 BlockDriverState *bs;
3162 const char *filename, *fmt, *output;
3163 int64_t length;
3164 MapEntry curr = { .length = 0 }, next;
3165 int ret = 0;
3166 bool image_opts = false;
3167 bool force_share = false;
3168 int64_t start_offset = 0;
3169 int64_t max_length = -1;
3170
3171 fmt = NULL;
3172 output = NULL;
3173 for (;;) {
3174 int option_index = 0;
3175 static const struct option long_options[] = {
3176 {"help", no_argument, 0, 'h'},
3177 {"format", required_argument, 0, 'f'},
3178 {"output", required_argument, 0, OPTION_OUTPUT},
3179 {"object", required_argument, 0, OPTION_OBJECT},
3180 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3181 {"force-share", no_argument, 0, 'U'},
3182 {"start-offset", required_argument, 0, 's'},
3183 {"max-length", required_argument, 0, 'l'},
3184 {0, 0, 0, 0}
3185 };
3186 c = getopt_long(argc, argv, ":f:s:l:hU",
3187 long_options, &option_index);
3188 if (c == -1) {
3189 break;
3190 }
3191 switch (c) {
3192 case ':':
3193 missing_argument(argv[optind - 1]);
3194 break;
3195 case '?':
3196 unrecognized_option(argv[optind - 1]);
3197 break;
3198 case 'h':
3199 help();
3200 break;
3201 case 'f':
3202 fmt = optarg;
3203 break;
3204 case 'U':
3205 force_share = true;
3206 break;
3207 case OPTION_OUTPUT:
3208 output = optarg;
3209 break;
3210 case 's':
3211 start_offset = cvtnum("start offset", optarg);
3212 if (start_offset < 0) {
3213 return 1;
3214 }
3215 break;
3216 case 'l':
3217 max_length = cvtnum("max length", optarg);
3218 if (max_length < 0) {
3219 return 1;
3220 }
3221 break;
3222 case OPTION_OBJECT: {
3223 QemuOpts *opts;
3224 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3225 optarg, true);
3226 if (!opts) {
3227 return 1;
3228 }
3229 } break;
3230 case OPTION_IMAGE_OPTS:
3231 image_opts = true;
3232 break;
3233 }
3234 }
3235 if (optind != argc - 1) {
3236 error_exit("Expecting one image file name");
3237 }
3238 filename = argv[optind];
3239
3240 if (output && !strcmp(output, "json")) {
3241 output_format = OFORMAT_JSON;
3242 } else if (output && !strcmp(output, "human")) {
3243 output_format = OFORMAT_HUMAN;
3244 } else if (output) {
3245 error_report("--output must be used with human or json as argument.");
3246 return 1;
3247 }
3248
3249 if (qemu_opts_foreach(&qemu_object_opts,
3250 user_creatable_add_opts_foreach,
3251 qemu_img_object_print_help, &error_fatal)) {
3252 return 1;
3253 }
3254
3255 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3256 if (!blk) {
3257 return 1;
3258 }
3259 bs = blk_bs(blk);
3260
3261 if (output_format == OFORMAT_HUMAN) {
3262 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3263 } else if (output_format == OFORMAT_JSON) {
3264 putchar('[');
3265 }
3266
3267 length = blk_getlength(blk);
3268 if (length < 0) {
3269 error_report("Failed to get size for '%s'", filename);
3270 return 1;
3271 }
3272 if (max_length != -1) {
3273 length = MIN(start_offset + max_length, length);
3274 }
3275
3276 curr.start = start_offset;
3277 while (curr.start + curr.length < length) {
3278 int64_t offset = curr.start + curr.length;
3279 int64_t n = length - offset;
3280
3281 ret = get_block_status(bs, offset, n, &next);
3282 if (ret < 0) {
3283 error_report("Could not read file metadata: %s", strerror(-ret));
3284 goto out;
3285 }
3286
3287 if (entry_mergeable(&curr, &next)) {
3288 curr.length += next.length;
3289 continue;
3290 }
3291
3292 if (curr.length > 0) {
3293 ret = dump_map_entry(output_format, &curr, &next);
3294 if (ret < 0) {
3295 goto out;
3296 }
3297 }
3298 curr = next;
3299 }
3300
3301 ret = dump_map_entry(output_format, &curr, NULL);
3302 if (output_format == OFORMAT_JSON) {
3303 puts("]");
3304 }
3305
3306 out:
3307 blk_unref(blk);
3308 return ret < 0;
3309 }
3310
3311 #define SNAPSHOT_LIST 1
3312 #define SNAPSHOT_CREATE 2
3313 #define SNAPSHOT_APPLY 3
3314 #define SNAPSHOT_DELETE 4
3315
3316 static int img_snapshot(int argc, char **argv)
3317 {
3318 BlockBackend *blk;
3319 BlockDriverState *bs;
3320 QEMUSnapshotInfo sn;
3321 char *filename, *snapshot_name = NULL;
3322 int c, ret = 0, bdrv_oflags;
3323 int action = 0;
3324 qemu_timeval tv;
3325 bool quiet = false;
3326 Error *err = NULL;
3327 bool image_opts = false;
3328 bool force_share = false;
3329
3330 bdrv_oflags = BDRV_O_RDWR;
3331 /* Parse commandline parameters */
3332 for(;;) {
3333 static const struct option long_options[] = {
3334 {"help", no_argument, 0, 'h'},
3335 {"object", required_argument, 0, OPTION_OBJECT},
3336 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3337 {"force-share", no_argument, 0, 'U'},
3338 {0, 0, 0, 0}
3339 };
3340 c = getopt_long(argc, argv, ":la:c:d:hqU",
3341 long_options, NULL);
3342 if (c == -1) {
3343 break;
3344 }
3345 switch(c) {
3346 case ':':
3347 missing_argument(argv[optind - 1]);
3348 break;
3349 case '?':
3350 unrecognized_option(argv[optind - 1]);
3351 break;
3352 case 'h':
3353 help();
3354 return 0;
3355 case 'l':
3356 if (action) {
3357 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3358 return 0;
3359 }
3360 action = SNAPSHOT_LIST;
3361 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3362 break;
3363 case 'a':
3364 if (action) {
3365 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3366 return 0;
3367 }
3368 action = SNAPSHOT_APPLY;
3369 snapshot_name = optarg;
3370 break;
3371 case 'c':
3372 if (action) {
3373 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3374 return 0;
3375 }
3376 action = SNAPSHOT_CREATE;
3377 snapshot_name = optarg;
3378 break;
3379 case 'd':
3380 if (action) {
3381 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3382 return 0;
3383 }
3384 action = SNAPSHOT_DELETE;
3385 snapshot_name = optarg;
3386 break;
3387 case 'q':
3388 quiet = true;
3389 break;
3390 case 'U':
3391 force_share = true;
3392 break;
3393 case OPTION_OBJECT: {
3394 QemuOpts *opts;
3395 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3396 optarg, true);
3397 if (!opts) {
3398 return 1;
3399 }
3400 } break;
3401 case OPTION_IMAGE_OPTS:
3402 image_opts = true;
3403 break;
3404 }
3405 }
3406
3407 if (optind != argc - 1) {
3408 error_exit("Expecting one image file name");
3409 }
3410 filename = argv[optind++];
3411
3412 if (qemu_opts_foreach(&qemu_object_opts,
3413 user_creatable_add_opts_foreach,
3414 qemu_img_object_print_help, &error_fatal)) {
3415 return 1;
3416 }
3417
3418 /* Open the image */
3419 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3420 force_share);
3421 if (!blk) {
3422 return 1;
3423 }
3424 bs = blk_bs(blk);
3425
3426 /* Perform the requested action */
3427 switch(action) {
3428 case SNAPSHOT_LIST:
3429 dump_snapshots(bs);
3430 break;
3431
3432 case SNAPSHOT_CREATE:
3433 memset(&sn, 0, sizeof(sn));
3434 pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3435
3436 qemu_gettimeofday(&tv);
3437 sn.date_sec = tv.tv_sec;
3438 sn.date_nsec = tv.tv_usec * 1000;
3439
3440 ret = bdrv_snapshot_create(bs, &sn);
3441 if (ret) {
3442 error_report("Could not create snapshot '%s': %d (%s)",
3443 snapshot_name, ret, strerror(-ret));
3444 }
3445 break;
3446
3447 case SNAPSHOT_APPLY:
3448 ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3449 if (ret) {
3450 error_reportf_err(err, "Could not apply snapshot '%s': ",
3451 snapshot_name);
3452 }
3453 break;
3454
3455 case SNAPSHOT_DELETE:
3456 ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3457 if (ret < 0) {
3458 error_report("Could not delete snapshot '%s': snapshot not "
3459 "found", snapshot_name);
3460 ret = 1;
3461 } else {
3462 ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3463 if (ret < 0) {
3464 error_reportf_err(err, "Could not delete snapshot '%s': ",
3465 snapshot_name);
3466 ret = 1;
3467 }
3468 }
3469 break;
3470 }
3471
3472 /* Cleanup */
3473 blk_unref(blk);
3474 if (ret) {
3475 return 1;
3476 }
3477 return 0;
3478 }
3479
3480 static int img_rebase(int argc, char **argv)
3481 {
3482 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3483 uint8_t *buf_old = NULL;
3484 uint8_t *buf_new = NULL;
3485 BlockDriverState *bs = NULL, *prefix_chain_bs = NULL;
3486 BlockDriverState *unfiltered_bs;
3487 char *filename;
3488 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3489 int c, flags, src_flags, ret;
3490 bool writethrough, src_writethrough;
3491 int unsafe = 0;
3492 bool force_share = false;
3493 int progress = 0;
3494 bool quiet = false;
3495 Error *local_err = NULL;
3496 bool image_opts = false;
3497
3498 /* Parse commandline parameters */
3499 fmt = NULL;
3500 cache = BDRV_DEFAULT_CACHE;
3501 src_cache = BDRV_DEFAULT_CACHE;
3502 out_baseimg = NULL;
3503 out_basefmt = NULL;
3504 for(;;) {
3505 static const struct option long_options[] = {
3506 {"help", no_argument, 0, 'h'},
3507 {"object", required_argument, 0, OPTION_OBJECT},
3508 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3509 {"force-share", no_argument, 0, 'U'},
3510 {0, 0, 0, 0}
3511 };
3512 c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3513 long_options, NULL);
3514 if (c == -1) {
3515 break;
3516 }
3517 switch(c) {
3518 case ':':
3519 missing_argument(argv[optind - 1]);
3520 break;
3521 case '?':
3522 unrecognized_option(argv[optind - 1]);
3523 break;
3524 case 'h':
3525 help();
3526 return 0;
3527 case 'f':
3528 fmt = optarg;
3529 break;
3530 case 'F':
3531 out_basefmt = optarg;
3532 break;
3533 case 'b':
3534 out_baseimg = optarg;
3535 break;
3536 case 'u':
3537 unsafe = 1;
3538 break;
3539 case 'p':
3540 progress = 1;
3541 break;
3542 case 't':
3543 cache = optarg;
3544 break;
3545 case 'T':
3546 src_cache = optarg;
3547 break;
3548 case 'q':
3549 quiet = true;
3550 break;
3551 case OPTION_OBJECT: {
3552 QemuOpts *opts;
3553 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3554 optarg, true);
3555 if (!opts) {
3556 return 1;
3557 }
3558 } break;
3559 case OPTION_IMAGE_OPTS:
3560 image_opts = true;
3561 break;
3562 case 'U':
3563 force_share = true;
3564 break;
3565 }
3566 }
3567
3568 if (quiet) {
3569 progress = 0;
3570 }
3571
3572 if (optind != argc - 1) {
3573 error_exit("Expecting one image file name");
3574 }
3575 if (!unsafe && !out_baseimg) {
3576 error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3577 }
3578 filename = argv[optind++];
3579
3580 if (qemu_opts_foreach(&qemu_object_opts,
3581 user_creatable_add_opts_foreach,
3582 qemu_img_object_print_help, &error_fatal)) {
3583 return 1;
3584 }
3585
3586 qemu_progress_init(progress, 2.0);
3587 qemu_progress_print(0, 100);
3588
3589 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3590 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3591 if (ret < 0) {
3592 error_report("Invalid cache option: %s", cache);
3593 goto out;
3594 }
3595
3596 src_flags = 0;
3597 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3598 if (ret < 0) {
3599 error_report("Invalid source cache option: %s", src_cache);
3600 goto out;
3601 }
3602
3603 /* The source files are opened read-only, don't care about WCE */
3604 assert((src_flags & BDRV_O_RDWR) == 0);
3605 (void) src_writethrough;
3606
3607 /*
3608 * Open the images.
3609 *
3610 * Ignore the old backing file for unsafe rebase in case we want to correct
3611 * the reference to a renamed or moved backing file.
3612 */
3613 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3614 false);
3615 if (!blk) {
3616 ret = -1;
3617 goto out;
3618 }
3619 bs = blk_bs(blk);
3620
3621 unfiltered_bs = bdrv_skip_filters(bs);
3622
3623 if (out_basefmt != NULL) {
3624 if (bdrv_find_format(out_basefmt) == NULL) {
3625 error_report("Invalid format name: '%s'", out_basefmt);
3626 ret = -1;
3627 goto out;
3628 }
3629 }
3630
3631 /* For safe rebasing we need to compare old and new backing file */
3632 if (!unsafe) {
3633 QDict *options = NULL;
3634 BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs);
3635
3636 if (base_bs) {
3637 blk_old_backing = blk_new(qemu_get_aio_context(),
3638 BLK_PERM_CONSISTENT_READ,
3639 BLK_PERM_ALL);
3640 ret = blk_insert_bs(blk_old_backing, base_bs,
3641 &local_err);
3642 if (ret < 0) {
3643 error_reportf_err(local_err,
3644 "Could not reuse old backing file '%s': ",
3645 base_bs->filename);
3646 goto out;
3647 }
3648 } else {
3649 blk_old_backing = NULL;
3650 }
3651
3652 if (out_baseimg[0]) {
3653 const char *overlay_filename;
3654 char *out_real_path;
3655
3656 options = qdict_new();
3657 if (out_basefmt) {
3658 qdict_put_str(options, "driver", out_basefmt);
3659 }
3660 if (force_share) {
3661 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3662 }
3663
3664 bdrv_refresh_filename(bs);
3665 overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3666 : bs->filename;
3667 out_real_path =
3668 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3669 out_baseimg,
3670 &local_err);
3671 if (local_err) {
3672 qobject_unref(options);
3673 error_reportf_err(local_err,
3674 "Could not resolve backing filename: ");
3675 ret = -1;
3676 goto out;
3677 }
3678
3679 /*
3680 * Find out whether we rebase an image on top of a previous image
3681 * in its chain.
3682 */
3683 prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path);
3684 if (prefix_chain_bs) {
3685 qobject_unref(options);
3686 g_free(out_real_path);
3687
3688 blk_new_backing = blk_new(qemu_get_aio_context(),
3689 BLK_PERM_CONSISTENT_READ,
3690 BLK_PERM_ALL);
3691 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs,
3692 &local_err);
3693 if (ret < 0) {
3694 error_reportf_err(local_err,
3695 "Could not reuse backing file '%s': ",
3696 out_baseimg);
3697 goto out;
3698 }
3699 } else {
3700 blk_new_backing = blk_new_open(out_real_path, NULL,
3701 options, src_flags, &local_err);
3702 g_free(out_real_path);
3703 if (!blk_new_backing) {
3704 error_reportf_err(local_err,
3705 "Could not open new backing file '%s': ",
3706 out_baseimg);
3707 ret = -1;
3708 goto out;
3709 }
3710 }
3711 }
3712 }
3713
3714 /*
3715 * Check each unallocated cluster in the COW file. If it is unallocated,
3716 * accesses go to the backing file. We must therefore compare this cluster
3717 * in the old and new backing file, and if they differ we need to copy it
3718 * from the old backing file into the COW file.
3719 *
3720 * If qemu-img crashes during this step, no harm is done. The content of
3721 * the image is the same as the original one at any time.
3722 */
3723 if (!unsafe) {
3724 int64_t size;
3725 int64_t old_backing_size = 0;
3726 int64_t new_backing_size = 0;
3727 uint64_t offset;
3728 int64_t n;
3729 float local_progress = 0;
3730
3731 buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3732 buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3733
3734 size = blk_getlength(blk);
3735 if (size < 0) {
3736 error_report("Could not get size of '%s': %s",
3737 filename, strerror(-size));
3738 ret = -1;
3739 goto out;
3740 }
3741 if (blk_old_backing) {
3742 old_backing_size = blk_getlength(blk_old_backing);
3743 if (old_backing_size < 0) {
3744 char backing_name[PATH_MAX];
3745
3746 bdrv_get_backing_filename(bs, backing_name,
3747 sizeof(backing_name));
3748 error_report("Could not get size of '%s': %s",
3749 backing_name, strerror(-old_backing_size));
3750 ret = -1;
3751 goto out;
3752 }
3753 }
3754 if (blk_new_backing) {
3755 new_backing_size = blk_getlength(blk_new_backing);
3756 if (new_backing_size < 0) {
3757 error_report("Could not get size of '%s': %s",
3758 out_baseimg, strerror(-new_backing_size));
3759 ret = -1;
3760 goto out;
3761 }
3762 }
3763
3764 if (size != 0) {
3765 local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3766 }
3767
3768 for (offset = 0; offset < size; offset += n) {
3769 bool buf_old_is_zero = false;
3770
3771 /* How many bytes can we handle with the next read? */
3772 n = MIN(IO_BUF_SIZE, size - offset);
3773
3774 /* If the cluster is allocated, we don't need to take action */
3775 ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n);
3776 if (ret < 0) {
3777 error_report("error while reading image metadata: %s",
3778 strerror(-ret));
3779 goto out;
3780 }
3781 if (ret) {
3782 continue;
3783 }
3784
3785 if (prefix_chain_bs) {
3786 /*
3787 * If cluster wasn't changed since prefix_chain, we don't need
3788 * to take action
3789 */
3790 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs),
3791 prefix_chain_bs, false,
3792 offset, n, &n);
3793 if (ret < 0) {
3794 error_report("error while reading image metadata: %s",
3795 strerror(-ret));
3796 goto out;
3797 }
3798 if (!ret) {
3799 continue;
3800 }
3801 }
3802
3803 /*
3804 * Read old and new backing file and take into consideration that
3805 * backing files may be smaller than the COW image.
3806 */
3807 if (offset >= old_backing_size) {
3808 memset(buf_old, 0, n);
3809 buf_old_is_zero = true;
3810 } else {
3811 if (offset + n > old_backing_size) {
3812 n = old_backing_size - offset;
3813 }
3814
3815 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3816 if (ret < 0) {
3817 error_report("error while reading from old backing file");
3818 goto out;
3819 }
3820 }
3821
3822 if (offset >= new_backing_size || !blk_new_backing) {
3823 memset(buf_new, 0, n);
3824 } else {
3825 if (offset + n > new_backing_size) {
3826 n = new_backing_size - offset;
3827 }
3828
3829 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3830 if (ret < 0) {
3831 error_report("error while reading from new backing file");
3832 goto out;
3833 }
3834 }
3835
3836 /* If they differ, we need to write to the COW file */
3837 uint64_t written = 0;
3838
3839 while (written < n) {
3840 int64_t pnum;
3841
3842 if (compare_buffers(buf_old + written, buf_new + written,
3843 n - written, &pnum))
3844 {
3845 if (buf_old_is_zero) {
3846 ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3847 } else {
3848 ret = blk_pwrite(blk, offset + written,
3849 buf_old + written, pnum, 0);
3850 }
3851 if (ret < 0) {
3852 error_report("Error while writing to COW image: %s",
3853 strerror(-ret));
3854 goto out;
3855 }
3856 }
3857
3858 written += pnum;
3859 }
3860 qemu_progress_print(local_progress, 100);
3861 }
3862 }
3863
3864 /*
3865 * Change the backing file. All clusters that are different from the old
3866 * backing file are overwritten in the COW file now, so the visible content
3867 * doesn't change when we switch the backing file.
3868 */
3869 if (out_baseimg && *out_baseimg) {
3870 ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt,
3871 true);
3872 } else {
3873 ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false);
3874 }
3875
3876 if (ret == -ENOSPC) {
3877 error_report("Could not change the backing file to '%s': No "
3878 "space left in the file header", out_baseimg);
3879 } else if (ret < 0) {
3880 error_report("Could not change the backing file to '%s': %s",
3881 out_baseimg, strerror(-ret));
3882 }
3883
3884 qemu_progress_print(100, 0);
3885 /*
3886 * TODO At this point it is possible to check if any clusters that are
3887 * allocated in the COW file are the same in the backing file. If so, they
3888 * could be dropped from the COW file. Don't do this before switching the
3889 * backing file, in case of a crash this would lead to corruption.
3890 */
3891 out:
3892 qemu_progress_end();
3893 /* Cleanup */
3894 if (!unsafe) {
3895 blk_unref(blk_old_backing);
3896 blk_unref(blk_new_backing);
3897 }
3898 qemu_vfree(buf_old);
3899 qemu_vfree(buf_new);
3900
3901 blk_unref(blk);
3902 if (ret) {
3903 return 1;
3904 }
3905 return 0;
3906 }
3907
3908 static int img_resize(int argc, char **argv)
3909 {
3910 Error *err = NULL;
3911 int c, ret, relative;
3912 const char *filename, *fmt, *size;
3913 int64_t n, total_size, current_size;
3914 bool quiet = false;
3915 BlockBackend *blk = NULL;
3916 PreallocMode prealloc = PREALLOC_MODE_OFF;
3917 QemuOpts *param;
3918
3919 static QemuOptsList resize_options = {
3920 .name = "resize_options",
3921 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3922 .desc = {
3923 {
3924 .name = BLOCK_OPT_SIZE,
3925 .type = QEMU_OPT_SIZE,
3926 .help = "Virtual disk size"
3927 }, {
3928 /* end of list */
3929 }
3930 },
3931 };
3932 bool image_opts = false;
3933 bool shrink = false;
3934
3935 /* Remove size from argv manually so that negative numbers are not treated
3936 * as options by getopt. */
3937 if (argc < 3) {
3938 error_exit("Not enough arguments");
3939 return 1;
3940 }
3941
3942 size = argv[--argc];
3943
3944 /* Parse getopt arguments */
3945 fmt = NULL;
3946 for(;;) {
3947 static const struct option long_options[] = {
3948 {"help", no_argument, 0, 'h'},
3949 {"object", required_argument, 0, OPTION_OBJECT},
3950 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3951 {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3952 {"shrink", no_argument, 0, OPTION_SHRINK},
3953 {0, 0, 0, 0}
3954 };
3955 c = getopt_long(argc, argv, ":f:hq",
3956 long_options, NULL);
3957 if (c == -1) {
3958 break;
3959 }
3960 switch(c) {
3961 case ':':
3962 missing_argument(argv[optind - 1]);
3963 break;
3964 case '?':
3965 unrecognized_option(argv[optind - 1]);
3966 break;
3967 case 'h':
3968 help();
3969 break;
3970 case 'f':
3971 fmt = optarg;
3972 break;
3973 case 'q':
3974 quiet = true;
3975 break;
3976 case OPTION_OBJECT: {
3977 QemuOpts *opts;
3978 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3979 optarg, true);
3980 if (!opts) {
3981 return 1;
3982 }
3983 } break;
3984 case OPTION_IMAGE_OPTS:
3985 image_opts = true;
3986 break;
3987 case OPTION_PREALLOCATION:
3988 prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3989 PREALLOC_MODE__MAX, NULL);
3990 if (prealloc == PREALLOC_MODE__MAX) {
3991 error_report("Invalid preallocation mode '%s'", optarg);
3992 return 1;
3993 }
3994 break;
3995 case OPTION_SHRINK:
3996 shrink = true;
3997 break;
3998 }
3999 }
4000 if (optind != argc - 1) {
4001 error_exit("Expecting image file name and size");
4002 }
4003 filename = argv[optind++];
4004
4005 if (qemu_opts_foreach(&qemu_object_opts,
4006 user_creatable_add_opts_foreach,
4007 qemu_img_object_print_help, &error_fatal)) {
4008 return 1;
4009 }
4010
4011 /* Choose grow, shrink, or absolute resize mode */
4012 switch (size[0]) {
4013 case '+':
4014 relative = 1;
4015 size++;
4016 break;
4017 case '-':
4018 relative = -1;
4019 size++;
4020 break;
4021 default:
4022 relative = 0;
4023 break;
4024 }
4025
4026 /* Parse size */
4027 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
4028 if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) {
4029 error_report_err(err);
4030 ret = -1;
4031 qemu_opts_del(param);
4032 goto out;
4033 }
4034 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
4035 qemu_opts_del(param);
4036
4037 blk = img_open(image_opts, filename, fmt,
4038 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
4039 false);
4040 if (!blk) {
4041 ret = -1;
4042 goto out;
4043 }
4044
4045 current_size = blk_getlength(blk);
4046 if (current_size < 0) {
4047 error_report("Failed to inquire current image length: %s",
4048 strerror(-current_size));
4049 ret = -1;
4050 goto out;
4051 }
4052
4053 if (relative) {
4054 total_size = current_size + n * relative;
4055 } else {
4056 total_size = n;
4057 }
4058 if (total_size <= 0) {
4059 error_report("New image size must be positive");
4060 ret = -1;
4061 goto out;
4062 }
4063
4064 if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
4065 error_report("Preallocation can only be used for growing images");
4066 ret = -1;
4067 goto out;
4068 }
4069
4070 if (total_size < current_size && !shrink) {
4071 error_report("Use the --shrink option to perform a shrink operation.");
4072 warn_report("Shrinking an image will delete all data beyond the "
4073 "shrunken image's end. Before performing such an "
4074 "operation, make sure there is no important data there.");
4075 ret = -1;
4076 goto out;
4077 }
4078
4079 /*
4080 * The user expects the image to have the desired size after
4081 * resizing, so pass @exact=true. It is of no use to report
4082 * success when the image has not actually been resized.
4083 */
4084 ret = blk_truncate(blk, total_size, true, prealloc, 0, &err);
4085 if (!ret) {
4086 qprintf(quiet, "Image resized.\n");
4087 } else {
4088 error_report_err(err);
4089 }
4090 out:
4091 blk_unref(blk);
4092 if (ret) {
4093 return 1;
4094 }
4095 return 0;
4096 }
4097
4098 static void amend_status_cb(BlockDriverState *bs,
4099 int64_t offset, int64_t total_work_size,
4100 void *opaque)
4101 {
4102 qemu_progress_print(100.f * offset / total_work_size, 0);
4103 }
4104
4105 static int print_amend_option_help(const char *format)
4106 {
4107 BlockDriver *drv;
4108
4109 /* Find driver and parse its options */
4110 drv = bdrv_find_format(format);
4111 if (!drv) {
4112 error_report("Unknown file format '%s'", format);
4113 return 1;
4114 }
4115
4116 if (!drv->bdrv_amend_options) {
4117 error_report("Format driver '%s' does not support option amendment",
4118 format);
4119 return 1;
4120 }
4121
4122 /* Every driver supporting amendment must have amend_opts */
4123 assert(drv->amend_opts);
4124
4125 printf("Amend options for '%s':\n", format);
4126 qemu_opts_print_help(drv->amend_opts, false);
4127 return 0;
4128 }
4129
4130 static int img_amend(int argc, char **argv)
4131 {
4132 Error *err = NULL;
4133 int c, ret = 0;
4134 char *options = NULL;
4135 QemuOptsList *amend_opts = NULL;
4136 QemuOpts *opts = NULL;
4137 const char *fmt = NULL, *filename, *cache;
4138 int flags;
4139 bool writethrough;
4140 bool quiet = false, progress = false;
4141 BlockBackend *blk = NULL;
4142 BlockDriverState *bs = NULL;
4143 bool image_opts = false;
4144 bool force = false;
4145
4146 cache = BDRV_DEFAULT_CACHE;
4147 for (;;) {
4148 static const struct option long_options[] = {
4149 {"help", no_argument, 0, 'h'},
4150 {"object", required_argument, 0, OPTION_OBJECT},
4151 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4152 {"force", no_argument, 0, OPTION_FORCE},
4153 {0, 0, 0, 0}
4154 };
4155 c = getopt_long(argc, argv, ":ho:f:t:pq",
4156 long_options, NULL);
4157 if (c == -1) {
4158 break;
4159 }
4160
4161 switch (c) {
4162 case ':':
4163 missing_argument(argv[optind - 1]);
4164 break;
4165 case '?':
4166 unrecognized_option(argv[optind - 1]);
4167 break;
4168 case 'h':
4169 help();
4170 break;
4171 case 'o':
4172 if (accumulate_options(&options, optarg) < 0) {
4173 ret = -1;
4174 goto out_no_progress;
4175 }
4176 break;
4177 case 'f':
4178 fmt = optarg;
4179 break;
4180 case 't':
4181 cache = optarg;
4182 break;
4183 case 'p':
4184 progress = true;
4185 break;
4186 case 'q':
4187 quiet = true;
4188 break;
4189 case OPTION_OBJECT:
4190 opts = qemu_opts_parse_noisily(&qemu_object_opts,
4191 optarg, true);
4192 if (!opts) {
4193 ret = -1;
4194 goto out_no_progress;
4195 }
4196 break;
4197 case OPTION_IMAGE_OPTS:
4198 image_opts = true;
4199 break;
4200 case OPTION_FORCE:
4201 force = true;
4202 break;
4203 }
4204 }
4205
4206 if (!options) {
4207 error_exit("Must specify options (-o)");
4208 }
4209
4210 if (qemu_opts_foreach(&qemu_object_opts,
4211 user_creatable_add_opts_foreach,
4212 qemu_img_object_print_help, &error_fatal)) {
4213 ret = -1;
4214 goto out_no_progress;
4215 }
4216
4217 if (quiet) {
4218 progress = false;
4219 }
4220 qemu_progress_init(progress, 1.0);
4221
4222 filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
4223 if (fmt && has_help_option(options)) {
4224 /* If a format is explicitly specified (and possibly no filename is
4225 * given), print option help here */
4226 ret = print_amend_option_help(fmt);
4227 goto out;
4228 }
4229
4230 if (optind != argc - 1) {
4231 error_report("Expecting one image file name");
4232 ret = -1;
4233 goto out;
4234 }
4235
4236 flags = BDRV_O_RDWR;
4237 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
4238 if (ret < 0) {
4239 error_report("Invalid cache option: %s", cache);
4240 goto out;
4241 }
4242
4243 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4244 false);
4245 if (!blk) {
4246 ret = -1;
4247 goto out;
4248 }
4249 bs = blk_bs(blk);
4250
4251 fmt = bs->drv->format_name;
4252
4253 if (has_help_option(options)) {
4254 /* If the format was auto-detected, print option help here */
4255 ret = print_amend_option_help(fmt);
4256 goto out;
4257 }
4258
4259 if (!bs->drv->bdrv_amend_options) {
4260 error_report("Format driver '%s' does not support option amendment",
4261 fmt);
4262 ret = -1;
4263 goto out;
4264 }
4265
4266 /* Every driver supporting amendment must have amend_opts */
4267 assert(bs->drv->amend_opts);
4268
4269 amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts);
4270 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4271 if (!qemu_opts_do_parse(opts, options, NULL, &err)) {
4272 /* Try to parse options using the create options */
4273 amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts);
4274 qemu_opts_del(opts);
4275 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4276 if (qemu_opts_do_parse(opts, options, NULL, NULL)) {
4277 error_append_hint(&err,
4278 "This option is only supported for image creation\n");
4279 }
4280
4281 error_report_err(err);
4282 ret = -1;
4283 goto out;
4284 }
4285
4286 /* In case the driver does not call amend_status_cb() */
4287 qemu_progress_print(0.f, 0);
4288 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err);
4289 qemu_progress_print(100.f, 0);
4290 if (ret < 0) {
4291 error_report_err(err);
4292 goto out;
4293 }
4294
4295 out:
4296 qemu_progress_end();
4297
4298 out_no_progress:
4299 blk_unref(blk);
4300 qemu_opts_del(opts);
4301 qemu_opts_free(amend_opts);
4302 g_free(options);
4303
4304 if (ret) {
4305 return 1;
4306 }
4307 return 0;
4308 }
4309
4310 typedef struct BenchData {
4311 BlockBackend *blk;
4312 uint64_t image_size;
4313 bool write;
4314 int bufsize;
4315 int step;
4316 int nrreq;
4317 int n;
4318 int flush_interval;
4319 bool drain_on_flush;
4320 uint8_t *buf;
4321 QEMUIOVector *qiov;
4322
4323 int in_flight;
4324 bool in_flush;
4325 uint64_t offset;
4326 } BenchData;
4327
4328 static void bench_undrained_flush_cb(void *opaque, int ret)
4329 {
4330 if (ret < 0) {
4331 error_report("Failed flush request: %s", strerror(-ret));
4332 exit(EXIT_FAILURE);
4333 }
4334 }
4335
4336 static void bench_cb(void *opaque, int ret)
4337 {
4338 BenchData *b = opaque;
4339 BlockAIOCB *acb;
4340
4341 if (ret < 0) {
4342 error_report("Failed request: %s", strerror(-ret));
4343 exit(EXIT_FAILURE);
4344 }
4345
4346 if (b->in_flush) {
4347 /* Just finished a flush with drained queue: Start next requests */
4348 assert(b->in_flight == 0);
4349 b->in_flush = false;
4350 } else if (b->in_flight > 0) {
4351 int remaining = b->n - b->in_flight;
4352
4353 b->n--;
4354 b->in_flight--;
4355
4356 /* Time for flush? Drain queue if requested, then flush */
4357 if (b->flush_interval && remaining % b->flush_interval == 0) {
4358 if (!b->in_flight || !b->drain_on_flush) {
4359 BlockCompletionFunc *cb;
4360
4361 if (b->drain_on_flush) {
4362 b->in_flush = true;
4363 cb = bench_cb;
4364 } else {
4365 cb = bench_undrained_flush_cb;
4366 }
4367
4368 acb = blk_aio_flush(b->blk, cb, b);
4369 if (!acb) {
4370 error_report("Failed to issue flush request");
4371 exit(EXIT_FAILURE);
4372 }
4373 }
4374 if (b->drain_on_flush) {
4375 return;
4376 }
4377 }
4378 }
4379
4380 while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4381 int64_t offset = b->offset;
4382 /* blk_aio_* might look for completed I/Os and kick bench_cb
4383 * again, so make sure this operation is counted by in_flight
4384 * and b->offset is ready for the next submission.
4385 */
4386 b->in_flight++;
4387 b->offset += b->step;
4388 b->offset %= b->image_size;
4389 if (b->write) {
4390 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4391 } else {
4392 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4393 }
4394 if (!acb) {
4395 error_report("Failed to issue request");
4396 exit(EXIT_FAILURE);
4397 }
4398 }
4399 }
4400
4401 static int img_bench(int argc, char **argv)
4402 {
4403 int c, ret = 0;
4404 const char *fmt = NULL, *filename;
4405 bool quiet = false;
4406 bool image_opts = false;
4407 bool is_write = false;
4408 int count = 75000;
4409 int depth = 64;
4410 int64_t offset = 0;
4411 size_t bufsize = 4096;
4412 int pattern = 0;
4413 size_t step = 0;
4414 int flush_interval = 0;
4415 bool drain_on_flush = true;
4416 int64_t image_size;
4417 BlockBackend *blk = NULL;
4418 BenchData data = {};
4419 int flags = 0;
4420 bool writethrough = false;
4421 struct timeval t1, t2;
4422 int i;
4423 bool force_share = false;
4424 size_t buf_size;
4425
4426 for (;;) {
4427 static const struct option long_options[] = {
4428 {"help", no_argument, 0, 'h'},
4429 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4430 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4431 {"pattern", required_argument, 0, OPTION_PATTERN},
4432 {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4433 {"force-share", no_argument, 0, 'U'},
4434 {0, 0, 0, 0}
4435 };
4436 c = getopt_long(argc, argv, ":hc:d:f:ni:o:qs:S:t:wU", long_options,
4437 NULL);
4438 if (c == -1) {
4439 break;
4440 }
4441
4442 switch (c) {
4443 case ':':
4444 missing_argument(argv[optind - 1]);
4445 break;
4446 case '?':
4447 unrecognized_option(argv[optind - 1]);
4448 break;
4449 case 'h':
4450 help();
4451 break;
4452 case 'c':
4453 {
4454 unsigned long res;
4455
4456 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4457 error_report("Invalid request count specified");
4458 return 1;
4459 }
4460 count = res;
4461 break;
4462 }
4463 case 'd':
4464 {
4465 unsigned long res;
4466
4467 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4468 error_report("Invalid queue depth specified");
4469 return 1;
4470 }
4471 depth = res;
4472 break;
4473 }
4474 case 'f':
4475 fmt = optarg;
4476 break;
4477 case 'n':
4478 flags |= BDRV_O_NATIVE_AIO;
4479 break;
4480 case 'i':
4481 ret = bdrv_parse_aio(optarg, &flags);
4482 if (ret < 0) {
4483 error_report("Invalid aio option: %s", optarg);
4484 ret = -1;
4485 goto out;
4486 }
4487 break;
4488 case 'o':
4489 {
4490 offset = cvtnum("offset", optarg);
4491 if (offset < 0) {
4492 return 1;
4493 }
4494 break;
4495 }
4496 break;
4497 case 'q':
4498 quiet = true;
4499 break;
4500 case 's':
4501 {
4502 int64_t sval;
4503
4504 sval = cvtnum_full("buffer size", optarg, 0, INT_MAX);
4505 if (sval < 0) {
4506 return 1;
4507 }
4508
4509 bufsize = sval;
4510 break;
4511 }
4512 case 'S':
4513 {
4514 int64_t sval;
4515
4516 sval = cvtnum_full("step_size", optarg, 0, INT_MAX);
4517 if (sval < 0) {
4518 return 1;
4519 }
4520
4521 step = sval;
4522 break;
4523 }
4524 case 't':
4525 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4526 if (ret < 0) {
4527 error_report("Invalid cache mode");
4528 ret = -1;
4529 goto out;
4530 }
4531 break;
4532 case 'w':
4533 flags |= BDRV_O_RDWR;
4534 is_write = true;
4535 break;
4536 case 'U':
4537 force_share = true;
4538 break;
4539 case OPTION_PATTERN:
4540 {
4541 unsigned long res;
4542
4543 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4544 error_report("Invalid pattern byte specified");
4545 return 1;
4546 }
4547 pattern = res;
4548 break;
4549 }
4550 case OPTION_FLUSH_INTERVAL:
4551 {
4552 unsigned long res;
4553
4554 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4555 error_report("Invalid flush interval specified");
4556 return 1;
4557 }
4558 flush_interval = res;
4559 break;
4560 }
4561 case OPTION_NO_DRAIN:
4562 drain_on_flush = false;
4563 break;
4564 case OPTION_IMAGE_OPTS:
4565 image_opts = true;
4566 break;
4567 }
4568 }
4569
4570 if (optind != argc - 1) {
4571 error_exit("Expecting one image file name");
4572 }
4573 filename = argv[argc - 1];
4574
4575 if (!is_write && flush_interval) {
4576 error_report("--flush-interval is only available in write tests");
4577 ret = -1;
4578 goto out;
4579 }
4580 if (flush_interval && flush_interval < depth) {
4581 error_report("Flush interval can't be smaller than depth");
4582 ret = -1;
4583 goto out;
4584 }
4585
4586 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4587 force_share);
4588 if (!blk) {
4589 ret = -1;
4590 goto out;
4591 }
4592
4593 image_size = blk_getlength(blk);
4594 if (image_size < 0) {
4595 ret = image_size;
4596 goto out;
4597 }
4598
4599 data = (BenchData) {
4600 .blk = blk,
4601 .image_size = image_size,
4602 .bufsize = bufsize,
4603 .step = step ?: bufsize,
4604 .nrreq = depth,
4605 .n = count,
4606 .offset = offset,
4607 .write = is_write,
4608 .flush_interval = flush_interval,
4609 .drain_on_flush = drain_on_flush,
4610 };
4611 printf("Sending %d %s requests, %d bytes each, %d in parallel "
4612 "(starting at offset %" PRId64 ", step size %d)\n",
4613 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4614 data.offset, data.step);
4615 if (flush_interval) {
4616